A new adaptive control algorithm based on signal characterization

The problem of the design and implementation of control strategies foe a class of systems with unknown or partially known dynamics is investigated. The proposed solution is based on input/output (I/O) signal characterization, as opposed to the familiar model-based (adaptive) algorithms. It is shown that the approximation of I/O signals by a set of orthogonal functions is adequate for generating control signals under little a priori information. It is argued that such characterization is well suited for implementation on neural networks, which offer a definite speed advantage. Several examples are given to illustrate the application of the proposed method