Deep Q-Learning-Driven Power Control for Enhanced Noma User Performance
By: Bach Hung Luu, Sinh Cong Lam, Nam Hoang Nguyen
Potential Business Impact:
Drones boost slow internet for people far away.
Cell-edge users (CEUs) in cellular networks typically suffer from poor channel conditions due to long distances from serving base stations and physical obstructions, resulting in much lower data rates compared to cell-center users (CCUs). This paper proposes an Unmanned Aerial Vehicles (UAV)-assisted cellular network with intelligent power control to address the performance gap between CEUs and CCUs. Unlike conventional approaches that either deploy UAVs for all users or use no UAV assistance, our model uses a distance-based criterion where only users beyond a reference distance receive UAV relay assistance. Each UAV operates as an amplify-and-forward relay, enabling assisted users to receive signals from both the base station and the UAV simultaneously, thereby achieving diversity gain. To optimize transmission power allocation across base stations, we employ a Deep Q-Network (DQN) learning framework that learns power control policies without requiring accurate channel models. Simulation results show that the proposed approach achieves a peak average rate of 2.28 bps/Hz at the optimal reference distance of 400m, which represents a 3.6% improvement compared to networks without UAV assistance and 0.9% improvement compared to networks where all users receive UAV support. The results also reveal that UAV altitude and reference distance are critical factors affecting system performance, with lower altitudes providing better performance.
Similar Papers
Power Control Based on Multi-Agent Deep Q Network for D2D Communication
Networking and Internet Architecture
Makes phones share airwaves without messing up calls.
Deep Graph Reinforcement Learning for UAV-Enabled Multi-User Secure Communications
Signal Processing
Drones learn to send secret messages safely.
Energy Efficient Task Offloading in UAV-Enabled MEC Using a Fully Decentralized Deep Reinforcement Learning Approach
Multiagent Systems
Drones fly smarter by talking to neighbors.