Learning-based double layer control method of yaw stability simulation research for rear wheel independent drive electric tractor plowing operation

Abstract

Complex field environments and variable soil conditions make tractor plowing operations susceptible to maneuvering instability, especially challenging in yaw stability. A learning-based double layer control (LDLC) method of yaw stability for rear wheel independent drive electric tractor (ET) plowing operation is proposed in this paper in order to solve this problem. Firstly, a sliding window self-learning traction resistance predictor is proposed based on Gaussian process regression (GPR) theory to predict the future traction resistance of plowing operation. Then, a discrete nonlinear model predictive control (NMPC) supervisor yaw stability controller is designed based on the dynamics model of the plowing unit and the tire-soil model to solve for the additional yaw torque required to maintain the yaw stability of the tractor. Finally, considering the additional yaw moment and driver demand torque, the subordinate controller composed of a particle swarm optimization algorithm optimally allocates the driving torque of the two driving wheels. Hardware-in-the-loop test results show that the maximum mean tracking errors of yaw rate and sideslip angle are 0.002 rad/s and 0.01 rad for the LDLC controlled ET during straight line and continuous steering under normal plowing operation, the maximum mean tracking errors of yaw rate and sideslip angle are 0.007 rad/s and 0.012 rad for straight line and continuous steering plowing operations under harsh plowing operation. The tracking accuracy is significantly higher than traditional cascade PID and equal distribution control methods. The proposed LDLC method effectively ensures the yaw stability of the ET during straight line and continuous steering under normal and harsh plowing operation conditions.