Artificial Intelligence Depot
Visiting guest. Why not sign in?
News, knowledge and discussion for the AI enthusiast.
FEATURES COMMUNITY KNOWLEDGE SEARCH  
Understanding Artificial Intelligence
Comprised of critically acclaimed essays by the world's leading experts on each topic in the series, these collections will become definitive texts on crucial issues of our technological times.
More information at Amazon US UK
Thesis - QuakeIII bot from scratch
 
Thesis - QuakeIII bot from scratch

Hellow all,

First let us introduce ourselves. We are two students living in Belgium and studying Industrial Engineering in Gent.
As we are in our 3th year of our course, we already have to pick a subject for our thesis.

Since the area of AI has always interested us, We've put in a proposition to make a bot for QuakeIII. We have to admit, we are still newbies and don't know anything yet, but we really want to learn....

We know that this has been done before and maybe we won't even code a bot thats halve as good as any both ever coded, but we still want to try.

We want to start off from scratch (besides the QIII-code of course, and a little help from previous algorithms and agents) and build a bot that actually has to learn everything. If its possible, we would like to do this mainly through Neural Networks and/or Genetic Algorithms...

Our teacher was fond of the proposition but he suggested we'd first checked if all this is possible within the QuakeIII-source code. As we know (or maybe don't know, as far as our knowledge goes) we would have to be able to use an exterior kind of memory to store the information for the NN and create algorithms that don't interfere with most of the existing QIII code itself.

As you can read, there are many "if"-s and "would be able"-s in this text. That's why we would like an opinion from the expert himself. If you've got the time, we would be very pleased to receive some answers...
Especially we would like to know if this could work or not, since we don't plan working on something we will never be able to accomplish...

Thanks...

Davy Berckers and Roel Arys



1 posts.
Thursday 07 March, 14:40
Reply
Bot AI and Games

Welcome to the AI Depot Davy and Roel.

Considering you two are sharing a keyboard, I have to say I'm impressed with your typing ;)

Short answer, Q3 bots are possible, yes.

Longer version.

It'll be a fair bit of work for you to get to terms with the code, and them to insert your own AI on top of that, but definitely possible. In fact, Q3 comes with default bots, so the framework is there to expand upon.

As your supervisor mentioned, it's crucial that you don't get tied down by the game development aspect, as a 3rd year project will go very quickly. Setting yourself strict guidelines will help, and making the AI modular is a must.

If your project starts next year, you may be in luck ;) I'm designing and co-developing an Embedded Agent Framework, which will allow AI code to be based on that only. That framework will provide input and output in a simple abstract fashion, so you don't even have to worry about the implementation. The initial framework will be in Q2, but Q3 and UT are planned.

That will save you a lot of time, and allow you to focus on whatever you want to put in the bot's brain. Though I'm not convinced about NN/GA more than any other approach to bot AI, to be perfectly honest!

Anyway, keep us upto date with your progress!

935 posts.
Saturday 09 March, 11:50
Reply
Another bot Thesis

Hello.
My name is Wojtek Pszczolkowski and I'm a Polish student. I study computer science at Wroclaw University of Technology. I'm on my 4th year now (actually it's nearly the end of 4th year). I've got more or less the same project to do. There are some differences though.
1. I've decided to make Q2 bot.
2. I've decided to mix Finite State Machine, Machine Learning, Fuzzy Logic, Opponent Modelling and maybe some CBR to do the work.

Well I'm very curious of the result of my work.

Maybe there will be some possibilities to exchange experiences some time.

Wojtek Pszczolkowski

1 posts.
Sunday 16 June, 11:47
Reply
Mix & Match

Welcome to the boards Wojtek.

I definitely think blending different technologies is the way to go. You'll realise which is best applied to which part of the bot as you experiment. There's no right solution of course, but some things just won't work ;)

Do let us know of your results and findings... a small tutorial maybe? :P

935 posts.
Thursday 20 June, 07:01
Reply