Strategy specification for teamwork in robot soccer
Bodey R. Baker, Mark A Reynolds, Wei Liu · 2006
We consider real-time environments that require robot agents to work together. In these situations the agents can be designed to coodinate explicitly with some form of negotiation or implicitly with precoded agreements. We propose a implicit method of co-ordinating a team of agents responses within the RoboCup Simulated Soccer league. This involves a list of multi-agent plans called plays that a soccer expert can easily define for the system. The aim is to have a system that is flexible yet powerful and executes the play as intended with very little performance overhead. One possible problem with this approach is that agents will have differing world models. We hypothesise that this approach will keep agents co-ordinated a reasonable percentage of the time even with their differing world models. This is supported by our preliminary results although further tests are required.