Sum Secrecy Rate Optimization in UAV Communications Using Quantum Actor-Critic Reinforcement Learning

Reinforcement learning (RL) has proven effective in wireless tasks like dynamic spectrum access and power control. Extending this, Quantum Reinforcement Learning (QRL) addresses quantum-specific challenges such as channel estimation and multi-agent UAV communications. We propose a hybrid Quantum Deep RL (QDRL) framework to optimize the average sum secrecy rate (SSR) in a mmWave UAV system with a Reconfigurable Intelligent Surface (RIS), under imperfect CSI and multiple eavesdroppers. Based on the DDPG framework, we explore quantum variants like QTDDPG and QTTD3. Simulations show QDRL achieves comparable SSR to classical methods with far fewer parameters. Notably, even a simple QTDDPG with a single-layer quantum encoder performs well, underscoring the promise of efficient quantum policies for secure wireless communication.