Design And Implement Of A Multi-Agent Scheduling System

A study on modeling and implementing a multi-agent scheduling system is carried out considering the core effect of scheduling system in manufacturing enterprises and the difficulties classic scheduling theories meet.The problems exist in traditional scheduling systems are analyzed firstly in this paper, then the logic reasoning model of agent, the architecture of multi-agent system and the interaction between agents are studied deeply. Combined with the theories that have been developed, a Multi-Agent Scheduling Model (MASM) is proposed, in which agents are created for each tasks and resources, and they are called TA and RA respectively. Two supervisor agents are added into the model to control TA and RA. The architecture of this model is hybrid, in which high-level agents can supervise low-level agents to improve the global perspective of the result. At the same time, the autonomous of agent is regarded to increase the robustness and flexibility of the system. The agents in this model have a reaction stnicture or hybrid structure according to practice needs.Agents negotiate with each other base on Contract-Net Protocol. After several times' interactions, the scheduling plan can be pursued. During the negotiation, RA waits a little moment after it receives the first request of the batch, therefore it can schedule according to the priority after all steps of this batch have been put into the queue. TA checks the conflicts between steps only after all the steps of this task have received a proposal from RA. The above mechanisms can greatly decrease the times of interaction among agents and improve system efficiency.The model is designed and described using UML and implemented with Java. RMI and multi-threads technologies are used to implement the distributed communication and parallel processing of the system. Thanks to the multi-agent technology, this system is inherently distributed and intelligent. It also offers adaptability and expansibility against disruptions and changes in manufacturing environment. Moreover, this system can be integrated to the legacy discrete ERP system smoothly. The test of the system shows that it can offer a reasonable scheduling plan, decrease the burden of management people and improve the efficiency of production and resources.