Quantum game-theoretic reinforcement learning for autonomous vehicle navigation and communication in vehicular networks

Abstract Autonomous vehicle (AV) navigation and communication in highly dynamic vehicular networks require fast decision-making, reliable coordination, and energy-efficient connectivity. However, traditional Reinforcement Learning (RL) and game-theoretic approaches often suffer from slow convergence and limited scalability in dense multi-agent environments. To address these challenges, this paper proposes a novel Quantum Game-Theoretic Reinforcement Learning (QGTRL) framework that integrates quantum-inspired policy encoding using Parameterized Quantum Circuits (PQCs) with game-theoretic multi-agent coordination. The proposed approach uses quantum superposition and complex probability amplitudes to enhance exploration efficiency and improve policy convergence stability. In addition, a communication-aware coordination mechanism is incorporated to support robust interaction in dynamic Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) environments. The QGTRL framework is benchmarked against state-of-the-art methods, including Deep Q-Network (DQN), Multi-Agent Deep Deterministic Policy Gradient (MADDPG), Quantum Long Short-Term Memory (QLSTM) Deep Recurrent Q-Network (QLSTM-DRQN), and Quantum Deep Reinforcement Learning (QDRL), under realistic simulation conditions. Experimental results demonstrate that QGTRL improves path-planning accuracy by 18.7%, reduces communication latency by 15.3%, and decreases energy consumption by 22.4% compared to baseline models, while maintaining higher mission success rates in dense and high-mobility scenarios. Furthermore, the framework operates on quantum-inspired classical computation, eliminating the need for specialized quantum hardware and enabling practical deployment. These results highlight QGTRL as a scalable, efficient, and robust solution for next-generation cooperative autonomous vehicle systems.