A TV-MPC Methodology for Uncertain Under-Actuated Systems: A Rotary Inverted Pendulum Case Study

This paper presents a systematic approach of formulating a Time-Varying Model Predictive Control (TV-MPC) framework for uncertain and under-actuated mechanical systems. The proposed methodology utilizes the nonlinear decomposed dynamics in conjunction with a special class of orthonormal basis functions – the Laguerre functions in the model structure. A possible numerical ill-conditioning problem, for large prediction horizons, has been coped with using the idea of exponential data weighting in the cost function, which results in condition number improvement, for the main TV-MPC algorithm. A rotary inverted pendulum is considered as a case study under-actuated system. The content of this research revolves around the TV-MPC treatment for cubic polynomial type reference position tracking problem using the decomposed nonlinear dynamics in the TV-MPC model structure and using Laguerre functions for future control trajectory modeling and motion predictions of the rotary servo arm and the pendulum bar. Finally, the applicability of TV-MPC algorithm is demonstrated with the help of simulation results for the subject benchmark system.