3D Sensing Character Simulation using Game Engine Physics

• Autor(a) principal: Rodrigues, João Filipe Pereira
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/151145
• topic: 3D Sensing Characters; Game Bot; Game Engin; 3D Animation; Simulation Visualization; Agent Simulation; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Creating visual 3D sensing characters that interact with AI peers and the virtual envi- ronment can be a difficult task for those with less experience in using learning algorithms or creating visual environments to execute an agent-based simulation. In this thesis, the use of game engines was studied as a tool to create and execute vi- sual simulations with 3D sensing characters, and train game ready bots. The idea was to make use of the game engine’s available tools to create highly visual simulations without requiring much knowledge in modeling or animation, as well as integrating exterior agent simulation libraries to create sensing characters without needing expertise in learning algorithms. These sensing characters, were be 3D humanoid characters that can perform the basic functions of a game character such as moving, jumping, and interacting, but also have simulated different senses in them. The senses that these characters can have include: touch using collision detection, vision using ray casts, directional sound, smell, and other imaginable senses. These senses are obtained using different game develop- ment techniques available in the game engine and can be used as input for the learning algorithm to help the character learn. This allows the simulation of agents using off-the- shelf algorithms and using the game engine’s motor for the visualizations of these agents. We explored the use of these tools to create visual bots for games, and teach them how to play the game until they reach a level where they can serve as adversaries for real-life players in interactive games. This solution was tested using both reinforcement learning and imitation learning algorithms in an attempt to compare how efficient both learning methods can be when used to teach sensing game bots in different game scenarios. These scenarios varied in both objective and environment complexity as well as the number of bots to access how each solution behaves in different scenarios. In this document is presented a related work on the agent simulation and game engine areas, followed by a more detailed solution and its implementation ending with practical tests and its results.