Accelerate the autonomous vehicles reliability testing in parallel paradigm

Wuling Huang, Yisheng Lv, Long Chen, Fenghua Zhu

Abstract: Autonomous vehicle testing is important. Testing tools are developed to carry out simulation or real field autonomous driving testing, which are both common used methods. However, real field on-road testing is with low efficiency and limited ability to test all critical scenarios due to safety and costs involved. How to accelerate the testing and achieve the full coverage is the goal of this research. Therefore, use the simulation methods to achieve the full coverage test, especial the corner cases, use the real field test methods to do the importance sampling verification, obtain the parameters estimate through iteration of testing data, and then the functional testing of Autonomous vehicle can be accomplished during the develop stage. This paper discusses the autonomous driving reliability accelerated testing in parallel paradigm and proves the importance of this method, which is an appropriate reference for autonomous vehicle testing.