Lyu, Hao (Southeast University), Guo, Yanyong (Southeast University), Liu, Pan (Southeast University), Wu, Yao (Nanjing University of Posts and Telecommunications), Yue, Quansheng (Southeast University), Wang, Ting (Tongji University)

KoopLCC: The Koopman Operator-Based Predictive Leading Cruise Control for Mixed Vehicle Platoons Considering the Driving Styles

Scheduled for presentation during the Poster Session "Automated vehicle platoons" (ThBT16), Thursday, September 26, 2024, 14:30−16:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

To address the issues of online computational burden and model-plant mismatch for mixed traffic control, this paper proposes a Koopman operator-based predictive leading cruise controller (KoopLCC) for mixed vehicle platoons. Specifically, the KoopLCC is based on a personalized human-driven vehicles (HDVs) dynamic deep Koopman model (PerKoopnet) and model predictive control (MPC) framework. Firstly, the PerKoopnet is developed, where the encoder adaptively extracts the short-term driving style from the historical HDVs state sequence, and then combines it with the current HDV state to form an observable high-dimensional linear space representation. Furthermore, the transition relationship between Koopman state and real state is utilized to structure multiple HDVs Koopman state space. Finally, the new linear state space-based KoopLCC is designed for real-time control of the mixed platoon. The HighD dataset is used in the experiment to validate the PerKoopnet. The results demonstrate that the PerKoopnet outperforms the baseline models in multi-step prediction performance. The simulation experiments are conducted to validate the KoopLCC. The results show that traffic oscillations are significantly reduced compared to the traditional MPC controller while ensuring the overall stability of the platoon. Moreover, the fuel consumption is reduced by 6.27% to 11.83%.