Wang, Zhaoyi (Tongji University), Li, Xincheng (Tongji University), Li, Shizhen (Tongji University), Yang, Shuo (Tongji University), Du, Jiatong (Tongji University), Zhang, Xinyu (Tongji University), Huang, Yanjun (Tongji University)

Safety Evaluation of Autonomous Driving Based on Safety-Critical Scenario Generation

Scheduled for presentation during the Invited Session "Data-driven and Learning-based Control Techniques for Intelligent Vehicles" (FrAT1), Friday, September 27, 2024, 11:10−11:30, Salon 1

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

Knowing the safety capacity of an autonomous driving system is crucial for evaluating and improving it. To achieve this, safety-critical scenarios, essential for safety evaluation, must be explored and studied. This paper proposes a framework for safety-critical scenario generation and safety evaluation based on given algorithms. Firstly, a method for fine-grained safety-critical scenario generation based on reinforcement learning is proposed for efficient vulnerability exploration. Then, scenario features are non-linearly extracted and clustered for safety analysis. In addition, based on the distribution map of scenario features, the safety boundary of the autonomous driving algorithm is defined, identified, and analyzed. Finally, two given algorithms are evaluated with the proposed framework. The main accident categories of given algorithms are summarized from the perspective of vehicle interaction relationships, and their safety boundaries are quantitatively described and analyzed in conjunction with the distribution map of scenario features. The case study demonstrates the effectiveness of the proposed framework for safety-critical scenario exploration and safety analysis. The proposed framework allows for a comprehensive safety analysis for any given algorithm, which is of great significance for the evaluation of autonomous vehicles.