Holonic MAXCS and its application to Hot Strip Roller Scheduling

This paper proposes HMAXCS, which refers to Holonic Multiagent learning classifier system. HMAXCS is characterized by recursive holonic organizational control to a set of XCS based agents. Agent consisting of sub-agents with the same inherent structure are Holonic Agent. Holonic MAS provides terminology and theory for the implementation and realization of dynamically organizing agents. Here HMAXCS is applied to the production scheduling optimization of Hot Strip Mills(HSM) of a steel plant. Because the operation is real time and the scheduling problem contains so many parameters, it requires manual monitoring and controlling for the sake of timely quality production outputs. Simulation experiments with HMAXCS have exhibited that, the production delay and manual rescheduling control constraints are minimized. The robust predictive reactive scheduling with respect to co-operative reward distribution provides emergent scheduling behavior.