hi, I have read Brian's book "AI Game Engine Programming", though it's a good book but I still have a lot of questions about FuSM, in the previous thread that explained about FuSM still don't solve my problems. I have two different types of state : command and flexible command type is a state that must be done by the agents coz it was decided by the agent's mission. an agent only get one command state at a time the command type state is such as : move, strike, return and the flexible type state is such as : evade, retreat, waiting, pursuit questions : 1. should my FuSM doing many states at the same time just like in the book? coz I only need one state at a time but using the Fuzzy logic to make agent decision 2. can FuSM calculate whether the agent will decide to do one of the flexible states or keep on doing the active command state ? for example : the agent will keep on doing the command state if the requirements fulfilled and the agent will switch doing one of the flexible state if the requirements fulfilled thank you [Edited by - wiwing on October 7, 2009 4:05:12 AM]

Quote: Original post by wiwing
questions :
1. should my FuSM doing many states at the same time just like in the book? coz I only need one state at a time but using the Fuzzy logic to make agent decision

Ok, then that's not really a FuSM, just an FSM with fuzzy logic in the transitions. You can do it that way if you like.

I can't answer your second question - that will come down to the individual state transition conditions you decide, and isn't really a property of the state machine algorithm.

now I get it ... what's FuSM and the difference between Fuzzy Logic and FuSM
coz some websites really make me confused, by its explanation
thanks dude :D

With fuzzy logic, instead of everything being either true or false, everything has a degree of truth between 0 and 1.

A fuzzy state machine extends this to states, so instead of your membership in one state being true and membership of all the other states being false, you are a member of all sets to some degree between 0 and 1.

I wouldn't advise trying a fuzzy state machine unless you really know what you are doing with fuzzy logic. It would be quite easy for the fuzzy states to degenerate to a point where you don't get any useful values out if you're not careful.

Most places/tutorials get fuzzy finite state machines completely confused.

Most often they mean a probabilistic finite state machine, which means you throw a dice randomly and pick a transition based on that. You could also use fuzzy logic to rank the different transition options for each state, and then pick the best transition.

However, technically a fuzzy finie state machine (FFSM) has all the states "active" at once. Then you calculate the next value of each state by going through all transitions and propagating the value of each previous state. So it becomes very much like a neural network, and there are loads of great papers about this.


Though, most likely what you want is a probabilistic finite state machine :-)

maybe my agent will use a FSM but with Fuzzy Logic transition just like kylotan said ... is it the same as Fuzzy Finite State Machine ? but I still confused about the transition itself.
what if my agent only do one state but before that it will evaluate all the active states with some fuzzy logic algorithm ?