Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Sensing, Vision, and Perception
Abstract: Accurate and reliable localization is a prerequisite for autonomous driving. Methods based on sparse landmarks, such as pole-like structures, have been widely studied because of their lower requirements for computing and storage. However, the number of landmarks of a single type is not always sufficient for reliable positioning. We propose a localization method using three different types of features in urban environments. The features we choose are poles, corners and walls which are persistent over time and can be reliably detected with LiDAR sensors. A pattern matching method for data association is introduced. Instead of using a filtering method, we adopt the graph-based optimization method to solve the pose estimation problem. Experiments conducted on two test roads show that the proposed method can provide accurate and reliable localization results in urban environments.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation
Abstract: Trajectory planning in the intersection is a challenging problem due to the strong uncertain intentions of surrounding agents. Conventional methods may fail in some corner cases when the ad-hoc parameters or predictions do not match the real traffic. This paper proposes a trajectory planning method, adaptive to the uncertain interactions, called Value-Estimation-Guild (VEG) trajectory planner. The method builds on the Frenet frame trajectory planner, in the meantime, uses the deep reinforcement learning to deal with the high uncertainty. The deep reinforcement learning learns from past failures and adjusts the sample direction of the optimal planner under the Frenet frame. In this way, the generated trajectory can be partially optimal and adapt to the stochastic as well. This method drives the automated vehicle through intersections and completes the unprotected left turn mission. During the testing, traffic density, surrounding vehicles’ types, and intentions are all generated randomly. The statistics results show that the proposed trajectory planner works well under high uncertainty. It helps the automatic vehicles to finish the unprotected left turn with a success rate of 94.4%, compared with the baseline method of 90%

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Sensing, Vision, and Perception, Cooperative Techniques and Systems
Abstract: Predicting the surrounding vehicles' behavior is an important requirement for automated driving as it enables estimating others' reactions to the own behavior during planning as well as the identification of critical situations.

This work proposes a recursive multi-step training scheme for neural networks that predict other vehicles' positions in a highway car-following scenario. We implement a neural network and compare the proposed approach to the commonly used single-step training as well as parametric models. For this, the Intelligent Driver Model (IDM) and its derivatives have been calibrated using the same approach.

Evaluation is performed on 10 hours of real-world car-following situations, extracted from the extensive HighD dataset. Given equal inputs, we show that a minimal neural network with two layers composed of three neurons each surpasses the prediction performance of both the parametric prediction models and the network trained with the standard single-step approach.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Data Mining and Data Analysis, Theory and Models for Optimization and Control
Abstract: Intelligent motion planning is one of the core components in automated vehicles, which has received extensive interests. Traditional motion planning methods suffer from several drawbacks in terms of optimality, efficiency and generalization capability. Sampling based methods cannot guarantee the optimality of the generated trajectories. Whereas the optimization-based methods are not able to perform motion planning in real-time, and they do not generalize to new scenarios. In this work, we propose a learning-based approach to handle those shortcomings. Mixed Integer Quadratic Problem (MIQP) is used to generate the optimal lane-change trajectories which served as the training dataset for learning-based action generation algorithms. A hierarchical supervised learning model is designed to make the fast lane-change decision. Numerous experiments have been conducted to evaluate the optimality, efficiency, and generalization capability of the proposed approach. The experimental results indicate that the proposed model outperforms several common used motion planning baselines.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Advanced Vehicle Safety Systems, Driver Assistance Systems
Abstract: Highway scenarios are highly dynamic environments where several vehicles interact following their own goal, leading to different combinations of scenes that also change over time. Human drivers adapt their driving behavior integrating current information with their former experiences. In a similar way, an autonomous system performing any driving activity should be able to integrate information learned from former interactions. Reinforcement Learning has shown promising results, but it should only be applied to autonomous vehicles if the system is also able to fulfill safety and integrity requirements on a deterministic and reproducible way. This paper presents a planning system that is able to learn over time, always complying to the safety requirements. Our planner integrates several layers interacting with each other, combining the advantages of Reinforcement Learning based systems and reactive systems. We present a planner that ensures driving safety on short horizons and integrates previous experiences to optimize the expected reward. We evaluate our method in simulation comparing different learning techniques. Our results show that the planning system is able to adaptively integrate this experience outperforming rule-based strategies

Keywords: Sensing, Vision, and Perception, Driver Assistance Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations
Abstract: 3D object detection is one of the vital tasks in many fields, especially for autonomous driving. Among all technology paths for this task, the detection based on monocular images has been proven an efficient way with low cost. However, the performance of most current algorithms is far from satisfactory. The current leading systems can’t achieve the accuracy which is comparable with LiDAR-based algorithms, while the real-time problems also exist. In this paper, we propose a novel anchor-free model for monocular 3D object detection. We choose the effective modified DenseNet as the feature extraction part. The Joint Pyramid Upsampling is applied to fuse features maps for multiply scales, and the Atrous Spatial Pyramid Pooling is used to maximize the context information. Finally, 5 convolution layers are connected with the feature map to predict the information. We call the model Dense-JANet and train it on the large autonomous driving dataset, called nuScenes, which has more scenes and data than KITTI. Experiments show that Dense-JANet’s performance exceeds SOTA model on small object prediction and orientation prediction, while the proposed model can fully meet the real-time requirements.

Keywords: Sensing, Vision, and Perception
Abstract: In the field of Advanced Driver-Assistance Systems, road traffic actors detection is a vital task in order to avoid human errors in driving. Unlike camera only-based convolutional neural networks for 2D object detection, multimodality using improve object detectors accuracy and robustness. In this paper, we propose Stacked Fusion Double RetinaNet (SFD-Retina) and Gated Fusion Double RetinaNet (GFD-Retina), two convolutional neural networks taking multimodal data (RGB, Depth from Stereo, Optical Flow, LIDAR Point Cloud) as input. These networks combine efficiently sensor specific properties by using both early fusion and middle fusion for detecting road objects and their 2D localization. Evaluation of SFD-Retina and GFD-Retina on the challenging KITTI object detection benchmark shows that using sensor fusion improve significantly object detection accuracy. Moreover, GFD-Retina with Gated Fusion Unit outperforms SFD-Retina with Stacked Fusion Unit, and obtain satisfying results against state-of-the-art algorithms.

Keywords: Sensing, Vision, and Perception, Driver Assistance Systems
Abstract: A millimeter-wave radar has major advantages in robustness under adverse weather and illumination conditions. However, it has some concerns regarding signal noise and resolution. They make it difficult for the radar to precisely recognize the driving environments. We are challenging to reconstruct parking scenes requiring precise obstacle shape information. We have developed the 24-layer Convolutional Neural Network (CNN) originally designed from scratch using the accumulated radar reflection map for the reconstruction. 676 radar reflection maps with the ground truth for parking cars, curbs and fences were generated as the dataset by the 79 GHz UWB radar installed on our experimental vehicle. The object shapes were reconstructed by our network under various conditions compared with conventional CNNs, SegNet and U-net. Our CNN achieved a high estimation accuracy of over 97% under all conditions. Maximum outline errors of the parking cars in adjacent area were within 7.3 cm. In contrast, other CNNs degraded to less than 94 % and the outline errors increased to over 9.7 cm. Most of them had a considerable collapse of the reconstructed shape.

Keywords: Sensing, Vision, and Perception, Sensing and Intervening, Detectors and Actuators, Driver Assistance Systems
Abstract: Three-dimensional object estimation from monocular imagery is a difficult task due to the geometric information loss induced by the perspective projection. Currently, most methods try to deal with this limitation by using deep learning models augmented with some form of 3D reasoning, which often leads to complicated inference pipelines. In this work, we argue that explicit 3D reasoning is not mandatory for good monocular 3D detection performance. To this end, we extend a 2D object detection framework with a small subnetwork responsible for 3D bounding box estimation. To train this module, we introduce a loss function based on the Generalized Intersection-over Union in which each degree of freedom is optimized separately. The resulting approach is simple, modular and straightforward to integrate into existing monocular 2D detection frameworks. Experiments on the challenging KITTI dataset show that our method achieves state-of-the-art performance on the 3D detection task.

Keywords: Sensing, Vision, and Perception
Abstract: Radar sensors are an important part of driver assistance systems and intelligent vehicles due to their robustness against all kinds of adverse conditions, e.g., fog, snow, rain, or even direct sunlight. This robustness is achieved by a substantially larger wavelength compared to light-based sensors such as cameras or lidars. As a side effect, many surfaces act like mirrors at this wavelength, resulting in unwanted ghost detections. In this article, we present a novel approach to detect these ghost objects by applying data-driven machine learning algorithms. For this purpose, we use a large-scale automotive data set with annotated ghost objects. We show that we can use a state-of-the-art automotive radar classifier in order to detect ghost objects alongside real objects. Furthermore, we are able to reduce the amount of false positive detections caused by ghost images in some settings.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Simulation and Modeling
Abstract: Eco-routing aims to find paths that minimize vehicles' energy consumption. We develop a new data-driven energy consumption model for plug-in hybrid vehicles and propose an algorithm that determines both the optimal path and powertrain strategy. To evaluate the performance of our proposed model and algorithm, we develop a calibrated simulation model of Ann Arbor in SUMO and implement our eco-routing algorithm with connected and automated vehicles in the simulation. Our result shows that we achieve an energy-saving around 7%-14%.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Driver Assistance Systems
Abstract: Video-only Automatic Cruise Control (ACC) is a low-cost alternative to radar-based and multi-sensor systems. Furthermore, (mono) cameras are becoming a standard vehicle equipment as part of lane support systems, e.g. in order to be able to achieve a 5 star rating in the EuroNCAP. In this case, ACC can be provided as value-added even without additional sensory costs. However, longitudinal control based on camera data is a challenging task because neither the relative distance nor the relative velocity to the target vehicle can directly be measured by the video sensor. Therefore, these quantities must be inferred using filtering methods if classical ACC controllers are to be used. This introduces errors and delays due to model assumptions of the filtering algorithms and leads to poor control performance. We address this issue by designing an ACC system that relies on quantities that can be directly extracted from the image data namely time to collision and the size of the target object detection. In order to leverage experience in classical ACC design, we learn a deep neural network target acceleration controller by copying the behavior of a classical controller. The proposed approach is demonstrated in a test vehicle on a German highway.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Aerial, Marine and Surface Intelligent Vehicles, Driver Assistance Systems
Abstract: Automated driving (AD) is one of the fastest-growing tendencies in the Intelligent Transportation Systems (ITS) field with some interesting demonstrations and prototypes. Currently, the main research topics are aligned with vehicle communications, environment recognition, control, and decision-making. A real-time trajectory planning method for Automated vehicles (AVs) is presented in this paper; the contribution is part of AD’s decision-making module. This novel approach uses the properties of the 3er order Bezier curves to generate fast and reliable vehicle trajectories. Online execution and vehicle tracking capacities are considered on the approach. A feasible trajectory is selected based on the criteria: (i) the vehicle must be contained by a collision-free corridor given by an upper decision layer, (ii) the vehicle must be capable to track the generated trajectory, and (iii) the continuity of the path and curvature must be preserved in the joints. Our approach was tested considering a vehicle length (automated bus) of 12 meters. The scenario has the dimension of a real test location with multiple roundabouts.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Driver Assistance Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations
Abstract: It is fair to assume that when humans drive, they rely on their prior knowledge to predict the behaviors of other vehicles. In order for a self-driving car to be driven safely with other surrounding cars, it must behave in a predictable way, similar to the behaviors of human-driven cars. It is common to adopt machine learning methods which have been considerably improved to predict behaviors of vehicles with big data. The more scenarios the model covers, the larger the model becomes. As a result, it is inevitable that the model loses interpretability. On the other hand, although a simple model represented by linear combination of features can be interpretable, it can cover only a limited scenario. In order to solve these limitations, we propose a method to build multiple linear models and a model to select appropriate one of them on the situation. This set of models is extracted from a trained machine learning model. The proposed approach was validated using real-world driving data, namely Next Generation Simulation (NGSIM).

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Driver Assistance Systems, Data Mining and Data Analysis
Abstract: Predicting the behavior of surrounding traffic participants is crucial for advanced driver assistance systems and autonomous driving. Most researchers however do not consider contextual knowledge when predicting vehicle motion. Extending former studies, we investigate how predictions are affected by external conditions. To do so, we categorize different kinds of contextual information and provide a carefully chosen definition as well as examples for external conditions. More precisely, we investigate how a state-of-the-art approach for lateral motion prediction is influenced by one selected external condition, namely the traffic density. Our investigations demonstrate that this kind of information is highly relevant in order to improve the performance of prediction algorithms. Therefore, this study constitutes the first step towards the integration of such information into automated vehicles. Moreover, our motion prediction approach is evaluated based on the public highD data set showing a maneuver prediction performance with areas under the ROC curve above 97% and a median lateral prediction error of only 0.18m on a prediction horizon of 5s.

Keywords: Sensing, Vision, and Perception, Advanced Vehicle Safety Systems, Driver Assistance Systems
Abstract: In this work, we solve the task of automotive traffic scene classification using a deep learning approach on low-level radar data. In contrast to existing approaches using 2D camera images, the input are complex-valued 3D range-beam-doppler tensors outputted by an automotive radar. We design, train and evaluate three different networks and analyze the effects of different nuances in processing this type of data in deep neural networks. Particular attention is paid to complex-valued convolutions and their usage on complex-valued 3D tensors. The resulting networks achieve a classification accuracy above 95% on split test datasets and show the usability of automotive radar data for traffic scene classification, which can be integrated into weather robust multi-sensor systems.

Keywords: Sensing, Vision, and Perception, Driver Assistance Systems
Abstract: In smart mobility based road navigation, object detection, depth estimation and tracking are very important tasks for improvement of the environment perception quality. In the recent years, a surge of deep-learning based depth estimation methods for monocular cameras has lead to significant progress in this field. In this paper, we propose an evaluation of state-of-the-art depth estimation algorithms based on single single input on both the KITTI dataset and the recently published NUScenes dataset. The models evaluated in this paper include an unsupervised method (Monodepth2) and a supervised method (BTS). Our work lies in the elaboration of novel depth evaluation protocols, object depth evaluation and distance ranges evaluation. We validated our new protocols on both KITTI and NUScenes datasets, allowing us to get a more comprehensive evaluation for depth estimation, especially for applications in scene understanding for both road and rail environment.

Keywords: Sensing, Vision, and Perception, Automated Vehicle Operation, Motion Planning, Navigation, Aerial, Marine and Surface Intelligent Vehicles
Abstract: Environment perception is a critical function used by driving automation systems, or self-driving cars, for detecting objects such as obstacles, lane markings, and road signs. In order to replace human drivers, self-driving cars will need to safely operate in parking lots, private roads, underground, or any other environment with poor GPS signals or uncharted infrastructure. While much attention has been spent on recognizing lane markings, non-lane road markings have received considerably less attention. Current perception systems can recognize only a small subset of markings and often only under favorable weather conditions. This limitation is exacerbated by the current quality of scene segmentation data sets. Only a select few of existing data sets have annotations for non-lane road markings, and the ones that do only have them for a small number of marking types. Additionally most of the data sets were generated under one type of driving condition. Finally, it is difficult to determine if current recognition systems can satisfy real-time requirements. This paper investigates the current limitations and challenges for non-lane road marking detection and recognition including recognition capabilities, data set quality, and inference times.

Keywords: Sensing, Vision, and Perception
Abstract: Complete perception of the environment and its correct interpretation is crucial for autonomous vehicles. Object perception is the main component of automotive surround sensing. Various metrics already exist for the evaluation of object perception. However, objects can be of different importance depending on their velocity, orientation, distance, size, or the potential damage that could be caused by a collision due to a missed detection. Thus, these additional parameters have to be considered for safety evaluation. We propose a new safety metric that incorporates all these parameters and returns a single easily interpretable safety assessment score for object perception. This new metric is evaluated with both real world and virtual data sets and compared to state of the art metrics.

Keywords: Sensing, Vision, and Perception
Abstract: Object tracking is one of the key challenges for perception systems of autonomous vehicles. Recursive Bayesian state estimation can be used to obtain object tracks. Both the measurement association and the object update within such Bayesian filters rely on sensor measurement models. These models offer an approximation of the expected sensor values that can be error-prone due to a mismatch between model and reality. The discrepancy is caused by the limited descriptive capacity of measurement models since sensor measurements are highly object and situation dependent. We address this problem in a data-driven approach by using Deep Neural Networks (DNNs) to learn situation dependent sensor measurement models. In detail, the DNN acts as a virtual sensor that uses current sensor measurements to regress necessary corrections of predicted object states. It can be directly plugged into existing tracking frameworks, substituting the previously hand-modeled association and update steps during Bayesian Filtering. We apply the proposed DNN-based measurement models to the problem of vehicle tracking using radar data in an Extended Kalman Filter setup and compare it to a classical closest reflex and an L-shape measurement update model. Extensive evaluation on a real-world dataset shows that our model improves performances significantly compared to state of the art methods.

Keywords: Data Mining and Data Analysis, Commercial Fleet Management, Human Factors in Intelligent Transportation Systems
Abstract: Recent studies on machine learning based driver identification have shown that leveraging deep neural networks to learn latent features from vehicle sensor data boosts the performance of the models to high levels of accuracies. However, models produced by deep neural networks are difficult to explain and the interpretability of their results is limited. Consequently, the reliability of the learned models heavily depends on the amount, quality and diversity of the training dataset as well as on the validation scenarios. In this work, we evaluate state-of-the-art deep learning networks for driver identification using a very rich dataset of more than 395,000 kilometres of different driving scenarios and environmental conditions (e.g. route, vehicle, traffic, weather) collected over two years. It turns out that the neural networks achieve high accuracy levels when training and testing on the same type of driving conditions. However, accuracy drops and importance of individual signals varies when testing on different driving conditions, although all best practice like stratification and cross validation have been applied. Our findings suggest that relying solely on the vehicle sensor data without taking the contextual information about driving conditions into account is not a practical solution.

Keywords: Data Mining and Data Analysis, Off-line and Online Data Processing Techniques
Abstract: The modern road network infrastructure (signal controllers and detectors) continuously generates data that can be transformed and used to evaluate the performance of signalized intersections. In order to automatically make meaningful observations about signal performance, we propose the application of data summarization and compression techniques in order to intelligently group together intersections and/or time intervals during the day and certain days of the week. This work details the use of linear and nonlinear dimensionality reduction techniques to achieve the aforementioned goals. The approach is also extended to perform change detection so that significant changes at intersections and corridors can be highlighted.

