Adaptive AI Engine for RTS Games

Discussing the theory and practice

Our Project’s Research Current Situation – 25 October 2009

Posted by merothehero on October 25, 2009

We have come to the most crucial part of our project’s life cycle. It’s This CREATIVE part concerned with creating new techniques or developing existing ones using one or more approaches.

It’s now clear to us that offline Learning (Learning in the company developing the game before releasing the game) is done through Evolutionary Algorithms, also through Cased-based Planning (based on case-based reasoning) through learning from simulation. At the moment we do not wish to enter this field, our main goal is Real-Time Learning.

On the other hand, Online Learning is done through Reinforcement Learning, as well as Case-Based planning as well. Most of the online learning papers make the learning just after the game is finished, but little theses of these are doing Real-Time Learning (learning while playing).

It seems that the most approaches used are Reinforcement Learning and Case-Based Reasoning.  Genetic Algorithms are extremely offline learning.

Since more than a week, we divided ourselves to two groups, me and Saqr studying Reinforcement Learning and trying to implement it, Abdurrahman and A. Atta studying Case-Based Planning and trying to implement it.

I and Saqr are currently:

  • Mastering RL through the ultimate Reinforcement Learning book “RL : An Intro”
  • Understanding Eric Kok’s Paper “Adaptive Reinforcement Learning agents in RTS Games – 2008” as well as its implementation
  • Understanding the paper “Transfer Learning in RTS Games using Hybrid CBR/RL”
  • Implementing a simulation for a Mini-RTS Game to be able to test our Techniques on it.

While Abdurrahman and A. Atta are currently:

  • Understanding more Case-Based Planning.
  • Understanding more CBP papers and their implementation.
  • Starting to think about a design for a platform with CBP to be used with Learning in BosWars (Our Open Source Game)

We have till the 11th of November to complete at least 80% of these Tasks. So may ALLAH be with us!


3 Responses to “Our Project’s Research Current Situation – 25 October 2009”

  1. ferasferas said

    السلام عليكم و رحمه الله و بركاته

    i talked with a. Zakria about the project and when he told me about learning while playing i am still confused cant see a real example
    or at least my mind cant imagine a real example for that

    also i dont think Computational Powers are enough for Learning while Playing

    any way,
    إن شاء الله ربنا معاكم و يوفقكم

    and still my motto is “iam monitoring ur WORK”

    😉 have Fun

  2. ferasferas said

    thx for the quick replay 😉

    its an honor for me to see a great project developing and Learn from it.

    yes its a very good example for real time learning :).

    ربنا معاكم إن شاء الله

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: