An application of quantitative techniques to conflict resolution in a multi-agent system

Abstract

The distributed artificial intelligence (DAI) can create multiple interacting systems and deal with the description. It looks alike that a group of experts’ cooperating together to solve a global problem, which is difficult and decomposed. Each expert can be implemented as an individual agent, which is a self-contained, independent, and active object that can communicate with other agents either indirectly or directly. For achieving a solution to a certain global problem, each agent has its own local goal to be fulfilled. As each agent is given control over its own actions, its own local goals, and its own utility functions, it will naturally pursue its own interests. When working on a shared goal each agent would like its priorities to be given the most consideration. Unfortunately, many agents’ priorities (arising from their own personal goals) will not be coherent with their fellow agents’, thus leading to the conflict phenomenon. Thus, before a satisfactory solution to the whole problem can be presented, this situation of conflict must be resolved. In this paper, we will develop a mathematical model of which tries to resolve the conflict in a multi-agent system by providing a possible solution to the problem. We concentrate on the linear programming model to develop its conflict resolution algorithm and implement it on AGENT-0. Finally, the linear programming algorithm was implemented on the Unix workstations within the AGENT-0 environment that is set under the common Lisp environment. Besides, we evaluated the whole system and proposed some suggestions for future work.