Requirements Driven Service Agent Collaboration
Liwei Zheng, Jian Tang, Zhi Gang Ji · 2009
Multiagent Systems 154 requesters submit their service requests, normally attached with payments, which are virtual and needed-to-be-fulfilled service agents with some desired functionalities.When there is a service request, service agents negotiate with the service request for deciding if they can make contribution and get benefits from the contribution.When a service request can be satisfied by a group of service agents, these service agents form a feasible coalition to fulfil the service request.That has been called on-demand collaboration of service agents (Zheng & Jin, 2007a; Zheng & Jin, 2007b).This chapter presents the computing mechanism for on-demand collaboration of service agents.There are several functional parts in this computing mechanism.First of all, for allowing the negotiation among the service requests and the available service agents, function ontology has been designed to provide the sharable terminology and serve as the background knowledge for both the available service agents and the service requests.After the service requests and the available service agents make agreement on their contribution relationship, these service agents form candidate feasible coalitions based on the strategy of automated mechanism design (Sandholm, 2003) for fulfilling the service requests.Normally, several feasible coalitions can be formed for one service request.Then after a negotiation process between the service request and its candidate coalitions, one of them can be chosen out and a model of a multi-agent system consisting of the service agents in the chosen coalition can be derived as the specification of the service request.The strategies adopted in our approach are reasonable and feasible.First of all, Ontology (Sycara & Paolucci, 2003) is often used for providing common knowledge in application domain.We use function ontology for allowing the service requests and the available service agents can understand each others' functionalities.So that they can match the required functions and the provided functions.In terms of the functional matchability between a service request and a set of service agents and the satisfiability of the service request, the set of service agents forms a coalition which can satisfy the service request.Secondly, automated mechanism design is a feasible technique for preventing agents from misreporting information and quitting an established coalition which would make the coalition unstable.That will guarantee those coalitions are feasible and stable.Thirdly, the service request needs to select a suitable coalition based on some criteria, such as the service quality.And on the other hand, the service agents need to decide what they are going to serve based on what they can earn for the contribution.A negotiation framework is designed for them to make the final decision and to build a solution for the service request.The rest of this chapter is organized as follows.Section 2 presents the function ontology.In section 3, the coalition formation of service agents has been given.And the coalition stability has been analyzed, and how to generate coalition constraints based on automated mechanism design has been discussed.Section 4 formulates the solution selection as a negotiation process and gives a negotiation framework to generate a good enough solution.A case study is given in section 5 for illustrating the whole idea for the service agent coalition formation and the negotiation-based decision-making.Section 6 presents some related works and concludes the chapter. Function ontology Ontology concepts and associationsFor enabling the requirements driven agent collaboration, service agents have to understand the service requests.Function Ontology is constructed for this purpose.This ontology gives