Hi. In here (animated movie) http://www.laseeb.org/vramos/SRSmovie.gif
these agents move on 3D grids while self-regulating their population,
but the method
can be easily applied to 2D grids.
Details can be obtained here (PDF available):
http://www.laseeb.org/vramos/ref61.html
+
http://www.laseeb.org/vramos/ref64.html
Best, v.
On Apr 2, 9:01 pm, "TheCornjerker" <add...@[EMAIL PROTECTED]
> wrote:
> Posted in COMP.AI.GAMES as well.
>
> Hello, I'm sure this has been vetted here before but I'd like some
> layman's advice on where to begin my quest into genetic algorithms.
> At least I think that is what I want to do. I'm new to this area, so
> please pardon any misstatement or misunderstanding of the topic.
>
> I'm an avid wargammer and a java, vb, web developer as well as a
> database admin. I have an interest in genetic algorithms. I played
> a
> game called Cloak and Dagger DNA (Goggle it with "CDDNA") a few years
> ago and it has stuck in my head ever since. I want to put this demon
> to bed by seeing if I can develop something similar.
>
> I want to develop a grid based (squares) game that uses GA for
> movement of pieces. I'm starting simple with only one type of piece,
> say tanks. I've read and implemented various path finding, line of
> sight, field of view, influence mapping, etc routines. But I've had
> this GA movement thing in my head. I'd like some help on
> representing
> the GA either in XML or a database or something.
>
> CDDNA seems to us an instruction set for its piece movement. The
> author explains a little in the help file as to what is happening,
> but
> it is hard to understand how the movement is based on the instruction
> set. The Evolution Lap crosses these instructions during the
> evolution process and runs a tournament and determines a winner. It
> costs food to enter the tournament and to be a parent. You win food
> by winning the tournament. I've run CDDNA's evolution lab for over
> 250,000 tournaments and the game does get a little better, although
> I'm still able to defeat it.
>
> I want to start simple. My game will have a 10x10 grid of squares.
> There are four players. Each player can create tanks at their corner
> of the grid. You gain dollars by owning squares in the grid. You
> start by owning one square and one tank. As your tank moves to an
> adjacent square in round 2, you own two squares and you get $2 per
> round. Tanks cost $5.
>
> First I just want my tanks to learn they need to capture squares to
> earn dollars to buy more tanks. Next I want to be able to introduce
> things like other factories on the grid that if they own them, they
> can build tanks there as well.
>
> I'm not for sure where to start. Each tank can only move North,
> South, East, or West. How do I represent my tank movements as a set
> of instructions that can evolve? How would they evolve? Maybe I
> need
> my GA to create a decision tree for each tank based on its
> surroundings. If so, how to I represent this decision tree?
>
> Eventually I'd like to be able to introduce things like battles for
> squares. If I can introduce a parameter or condition that if you
> attack a tank from the left or right side you get a 1.5x bonus in
> damage or from the back a 2x bonus. This could hopefully teach the
> tanks to perform flanking maneuvers. Another possibility is giving a
> bonus for moving multiple tanks into one square before attacking.
> This could evolve into flocking maneuvers of the tanks.
>
> Any help, directions, or examples would be greatly appreciated.
>
> Thanks.
|
