Abstract
INTRODUCTION: Accurate path tracking is essential for achieving intelligent operation in unmanned agricultural machinery. To address the limitations of traditional agricultural machine path tracking methods, which are susceptible to high frequency oscillations and external disturbances, this study proposes a fixed time super-twisting sliding mode adaptive path-tracking control for unmanned agricultural vehicles. METHODS: The approach utilizes a Regularized Least Squares Extreme Learning Machine (RLS-ELM) to improve robustness and adaptability under certain operating conditions. A generalized terminal sliding mode surface is first designed by incorporating both lateral and heading deviations of the vehicle. Next, a Super-Twisting Sliding Mode control law is developed to perform path tracking, while the RLS-ELM is used to estimate and compensate for unknown disturbances. The stability of the proposed control system is verified through the construction of a new Lyapunov function. RESULTS: The control algorithm is validated via field experiments on an agricultural platform. Results show that, compared to the Fixed-Time Generalized Terminal Super Twisting control method (FGST) and the Fixed-Time Sliding Mode Controller (FTSMC), the Extreme Learning Machine-Adaptive Fixed-Time Generalized Super-Twisting (ELM-AFGST) method reduces lateral mean absolute errors by 24.5% and 27.4%, respectively, and decreases heading mean absolute errors by 5.4% and 30.8%, respectively. These findings demonstrate that the proposed path tracking method provides a solid theoretical framework for high-precision path tracking of unmanned agricultural machines.