Multi-party negotiation with preferences rather than utilities

Abstract

We propose here a multi-party negotiation game where agents exchange offers over a single issue. Contrary to most of existing works, we capture the agents' rationality with incomplete (weak and partial) preferences rather than with utility functions. We prove that our negotiation process is always successful and reach fair agreements when the agents are cooperative. In order to illustrate how our negotiation game can tackle distributed problem solving, we introduce a toy problem: two or more agents are paratroopers landed in an unknown maze. Since they aim to meet as soon as possible, they negotiate a rendezvous during their exploration. Our experimentations show that the added value of negotiation is a fast resolution since this adaptive heuristic shares the efforts performed by the agents.