Keywords: Data Mining and Data Analysis, Off-line and Online Data Processing Techniques, Roadside and On-board Safety Monitoring
Abstract: Predictive maintenance is a key component regarding cost reduction in automotive industry and is of great importance. It can improve both feeling of comfort and safety, by means of early detection, isolation and prediction of prospective failures. That is why automotive industry and fleet managers are turning to predictive analytic to maintain a lead position in industry. A patent application has been recently submitted, proposing a two stage solution, including a real-time solution (on-board diagnostic system) and an offline solution (in the back-end), for health monitoring/assessment of different chassis components. Hardware faults are detected based on changes of the fundamental eigen-frequencies of the vehicle where time series of interest, from in-car sensory system, are collected/reported for advanced data analytic in the back-end. The main focus of this paper in on the latter solution, using an unsupervised machine learning approach. A clustering approach based on Mixture of Hidden Markov Models, is adopted to conduct automatic diagnosis and isolation of faults. Detection and isolation of tire and wheel bearing faults has been considered for this study but same framework can be used to handle other components faults, such as suspension system faults. In order to validate the performance of the proposed approach tests were done at Hällared test track in Gothenburg, and data were collected for two faulty states (for faulty wheel bearing and low-tire pressure) and no-fault state.

Keywords: Data Mining and Data Analysis, Commercial Fleet Management, Sensing, Vision, and Perception
Abstract: Vehicle viewpoint estimation from vehicle cameras is a crucial component of road scene understanding. In this paper, we propose a deep lightweight method to predict vehicle viewpoint from a single RGB dashcam image. To this aim, we customize and adapt state-of-the-art deep learning techniques for general object viewpoint estimation to the vehicle viewpoint estimation task. Furthermore, we define a novel objective function that takes into account errors at different granularity to improve neural network training. To keep the model lightweight and fast, we rely upon MobileNetV2 as backbone. Tested both on benchmark viewpoint estimation data(Pascal3D+) and on actual vehicle camera data(nuScenes), our method is shown to outperform the state of the art in vehicle viewpoint estimation, in terms of both accuracy and memory footprint.

Keywords: Data Mining and Data Analysis, Theory and Models for Optimization and Control, Off-line and Online Data Processing Techniques
Abstract: Special event, including sporting events, concerts, fairs and festivals, can generate large volumes of traffic such that congestion and associated problems occur. Quantifying and understanding the traffic characteristics of sport events would be useful in predicting how such events will affect traffic flow. In this communication we propose a quantitative methodology in order to gauge the effects of mayor events on the traffic of a city. In particular we apply the methodology to the city of Madrid and the games of two of the main soccer teams of the city for three different competitions. Furthermore it has been found that the intensity of the effect of the games decay as an inverse power law of the distance to the stadium.

Keywords: Rail Traffic Management, Data Mining and Data Analysis, Transportation Security
Abstract: This paper proposes a data-driven method that combines eXtreme Gradient Boosting (XGBoost) and Bayesian optimization algorithm to predict train arrival delays. First, 11 characteristics that may affect train arrival delay at the next station are identified as independent variables. Second, an XGBoost prediction model that capture the relation between train arrival delays and various characteristics of a railway system is established. Third, the Bayesian optimization algorithm is applied to the hyper-parameter optimization of XGBoost model to improve the prediction accuracy. Finally, a case study is illustrated to show the prediction accuracy of the proposed method. The results demonstrate that the proposed method has a higher prediction precision and outperforms other benchmark methods (i.e., Random Forest, Deep Extreme Learning machine and Gradient Boosting Regression Tress).

Keywords: Rail Traffic Management, Transportation Security, Other Theories, Applications, and Technologies
Abstract: To obtain refined safety requirements during system architecture design stage, the traditional static fault tree analysis mothed is widely used to analyse the logical relationship between basic hazardous events leading to system hazards in railway signal system, and to identify the weak and key equipment of the system. In order to accurately characterize the dynamic behaviours of the system and improve the accuracy and credibility of analysis results, a dynamic fault tree analysis mothed of railway signal system based on failure propagation modelling was proposed. According to failure logic modelling of dynamic failure behaviours, conforming to signal system function design, component failure model was established within SimFIA platform, and fault trees of system hazards were attained by model simulation. The result of Communication-Based Train Control (CBTC) System on-board equipment case study shows that the fault tree of complex system function were generated by the dynamic fault tree analysis mothed based on Failure Propagation and Transformation Notation (FPTN) modelling, which guaranteed the safety analysis outcomes was accurate and credible.

Keywords: Rail Traffic Management, Theory and Models for Optimization and Control, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations
Abstract: In vehicle-centralized train control design, the routing function is transferred from interlocking equipment to vehicle, which asks for the ability of train to schedule feasible route autonomously. This paper describes an autonomous routing schedule (ARS) model based on graph theory. Firstly, by converting key elements of rail network into directed graph nodes and marking the weights of arcs based on section transit time’s prediction, a topological graph model reflecting railway structure’s characteristic is set up. According to the graph model, a heuristic algorithm is used to search the feasible route with shortest transit time. Considering the limitation of rail state’s information gathered by trains in decentralized control design, arc’s weight (the prediction of transit time in rail section) is updated in real time based on the communication between trains so that the routing schedule can be dynamically adjusted based on section’s availability. The computational tests are performed on Beijing Daxing Airport Station. The result shows the feasibility of model in searching route with reasonable transit time. It’s potential for rerouting based on disturbances and resolve delays is also analyzed.

Keywords: Rail Traffic Management, Public Transportation Management, Cooperative Techniques and Systems
Abstract: Virtual Coupling (VC) is a newly introduced concept of train-centric signalling technology that conceives trains to run autonomously in radio-connected platoons. These trains move synchronously at a relative braking distance to significantly improve railway capacity and address the forecasted increase in railway demand. The technical feasibility of VC depends on its strengths, weaknesses, opportunities and threats which can introduce radical changes to current train services, technologies and procedures. This paper investigates demand trends and operational scenarios of future train-centric signalling systems. To this end, stated travel preferences have been collected by means of a survey to have more insight on modal shares in the case of future VC applications. In addition, a Delphi method has been applied where another extensive survey has collected expert opinions about benefits and challenges of VC. Results show that VC can be very attractive to customers of high-speed and main line railways and have special benefits to the regional market where a manifest willing to pay more for using a more frequent train service was found. This concept therefore calls for a deeper understanding of possible Virtual Coupling operational scenarios and the impact on the railway industry.

Keywords: Rail Traffic Management, Theory and Models for Optimization and Control
Abstract: The passenger demand of urban rail transit is dynamic and uneven in time and space, and traditional train plan of single train formation cannot adapt to dynamic passenger demand. In order to solve the redundancy of train capacity caused by uneven passenger demand in bi-directions, we proposed a mixed-integer linear programing model (MILP) to optimize the train formation plan and rolling stock scheduling integrally based on known passenger demand and timetable for an urban rail transit line. The turnaround operation, coupling/decoupling operation, the entering/exiting depot operation of train services, the number of available trains and the capacity of depot are involved. The model is solved by the CPLEX solver. As illustration, the model is applied to Beijing Batong line to verify its effectiveness and performance. The results show that through this integrated approach the number of operation formations can reduce 44% and the number of rolling stocks can reduce 20%. It demonstrated that the proposed model can effectively reduce the operation cost while satisfy the uneven demand.

Keywords: Simulation and Modeling, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Network Management
Abstract: Highway work zones can induce significant delays and undermine traffic safety. The recent advent of connected and automated vehicles (CAVs) can pose an additional threat to traffic flow performance and safety around highway work zones. CAVs equipped with low – medium level automation systems that cannot reliably address work zone scenarios under all circumstances could induce control transitions and imminent Minimum Risk Manoeuvers (MRMs) that would result in significant traffic disruption and multiple safety critical events. The latter negative effects could be mitigated via the introduction of highly automated vehicles that could utilize sophisticated infrastructure assistance to traverse highway work zones without disengaging automation systems. This study develops novel and utilizes existing vehicle-driver models to simulate manual driving, mixed traffic and infrastructure-assisted highly automated traffic around highway work zones. Traffic operations are evaluated for the latter fleet mixes and three different traffic demand levels. Simulation results indicate that joint deployment of infrastructure-assisted traffic management and cooperative driving can ensure increased traffic efficiency and safety levels for high traffic intensity in a fully connected and automated road environment.

Keywords: Travel Information, Travel Guidance, and Travel Demand Management, Sensing and Intervening, Detectors and Actuators, Other Theories, Applications, and Technologies
Abstract: Predicting driving conditions is a complex endeavour, both in terms of defining a metric that reliably describes driving conditions as well as for quantifying and predicting anticipated future conditions. In this paper, we introduce the concept of using measured vehicle speeds as a ground truth for driving conditions at specific locations at mountain crossings, and present a prediction model that estimates future expected driving conditions using deep neural networks. Our network inputs consist of time series measurements from weather stations and vehicle speeds together with weather predictions from national forecasts. Applying multi-step hybrid Convolution and Long Short-Term Memory networks, we predict vehicle speeds six hours into the future to be used as an indicator for the expected driving conditions. The model is trained on historical data and then applied to real-time weather and vehicle data to set up a live system that can be used as a decision-making tool for road entrepreneurs to help them assess whether or not extraordinary measures, such as convoy driving or road closure, may need to be implemented.

Keywords: Emission and Noise Mitigation, Electric Vehicles, Simulation and Modeling
Abstract: This paper presents a city-scale evaluation of the effects of road transport electrification and automation on heat and carbon emissions. We present a case study for the city of Singapore examining the spatio-temporal profile of the heat emissions due to road transport for a typical day. We calibrate and validate our simulation model which is later used for analysis of future electrification and automation scenarios. Furthermore, we also evaluate the temporal energy demand associated with the electrification of transport and assess the heat released for the production of this energy. Our results show a sixfold decrease of the energy usage, and thus heat production, of the road transport sector in case of a complete electrification of all vehicle classes, which include lorries and vans, private vehicles, taxis, buses, and motorcycles. We find that while autonomous mobility greatly reduces the overall trip durations as it mitigates congestion, the energy consumption of the sector remains almost unchanged compared to the fully electric scenario due to the overall increase of average speed in the transport system. Finally, we perform a carbon emission analysis comparing the current scenario to a fully electrified road system. Our results show that, for the case of Singapore, while electricity generation produces twice as much CO2 as the cradle-to-gate emissions for petrol/diesel, electric vehicles still reduce the total carbon emissions of the road transport sector by 40%.

Keywords: Emission and Noise Mitigation, Simulation and Modeling, Road Traffic Control
Abstract: The design of Intelligent Intersection Management (IIM) schemes for fully Autonomous Vehicles (AVs) and mixed with Human-driven Vehicles (HVs) has focused mainly on throughput maximization and users' safety. However, new IIM strategies should consider environmental factors and human health conditions in their design, given their impact on fuel wastage and emission of dangerous air pollutants. In this paper, we compare the ecological footprint of two IMM protocols that follow opposite paradigms in handling AVs and HVs with internal combustion engine. We consider Round-Robin (RR) that favors the crossing of multiple consecutive cars from one road at a time and the recently proposed Synchronous Intersection Management Protocol (SIMP) that favors the crossing of multiple cars simultaneously, one from each road. Through experiments in the SUMO simulator, we observe that SIMP promotes more fluid traffic flows, causing traffic throughput to be up to 3.7 times faster and consume less fuel than the RR schemes, with similar results for vehicular emissions (PMx, NOx, CO, CO2, and HC).

Keywords: Sensing and Intervening, Detectors and Actuators, Traffic Theory for ITS, Theory and Models for Optimization and Control
Abstract: In this work we develop a kalman filtering approach for the problem of traffic state estimation in urban networks. The proposed approach employs concepts developed in the field of traffic flow observability, to derive both i) a minimal set of locations wherein traffic sensing infrastructure the network should be equipped and ii) topological relationships to be employed in the filtering technique's error covariance matrices, to improve the estimation process. A Linear Time-Variant formulation of first-order traffic flow theory is employed to model node-node vehicle propagation, allowing to predict the evolution of Cumulative Vehicle Numbers at intersections. This model is then embedded in the proposed cascading Kalman Filter framework. Validation of the proposed filtering approach is performed on a simple grid-like network, bearing considerable congestion, spillback and rerouting behaviour. We generate experimental data through a microscopic simulation software (SUMO).

Test results showcase how the proposed approach successfully exploits observability-based information to reconstruct data in unmeasured segments of the network. Particular care should however be devoted to appropriate inference of turning fractions at intersections, in order to achieve the lowest possible estimation error.

Keywords: Infrastructure for Charging, Communication and Controls, Electric Vehicles, Simulation and Modeling
Abstract: The major challenges for electric vehicle deployment include the prohibitive battery cost and a large amount of time required for energy refueling. Although the concept of battery swapping can solve the latter problem, it is difficult to overcome the former problem while ensuring a long-range drive. Thus, this paper reports upon a conceptual electric vehicle (BaaS-EV) that shares a part of the battery and presents an optimization model for its service infrastructure, battery as a service (BaaS). The BaaS-EV has two different batteries and operation modes. For short-range drive, the BaaS-EV can be driven using only one of the batteries, and to operate in a long range, an additional shared battery can be employed. Such shareability can increase the utilization rate of the battery and reduce the battery cost. The BaaS infrastructure provides BaaS-EV users with recharged batteries on demand. A cost-efficient operation of BaaS requires simultaneous optimization of the number of batteries, charging techniques, and transportation costs. To realize this aspect, an optimization model is established based on the minimum cost flow problem. Results of numerical simulations and case studies based on the model highlight the possibility of cost-benefit pertaining to the battery, user opportunity, battery charging, and transportation costs.

Keywords: Theory and Models for Optimization and Control, Driver Assistance Systems, Electric Vehicles
Abstract: Accurate long-distance route planners for electric vehicles could help to alleviate driving range anxiety. A time- and energy-optimal routing strategy with consideration of battery charging constraints is presented here. The accuracy of the approach is improved by modeling the impact of weather and traffic conditions on the vehicle's energy consumption, as well as by considering realistic charging functions. The routing problem is cast as a multi-objective optimization and solved with different algorithms to assess the accuracy and the tractability of each method. Results show that appealing trade-offs in terms of trip time and energy consumption appear when speed becomes a decision variable and is adjusted on some portions of the route.

Keywords: Simulation and Modeling, Network Management, Cooperative Techniques and Systems
Abstract: This study investigates the effects of the “white phase” on the performance of isolated signalized intersections. During the white phase, connected automated vehicles (CAV) control traffic flow through an intersection, and connected human-driven vehicles (CHV) follow their front vehicle (either CAV or CHV). Traffic controller ensures collision-free movement of vehicles through the intersection by determining 1) the sequence and duration of phases (green and white) and 2) trajectory of CAVs during white phases. White phases can be assigned to conflicting movements simultaneously. We have formulated this problem as a mixed-integer non-linear program (MINLP) and solved it using a receding horizon algorithm. Two demand patterns with five different CAV market penetration rates are used to evaluate the effects of the white phase on mobility and safety in an isolated intersection. Each case study is tested with three different control scenarios: 1) No-white-phase, 2) white-phase-only, and 3) optimal-white-phase activation (combination of white, green, and red phases). The results indicate that the white phase yields significant improvement in intersection performance while maintaining the same safety level

Keywords: Electric Vehicles, Infrastructure for Charging, Communication and Controls
Abstract: The environmental and economic advantages of renewable-energy technologies inspire efforts to encourage the use of electric vehicles (EVs) by business owners and individuals. Large-scale electric mobility is affected by inadequate charging infrastructure and battery technology. This study formulates a bi-level optimization program that aims to minimize the cost of EV charging facility deployment and utilization considering EV users’ travel and charging expenses. A hybrid methodology is developed that (i) converts the proposed formulation into an equivalent single-level formulation, (ii) implements an active-set based technique, and (iii) estimates travel costs using a macroscopic fundamental diagram (MFD) concept. Numerical experiments on an empirical case study show the performance of the proposed algorithm and some managerial insights. The results are also compared to a benchmark algorithm, which indicate that the proposed methodology can determine near-optimal solutions efficiently.

Keywords: Electric Vehicles, Personalized Public Transit, Network Management
Abstract: In the current paradigm of rapid urban growth, it is critical to create new and innovative mobility solutions that ensure an efficient, inexpensive and reliable flow of people throughout the city. In this paper, we propose the implementation of autonomous driving technology to address some of the current challenges of bicycle-sharing systems. We also propose a solution to provide bicycles of the necessary lateral stability to drive autonomously. Under this view, a fleet of autonomous shared bicycles would work as a mobility-on-demand system: when a user requested a ride, the nearest available bike would drive autonomously to wherever they are. Then, the user would ride it like a regular bike, and upon arrival to the destination, the bike would leave autonomously again for the next user. Such a system would eliminate the need for rebalancing and docking stations, and the needed fleet size would be smaller than in current station-based and dockless systems. In addition, it would solve the difficulty of finding available bicycles or docks and eliminate walking distance, improving the user experience and incentivizing its use.

Keywords: Energy Storage and Control Systems, Electric Vehicles, Simulation and Modeling
Abstract: Safety of the battery and charging systems are critical concerns for intelligent transportation systems.The increased power capacity and networking requirements in Extremely Fast Charging (XFC) systems and the resulting increase in the adversarial attack surface also call for security measures to be taken in the involved cyber-physical system (CPS) as a whole. This study proposes a moving-target defense (MTD) based novel approach for the battery management system (BMS) of an electric vehicle during the charging process, during which a compromised vehicle may contaminate the whole grid. In that sense, to our best knowledge, this is the first study in the literature on security-hardened BMS and charging process, aiming to increase the security of operations between the charging station, batteries and BMS of electric vehicles. The developed MTD strategies make use of redundancies in the controller and feedback design with novel switching strategies. The performed simulations demonstrate increased security and optimized charging performance under adversarial attacks.

Keywords: Infrastructure for Charging, Communication and Controls, Data Mining and Data Analysis, Electric Vehicles
Abstract: The aim of this paper is the technical and economic analysis of state-of-the-art charging infrastructure concepts for three typical commercial battery electric fleets. The investigated charging infrastructure concepts are based on systems that are currently available on the market and are thus an essential part of the current mobility change towards an emission-free mobility system. An iterative evaluation of the charging infrastructure concepts under variation of the mobility behavior of the three fleets enables statistical statements to be made about the technical and economic suitability. Based on the results, the authors can derive initial recommendations for the design of charging infrastructure for commercial vehicle fleets.

