Coalition Formation for Resource Coallocation Using BDI Assignment Agents

Abstract

A new distributed agent algorithm for resource co-allocation to different tasks is proposed. The algorithm extends a belief-desire-intention (BDI) assignment algorithm with resource capability reasoning. It enables resource agents to form coalitions via iterative BDI reasoning and negotiation given the limited capabilities of the resources vis-a-vis task requirements, without directly limiting the coalition size. In the worst case analysis, the number of negotiation rounds required by the algorithm is shown to be of a polynomial order in the number of agents. Empirical evidence from simulations shows that the algorithm yields favorable results in terms of the number of effective coalitions formed for different tasks. Fundamental differences between the proposed algorithm and related work are also discussed.