Multi-agent collaboration in competitive scenarios

For many multi-agent scenarios one can assume that the agents behave cooperatively and contribute to a common goal according to their design. However, our work focuses on competitive scenarios which are characterized by the agents' strong local interests, their high degree of autonomy, and the lack of global goals. Therefore, two agents will cooperate if, and only if, both will gain - or at least expect to gain - from that cooperation. This paper presents a conflict resolution mechanism which is appropriate for competitive resource allocation in dynamic environments which is based on compromising. It integrates a goal relaxation mechanism, negotiation histories, and multilateral negotiations.