Publication: Modelado interno de personajes en entornos virtuales usando lógica difusa
Loading...
Official URL
Full text at PDC
Publication Date
2018
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
A lo largo de la historia de la Inteligencia Artificial, se ha llegado a la conclusión de que las emociones son un factor crítico para desarrollar el comportamiento inteligente de un sistema. Simular la reacción humana ante distintos eventos que suceden en el mundo real siempre fue un reto para este campo de investigación. Por ello, muchos modelos psicológicos han contribuido a describir en qué consiste el proceso emocional centrándose en el efecto que pueden causar estados emocionales como el dolor, el miedo o el hambre. A partir de esta base se crearon lo que se conoce como modelos computacionales de la emoción para agentes virtuales, con múltiples aplicaciones y presentes en distintos campos, como por ejemplo el de los videojuegos.
Sin embargo, los modelos computacionales de la emoción más actuales e influyentes son muy teóricos y utilizan procesos de aprendizaje automático para un entorno de trabajo determinado. Comprenden una compleja base teórica pero la labor de implementación no queda demasiado clara ya que los pocos detalles que se ofrecen no van más allá de los que se hicieron para realizar la evaluación del sistema. Además, ningún modelo ha conseguido simular emociones combinando elementos como el estado de ánimo y rasgos de personalidad y utilizando técnicas de lógica difusa (fuzzy logic). El único modelo destacable de lógica difusa aplicada a la simulación de emociones no contempla la personalidad y no termina de desarrollar un complejo estado de ánimo.
En este trabajo se presenta una alternativa más a ellos, un modelo computacional de la emoción para agentes virtuales que combina la simulación de emociones con el estado de ánimo y rasgos de personalidad mediante lógica difusa. Primero, se pretende realizar un diseño teórico explicando los fundamentos del modelado de emociones de agentes virtuales para que cualquier usuario pueda entender cómo funciona el sistema. Segundo, se propone una implementación en Unity, uno de los motores de videojuegos más utilizados actualmente, explicando los métodos y procedimientos que se van a llevar a cabo en cada una los componentes que conforman el modelo dentro del entorno virtual. La implementación se realizará de forma modular, de manera que el usuario pueda decidir si incorpora componentes de estado de ánimo o personalidad a su propio sistema, o por el contrario sólo conserva la parte más básica del modelo.
Por último, se ha comprobado su correcto funcionamiento a través del desarrollo de un videojuego de aventura conversacional en primera persona. Este videojuego servirá como método de evaluación para que usuarios puedan dar su valoración respecto la eficacia con la que actúa el modelo y mejorar en la medida de lo posible su funcionamiento. En los resultados finales y las conclusiones extraídas se verá que el modelo, combinando las emociones con la personalidad y el estado de ánimo, obtiene mayor nivel de coherencia en la simulación de las emociones por parte de los evaluadores con respecto a su forma más básica, sin personalidad ni estado de ánimo.
Throughout the history of Artificial Intelligence, it has been concluded that emotions are a critical factor in developing the intelligent behavior of a system. Simulating the human reaction to different events that happen in the real world was always a challenge for this field of research. For this reason, many psychological models have contributed to describe what the emotional process consists of, focusing on the effect that emotional states such as pain, fear or hunger can cause. From this base we created what are known as computational models of emotion for virtual agents, with multiple applications and present in different fields, such as videogames. However, the most current and influential computer models of emotion are highly theoretical and use machine learning processes for a given work environment. They have a complex theoretical base but that the implementation is not too clear because the few details that are offered do not go beyond those that were made to carry out the evaluation of the system. In addition, no model has managed to simulate emotions by combining elements such as mood and personality traits and using fuzzy logic techniques. The only remarkable model of fuzzy logic applied to the simulation of emotions does not contemplate the personality and does not finish developing a complex mood. In this document we present an alternative to them, a computational model of emotion for virtual agents that combines the simulation of emotions with mood and personality traits through diffuse logic. First, it is intended to perform a theoretical design explaining the fundamentals of modeling emotions of virtual agents so that any user can understand how the system works. Second, an implementation is proposed in Unity, one of the most commonly used video game engines, explaining how the components that make up the model within the virtual environment are connected. The implementation will be done in a modular way, so that the user can decide if he incorporates mood or personality components into his own system, or on the contrary, only conserves the most basic part of the model. Finally, its correct performance will be verified through the development of a first-person conversational adventure videogame. This video game will serve as an evaluation method so that users can give their assessment of the effectiveness of the model and improve the simulation as much as possible. In the final results and the extracted conclusions it will be seen that the model, combining the emotions with the personality and the mood, obtains a greater level of coherence in the simulation of the emotions than its most basic form, without personality or mood.
Throughout the history of Artificial Intelligence, it has been concluded that emotions are a critical factor in developing the intelligent behavior of a system. Simulating the human reaction to different events that happen in the real world was always a challenge for this field of research. For this reason, many psychological models have contributed to describe what the emotional process consists of, focusing on the effect that emotional states such as pain, fear or hunger can cause. From this base we created what are known as computational models of emotion for virtual agents, with multiple applications and present in different fields, such as videogames. However, the most current and influential computer models of emotion are highly theoretical and use machine learning processes for a given work environment. They have a complex theoretical base but that the implementation is not too clear because the few details that are offered do not go beyond those that were made to carry out the evaluation of the system. In addition, no model has managed to simulate emotions by combining elements such as mood and personality traits and using fuzzy logic techniques. The only remarkable model of fuzzy logic applied to the simulation of emotions does not contemplate the personality and does not finish developing a complex mood. In this document we present an alternative to them, a computational model of emotion for virtual agents that combines the simulation of emotions with mood and personality traits through diffuse logic. First, it is intended to perform a theoretical design explaining the fundamentals of modeling emotions of virtual agents so that any user can understand how the system works. Second, an implementation is proposed in Unity, one of the most commonly used video game engines, explaining how the components that make up the model within the virtual environment are connected. The implementation will be done in a modular way, so that the user can decide if he incorporates mood or personality components into his own system, or on the contrary, only conserves the most basic part of the model. Finally, its correct performance will be verified through the development of a first-person conversational adventure videogame. This video game will serve as an evaluation method so that users can give their assessment of the effectiveness of the model and improve the simulation as much as possible. In the final results and the extracted conclusions it will be seen that the model, combining the emotions with the personality and the mood, obtains a greater level of coherence in the simulation of the emotions than its most basic form, without personality or mood.
Description
Facultad de Informática, Departamento de Ingeniería del Software e Inteligencia Artificial, curso 2017-2018