Using a priori knowledge to design a controller for flexible joint robots

A unique controller for robot manipulators that exhibit joint compliance is developed by appealing to the engineer´s understanding of the problem´s physical dynamics rather than relying upon a more general design approach. The control approach is decentralized in the sense that resonant effects are controlled using local feedback at each joint. Using a lumped-parameter linear model, the controller is designed to eliminate: (1) the effects of joint compliance acting upon the actuator; and (2) the actuator dynamics. The design then proceeds to robustly stabilize the reduced-order system against the nonlinear robot dynamics that were initially ignored. The method is significantly less complex than the feedback linearization and inverse dynamic approaches. A single-link case is discussed in an example, but the analytical results are for the multilink case