Área: Engenharia Informática e de Computadores

Despacho de nomeação de Júri

Título da Tese: Improving Believable Interactions in Real-time Multi-party Interactive Experiences

Local da Prova: https://videoconf-colibri.zoom.us/j/91735143620

Data: 09/07/2024

Hora: 14h00
Abstract: Believable interactions between synthetic characters are an essential factor defining the success of a virtual environment relying on human participants being able to create emotional bonds with such characters. It is important not only for the characters to be believable but also for interactions with and between them too. This work proposes 3motion, a model for synthetic characters' interactions based on anticipation and emotion, in which the traditional atomic action is deconstructed into three stages: anticipation, action, and follow-through. With these divisions, the model allows for precise non-verbal effective communication, which improves the believability of both synthetic characters and their actions. We improve the model by expanding on its emotion expression capabilities while adapting it to Adfectus, a 3D real-time video game use case in which two characters battle one another. While the characters are controlled by the players, their emotions are expressed independently. To validate and understand the emotion expressions implemented, we created the TAI method, which through the use of a triad comparison and multiple data analysis techniques, creates a model of how people perceive expressions. To validate whether 3motion in Adfectus actually improves believability, we perform several evaluations in which we compare our emotional models to a more reactive approach common in video games. The results show a high level of interaction believability regardless of the version, but no significant differences between emotional models. We did however find that anticipatory models (taking advantage of action subdivision) create characters that are perceived as more aware of others and more capable of expressing unique personalities.