Abstract: By using learning control scheme the paper discusses tracking control for n joint robotic systems driven by n actuators. First, corresponding systems are turned into lower order systems by transformational reduction of order, and then the lower order systems are designed. The scheme eliminates the restriction of passing the same initial value for every learning in Kawamara's scheme and thus solves Kawamara's problem of applying his scheme to "the existence of initial deviation in learning". Simulation results show the efficiency and superiority of the new scheme.