A Multiple-Nash-SA Mediator for Huge Design Utility Spaces in Automated Negotiations

Takayuki Ito

Abstract: Automated negotiation is crucial in multi-agent systems, especially for design processes involving multiple stakeholders collaborating on new products or plans while satisfying complex, interdependent constraints. These design spaces are inherently intractable due to the vast number of issues, options, and constraints, often leading to sparse utility spaces with extensive plateaus that challenge traditional negotiation methods. Previous research has primarily focused on simple utility spaces or introduced mediator protocols like Simulated Annealing SA-based or auction-based mediation to handle nonlinear utilities. However, these approaches are insufficient for the intractability and uncertainty present in real-world design scenarios, such as car design, where issues and constraints are overwhelming. We adopt the Nash bargaining solution to define the best possible outcome, acknowledging the necessity of addressing numerous trade-offs among issues. Recognizing the limitations of single-search methods that often fail in flat solution spaces, we propose a multiple SA-based mediated agent. This mediator agent performs multiple Simulated Annealing searches within the same utility space, improving the exploration process and increasing the likelihood of escaping plateaus to find better solutions. By leveraging multiple searches, this method effectively tackles the inherent challenges of complex design negotiations, offering a more robust solution for automated negotiation in design processes.