Zhang, Chenxi (Xi'an Jiaotong University), Chen, Shitao (Xi'an Jiaotong University, Xi'an, China), Shi, Jiamin (Xi'an Jiaotong University), Xin, Jingmin (Xi'an Jiaotong University), Zheng, Nanning (Xi'an Jiaotong University)

Socially-Compatible Motion Planning of Autonomous Vehicles with Verification on Real Human Data

Scheduled for presentation during the Invited Session "Towards Human-Inspired Interactive Autonomous Driving II" (ThBT2), Thursday, September 26, 2024, 16:10−16:30, Salon 5

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 3, 2024

Abstract

One of the fundamental challenges for autonomous vehicles in complex dynamic scenarios is planning human-like trajectories in a socially acceptable manner. Existing methods, focused solely on self-interest and implicitly modeling a singular driving style, prove inadequate for the deployment of autonomous vehicles in mixed traffic environments while maintaining social acceptability. This paper presents the Socially-Compatible Motion Planning (SCMP) framework, which dynamically captures the spatiotemporal interaction characteristics of traffic participants for various driving tasks and adaptively adjusts the driving style of autonomous vehicles. Guided by driving style, the motion planning module considers vehicle physical constraints, safety constraints and comfort constraints, thereby generating human-like trajectories that incorporate multiple driving styles. The proposed method was rigorously validated and evaluated using the real-world dataset nuPlan, where it demonstrated exceptional performance in various driving scenarios and tasks.