A Reinforcement Learning Routing Protocol for UAV Aided Public Safety Networks

Area of Science:

• Computer Science
• Electrical Engineering
• Network Engineering

Background:

• Public Safety Networks (PSNs) require reliable, long-term connectivity for first responders.
• Limited device energy and transmit power necessitate efficient network operation.
• Multipath routing and cooperation among devices are crucial for sustained connectivity.

Purpose of the Study:

• To enhance Energy Efficiency (EE) and extend the network lifetime of PSNs.
• To develop a novel routing scheme utilizing Reinforcement Learning (RL) and Unmanned Aerial Vehicles (UAVs).
• To analyze the impact of clustering, RL-based routing, and UAV deployment on network performance.

Main Methods:

• Network configurations generated using various clustering schemes.
• Reinforcement Learning (RL) applied to optimize routing topology based on energy and distance costs.
• Performance evaluation considering throughput, energy consumption, packet delivery ratio, and network lifetime.
• Analysis of UAV trajectory and number of UAVs on network Energy Efficiency (EE).

Main Results:

• Clustering schemes improved Energy Efficiency (EE) by approximately 42% compared to non-clustering approaches.
• A two-UAV configuration (TUOAO) further boosted EE by 27% over a single UAV.
• Comparable control packet numbers were observed between single and dual UAV scenarios, but cluster head changes differed significantly.

Conclusions:

• The proposed RL and UAV-aided multipath routing scheme effectively enhances PSN Energy Efficiency (EE) and network lifetime.
• Clustering and optimized UAV deployment are vital strategies for improving PSN performance in critical situations.
• The findings support the use of intelligent routing and aerial platforms for resilient public safety communications.