PaperlayerView exact revision on arXiv

Source-linked AI summary

Understanding UAV Cellular Communications: From Existing Networks to Massive MIMO

Giovanni Geraci, Adrian Garcia-Rodriguez, Lorenzo Galati Giordano, David López-Pérez, Emil Björnson

arXiv:1804.08489v1cs.ITeess.SPmath.IT

TL;DR

Reliable UAV command-and-control links are important for remote operation, but aerial propagation can disrupt communication with both UAVs and ground users. The article compares realistic 3GPP-based existing cellular and multi-user massive MIMO deployments, finding higher C&C reliability with massive MIMO while identifying pilot contamination as a concern for ground users.

  • Problem

    Reliable downlink C&C links are needed for safe remote UAV operation beyond visual line-of-sight constraints, while aerial links can create substantial interference for neighboring base stations and users.

    Source: A. Background and Motivation · p. 1

  • Method

    The study compares existing single-user cellular networks with multi-user massive MIMO using realistic 3GPP models and a 10 MHz bandwidth.

    Source: §VI · p. 10

  • Results

    74% with pilot reuse and 96% without it: massive MIMO supports a 100 kbps C&C channel for UAVs up to 300 m, versus 1% reliability at 300 m in existing single-user networks.

    Source: B. Approach and Summary of Results · p. 2

  • Takeaways & Limitations

    UAV pilots can severely contaminate ground-user pilots and degrade their performance, so uplink power control is required to protect ground users.

    Source: B. Approach and Summary of Results · p. 2

Abstract

from arXiv · show

The purpose of this article is to bestow the reader with a timely study of UAV cellular communications, bridging the gap between the 3GPP standardization status quo and the more forward-looking research. Special emphasis is placed on the downlink command and control (C&C) channel to aerial users, whose reliability is deemed of paramount technological importance for the commercial success of UAV cellular communications. Through a realistic side-by-side comparison of two network deployments -- a present-day cellular infrastructure versus a next-generation massive MIMO system -- a plurality of key facts are cast light upon, with the three main ones summarized as follows: (i) UAV cell selection is essentially driven by the secondary lobes of a base station's radiation pattern, causing UAVs to associate to far-flung cells; (ii) over a 10 MHz bandwidth, and for UAV heights of up to 300 m, massive MIMO networks can support 100 kbps C&C channels in 74% of the cases when the uplink pilots for channel estimation are reused among base station sites, and in 96% of the cases without pilot reuse across the network; (iii) supporting UAV C&C channels can considerably affect the performance of ground users on account of severe pilot contamination, unless suitable power control policies are in place.

I. INTRODUCTION

The article examines how cellular networks can provide reliable downlink command-and-control links to UAVs, whose aerial propagation can disrupt both UAV and ground-user communications. Using realistic 3GPP models, it compares existing single-user networks with multi-user massive MIMO and evaluates UAV cell selection, C&C reliability, and effects on ground users.

A. Background and Motivation

  • Reliable downlink C&C is required for safe remote UAV operation beyond visual line-of-sight constraints, while aerial line-of-sight links can create substantial interference to neighboring base stations.3GPP standardization and academic research motivate the article's effort to connect current practice with forward-looking study.Source: A. Background and Motivation · p. 1

B. Approach and Summary of Results

  • The study compares existing single-user cellular deployments with multi-user massive MIMO using a 10 MHz bandwidth and a 100 kbps C&C target.The article also assesses UAV cell selection and effects on conventional ground users.Source: B. Approach and Summary of Results · p. 2
  • At heights of 75 m and above, secondary antenna lobes largely drive cell selection, so UAVs generally associate with farther rather than nearby base stations.The passage links this behavior to nearly free-space propagation.Source: B. Approach and Summary of Results · p. 2
  • In existing single-user networks, 100 kbps C&C reliability falls from 87% at 1.5 m to 35% at 50 m, 2% at 150 m, and 1% at 300 m.The passage attributes the decline to strong line-of-sight interference from many visible cells.Source: B. Approach and Summary of Results · p. 2
  • 74% with pilot reuse and 96% without it: multi-user massive MIMO supports a 100 kbps C&C channel for UAVs up to 300 m.The gains are attributed to mitigated interference, a stronger carrier signal, and spatial multiplexing.Source: B. Approach and Summary of Results · p. 2
  • UAV transmissions can significantly degrade massive MIMO ground-user performance because overlapping UAV pilots severely contaminate ground-user pilots; uplink power control is required to protect them.Source: B. Approach and Summary of Results · p. 2

II. 3GPP SYSTEM SETUP

The system setup uses a 3GPP-compliant cellular topology and channel model to evaluate ground and aerial users. It defines user association and scheduling, then specifies the propagation, signal, and rate models used in the analysis.

A. Cellular Network Topology

  • The network uses a hexagonal layout with three co-located 120-degree sectors per site; users associate with the base station providing the highest reference signal received power.UAVs range from 1.5 m to 300 m in height, and scheduling distinguishes single-user from multi-user operation.Source: A. Cellular Network Topology · p. 2

B. 3D Propagation Channel Model

  • The 3GPP channel model includes antenna gain, path loss, shadow fading, small-scale fading, 3D directionality, and height-aware aerial propagation parameters.The model also accounts for spatially correlated shadow fading and time- and frequency-correlated small-scale fading.Source: B. 3D Propagation Channel Model · p. 2
  • The received signal model distinguishes the useful signal, intra-cell interference in multi-user operation, inter-cell interference, and thermal noise.Precoding is normalized to satisfy the total power constraint; the setup uses identical-scalar precoding for single-user operation and digital zero-forcing for multi-user operation.Source: B. 3D Propagation Channel Model · p. 3
  • User rates are obtained by mapping SINR to the maximum modulation and coding scheme meeting a 10^-1 block error rate, with control-signaling overhead included.The stated spectral-efficiency range is 0.22 to 7.44 b/s/Hz.Source: B. 3D Propagation Channel Model · p. 3

III. INTEGRATING UAVS INTO EXISTING NETWORKS

The existing-network setup models sectorized base stations serving users in single-user mode, without digital precoding. It examines UAV association and downlink C&C performance as UAV height changes.

