基于卡尔曼滤波与可达集的自主车辆安全性验证方法

Autonomous vehicle safety verification method based on Kalman filtering and reachable set

针对目前自动驾驶中车辆安全性验证缺少动态验证过程以及安全性验证结果保守度过高的问题,构建车辆动力学和车辆跟踪偏差模型,结合卡尔曼滤波与传感器数据对目标集动态状态进行预测更新,通过可达性分析计算可达集;提出准确度指标、优化度指标以及风险指标评价可达集,通过优化各个指标参数以平衡可达集保守度和危险度,在此基础上建立车辆安全等级,实现了自主车辆动态安全性验证任务。结果表明,基于卡尔曼滤波与可达集的自主车辆安全性验证方法不仅可以在平衡保守度和风险度的情况下验证自主车辆安全等级,还可以为车辆监控预警系统提供依据。

Aiming at the current problem of the lack of dynamic verification process of vehicle safety verification in autonomous driving and the excessively conservative safety verification results,the vehicle dynamics and vehicle tracking deviation model are constructed.The dynamic state of the target set is predictively updated by Kalman filter combining with sensor data,and the reachable set is calculated by accessibility analysis.The accuracy index,optimization index and risk index evaluation set are proposed,and the vehicle safety level is established on the basis of optimizing the parameter of each index to balance the conservatism and danger of the attainment set,and the dynamic safety verification task of autonomous vehicles is realized.The results show that the autonomous vehicle safety verification method based on Kalman filtering and reachable set can not only verify the safety level of autonomous vehicles without balancing conservatism and risk,but also provide a basis for vehicle monitoring and early warning system.