Robust control of robots by the computed torque law (Q1175506)

A method is presented for the selection of feedback gains in order to make the computed torque control law, applied to robotic manipulators, robust under bounded, unknowns dynamics. The computed value of the inertia matrix is chosen to be either the estimate of the inertia matrix or any other positive definite matrix. The value of the nonlinear feedforward term is computed using the estimates of the nonlinear dynamics. The feedback gains are constant, and they depend only on the coefficients of a polynomial bound for the unknown dynamics. The resulting error system is shown via Theorem 2 to be uniformly ultimately bounded and stable if a certain constant gain is chosen large enough. Moveover, the tracking error is shown to approach zero asymptotically not only in the limit as the gains go to infinity, but also for the case of finite gains where the static balancing forque inputs are known. Finally a computed torque control law is presented that satisfies the outlined requirements and is simple and easily implementable.