Nucleolus-based compensation payments for automated negotiations of complex contracts

Abstract

Automated negotiation mechanisms (ANMs) can be used to semi-automatically negotiate well-structured but complex contracts. Such negotiations among multiple negotiators, who are represented by software agents, can easily reach a deadlock because they block each others proposals. This of course leads to inferior results. Our aim is to improve the performance of ANMs. In this paper, we try to overcome deadlocks during automated single negotiation text by using compensation payments. Compensation payments are calculated by a mediator according to the Nucleolus method from cooperative game theory. It guarantees that a unique payment is calculated in any case. Furthermore, it lies in the Core, if the Core exists. The proposed ANM can be vulnerable against shading the desired compensation payments. However, our computational experiments suggest that the negotiation results are both superior and faster compared to an ANM without compensation payments. The dominance increases with an increasing number of negotiators.