Kim, Pooreumoe (Kakao Mobility), Ham, Seung Woo (Seoul National University), Park, Jung Soo (Kakao Mobility), Kim, Jungmin (Kakao Mobility)

Accurate Estimated Time of Arrival Prediction Using Real-Time Path-Level Speed Computation

Scheduled for presentation during the Poster Session "Travel Information, Travel Guidance, and Travel Demand Management" (ThAT16), Thursday, September 26, 2024, 10:30−12:30, 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 October 14, 2024

Abstract

Modern navigation typically measures vehicle speeds based on road segments, which may not accurately reflect actual traffic conditions, particularly on complex roads with merging or diverging lanes. On such complicated roads, traffic speeds vary among different paths taken by drivers, resulting in a mixture of traffic flows with distinct speeds within one road segment. Existing studies attempted to address this issue by computing speeds at the lane-level, while exhibiting limitations.

As a solution, we propose a methodology for computing speeds at the path level. Our algorithm automatically analyzes a national-scale road network, identifying sub-graphs where the mixture of different traffic flows occurs and distinguishes their paths. Subsequently, we construct a path set whose element is a path representing single traffic flow. Then, we compute the average speed along these paths using real-time GPS traces.

Finally, we applied our methodology to real-world navigation and conducted a fully online test. When predicting drivers' speed (in km/h) on the urban complex roads, the Mean Absolute Error (MAE) decreased by 52.09%, and the Root Mean Squared Error (RMSE) by 28.26% compared to the existing methods. Also, we assessed the accuracy of Estimated Time of Arrival (ETA) from drivers' origination to destination who traversed four complex roads in Seoul. The most notable finding emerged when drivers traversed an intersection located in a central downtown, where ETA MAE dropped by 18.56% and RMSE by 16.67%.