Research Of Resource Service Aggregation Model Based On Multi-agent System

With the development and extensive application of computer and network technology, the type and quantity of resources dispersed in the network become much great. Through dynamic aggregation of resources, users can reorganize these distributed and heterogeneous resources, then these resources can provide specific services to meet users’ needs. However, hugeness in quantity, heterogeneousness in distribution and other factors have brought problems for the aggregation of resource service. The problem of resource service aggregation is studied in this paper, and a method of using multi-agent technology to achieve the resource service aggregation has been analyzed and discussed during the resource service aggregation.A resource service aggregation model which is based on the multi-agent system is presented to solve the problem of distributed resource aggregation. The hierarchical structure of the model, the division of functions of each level, the relationships between the hierarchies and the structure of various agents in this model are analyzed. In allusion to the situation that the aggregation tasks submitted by users are usually difficult for the agents in this model to solve independently, a task decomposition method is presented. Then the decomposed tasks can be performed by these agents in the model, and complex tasks can be completed by multiple agents by means of the collaboration between them.According to the features of the resource service aggregation model, the FIPA contract net protocol has been extended to support the collaboration and task allocation between the agents in the model. In the given extended contract net protocol, a tender strategy which based the public information board is used at the tender stage, the communication traffic is reduced observably; then an autonomous bidding strategy is adopted at the bidding stage, agents decide whether to tender on the basis of their real-time status; and case-based reasoning thought is imported at the winning stage, combining with the experience of solving problems within the historical period and the coefficients such as trust degree, activity degree and degree of availability which reflect the real-time capability of agents to make decision on the current problems. At the same time, the treatment mechanism of dealing with the failure of task execution is improved. Thus,the collaboration efficiency has been promoted.Finally, a battlefield target processing system based on the Java Agent Development Environment(JADE) is presented as a instance to validate the resource service aggregation model and the extended contract net protocol. Experimental results show that: the model and the extended contract net protocol have capacity of completing dynamic resource service aggregation and the ability to distinguish similar tasks. They can also reappoint the winning Agent to avoid re-initiate tasks, and spend less communication traffic and time than FIPA contract net.