INTELIGÊNCIA ARTIFICIAL PARA AMBIENTES DE JOGO: UM FRAMEWORK PARA AGENTES AUTÔNOMOS DE TOMADA DE DECISÃO

• DOI: https://doi.org/10.22533/at.ed.394122608016

• Palavras-chave: Inteligência artificial em jogos; Aprendizado por reforço profundo; Agentes autônomos; Sistemas multiagente; pensamento computacional.

• Keywords: Artificial Intelligence in Games; Deep Reinforcement Learning; Autonomous Agents; Multi-Agent Systems; Computational Thinking

• Abstract: The integration of artificial intelligence (AI) into digital game environments has evolved from deterministic, script-based behaviors to autonomous systems capable of adaptive decision-making. This advancement has been driven by machine learning techniques, particularly deep reinforcement learning and multi-agent systems. This article proposes a conceptual framework for intelligent agents in games, combining reinforcement learning, hierarchical planning mechanisms, and abstract state representation. Initially, a review of the evolution of AI in games is presented, ranging from classical approaches to contemporary techniques based on deep learning. Subsequently, research gaps related to generalization, computational cost, and integration with game engines are identified. As a contribution, a modular architecture composed of perception, decision-making, planning, and execution layers is proposed. The framework aims to provide a foundation for the development of adaptive agents that can be integrated into modern game engines, such as Unity and Unreal Engine. Finally, experimental evaluation methodologies and implications for game design and player experience are discussed. As an analytical complement, a predator–prey scenario is considered, in which agents interact in a dynamic environment, allowing the observation of how internal variables influence adaptive behaviour and decision-making. In this context, the study also establishes a connection with computational thinking, highlighting processes such as decomposition, abstraction, pattern recognition, and algorithm design.