Practical consequences of inertia shaping for interaction and tracking in robot control

In trajectory tracking and interaction control of robots, two fundamentally different concepts define the boundaries within which most nonlinear model-based approaches can be located. On the one hand controllers such as the PD+ preserve the natural inertia and avoid feedback of external forces and torques. On the other hand controllers based on feedback linearization, as used in most inverse dynamics approaches, enforce linear closed-loop dynamics by means of external force/torque feedback. Here, these two basic concepts of keeping and shaping of the natural inertia are investigated and compared including aspects such as interaction behavior, tracking performance, tuning parameters, influence of modeling errors, and effective feedback gains. Exemplary case studies on a standard torque-controlled robot are performed. The understanding of these features and differences is of major importance for the proper selection and deployment of interaction and tracking controllers in practice.