Pointwise control of dynamical systems using an optimal decision strategy with neural network trajectory learning

Pointwise control algorithms termed optimal decision strategies (ODSs) have proven effective for the control of nonlinear dynamical systems having control variable magnitude constraints. The authors propose a specific neural network structure and, through application of a certain ODS algorithm, show the stabilization of otherwise unstable systems. External decision strategies provide an alternative to potentially difficult function space optimization by controlling the system on a pointwise basis. That is, at each instant of time, a preferred velocity is selected from a set of permissible or achievable velocities. It is the relation of the achievable velocity to the desired velocity that determines the performance of the controlled system. The use of a neural network for providing desired velocity vectors is explored