Robust Model Predictive Control for Robot Manipulators

Inherent nonlinearities, external disturbances and model uncertainties hinder the performance of controlling real-world systems. In the present study, we proposed a robust model prediction-based virtual decomposition control method (RMP-VDC) as a modification of the VDC using the model predictive control (MPC) to offer a practical solution for the real system control problem. The proposed method deals with uncertainties and external forces, as well as constraint matters, for complex nonlinear robot manipulators. By modifying the ideas from the VDC with MPC techniques, the time-varying state feedback control law for the ancillary controller is provided. The proposed method benefits from the introduction of a prediction horizon, which induces robustness and increases accuracy. The constrained optimization problem is analytically solved online by the continuous linearization of the nonlinear model and by employing the active set method. To validate the proposed controller, we performed the implementation on a real 7-degrees-of-freedom upper body exoskeleton robot, and the results were compared with those obtained using the adaptive VDC. The experimental results revealed increased accuracy for the proposed RMP-VDC in dealing with model uncertainties and interaction forces between humans and exoskeleton robots.