There is growing interest in the use of unmanned aerial vehicles (UAVs), or drones, across a range of applications, including public safety, disaster response, environmental monitoring, package delivery, and wireless communications. A particularly promising direction is the use of drones as aerial base stations to improve the performance and coverage of terrestrial cellular networks, especially in scenarios where infrastructure is damaged, unavailable, or sparse. However, because these aerial platforms often operate in the same radio frequency (RF) spectrum as existing terrestrial systems, their integration introduces new challenges related to interference management and spectral coexistence. The goal of this collaborative SpecEES project was to develop a rigorous scientific foundation for analyzing and enabling the coexistence of drone and terrestrial wireless networks under realistic deployment and hardware constraints. To this end, the project team employed tools from communication theory, stochastic geometry, propagation modeling, and machine learning to study the fundamental trade-offs, performance limits, and design principles for these emerging hybrid systems, which we refer to as DroTerNets.

A key contribution of the VT team was demonstrating the applicability of determinantal point processes (DPPs) to wireless network optimization by developing a DPP-based learning framework (DPPL) for solving optimization problems that require balancing quality and similarity in subset selection. Specifically, DPPs allow for the selection of elements that are individually high quality (such as wireless links with high data rates) while also being mutually dissimilar, for example, by selecting links that are spatially well separated. This trade-off is essential for problems such as scheduling, channel assignment, and resource allocation in interference-limited environments. As two representative extensions of this framework, we first introduced a method for defining similarity matrices that are not necessarily symmetric, enabling the modeling of asymmetric interactions arising from directional antennas in drone networks. Second, we developed an approach based on k-determinantal point processes to select subsets of fixed size, which we applied to dynamic channel assignment under practical physical-layer impairments, including receiver nonlinearity and adjacent-channel interference. To support the design and analysis of DroTerNets, we introduced new mathematical models based on random geometric graphs that account for key features such as drone mobility and route-induced spatial correlations. Complementing these theoretical developments, we carried out a range of propagation and application-oriented investigations, including the development of a unified air-to-ground channel model that captures the joint impact of UAV wobbling and RF hardware impairments, and the use of diffraction-based propagation models to support outdoor-to-indoor communication and localization. As a natural extension of the project's broader goals, we also explored a representative case study in public safety applications by developing a 3GPP-compliant framework that repurposes uplink sounding reference signals (SRS) for efficient indoor positioning, leading to low-complexity algorithms suitable for UAV-enabled emergency response. Building on this, we also contributed to early efforts evaluating localization using low Earth orbit (LEO) satellites within the 3GPP framework, including location verification in non-terrestrial networks and the potential of multi-LEO and hybrid LEO-GNSS configurations for positioning in future 6G systems. Portions of this work, particularly in propagation modeling, were conducted in collaboration with our partner institutions. Together, these results represent a sample of the project's broader intellectual contributions, which span learning-based optimization, stochastic modeling, channel characterization, and emerging use cases in 5G and beyond.

The broader impacts of this project include the training and mentorship of three Ph.D. students, two of whom have already completed their degrees and transitioned to successful careers in research. The research results were disseminated widely through journal and conference publications, invited talks, tutorials, and code releases via GitHub and IEEE DataPort. Key ideas from the project were integrated into graduate and undergraduate courses and shared through a variety of outreach activities, including a lecture on wireless systems delivered to more than 100 elementary and middle school students through Virginia Tech's Kids' Tech University. Throughout the duration of the project, the PIs organized an annual workshop on drone-assisted wireless communications, creating a dedicated forum for researchers and practitioners to exchange ideas and stay informed about the latest developments in this research direction. In addition, the PIs delivered dozens of seminars, colloquia, and invited talks at universities, government agencies, and industry research labs. These engagements helped shape conversations on spectrum sharing and drone integration, and also led to close collaborations with industry partners, resulting in joint publications and continued technical exchange. Through these efforts, the project has not only advanced the technical foundations of drone-terrestrial coexistence but has also contributed meaningfully to the broader wireless research and policy ecosystem.

Last Modified: 04/01/2025
Modified by: Harpreet S Dhillon