Echtzeit-optimierung von kontaktkräften in mehrfingrigen manipulationssystemen

In this article a realtime optimization method to obtain optimal robot hand grasping forces in closed-loop control is presented. The algorithm is based on a problem formulation of unilateral contact constraints as a semidefinite program. This can be solved in a few milliseconds by a discrete-time gradient flow numerical iteration scheme. Experimental results obtained with the Tokyo Institute of Technology robot hand are presented validating the optimization approach to grasp manipulation. A multimedia appendix to the article provides a Matlab implementaton of the optimization algorithm, experimental data, and short movie clips of the experiments.