The event-triggered differential drive AGV model predictive robust trajectory tracking control
摘要
This paper addresses the trajectory tracking control problem of differentially driven automated guided vehicles (AGVs) under model parameter uncertainty and external environmental disturbances. A key challenge in this setting is to simultaneously satisfy kinematic constraints and achieve disturbance-resilient dynamic tracking when the AGV model is imperfect and disturbances are time varying. In practice, purely kinematic MPC may produce feasible velocity commands but cannot guarantee accurate velocity/torque realization under dynamic uncertainties, while purely robust dynamic controllers may reject disturbances but do not explicitly enforce kinematic constraints and can induce chattering or excessive control effort. This motivates a cascaded design that couples constraint-aware MPC with an online disturbance/uncertainty estimator. A cascaded robust control strategy integrated with a fixed-threshold event triggering mechanism is proposed. At the kinematic level, an AGV trajectory tracking error model is established and transformed into a quadratic programming problem. The model predictive control (MPC) method is introduced to generate the desired velocity commands (linear and angular velocities) from the pose tracking errors. Meanwhile, an event triggering mechanism is incorporated to enhance computational efficiency without sacrificing control accuracy. At the dynamic level, a nonsingular fast terminal sliding mode control with a modified reaching law is adopted to effectively reduce chattering. A Radial Basis Function neural network is introduced to perform online estimation and compensation of model uncertainties and unknown disturbances, and an adaptive update rate integrated with sliding mode control is designed. This integration not only counteracts disturbances but also ensures the stability of the closed-loop system, thereby significantly improving the robustness of the controller. A saturation function is incorporated into the sliding mode control to constrain control outputs exceeding the AGV motion limits, ensuring consistency with the MPC constraints. Simulation results for straight, sinusoidal, and circular trajectories demonstrate that, under the same event-triggered framework, the proposed method significantly outperforms ESO- and ELM-based schemes in both tracking accuracy and triggering efficiency. Compared to ESO-based schemes, the average position error is reduced by