Adaptive Negotiation for Agent Based Distributed Manufacturing Scheduling

Manufacturing scheduling problem is typically NP-hard. While traditional heuristic search based approaches have been considered not suitable for dynamic environments because of their inherent centralized nature, agent based approaches are promising for their decentralized, autonomous, coordinated, and rational natures. However, many challenging issues still need to be addressed when applying agent based approaches to complex distributed manufacturing scheduling environments. One of them, namely adaptive negotiation [2], is to integrate intelligence and rationality into negotiation mechanisms and make the system more adaptive in dynamic environments. This is very important to the manufacturing scheduling problem because the agent based scheduling process is essentially a coordination process among agents. Adaptive negotiation is a way to achieve the coordination in a dynamic scheduling environment. By adaptive negotiation we mean that more intelligence and rationality are integrated into the negotiation mechanism, thus make it adaptive to the changes of the dynamic scheduling environment. To achieve this, issues at three levels have to be addressed: system architecture, agent architecture, and heuristics. At the system architecture level, the system must have the architecture with corresponding characteristics to support adaptive negotiation among agents. At the agent architecture level, an agent must have rational abilities (decision making mechanisms) embedded in the architecture to transfer the knowledge (in our case, negotiation heuristics) and environment conditions into specific negotiation behaviors. The third level is the heuristics level. By this we mean the knowledge that needs to be integrated into the agent negotiation mechanism and can be used by agents in terms of rational decision making.