Driving Safety Field-Based Scenario Generation of Key High-Risk Test Scenario Library for Intelligent Connected Vehicles

The construction of scenario library for autonomous driving algorithm testing has always been a difficult problem in intelligent connected vehicles. Aiming at the problem that the proportion of key high-risk test scenarios that can effectively test the automatic driving algorithm in the natural driving scenario library is too low, this paper proposes a method for generating key high-risk test scenario library of intelligent connected vehicles based on the driving safety field to improve the proportion of key high-risk scenarios in the test scenario library. Firstly, based on the theory of the driving safety field, the Scenario Potential Risk-Driving Safety Field (SPR-DSI) is taken as the risk evaluation index representing the potential risk of automatic driving scenarios and verified. Then, the threshold method is used to extract the scenario elements, and the correlation analysis between the scenario elements and SPR-DSI is carried out to determine the key scenario elements. After that, based on the Genetic Algorithm (GA), SPR-DSI is used as the risk evaluation index for generating scenarios, and the key scenario elements are taken as chromosomes to generate a database of key high-risk test scenarios for autonomous vehicles. Finally, compared with the scenario library generated by Markov Chain Monte Carlo (MCMC) and Automatic Efficient Test Generator (AETG), the results show that, the high-risk test scenario library generated based on driving safety field effectively increases the proportion of key high-risk test scenarios, which helps to fully verify the safety and robustness of automatic driving algorithm.