Unmanned aerial vehicles (UAVs) were recently proposed as on-demand mobile wireless infrastructure. While most research focused on single UAVs or multiple independent UAVs, this project explored the benefits of coordinated UAV swarms within Fog radio access networks (Fog-RAN). The research leveraged UAVs with on-board radios and processors to enable IoT data aggregation, MIMO wireless access, and device-to-device communications through novel algorithms and hardware solutions.

The project addressed three primary challenges: swarm array synchronization and backhaul, distributed processing on edge devices, and UAV swarm placement for channel optimization and target localization. Two swarm systems were explored: weakly coordinated swarms with light synchronization and strongly coordinated swarms requiring tight synchronization and distributed array processing.

Research activities included studying closed loop adaptive distributed beamforming in weakly coordinated swarms, developing destination-less beamforming protocols in strongly coordinated swarms, creating deep learning approaches for signal strength prediction, and investigating cooperative UAV-based localization using distributed MIMO radar systems. The project focused on optimizing UAV placement to maximize wireless link capacity in urban environments where line-of-sight connections were not guaranteed.

The team developed and analyzed a destination-led distributed beamforming protocol for maximum communication range. Results showed that at large distances, synchronization and channel estimation errors dominated performance limitations. Increasing the number of beamforming radios provided little improvement, while extending preamble duration or increasing destination power had more significant impact on range extension.

Experimental verification used USRP B205-mini software-defined radios mounted on DJI Phantom 3 drones. Despite the small channel coherence time of 84ms due to UAV motion, the beamforming setup achieved 80% of ideal beamforming gains. Guided distributed beamforming demonstrated a 9dB SNR improvement along the beamforming direction.

For target localization, two algorithms were developed to optimize UAV placement: a heuristic search algorithm and a gradient descent-based approach. Both algorithms significantly improved localization accuracy compared to random placement by minimizing the mutual coherence of the dictionary matrix. The results showed that performance could be further improved by increasing the number of UAVs.

Key outcomes included distributed beamforming algorithms for long-range communications with predictable SNR probability, confirmation that synchronization and channel estimation errors were dominant limiting factors at large distances, experimental verification achieving 80% of ideal beamforming gains despite challenging conditions, successful implementation of guided distributed beamforming, development of optimization algorithms that significantly improved localization accuracy, and demonstration that deep learning approaches outperformed state-of-the-art methods for signal strength prediction.

The project successfully addressed UAV swarm communication challenges through novel algorithms and hardware solutions. The research showed that coordinated UAV swarms could significantly enhance wireless communication and sensing capabilities in challenging environments. The developed techniques for beamforming, synchronization, localization and optimal placement established a foundation for future applications in emergency response, IoT data aggregation, and wireless access in underserved areas.

Last Modified: 03/15/2025
Modified by: Danijela Cabric