In response to the challenges in mobile robot path tracking using model predictive control, where the predictive model weakens the controller’s ability to respond to sudden changes in the reference path curvature and heading, this paper proposes a composite control strategy suitable for agricultural robots. The strategy combines the maximum torque per ampere control and an Extended State Observer (ESO). The paper initially establishes a mathematical model for a Permanent Magnet Synchronous Motor (PMSM) considering aggregated disturbances. It designs a position tracking controller based on a non-singular terminal sliding mode and convergence law. This controller, employing a non-cascaded structure, replaces traditional position and velocity loop controllers and is proven to be stable with finite-time convergence through Lyapunov’s theorem. To enhance the system’s disturbance rejection capabilities further, the paper introduces an ESO to estimate system disturbances and applies it for feedforward compensation. The paper concludes by providing stability proof for the overall PMSM servo system in agricultural robots. Finally, the paper conducts simulations and experimental verifications based on the designed controller, demonstrating that the controller exhibits excellent path tracking performance, fast convergence, and robustness against external disturbances.