Keywords: Other Theories, Applications, and Technologies, Sensing, Vision, and Perception, Simulation and Modeling
Abstract: Neuromorphic event cameras are useful for dynamic vision problems under difficult lighting conditions. To enable studies of using event cameras in automobile driving applications, this paper reports a new end-to-end driving dataset called DDD20. The dataset was captured with a DAVIS camera that concurrently streams both dynamic vision sensor (DVS) brightness change events and active pixel sensor (APS) intensity frames. DDD20 is the longest event camera end-to-end driving dataset to date with 51h of DAVIS event+frame camera and vehicle human control data collected from 4000 km of highway and urban driving under a variety of lighting conditions. Using DDD20, we report the first study of fusing brightness change events and intensity frame data using a deep learning approach to predict the instantaneous human steering wheel angle. Over all day and night conditions, the explained variance for human steering prediction from a Resnet-32 is significantly better from the fused DVS+APS frames (0.88) than using either DVS (0.67) or APS (0.77) data alone.

Keywords: Sensing, Vision, and Perception, Automated Vehicle Operation, Motion Planning, Navigation
Abstract: Reliable outdoor deployment of mobile robots requires the robust identification of permissible driving routes in a given environment. The performance of LiDAR and vision-based perception systems deteriorates significantly if certain environmental factors are present e.g. rain, fog, darkness. Perception systems based on FMCW scanning radar maintain full performance regardless of environmental conditions and with a longer range than alternative sensors. Learning to segment a radar scan based on driveability in a fully supervised manner is not feasible as labelling each radar scan on a bin-by-bin basis is both difficult and time-consuming to do by hand. We therefore weakly supervise the training of the radar-based classifier through an audio-based classifier that is able to predict the terrain type underneath the robot. By combining odometry, GPS and the terrain labels from the audio classifier, we are able to construct a terrain labelled trajectory of the robot in the environment which is then used to label the radar scans. Using a curriculum learning procedure, we then train a radar segmentation network to generalise beyond the initial labelling and to detect all permissible driving routes in the environment.

Keywords: Sensing, Vision, and Perception, Sensing and Intervening, Detectors and Actuators, Other Theories, Applications, and Technologies
Abstract: This paper proposes the addition of a thermal camera to an RGB system with the goal of improving person and road detection reliability in unfavorable weather and illumination conditions. Custom data is gathered on an experimental vehicle and used for development and testing. For person detection, we propose a novel multi-modal approach, where bounding boxes are initially obtained from RGB and thermal images using YOLOv3-tiny. We then identify high-intensity connected components in thermal images to compensate for missed detections. Detections from the two cameras and the two algorithms are finally weighed and combined into a confidence map. Using the proposed fusion method, recall and precision are improved compared to using RGB only, without the need to retrain the network. For thermal-based road segmentation, we achieve an average precision of 94.2% after re-training MultiNet's KittiSeg decoder on a small thermal dataset, while using pre-trained weights for MultiNet's VGG-based encoder. These results show that the addition of thermal cameras to perception systems of autonomous vehicles can bring substantial benefits with minimal labelling, implementation effort and training requirements.

Keywords: Rail Traffic Management, Travel Information, Travel Guidance, and Travel Demand Management, Theory and Models for Optimization and Control
Abstract: Interconnection operation of rail transit line has become a key means to address the long-distance and diversified travel de-mands in the rail transit region. The influencing mechanism of the interconnection operation on the capacity of urban rail transit network is complex. In this study, the operation of off-line train is analyzed after interconnection. The capacity is mainly impacted during the stages of transportation scheduling and rescheduling. Three types of off-line train operation modes, namely, running only one off-line train (M-1), continuously running multiple off-line trains (M-2), and operating off-line and in-line trains at a certain proportional interval (M-3), are studied. The methods to reduce the impact on capacity during transportation planning and operation scheduling are put for-ward. A suburban railway and an urban rail line in Cheng-du-Chongqing area are selected as the research cases to verify our research conclusions. We found that the operation of off-line trains has strict requirements on the departure interval of the line. We further found that the mode M-2 is suitable for the lines with tight capacity, and the mode M-3 is suitable for off-peak periods when the transportation capacity is relatively abundant.

Keywords: Theory and Models for Optimization and Control, Rail Traffic Management, Simulation and Modeling
Abstract: This paper investigates the distributed optimization of train scheduling and waiting passengers control problem for metro lines to improve the operational efficiency and service level. Integrating the train departure time and waiting passengers, a multi-train coupling state-space model is established. To improve computational efficiency, we propose a distributed optimization algorithm to decompose the complex joint optimization problem of multiple trains into several subproblems that can be computed in parallel. Based on the alternating direction method of multiplier (ADMM) algorithm, the coupling constraints between trains are effectively eliminated, and each subproblem is solved through iterative coordination with other subproblems. Numerical examples illustrate the effectiveness of the proposed distributed optimization algorithm for train automatic train regulation and waiting passengers control.

Keywords: Rail Traffic Management, Data Mining and Data Analysis, Theory and Models for Optimization and Control
Abstract: In urban rail transit systems, the intelligent driving system is gradually replacing manual driving for its high safety, punctuality, and stopping accuracy. With the development of big data analytics, the data-driven intelligent driving system becomes a research focus. Traditional data-driven intelligent driving systems suffer from inadequate data. Due to lacking effective incentives and trust, data from different urban rail operators cannot be shared directly. In this paper, we propose a framework that uses blockchain technology to realize sharing and collaborative training of intelligent driving models between operators. In this framework, we use blockchain-based distributed federated reinforcement learning methods to complete intelligent driving calculations. We use smart contracts to implement the management of the entire federal reinforcement learning. Operators use local historical data to participate in intelligent driving training based on Q-network by exchanging encrypted model parameters, and optimize the safety distance, energy consumption, and punctuality of urban rail transit systems. Simulation results show that our proposed distributed federated reinforcement learning method can significantly improve the intelligent driving system performance.

Keywords: Data Mining and Data Analysis
Abstract: Although much research work for the operation safety has been taken in the railway domain, some accidents still occur because past experiences of accident analysis were not fully accumulated for safety improvement. This study aims to identify potential causal relationships among the many factors playing a role in railway accidents. A new interestingness measure, Confidence_interestingness ("C_" Inter) and corresponding improved algorithm, Positive and Negative Association Rules Algorithm based on "C_" Inter (PNARA_CI) were put forward in our study. Compared with traditional association rule mining algorithms, the PNARA_CI does not generate candidate association rules by means of frequent itemsets, but by the combination between every two accident factors, which can mine the positive and negative association rules with practical value to the maximum. And they were applied to railway accidents data to explore the association rules of the causal factors in the case study. The effectiveness of the algorithm was verified.

Keywords: Rail Traffic Management, Theory and Models for Optimization and Control, Simulation and Modeling
Abstract: Aiming at the difficult point of generating driving strategy of heavy-haul trains under moving block system, a driving strategy generation method based on fuzzy predictive control is proposed for alleviating the labor intensity of drivers and better guaranteeing the operation safety of heavy-haul trains. Firstly, by analyzing the difficulties of driving heavy-haul trains in natural sections and steep slopes, constraint models of driving strategy are established, and a speed curve optimization algorithm based on adaptive step size is designed. Then, the target speed curve is obtained by locally optimizing the driving strategy in difficult sections based on the principle of time equivalence and a fuzzy predictive controller is designed. Finally, use the actual train data and line data to validate the method. The results show that compared to the traditional PID control method, the method proposed in this paper can control the train to run more placidly and the maximum error of tracking target speed is ±0.23m∙s-1, which proves that the proposed method is feasible.

Keywords: Data Mining and Data Analysis, Network Modeling, Other Theories, Applications, and Technologies
Abstract: Urban rail transit systems produce big automatic fare collection (AFC) data. With the support of big data analytics, more accurate forecasting of passenger flow in the urban rail transit system can help optimize the train operation timetable and alleviate urban traffic congestion. In the lack of adequate data, most of the existing works study the urban rail passenger forecasting using time series analysis. The historical passenger data at each station is used to predict the future passenger flow at each independent station, where the passenger flow correlation between time and station is largely ignored. In this paper, we predict the passenger flow using the data fusion of time and space. We obtain the historical pristine passenger flow data from the real automatic fare collection system. Teradata big data platform is used to process the data and obtain the spatiotemporal fusion data. Spatiotemporal passenger flow dynamics is converted to a two-dimensional time-space matrix describing the time and space relations of passenger flow. A convolutional neural network model is established after the two-dimensional time-space matrix. We obtain the optimal hyperparameter combinations of CNN models using the grid search algorithm. The performance of the CNN model is evaluated using real metro data. The simulation results demonstrate the efficiency and accuracy of the proposed method.

Keywords: Rail Traffic Management, Data Mining and Data Analysis, Network Modeling
Abstract: Accurate passenger flow forecasting is vital for passenger flow management and planning. However, it is a challenging task in practice as passenger flow of a certain transportation network is affected by complex factors including the unstructured spatial dependencies constrained by the transportation network topological structure, intra-location correlations (inflow relates to outflow), temporal dependencies, and exogenous factors. To cope with the aforementioned challenges, this paper proposes a novel deep learning-based spatiotemporal passenger flow forecasting model, named Graph Convolutional-Long Short Term Memory (GC-LSTM). The designed architecture of GC-LSTM extends convolution with Graph Convolutional Network (GCN) to handle graph-based spatial dependencies, while LSTM in the architecture is employed to capture the long-term temporal dependencies as well as nonlinear traffic dynamics. The proposed method also enables collectively forecasting of inflow and outflow at the location of interest within transportation network by capturing the intra-location correlations in parallel views. Then the proposed method is validated by the real-world passenger flow data of China High-Speed Rail (HSR) network, and the experimental results show that GC-LSTM can well capture the graph-based spatial and temporal dependencies and outperform state-of-art baselines in terms of forecasting accuracy.

Keywords: Intelligent Logistics, Theory and Models for Optimization and Control, Off-line and Online Data Processing Techniques
Abstract: Multitasking optimization is a recently introduced paradigm, focused on the simultaneous solving of multiple optimization problem instances (tasks). The goal of multitasking environments is to dynamically exploit existing complementarities and synergies among tasks, helping each other through the transfer of genetic material. More concretely, Evolutionary Multitasking (EM) regards to the resolution of multitasking scenarios using concepts inherited from Evolutionary Computation. EM approaches such as the well-known Multifactorial Evolutionary Algorithm (MFEA) are lately gaining a notable research momentum when facing with multiple optimization problems. This work is focused on the application of the recently proposed Multifactorial Cellular Genetic Algorithm (MFCGA) to the well-known Capacitated Vehicle Routing Problem (CVRP). In overall, 11 different multitasking setups have been built using 12 datasets. The contribution of this research is twofold. On the one hand, it is the first application of the MFCGA to the Vehicle Routing Problem family of problems. On the other hand, equally interesting is the second contribution, which is focused on the quantitative analysis of the positive genetic transferability among the problem instances. To do that, we provide an empirical demonstration of the synergies arisen between the different optimization tasks.

Keywords: Simulation and Modeling, Electric Vehicles, Data Mining and Data Analysis
Abstract: The electric vehicle (EV) is experiencing rapid growth in today's commercial mobility service market, from carrying passengers to the delivery of goods. One cardinal issue concerning the operation of EVs is the optimal routing of EV fleets with limited battery capacity. In this study, we investigate the energy-optimal online routing problem for the fleet of EVs, which focuses on identifying real-time minimum electricity consumption paths (MECP) for multiple OD pairs with limited information. We develop a multi-OD combinatorial multi-arm semi-bandit model (MCMAB) that uses the fleet of EVs as sensors in the transportation network and promotes the utilization of common information shared by different OD pairs. We further enrich the model with the path elimination policy to obtain MECP of high confidence while significantly reducing the number of learning iterations and the number of explorations needed. We demonstrate the effectiveness of the MCMAB and the efficiency of the path elimination policy with comprehensive numerical experiments in Manhattan, NYC. The results show that the proposed online routing algorithms can achieve near-optimal MECPs efficiently, and the quality of the solutions is significantly better than using the shortest travel time paths as approximate MECPs.

Keywords: Data Mining and Data Analysis, Off-line and Online Data Processing Techniques, Emission and Noise Mitigation
Abstract: This paper describes a novel method to estimate vehicle speed trajectory at a road-link scale. Road network infrastructure and topology have a significant impact on the vehicle's driver behavior. We propose a dynamic speed estimation based on road macroscopic features, available anywhere through digital maps webservices. The method combines machine learning techniques and stochastic approaches to construct dynamic speed trajectories. This allows the estimation of several speed trajectories per road-segment to take into account several possible driving behaviors. It has been trained and validated on a database of 50 million kilometers of 1 Hz driving recordings coming from a crowdsensing project. The model is compared to these driving recordings and promising results are obtained in terms of reproducing real driving behavior. A map of traffic pollutant emission is presented as a first application.

Keywords: Driver Assistance Systems, Automated Vehicle Operation, Motion Planning, Navigation, Simulation and Modeling
Abstract: Car-following is a fundamental operation for vehicle control for both ADAS on modern vehicles and vehicle control on autonomous vehicles. Most existing car following mechanisms react to the observations of nearby vehicles in real-time. Unfortunately, lack of capability of taking into account multiple constraints and objectives, these mechanisms lead to poor efficiency, discomfort, and unsafe operations. In this paper, we design and implement a proactive car-following model to take into account safety regulation, efficiency, and comfort using deep reinforcement learning. The evaluation results show that the proactive model not only reduces the number of inefficient and unsafe headway but also eliminates the traffic jerk, compared to human drivers. The model outperformed 79% human drivers in public data set and the road efficiency is only 2% less than the optimal bound.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Sensing, Vision, and Perception, Other Theories, Applications, and Technologies
Abstract: Global Navigation Satellite Systems (GNSS) is the positioning technology of choice outdoors but it has many limitations to be used in safety-critical applications such Intelligent Transportation Systems (ITS). Namely, its performance clearly degrades in harsh propagation conditions, these systems are not reliable due to possible attacks, may not be available in GNSS-denied environments, and using standard architectures do not provide the precision needed in ITS. Among the different alternatives, Ultra-WideBand (UWB) ranging is a promising solution to achieve high positioning accuracy. The key points impacting any time-of-arrival (TOA) based navigation system are i) transmitters’ geometry, and ii) a perfectly known transmitters’ position. In this contribution we further analyze the performance loss of TOA-based navigation systems in real-life applications where we may have both transmitters' position mismatch and harsh propagation conditions, i.e., measurements corrupted by outliers. In addition, we propose a new robust filtering method able to cope with both effects. Illustrative simulation results are provided to support the discussion and show the performance improvement brought by the new methodology with respect to the state-of-the-art.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Accurate Global Positioning
Abstract: In standard two-step Global Navigation Satellite Systems (GNSS) receiver architectures the precision on the position, velocity and time estimates is driven by the precision on the intermediate parameters, i.e., delays and Dopplers. The estimation of the time-delay is in turn driven by the baseband signal resolution, that is, by the type of broadcasted signals. Among the different GNSS signals available the so-called AltBOC modulated signal, appearing in the Galileo E5 band and the new GNSS meta-signal concept, is the one which may provide the better time-delay precision. In order to meet the constraints of safety-critical applications such as Intelligent Transportation Systems or automated aircraft landing, it is fundamental to known the ultimate code-based precision achievable by standalone GNSS receivers. The main goal of this contribution is to assess the time-delay precision of AltBOC type signals. The analysis is performed by resorting to a new compact closed-form Cramér-Rao bound expression for time-delay estimation which only depends on the signal samples. In addition, the corresponding time-delay maximum likelihood estimate is also provided to assess the minimum signal-to-noise ratio that allows to be in optimal receiver operation.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Accurate Global Positioning, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations
Abstract: To accommodate the increasing intensity of port operations, the introduction of automated straddle carriers (ASCs) is considered an innovative direction in the management of ports. However, ASCs rely on satellite navigation systems for positioning, which is accompanied by limitations. The purpose of this paper is to present an algorithmic approach to fleet management, routing and collision avoidance and use the algorithm to study the effects of GNSS limitations to its results. The designed algorithm provides information on assignment, routing and speed profiles for the Automated Straddle Carriers with the purpose of avoiding collisions. Five different scenarios were simulated with experimental data. The results indicated that when noise is introduced to more than one places, then more time is necessary to complete the tasks. Finally, differences in the results may appear insignificant, however, in an automated environment they could increase the volatility and create cascading effects that increase the risk.

Keywords: Cooperative Techniques and Systems, Accurate Global Positioning, Automated Vehicle Operation, Motion Planning, Navigation
Abstract: Global Navigation Satellite Systems (GNSS) constitute the cornerstone for outdoor positioning, which is essential information for prospective automated vehicles. The combination of GNSS with terrestrial ranging, for instance in the form of 5G or UWB, will make accurate positioning a reality even in urban canyon scenarios where GNSS is likely to fail. Thus, hybrid GNSS-terrestrial localization in collaborative networks has become a hotspot for the research community. This paper discusses the Cramér-Rao Bound (CRB) as lower bound for location estimates and evaluates two snapshot estimators, one deterministic and the other Bayesian, for distributed and centralized localization in cooperative networks. The performance of the estimators is evaluated with respect to the presented CRB in a simulated network of mobile and anchor agents, and the role played by agent-to-agent and agent-to-anchor ranging is discussed.

