The planning of tracking control via search

The issue of computability which is associated with the design of systems combining feedforward and feedback controls is addressed. Hybrid systems of this kind are shown to be suitable for the separation of the control problem into two parts, i.e., planned operation and compensation for unmodeled dynamics, which permits better tracking of assignments than any feedback scheme without prediction. It is shown that the synthesis of feedforward commands relies on the solution of (often ill-posed) inverse problems, and that there exists a need for methods which provide approximate solutions in such cases. A definition of approximate inversion is constructed and a simple algorithm for performing this operation is described. Results are shown for SISO and nonlinear MIMO cases, and avenues for future research are suggested