Home

Project Title: SpecEES: Collaborative Research: DroTerNet: Coexistence between Drone and Terrestrial Wireless Networks

Funding Agency: National Science Foundation (Division of Computer and Network Systems)

Award Number: CNS-1923807 (Virginia Tech), CNS-1923601 (USC), CNS-1923774 (Cal State LA)

Project Summary:

There is tremendous recent interest in drones with applications ranging from public safety, first responders, surveillance, to package delivery. Drones are also being considered as flying wireless nodes to augment the capabilities of current terrestrial communication networks. Irrespective of the application, drones need radio frequency (RF) spectrum to communicate with their ground control stations as well as with other drones and terrestrial nodes. Since transmissions from higher altitude have the potential of interfering with other wireless services over a large area, it is currently being debated whether and under what rules should drones share spectrum with existing networks or whether it is better to operate them over specifically licensed frequencies. In order to answer such important and timely questions, this project develops a new cross-disciplinary approach to the design and analysis of coexisting drone and terrestrial networks (DroTerNets) by blending ideas from multiple disciplines, such as spectrum sharing, communication theory, propagation science, test-bed development, machine learning, and stochastic network modeling. This research will inform both industry and government on spectrum usage by providing a scientific basis for the high-stakes ruling on spectrum for drones. Further broader impacts will be through student training and wide dissemination of results.

The overarching goal of this research is to develop a holistic new approach to the spectral and energy efficiency analysis of DroTerNets, yielding the following key innovations: (i) A new learning framework based on the idea of determinantal point processes (DPPs) will be developed to facilitate both simulation-based and analytical characterization of the locations of simultaneously active nodes in a given frequency band for a variety of coexistence schemes, (ii) Drawing on multi-label classification in machine learning, a novel deep DPP-based channel assignment algorithm will be developed by utilizing the structure of DPP kernels to limit the search space, (iii) Non-linear receiver characteristics will be included in the learning framework to both quantify their effect on the energy and spectral efficiency of DroTerNets and to develop novel receiver-aware channel assignment schemes, (iv) Mobility constraints and characteristics of drones that result from the opportunistic access of the channel will be characterized and incorporated in the analysis, (v) Measurements and models of air-to-ground (A2G) channels in a variety of environments with particular emphasis on directional characteristics that determine the effectiveness of multi-antenna receivers will be obtained, and (vi) Experimental investigation and modeling of the correlation between terrestrial and A2G links will be performed to provide a solid foundation for coexistence margins.