Adaptive model-based control of the Utah/MIT hand

Abstract

Although previous case studies rigorously investigated the feasibility of implementing passivity-based inner loop adaptation on an industrial manipulator, an unanswered question was the general validity of those results. That issue is addressed by repeating the analysis on a small serial robot with a pneumatic tendon drive system. Experimental analysis reverified the algorithm's adaptation/learning capabilities and provided insight into how the implementation issues of tuning and parameter convergence depend on specific manipulator structure. Direct adaptive control enhances trajectory efficacy for a wide range of manipulators.<>