Chapter 10 - Developments in Learning Control Systems

Abstract

This chapter focuses on the developments in learning control systems in general, and on the recent advances in iterative learning control (ILC) and direct learning control (DLC) systems in particular. The objective of the chapter is to present a number of new developments with two types of learning control methods: iterative learning control and direct learning control. ILC is one of the repetition-based learning schemes, whereas DLC is one of the pattern-based learning schemes. Iterative learning control differs from most of the existing control methods in the sense that it exploits every possibility to incorporate past control information, such as tracking errors and control input signals, into the construction of the present control action. A current cycle P-type learning scheme is developed and analyzed for nonlinear servo control systems. It shows that the existing feedback controller is very helpful in improving the control performance. A high-order PID-type learning scheme is then developed for nonlinear uncertain systems with state delays. A dual- PID scheme in both time and iteration number is suggested in the chapter, which shows the potential link with the well-established PID tuning methods. Direct learning control also differs from other pattern-based learning schemes in that it uses all available system structure knowledge and hence ensures exact tracking for a new trajectory by learning in a point-wise manner. DLC is able to generate the desired control profile from previous input–output patterns with either nonuniform time scales or nonuniform magnitude scales. Three typical applications of the learning control methods are described in the chapter.