Joint Optimization of Deployment and Caching for HAPS Assisted Multi-UAV Communications

The Air Access Network (AAN) provides rapid emergency communication, leveraging its flexible deployment and extensive coverage. Specifically, the High-Altitude Platform Station (HAPS) aids multi-Unmanned Aerial Vehicle (UAV) communications, delivering superior service by reducing transmission distance and latency. This is achieved through real-time UAV deployment and content caching. The optimization of this system, modeled as a continuous-discrete Markov Decision Process (MDP), is addressed using the Multi-Agent Hybrid Double-Layer Soft Actor Critic (MAHDSAC) algorithm. This algorithm integrates continuous and discrete SAC to effectively manage mixed UAV deployment and cache control. Simulation results indicate that MAHDSAC outperforms traditional Multi-Agent Deep Reinforcement Learning (MADRL) methods in performance.