Keywords: Accurate Global Positioning, Automated Vehicle Operation, Motion Planning, Navigation, Sensing, Vision, and Perception
Abstract: The knowledge of the map-relative pose of a vehicle is crucial for automated driving. This paper presents a localization approach using a graph-based sliding window optimization with non-semantic LiDAR features. In short, this paper introduces and evaluates our overall concept of non-semantic localization, whose idea and steps were depicted in previous publications. In doing so, we do not rely on static, semantic landmarks, like building corners, thus we achieve independence of these infrastructure elements. Our approach consists of three main steps. At first, we collect the input data, i.~e. extracting on-board features from LiDAR point clouds. Afterward, the on-board features are associated over time and matched with the features stored in the map. Then, we use the on-board detected features, map matched features, odometry and GNSS measurements to build a factor graph. The factor graph is optimized to estimate the vehicle pose. Our approach is tested on real-world data. The experiment suggests that our method of applying non-semantic elements provides a working localization solution with practical relevance. It also displays that our approach is persistent examining test drives over one and half years along the same route.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Simulation and Modeling
Abstract: While on-board sensing equipment of CAVs can reasonably characterize the surrounding traffic environment, their performance is limited by the range of the sensors. By integrating short- and long-range information, a CAV can comprehensively construct its surrounding environment, thereby allowing it to plan both short and long-term maneuvers. Coalescing local information and downstream information is critical for the CAV to make safe and effective driving decisions. While literature is replete with CAV control approaches that use information sensed from the local traffic environment, studies that fuse information from various temporal-spatial instances to facilitate CAV movements is limited. In this paper, we propose a Deep Reinforcement Learning (DRL) based approach that fuses information obtained (via sensing and connectivity) on the local downstream environment for CAV lane changing decisions. We adopt Artificial Intelligence (AI) techniques to provide an end-to-end solution that incorporates the information fusion and movement-decision processor. We also determine the optimal connectivity range for each operating traffic density range. We anticipate that deployment of the proposed algorithm in a CAV will facilitate reliable proactive driving decisions and ultimately enhance the overall operational efficiency of CAVs in terms of safety and mobility.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation
Abstract: Planning is an essential topic in the realm of automated driving. Besides planning algorithms that are widely covered in the literature, planning requires different software tools for its development, validation, and operation. This paper presents a survey of such tools including map representations, communication, traffic rules, open-source planning stacks and middleware, simulation, and visualization tools as well as benchmarks. We start by defining the planning task and different supporting tools. Next, we provide a comprehensive review of state-of-the-art developments and analysis of relations among them. Afterwards, a systematic method to opt for superior tools with respect to specific planning tasks is proposed. Finally, we discuss the current gaps and suggest future research directions. The survey as well as methodology of selecting tools can speed up the pace of planning research for automated driving.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation
Abstract: In this paper, a nonlinear model predictive control (NMPC) scheme for path tracking of autonomous vehicles using discrete previewed points in the inertial coordinate is presented. The control object is to improve the tracking accuracy under small lateral acceleration in the scenario of low speed and narrow space. The path tracking problem is formulated as a nonlinear model predictive control model, in which the vertical distance between the vehicle position and the tangent of previewed point is adopted to evaluate the tracking error. The discrete previewed points are generated from the path points independently, which does not rely on the approximate path functions. The iterative initial values of the optimization model are appropriately selected by combining with the Stanley method to accelerate optimization computations. A simulation comparison between the proposed NMPC controller and a linear model predictive control (LMPC) controller is conducted through Carsim-Matlab/Simulink co-simulations. Simulation results show that the proposed controller exhibits better tracking accuracy than the LMPC controller under small lateral acceleration, especially when the path curvature is large and continuously changing.

Keywords: Automated Vehicle Operation, Motion Planning, Navigation, Advanced Vehicle Safety Systems, Driver Assistance Systems
Abstract: The safety validation of motion planning algorithms for automated vehicles requires a large amount of data for virtual testing. Currently, this data is often collected through real test drives, which is expensive and inefficient, given that only a minority of traffic scenarios pose challenges to motion planners. We present a workflow for generating a database of challenging and safety-critical test scenarios that is not dependent on recorded data. First, we extract a large variety of road networks across the globe from OpenStreetMap. Subsequently, we generate traffic scenarios for these road networks using the traffic simulator SUMO. In the last step, we increase the criticality of these scenarios using nonlinear optimization. Our generated scenarios are publicly available on the CommonRoad website.

Keywords: Accurate Global Positioning, Automated Vehicle Operation, Motion Planning, Navigation, Driver Assistance Systems
Abstract: Autonomous vehicles and driver assistance systems utilize maps of 3D semantic landmarks for improved decision making. However, scaling the mapping process as well as regularly updating such maps come with a huge cost. Crowdsourced mapping of these landmarks such as traffic sign positions provides an appealing alternative. The state-of-the-art approaches to crowdsourced mapping use ground truth camera parameters, which may not always be known or may change over time. In this work, we demonstrate an approach to computing 3D traffic sign positions without knowing the camera focal lengths, principal point, and distortion coefficients a priori. We validate our proposed approach on a public dataset of traffic signs in KITTI. Using only a monocular color camera and GPS, we achieve an average single journey relative and absolute positioning accuracy of 0.26 m and 1.38 m, respectively.

Keywords: Sensing, Vision, and Perception, Advanced Vehicle Safety Systems
Abstract: Understanding and predicting pedestrian behavior is an important and challenging area of research for realizing safe and effective navigation strategies in automated and advanced driver assistance technologies in urban scenes. This paper focuses on monocular pedestrian action recognition and 3D localization from an egocentric view for the purpose of predicting intention and forecasting future trajectory. A challenge in addressing this problem in urban traffic scenes is attributed to the unpredictable behavior of pedestrians, whereby actions and intentions are constantly in flux and depend on the pedestrians pose, their 3D spatial relations, and their interaction with other agents as well as with the environment. To partially address these challenges, we consider the importance of pose toward recognition and 3D localization of pedestrian actions. In particular, we propose an action recognition framework using a two-stream temporal relation network with inputs corresponding to the raw RGB image sequence of the tracked pedestrian as well as the pedestrian pose. The proposed method outperforms methods using a single-stream temporal relation network based on evaluations using the JAAD public dataset. The estimated pose and associated body key-points are also used as input to a network that estimates the 3D location of the pedestrian using a unique loss function. The evaluation of our 3D localization method on the KITTI dataset indicates the improvement of the average localization error as compared to existing state-of-the-art methods. Finally, we conduct qualitative tests of action recognition and 3D localization on HRI's H3D driving dataset.

Keywords: Sensing, Vision, and Perception
Abstract: On-board 3D object detection in autonomous vehicles often relies on geometry information captured by LiDAR devices. Albeit image features are typically preferred for detection, numerous approaches take only spatial data as input. Exploiting this information in inference usually involves the use of compact representations such as the Bird's Eye View (BEV) projection, which entails a loss of information and thus hinders the joint inference of all the parameters of the objects' 3D boxes. In this paper, we present a fully end-to-end 3D object detection framework that can infer oriented 3D boxes solely from BEV images by using a two-stage object detector and ad-hoc regression branches, eliminating the need for a post-processing stage. The method outperforms its predecessor (BirdNet) by a large margin and obtains state-of-the-art results on the KITTI 3D Object Detection Benchmark for all the categories in evaluation. Source code is available at https://github.com/AlejandroBarrera/BirdNet2.

Keywords: Sensing, Vision, and Perception, Other Theories, Applications, and Technologies
Abstract: 3D object detection based on monocular camera data is a key enabler for autonomous driving. The task however, is ill-posed due to lack of depth information in 2D images. Recent deep learning methods show promising results to recover depth information from single images by learning priors about the environment. Several competing strategies tackle this problem. In addition to the network design, the major difference of these competing approaches lies in using a supervised or selfsupervised optimization loss function, which require different data and ground truth information. In this paper, we evaluate the performance of a 3D object detection pipeline which is parameterizable with different depth estimation configurations. We implement a simple distance calculation approach based on camera intrinsics and 2D bounding box size, a self-supervised, and a supervised learning approach for depth estimation.

Ground truth depth information cannot be recorded reliable in real world scenarios. This shifts our training focus to simulation data. In simulation, labeling and ground truth generation can be automatized. We evaluate the detection pipeline on simulator data and a real world sequence from an autonomous vehicle on a race track. The benefit of training on simulation data for the application of the network on real world data is investigated. Advantages and drawbacks of the different depth estimation strategies are discussed.

Keywords: Sensing, Vision, and Perception
Abstract: 3D object detection serves as a crucial basis of visual perception, motion prediction, and planning for automated driving. To apply an algorithm for this purpose, the detection of all types of road users in real-time is an essential condition. In this paper, we propose an approach that projects the 3D points of image-based bounding box proposals into so-called grid map patches. These patches are used to estimate the exact dimensions of the 3D box with the help of a lightweight CNN. The complete proposed processing chain is parallelized and implemented on a GPU. This makes our approach the fastest stereo-based 3D object detector on the KITTI benchmark while still achieving results that are within the range of the best image-based algorithms.

Keywords: Sensing, Vision, and Perception, Modeling, Simulation, and Control of Pedestrians and Cyclists, Driver Assistance Systems
Abstract: Pedestrian trajectory prediction plays an important role in both pedestrian collision avoidance systems and autonomous driving. However, most of the previous works have ignored the interaction between traffic participants or only take it into account implicitly based on neural networks, which need a large number of training data and hold poor scenario adaptability. Meanwhile, pedestrian changeable behaviors are also always overlooked in trajectory prediction. In this paper, we present a novel pedestrian trajectory prediction method that involves pedestrian intention and behavior information into prediction. Verification of this method has been conducted in our provided BPI dataset. Without previous training of pedestrian trajectories, the method shows good scenario adaptability and provides accurate path prediction results for eight defined typical pedestrian crossing-road scenarios in 1s prediction horizon, especially for stopping scenarios.

Keywords: Cooperative Techniques and Systems, Automated Vehicle Operation, Motion Planning, Navigation, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations
Abstract: In order to generate favorable trajectories, road users need to cope with interaction among them, especially in dense traffic. Thus, for autonomous cars, the intention of involved vehicles needs to be considered in their motion planning. This paper proposes a general framework for cooperative interaction aware trajectory generation based on multi-agent trajectory planning. Possible intentions are distinguished by different cost functions, resulting in different behaviors such as cooperative or non-cooperative. Given observations, Bayesian estimation is used to obtain a probability distribution of the intention models. Considering these probabilities during prediction and planning results in trajectories taking the uncertain interaction with surrounding vehicles into account. The performance of the approach is demonstrated via numerical experiments for a lane change scenario in dense traffic.

Keywords: Cooperative Techniques and Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Advanced Vehicle Safety Systems
Abstract: Vehicular ad-hoc networks (VANET) are often assumed in cooperative driving scenarios, such as vehicle platooning, in order to achieve better control quality and enhanced autonomy. The increased connectivity and software programmability required for this features exposes new attack surfaces, escalating vehicular security risks and potentially compromising the safety of road users. Given the safety critical nature of such applications, the ability to verify the authenticity and integrity of the communicated data as well as to detect malicious driving behaviors is of paramount importance. In practice, such assurances are hard to derive since carefully crafted stealthy attack signals aimed at violating safety properties, e.g. safe inter-vehicle distance, are difficult to distinguish from normal operation if only data from on-board sensors is used by the detection system. To cope with the problem of detecting stealthy attacks on communicated data and the control algorithm in cooperative vehicle platoons, we propose a novel detection technique based on a state estimator that takes advantage of communicated data from multiple vehicles connected over VANET.

Keywords: Cooperative Techniques and Systems, ITS Field Tests and Implementation, ITS Policy, Design, Architecture and Standards
Abstract: Cooperative Intelligent Transportation Systems (C-ITS) are technologies that enable vehicles to communicate with each other and with the road infrastructure. These innovative technologies enable road users and traffic managers to share useful information, assisting the coordination of their actions. During the last years various initiatives providing policy rules for C-ITS deployment and a large number of projects demonstrating C-ITS implementation have taken place in Europe and USA. However, the identification of the status of C-ITS deployment remains ambiguous at binational level. The purpose of this paper is to provide an overview of the European and US milestones, that have been reached so far in the field of C-ITS, by identifying and reporting the policy framework, as well as the projects concerning C-ITS deployment in Europe and USA.

Keywords: Cooperative Techniques and Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Advanced Vehicle Safety Systems
Abstract: The cooperative perception among connected autonomous vehicles extends the field-of-view of the individual cars and adds significantly to their sensing and collision avoidance capabilities. This feature is particularly useful and essential in avoiding collisions with pedestrians, vulnerable road users, and other objects or cars which are obscured in the typical field-of-view of an ego vehicle. This paper proposes a simple to implement but effective longitudinal control algorithm to avoid collisions in a dynamic environment for cooperative autonomous vehicles. The algorithm is applied to ego and lead vehicles to control longitudinal dynamics with appropriate braking based on safety distance modeling. Simulations using dynamic models for both vehicles and pedestrians on a hazardous traffic scenario are presented to illustrate the effectiveness of the proposed control algorithm. The proposed method is also capable of warning and avoiding collisions for several other critical situations that may appear in autonomous driving. The results demonstrate a promising solution for cooperative collision avoidance, which can be further expanded to more complex scenarios.

Keywords: Cooperative Techniques and Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Automated Vehicle Operation, Motion Planning, Navigation
Abstract: The advanced controls of Connected Autonomous Vehicles (CAVs) have enabled complex traffic management for higher efficiency and safety, such as dynamic lane assignment and non-signalized autonomous intersection management (AIM). Unlike traditional intersections, non-signalized AIM does not rely on the settings of signal phases or traffic-light cycles for controlling the traffic, and the measurements of lane congestion can no longer be based on these two settings. Moreover, in the CAV environment, lane use can be more dynamic in response to real-time traffic demands and the policy of AIM. Therefore, we proposed a new AIM system (named Roadrunner) that combines dynamic lane assignment and autonomous intersection management. In Roadrunner, lane use is not limited by turns, and lanes are dynamically assigned to CAVs, based on our lane-assignment policy and a novel approach of measuring lane congestion levels over non-signalized intersection controls. We implemented Roadrunner in a SUMO simulator and compared the performance with different intersection management systems, where lane use is predefined. The result shows that Roadrunner increases the intersection capacity by more than 11%, and CAVs have lower average traveling delay.

Keywords: Rail Traffic Management, Management of Exceptional Events: Incidents, Evacuation, Emergency Management, Other Theories, Applications, and Technologies
Abstract: Complicated railway networks and high-density train service in common urban areas have led to difficulties in railway operation after incidents. Supporting crew rescheduling task in disruption management is important for realizing reliable services. In crew rescheduling task, flexible responses are required depending on the situation, but with common methods whose objectives and constraints are given, getting an appropriate crew plan in every case is hard. Therefore, we developed a support function by which railway operators can compare multiple feasible crew plans and select one that meets the current situations. To realize this function, we propose a method of constructing a zero-suppressed binary decision diagram (ZDD) that expresses a set of multiple feasible crew plans. We compared the performance of the proposed function with the conventional way in the case of medium-sized railway line and confirmed that our function is suitable for situations in which interactive comparisons are performed.

Keywords: Rail Traffic Management
Abstract: For the driving curve optimization of heavy haul trains with pneumatic braking, one of the biggest challenges is how to generate a driving curve on the long steep downward slope. Since the train velocity will increase rapidly due to a long releasing time and unsuitable position of applying the air braking will influence the operational efficiency or may cause accidents. In this paper, the optimization method for the driving curve of heavy haul trains is investigated. Based on the train dynamic model, an optimization model is firstly formulated, taking the energy consumption, running time and distance of pneumatic braking as the objectives. Practical constraints are considered, including the releasing time, velocity, minimum hold-on time and conversion rules of regimes. To solve this problem, the complex constraints are combined into the objective function by using penalty functions. Then, the artificial bee colony (ABC) algorithm is introduced to find the proper switching points of different regimes. Simulations are conducted based on the practical data of Shuohuang railway line and the results confirmed the effectiveness of the proposed algorithm.

Keywords: Rail Traffic Management, Theory and Models for Optimization and Control, Cooperative Techniques and Systems
Abstract: According to vehicle-centralized train control de- sign, instead of traditional interlocking equipment’s unified management of rail resources, train can control resources based on plan autonomously, which asks for the capability of trains to detect and resolve the potential requisition conflict caused by multi-train decentralized control. This paper proposes a real- time autonomous conflict detection and resolution (rtACDR) model based on cooperative game theory. Firstly, the occupancy time of trains at sections is predicted by using blocking time theory, a criterion for the existence of conflict is also described. When potential conflict is detected, factors considered during dispatching and driving are weighted and described as an individual characteristic function, which represents relevant train’s gaming revenue. At last, based on the Shapely theory, train’s reordering at conflict area is transformed into the optimization problem of cooperative alliance’s revenue, and solved by corresponding driving strategy’s generation. The computational tests are performed on rail network of Cao Qiao Station in Beijing with various perturbations. Based on the detailed analysis of trains’ driving strategies generated by model and the comparison of model and traditional dispatch- ing strategy’s performances, feasibility of proposed model is estimated preliminarily.

Keywords: Rail Traffic Management, Network Management, Incident Management
Abstract: This paper tackles railway timetabling with infrastructure work possessions. It introduces the integrated Passenger and Freight Train Timetable Adjustment Problem (PF-TTAP) which handles both passenger as well as freight trains. To deal with possessions, passenger trains are typically retimed, reordered or partially cancelled, while for freight trains it is important to reach their destination, possibly using an alternative path. Alternative paths for freight trains are generated using the k-shortest path algorithm. To solve the PF-TTAP, a mixed integer linear programming (MILP) problem is developed to simultaneously retime, reroute and cancel trains in the network. The model aims at minimizing deviations from the original timetable and in particular selecting alternative freight paths with the least turning activities and non-commercial stops. The model was tested on the Dutch national railway network. The PF-TTAP model successfully created an alternative hour pattern satisfying all the railway stakeholders.

Keywords: Rail Traffic Management, Theory and Models for Optimization and Control, Simulation and Modeling
Abstract: In metro system, unpredictable disturbances influence the normal operation and bring much inconvenience to passengers. This paper focuses on train timetable rescheduling (TTR) problem with considering practical operations in the metro management. At first, an optimization model that takes the deviation of rescheduled timetable, the total delay time of passengers and energy consumption as objective is developed. Meanwhile, the constraints, together with some practical rescheduling rules (e.g., the preprogrammed speed profiles, train detention strategy) are introduced. Secondly, the model is reformulated into an Markov decision process (MDP) with well defining the state, action and reward function, which is then solved by the proposed Q-learning approach. Finally, some case studies using the operational data of Beijing Yizhuang Subway Line are carried out to demonstrate the effectiveness of the proposed approach. The results indicate that a tradeoff solution among the optimization objectives can be obtained within a short time.

Keywords: Network Modeling, Theory and Models for Optimization and Control, Public Transportation Management
Abstract: The transits constitute the traffic skeleton of the city. With the installation of the high-tech sensors and the emergence of the high-speed communication, interest to grant bus priority via traffic signals grows rapidly during recent years. Pre-signal is an additional traffic signal located upstream of the main signal, which is used to discontinue the private cars and accordingly allow buses jumping the car queue. Compared with the traditional dedicated bus lane through the whole link, the pre-signal can provide bus priority meanwhile minimizing the negative impacts on car traffic. In this paper, we propose an adaptive signal strategy for a traffic network by adjusting the pre-signal as well as the main signal in order to minimize the total passenger delay. A macroscopic model based on cell transmission model is developed, where the link is partitioned into the car cell, the bus cell and the mixed cell. The merging and the diverging of the car flow and the bus flow are elaborated modeled for each type of the cell. With the aim to obtain the signal setting in real time, the harmony search algorithm is adopted to solve the optimization problem. Finally, the case studies illustrate the efficacy of the proposed strategy.

