Desarrollo de un agente inteligente para apoyar el desarrollo del pensamiento computacional en un videojuego
Cargando...
Archivos
Fecha
2021
Autores
Profesor/a Guía
Facultad/escuela
Idioma
es
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Andrés Bello
Nombre de Curso
Licencia CC
Licencia CC
Resumen
El pensamiento computacional permite adquirir diferentes habilidades de solución de
problemas y racionalización, este set de habilidades es de suma importancia para los
jóvenes en una sociedad digital, por lo que enseñarlo de manera efectiva es primordial.
Es difícil enseñarlo de la forma tradicional, ya que no se han tenido buenos resultados,
por lo que se ha intentado su desarrollo por medio de nuevas formas de enseñar, por
ejemplo, los videojuegos, los cuales son la opción elegida en este trabajo dada su
efectividad, la cual es avalada por diversos estudios que se muestran en este trabajo.
Mediante el videojuego, se puede facilitar el desarrollo inicial del pensamiento
computacional, para ello, es de suma importancia de igual forma ajustar el equilibrio
desafío-habilidad de cara al jugador, ya que éste debe ser capaz de poder afrontar los
desafíos que plantea el videojuego con el fin último de desarrollar sus habilidades de
pensamiento computacional.
Es por todo lo anterior que se plantea el proyecto de desarrollar un agente inteligente de
software para soportar el equilibrio desafío - habilidad del jugador, prestarle apoyo y
asistencia en forma de pistas para poder resolver un desafío, con el fin de que el proceso
de desarrollar el pensamiento computacional sea de mejor forma, más intuitivo y óptimo
para todos los jugadores.
Este agente inteligente debe ser flexible y utilizar tecnologías que se adapten a diferentes
desafíos, es por ello que se opta por utilizar métodos de Inteligencia Artificial, en
particular las técnicas de Aprendizaje por refuerzo.
Para el desarrollo del trabajo se implementó un videojuego 3D en donde los puzles
requerían la aplicación del pensamiento computacional para su solución. Los puzles
consistían en plataformas que se debían desplazar por un entorno 3D usando instrucciones
dadas por una consola. El agente implementado se entrenó en una representación 2D del
mundo del videojuego. Se probaron múltiples variaciones de condiciones de
entrenamiento lográndose un reward máximo de 0.99 en la simulación y un máximo de
0.93 en el entorno del juego.
Computational thinking is a crucial problem-solving and thinking skill that is of utmost importance for young people in a digital society, so teaching it effectively is paramount. It is difficult to teach it in the traditional way, since there have not been good results in all kinds of students, so its development has been tried through new ways of teaching, for example, video games, which are the chosen option in this work given their effectiveness, which is supported by various studies in the literature. Using a video game, the initial development of computational thinking skills can be facilitated, but to be effective it is important to adjust the challenge-skill balance for the player, since he must be able to face the challenges posed by the video game with the ultimate goal of developing their computational thinking skills. It is for all of the above that this project proposes the development of an intelligent software agent to support the challenge-skill balance of the players, through the provision of support and assistance in the form of clues to help them solve game challenges. This help mechanism should ease the development of computational thinking in an intuitive and optimal way for all players. This intelligent agent must be flexible and use technologies that adapt to different challenges, which is why this work proposes the use of Reinforcement Learning techniques. For the development of the work, a 3D video game was implemented where the puzzles required the application of computational thinking for their solution. The puzzles consisted of platforms that had to be moved through a 3D environment using instructions given by a console. The deployed agent was trained in a 2D representation of the video game world. Multiple variations of training conditions were tested, achieving a maximum reward of 0.99 in the simulation and a maximum of 0.93 in the game environment.
Computational thinking is a crucial problem-solving and thinking skill that is of utmost importance for young people in a digital society, so teaching it effectively is paramount. It is difficult to teach it in the traditional way, since there have not been good results in all kinds of students, so its development has been tried through new ways of teaching, for example, video games, which are the chosen option in this work given their effectiveness, which is supported by various studies in the literature. Using a video game, the initial development of computational thinking skills can be facilitated, but to be effective it is important to adjust the challenge-skill balance for the player, since he must be able to face the challenges posed by the video game with the ultimate goal of developing their computational thinking skills. It is for all of the above that this project proposes the development of an intelligent software agent to support the challenge-skill balance of the players, through the provision of support and assistance in the form of clues to help them solve game challenges. This help mechanism should ease the development of computational thinking in an intuitive and optimal way for all players. This intelligent agent must be flexible and use technologies that adapt to different challenges, which is why this work proposes the use of Reinforcement Learning techniques. For the development of the work, a 3D video game was implemented where the puzzles required the application of computational thinking for their solution. The puzzles consisted of platforms that had to be moved through a 3D environment using instructions given by a console. The deployed agent was trained in a 2D representation of the video game world. Multiple variations of training conditions were tested, achieving a maximum reward of 0.99 in the simulation and a maximum of 0.93 in the game environment.
Notas
Tesis (Ingeniero Civil Informático)
Palabras clave
Videojuegos, Pensamiento Computacional