Drones need wireless connectivity to transfer commands, images and videos, sensor measurements, and other mission-oriented data [9, 19, 13]. The requirements on the wireless technology to be used are very application-specific . Commonly used short-range unlicensed radio technologies — such as Wi-Fi, Bluetooth, and Zigbee — are not fully suited for some commercial drone applications due to demands in terms of radio coverage, throughput, latency, and scalability (see [9, 18]). Existing fourth generation (4G) cellular networks offer wide-area coverage with data rates acceptable for certain drone systems but would still be insufficient for advanced, autonomously organizing multi-drone systems that require high-rate real-time communication and computation offloading. The ongoing rollout of fifth generation (5G) cellular networks is expected to improve the situation with new features like enhanced mobile broadband, ultra-reliable low latency communication, network slicing, and mobile edge computing. In fact, standardization efforts are underway to incorporate drones into 5G networks [7, 10].
This paper aims to examine the issue of 5G connectivity for drones experimentally, taking into account the current state of 5G deployment. In simple terms, our key question is: What happens if we connect a flying drone to one of today’s commercially operated 5G systems? Although cellular-connected drones have been investigated in the past few years (see, e.g., [3, 2, 4, 11]), to our knowledge, this is the first paper with experimental results on radio link quality and throughput between a drone and a 5G base station in commercial operation. The measurements are carried out using our Cellular Drone Measurement Tool (CDMT)  in one of the 5G new radio systems deployed by Magenta Telekom in Austria. To do so, we fly a quadrocopter carrying a 5G user equipment at different heights and different distances in the vicinity of a 5G base station with three sectors (Fig. 1). The downlink throughput is about Mbit/s on average with peaks above Mbit/s. However, frequent handovers to 4G are observed, causing high fluctuations of the overall throughput. As there are no other 5G base stations in the vicinity of our experiments, handovers to 4G are inevitable. This setup reflects the actual situation during the transition phase from 4G to 5G. The results are not representative for a fully-deployed 5G network, but they are valuable as they provide basic insight and serve as a benchmark for further experiments to be conducted in a full 5G network.
The remainder of the paper is structured as follows. Section 2 covers related experimental work on cellular-connected drones and summarizes standardization efforts. Section 3 describes the experimental setup. Section 4 is the core part of the paper: it presents and discusses the experimental results for three different flights. Finally, Section 5 draws general conclusions and states open issues.
2 Related Work
Previous experimental studies on cellular-connected drones study throughput, interference, and handovers in 4G networks (see, e.g., [2, 8, 6, 15]). An average throughput of a few tens of Mbit/s can be achieved both in uplink and downlink. One of the main issues is that drones experience line-of-sight (LoS) links to far distant base stations, which are normally “invisible” to regular ground users. These LoS links cause interference in base stations and aerial drones  and lead to frequent and unnecessary handovers . Interference degrades the throughput for both terrestrial and aerial users [8, 14]. A degraded performance implies that data transmission takes longer and hence requires more resources .
The 3GPP (3rd Generation Partnership Project) Release 15 addresses some of these issues. The technical specification defines the requirements for enhanced support of aerial vehicles . Data rates between and kbit/s are suggested for command-and-control signaling with a maximum one-way latency of ms to ensure proper operational control of the aerial vehicles. The uplink application data rate of Mbit/s is suggested with a latency similar to 4G terrestrial users .
Potential enhancements to support connectivity for aerial devices includes reporting measurements (such as height estimation and radio link quality) from the aerial devices. Such information would be useful for base stations to identify a drone that might experience interference. Furthermore, flight paths can be reported to enable the network to plan for required resources[14, 17].
3 Experimental Setup
The purpose of our study is to analyze 5G radio links for use with drones. The experiments are performed at a site in Feichtendorf, Austria, where Magenta operates a 5G base station (BS). We fly an Asctec Pelican quadrocopter in an open field adjacent to this BS (see Fig. 2). Three types of flights are performed: a liftoff from the ground to a height of m (vertical flight m away from the BS) and two horizontal flights where the drone moves at a height of m and m above ground, respectively. The drone starts m away from the BS, flies m toward the BS with a speed of m/s, and returns to the starting point.
