Research on Optimization of RIS-Assisted Air-Ground Communication System Based on Reinforcement Learning

Area of Science:

• Wireless Communication
• Network Engineering
• Artificial Intelligence

Background:

• Urban environments pose challenges to base station (BS) communication networks due to building obstructions.
• Unmanned aerial vehicles (UAVs) offer a flexible solution for enhancing aerial communication networks.
• Reconfigurable intelligent surfaces (RIS) can dynamically adjust wireless signal propagation.

Purpose of the Study:

• To propose an air-ground wireless network using UAVs and RIS to overcome urban communication barriers.
• To enhance the performance of UAV-enabled Multiple-Input Single-Output (MISO) networks in challenging urban settings.
• To develop advanced algorithms for optimizing UAV movement, RIS phase shifts, and user power allocation.

Main Methods:

• A UAV is modeled as an intelligent agent capable of 3D movement and channel sensing.
• Zero-forcing (ZF) precoding is employed to mitigate interference from ground users.
• Two deep reinforcement learning (DRL) algorithms, D3QN-WF and DDQN-WF, are proposed for joint design optimization.

Main Results:

• The D3QN-WF algorithm demonstrated a 15.9% increase in sum rate compared to the DDQN-WF baseline.
• A 50.1% greater throughput was achieved with the D3QN-WF algorithm.
• The D3QN-WF algorithm exhibited significantly faster convergence rates in simulations.

Conclusions:

• The proposed air-ground network with UAVs and RIS effectively improves urban emergency communication resilience.
• DRL-based optimization, particularly with D3QN-WF, offers a superior approach to managing complex wireless environments.
• The system shows promise for enhancing mobile network performance in obstructed urban areas.