Path Tracking of Differential-Drive Tracked Agricultural Vehicles Using CPO-STSMC with Multi-Criteria Parameter Optimization

Differentially driven unmanned tracked agricultural vehicles operating on unpaved field roads are vulnerable to track micro-slip, lateral disturbances, and actuation lag, while manually tuned super-twisting sliding mode control often cannot simultaneously ensure tracking accuracy and steering smoothness. This study proposes a Crested Porcupine Optimizer-based super-twisting sliding mode controller (CPO-STSMC) for robust path tracking. Track slip and unmodeled dynamics are represented as bounded lumped disturbances in a coupled lateral–heading error model. A yaw-rate compensation law is developed with reference yaw-rate feedforward, a saturation boundary layer, and constraints on yaw-rate magnitude and rate. CPO is used offline to optimize key parameters through a fitness function that combines lateral and heading errors, peak errors, yaw-rate increments, and constraint penalties. Simulations and field experiments on unpaved farm roads are conducted using S-shaped and double-lane-change paths at 0.5, 0.8, and 1.0 m/s, with PP, SMC, and manually tuned STSMC as benchmarks. Compared with STSMC, CPO-STSMC reduces the root mean square lateral error by 23.9–30.0%, heading error by 23.9–28.0%, and yaw-rate increment by 11.8–22.3% in field experiments, while generally outperforming PP and SMC. The proposed method provides a practical balance among tracking accuracy, robustness, and steering smoothness for low-speed autonomous tracked agricultural vehicles.