An Experimental Study on MPC based Joint Torque Control for Flexible Joint Robots

Abstract

This paper presents an experimental evaluation of joint torque control in a flexible joint robot using a model predictive control approach. The control of elastic robots is challenging due to the underactuated nature of the connected link and motor dynamics. A typical solution is to formulate the controller in a cascaded way with an outer loop controller for the nonlinear link side dynamics in combination with an inner loop controller for the joint torque. In this paper the inner loop torque controller utilizes a model predictive control design in order to take actuator constraints into account and, in case of a special choice of the torque reference and control gains, to avoid the computation of higher time derivatives in the desired joint torque. We compare two different controller implementations, which differ in the choice of the reference model. In particular, we investigate on the use of the natural open loop dynamics for the reference model. The controllers are evaluated in detail using a simulation study as well as in experiments on a single joint testbed.