A SET-POINT CONTROLLER-DESIGN FOR NONLINEAR STOCHASTIC-SYSTEMS

A method of nonlinear system controller design is presented for the class of systems that can be represented by nonlinear memoryless elements embedded in a linear dynamic system. The approach is based on a sampled-data technique for simulating nonlinear systems that involves the use of one additional state variable for each nonlinear element. The controller consists of an estimator, which was developed along the lines of the basic Kalman filter, and a neural net trained from data collected using selected sets of feedback gains that are established for a predetermined number of operating points of the plant.