Techniques and Directions for Building Very Large Agent Teams
Any type of content formally published in an academic journal, usually following a peer-review process.
We have developed probabilistic algorithms that leverage the associates network for distributed plan instantiation, role allocation, information sharing and adjustable autonomy with a team. By developing such new algorithms, we have been able to build teams of hundreds of cooperating agents, and test specific behaviors among tens of thousands of agents. In this paper, we describe the algorithms that we have developed, the tests that we subjected them to, and sketch some of the key challenges that remain to be addressed.