Coordinated work of large collections of agents using COllective INtelligence

It is a challenging and difficult task that designing large collections of agents without centralized control to support coordinated work. Previous methods, such as rule-based systems, artificial life and market-based systems, used to address this problem have proved to be either too brittle or not scalable to large system. To address this problem, a new method based on COllective INtelligence(COIN) which presented by NASA/Ames research center is proposed. Firstly, a hierarchical model is defined which expand the conventional COIN framework to the heterogeneous multi-agent system; then the forms and the constructing process of the global utility constructed by the hierarchical model are discussed. Secondly, based on an air defense scenario where weapons(agents) needed to assign to the targets without centralized control, a concrete world utility is constructed to solve the rapid self-response of agents. Finally, the impacting factors, effect set, which impacting the coordination effect of the system are discussed. Simulation experiment results show our model and algorithm are effective by comparing with the original methods.