Hybrid control of nonlinear dynamical systems using neural nets and conventional control schemes

A hybrid control scheme for the set-point change of nonlinear systems is described. The essence of the scheme is to divide the control into two different stages, namely, coarse control and fine control, and to use different controllers to accomplish the specific control action at each stage. For coarse control, a modified backpropagation neural network is used, which drives the system output into a pre-defined neighborhood of the set-point. The controller then switches to the fine control stage, at which time a linearization of the system model is identified around the set-point, and is controlled with an appropriate PID controller. Certain considerations are given to achieve smooth transition between the two different controllers. A simulation example is presented