基于无迹卡尔曼滤波的智能汽车驾驶员转向手力矩估算方法

Driver Steering Hand Torque Estimation Method of Intelligent Vehicle Based on Unscented Kalman Filter

中文关键词:  驾驶员手力矩  驾驶员干预识别  无迹卡尔曼滤波  未知输入观测

英文关键词:driver hand torque  driver intervention identification  unscented Kalman filter  unknown input observer

中文摘要:

      为了在L3、L4级智能驾驶车辆中有效地感应驾驶员的转向手力矩，使驾驶员在必要时能够快速平顺地接管车辆控制，提出一种基于无迹卡尔曼滤波的驾驶员手力矩动态估计方法。设计了有偏无迹观测器和未知输入估算方法，将估算结果和状态量通过无迹卡尔曼滤波器滤波。在Carsim和MATLAB/Simulink环境下搭建了系统的联合仿真模型进行仿真验证，并在PXI实时系统中进行了硬件在环实验。实验结果表明：在自动驾驶模式中，驾驶员手力矩为0时，估算误差小于0.2 N·m，当驾驶员以2 N·m手力干预自动驾驶时，估算误差小于0.7 N·m；当转向电机为助力模式时，估算手力矩均方根误差（Root Mean Squared Error,RMSE）为0.198 N·m，占输入范围的1.6%。上述结果误差可以有效满足对驾驶员手力矩的识别要求，因此能够识别自动驾驶过程中的驾驶员干预行为。

英文摘要:

      In order to effectively sense the steering hand torque of the driver in L3 and L4 level intelligent driving vehicle,so that the driver can take over the vehicle control quickly and smoothly when necessary,a dynamic estimation method of driver torque based on unscented Kalman filter is proposed.A biased unscented observer and unknown input estimation method are designed.The estimated result and state variables are filtered by unscented Kalman filter.The co-simulation model of the system in Carsim and MATLAB/Simulink environment is set up,and the hardware-in-the-loop experiment is carried out in the PXI real-time system.The results show that the estimated error is less than 0.2 N·m when the driver has zero manual torque in autonomous driving mode and less than 0.7 N·m when the driver intervenes in autopilot with 2 N·m manual torque.When the steering motor is in power assist mode,the root mean squared error (RMSE) of the estimated hand torque is 0.198 N·m,accounting for 1.6% of the input range.The results can effectively meet the requirement of driver hand torque identification,so it can effectively identify the drivers intervention behavior in the course of automatic driving.