Evolutionary game-theoretic approach for shop floor control

Shop floor control has been a key area for research for several decades. Most of the control methods developed are centralized and static. This paper presents a real-time distributed shop floor control system which dynamically routes and schedules the jobs among several nonidentical candidate machines using real-time information. The approach used is based on game-theory where the operations to be performed on the parts act as players with the alternate machines used as their strategies. Each machine exchanges information about its status with the other machines, and each machine displays some characteristic feature that affects its behavior. The objective of the players is to maximize their payoff. The payoff to an operation using a particular strategy considers the present system status and the resulting future status, thus avoiding myopic decisions. The results of the games-theoretic approach are compared to four dispatching rules under varying arrival rates and failures rates of the machines. A simulation study shows that each dispatching rule performs well within a small range of system parameters. The game theoretic approach, however, adapted itself well to the various system conditions, and performs well over the entire range of arrival rates and failure rates.