基于LGSVL/Apollo的网络延迟攻击下自动驾驶车辆定位估算

为提高网络延迟攻击下自动驾驶车辆定位估计算法的精确度,研究了延迟模型下自动驾驶车辆定位的无偏差有限脉冲响应(UFIR)估计器设计方法,并仿真实验。搭建延迟攻击下的车辆运动学模型,拓展模型至有限长度的时间窗口,推导UFIR算法按批处理式与迭代式表达形式,分析Apollo系统各功能模块的数据流动,基于LG开源自动驾驶仿真器(LGSVL)与Apollo系统,搭建联合仿真测试平台并开展实验。结果表明:与Kalman滤波器(KF)相比,该算法估计精确度更高;当延迟数据出现较大变化时,算法响应速度更快,波动幅值更小,鲁棒性更强。当数据延迟时间小于等于1 s时,估计效果良好。因而,验证了基于LGSVL与Apollo系统进行自动驾驶仿真实验的可行性。

Localization estimation algorithm under cyber delay attack for autonomous vehicle based on LGSVL/Apollo

An unbiased finite impulse response filter(UFIR)under delay model was proposed to improve the accuracy of autonomous vehicle location estimation algorithm under delayed attack.A vehicle kinematics model under delayed attack was established and extended to a finite length time window.A batch and iterative forms of UFIR algorithms were derived.The embedding position of the algorithm was selected by analyzing the data flow of Apollo functional modules.A co-simulation test platform was built based on LG Silicon Valley Lab(LGSVL)Simulator and Apollo system,and conducted experiments.The results show that compared with the Kalman filter(KF),the algorithm has higher estimation accuracy,faster response speed,smaller fluctuation amplitude,and stronger robustness when the delay data changes greatly.The estimation effect is great when the data delay time is less than or equal to 1 s.Therefore,the result verifies the feasibility of the autonomous driving simulation experiment based on LGSVL and Apollo system.