Direct Adaptive Inverse Control Based on Variable Radial Basis Function Network

This paper presents a neural direct adaptive inverse control strategy for nonlinear dynamical system. Based on identification of the nonlinear autoregressive model with exogenous inputs (NARX), the direct inverse control law can be derived and implemented for a process in a straightforward way. The identification is performed by variable radial basis function network (RBFN), which is realized by the minimal regularized orthogonal least squares (MROLS) algorithm for growing and pruning basis functions according to a design strategy. To improve network sensitivity to the change of system dynamics, the gradient messages of system output are also included into the RBFN. Simulation results demonstrate that the proposed strategy has strong adaptive ability and can be accomplished on-line effectively