Optimal control concepts for intrinsic elastic joint robots

Abstract

A huge part of (industrial-) robots is still modeled and designed as rigid and stiff multibody systems. In contrast to this common rigid mechanical systems, those with passive stiffness or even with adjustable variable stiffness are capable of storing potential energy. Humans and other biological systems use this intrinsic compliance to enhance energy efficiency, performance, and robustness. Since some years a paradigm change towards elastic actuation is taking place in robotics. Elasticity is no longer considered as unwanted on the contrary, it becomes clear that significant joint elasticity can act as a key feature for robotic design. Besides the regularly mentioned aspects of beneficial safety characteristics and robustness, very little work has been carried out for exploiting the natural dynamics of elastic joint robots up to now. This thesis investigates how the potential energy in the joint can be exploited in order to reach high link side velocities, that are potentially significantly higher than the maximum motor speed. In particular, the performance of a rigid and a elastic robot are compared on this basis. As an exemplary high speed task, a throwing motion shall be realized by different types of robots. For that reason optimal control techniques are adapted to solve the system equations for an appropriate input control trajectory. The relevant real-world input and state constraints are considered. It can be shown that an optimization of the end effector velocity leads to an excitation motion and in consequence to an energy transfer from the elastic joint energy into link side velocity. This enables that the robot moves faster on the link side than the maximal achievable motor speed. The developed methods and simulation results are experimentally verified with the new DLR-Hand-Arm System (HASY).