自动驾驶汽车状态估计的矩阵加权自适应扩展卡尔曼滤波方法

Matrix-Weighted Fusion AEKF Method for Autonomous Vehicle States Estimation

针对自动驾驶汽车状态和故障估计问题,基于自适应扩展卡尔曼滤波(AEKF)理论提出了一种矩阵加权多传感器信息融合估计方法。首先,将自动驾驶汽车运动学模型进行离散化处理;然后,以极小化多传感器量测估计误差向量均方和为目标,采用拉格朗日极值求解方法设计了矩阵加权多传感器信息融合方法;最后,在低精度传感器、高精度传感器、融合估计3种条件下仿真验证该方法的正确性。结果表明,与单个传感器情形相比较,所提出的方法能较好地估计自动驾驶汽车的位置、航向角、速度信息以及执行机构故障信息,为自动避障、自动泊车等精准测控任务提供参考。

For the estimation of state and fault of autonomous vehicles, this paper proposes a matrix weighting multisensor information fusion method based on Adaptive Extended Kalman Filtering(AEKF) theory. Firstly, the dynamics model of the automatic driving vehicle is discretized. Then, focusing on minimizing the mean square sum of multi-sensor measurement estimation error vectors, a matrix weighted multi-sensor information fusion method is designed by using Lagrange extreme value solution method. Finally, 3 simulation scenarios of low precision sensor/high precision sensor/fusion estimation are designed to verify the correctness of this method. The simulation results show that, compared with the single sensor case, the proposed method achieves better estimation of the position, direction and speed information of the automobile, and the fault information of the actuator, so as to provide reference for automatic obstacle avoidance and automatic parking.