hello, what are the basis needed for an IA for a race ? i''ve got all the verticies of the terrain and i thing this is not suffisant. maybe the direction of the circuit ? thanx Fratt

quote: Original post by Fratt
hello,

what are the basis needed for an IA for a race ?
i''ve got all the verticies of the terrain and i thing this is not suffisant.

maybe the direction of the circuit ?

thanx

Fratt

Take a gander at R.A.R.S...the Robotic Auto Racing Simulator. That should answer your questions....




Ferretman

ferretman@gameai.com
www.gameai.com
From the High, Cold, Snowy Mountains of Colorado

The more information, the easier it will be for you to write a script and/or program to control the car.

However, it is possible with very little information: i.e. just tell the car when it collides. A track can be learnt that way -- using reinforcement learning for example.

My guess is you''ll need the position of the centre of the road at regular intervals in front of the car. It''s downhill from there.

Alex


Artificial Intelligence Depot - Maybe it''s not all about graphics...

Actually, a fuzzy finite state machine might be appropriate. I recall a paper on applying a fuFSM to controlling a car in a game in Game Programming Gems. Check that out if you have the opportunity.

Timkin

quote: Original post by Timkin
Actually, a fuzzy finite state machine might be appropriate. I recall a paper on applying a fuFSM to controlling a car in a game in Game Programming Gems. Check that out if you have the opportunity.

Timkin

I believe that was me. I wrote the article on Fuzzy State Machines in Game Programming Gems 2
(and the article on Finite State Machines in Game Programming Gems 1 ). I have also developed
Car AI for a trans-am racing game and a drag racing game.

Anyway, for racing game Car AI, I offer 2 approaches. If you have the geometry of the track
available and organized so that finding the track edges and curve radii is fast and convenient,
and you are not pressed for CPU (ie. you don''t have to control bunches of cars) then it is
possible to calculate where on the track the car should be steering towards at any given time
and at any given speed, based on the physics that have been applied to the car.

Another less CPU intensive approach, is to create one or more driving paths on the track,
which are basically a linked series of points that describe where the car should be on the track
at given spots on the track, and have the Car AI follow the paths (switching back and forth from
several paths based on situations the car might be in).

As Steve pointed out, the R.A.R.S. system is a great place to start becoming familar with
racing game Car AI and I would recommend that anyone interested in this, look at it.

Eric