Research on Trajectory Tracking Control of Distributed Drive Electric Vehicles With Four‐Wheel Steering Based on Iterative Linear Quadratic Regulator

ABSTRACT Four‐wheel‐steering distributed‐drive electric vehicles (4WS‐DDEVs) provide multiple control degrees of freedom for improving path‐tracking and lateral‐motion performance. However, the coupling among front‐axle steering, rear‐axle steering, additional yaw moment, and nonlinear vehicle dynamics increases the difficulty of coordinated control. To address this issue, this study develops a coordinated path‐tracking framework based on an iterative linear quadratic regulator with particle swarm optimization–gray wolf optimization (PSO‐GWO‐ILQR). At the upper level, an online receding‐horizon ILQR controller determines the front‐axle steering angle, rear‐axle steering angle, and additional yaw moment. While retaining the conventional ILQR recursion, a differentiable hyperbolic‐cosine penalty is incorporated into the running cost to regulate excessive variations in these control commands. The state and control weighting parameters are calibrated offline using an update‐level PSO‐GWO mechanism that combines the multi‐leader guidance of GWO with the historical velocity information of PSO. A tracking‐ and stability‐oriented fitness function incorporating lateral error, heading angle error, sideslip angle, and yaw rate error is used to evaluate candidate weighting configurations. At the lower level, the axle‐level steering commands are converted into four individual wheel angles, and the wheel torques are allocated by considering the instantaneous steering geometry and wheel‐specific tire and actuator conditions. The proposed framework is evaluated through MATLAB/Simulink–Carsim co‐simulation under high‐speed/high‐adhesion and low‐speed/low‐adhesion operating conditions. Compared with the evaluated fixed‐weight ILQR baseline, PSO‐GWO‐ILQR reduces the maximum absolute heading angle error by 52.8% and 58.6%, respectively, and the maximum absolute yaw rate by 26.5% and 61.6%, respectively. The results indicate that the proposed framework provides a favorable compromise among tracking‐error attenuation, yaw‐response regulation, and control‐input smoothness under the investigated operating conditions.