AMD Based Service Agent Collaboration and Specification

With the emergence of Big Data in Internet, composing existing web services for satisfying new requirements, such as data quality enhancing, effective data choosing, knowledge discovering etc, has gained daily expanding attentions and interests. Many efforts have been pursued for supporting the essential activities in service composition. However, the existing techniques only focus on passive services which are waiting there for being discovered and invoked. We argue that it might be more attractive when Web services become active entities (Service Agent) distributed in Internet which can recognize the newly emergent requirements and compete with others for realize (part of) the requirements. Retreating or refinement of Big data will hardly be accomplished by one or two data handling center, Service Agent collaboration would be a competitive method for the big data handling problem. Mostly more than one service agents have to collaborate to satisfy requirements in current internet environment especially with social networks. That could be called as the requirement driven agent collaboration. Research on such collaboration might be useful for the previous problem. We have given a preliminary model for the requirement driven agent collaboration based on a function ontology and the automated mechanism design in the earlier work. This paper extended the Function Ontology, and enhanced the AMD model. That makes the interactions in MAS generated by agent collaboration can be described. A negotiation frame for the evaluation and choice of collaboration solutions is also given in this paper. It helps the requester evaluate the possible MAS systems, and helps the service agents make decisions to choose a good enough solution by negotiation. According to the dependencies provided in Function Ontology, a specification is given to describe the execution process of the chosen MAS. And also a method is given to translate the specification to BPEL which is more standard, acceptab- e, and easier to understood.