Keywords: Network Modeling, Off-line and Online Data Processing Techniques, Data Mining and Data Analysis
Abstract: Lane line is a very important element in HD maps, and map updating based on information can effectively reduce production cost. We use the images obtained by crowdsourcing for information mining. Most of these images are discontinuous and there are no internal or external parameters. However, lane detection algorithms are mostly applied to the vehicle, which are not suitable to detect road changed information. We propose a lane line perception network for information discovery, which directly takes the returned image as input and outputs the number of lane lines, as well as the color and type attributes of each lane. In contrast to previous works, we have solved the gradient explosion problem and specially optimized type segmentation. Finally, the proposed method is applied to mine information about lane changes.

Keywords: Network Modeling, Simulation and Modeling, Other Theories, Applications, and Technologies
Abstract: Floating Car Data (FCD) represents a data source with exclusive features not available in conventional sources. FCD serves as motion sensors that can provide rich input for traffic demand models. An accurate traffic demand estimation is fundamental for many transportation-related applications. The goal of this research is to exploit the valuable information provided by FCD to enhance the accuracy and reduce the complexity of the traffic demand estimation process. We used the information minimization model to estimate origin-destination matrices. This model requires multiple inputs; namely, a seed matrix, route choice information, and traffic counts. We obtain the first two inputs directly from FCD. We also used FCD as a measure of attractiveness in the gravity model to calculate traffic volumes on turns to increase the number of available link traffic counts. A simulation and a field study are performed to depict the effect of FCD on the estimation process. All of the inputs and the reference data of the field study are real data gathered from different sources. The results of the GEH test and the correlation coefficient confirm that the use of FCD in the proposed model leads to an improvement of the OD-estimation in terms of accuracy and calculation time.

Keywords: Network Modeling, Theory and Models for Optimization and Control
Abstract: A traffic flows optimal control problem in an urban road network is considered. It is assumed that all information on the traffic flows and maneuvers is known. The optimal control problem is to find duration of signal phases at controlled intersections that provide optimal traffic estimation in the considered road network. Optimization criteria include penalties for violation of constraints. To find the optimal control program a variation genetic algorithm is used. The traffic flow model under study is a mathematical model based on the controlled networks theory. Some properties of the model are discussed. The problem of determining the critical states is formulated. An optimal control problem for a network of four controlled intersections is solved.

Keywords: Public Transportation Management, Other Theories, Applications, and Technologies
Abstract: This study proposes an express mode of urban bus systems. The new mode features conventional buses serving as a feeder to the express bus system that connects two distant spots with intense demand (e.g. transport hubs). Operated as ground transit service with very few stops, the express bus is designated to run on urban expressways to guarantee a high cruising speed. Scattered around the globe are mostly bus rapid transit (BRT) or express buses that skip stops on conventional transit lines, yet these operational strategies cannot avoid high delays at, for instance, signalized intersections and bus stops. Herein lies a simplified transit corridor model developed to analyze people’s travel cost (total travel time) for the express/conventional mode of bus systems. The travel cost in the express mode is proved to be much smaller than the conventional mode in most cases. Furthermore, we study a Hotelling game in this corridor and prove the existence and uniqueness of Nash Equilibrium. Our result shows that it is impractical to achieve the Hotelling Nash Equilibrium (HNE) in conventional bus systems, but not in the proposed express system that would, shown by our model, incentivize the development of commercial centers near express bus terminals. Therefore, we conclude that the express mode could save travel cost and concentrate urban commerce.

Keywords: Public Transportation Management, Data Mining and Data Analysis
Abstract: Dealing with bus bunching is a challenge for bus operators, as it strongly affects perception on the level of service provided by buses. Despite the growing interest on bus bunching, existing studies have focused on its frequency, neglecting potential differences in the intensity of bus bunching events and, thus, in the associated effect on passengers. This study focuses on the duration of bus bunching events, as a measure of both the problem magnitude and its impact on passengers. Hazard-based duration models are employed to bus bunching event data from the Athens Public Transport Network to estimate the effect of and understand the factors that affect bus bunching event durations. The Lognormal Accelerated Failure Time model performs best among parametric models, with deep survival analysis achieving superior predictive accuracy. Results show that temporal factors, such as the afternoon peaks and weekends affect bunching duration. Further, physical and operational characteristics of the initial point in a sequence of bus stop instances also have an impact on the duration of bunching. Overall, results can shed light at the contributing factors of bus bunching events duration and help operators prioritize interventions at the level of bus stops.

Keywords: Public Transportation Management, Data Management and Geographic Information Systems, Data Mining and Data Analysis
Abstract: Public transport is one of the major forms of transportation in the world. This makes it vital to ensure that public transport is efficient. This research presents a novel real-time GPS bus transit data for over 500 routes of buses operating in New Delhi. The data can be used for modeling various timetable optimization tasks as well as in other domains such as traffic management and travel time estimation. The paper also presents an approach to reduce the waiting time of Delhi buses by analyzing the traffic behavior and proposing a timetable. This algorithm serves as a benchmark for the dataset. The algorithm uses a constrained clustering algorithm for classification of trips. It further analyses the data statistically to provide a timetable that is efficient in learning the inter- and intra-month variations.

Keywords: Public Transportation Management
Abstract: Cableway systems are often associated with mountains and skiing. Nonetheless, they established in several cities for public transportation in the last years. This paper describes the state of the art in urban cable car systems and focuses especially on advances and experiences of planning processes in Germany. For this purpose, we conduct a systematic review of academic literature regarding urban ropeway systems, including analyses of the international context, the methodologies used and different perspectives on the topic in an international context. Additionally, we investigate the Emirates Air Line which operates since 2012 in London in order to identify the challenges for urban ropeway systems for Germany. The results indicate that urban ropeway systems have a potential to establish in public transportation in German cities, but still exist several obstacles that need to be settled.

Keywords: Public Transportation Management, Network Modeling, Simulation and Modeling
Abstract: Public transport of a region is extremely important for connecting the commuters from their origins to destinations. Public transport systems with large fleets cannot be guaranteed to perform efficiently, unless it is well connected and accessible to maximum possible population. In this regard, the localization of public transport stops (nodes) are highly important, since access to public transit systems is only possible through these nodes. In this framework, this paper focuses on the formulation of a general methodology for the evaluation of public transit nodes of a region based on transit system characteristics, spatial coverages and characteristics of zones using the concepts of connectivity and accessibility. Similarly, connectivity and accessibility index are calculated and enhanced based on distribution of public transport trips in zones and compared with each other to determine the critical nodes. To show the capability of the proposed approach, an application of this methodology in terms of a case study is analyzed in order to show the effects of trip distribution in the zones on the connectivity and accessibility index values during different time periods.

Keywords: Road Traffic Control, Multi-modal ITS, Cooperative Techniques and Systems
Abstract: Transit signal priority (TSP) is recognized as an innovative technology solution capable of enhancing traditional transit services. This paper presents a novel de-centralized TSP system based on a Nash bargaining solution, with a variable phasing sequence and free cycle length to obtain an optimal control strategy (DNB-TSP controller). The developed system was implemented and evaluated in INTEGRATION microscopic traffic assignment and simulation software. The developed DNB-TSP system was compared to the operation of an optimum fixed time plan controller, a centralized adaptive phase split controller, a decentralized phase split and cycle length controller, and a de-centralized Nash bargaining (DNB) controller without TSP to evaluate the developed controller’s performance. The developed DNB-TSP system was applied to obtain an optimal control strategy on an isolated intersection and on an arterial corridor. Findings revealed that the developed DNB-TSP system produces significantly better performance for both passengers and transit vehicles compared to other controllers on an isolated intersection and arterial network.

Keywords: Road Traffic Control, Automated Vehicle Operation, Motion Planning, Navigation, Simulation and Modeling
Abstract: Road traffic management experts are constantly striving to develop, implement, and test a number of novel strategies to reduce traffic congestion impact on the economy, society, and the environment. Despite their efforts, these strategies are still inefficient and a call for advanced multidisciplinary approaches is needed. We, therefore, introduce in this paper an original traffic congestion mitigation strategy inspired by a well-known technology in wireless communications, i.e. cognitive radio technology. Our strategy exploits Connected Vehicles technology along with the often under-utilized reserved lanes, such as bus and carpool lanes, to virtually inflate the road network capacity to ease traffic congestion situations. Two variants of our strategy have been evaluated using simulation and the obtained results are very promising in terms of the achieved reduction in average travel time for different vehicle classes including buses as well.

Keywords: ., Road Traffic Control, Traffic Theory for ITS
Abstract: With unrelenting growth in population and urbanization, cities face escalating challenges in providing fast and reliable mobility. A significant source of delay for urban commuters is related to signalized intersections. Optimizing traffic signals to maximize the capacity of intersections is of critical importance for cities. Although this topic has been around for decades, recently, Deep Reinforcement Learning (DRL) approaches have begun to be used for intelligent traffic signal control. The result is a promising new generation of traffic signal controllers (TSCs). They are model-free, adaptive, and capable of exploiting newer pervasive sensory technologies (e.g., connected vehicles). While most of the RL-based TSCs focus on second-by-second level decision making, the industry favors controllers that manipulate signals less frequently. This changes the RL control problem from being in a discrete to a continuous action space. With the lack of RL-based TSCs that can handle a continuous action space, we propose a novel RL-based cycle-level TSC that determines the phase timings once every cycle. Our controller uses Proximal Policy Optimization (PPO) as one of the most promising continuous action RL algorithms to produce signal timings in a cycle. We test our proposed controller against one RL-based second-level controller as well as an optimized fixed-time traffic signal controller and compare the results.

Keywords: Road Traffic Control, Theory and Models for Optimization and Control, Simulation and Modeling
Abstract: Reinforcement Learning (RL) is being rapidly adopted in many complex environments due to its ability to leverage neural networks to learn good strategies. In traffic signal control (TSC), existing work has focused on off-policy learning (Q-learning) with neural networks. There is limited study on on-policy learning (SARSA) with neural networks. In this work, we propose a deep dueling on-policy learning method (2DSARSA) for coordinated TSC for a network of intersections that maximizes the network throughput and minimizes the average end-to-end delay. To describe the states of the environment, we propose traffic flow maps (TFMs) that capture head-of-the-line (HOL) sojourn times for traffic lanes and HOL differences for adjacent intersections. We introduce a reward function defined by the power metric which is the ratio of the network throughput to the average end-to-end delay. The proposed reward function simultaneously maximizes the network throughput and minimizes the average end-to-end delay. We show that the proposed 2DSARSA architecture has a significantly better learning performance compared to other RL architectures including Deep Q-Network (DQN) and Deep SARSA (DSARSA).

Keywords: Road Traffic Control, Simulation and Modeling, Theory and Models for Optimization and Control
Abstract: Shared left-turn lanes which serve for both through and left-turn vehicles, are typically installed and operated under permitted phase with an adjacent pedestrian crosswalk in Japan. Because left-turn vehicles have to decelerate and often yield to crossing pedestrians, the queue discharge on such a lane tends to be disturbed. However, previous studies have only focused on the estimation of the expected capacity, without paying attention to the impacts of its variations on performance. Therefore, this study addresses the random nature of the shared left-turn lane. We developed a Monte Carlo simulation that can reproduce the departure of vehicles in shared left-turn lanes, while also considering the impact of pedestrian blockages, and the random order of left-turn and through vehicles, then computed the performance indices such as the number of departures per cycle, delays, and occurrence of oversaturation. The results exhibited that although demand does not exceed the capacity in terms of their average, the random nature in the shared left-turn lane sometimes causes severe oversaturation, which cannot be anticipated by current performance evaluation methods. That provides important insights for intersection layouts and controls, and shows the necessity of the real-time intelligent signal control scheme.

Keywords: Road Traffic Control, Travel Information, Travel Guidance, and Travel Demand Management, Data Mining and Data Analysis
Abstract: Estimating Arterial Travel Time distributions on a regular basis is a challenging task for urban traffic agencies. Even though high-resolution loop detector data is usually available, the lack of other data modes and ground-truth labels hinders travel time estimation approaches that rely on such additional information.

Among the approaches relying exclusively on loop detector data are deterministic physics-based approaches (which use the mechanics of a virtual probe moving down a signalized arterial) and cost-minimization approaches (which directly estimate travel times from detector counts).

In this work, we propose and evaluate a hybrid model that uses virtual probe trajectories to indicate probable arrival and departure windows within which we apply a sequence alignment algorithm on high-resolution loop detector data to match platoons. Our approach naturally generalizes to real probe trajectory data, improving accuracy when such data is available.

Simulation-based results show that our approaches enable us to calculate a good estimate of arterial travel time distributions and are robust to noise. Thus, our methods can be used both with standalone archived loop detector data and in conjunction with trajectory data from connected vehicles.

Keywords: Travel Information, Travel Guidance, and Travel Demand Management, Network Management, ITS Field Tests and Implementation
Abstract: Cooperative Intelligent Transportation Systems (C-ITS) enable vehicles’ communication with each other (Vehicle-to-Vehicle, V2V) and with roadside infrastructure (Vehicle-to-Infrastructure, V2I). In the context of traffic efficiency C-ITS technologies could assist through data exchange, improving traffic control and traffic management implementation. Coordinated provision of C-ITS services refers to the provision of several C-ITS services as one combined service. The purpose is to harvest the usability of C-ITS services by developing a strategy for the operation and exploitation of services in real-time and within varying geographical areas. Two different dimensions of the coordinated provision of C-ITS services have been recognized covering: 1) the drivers, and 2) the road network operators and managers. The present paper deals with the road network operators and managers dimension, the objective of which is the integration and coordinated provision of C-ITS services in operational dynamic traffic management.

Keywords: Network Management, Cooperative Techniques and Systems, Theory and Models for Optimization and Control
Abstract: Cooperative Intelligent Transportation Systems (C-ITS) enable vehicles’ communication with each other (Vehicle-to-Vehicle, V2V) and with the roadside infrastructure (Vehicle-to-Infrastructure, V2I). In the context of traffic efficiency, C-ITS technologies can assist through data exchange towards improved traffic control organization and traffic management implementation. One of the major questions concerning this innovative array of technologies is how to integrate C-ITS services which are provided by private sector business-to-consumer vendors, in a coordinated manner to operational processes of traffic management. In this work an approach based on game theory principles is proposed for a framework that will support the coordination and exploitation of C-ITS services from Traffic Management Authorities. The objective of the proposed framework is to address cooperation or conflicts schemes, which are associated with the optimal implementation of C-ITS services from Traffic Management Authorities operating in a common area of the road network through the Nash bargaining problem and solution.

Keywords: Network Management, Other Theories, Applications, and Technologies
Abstract: We present a cap-and-trade scheme for the regulation of ridesharing. As opposed to marginal-pricing schemes, cap and-trade schemes limit the quantity of transportation. Recognizing that a central authority may not be able to adequately regulate quantity, we let the quantity be determined according to demand for ridesharing. We use demand to compute the social cost of selfish driving in a virtual world where ridesharing does not exist and set this cost as a limit on the amount of social cost that a transportation network company (TNC) can incur. We perform analysis in the static case to show that our scheme has the effect of incentivizing the positive effects of ridesharing, i.e., carpooling, while limiting its negative effects, e.g., deadheading.

We also present and discuss a practical implementation of the scheme. In implementation, the virtual social costs would be issued as credits through a central service and the actual social costs would be issued as debits; a net-positive balance would be imposed by the central service and TNCs could trade credits and debits on the open market.

Keywords: Theory and Models for Optimization and Control, Road Traffic Control, Automated Vehicle Operation, Motion Planning, Navigation
Abstract: The development of connected vehicle technology is beneficial to traffic operations at intersections. Automated vehicles are expected to emerge in the urban network gradually. The mixed traffic environment will exist for a long period. This paper presents a signal optimization model for an isolated intersection under the mixed traffic environment including connected and human-driven vehicles (CHVs) and connected and automated vehicles (CAVs). The problem is formulated as a mixed integer linear programming (MILP) model. Phase sequence and phase duration are optimized without the concept of cycle length for total vehicle delay minimization. Vehicle motion is analyzed and the passing states of CHVs and CAVs at stop bars are predicted to avoid the prior information of specific trajectory planning strategies of CAVs. A rolling horizon procedure is designed for the dynamic implementation of the proposed model with varying traffic conditions. The simulation results show that the proposed signal timing model can reduce average vehicle delay significantly compared to vehicle-actuated and fixed time control, especially with under-saturated traffic demand and more conflicting traffic streams. The sensitive analysis further shows that average vehicle delay decreases notably when the penetration rate of CAVs increases from 0% to 30%.

Keywords: Cooperative Techniques and Systems, Transportation Security, Network Management
Abstract: V2X (Vehicle-to-X) network is an important support to automated vehicle. However, for automated vehicle safety, if the security of V2X network cannot be guaranteed, automated vehicles may not obtain required correct information, which is a fatal problem. In V2X network, black hole and gray hole attacks are critical threats of network availability and may prevent legitimate users from receiving packets. In this review, the definition, classification and impact of the attacks on V2X network are presented, and the detection and prevention technologies are classified and analyzed, respectively. Finally, three open issues and future trends are proposed, which are considering specific characters of V2X network, paying attention to artificial intelligent technologies in detection and utilizing edge computing architecture as the local trusted environment in detection and prevention.

Keywords: ITS Field Tests and Implementation
Abstract: With the evolution of autonomous vehicle (AV), especially its ability to cope with harsh conditions, which is attracting intense attention from both academia and industry practitioners. Due to the high risk, high cost, and great difficulties of real road testing for AV, the hardware-in-the-loop (HIL) simulation platform is considered as a safe, economical, and effective testing method. In this paper, we developed a novel indoor virtual-reality combined HIL simulation platform for AV testing. This platform consists of a series of subsystems including math models, physical hardware and software components, on which all parts of an AV can be tested separately and jointly. To validate the functionality and performance of the platform, we adopt Udwadia–Kalaba(U-K) approach to build autonomous emergency braking (AEB) control algorithm as a case study due to the explicitness and simplicity of U-K approach. Further, a group of comparison experiments on both real road environment and the proposed indoor HIL platform is conducted. The experimental results show that the testing data obtained from the proposed HIL platform have a high similarity to those of the real road tests. Simultaneously, the testing time and repeatability of the former perform better, which confirms its feasibility and effectiveness on the vehicle AEB testing.

