Output trajectory tracking based on feedforward learning

This paper presents an iterative scheme for the feedforward output tracking control of a nonlinear system. In each iteration, the system is probed with a finite number of independent input signals and then the corresponding output signals are fitted to the desired output trajectory. For linear time invariant systems, a standard least square problem is solved. For nonlinear systems, this procedure amounts to a Newton-Raphson iteration with an approximate gradient. Simulation results involving the application to a 6-DOF PUMA 560 robot are included.