A formal model for strategic planning in cooperative and competitive environments case study: design and implementation of a basketball simulator

• Ano de defesa: 2017
• Autor(a) principal: Otranto, Guilherme Fernandes
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Artificial intelligence; Behavior simulation; Inteligência artificial; Planejamento; Planning; Simulação de comportamento
• Link de acesso: http://www.teses.usp.br/teses/disponiveis/45/45134/tde-13112017-160053/

The motivation that originated this work was the desire to create an invasion team sports simulator capable of applying user defined strategies to guide the behavior of the agents in the simulation. With this objective in mind we created a formal strategy model to describe complex team behavior and developed methods of using that model to calculate collective plans. We defined both the strategy model and the planning methods in a broad manner that can be applied in many different domains. Then we defined a basketball simulation domain and implemented our methodology to develop a simulator. We also present a control system architecture that is compatible with our proposed planner and show how we implemented it to create the basketball simulator. The formal strategy model we developed can be used to represent team behavior, analyze real world events and create simulations. We developed a strategy design tool that allows the end user to create and visualize team strategies for basketball. Finally, we developed a system that interprets the user generated strategies and creates a basketball match simulation of the described behavior. We also proposed a methodology for the development of simulation systems involving multiple intelligent agents. Our recommended control system architecture separates the many layers of control, which simplifies the development process and results in a naturally expansible system. In this thesis we have provided a novel approach to collective behavior simulation utilizing user input as a guide to the strategy planning. Both the theory and methods developed have been tested through the implementation of a basketball simulator and the results were satisfactory. We believe this is a seminal work that will lead to many interesting developments, both in the realm of sports and in broader domains.