An approach to nonlinear feedback control with applications to robotics

A control problem involving a mechanical system with generalized coordinates q∈R^m is considered. The error in tracking a desired input y^d∈R^p is e=E(q,y^d)∈R^m. If E satisfies simple conditions, it leads to a nonlinear control law that assures e(t)→0 as t→∞. The law is robust in that small changes in it do not produce large steady-state errors or loss of stability. In this theory a unified framework is presented for treating a number of problems in the control of mechanical manipulators.