Computationally Efficient Predictive Learning Control for Wheeled Mobile Robots: An Offline Sampling Approach

This article develops a practical predictive iterative learning framework for repetitive trajectory tracking tasks of wheeled mobile robots with lightweight computational resources. To improve input smoothness and reduce computational cost, an offline sampling-based predictive learning control (S-PLC) approach is proposed. Moreover, the design of S-PLC is analyzed through constructing the cost function term by term in accordance with the control objectives of high tracking accuracy, smooth control commands, and minimal computational burden. By leveraging the offline iteration-by-iteration execution property, additional initialization and filtering strategies are incorporated to improve its practical applicability. Both comparative simulations and real-world experiments validate the effectiveness and robustness of the proposed S-PLC.