Keywords: Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Driver Assistance Systems
Abstract: In this study, a Human-in-the-Platoon CACC (HiP-CACC) controller is proposed for connected and automated vehicles to “include” human drivers into platooning process. The goal is to form a platoon between automated vehicles and human drivers so that turbulences caused by human drivers could be smoothed out by automated vehicles. Unlike the conventional CACC where only longitudinal control is automated, the proposed HiP-CACC regulates both longitudinally and laterally. In other words, the followers in a HiP-CACC platoon are fully autonomous. The controller is formulated as a model predictive control (MPC) solved by Chang-Hu’s method. The technology has the following advantages: i) take advantage of human drivers’ perception to enable conditional full autonomy; ii) accommodate actuator delay in system dynamics to improve actuator control accuracy; iii) automates both longitudinally and laterally; iv) ensures string stability in partially connected and automated vehicles environment; Field tests were conducted to verify the effectiveness of the proposed algorithm. The results show that the platoon is able to keep a constant distance gap with maximum 40 cm error longitudinally and maximum 40 cm error laterally.

Keywords: Accurate Global Positioning, Data Mining and Data Analysis
Abstract: Accurate vehicle position data is essential information for active traffic management in connected freeway. Traffic data of connected vehicles can be collected in real time while the one of the human drive vehicle have to be estimated in connected environment. A vehicle position estimation was proposed for human driving vehicle which are not adjacent to communicated vehicles, where the car-following equation was trained by a complex neural network. An improved recurrent neural network(RNN) based on gated recurrent unit (GRU) was adopted in the modeling to solve long-term dependencies. Both historical and present movement data the preceding vehicle were considered in the improved RNN model. Performance of the method was evaluated by vehicle-pair data extracted from NGSIM. The results indicated that the proposed method has higher accuracy than the method based on traditional car-following models.

Keywords: Cooperative Techniques and Systems, Theory and Models for Optimization and Control, Human Factors in Intelligent Transportation Systems
Abstract: There are problems such as complex control objects, multiple optimization objectives and control variables in traffic control under intelligent Vehicle-Infrastructure Cooperative System(i-VICS). In order to overcome the above problems, a control algorithm considering the penetration rate of connected automated vehicle (CAV) was established in this paper. To be able to control spatiotemporal-right and signal timing as a whole, a spatiotemporal-right solving method based on decision tree, a high-dimensional based genetic (MV-A1) and a low-dimensional based enumerated (MV-A2) signal-solving method were proposed; In order to achieve coordinated optimization of traffic efficiency and safety at intersections, a solution method based on Pareto optimality was proposed; Considering the influence of human-driven vehicle(HV), a CAV lane change and speed control method under mixed traffic conditions is proposed. A simulation experiment platform was developed by python3.7 to simulate and analyze the algorithm proposed in the paper. The results show that when the flow intensity is 0.23-0.45, the control effect is the best; 50% CAV penetration rate is the best benefit point of the control effect; the control algorithm has better adaptability to the time-varying of traffic demand.

Keywords: Traffic Theory for ITS, Simulation and Modeling
Abstract: In this study, we develop a traffic model to simulate network traffic evolution under the impact of controlled autonomous vehicles acting as moving bottlenecks. We first extend the Newell-Daganzo method to track the trajectories of moving bottlenecks and calculate the cumulative number of vehicles passing moving bottlenecks. By integrating the solutions to the cumulative number of vehicles passing moving bottlenecks and link nodes as boundary conditions in the link-transmission models, we can incorporate the impact of moving bottlenecks into the flow of traffic at a network scale. The numerical simulation results demonstrate the effectiveness of the developed model to track trajectories of the moving bottlenecks and simulate their impact on freeway traffic.

Keywords: Data Management and Geographic Information Systems, Other Theories, Applications, and Technologies
Abstract: The use of Global Positioning System (GPS) technology for travel data collection has become of interest to address requirements regarding modern transport modelling. However, such use is associated with several practical limitations including, among others, a need for significant participant involvement and considerable cost of implementation. These limitations restrict large-scale deployment of GPS-assisted data collection methods. Recent developments of positioning and communication technologies in smartphones have provided opportunities to address some of these limitations. This paper evaluates potential advantages of using smartphones through an experimental travel data collection application. The results reveal some capabilities of smartphones in improving the spatiotemporal accuracy of collected data. The paper provides a series of recommendations for more efficient deployment of GPS-assisted travel data collection using smartphones.

Keywords: Data Management and Geographic Information Systems, Data Mining and Data Analysis, Off-line and Online Data Processing Techniques
Abstract: Traffic sign inventory is an important part of road safety and traffic management. Road maintenance services perform this task with periodical on-site inspections, monitoring the physical presence and integrity of each object. Due to the increasing complexity of road infrastructure and the high number of objects, this task is usually time-consuming. Object detection based on artificial neural networks and state-of-the-art 360° camera equipment may be used to enhance the work of road maintenance services. By automatically capturing images during mandatory inspection drives, detecting traffic signs in those images and linking them with corresponding database entries, semi-automatic inventory and off-site inspections will be possible. This paper describes a system for semi-automatic traffic sign inventory that, in contrast to other approaches, automatically detects and classifies various traffic sign types in 360° images. It then references those signs in an existing inventory database, using GPS and distance estimation and allows for a virtual reality like off-site inspections with a 360° view.

Keywords: Data Management and Geographic Information Systems, Travel Information, Travel Guidance, and Travel Demand Management, Off-line and Online Data Processing Techniques
Abstract: Frequently, various sources of geographic street related data are covering the same space. Many geospatial traffic services require interoperability of the different datasets, which can be achieved by road network matching. A prominent use case is map conflation. More recently, the authors have suggested approaches to dynamic location referencing between maps (GIMME) and to automatic relocation of link related data in updated street maps within a framework called Map2Map. In this paper, an update on the recent advances in GIMME and Map2Map is given. Methodologically, path contraction is used to obtain a simplified version of the digital road network. This pre-processed version then augments the original network, and serves as a guide for finding routes covering the entire network, and facilitating the inter-map matching process. Path contraction also helps to reduce the complexity of the core inter-map matching method GIMME without loss in matching quality. On the conceptual side, a general strategy called “calibration-preserving pre- or post-processing" (C-3PO) is introduced. The aforementioned path contraction and two more post-processing methods used in Map2Map give examples for an implementation of C-3PO. Experimental results demonstrate the effectiveness of the presented approach.

Keywords: Accurate Global Positioning, Communications and Protocols in ITS, Data Management and Geographic Information Systems
Abstract: Highly accurate localization technologies are a key enabler of future intelligent transport system applications including automated and cooperative driving as well as location based services. One major challenge is positioning in urban areas, where high buildings with reflecting surfaces lead to signal blockage, multipath effects and Non-Line-of-Sight (NLOS) reception. As a countermeasure, deterministic NLOS modelling approaches using 3D map aided GNSS (3DMA-GNSS) have been developed and show promising results in mitigating those effects. However, most prior studies focused on advancing the method itself, while the implementational aspects of such a system regarding real-world applications remain unsolved. Major challenges lie within the generation and dissemination of suitable 3D maps as well as the computational cost of the classification of measurements into LOS resp. NLOS behaviour. The aim of this paper is to connect the parts of highly precise GNSS based urban localization, digital maps, Vehicle-to-Everything (V2X) based cooperative positioning and data dissemination. A map dissemination scheme for precomputed sky occupancy masks using V2X communication technology is evolved and discussed. Furthermore, the data rate requirements are compared with the capability of the existing IEEE802.11p (ITS-G5) V2X standard in terms of data rate and the range of the communication link.

Keywords: Off-line and Online Data Processing Techniques, Data Mining and Data Analysis, Driver Assistance Systems
Abstract: For the introduction of new automated driving functions, the systems need to be verified extensively. A scenario-driven approach has become an accepted method for this task. But, to verify the functionality of an automated vehicle in the simulation in a certain scenario such as a lane-change, relevant characteristic of scenarios need to be identified. That, however, requires to extract these scenarios from real-world drivings accurately. For that purpose, this work proposes a novel framework based on a set of unsupervised learning methods to identify lane-changes on motorways. To represent various types of lane-changes, the maneuver is split up into primitive driving actions with a Hidden Markov Model (HMM) and Divisive Hierarchical Clustering (DHC). Based on this, lane-change maneuvers are identified using Dynamic Time Warping (DTW). The presented framework is evaluated with a real-world test drive and compared to other baseline methods. With a F₁ score of 98.01% in lane-change identification, the presented approach shows promising results.

Keywords: Off-line and Online Data Processing Techniques, Roadside and On-board Safety Monitoring
Abstract: Railway turnouts monitoring is critical for ensuring the safety of railway systems. Owing to the scarcity of faulty samples, many approaches focus on the fault detection of railway turnouts only using normal samples. However, although the faulty samples are insufficient, they can still provide useful information for fault detection. To improve the fault detection performance of railway turnouts, this paper proposes an enhanced fault detection method incorporating prior faulty information, which deals with the limitation of existing fault detection methods of turnouts that they fail to take insufficient faulty samples into account. In our method, a novel model which shares the architecture of deep autoencoders but has a different training objective is presented. By minimizing the difference between the averaged reconstruction error of normal samples and that of faulty samples, the proposed model enlarges the gap between the normal and faulty classes and making them more separable. The field data collected from a real high-speed railway in China is used to evaluate the proposed method, and a comparative study with several existing approaches is conducted. Experimental results show the effectiveness of the proposed method.

Keywords: Roadside and On-board Safety Monitoring, Off-line and Online Data Processing Techniques, Other Theories, Applications, and Technologies
Abstract: In this study, we propose an ad-hoc framework for real-time turn count and trajectory reconstruction for vehicles passing through urban intersections. Our proposed framework utilizes virtual lanes representing the 8 standard National Electrical Manufacturers Association (NEMA) movements within an intersection. A Python Graphical User Interface (GUI) was developed is utilized to identify entry planes for each NEMA movement, obtaining the trajectories and counts for the vehicles that are detected and identified at those planes which are then used as identifiers for other vehicles detected inside the intersection using a nearest neighbors search algorithm. Our proposed framework runs as an additional layer to any multi-object tracker with minimal additional computation, and the results of this preliminary assessment indicate the high potential that this proposed framework has in obtaining reliable turn count and in mitigating identity switches by resolving the vehicle re-identification that occurs within the intersection due to detection errors and occlusion, resulting in more accurate vehicle trajectories from video data which will aid developing more reliable intersection safety surrogate measures.

Keywords: Roadside and On-board Safety Monitoring, Advanced Vehicle Safety Systems, Data Mining and Data Analysis
Abstract: High risk driving scenarios are critical for the deployment of highly automated vehicles virtual test. In this study, we have proposed a deep learning method to identify high risk scenarios from the field operation test (FOT) data. The proposed method tries to overcome the shortcomings of existing relevant studies for their limited utilizations of video data and mainly based upon instant kinematic indicators, which has led to high false alarm rate issue. In this study, a combined video analysis method (Convolutional Neural Network, CNN) and temporal feature analysis model (Long Short-Term Memory, LSTM) was proposed. To be specific, we used CNN-LSTM and Convolutional Neural Networks and Long Short-Term Memory (Resnet-LSTM) to perform the classifications for high risk scenarios and non-conflict scenarios. The empirical analyses have been conducted using commercial vehicle FOT data. And the results showed that the overall model performance (AUC index) in the test set could reach 0.91 with 83% accuracy rate. Finally, the future works have been discussed from the aspects of further extractions of video data and investigations of LSTM modelling results.

Keywords: Roadside and On-board Safety Monitoring, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Sensing, Vision, and Perception
Abstract: An important component for the realization of the automated driving task is a holistic environment model. Connected and Autonomous Vehicles (CAVs) must be capable of detecting other vehicles, road markings, dangerous obstacles and upcoming road conditions. Apart from the comfort dependency on the road condition, friction values are calculated on the basis of road properties, which in turn are relevant for e.g. breaking and safety distances of CAVs. Due to the substitution of the human control task by the machine, this information must in future be detected by the vehicle itself. Based on the existing Vehicle Level Sensors (VLSs) and Acceleration Sensors (ASs) data, which are standard components in modern vehicles, a machine-learning approach of determining road surface materials and road hazards is presented. Our software solution of determining different road surface materials as asphalt, concrete, cobblestone or gravel with a total accuracy of 92.36 % is presented. Furthermore, the results of the road hazards detection as potholes and speed bumps with a total accuracy of 92.39 % is stated. Additionally to the edge calculations in the vehicle, our idea resolves in connected vehicles being capable of classifying road conditions enabling them to provide road analyses to a cloud platform. The goal is to establish a holistic cloud solution for road conditions to enable CAVs for the consumption of road condition data of upcoming road segments and empower them to adjust to those.

Keywords: Roadside and On-board Safety Monitoring, Driver Assistance Systems, Off-line and Online Data Processing Techniques
Abstract: Driver is the most important part of safety in transportation as risky driving behavior is the leading cause of accidents. Driver performance analysis enables us to identify risky drivers and improve safety by providing in time and specialized coaching. It can be achieved by analyzing driver actions. In this work, we propose a framework for context-aware driver performance analysis by validating the actions drivers take on the road by considering the context in which it has occurred. The framework is tested on tailgating examples from long haul trucks and the results show that it can differentiate driving behavior where a driver was at fault from the ones where she was not by incorporating context. The framework can be used to identify risky drivers by analyzing their context-aware actions.

Keywords: ITS Field Tests and Implementation, Electronic Payment, Data Mining and Data Analysis
Abstract: This study proposes an algorithm to integrate bay-based parking occupancy, captured using an image processing system, with information from a common conventional parking payment management system. The algorithm enables the use of simple and inexpensive cameras to collect parking utilization to complement conventional payment transaction data. Details about the design, implementation, testing, and validation of the algorithm are provided. Validation was performed using data collected through an accurate observational survey, including plate numbers of all vehicles parked in the areas of analysis. Results from a case study using the proposed algorithm provided an accuracy of 76% in terms of correct data integration. Logistic regression analysis was used to improve parameters used by the proposed algorithm. To illustrate the value of the algorithm, the integrated data were used to evaluate parking payment behavior. Various performance measures can be proposed and estimated using the integrated data. The integrated data can be used to address questions of high importance for more efficient and effective management of parking facilities.

Keywords: Sensing, Vision, and Perception
Abstract: Depth estimation is a basic problem in computer vision, which provides three-dimensional information by assigning depth values to pixels. With the development of deep learning, researchers have focused on estimating depth based on a single image, which is known as the "monocular depth estimation" problem. Moreover, panoramic images have been introduced to obtain a greater view angle recently, but the corresponding model for monocular depth estimation is scarce in the state of the art. In this paper, we propose PADENet for panoramic monocular depth estimation and re-design the loss function adapted for panoramic images. We also perform model transferring to panoramic scenes after training. A series of experiments show that our PADENet and loss function can effectively improve the accuracy of panoramic depth prediction while maintaining a high level of robustness and reaching the state of the art on the CARLA Dataset.

Keywords: Sensing, Vision, and Perception, Other Theories, Applications, and Technologies, .
Abstract: In this paper, a multi-modal 360◦ framework for 3D object detection and tracking for autonomous vehicles is presented. The process is divided into four main stages. First, images are fed into a CNN network to obtain instance segmentation of the surrounding road participants. Second, LiDAR-to-image association is performed for the estimated mask proposals. Then, the isolated points of every object are processed by a PointNet ensemble to compute their corresponding 3D bounding boxes and poses. Lastly, a tracking stage based on Unscented Kalman Filter is used to track the agents along time. The solution, based on a novel sensor fusion configuration, provides accurate and reliable road environment detection. A wide variety of tests of the system, deployed in an autonomous vehicle, have successfully assessed the suitability of the proposed perception stack in a real autonomous driving application.

Keywords: Sensing, Vision, and Perception, Off-line and Online Data Processing Techniques
Abstract: While neural networks trained for semantic segmentation are essential for perception in autonomous driving, most current algorithms assume a fixed number of classes, presenting a major limitation when developing new autonomous driving systems with the need of additional classes. In this paper we present a technique implementing class-incremental learning for semantic segmentation without using the labeled data the model was initially trained on. Previous approaches still either rely on labels for both old and new classes, or fail to properly distinguish between them. We show how to overcome these problems with a novel class-incremental learning technique, which nonetheless requires labels only for the new classes. Specifically, (i) we introduce a new loss function that neither relies on old data nor on old labels, (ii) we show how new classes can be integrated in a modular fashion into pretrained semantic segmentation models, and finally (iii) we re-implement previous approaches in a unified setting to compare them to ours. We evaluate our method on the Cityscapes dataset, where we exceed the mIoU performance of all baselines by 3.5% absolute reaching a result, which is only 2.2% absolute below the upper performance limit of single-stage training, relying on all data and labels simultaneously.

Keywords: Sensing, Vision, and Perception, Automated Vehicle Operation, Motion Planning, Navigation
Abstract: Place recognition refers to the problem of finding the position of a query image based on a series of images acquired at different places. Yet the day and night place recognition problem is hard to solve due to the illumination and appearance changes. Image-to-image translation methods have been introduced to solve the place recognition problem by synthesizing daytime images from the night ones. However, these methods cannot achieve good translation performance with low-quality night-time images. In this paper, a new method is introduced to improve the quality of night-time restored images by combining image enhancement and image inpainting methods together. Three kinds of enhanced night-time images are generated based on the proposed method.

Our place recognition system includes a model of GoogleNet to generate deep features of input images and nearest neighbor searching for image retrieval process. The approach is tested on Oxford RobotCar dataset, where three low-quality night sequences are selected as query sequences and one day sequence is selected as reference sequence. The results obtained with the approach based on the three proposed enhanced night-time images are better than those obtained with the raw night-time images. The results of our proposed place recognition system are also compared with two state-of-art place recognition methods: ToDayGAN and densevlad.

Keywords: Sensing, Vision, and Perception
Abstract: A crucial part of safe navigation of autonomous vehicles is the robust detection of surrounding objects. While there are numerous approaches covering object detection in images and LiDAR point clouds, this paper addresses the problem of object detection in radar data. For this purpose, the fully convolutional network YOLOv3 is adapted to operate on sparse radar point clouds. In order to apply convolutions, the point cloud is first transformed into a grid-like structure. The impact of this representation transformation is shown by comparison with a network based on Frustum PointNets which directly processes point cloud data. The presented networks are trained and evaluated on the public nuScenes dataset. While experiments show that the point cloud-based network outperforms the grid-based approach in detection accuracy, the latter has a significantly faster inference time neglecting the grid conversion which is crucial for applications like autonomous driving.

