Handling Agents' Incomplete Information in a Coalition Formation Model

Coalition formation is a problem of great interest in AI, allowing groups of autonomous rational agents to form suitable teams. Our work specially focuses on agents which are self-interested and want to negotiate for executing actions in their plans. Depending on its capabilities, an agent may not be able to perform actions alone. Then the agent needs to find partners, interested in the same actions, and agree to put their resources in common, in order to perform these actions all together. We propose in this paper a coalition formation mechanism based on: (1) an action selection algorithm, which allows an agent to select the actions to propose and deal with the incomplete information about other agents in the system and (2) a coalition evaluation algorithm, which allows an agent to select a group of agents to perform with these actions. Our coalition evaluation algorithm is designed for structured-preference context, based on the use of the information gathered in the previous interactions with other agents. It allows the agents to select partners, which are more likely interested in the actions. These algorithms are detailed and exemplified. We have studied the quality of the solution, we have implemented and tested them, and we provide the results of their evaluation.