Critical Scenario Generation for Realistic Testing of Autonomous Driving System via Multi-vehicle Cooperative Control

Among the three major safety assessment processes (i.e. simulation, closed-course test, and public road test) for highly automated driving systems (ADS), closed-course tests offer a high level of fidelity and provide a secure and controllable testing environment. However, due to the lack of realistic background traffic, scenarios in closed-course tests are usually static and limited. To address this limitation, this paper proposes a new safety assessment framework, which integrates cooperative speed optimization of multiple traffic participants and active conflict generation along the reference path. Through the speed planning for background vehicles and background dummies, the conflict between the background participants and the tested autonomous vehicle is generated to test the performance of the autonomous vehicle in corner cases. The effectiveness of the proposed conflict generation algorithm is validated through MATLAB simulations and implemented at Tongji University’s intelligent connected vehicle evaluation base. Test results demonstrate the capability of the proposed framework to assess the safety performance of highly ADS in a cost-effective manner.