Automatic Generation of Intelligent Vehicle Testing Scenarios at Intersections Based on Natural Driving Datasets

作     者：Hu, Lin Lu, Tao Li, Gen Zhang, Xin Cai, Hai 

作者机构：Changsha University of Science and Technology School of Automotive and Mechanical Engineering Changsha410114 China Changsha University of Science and Technology Hunan Province Key Laboratory of Safety Design and Reliability Technology for Engineering Vehicle Changsha410114 China 

出 版 物：《IEEE Transactions on Intelligent Vehicles》 (IEEE Trans. Intell. Veh.)

年 卷 期：2024年第9卷第9期

页      面：5448-5460页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 081203[工学-计算机应用技术] 0837[工学-安全科学与工程] 080204[工学-车辆工程] 0835[工学-软件工程] 0714[理学-统计学（可授理学、经济学学位）] 0802[工学-机械工程] 0811[工学-控制科学与工程] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 0823[工学-交通运输工程] 082301[工学-道路与铁道工程] 

基　　金：This work was supported in part by the National Natural Science Funds for Distinguished Young Scholar under Grant 52325211  in part by the National Natural Science Foundation of China under Grants 52172399  and 52175088  and in part by the Natural Science Foundation of Changsha under Grant KQ2208235 

主　　题：Trajectories 

摘      要：Turning right at a junction without the aid of traffic signals is a frequent cause of intersection accidents and a situation that intelligent automobiles will need to handle well in the future. This allows this study to choose the traffic circumstances of turning right at junctions and traveling straight through them. It then suggests an automatic approach for creating test scenarios for intelligent vehicles based on Monte Carlo simulation. Through the examination of scene feature parameters and a natural driving dataset, a Markov chain-based right-turn trajectory model was created. Then, test vehicle trajectories and right-turn trajectories were produced using Monte Carlo simulations. By figuring out the average DTW (Dynamic Time Warping) value between the created and actual right-turn trajectories in the four speed intervals with the highest occurrence probabilities, the authenticity of the generated right-turn trajectories was confirmed. The produced trajectories were simultaneously loaded into the software for cooperative simulation. The AEB (Automatic Emergency Braking) system s shortcomings were shown by extracting the relative positions of the two cars before to the incident. The two popular varieties of AEB algorithm models were then compared using safety and comfort parameters. Furthermore, the test scenario s risk level can be managed by changing the scenario feature settings. There are some potential uses for the suggested test scenario generation method in the testing and assessment of algorithms for intelligent vehicles. © 2016 IEEE.