Advances and Challenges on Intelligent Learning in Control Systems

Abstract

Although there are many interesting and emerging technologies in the specialized area of intelligent learning in control systems, this chapter is devoted to four significant technical areas: reinforcement learning (RL), evolutionary learning control (ELC) using evolutionary algorithms (EAs), intelligent learning control, and learning via case-based reasoning (CBR). The resolving framework about reinforcement learning is well recognized as a machine learning paradigm especially well-suited to learning action policies for mobile robots. The chapter presents some experimental results from applying this framework to a mobile robot control task: behavior learning of soccer robots. The chapter discusses some stable intelligent learning approaches using Lyapunov stability theory for various fuzzy-neural-networks (FNN) control systems. CBR is a decision-making and learning method widely used in artificial intelligence. The purpose of using CBR is to provide a way to accumulate past experiences for future use, and CBR is sometimes considered as using analogy to solve problems.