The testing of autonomous vehicle (AV) safety is receiving a lot of attention due to several high-profile crashes caused by AVs. Popular AV testing techniques include test matrix and Naturalistic Field Operational Test (N-FOT) which are either not comprehensive or too costly and time-consuming. In this paper, a new method based on Subset Simulations (SS), is proposed as an alternative method for evaluating the safety of AV. The lane change scenario is used to demonstrate the performance of SS, and a new Gaussian Mixture Model with 8 parameters describing the stochastic nature of lane change scenario is developed. The new testing procedure achieves a similar level of accelerated rate compared with IS, and SS can easily handle black-box control systems.
Accelerated Evaluation of Autonomous Vehicles in the Lane Change Scenario Based on Subset Simulation Technique
2018-11-01
787526 byte
Conference paper
Electronic Resource
English
Framework for Simulation-Based Lane Change Control for Autonomous Vehicles
British Library Conference Proceedings | 2017
|