Article View: rec.games.programmer
Article #755Re: Computer-Controlled Players - simulated evolution
From: Ross E Milward
Date: Mon, 17 Aug 1992 06:30
Date: Mon, 17 Aug 1992 06:30
52 lines
2846 bytes
2846 bytes
In article <psm.714022271@murlibobo> psm@mullian.ee.mu.OZ.AU (Phil Malin) writes: >Hi everyone...I just had a couple of thoughts about this concept. > >What if we wanted the computer to construct the machines/monsters from the >ground up. We could specify rules describing (I'll use machines to illustrate) >armour, weapons, means of movement, etc... and the relationships between them. >For example strong armour would mean slow movement, and so on. For different >terrains the machine would have an optimal configuration. > I was thinking along similar lines. Given the list of attributes, the program can build random CCPs. By using an evaluation function which would determine the relationship between the attributes it would be possible to evolve an optimal configuration using the GA. You could either have it doing this in the initial game setup, or else have the machines evolve into optimal configuration as the game progresses. >What all this means is that for a given terrain we could define a fitness >function, and from this the genetic algorithm could construct the optimal >machine configuration. > Yep, it would be quite easy actually. >I suppose we could also have the GA design the strategy for this machine in >the specified terrain but this would probably get VERY JUICY. > Hmmm, I'm not sure how easy it would be to create an evaluation function for this, although with a defined environment it should be ok. Another good addition for choosing strategies could be the use of neural networks. This would give the machines a learning ability. For example based on a given set of inputs determined by their current location in the environment, it would be then possible for them to choose the best action based on previous experience. >Anyone have any thoughts on this? Now we've got this discussion rolling, is anyone actually going to begin coding a game with intelligent/evolvable CCPs. It would be great to see a game that had a large variety of CCPs that can evolve into stronger opposition as the game develops. > >Phil >psm@mullian.ee.mu.oz.au Ross. -- Ross Milward, ross@coral.cs.jcu.edu.au, cprem@marlin.jcu.edu.au Honours Student - Comp. Sci. Dept., James Cook Uni, Qld, 4811, Australia. "The way the rain comes down hard, that's how I feel inside." - The Cure
Technical Details
Message-ID:
<1992Aug17.063056.29938@marlin.jcu.edu.au>Path:
arc1-ca.pugleaf.net!sparky!uunet!munnari.oz.au!bunyip.cc.uq.oz.au!marlin.jcu.edu.au!cpremReferences:
<Bt3G69.5Cp@dptspd.sat.datapoint.com> <1992Aug17.003402.20999@marlin.jcu.edu.au> <psm.714022271@murlibobo>Show Full Headers
Complete Article Headers
Newsgroups: rec.games.programmer Path: arc1-ca.pugleaf.net!sparky!uunet!munnari.oz.au!bunyip.cc.uq.oz.au!marlin.jcu.edu.au!cprem From: cprem@marlin.jcu.edu.au (Ross E Milward) Subject: Re: Computer-Controlled Players - simulated evolution Message-ID: <1992Aug17.063056.29938@marlin.jcu.edu.au> Organization: James Cook University References: <Bt3G69.5Cp@dptspd.sat.datapoint.com> <1992Aug17.003402.20999@marlin.jcu.edu.au> <psm.714022271@murlibobo> Date: Mon, 17 Aug 92 06:30:56 GMT Lines: 52