Neural-net based robust adaptive control of uncertain nonlinear composite systems

Abstract

A new robust adaptive hybrid control design, employing both math-analytical model and neural networks, for a class of uncertain nonlinear composite systems possessing similarity property has been derived. This design technique makes an adequate use of the structural characteristics of similar composite systems, and resolves the uncertainty issues in gains and interconnections by on-line updating the weights of the respective artificial neural nets. It depends on little a-priori knowledge and assures the system stability in closed-loop. It can be readily implemented within computer control systems, and it requires little computation effort and time. The application of this technique is illustrated by the real-world axis-tray drive mechatronic system.