Planning Coalition Formation under Uncertainty: Auction Approach

Coalition formation approach is very important for applications where an agent can't efficiently realize a task by him-self. Thus agents have to form coalitions in order to execute tasks and to obtain some rewards. Coalition formation problem has widely been studied and many solution approaches were proposed to maximize agents reward and to provide some stability. In this study, we address the problem of coalition formation in environments where agents can not control perfectly the consumption of their resources. In such situations, an agent is uncertain regarding tasks execution. Thus, forming coalitions to maximize the agents' reward is an unfeasible operation. Our idea is that before starting negotiations, the agent must select subtasks to execute taking into account the uncertain task execution and the results of each selection on the future possible execution. We will formalize the subtask selection using a Markov Decision Process. Next, agents can use any auction protocol to coordinate their decisions and form coalition.