Kinematics-Dynamics-Coupled Learning and Control of Discrete-Time Nonlinear Mechanical Systems with Completely-Unknown Dynamics

This paper proposes a novel kinematics-dynamics-coupled learning and tracking control scheme for a class of discrete-time mechanical systems with completely-uncertain and highly-nonlinear dynamics. This scheme is designed based on the deterministic learning technique for learning kinematics-dynamics-coupled function that contains all the crucial information of system model in control process. It can achieve the exponential convergence of both the parameter estimation error and state tracking error, with guaranteeing the satisfaction of partial persistent excitation condition. Empowered with this, the proposed approach develops distinctive capabilities of not only achieving locally-accurate identification/learning for system's completely-unknown and nonlinear dynamics through the online control process, but also realizing accurate tracking control for the system state with online using the learned knowledge. Extensive simulations have been conducted to demonstrate the effectiveness of the proposed approaches.