Disturbance rejection sliding mode control for robots and learning design

The control of a robot to achieve both good dynamic and static performance against external disturbances is a challenging task, especially when high speed and a wide range of motion is required. In this paper, a disturbance rejection sliding mode control (SMC) methodology is designed for a robot manipulator. This methodology synthesizes the SMC design with the active disturbance rejection control (ADRC) technique. An extended state observer is employed to estimate unknown disturbances, which is difficult to deal with in a conventional SMC design, and to simplify the SMC law design. A learning-based parameter tuning methodology is presented to autonomously obtain the control parameters offline. To develop a robust and transferring controller, a neural network is used to learn the joint actuation ability for the controller optimizing process. Compared with other state-of-the-art controllers, both numerical simulations and experiments of a 6-DOF robot are provided to demonstrate the proposed control method and design methodology. These results reveal that the proposed control method has a satisfying tracking performance and strong disturbance rejection ability.