| Popis: |
There has been an increasing interest in automated negotiation and particularly negotiations that involves multiple interdependent issues, which yield complex nonlinear utility spaces. However, none of the proposed models were able to find a high-quality solution within a realistic time. In this paper we presents a dependency-based mediation mechanism for complex negotiations. In the complex negotiation field, there have been works on issue-by-issue negotiation models. When considering a real world negotiation with a mediator, a mediator proposes issues, and players negotiate on those issues.Then if there is no agreement, the other issues are proposed by the mediator so that players can find a possible agreement. The sequence of proposing issues should be based on the dependency of those issues. In this paper, we adopt a dependency-based complex utility model (Hadfi and Ito, Modeling decisions for artificial intelligence (2014) [12]), which allow a modular decomposition of the issues and the constraints by mapping the utility space into an issue-constraint hypergraph. Based on this model, we propose a new mediation model that can efficiently mediate an issue-by-issue negotiation while keeping privacy on the utility values. Our experimental results show our new mediation model can find agreement points that are close to Pareto front in efficient time periods. |
