NEURAL APPROXIMATIONS FOR MULTISTAGE OPTIMAL-CONTROL OF NONLINEAR STOCHASTIC-SYSTEMS

Two main approximations are used to solve a nonlinear-quadratic-Gaussi an (LQG) optimal control problem: the control law is assigned a given structure in which a finite number of parameters have to be determined to minimize the cost function (the chosen structure is that of a mult ilayer feedforward neural network), and the control law is given a ''l imited memory.'' The errors resulting from both assumptions are discus sed. Simulation results show that the proposed method may constitute a simple and effective tool for solving, to a sufficient degree of accu racy, optimal control problems traditionally regarded as difficult one s.