'An AI Framework for Real-Time Strategy Games'
Authors
Frederiksen, Kasper ; Kristensen, Kasper G. ; Lauritsen, Anders
Term
4. term
Education
Publication year
2006
Abstract
Mange spillere lægger i stigende grad vægt på selve spiloplevelsen. Det har øget interessen for multiplayer, fordi menneskelige modstandere giver en udfordring, som computerstyrede modstandere i singleplayer ofte ikke kan matche. Singleplayer-AI'er er typisk lette at slå, fordi de mangler dynamisk tilpasning under spillet. Dette projekt udvikler en AI-rammeværksarkitektur til Real-Time Strategy (RTS)-spil – strategispil i realtid, hvor man løbende styrer enheder og ressourcer – for at skabe modstandere med mere menneskelignende evner og adfærd. Rammeværket bygger på en model af, hvordan spillere faktisk spiller RTS-spil. Vi tester det ved at koble det til et game development framework, altså en softwareplatform til at udvikle spil. Vi viser, at rammeværket kan konfigureres af både programmører og ikke-programmører, og at det kan bruges til at levere komplette AI-løsninger i almindelige RTS-spil. Foreløbige resultater tyder på, at ideerne bag designet og en ny datastruktur til håndtering af strategier er lovende.
Players are increasingly focused on the quality of gameplay. This has boosted interest in multiplayer, because human opponents pose challenges that computer-controlled opponents in single-player often cannot. Single-player AIs are typically easy to beat because they lack dynamic adaptability during play. This thesis develops an AI framework architecture for Real-Time Strategy (RTS) games (strategy games played in real time where you manage units and resources) to build opponents with more human-like capabilities and behavior. The framework is based on a model of how players actually play RTS games. We test it by connecting it to a game development framework, a software environment used to build games. We show that the framework can be configured by both programmers and non-programmers and can be used to provide complete AI solutions within mainstream RTS games. Preliminary results indicate that the framework’s design and a new data structure introduced to handle strategies are promising.
[This abstract was generated with the help of AI]
Documents