Model-Free Adaptive Robust Iterative Learning Predictive Control for MIMO Nonaffine Nonlinear Systems

This paper presents a novel model-free adaptive robust iterative learning predictive control (ILPC) method for MIMO nonaffine nonlinear systems. First, a dynamic linearization data predictive model is developed using dynamic linearization and a multi-level hierarchical prediction approach to estimate the system’s behavior. Then, the estimation of the unknown pseudo-Jacobian matrix (PJM) in the data model is reformulated as a zero-seeking problem along the iteration axis, solved with robust noise-reducing zeroing neural network (NR-ZNN). Theoretical analysis shows that NR-ZNN achieves residual-free convergence and demonstrates strong noise suppression capabilities. Next, an ILPC scheme is developed, with theoretical proof of its asymptotic convergence along the iteration axis. Finally, the proposed scheme is validated through a typical example, demonstrating its excellent performance in completing tracking tasks in the presence of noise pollution.