考虑滑移影响的履带式收获机MPPI路径跟踪控制方法

MPPI-based Path Tracking Control Method for Tracked Combine Harvesters Considering Slip Effects

摘要:

在田间作业过程中履带式联合收获机常因纵向打滑与横向漂移导致实际运行轨迹偏离规划路径，严重影响作业效果。而传统路径跟踪控制仍多基于理想无滑假设，难以准确反映履带-地表之间滑移导致的非线性相互作用。目前尚缺乏能够有效抑制滑移扰动并保证收获机持续稳定路径跟踪的实用方法，为此，本文提出一种考虑滑移影响的履带式收获机模型预测路径积分（SLIP-MPPI）路径跟踪方法。在运动学建模中引入纵向滑移率与横向位移近似模型，以修正状态传播;并在代价函数中增加滑移惩罚项，从而在采样优化过程主动规避滑移过大的不稳定轨迹。结合响应面方法对代价函数权重进行多目标优化，获得更合理的权重组合。田间试验结果表明，在复杂转向工况下SLIP-MPPI横向误差与航向误差均显著优于传统模型预测路径积分（MPPI）。正弦曲线路径试验结果表明，SLIP-MPPI最大横向误差和标准差分别降至0.029 m和0.010 m，比MPPI下降40.8%和41.2%;最大航向误差与标准差分别为1.84°和0.91°，较MPPI降低38.5%和27.8%。此外，直线路径收获试验结果表明，该方法能够保持稳定作业，作业带边界整齐，无明显偏移或蛇形轨迹，验证了SLIP-MPPI方法在田间复杂工况下有效性与鲁棒性。研究结果为履带式收获机在复杂土壤条件下实现高精度路径跟踪提供了一种可行方案。

Abstract:

During field operations, tracked combine harvesters often experience longitudinal slip and lateral drift, which cause deviations of the actual trajectory from the planned path and severely affect operational performance. However, conventional path tracking control is still mostly based on the ideal no-slip assumption, which fails to accurately capture the nonlinear interactions caused by slip between the tracks and the ground surface. At present, practical methods that can effectively suppress slip disturbances and ensure continuous and stable path tracking of harvesters remain lacking. To address this issue, a slip-aware model predictive path integral (SLIP-MPPI) path tracking method was proposed for tracked combine harvesters. In the kinematic modeling, the method incorporated an approximate model of longitudinal slip ratio and lateral displacement to correct state propagation. Additionally, a slip penalty term was introduced into the cost function, enabling the sampling-based optimization to actively avoid unstable trajectories with excessive slip. Furthermore, a response surface methodology was employed to perform multi-objective optimization of the cost function weights, yielding more reasonable combinations of weight coefficients. Field experiments demonstrated that SLIP-MPPI achieved significantly lower lateral and heading errors than conventional MPPI under complex steering conditions. In sinusoidal path tests, the maximum lateral error and standard deviation of SLIP-MPPI were reduced to 0.029 m and 0.010 m, respectively, representing decreases of approximately 40.8% and 41.2% compared with MPPI. The maximum heading error and standard deviation were 1.84° and 0.91°, corresponding to reductions of 38.5% and 27.8%. In addition, straight-line harvesting tests showed that the method maintained stable operation, with well-aligned working strip boundaries and without noticeable deviations or serpentine trajectories. These results verified the effectiveness and robustness of the SLIP-MPPI method under complex field conditions. The proposed SLIP-MPPI method provided a feasible solution for achieving high-precision path tracking of tracked combine harvesters under complex soil conditions, contributing to the advancement of agricultural machinery automation and the development of smart agriculture.