I'm interested in how to get multiple agents to work together as a team in an intelligent way. Hopefully in a way that makes them much more efficient than one "big" agent working alone. I'd also like to know if there're way's to do this without tons of predeterined rules. I searched for a while with google but didn't find anything good. Can anyone reccomend links? Just to make myself clear, I'm not looking for something like steering behaviours - I'm looking for a way to use multiple agents to solve problems. Thanks.

Basically, you have two ways of doing this: unilateral or bilateral.

Unilateral is usually synonymous with hierachical and authoritarian. An example would be a leader setting tasks for all of the other agents. The leader doesnt mean to be an agent itself, it can be an abstract entity... A good way to implement this would be some form of conccurent planning.

Bilateral means that instead of looking for a centralized directions, agents communicate their intentions and make requests with each others. For example, an agent might want pass a door, which requires a switch situated in another room to be open. The agent stays at the door and make a request for another agent to hold the switch. You'll need some good form of either task priority or planning to avoid conflicts. A popular way to implement this is through a team blackboard system.

You can have any mix of both systems, of course.

Thanks. Do you have any ideas for a project I can work on to test these concepts?

Ideas:

- Capture the flag teamwork
- Searching a large building collaboratively
- Herding many sheep with multiple dogs
- A sport game, "turn-based" like netball
- Robocup soccer, a "real-time" sport sim

Does that help?