Self organization of manufacturing systems-stochastic matrix learning automata approach

The authors discuss the self-organization of distributed manufacturing controllers which employ Stochastic Matrix Learning Automata (SLA) theory (Narendra et al.) Learning Automata, Prentice Hall, 1989 as a learning method. An SLA based distributed learning controller is first proposed. One faces the following two problems when applying SLA to the distributed control: (1) the local rule updating does not prove the convergence in the global optimization, and (2) the explosion of state spaces causes insufficiency of learning. The authors discuss how to solve these problems applying genetic algorithms. The experimental results illustrate that the system is expected to automatically acquire the feasible knowledge, and that a genetic search can effectively solve the state explosion problem