Lateral and longitudinal tracking control of tracked orchard transport vehicles based on adaptive MPC

Tracked transport vehicles offer strong terrain adaptability for hilly and mountainous orchards, yet path tracking remains challenging due to complex terrain and load variations. This study proposes a cooperative lateral and longitudinal tracking control strategy based on adaptive Model Predictive Control (MPC). A coupled dynamic-kinematic model of the tracked vehicle is constructed, while Global Navigation Satellite System (GNSS)/Inertial Navigation System (INS) fusion and a two-stage recursive least squares algorithm with a forgetting factor are employed for online estimation of vehicle states and key parameters. The control framework employs Nonlinear MPC (NMPC) for lateral trajectory tracking and mass-adaptive Linear Time-Varying MPC (LTV-MPC) for longitudinal speed control. Simulation results based on mathematical paths and measured orchard trajectories demonstrated that the proposed adaptive MPC outperformed traditional control methods in tracking performance. In field tests in orchards, the average lateral trajectory deviation of the adaptive MPC under full load conditions (0.085-0.092 m) was better than under empty load conditions (0.125-0.140 m), while the traditional algorithm had a worse average lateral trajectory deviation under full load conditions than under empty load conditions. This advantage was attributed to the proposed algorithm's ability to adaptively account for changes in the vehicle's total mass and converge rapidly, improving control accuracy on paths with sharp curvature and varying loads. This study provides an effective path tracking solution for tracked vehicles in complex terrain, supporting automated transport in smart orchards.