High-Order Disturbance Observer Enhanced Explicit Nonlinear MPC for Robust Trajectory Tracking of Agricultural Robots

Autonomous robots operating in unstructured agricultural environments are often subjected to unmodeled dynamics and time-varying external disturbances caused by uneven and slippery terrain. To address these challenges, this article proposes a novel high-order disturbance observer-enhanced explicit nonlinear model predictive control (HODO-ENMPC) scheme for robust trajectory tracking of agricultural robots. Unlike most existing composite MPC frameworks, the proposed approach directly incorporates disturbance estimates and their predicted dynamics into the predictive model, thereby improving prediction accuracy and enabling offset-free control under both matched and unmatched disturbances. An extended kinematic model is developed to explicitly account for these uncertainties and disturbances. Based on this model, the analytical solution of the ENMPC is derived by a Taylor series expansion of the tracking error up to a specified order. Subsequently, a nonlinear HODO is designed to estimate these disturbances and their derivatives in real-time. Due to the distinct design philosophy, online optimization is not required in the proposed scheme, allowing for highbandwidth control of fast robot dynamics. The stability of the composite controller is rigorously established. Extensive simulations and field experiments on an agricultural robot validate the effectiveness of the method, demonstrating superior tracking accuracy and enhanced disturbance rejection compared with conventional NMPC, DO-NMPC, and DO-based ENMPC strategies.