Keywords: Cooperative Techniques and Systems, Sensing, Vision, and Perception, Communications and Protocols in ITS
Abstract: Using vehicles and roads to perceive pedestrians cooperatively can eliminate the visual limitation of a single vehicle, thus improve the pedestrian collision warning capability of the driving assistance system. In this paper, we propose a vehicle-road pedestrian re-identification (ReID) system for vehicle-road cooperative perception of pedestrians, which can provide real-time and reliable ReID results under the premise of minimizing the occupation of communication resources. Among the system, we propose the CNN-LSH feature extraction algorithm and adaptive feature dimension reduction algorithm to adjust the dimension of pedestrian images features effectively according to the status of wireless channel, and use Hamming distance metric to simplify the calculation of feature similarity, which accelerates the speed of image feature matching in ReID. Experimental results demonstrate that the vehicle-road ReID system can significantly reduce the communication overhead and computation time while providing reliable ReID results in real time.

Keywords: Cooperative Techniques and Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Road Traffic Control
Abstract: Cooperative driving is one of the important research topics in the transportation field due to the positive impact on road capacity and safety as well as fuel consumption. One of the most well-known applications of cooperative driving is vehicle platooning, which has proven its ability to improve the fuel consumption of its members. However, the fuel consumption of a vehicle during a highway journey depends on many other factors; such as destinations and speeds of the solo vehicles as well as the platoon members. Hence, a joining decision-making algorithm is proposed in this paper to estimate the potential fuel savings from joining the candidate platoons on a highway based on optimizing the joining speed profile using Sequential Quadratic Programming as well as Genetic Algorithm. Six different scenarios tested the performance of the decision-maker in terms of fuel and time savings as well as the computational cost. Furthermore, a deep neural network model is proposed to try to maintain the optimizer accuracy and reduce the computational cost, which was successfully obtained from the testing experiments.

Keywords: Cooperative Techniques and Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations
Abstract: Platoon control of connected vehicles (CVs) can greatly improve fuel efficiency and traffic throughput. This paper proposes a unified hierarchical framework for platoon control of CVs with two different types of control modes, i.e., desired acceleration control and desired velocity control. By separating neighboring information interaction from local dynamics control, the framework divides the task of distributed platoon control into two layers, i.e., an upper-level observing layer and a lower-level tracking control layer, to address vehicle dynamics heterogeneity. Within the proposed framework, an observer is designed for following vehicles to observe the leading vehicle’s states through vehicle-to-vehicle communication, while a tracking controller is designed to track the leading vehicle using local observation information. A necessary and sufficient condition is further derived to guarantee asymptotic stability of the platoon control system. Numerical simulation results demonstrate the effectiveness of the proposed hierarchical platoon controller.

Keywords: Cooperative Techniques and Systems, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Driver Assistance Systems
Abstract: With the advances in communication technologies, a fully connected and automated traffic environment has become possible to be implemented. However, it is a fact that a road network that is ready to be operated in such a fashion needs several systems to handle different maneuvers in traffic flow. Therefore, in the study we summarize in the present paper, we propose an integrated system that is composed of three elements, which organizes longitudinal and lateral movements with the intention of mitigating shockwaves due to merging maneuvers. A car-following based (Cooperative Adaptive Cruise Control) CACC model is used for longitudinal motion modeling, while vehicles approaching to the merging zone are controlled using a speed harmonization algorithm and, in cases of zone conflicts, a merging strategy that is based on the game theory is applied for flow management. Simulation results from a single lane road segment show that the proposed approach outperforms the performance of the system in which cooperative merging is not adopted. I

Keywords: Cooperative Techniques and Systems, Driver Assistance Systems
Abstract: In this paper, we re-examine the notion of string stability as it relates to safety by providing an upper bound on the maximum spacing error of any vehicle in a homogeneous platoon in terms of the input of the leading vehicle. We reinforce our previous work on lossy CACC platoons by accommodating for burst-noise behavior in the V2V link. Further, through Monte Carlo type simulations, we demonstrate that connectivity can enhance traffic mobility and safety in a CACC string even when the deceleration capabilities of the vehicles in the platoon are heterogeneous.

Keywords: Traffic Theory for ITS, Rail Traffic Management
Abstract: Nowadays, the transport capacity of rail transit lines has reached saturation, so it is urgent to improve the utilization and capacity of existing lines. In another area of intelligent transportation, cars can significantly improve capacity by cooperative driving, and researchers have been inspired to propose a virtual coupling scheme. The scheme improves the capacity of the line by reducing the headway. However, the design of train control system based on virtual coupling is lack of mature standard, and the traditional safety analysis method is difficult to carry out. The system theoretic process analysis (STPA) method can build the control structure for the function and organization of the system, and then carry out the safety analysis, put forward the safety constraints, reaction to the system safety design. In this paper, the STPA method is used to construct a train control system based on virtual coupling, and some potential hazards not easily detected by traditional safety analysis methods are found. Based on this, a visual research method of the system safe state is proposed, which is helpful to show the safety constraint intuitively, and can assist the decision-making in the safety design of the system.

Keywords: Simulation and Modeling, Multi-modal ITS, Rail Traffic Management
Abstract: To achieve the climate protection targets despite the increasing transport demand, the shift from carbon-intensive to more environmentally friendly modes, such as rail, is indispensable in the field of freight transport. The Next Generation Train CARGO concept is intended to improve the competitiveness of rail freight, especially for low-density high value goods. However, the corresponding transhipment infrastructure has not yet been analysed in detail. In this work, we introduce a Model-Based Systems Engineering approach for the closer analysis and specification of an intermodal freight terminal for this high-speed freight train concept. This includes the elaboration of the system idea and context, the most important stakeholders and their requirements as well as the identification of the essential system functions. The systematic approach reveals a broad diversity of stakeholders and points out the complexity of the procedures taking place at the terminal. The chosen approach applied in this work has proven to be promising for the holistic system analysis of an intermodal transport node.

Keywords: Rail Traffic Management, Data Mining and Data Analysis, Simulation and Modeling
Abstract: Cascading delays that propagate from a primary source along a railway network are an immediate concern for British railway systems. Complex nonlinear interactions between various spatio-temporal variables govern the propagation of these delays which can quickly spread throughout railway networks, causing further severe disruptions. To better understand the effects of these nonlinear interactions, we present a novel, graph-based formulation of a subset of the British railway network. Using this graph-based formulation, we apply the Spatial-Temporal Graph Convolutional Network (STGCN) model to predict cascading delays throughout the railway network. We find that this model outperforms other statistical models which do not explicitly account for interactions on the rail network, thus showing the value of a Graph Neural Network (GNN) approach in predicting delays for the British railway system.

Keywords: Simulation and Modeling, Emission and Noise Mitigation, Rail Traffic Management
Abstract: Nowadays, energy conservation and emission reduction have become the top priority of metro systems. This paper concentrates on the application of regenerative braking technology in the energy-saving operation of metro systems. The research focus of this paper is maximizing the utilization of regenerative energy. We propose an energy-saving optimization method for train operation plans. Firstly, the train operation plan optimization model is constructed based on the adjustment of train operation interval and opposite departure interval. Secondly, we design a simulation algorithm that can effectively solve such a problem. Finally, the real-world line structure and speed profiles are used to verify the effectiveness of the optimization model and the algorithm. The results show that the optimization effects of this method are 0.8% and 2.2%, respectively, during peak and off-peak hours when the power grid is separated, and the optimization effects are 3% and 3.6%, respectively, during peak and off-peak hours when the power grid is connected. The method improves the utilization of regenerative energy and reduces the waste of resources significantly. Compared with taking the overlapping time to represent available regenerative energy, the energy calculation method proposed in this paper is more accurate.

Keywords: Network Modeling, Data Mining and Data Analysis
Abstract: Taxi demand prediction is an intensively studied topic in intelligent transportation research. Recently, deep learning models have been widely applied and have shown good performances. However, these methods overlook the existence of hyper-imbalanced taxi demand, which may result in good indicators in numerical experiments but weak performance in real scenarios. In this paper, we focus on the hyper-imbalance data and improve deep learning abilities for taxi demand prediction. To accomplish this task, slice indicators are introduced to fairly evaluate prediction performance at each taxi demand level. Then, through the lens of the slice indicators, a new form of loss called slice-weighted loss (SWL) is developed to improve the prediction of high taxi demand. Combining the SWL with an improved convolutional long short-term memory (Conv-LSTM) model, a spatiotemporal network called slice-wighted-Conv-LSTM (SW-CLSTM) is proposed. It can overcome the problem of data hyper-imbalance and make considerable improvements in taxi demand prediction. By conducting extensive experiments on large-scale TLC trips, we validate the power of slice-indicators and demonstrate the effectiveness of our approach over state-of-the-art methods.

Keywords: Network Modeling, Simulation and Modeling
Abstract: The growing popularity of mobility-on-demand fleets increases the importance to understand the impact of mobility-on-demand fleets on transportation networks and how to regulate them. For this purpose, transportation network simulations are required to contain corresponding routing methods. We study the trade-off between computational efficiency and routing accuracy of different approaches to routing fleets in a dynamic network simulation with endogenous edge travel times: a computationally cheap but less accurate Network Fundamental Diagram (NFD) based method and a more typical Dynamic Traffic Assignment (DTA) based method. The NFD-based approach models network dynamics with a network travel time factor that is determined by the current average network speed and scales free-flow travel times. We analyze the different computational costs of the approaches in a case study for 10,000 origin-destination (OD) pairs in a network of the city of Munich, Germany that reveals speedup factors in the range of 100. The trade-off for this is less accurate travel time estimations for individual OD pairs. Results indicate that the NFD-based approach overestimates the DTA-based travel times, especially when the network is congested. Adjusting the network travel time factor based on pre-processed DTA results, the NFD-based routing approach represents a computationally very efficient methodology that also captures traffic dynamics in an aggregated way.

Keywords: Network Modeling, Network Management, Other Theories, Applications, and Technologies
Abstract: Recently, the demand for urban transportation is increasing, and the construction of transport hubs is also ongoing. In view of the large, complex and uneven distribution of urban hubs in city, how to represent the multi-layer and multi-attributes of urban hubs is quite essential for the improvement of the coordination among hubs and operation efficiency. In term of the topological characteristics of hub network, this paper adopts Super Network (SN) theory. In traditional complex network research, the performance characteristics of nodes in network were ignored, which treated the nodes as undifferentiated single nodes. In this paper, the Urban Hub Super Network (UHSN) model is constructed and the characteristic of UHSN is also studied based on new element of Super-edge (SE). Besides, the topological indexes of Hub Degree of SE (HDSE), Zero Hub Degree of SE (ZHDSE), and the number of sub-nodes in subway sub-network to SE (NSSE) are presented, and based on which the heterogeneity, functionality and robustness are discussed. Based on the case study of Beijing, it found that ZHDSE is an important index and can well represent functionality. In addition, HDSE and NSSE can well express heterogeneity and robustness respectively after ZHDSE is considered.

Keywords: Travel Behavior Under ITS, Human Factors in Intelligent Transportation Systems, Ride Matching and Reservation
Abstract: Value-of-time (VOT) and willingness-to-pay (WTP) measures are valuable in a wide range of transport policies and planning applications. The purpose of the present research is to estimate these measures in Thessaloniki, where a pilot mobility scheme inspired by the concept of sharing economy is implemented. The pilot focuses on reducing the commuting trips from the eastern part of the city to the city centre by using a taxi sharing service. A questionnaire including a stated-preference (SP) experiment has been developed and administered to a random sample of 90 people. The survey combines trip-based characteristics (mode, travel time, and travel cost), with socioeconomic characteristics, such as profession, education, and car ownership. Discrete choice models are developed within a methodological framework and the estimated coefficients have been used to estimate VOT. A second sample consisted of users of the pilot service is selected for the estimation of WTP through the development of a Price Sensitivity Model. The model results in a range of acceptable prices from 2.00 to 3.50€ for the taxi-sharing service use supporting the long-term sustainability of the service.

Keywords: Public Transportation Management, Travel Information, Travel Guidance, and Travel Demand Management, Travel Behavior Under ITS
Abstract: Public transport system needs to serve passengers continuously, accurately and effectively. Service quality in public transportation has been widely researched since it can evaluate public transportation systems within the passengers' aspect. At present, however, the study of public transport system service quality is usually based on survey, which requires a lot of time and economic cost. Besides, the result is insufficient in some cases since few passengers take the survey. Commuter traffic in peak hours is always a hot topic in public transport research. Based on field experiments, this paper proposes an evaluation method for public transportation service quality based on the energy cost of passengers. This method utilizes the heart rate, acceleration and speed data automatically collected by the experimenters when they are walking in subway transfer stations, fits these data to Physical Activity Intensity and uses it as the index of travel energy cost. Subsequently, the accuracy, theoretical and practical prospect of this method are verified by the transfer passenger data of Beijing Subway Line 1 and Line 2 in May, 2017. The results show that the service quality evaluation method can accurately perceive the change of system service efficiency and its recovery ability according to different travel demands of the passengers. At the same time, this method uses automatic data collection to analyze, improves its accuracy and analysis adaptability compared with the traditional methods.

Keywords: Public Transportation Management, Theory and Models for Optimization and Control, Simulation and Modeling
Abstract: Transfers between public transport modes hinder the traveling by public transport due to the unreliability and long waiting times. Nonetheless, public transport operators are mainly interested in maintaining the regularity of different lines without accounting for the transfer waiting times. This work focuses on the application of bus holding to reduce the transfer waiting times, in-vehicle waiting times, and the deviation from the target headways. The determination of optimal holding times is performed in time horizons (e.g., hourly time periods) where past events are fixed, while future events are considered in the optimization process. This study models the bus holding problem which considers transfer synchronization as a nonlinear convex program (NLP) that can be solved by off-the-shelf solvers. Our approach is tested against state-of-the-art bus holding methods in a case study on the bus network of Almere (Netherlands) demonstrating a significant reduction in the transferring waiting times.

Keywords: Public Transportation Management, Network Management, Theory and Models for Optimization and Control
Abstract: Public Transportation Buses are an integral part of our cities, which relies heavily on optimal planning of routes. The quality of the routes directly influences the quality of service provided to passengers, in terms of coverage, directness, and in-vehicle travel time. In addition, it affects the profitability of the transportation system, since the network structure directly influences the operational costs. We propose a system which automates the planning of bus networks based on given demand. The system implements a paradigm, Deep Reinforcement Learning, which has not been used in past literature before for solving the well-documented multi-objective Transit Network Design and Frequency Setting Problem (TNDFSP). The problem involves finding a set of routes in an urban area, each with its own bus frequency. It is considered an NP-Hard combinatorial problem with a massive search space. Compared to state-of-the-art paradigms, our system produced very competitive results, outperforming state-of-the-art solutions.

Keywords: Public Transportation Management, Rail Traffic Management, Travel Information, Travel Guidance, and Travel Demand Management
Abstract: Understanding real-time train occupancy is a critical problem for public transport management, especially in the service disruption scenarios. To address this problem, this paper proposes a public transport passenger assignment method for estimating the time-dependent train occupancy comprising of a three-step modelling approach. Firstly, we make use of train station tap-on and tap-off information collected by Automated Fare Collection systems to estimate the initial time- dependent Origin-Destination matrix (OD) of the train network. Secondly, we take advantage of real-time train scheduling data to calibrate the initial OD matrix according to travel time, transfer time and waiting times across train lines. Thirdly, the calibrated OD matrix together with train scheduling data are used to generate entire passenger travel trajectories from origins to destinations including all path segments, by following a probabilistic hybrid Markov-driven approach. Lastly, after knowing all passenger trajectories, we further estimate the passenger occupancy for every train in the entire network in a given short time window. The results are applied over the real Sydney train network in Australia, and showcase that the proposed method can accurately quantify time-dependent passenger flows at a station platform level of granularity.

Keywords: Public Transportation Management, Data Mining and Data Analysis, Travel Information, Travel Guidance, and Travel Demand Management
Abstract: Accurate prediction of short-term passenger flow is of great significance for metro managers to organize passenger flow and allocate capacity resources high-efficiently. In this paper, we propose a spatiotemporal graph convolution gated recurrent unit neural network (GCGRUA) combined with attention mechanism to predict short-term passenger flow in metro systems. Graph convolutional network is applied to extract spatial feature of passenger flow in the metro network. Gated recurrent unit is introduced to extract temporal feature of passenger flow. Attention mechanism is proposed to identify the more relevant time step inputs to improve the performance in temporal prediction. The proposed model can handle spatiotemporal correlation of passenger flow prediction in large-scale metro network. A case study of Beijing metro system is illustrated to verify the performance of the proposed model. The results show that the proposed model can well deal with the spatial-temporal relationship of passenger flow in metro networks and is superior to other baseline models.

Keywords: Road Traffic Control, Simulation and Modeling, Theory and Models for Optimization and Control
Abstract: Variable Speed Limit (VSL) is a traffic control approach that optimises the mainstream traffic on motorways. Reinforcement Learning approach to VSL has been shown to achieve improvements in controlling the mainstream traffic bottleneck on motorways. However, single-agent VSL, applied to a shorter motorway segment, can produce a discontinuity in traffic flow by causing the significant differences in speeds between the uncontrolled upstream flow and the flow affected by VSL. A multi-agent control strategy can be used to overcome these problems by assigning speed limits in multiple upstream motorway sections enabling smoother speed transition. In this paper, we proposed a novel approach to set up multi-agent RL-based VSL by using the W-Learning algorithm (WL-VSL), in which two agents control two segments in the lead up to the congested area. The reward function for each agent is based on the agent's local performance as well as the downstream bottleneck. WL-VSL is evaluated in a microscopic simulation on two traffic scenarios using dynamic and static traffic demand. We show that WL-VSL outperforms base cases (no control, single agent, and two independent agents) with the improvement of traffic parameters up to 18 %.

