Adaptive AI Engine for RTS Games

Discussing the theory and practice

Archive for the ‘CaS-RTS Platform’ Category

Contains our own design to the CBP system for RTS Games

CaS System – An Introduction

Posted by Ogail on December 14, 2009

Hello all,

I’d like to introduce CaS (Case-Based Strategies) its a CBR system that is able to adapt itself depending of the opponent strategies. We consider CaS as a future work for CaT (CBR system developed bt David W. Aha). CaS is a mixture of CBR techniques and is mapped to BDI model. Till now we’ve finished the design of:

  • Case representation technique.
  • Retrieval method.
Advertisements

Posted in CaS-RTS Platform | Leave a Comment »

Case Based Planner Platform For RTS Games

Posted by Ogail on December 6, 2009

Posted in CaS-RTS Platform, Presentations | Tagged: , , , , , , , | Leave a Comment »

Cases Representation Techniques

Posted by Ogail on December 2, 2009

  1. Hash Map.

The key (identifier) of each element will be determined using a hash function depends on the identifier features of the case.

  1. A Simplex Plan.

The problem of acquiring amount of peons to train will be formalized in a Simplex method form and then solved.

  1. Hierarchical Memory Structure.

The most important features are placed in the top of this hierarchy and the less important ones led us to certain sub groups of cases till we reach the most similar cases to the new one we have.

Posted in CaS-RTS Platform | Leave a Comment »

Case Retrieval Technique

Posted by Ogail on December 1, 2009

  • Abstract:
    • Any case-based planning system deals with 5 problems: Case representation, retrieval, reuses, revise and retain. In this article we discuss the case retrieval of the CBP system.
  • In this article we are only restricted with example of collecting a resource for a certain amount.
  • We have three alternative techniques of retrieving cases:
    • Simplex algorithm:
      • The whole cases are just represented in one case.
      • Using Simplex method algorithm you can calculate number of peons required to collect the resource.
    • Hashing:
      • The cases will be retained in a map data structure and its index will be calculated from equation:
      • I = aX + bY + cZ
      • Where:
        • I: is the key of case in the map.
        • X: is number of gold mines in the map.
        • Y: is the number of free or useless peon.
        • Z: amount of gold required.
        • a, b, c are weights for the features.
    • Hierarchal Structure:
      • The cases will be represented as a hierarchy.
      • The most important features are placed in the top of this hierarchy and the less important ones led us to certain sub groups of cases till we reach the most similar cases to the new one we have.

Posted in CaS-RTS Platform | Leave a Comment »

Case Representation Technique

Posted by Ogail on December 1, 2009

  • Abstract:
    • Any case-based planning system deals with 5 problems: Case representation, retrieval, reuses, revise and retain. In this article we discuss the case representation (the plan contents and the case base design) of the CBP system.
  • We define a case C as a tuple of four objects:
    • C = <Goal(s), WorldState, Plan, Performance>
    • Where:
      • Goal(s): goal(s) that this case satisfies.
      • WorldState: is the state of the world in the current situation.
      • Plan: sequence of actions that will achieve the goal(s).
      • Performance: is a real value in the range [0,1] that reflects the utility of choosing that case for that state, where higher values indicate higher performance.
  • We’ve found that it’s wasting of memory and time if we make all of the cases contain the same WorldState because certain cases just require a little knowledge of the world to work. As an example suppose that we need to represent a case for solving problem of collecting a resource with a certain amount, what would the number of assaults I have will help in this problem? Rather we just need a little knowledge that will help. Below the case representation of the resource collecting goal:
    • Goal:
      • CollectResource (Gold, MaxInfluence).
    • WorldState:
      • Number of gold mines.
      • Distance between gold mines and nearest gold mine collector.
      • Number of free or useless peons.
    • Plan:
      • Train(3, peon).
      • Assign(3, peons, GoldMiner).
      • Build(1, Farm).
    • Performance:
  • Number of gold collected in 1 minute.

Posted in CaS-RTS Platform | Leave a Comment »