Neural network-assisted PI^λD^µ control

Realization of PI^λD^µ controllers by feedforward neural network structures is studied in this chapter. The motivation of the study is the difficulty of realizing the fractional order operators in real time. A good approximation entails tens of poles and zeros scheduled appropriately in the frequency spectrum and the solution of such a high order system is needed later. This chapter remedies this computational problem by introducing a neural network model that imitates a given order fractional differintegration operator. Several simulations have been presented to assess the performance that can be obtained using neural alternatives of PI^λD^µcontrollers.