Reaching desired states time-optimally from equilibrium and vice versa for visco-elastic joint robots with limited elastic deflection

Recently, intrinsically elastic joints became increasingly popular due to several reasons. Most importantly, elasticity improves impact robustness and, if used wisely, energy efficiency. Potential energy storage and release capabilities in the joints allow to outperform rigid manipulators by means of achievable peak link velocity. It has therefore been of great interest to find explosive or cyclic motions, similar to those of humans or animals, that make systematic use of joint elasticity. In this context, we address two important control problems in the present paper. First, we find all potential system states that a visco-elastic joint with constrained deflection may reach from its equilibrium state and analyze the influence of system parameters on the according reachable set. While high link velocities are certainly desirable in terms of performance, they may also increase the robot's level of dangerousness and/or the risk of self damage during potentially unforeseen collisions. Thus, we tackle the problem of how to brake a visco-elastic joint in minimum time. Furthermore, the results are extended to a near-optimal real-time control law for elastic n-DOF manipulators. The proposed braking controller is experimentally verified on a KUKA/DLR LWR4 in joint impedance control.