Keywords: Road Traffic Control, Theory and Models for Optimization and Control
Abstract: Traffic signal control plays an essential role in the Intelligent Transportation Systems (ITS). Due to the intrinsic uncertainty and the significant increase in travel demand, in many cases, a traffic system still has to rely on human engineers to cope with the complicated and challenging traffic control and operation problem, which cannot be handled well by the traditional methods alone. Thus, imitating the good working experience of engineers to solve traffic signal control problems remains a practical, smart, and cost effective approach. In this paper, we construct a modelling framework to imitate how engineers cope with complex scenarios through learning from the historical record of manipulations by traffic operators. To extract spatial-temporal traffic demand features of the entire road network, a specially designed mask and a graph convolutional neural network (GCNN) are employed in this framework. The simulation experiments results showed that, compared with the original deployed control scheme, our method reduced the average waiting time, average time loss of vehicles, and vehicle throughput by 6.6%, 7.2%, and 6.85%, respectively.

Keywords: Cooperative Techniques and Systems, Road Traffic Control
Abstract: Intersection management of Connected Autonomous Vehicles (CAVs) has the potential to significantly improve safety and mobility. While numerous intersection management designs have been proposed in the past few decades, most of them assume that the CAVs will precisely follow the directions of the Intersection Manager (IM) and prove the safety and demonstrate the efficiency based on this assumption. In real life, however, a CAV that is crossing the intersection may break down, accelerate out-of-control or lie about its information (e.g. intended outgoing lane) and cause an accident. In our point of view, a real-life intersection should be robust even under these circumstances. In this paper, we first define a fault model called ``rogue vehicle'', which is essentially a CAV that either is dishonest or does not follow the IM's directions and then, propose a novel management algorithm (R<sup>2</sup>IM) that will ensure safe operation, even if a CAV becomes ``rogue'' at any point in time. We prove that there can be no accidents inside the intersection, as long as there is no more than one ``rogue vehicle'' at a time. We demonstrate the safety of R²IM by performing experiments on 1/10 scale model CAVs and in simulation. We also show that our approach can recover after the rogue vehicle leaves/is removed.

Keywords: Data Mining and Data Analysis, Travel Information, Travel Guidance, and Travel Demand Management, Travel Behavior Under ITS
Abstract: Predicting urban traffic flow is significant to intelligent transportation systems. Urban traffic flow data is a type of time-series data, which collects the traffic flow of a road section or area. So in this paper, we treat urban traffic flow prediction as a time series problem. The traditional method to tackle traffic flow prediction is difficult, because of the complex influence factors and nonlinear dependencies. Recently, LSTM based network has been widely used to model long-term series, but the memory of LSTM is typically too small and is not enough to accurately remember facts from the past. In this paper, we using memory time-series network with additional memory mechanisms to address urban traffic prediction problems. Historical data were divided into long-term and short-term two parts, long-term historical data models the overall trend and short-term historical data takes into account recent changes. The experiment results on two urban traffic flow datasets demonstrate the model is effective and outperforms baselines.

Keywords: Road Traffic Control, Cooperative Techniques and Systems, Automated Vehicle Operation, Motion Planning, Navigation
Abstract: This paper proposes a Trajectory based Traffic Management (TTM) decision maker for connected and automated traffic. The goal is to enable active traffic management on a microscopic level with the help of CAVs. The proposed decision maker bears the following features: i) enable real-time traffic demand management; ii) avoid the curse of dimensionality; iii) fast in computation time; iv) be potential for extension to partially connected and automated traffic. The proposed decision maker is evaluated against the conventional active traffic management technique, named variable speed limit. The results demonstrate that the proposed method outperforms the conventional method. It improves traffic flow stability by 63.8%, lane change smoothness by 78.6%, impact range by 76.9% and control stability by 11 times.

Keywords: Sensing and Intervening, Detectors and Actuators, Roadside and On-board Safety Monitoring, Sensing, Vision, and Perception
Abstract: The real-time detection of traffic rule violators in a city-wide surveillance network is a highly desirable but challenging task because it needs to perform computationally complex analytics on the live video streams from a large number of cameras, simultaneously. In this paper, we propose an efficient framework using edge computing to deploy a system for automatic detection of bike-riders without a helmet. First, we propose a novel robust and compact method for the detection of the motorcyclists without helmet using convolutional neural networks (CNNs). Then, we scale it for the real-time performance on an edge-device by dropping redundant filters and quantizing the model weights. To reduce the network latency, we place the detector module on edge-devices in the cameras. The edge-nodes send their detected alerts to a central alert database where the end-users access these alerts through a web interface. To evaluate the proposed method, we collected two datasets of real traffic videos, namely, IITH_Helmet_1 which contains sparse traffic and IITH_Helmet_2 which contains dense traffic. The experimental results show that our method achieves a high detection accuracy of ~95% while maintaining the real-time processing speed of ~22fps on Nvidia-TX1.

Keywords: Sensing and Intervening, Detectors and Actuators, Simulation and Modeling, ITS Field Tests and Implementation
Abstract: Large driving datasets are a key component in the current development and safeguarding of automated driving functions. Various methods can be used to collect such driving data records. In addition to the use of sensor equipped research vehicles or unmanned aerial vehicles (UAVs), the use of infrastructure sensor technology offers another alternative. To minimize object occlusion during data collection, it is crucial to record the traffic situation from several perspectives in parallel. A fusion of all raw sensor data might create better conditions for multi-object detection and tracking (MODT) compared to the use of individual raw sensor data. So far, no sufficient studies have been conducted to sufficiently confirm this approach. In our work we investigate the impact of fused LiDAR point clouds compared to single LiDAR point clouds. We model different urban traffic scenarios with up to eight 64-layer LiDARs in simulation and in reality. We then analyze the properties of the resulting point clouds and perform MODT for all emerging traffic participants. The evaluation of the extracted trajectories shows that a fused infrastructure approach significantly increases the tracking results and reaches accuracies within a few centimeters.

Keywords: Sensing and Intervening, Detectors and Actuators, Data Mining and Data Analysis, Off-line and Online Data Processing Techniques
Abstract: In this paper, we present a reliable and scalable approach for real-time estimation of link speeds (i.e., traffic speeds on specific road segments) based on video feeds coming from monocular cameras. We detect and track vehicles of specific types, identify anchor points (or keypoints) on them, compute their poses, and use this information to estimate their speeds. We use deep learning methods for vehicle detection, tracking, keypoint detection and localization, and traditional 3D pose estimation techniques for which precise mathematical solutions are available. Thus, our approach exploits the best of both worlds. The proposed approach does not require any physical measurements (extrinsics) in the road scene, making it scalable and easy to install. Our results on video feeds from Bangalore, India, show that the method is able to generalize well for cameras mounted on street light poles, congested traffic situations, and various lighting conditions. Thus, the solution is suitable for emerging market scenarios where traffic tends to be chaotic and dense, and mounting speed sensors or strategically located downward-facing cameras is not feasible. The code and dataset for this work are being made available.

Keywords: Sensing and Intervening, Detectors and Actuators
Abstract: The challenge of traffic state estimation (TSE) lies in the sparsity of observed traffic data and the sensor noise present in the data. This paper presents a new approach — physics informed deep learning (PIDL) method — to tackle this problem. PIDL equips a deep learning neural network with the strength of the physical law governing traffic flow to better estimate traffic conditions. A case study is conducted where the accuracy and convergence-time of the algorithm are tested for varying levels of scarcely observed traffic density data — both in Lagrangian and Eulerian frames. The estimation results are encouraging and demonstrate the capability of PIDL in making accurate and prompt estimation of traffic states.

Keywords: Sensing and Intervening, Detectors and Actuators, Off-line and Online Data Processing Techniques, ITS Field Tests and Implementation
Abstract: The use of infrastructure sensor technology for traffic detection has already been proven several times. However, extrinsic sensor calibration is still a challenge for the operator. While previous approaches are unable to calibrate the sensors without the use of reference objects in the sensor field of view (FOV), we present an algorithm that is completely detached from external assistance and runs fully automatically. Our method focuses on the high-precision fusion of LiDAR point clouds and is evaluated in simulation as well as on real measurements. We set the LiDARs in a continuous pendulum motion in order to simulate real-world operation as closely as possible and to increase the demands on the algorithm. However, it does not receive any information about the initial spatial location and orientation of the LiDARs throughout the entire measurement period. Experiments in simulation as well as with real measurements have shown that our algorithm performs a continuous point cloud registration of up to four 64-layer LiDARs in real-time. The averaged resulting translational error is within a few centimeters and the averaged error in rotation is below 0.15 degrees.

Keywords: Accurate Global Positioning, Sensing and Intervening, Detectors and Actuators, Aerial, Marine and Surface Intelligent Vehicles
Abstract: Precise positioning of the Unmanned Aerial Vehicle (UAV) is critical to conduct many sophisticated civil and military applications in challenging environments. Many of the-state-of-the-art positioning methods rely on active range sensors. Among many available ranging sensors, Ultra-wideband (UWB) can provide many benefits such as high precision, power efficiency, and not prone to multipath propagation and noise. Thus, the UWB has recently been attracting many interests from the research community as a complementary positioning sensor. However, there is a significant lack of UWB benchmark data available to support developing, testing and generalizing their own positioning methods using UWB sensors. In this paper, we present a unique benchmark dataset that provides UWB and IMU signals acquired by a Q-Drone system in a diverse environment, including an indoor, open field, close to buildings, underneath the bridge and semi-open tunnel. This benchmark also provides ground truth of UAV positions independently measured with robotic total stations. In this paper, we present the characteristics of the UWB benchmark dataset, Q-Drone UAV platform and the results of the quality assessment conducted by a baseline positioning algorithm implemented with multilateration principal and non-linear optimization.

Keywords: Traffic Theory for ITS, Theory and Models for Optimization and Control, Data Mining and Data Analysis
Abstract: This study is an attempt to analyze the single-regime speed-density relations for urban expressways where traffic characteristics vary depending on time, vehicle, and location. A model is considered to be good and utilizable if it’s less complex in terms of number of parameters and numerically close to empirical data. For the analysis, high resolution Zen Traffic Data (ZTD), developed from the image sensing technology by the Hanshin Expressway Co. Ltd., Japan, is utilized which consists of all vehicle trajectory data. Firstly, the steady-state traffic data have been evaluated for varying spatiotemporal resolutions, which is primarily possible because of voluminous ZTD, followed by parameter estimation from empirical data. Using the values of jam density, approximated based on expressway composition, and kinematic-wave-speed, the value of another behavioral parameter: proportionality factor, is calculated. Functional and shape parameters are estimated using the Levenberg-Marquardt algorithm. Statistical metrics such as RMSE and ARE are used for assessing the model fitness in all categories of linear, exponential and complex forms of models and the performance of models at various resolutions. The analysis show that highly parameterized models have the lowest values of errors. However, May and Keller’s linear form of model can be considered of high potential for applications.

Keywords: Traffic Theory for ITS, Transportation Security, Other Theories, Applications, and Technologies
Abstract: Existing traffic flow forecasting technologies achieve great success based on deep learning models on a large number of datasets gathered by organizations. However, there are two critical challenges. One is that data exists in the form of “isolated islands”. The other is the data privacy and security issue, which is becoming more significant than ever before. In this paper, we propose a Federated Learning-based Gated Recurrent Unit neural network framework (FedGRU) for traffic flow prediction (TFP) to address these challenges. Specifically, FedGRU model differs from current centralized learning methods and updates a universe learning model through a secure aggregation parameter mechanism rather than sharing data among organizations. In the secure parameter aggregation mechanism, we introduce a Federated Averaging algorithm to control the communication overhead during parameter transmission. Through extensive case studies on the Performance Measurement System (PeMS) dataset, it is shown that FedGRU model can achieve accurate and timely traffic prediction without compromising privacy.

Keywords: Traffic Theory for ITS, Simulation and Modeling, Sensing and Intervening, Detectors and Actuators
Abstract: In this article, a method of traffic state prediction at intersection based video data is proposed. The method inherits the basic assumption of modified cell transmission model(MCTM) and depends on back propagation neural network (BPNN). The training set of the neural network consists of trafficdata from the video. In order to verify the good generalization of the prediction method, novel data is used as the verification set.The experimental results exhibit that the model has virtuous generalization. Especially, the model is suitable for short-term traffic prediction at intersections. The prediction results of the method serve to construct the traffic state prediction model (TSPM) of the urban traffic network. Moreover, making route arrangement and traffic guidance strategy also require them.

Keywords: Traffic Theory for ITS, Data Mining and Data Analysis, Travel Information, Travel Guidance, and Travel Demand Management
Abstract: Reliable estimates of typical travel times allow road users to forward plan journeys to minimise travel time, potentially increasing overall system efficiency. On busy highways, however, congestion events can cause large, short-term spikes in travel time. These spikes make direct forecasting of travel time using standard time series models difficult on the timescales of hours to days that are relevant to forward planning. The problem is that some such spikes are caused by unpredictable incidents and should be filtered out, whereas others are caused by recurrent peaks in demand and should be factored into estimates. Here we present the Wavelet Augmented Regression Profiling (WARP) method for long-term estimation of typical travel times. WARP linearly decomposes historical time series of travel times into two components: background and spikes. It then further separates the spikes into contributions from recurrent and residual congestion. This is achieved using a combination of wavelet transforms, spectral filtering and locally weighted regression. The background and recurrent congestion contributions are then used to estimate typical travel times with horizon of one week in an accurate and computationally inexpensive manner. We train and test WARP on the M6 and M11 motorways in the United Kingdom using 12 weeks of link level travel time data obtained from the UK's National Traffic Information Service (NTIS). In out-of-sample validation tests, WARP compares favourably to estimates produced by a simple segmentation method and to the estimates published by NTIS.

Keywords: Traffic Theory for ITS, Simulation and Modeling, Off-line and Online Data Processing Techniques
Abstract: This paper compares three filtering approaches to real-time freeway traffic state estimation based on the first-order and second-order macroscopic traffic flow models using fixed-sensing and mobile-sensing data together. The significance of online estimation of traffic flow model parameters along with traffic flow variables is also investigated in the context of heterogeneous sources of sensing data. The results indicate that the traffic state estimator based on a second-order model outperforms those based on two first-order models. In addition, the online model parameter estimation is significant for traffic state estimation especially when the penetration rate of mobile sensors is lower than 20%.

Keywords: Off-line and Online Data Processing Techniques, Data Mining and Data Analysis, Simulation and Modeling
Abstract: This work aims at unveiling the potential of Transfer Learning (TL) for developing a traffic flow forecasting model in scenarios of absent data. Knowledge transfer from high-quality predictive models becomes feasible under the TL paradigm, enabling the generation of new proper models with few data. In order to explore this capability, we identify three different levels of data absent scenarios, where TL techniques are applied among Deep Learning (DL) methods for traffic forecasting. Then, traditional batch learning is compared against TL based models using real traffic flow data, collected by deployed loops managed by the City Council of Madrid (Spain). In addition, we apply Online Learning (OL) techniques, where model receives an update after each prediction, in order to adapt to traffic flow trend changes and incrementally learn from new incoming traffic data. The obtained experimental results shed light on the advantages of transfer and online learning for traffic flow forecasting, and draw practical insights on their interplay with the amount of available training data at the location of interest.

Keywords: Off-line and Online Data Processing Techniques, Traffic Theory for ITS, Data Mining and Data Analysis
Abstract: Traffic state estimation (TSE), which reconstructs complete traffic states from partial observation data, is an essential component in intelligent transportation systems. In this study, a novel traffic state estimation method using connected vehicles and a single detector based on Kalman filtering and Rauch--Tung--Striebel (RTS) smoothing is proposed. To the author's knowledge, while filtering is common approach for TSE, smoothing has not been employed to TSE in the literature. The important features of the proposed method are twofold. First, thanks to RTS smoothing, it can estimate accurate traffic state using a single detector, and it does not require detectors in every entries and exits of a road section. In addition, the estimation accuracy is not significantly sensitive to detector location. Second, it does not require parameter calibration thanks to the method's data-driven nature. These features will make the method flexibly applicable for practical conditions. Estimation accuracy of the proposed method was empirically evaluated by using actual vehicle trajectories data, and the effectiveness of the above two features was confirmed.

Keywords: Off-line and Online Data Processing Techniques, Data Mining and Data Analysis, Simulation and Modeling
Abstract: Traffic state estimators (TSEs) based on the cell transmission model (CTM) are vulnerable to biased initial estimates of traffic flow parameters, in particular the critical density. For example, an overestimated (underestimated) initial estimate of critical density can cause a delayed (premature) switching of a TSE from free-flow working mode to congestion working mode, hence distorting estimates of traffic states. Only augmenting the traffic flow parameters into the state vector cannot resolve the issue of biased initial estimate of the critical density, because the critical density is unobservable under free-flow mode. To overcome this issue, this paper proposes an innovative supervisory observer to inform the TSE of correct instants for mode switching. In particular, the proposed supervisory observer requires no a priori knowledge of any traffic flow parameter. The idea is to make decisions for mode switching through capturing anomalies in the residuals of the extended Kalman filter (EKF) of the TSE. This paper also proposes, for the first time in relevant literature, to augment capacity drop proportion into state vector so that its value can be calibrated online. Simulation experiments show that the proposed method can correctly capture mode switching times and generate satisfactory estimates of traffic states and track time-varying traffic flow parameters.

Keywords: Off-line and Online Data Processing Techniques, Data Mining and Data Analysis
Abstract: The advent of new traffic data collection tools such as high-resolution signalized intersection controller logs opens up a new space of possibilities for traffic management. In this work, we describe the high-resolution datasets, apply appropriate machine learning methods to obtain relevant information from the said datasets and develop visualization tools to provide traffic engineers with suitable interfaces, thereby enabling new insights into traffic signal performance management. The eventual goal of this study is to enable automated analysis and help create operational performance measures for signalized intersections while aiding traffic administrators in their quest to design 21st-century signal policies.

Keywords: Off-line and Online Data Processing Techniques, Data Mining and Data Analysis, Travel Behavior Under ITS
Abstract: Being able to assess the state-of-health of a vehicle opens of course many possible applications. All the more so if the ongoing degradation of the monitored components can be provided continuously as the vehicle life extends over time. In modern shared mobility systems, thanks to which migration from ownership to usership models should eventually take place, developing means to actively monitor the state of the vehicle fleet is crucial to improve business models and feasible and predictive maintenance plans. Within this challenging context, the present paper focuses on the monitoring of the vehicle vertical dynamics, to understand, from the analysis of measured data, which is the combined effect of driving style and introduce road pavement roughness in determining the usage profile of the vertical-dynamics-related components of the vehicle, mostly the suspensions system. The proposed cost function concisely represents such wearing process, with the advantage of not requiring detailed parametric models of the vehicle dynamics and of the components themselves. The approach is tested on more than 9.000 km of trips carried out on four different vehicles, allowing to prove the effectiveness and generality of the approach.