Collabration in Soccer and similar games

[Guest]

Archived from groups: comp.ai.games (More info?)

Hello,
I learned that many soccer games use state machine to control the
behavior of players, such as pass or shoot. But in a higher
perspective, how the players of a team collaborate together, what kind
of model they are using?

 

[Guest]

Archived from groups: comp.ai.games (More info?)

We classify the state machines u mentioned as the low levels actions
and the collaboration as the strategy level. At the strategy level,
cooperation can be done through several models. Simple models like
petri-nets are like state-machines but at a general higher level
perspective. There are also the family of reinforcement learning that
randomly select an action and updates its performace value over a
period of time so that the team gets "smarter and smarter" over the
period.

If u are going for competitions.. Remember the simpler the better. A
fixed strategy or a simple meta-strategy works best where it is more
generic, simple, and easy to debug. However, if u are going into
research, the adaptive researches such as GA, neural networks,
Reinforcement learning etc. are good areas to go into.

REgarding how the team collaborate together? u can check out on
websites on the robocup compeititon or fira website. In fact, for the
low level behaviour u can also dun use state machines. Fuzzy methods
seems to work better.