10.4.1 The General Philosophy

This section briefly introduces the basic ideas of a framework that has been highly popular in the artificial intelligence community in recent years. It was developed and used primarily by machine learning researchers [19,930], and therefore this section is called reinforcement learning. The problem generally involves computing optimal plans for probabilistic infinite-horizon problems. The basic idea is to combine the problems of learning the probability distribution, $P(\theta\vert x,u)$, and computing the optimal plan into the same algorithm.