UAV-aided Fast Data Collection via Machine Learning Using AERPAW's Digital Twin
Conference
Sadique, JJ, Khan, MR, Ibrahim, AS. (2025). UAV-aided Fast Data Collection via Machine Learning Using AERPAW's Digital Twin
. 89-96. 10.1145/3737895.3768305
Sadique, JJ, Khan, MR, Ibrahim, AS. (2025). UAV-aided Fast Data Collection via Machine Learning Using AERPAW's Digital Twin
. 89-96. 10.1145/3737895.3768305
Unmanned aerial vehicles (UAVs) have emerged as essential components for 5G and beyond networks. Particularly, UAVs serve as aerial data mules, capable of efficiently collecting and transferring data from geographically dispersed ground-based nodes, such as ground base stations. However, practical limitations like battery life constraints, stringent mission timeliness, and varying wireless conditions present significant challenges for efficient data collection. To reduce the mission completion time, optimizing UAV's trajectory is often considered a primary challenge. In addition, data collection time under hovering mode also greatly affect the mission time. In this work, we propose machine learning (ML)-based UAV-aided data collection model that leverages K-means clustering to identify optimal waypoints near each base station within Aerial Experimentation and Research Platform for Advanced Wireless (AERPAW) geofence area, located in Raleigh, North Carolina, USA. By adopting the clustering strategy, our primary goal is to reduce flight time without compromising the high signal-to-noise ratio (SNR) across trajectories. Our proposed model also tackles the challenge of dynamic hovering time adjustment at each waypoint by incorporating an adaptive approach based on instantaneous link quality and data volume. To validate the proposed data collection model, we utilize a digital twin simulation environment from AERPAW. The results emulated over the AERPAW's digital twin demonstrate the effectiveness of the proposed algorithm, ensuring enhanced efficiency in the UAV-aided fast data collection strategy.