The BS uses 5G New Radio (NR) with 3GPP Release 15 operating at frequencies between and GHz using a MHz band. The system shares one band for downlink (DL) and uplink (UL) and assigns time slots for transmitting and receiving through time-division duplexing (TDD). The 5G BS has massive MIMO (multiple-input multiple-output) antennas with beamforming capabilities, which can electrically change the tilt from to . The horizontal dB beamwidth is and the vertical one is . Besides the 5G NR, the site is also connected to the Magenta Telekom 4G network. It runs the 3GPP Release in the and MHz bands. These two bands can be combined with the carrier aggregation feature to provide higher data rates.
The drone carries a Wistron NeWeb non-standalone mobile test platform based on the Qualcomm Snapdragon X50 5G modem (Fig. 3). The platform supports 5G NR at sub- GHz frequency and uses QAM (quadrature amplitude modulation) with MIMO that allows a maximum data rate of Gbit/s . It also supports 4G and has the ability to connect to both 4G and 5G networks simultaneously through E-UTRAN New Radio Dual Connectivity (ENDC). A custom-built measurement tool  based on an Android API (application programming interface) is used to record the measurements. The measurement tool employs a client-server model, where the client is the application on the drone, and the server resides at the Lakeside Science and Technology Park in Klagenfurt, Austria. The server is accessible from the public Internet via a symmetrical Gbit/s connection. We refer to the communication of the data from (to) the server via 5G to (from) the drone as DL (UL).
4 Experimental Results
|maximum||mean||stddev||5G mean||in 5G|
|Horizontal flight at 30 m||DL||Mbit/s||Mbit/s||Mbit/s||Mbit/s||%|
|Horizontal flight at 100 m||DL||Mbit/s||Mbit/s||Mbit/s||Mbit/s||%|
The communication of the 5G-connected drone is assessed in terms of the reference signal received power (RSRP), signal-to-noise ratio (SNR), throughput, 5G connectivity, and number of handovers at different flight heights. Table1 summarizes some measurement results. They are based on data collected from a single drone flight for each setup. Repetitions of a given setup led to similar qualitative results.
Fig. 4 shows the RSRP, SNR, and throughput during a liftoff from the ground to a height of m. In the DL, the drone initially connects to 4G but performs a handover to 5G at a height of m. A similar behavior occurs in the UL: the drone initially connects to 5G, switches to 4G at a height of m but switches back to 5G at m. The DL throughput over 5G is Mbit/s on average but shows high fluctuations with peaks above Mbit/s. Although the antennas reconfigure electrically for beamforming, the high fluctuation could be caused by connectivity with the side lobes (which have lower radiation intensity than the main lobe). The DL throughput in 4G is Mbit/s on average with peaks up to Mbit/s. The average UL throughput is Mbit/s in 4G and only Mbit/s in 5G. This discrepancy requires further investigation.
4.2 Horizontal flights
Fig. 5 shows the RSRP, SNR, and throughput for a horizontal flight at m height. In the DL, the drone first connects to 5G but immediately switches to 4G due to a disconnection to 5G. As soon as the 5G connectivity is reestablished, a handover back to 5G occurs. Interestingly, such handovers do not occur in the UL: the drone always remains connected to 5G with an average SNR of about dB. Similar to the results during the liftoff, the DL throughput is much higher in 5G than in 4G, namely Mbit/s on average with peaks above Mbit/s, compared to an average 4G throughput of Mbit/s with peaks of about
Mbit/s. However, compared to the liftoff, the fluctuations of the throughput are less severe: the standard deviation isMbit/s compared to more than Mbit/s for the liftoff. In the UL, the throughput remains stable with an average of Mbit/s and a standard deviation of Mbit/s.
Finally, Figs. 6 and 7 show the results when flying at m. The increased height leads to more handovers (DL: 3; UL: 5). This is most likely because the drone connects to distant 4G BSs via side lobes that have better signal quality than the physically closest BS (as in ). The average DL throughput is Mbit/s for 5G and Mbit/s for 4G. The average UL throughput is again lower for 5G ( Mbit/s) than for 4G ( Mbit/s).
5 Conclusions and Outlook
We performed radio link measurements with a drone in a region with a 5G base station that is otherwise covered by 4G. The link quality, throughput, and handovers were analyzed for the liftoff and horizontal flights at two typical heights.
Some specific quantitative results can be generalized as follows. First, the connectivity to the 5G base station cannot be maintained during the entire flight but handovers to 4G occur. As the flight height increases, the throughput suffers and handovers become more frequent.
