Modeling and feed-forward control of structural elastic robots

In this paper an approach for modeling and control of robots with elasticities in power trains and in structural parts is presented and experimentally verified. For this purpose object-oriented, nonlinear models are developed in the modeling language Modelica. A system theoretical study of the generated models shows that a direct inversion of the models, due to the unstable zero dynamics, is not possible. Therefore an algorithm for the approximate inversion is developed. With this inversion method an approximate inverse model considering structural elasticity for a 6-axis robot is created and verified for the control of the robot. The new control leads to a considerable improvement of the driving characteristics of the robot in the experiment.