Control Algorithms for Improving the Positioning Accuracy of Highly Compliant Robot Systems

Articulated soft robots (ASRs) increasingly gain importance due to their remarkable advantages such as mechanical robustness and temporal energy storage. Besides these properties, they are required to offer an appropriate interaction behavior and high po- sitioning accuracy in order to be versatile. Under-actuation and intrinsic oscillatory dynamics are some of the greatest challenges in achieving these goals. This work proposes a control concept that fulfills both requirements. For this purpose, an automatic switching between two control modes is introduced. One mode is a controller of the elastic structure preserving (ESP) framework that guarantees an appropriate in- teraction behavior by the direct adoption of impedance control techniques to ASRs, but does not compensate steady-state errors. For the second mode a combination of a linear PID controller for rigid robots and the ESP control concept is developed. While the integrator term of the PID controller can eliminate steady-state errors, it interferes with the desired interaction behavior. Therefore, the second mode is only selected when no interaction is detected and a steady-state error has to be eliminated. Saturation of the integrator term ensures safety while an anti-windup method prevents undesired effects caused by this saturation. Experiments on the single highly flexible test bed Softy show a significant improve- ment of the positioning accuracy in comparison to the one achieved by the original ESP controller. A statistical analysis of set-point regulation experiments on Softy and on a single rigid joint test bed of the Compliant Assistance and Exploration SpAce Robot (CAESAR) is provided. It proves that the positioning accuracy of the ASR Softy achieved by the proposed control concept is very similar to the one of the rigid joint test bed achieved by the standard controller of CAESAR. Furthermore, the proper function- ing of the control mode switch is verified by a corresponding evaluation of two of the conducted experiments. Hence, the proposed control concept can overcome main challenges of highly flexible joint robots such as intrinsic oscillatory dynamics to ensure high positioning accuracy and an appropriate interaction behavior