Home → Magazine Archive → January 2021 (Vol. 64, No. 1) → Technical Perspective: SkyCore's Architecture Takes... → Full Text

Technical Perspective: SkyCore's Architecture Takes It to the 'Edge'

By Richard Han

Communications of the ACM, Vol. 64 No. 1, Page 115

Save PDF

The following SkyCore paper addresses an exciting use case for Unmanned Aerial Vehicles (UAVs) or drones in which UAVs can act as mobile base stations for the cellular network, flying to areas in the cellular network in order to improve wireless connectivity in those areas. This adaptive capability to patch the network capacity on demand would be useful to address hotspots, such as at sporting venues or other temporary events that have insufficient network capacity, and/or emergency scenarios when parts of the cellular network are incapacitated. The paper uses the Long-Term Evolution (LTE) standard as a case study for providing on-demand adaptive cellular connectivity via UAVs.

The challenge of this work is how to adapt the existing LTE standard to support the concept of UAV-based mobile base stations, especially in the presence of multiple UAVs. In current cellular networks, base stations employ a Radio Access Network (RAN) to communicate with clients, for example, cell phones. Packets are then routed over a high-speed wired network of gateways comprising the Evolved Packet Core (EPC) network to the Internet. The paper observes that current cellular operators typically deploy UAV base stations with an architecture in which the UAVs contain the RAN while the EPC is ground-based. In order to connect the UAV-based RAN to the EPC, the UAV is either tethered via wire to the UAV base stations (limiting their mobility and range) or connected wirelessly to the UAV, exposing EPC communication to the unreliability of the wireless link.

Instead, the authors propose a novel Edge-EPC network architecture called SkyCore in which EPC functionality is pushed into the extreme edge, namely, the UAV itself. This avoids the tethering and wireless unreliability problems noted earlier but introduces two new challenges. First, the UAVs have limited computational resources. Second, the hierarchical nature of the standard EPC network provides a global view that can manage hand-off of a mobile client from one base station to the next, whereas UAV-based EPC functionality will not have a global view. The authors propose novel solutions to these problems respectively: software refactoring to reduce the EPC's footprint on the UAV; and proactive inter-UAV communication for EPC agents via a new software-defined networking (SDN) control-data interface.

The paper makes the following contributions: First, it builds a real-world prototype of the Skycore system consisting of a two UAV LTE network that seamlessly works with commercial off-the-shelf RANs and mobile LTE clients. Second, the paper shows the feasibility of SkyCore UAVs acting as adaptive LTE-hotspots, providing improved on-demand network capacity to clients. Third, the paper demonstrates the feasibility of Skycore to act as an independent ad hoc LTE network, connecting geographically separated clients through two different UAVs, while also allowing for seamless hands-off. Fourth, the work experimentally shows that when compared to a generic Edge-EPC architecture, SkyCore's enhanced Edge-EPC features lower control plane latencies by an order of magnitude, and lower CPU utilization by a factor of five.

The SkyCore system introduces a new edge-centric cellular network architecture that opens up the possibility for efficiently supporting mobile drone-based base stations in future hotspot and emergency scenarios. Skycore's Edge-EPC architecture also has the virtue that it is not limited to LTE networks and can be generalized to 5G cellular networks and beyond.

Back to Top


Richard Han is a professor in the Department of Computer Science at the University of Colorado, Boulder, CO, USA.

Back to Top


To view the accompanying paper, visit doi.acm.org/10.1145/3434161

Copyright held by author.
Request permission to (re)publish from the owner/author

The Digital Library is published by the Association for Computing Machinery. Copyright © 2021 ACM, Inc.


No entries found