DOI: 10.11908/j.issn.0253-374x.21004

作者:

作者单位: 同济大学 汽车学院，上海 201804


作者简介: 邢星宇（1991—），男，博士生，主要研究方向为自动驾驶汽车测试技术，车辆预期功能安全. E-mail： xingxingyu@tongji.edu.cn


通讯作者: 陈君毅（1980—），女，讲师，工学博士，主要研究方向为自动驾驶汽车安全性分析与性能研究. E-mail： chenjunyi@tongji.edu.cn

中图分类号: U467.3;TP391.9


基金项目: 国家重点研发计划（2018YFB0105103）




Automatic Testing Method Based on Optimization Algorithms for the Decision and Planning System of Autonomous Vehicles
Author:

Affiliation: School of Automotive Studies， Tongji University， Shanghai 201804， China


Fund Project:




摘要
| 图/表
| 访问统计
| 参考文献
|相似文献
| 引证文献
| 资源附件

摘要:利用仿真技术的场景测试方法已成为国内外研究热点，其中如何在大量场景中找到有价值的关键场景至关重要。针对上述问题，基于优化搜索算法提出了一种面向决策规划系统的关键场景自动化测试方法，能够克服传统场景测试方法的盲目性，提高测试效率。基于决策规划系统硬件在环测试平台，验证了该方法的有效性，并对比了不同搜索算法的关键场景生成效果。实验结果表明，贝叶斯优化算法和遗传算法相比于随机搜索算法产生危险关键场景的数量提高了3.3倍和2.5倍，配合自动化测试手段，方法能够有效提高场景测试效率。



Abstract:Simulation-based scenario testing methods have drawn significant research interests， and how to find critical testing scenarios among the infinite number of concrete scenarios becomes a critical issue. To solve this problem， this paper proposed a critical scenario generation and automatic testing method for the decision and planning system based on optimization and search algorithms. The method was verified through a hardware-in-the-loop platform， and the efficiencies of different search algorithms were compared. The experiment results show that the number of critical scenarios generated by the Bayesian optimization algorithm and the genetic algorithm is increased by 3 times and 2.5 times， compared with the random search algorithm. Combined with automatic testing， the method can quite improve testing efficiency.





PDF全文下载地址:

点我下载PDF