A dynamic decentralised coalition formation approach for task allocation under tasks priority constraints

In some real systems, individual agents often need to form coalitions in order to achieve an overall mission that would be impossible for a single robot. Due to communication and computation constraints, it is infeasible for agents to interact directly with all other agents to form coalitions. This problem becomes more complex and challenging when tasks have different priority levels of execution. Toward this end, in this paper, a decentralized dynamic coalition formation approach is presented. The proposed mechanism operates in a neighborhood agent network. Based on self-adaptation principles, this technique enables agents to dynamically join new coalitions with a higher priority at any time without degrading the system. We empirically evaluate our method through a comparison between a centralized and a decentralized approaches. Experimental results demonstrate the good performance of our proposed approach in terms of computation time with respect to the state-of-the-art approach.