A new method for designing stable neural control systems

摘要

In this paper, the conicity input-output stability criterion is applied as a design method to derive a neural controller for an unknown nonlinear dynamic system, so that the neural control system is stable. A reformulation is introduced for applying the conicity criterion to nonlinear plants controlled by a neural network, and an identification method is utilized to estimate the dynamics of the plant prior to the design of the neural controller. Application to a second order experimental plant demonstrates the effectiveness of the proposed method.