Nonlinear Model Predictive Control for Trajectory Tracking of Omnidirectional Robot Using Resilient Propagation

This paper proposes an enhanced Nonlinear Model Predictive Control (NMPC) framework that incorporates a robust, convergent variant of the resilient propagation (RPROP) algorithm to efficiently solve the Nonlinear Optimization Problem (NOP) in real time. The controller is developed for both constrained and unconstrained trajectory tracking of Wheeled Mobile Robots (WMRs), with operational constraints handled via the external penalty method. The proposed method introduces adaptive step sizes and a backtracking mechanism, significantly improving convergence speed without compromising accuracy. Simulation results show that, even under constraints, the proposed method reduces computational time by a factor of 6 to 11 compared to the Interior Point method and 2 to 4 compared to the Active Set method. In addition, it achieves superior tracking accuracy, with root mean square (RMS) position tracking errors reduced by approximately 50% relative to the benchmark methods. Real-time experiments conducted on the Robotino Festo Omnidirectional Mobile Robot (OMR) validate the method’s practical effectiveness, demonstrating faster convergence and improved velocity tracking performance, while maintaining comparable or better position tracking. These findings establish the proposed controller as a computationally efficient and accurate solution for real-time WMR trajectory tracking.