Adaptive Generalized Predictive Controller and Cartesian Force Control for Robot Arm Using Dynamics and Geometric Identification

Abstract

In realistic situations such as human-robot interactions or contact tasks, robots must have the capacity to adapt accordingly to their environment, other processes and systems. Adaptive model based controllers, that requires accurate dynamic and geometric robot’s information, can be used. Accurate estimations of the inertial and geometric parameters of the robot and end-effector are essential for the controller to demonstrate a high performance. However, the identification of these parameters can be time-consuming and complex. Thus, in this paper, a framework based on an adaptive predictive control scheme and a fast dynamic and geometric identification process is proposed. This approach was demonstrated using a KUKA lightweight robot (LWR) in the performance of a force-controlled wall-painting task. In this study, the performances of a generalized predictive control (GPC), adaptive proportional derivative gravity compensation, and adaptive GPC (AGPC) were compared. The results revealed that predictive controllers are more suitable than adaptive PD controllers with gravitational compensation, owing to the use of well-identified geometric and inertial parameters.