Informing Coalition Structure Generation in Multi-Agent Systems Through Emotion Modelling

We propose a hybrid coalition formation method for multi-agent systems that combines a rational mechanism and an emotionally-inspired mechanism to reduce the associated computational cost. To initialise coalition formation, the rational mechanism is used and in subsequent iterations, the emotional mechanism (that forms coalitions resulting from emotional reactions to aspects of interactions between agents) is used. The emotions of anger and gratitude are modelled and used as a basis to model trust which is in turn used to restrict the coalition state-space. We offer some discussion as to how this hybrid method offers an improvement over using a method that only considers payoff maximisation and we propose some direction for future work. 1998 ACM Subject Classification I.2.11 Distributed Artificial Intelligence