The goal of configuring a massive, complex multi-agent system can be viewed as a distributed search problem in which each agent attempts to choose a correct configuration. This research presents a technique that can simultaneously Junction as the problem decomposition and solution aggregation components in such a distributed search environment. The method is tested in a number of large multi-agent simulations to demonstrate its feasibility. The agents in the system are shown to achieve optimal or near optimal configurations in significantly less time than agents configuring themselves individually.
展开▼