Optimizing coverage of simulated driving scenarios for the autonomous vehicle

Abstract

Self-driving cars and advanced driver-assistance systems are perceived as a game-changer in the future of road transportation. However, their validation is mandatory before industrialization; testing every component should be assessed intensively in order to mitigate potential failures and avoid unwanted problems on the road. In order to cover all possible scenarios, virtual simulations are used to complement real-test driving and aid in the validation process. This paper focuses on the validation of the command law during realistic virtual simulations. Its aim is to detect the maximum amount of failures while exploring the input search space of the scenarios. A key industrial restriction, however, is to launch simulations as little as possible in order to minimize computing power needed. Thus, a reduced model based on a random forest model helps in decreasing the number of simulations launched. It accompanies the algorithm in detecting the maximum amount of faulty scenarios everywhere in the search space. The methodology is tested on a tracking vehicle use case, which produces highly effective results.