Data-Driven Robust Model Predictive Control for Pose Tracking of Redundant Manipulators

In this paper, we present a data-driven robust model predictive control (DD-RMPC) method for pose tracking of redundant manipulators. Firstly, a MPC-based trajectory tracking framework is established, in which the joint limits at three different levels (angle, velocity, and acceleration) are satisfied. Secondly, to deal with the system model's uncertainty, accelerate the error convergence speed, and reduce tracking errors, a novel data-driven RMPC is proposed, in which the system's input-output data are used to compensate for the conservatism of the system. Finally, the simulation results on a seven degrees of freedoms (DOFs) redundant manipulator show that DD-RMPC provides a faster error convergence rate and achieves smaller pose tracking errors than that of the comparison methods.