Implementation of the adaptive control algorithm for the KUKA LWR 4+ robot

Abstract

Model-based control methods are very attractive in the field of robotics as their tracking performance can exceed the classical controllers (such as the independent joint PID controllers). Using the dynamic model of the manipulator, however, requires detailed knowledge about the manipulator’s dynamic parameters such as link masses and inertias or joint friction properties. These parameters are not always easily identifiable and, to some degree, might vary between robots of one kind (e.g. slight differences in masses/inertias) or during the robot operation (e.g. friction changes related to the temperature). Thus, the identified model might not always be suitable for the desired control tasks. A possible method to overcome the aforementioned problems is to use the adaptive control scheme. In that approach, the parameters of the model are constantly updating their values in real-time to assure good tracking performance. This paper deals with the implementation of such an adaptive controller for the KUKA LWR 4+ robot. Using the KUKA’s communication protocols, a C++ implementation of the outer-loop adaptive controller (which feeds the KUKA controller with the desired joint torques) was created and its quality evaluated.