Adaptive control based on RBF networks

This paper proposes a nonlinear direct adaptive controller based on radial basis function (RBF) networks and gives a new online learning algorithm, which modified the RLS algorithm with proportional, integral and derivative terms. The effect of these terms on the convergence behaviour is studied. The proposed control scheme is robust, reliable, efficient and simple. Compared with controllers based on BP networks, the proposed algorithm converges much more quickly without the problem of local minima. Simulation examples demonstrate the simplicity of the design procedure and the good characteristics of the control strategy.