Forming Coalitions in the Face of Uncertain Rewards

When agents are in an environment where they can interact with each other, groups of agents may agree to work together for the benefit of all the members of the group. Finding these coalitions of agents and determining how the joint reward should be divided among them is a difficult problem. This problem is aggravated when the agents have different estimates of the value that the coalition will obtain. A "two agent auction" mechanism is suggested to complement an existing coalition formation algorithm for solving this problem.