Back to search

University of Nevada - Reno

Coevolutionary Approaches to Generating Robust Build-Orders for Real-Time Strategy Games

Abstract

dc:description.abstract

We aim to find winning build-orders for Real-Time Strategy games. Real-Time Strategy games provide a variety of challenges, from short-term control to longer term planning. We focus on a longer-term planning problem; which units to build and in what order to produce the units so a player successfully defeats the opponent. Plans which address unit construction scheduling problems in Real-Time Strategy games are called build-orders. A robust build-order defeats many opponents, while a strong build-order defeats opponents quickly. However, no single build-order defeats all other build-orders, and build-orders that defeat many opponents may still lose against a specific opponent. Other researchers have only investigated generating build-orders that defeat a specific opponent, rather than finding robust, strong build-orders. Additionally, previous research has not applied coevolutionary algorithms towards generating build-orders. In contrast, our research has three main contributions towards finding robust, strong build-orders. First, we apply a coevolutionary algorithm towards finding robust build-orders. Compared to exhaustive search, a genetic algorithm finds the strongest build-orders while a coevolutionary algorithm finds more robust build-orders. Second, we show that case-injection enables coevolution to learn from specific opponents while maintaining robustness. Build-orders produced with coevolution and case-injection learn to defeat or play like the injected build-orders. Third, we show that coevolved build-orders benefit from a representation which includes branches and loops. Coevolution will utilize multiple branches and loops to create build-orders that are stronger than build-orders without loops and branches. We believe this work provides evidence that coevolutionary algorithms may be a viable approach to creating robust, strong build-orders for Real-Time Strategy games.

Degree

thesis:*
Level thesis:degree_level
Doctorate Degree
Department thesis:degree_department
Computer Science and Engineering
Year dc:date.issued
2014

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Ballinger, Christopher A.
Advisor dc:contributor.advisor
  • Louis, Sushil J.
Committee members dc:contributor.committeemember
  • Dascalu, Sergiu
  • Nicolescu, Monica
  • Naik, Swatee
  • Dailey, Larry

Subjects

dc:subject × 5

Rights

dc:rights
Statement dc:rights
  • In Copyright(All Rights Reserved)
  • Author(s)

Identifiers

dc:identifier.*
Handle dc:identifier.uri
http://hdl.handle.net/11714/2397

Chain of custody

source
Harvested from
University of Nevada - Reno
Base URL
scholarwolf.unr.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Ballinger, Christopher A.. Coevolutionary Approaches to Generating Robust Build-Orders for Real-Time Strategy Games. Doctorate Degree thesis, 2014. http://hdl.handle.net/11714/2397