Intelligent Games for UAV Systems: A Survey of Game-Theoretic and AI-Enabled Methods

Abstract

Game theory provides a mathematical framework for strategic interactions involving conflict and cooperation, while artificial intelligence offers computational tools for learning, inference, and decision-making. Driven by the increasing complexity of autonomous missions, their convergence has fostered intelligent game approaches that are increasingly explored in unmanned aerial vehicle (UAV) systems. This survey reviews theoretical foundations and recent progress, organizing the literature along two complementary dimensions: (i) AI-enhanced game paradigms, including game-theoretic deep reinforcement learning, GAN-style adversarial learning, and large language model–enabled agents; and (ii) advanced game-theoretic models for complex environments, spanning perfect- and imperfect-information settings, differential games, evolutionary games, and mean field games for large-scale swarms. Representative UAV scenarios such as pursuit–evasion, cooperative control, and resource scheduling are summarized to highlight methodological trends and practical considerations. Finally, we outline open problems and future directions, emphasizing key challenges in real-time equilibrium computation, partial observability and communication constraints, sim-to-real gaps, and multi-objective trade-offs in safety–energy–mission performance.