Redundancy Resolution of an Index Finger Exoskeleton using Self Organizing Map

This paper discusses the redundant control of a 3 degrees of freedom (DOF) index finger exoskeleton with a neural network based learning strategy. More specifically, the Kohonen's self-organizing map (KSOM) based control scheme is proposed for this redundant exoskeleton. The inverse kinematic relationship between the joint space and Cartesian space of the exoskeleton is learned using the KSOM. The result of this learning is the approximate estimation of the Jacobian matrix inverse, making the algorithm more efficient. Straight line object translation task is considered to test the performance of the proposed redundant control strategy through simulation and experimental studies.