基于滑动参数实时估计的履带车辆运行轨迹预测方法研究

Research on Trajectory Prediction of Tracked Vehicles Based on Real Time Slip Estimation

要实现履带车辆的无人驾驶,在轨迹规划阶段需要准确预测其未来一段时间内的运动轨迹,然而履带与地面之间的滑动使车辆运动轨迹的准确预测变得非常困难。通过研究转向过程中履带接地段的运动,建立基于瞬时转向中心的履带车辆运动学模型。针对车辆的相对位姿是滑动参数的泛函,雅可比矩阵难以求解的问题,通过对泛函微分方程线性化,推导了雅可比矩阵的解析解。根据车辆相对位置计算值和测量值的差值,运用Levenberg-Marquardt算法迭代求解滑动参数,并结合给定控制序列预测未来一段时间内车辆的运动轨迹。该方法不需要提前知道土壤参数,并且能够实时估计滑动参数,以适应路面变化。实车试验结果表明,与传统轨迹预测方法相比,利用该方法预测车辆轨迹时,车辆位置偏差减少30%以上。

In order to realize the unmanned driving of tracked vehicle, its future motion trajectory within a period of time should be accurately predicted in trajectory planning. It is difficult to predict the future motion trajectory of vehicle due to the slippage between tracks and ground. A kinematics model based on the instantaneous steering center is developed by studying the interaction of track and ground. The rela- tive pose of vehicle is a function of the slippage parameters, and the Jacobi matrix is difficult to solve. For this problem, the analytical solutions of Jacobi matrix are deduced by linearizoffing the functional dif- ferential equations. Slippage parameters are solved iteratively using Levenberg-Marquardt method accord- ing to the calculated and measured pose errors, and a set of input commands is given to predict the future motion trajectory within a period of time. The proposed method is used to update the slip parameters in real time without prior knowledge of terrain parameters. The real vehicle tests show that the position errors predicted using this method are reduced by more than 30% compared to the traditional trajectory predic- tion method.