Skip to Main Content
In these days, many production systems are consist of several factories. Such factories are dispersed in wide area and form ldquoproduction networksrdquo. In such networks, each factory produces intermediate materials for other factories. In order to operate production networks efficiently, some rational and sound operational strategy is needed for realizing cooperative operation. In the previous work, ldquobehavior modelrdquo of scheduling activities in decentralized production networks was developed and the validity was confirmed. Also, ldquonetwork-based support system for decentralized scheduling of distributed production systems through man-machine collaborationrdquo was developed based on the model together with technologies of Servlet, RDBMS, etc. However, performances of production systems obtained through the support environment were discussed insufficiently. In this work, an attempt was made to obtain proper scheduling rules by means of reinforcement learning. Concretely, profit Sharing was adopted in order to obtain rules for selecting the factory where intermediate materials are manufactured under the operational model proposed in the previous work. The state space of production systems were introduced so as to apply this approach to production networks discussed here. A series of experiments was carried out using this system in order to confirm whether proper rules can be learned.