Abstract. Automated negotiation techniques have received considerable attention over the past decade, and much progress has been made in developing negotiation protocols and strategies for use by software agents. However, comparatively little effort has been devoted to understanding the computational complexity of such protocols and strategies. Building on the work of Rosenschein, Zlotkin, and Sandholm, we consider the complexity of negotiation in a particular class of task-oriented domains. Specifically, we consider scenarios in which agents negotiate to achieve a more favourable redistribution of tasks amongst themselves, where the tasks involve visiting nodes in a graph. Focussing on a particular representation of the domain (as a spanning tree), we establish a number of complexity results pertaining to the complexity of negotiation in this scenario, with our main result to the effect that the problem of deciding whether a given deal could be reached by a chain of rational proposal...
Paul E. Dunne, Michael Laurence, Michael Wooldridg