Publication: Implementación de una arquitectura para el desarrollo de comportamientos para StarCraft apoyado en PELEA
Loading...
Identifiers
Publication date
2013
Defense date
2013-10-02
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Los juegos de estrategia en tiempo real suponen un cambio en el concepto de jugabilidad al ofrecer a los participantes la posibilidad de realizar sus movimientos al mismo tiempo. Además, se distinguen de los juegos de estrategia en turnos cómo puede ser el ajedrez, en que el número de unidades que cada jugador puede controlar varían conforme el tiempo avanza tanto en número como en variedad de tipos según el jugador decida progresar en la partida. Estas características componen un entorno de trabajo muy complejo a la hora de desplegar jugadores que utilizan inteligencia artificial para tomar decisiones. Estos jugadores artificiales deben ser capaces de adaptarse a las variaciones en el entorno así como a las estrategias empleadas por los jugadores humanos y el resto de jugadores artificiales. Sin embargo, este aspecto ha sido relevado a favor de otros tales como el modelado 3D y los gráficos de los juegos. Esto ha llevado a que cada vez más jugadores demanden una IA más competitiva que sea capaz de suponer una dificultad cuando ya se han jugado varias partidas.
Una forma de mejorar los comportamientos de los jugadores automáticos, es mediante la utilización de la Inteligencia Artificial. En este documento se presenta el desarrollo de un sistema de control basado en Planificación automática para el juego StarCraft. La planificación es un área de la inteligencia artificial que permite la generación de planes de acciones para resolver un problema dada una representación del entorno. Con el fin de desplegar la Planificación Automática como modelo de razonamiento para StarCraft se ha utilizado la arquitectura PELEA, para la cual se ha implementado un conector con el fin de interactuar con el juego capaz de transformar la información del juego en un estado PDDL necesario para su posterior transformación en XPPDL, el formato que PELEA utiliza internamente además de ser capaz de ejecutar las acciones que PELEA le indique en el juego. Para permitir esta interacción se ha utilizado BWAPI y ChaosLauncher. BWAPI es un API para la interacción con StarCraft y ChaosLauncher es el programa que permite la interacción.
Real time strategy games suppose a change in the concept of gameplay by offering the players the possibility to perform their moves at the same time. In addition, they differ from the turn based strategy games like chess in that the numbers of units each player can control vary as time goes on in quantity and types as the player makes decisions towards the match. These characteristics create a complex framework when deploying players who use Artificial intelligence to make decisions. These artificial players must adapt to changes in the environment and the strategies the human players and the rest of the artificial players. However, this aspect has been replaced for others like 3D modeling and game graphics. This produced a growing number of players asking for more competitive AI able to suppose a challenge after several matches have been played. One option to improve the behavior of automated players is by using Artificial Intelligence. This document presents the development of a control system based on Automated Planning for the StarCraft game. Planning is a field of Artificial Intelligence which allows the generation of action plans to solve a problem given a representation of the environment. With the goal of deploying Automated Planning as the reasoning model for StarCraft the PELEA architecture has been used and an adaptor has been implemented. This adaptor is able to interact with the game and capable of transforming the information of the game into a PDDL state required for its later translation to XPDDL, the format PELEA used internally in addition to being able to execute into the game the actions PELEA orders. To allow this interaction BWAPI and ChaosLauncher have been used. BWAPI is an API for the interaction with StarCraft and ChaosLauncher is the program who allows the interaction.
Real time strategy games suppose a change in the concept of gameplay by offering the players the possibility to perform their moves at the same time. In addition, they differ from the turn based strategy games like chess in that the numbers of units each player can control vary as time goes on in quantity and types as the player makes decisions towards the match. These characteristics create a complex framework when deploying players who use Artificial intelligence to make decisions. These artificial players must adapt to changes in the environment and the strategies the human players and the rest of the artificial players. However, this aspect has been replaced for others like 3D modeling and game graphics. This produced a growing number of players asking for more competitive AI able to suppose a challenge after several matches have been played. One option to improve the behavior of automated players is by using Artificial Intelligence. This document presents the development of a control system based on Automated Planning for the StarCraft game. Planning is a field of Artificial Intelligence which allows the generation of action plans to solve a problem given a representation of the environment. With the goal of deploying Automated Planning as the reasoning model for StarCraft the PELEA architecture has been used and an adaptor has been implemented. This adaptor is able to interact with the game and capable of transforming the information of the game into a PDDL state required for its later translation to XPDDL, the format PELEA used internally in addition to being able to execute into the game the actions PELEA orders. To allow this interaction BWAPI and ChaosLauncher have been used. BWAPI is an API for the interaction with StarCraft and ChaosLauncher is the program who allows the interaction.
Description
Keywords
Juegos de ordenador, Videojuegos, Inteligencia artificial, Desarrollo de software