Jump to content
  • Advertisement
Sign in to follow this  
sidhantdash

Agents that 'learn'

This topic is 4813 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi I was wanting to know if Neural Networks are used in games to build agents that learn from their past experiences, like typical attack techniques/moves of their human opponents, and keep getting smarter with every game. Or are other techniques used? I would want to program such agents, and am currently studying NNs from Lauren Fausett's Fundamentals of Neural Networks. Please tell me if what I am doing is right. The "applications" section in the book, doesnot talk about any application of NNs in computer games/intelligent agents. If any other techniques are used, please post them, and also if you know any good textbooks that are focusses on these techniques. thanks

Share this post


Link to post
Share on other sites
Advertisement
Quote:
Original post by sidhantdash
I was wanting to know if Neural Networks are used in games to build agents that learn from their past experiences, like typical attack techniques/moves of their human opponents, and keep getting smarter with every game.


Black & White utilized perceptrons to model the creature desires (which were part of a belief-desire-intention model, BDI for short). The weights of the perceptrons were adjusted based on the pain-pleasure principle.

ANNs have been used sparsely in computer games. Where they have been used, they have been trained off-line and the trained ANN is used in the published game. Like Colin McCrae's Rally (follow the link at my website for more detail of this)

Allowing agents to learn in-game is very risky since it becomes difficult to test properly and tweak gameplay.

Share this post


Link to post
Share on other sites
Project NERO used a real time version of neuro evolution and neural networks to generate units that learn from their experiences. I'd suggest checking out that free download and the papers behind it if you are interested.

Share this post


Link to post
Share on other sites
Quote:
Original post by Jbs
Project NERO used a real time version of neuro evolution and neural networks to generate units that learn from their experiences. I'd suggest checking out that free download and the papers behind it if you are interested.


Ah yes. It was extremely remiss of me to forget NERO (sorry Ken). Go check it out; it's an exciting development.

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!