A stable and efficient scheme for task allocation via agent coalition formation

Task execution in a multi-agent, multi-task environment often requires allocation of agents to different tasks and cooperation among agents. Agents usually have limited resources that cannot be regenerated, and are heterogeneous in capabilities and available resources. Agent coalition benefits the system because agents can complement each other by taking different functions and hence improve the performance of a task. Good task allocation decision in a dynamic and unpredictable environment must consider overall system optimization across tasks, and the sustainability of the agent society for the future tasks and usage of the resources. In this chapter we present an efficient scheme to solve the real time team/coalition formation problem. Our domain of applications is coalition formation of various Unmanned Aerial Vehicles (UAVs) for cooperative sensing and attack. In this scheme each agent bids the maximum affordable cost for each task. Based on the bidding information and the cost curves of the tasks, the agents are split into groups, one for each task, and cost division among the group members for each task is calculated. This cost sharing scheme provably guarantees the stability in cost division within each coalition in terms of the core in game theory, therefore achieves good sustainability of the agent society with balanced resource depletions across agents. Simulation results show that, under most conditions, our scheme greatly increases the total utility of the system compared with the traditional heuristics.