REPOSITIONING CONTROL OF ROBOTIC ARMS BY LEARNING

The problem of moving a rigid robot between equilibrium points by mean s of a learning algorithm, which uses only the positioning error at th e end of a trial, is investigated. After high gain feedback linearizat ion of the robot dynamics, it is shown that simple, robust, finite dim ensional learning algorithms can be set up to accomplish this task for unconstrained robots and robots subject to smooth bilateral constrain ts for which hybrid force control is of interest.