Publication: UN ACERCAMIENTO BASADO EN HIPER-HEURÍSTICAS PARA LA GENERACIÓN DE BOTS PARA JUEGOS DE ESTRATEGIA EN TIEMPO REAL
No Thumbnail Available
Date
2018
Authors
BERMEO GARCÍA, LAURA
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
En esta memoria se presenta un algoritmo basado en Hill Climbing para construirbots de estrategias combinadas que se enfrentarán en un juego simplificado de estrategia entiempo real llamado Planet Wars, el cual fue el objeto de interés en el Desafío de InteligenciaArtificial de Google en el año 2010.Dados distintos enemigos y mapas, el objetivo es construir estrategias combinadas de maneraque individualmente no sean efectivas, pero al entrar distintos bots en puntos diferentesde una partida, las estrategias entre sí colaboren para terminar con una victoria, símil a loque es el uso de una hiperheurística.Usando Hill Climbing para armar mejores hiperheurísticas, se encontraron solucionescombinadas efectivas. Desde ahí, se estudió qué diferencia a una estrategia combinada deotra y qué funciona mejor en distintos casos.
In this memoir, a Hill Climbing-based algorithm is presented to build combinedstrategies that will fight in a simplified real-time strategy game named Planet Wars, whichwas the subject of interest during the Google AI Challenge in 2010.Given different enemies and maps, the objective is to build combined strategies so that whenused individually they aren’t effective, but then used at different points in the game theycollaborate with each other in order to get a win, similar to the use of hyperheuristics.Using Hill Climbing to get better hyperheuristics, effective combined solutions were found.Starting from there, the differences between strategies were studied in different cases.
In this memoir, a Hill Climbing-based algorithm is presented to build combinedstrategies that will fight in a simplified real-time strategy game named Planet Wars, whichwas the subject of interest during the Google AI Challenge in 2010.Given different enemies and maps, the objective is to build combined strategies so that whenused individually they aren’t effective, but then used at different points in the game theycollaborate with each other in order to get a win, similar to the use of hyperheuristics.Using Hill Climbing to get better hyperheuristics, effective combined solutions were found.Starting from there, the differences between strategies were studied in different cases.
Description
Catalogado desde la version PDF de la tesis.
Keywords
BOT , ESTRATEGIA COMBINADA , HILL-CLIMBING , HIPERHEURISTICA , JUEGO DE ESTRATEGIA EN TIEMPO REAL