Second, drones at typical flight heights can receive several hundred Mbit/s over 5G, which is sufficient for many applications. However, we have to keep in mind that the 5G DL performance is biased by the fact that there is only one single 5G BS. Broader deployment of 5G will change the setup with unclear consequences. On the one hand, more 5G BSs on the same frequency are expected to cause interference to airborne drones, due to line-of-sight links, which would lower the 5G throughput (as discussed for 4G in ). On the other hand, further 5G BSs will diminish handovers to 4G and thus in turn improve the overall DL throughput.
Third, the 5G implementation used does not improve the UL throughput compared to 4G. This issue needs further investigation, especially as the UL is important for drones that send pictures and videos to the BS.
This work is performed in a collaboration between the University of Klagenfurt, Lakeside Labs GmbH, and Magenta Telekom GmbH (previously T-Mobile Austria GmbH). R. Muzaffar is funded by the security research program KIRAS of the Federal Ministry for Transport, Innovation, and Technology (bmvit), Austria, under grant agreement n. 854747 (WatchDog). C. Bettstetter is faculty member in the Karl Popper Kolleg on networked autonomous aerial vehicles at the University of Klagenfurt.
-  (2014-05) Application-driven design of aerial communication networks. IEEE Commun. Mag. 52 (5), pp. 129–137. Cited by: §1.
-  (2019) LTE ground-to-air field measurements in the context of flying relays. IEEE Wireless Communications 26 (1), pp. 12–17. Cited by: §1, §2.
-  (2019) Cellular connectivity for uavs: network modeling, performance analysis, and design guidelines. IEEE Transactions on Wireless Communications (7), pp. 3366–3381. Cited by: §1.
-  (2017) Performance of mobile networks with UAVs: can flying base stations substitute ultra-dense small cells?. In Proc. European Wireless, Cited by: §1.
-  (2016) LTE in the sky: trading off propagation benefits with interference costs for aerial nodes. IEEE Communications Magazine 54 (5), pp. 44–50. Cited by: §2, §5.
-  (2019) Handover challenges for cellular-connected drones. In Proc. ACM Workshop on Micro Aerial Vehicle Networks, Systems, and Applications (DroNet), Cited by: §2, §4.2.
-  (2017)(Website) External Links: Cited by: §1, §2.
-  (2019) An experimental evaluation of LTE-A throughput for drones. In Proc. ACM Workshop on Micro Aerial Vehicle Networks, Systems, and Applications (DroNet), Cited by: §2.
-  (2016) Survey on unmanned aerial vehicle networks for civil applications: a communications viewpoint. IEEE Communications Surveys & Tutorials 18 (4), pp. 2624–2661. Cited by: §1.
-  (2017) Placement of 5G drone base stations by data field clustering. In Proc. IEEE Vehicular Technology Conf. (VTC), Cited by: §1.
-  (2019) UAV communications for 5g and beyond: recent advances and future trends. IEEE Internet of Things Journal 6 (2), pp. 2241–2263. Cited by: §1.
-  (2019-09) Real-world performance of 5G. Technical report Society of Cable Telecommunications Engineers-The International Society of Broadband Experts (SCTE-ISBE) and National Cable & Telecommunications Association (NCTA). Cited by: §3.
-  (2019) A tutorial on UAVs for wireless networks: applications, challenges, and open problems. IEEE Communications Surveys & Tutorials. Cited by: §1.
-  (2018) An overview of 3GPP Release-15 study on enhanced LTE support for connected drones. arXiv preprint arXiv:1805.00826. Cited by: §2, §2.
-  (2018) How to ensure reliable connectivity for aerial vehicles over cellular networks. IEEE Access 6, pp. 12304–12317. Cited by: §2.
-  (2019) A performance evaluation tool for drone communications in 4G cellular networks. In Proc. Intern. Symp. on Wirelss Communication Systems (ISWCS), Cited by: §1, §3.
-  (2018)(Website) External Links: Cited by: §2.
-  (2018) A telecom perspective on the Internet of drones: from LTE-Advanced to 5G. arXiv preprint arXiv:1803.11048. Cited by: §1.
-  (2019) Accessing from the sky: a tutorial on UAV communications for 5G and beyond. Proceedings of the IEEE 107 (12), pp. 2327–2375. Cited by: §1.