III. INTEGRATING UAVS INTO EXISTING NETWORKS

  • Base stations use 16 antennas in an eight-element cross-polarized vertical array, with 65° half-power beamwidth, 12° mechanical downtilt, and one RF chain.The configuration exemplifies existing cellular networks.Source: §III · p. 3
  • Each base station serves at most one user per PRB, without digital precoding; the paper calls this setup single-user mode.The analysis examines how UAV height affects cell association and downlink C&C performance in this mode.Source: §III · p. 3
  • Table II is identified as a table of system parameters, but the supplied passage does not provide its parameter values.Source: Table II · p. 4

A. UAV Association

UAV height changes cell-association distances and the antenna-pattern mechanisms behind association. At higher heights, secondary lobes can lead UAVs to associate with distant base stations.

A. UAV Association

  • The association analysis compares 3D serving-base-station distance and antenna gain versus 2D distance across UAV heights of 1.5, 50, 75, 150, and 300 m.Source: A. UAV Association · p. 4

UAVs at 1.5 m:

  • At 1.5 m, a UAV is in a base station’s main lobe when its 2D distance exceeds 52 m, while antenna gain is maximized at 80–180 m.Associations to non-closest base stations are mostly due to shadow fading and line-of-sight conditions.Source: UAVs at 1.5 m · p. 4
  • At 1.5 m, UAVs generally associate with a base station within about 250 m; association at 500 m or more occurs in 9% of cases.Their association distance is smoothly distributed between 35 m and 1 km.Source: UAVs at 1.5 m · p. 4

UAVs at 50 m:

  • For 3D distances of 250 m or less, 1.5 m and 50 m UAVs have similar association-distance distributions and antenna-gain patterns.Their similar vertical distances to 25 m base stations yield similar viewing angles.Source: UAVs at 50 m · p. 4
  • At 50 m, UAVs are more likely than 1.5 m UAVs to associate with base stations 500 m away when their 3D distance exceeds 250 m: 17% versus 9%.The passage attributes this difference to antenna-gain trends and the greater likelihood of multiple neighboring base stations in line of sight.Source: UAVs at 50 m · p. 4

A. UAV Association

  • For UAVs at 75 m and above, secondary-lobe antenna gains can outweigh path-loss differences, leading to association with distant base stations.Near-free-space propagation makes the path-loss difference between nearby and farther base stations less significant than their antenna-gain difference.Source: UAVs at 75 m and above · p. 4

UAVs at 75 m and above:

At heights of 75 m and above, secondary lobes shape the distance ranges at which UAVs associate with base stations. The analysis also reports how rising height affects coupling loss and SINR per PRB.

UAVs at 75 m and above:

  • For UAVs at 150 m, associated-UAV locations show distinct distance ranges, each corresponding to a step in the serving-BS distance CDF and, equivalently, to a secondary antenna lobe.Source: UAVs at 75 m and above · p. 4

Carrier signal strength and interference:

  • Figure 6 reports coupling loss and SINR per PRB versus UAV height, with 5%-best, average, and 5%-worst values for each metric.Coupling loss includes antenna gain, path loss, and shadow fading; SINR also includes small-scale fading and interference.Source: Carrier signal strength and interference · p. 5
  • As UAVs rise to around 25 m, average coupling loss improves, but continued height increases distance to the serving base station and degrades coupling loss.The 5%-best UAVs instead experience degraded coupling loss up to around 25 m as antenna gain diminishes.Source: Carrier signal strength and interference · p. 5

UAVs at 75 m and above:

Carrier signal strength and interference:

  • At around 25 m, average SINR per PRB generally degrades as more neighboring base stations become visible as strong line-of-sight interferers.As height continues increasing, SINR decreases more slowly than coupling loss, alongside a slight reduction in interference.Source: Carrier signal strength and interference · p. 5
  • Beyond 100 m, average SINR per PRB falls below the minimum MCS threshold of -5.02 dB; beyond 150 m, even the 5%-best SINR falls below it.Source: Carrier signal strength and interference · p. 5

C&C channel data rates:

In single-user networks, the 100 kbps C&C target becomes much less reliable as UAV height increases. The section then introduces massive MIMO, where beamforming and spatial multiplexing support multiple users per PRB.

C&C channel data rates:

  • In single-user mode, even the 5%-best UAVs experience very low SINR, so opportunistic proportional-fair scheduling alone would not make C&C performance acceptable.The comparison uses round-robin scheduling.Source: C&C channel data rates · p. 5
  • At 150 m and 300 m, the 100 kbps target is met just 2% and 1% of the time, respectively, versus 87% at 1.5 m.At around 50 m and 75 m, the target is met 35% and 40% of the time, and rates exceed 1 Mbps just 0.3% of the time in that height range.Source: C&C channel data rates · pp. 5–6

IV. SUPPORTING UAVS THROUGH MASSIVE MIMO

  • The massive MIMO setup uses 128 antennas and lets each base station serve up to eight users per PRB through digital zero-forcing precoding.The antennas form an 8 × 8 planar array of cross-polarized elements with 128 RF chains; the system uses TDD and channel reciprocity for pilot-based channel estimation.Source: §IV · p. 6

Channel state information acquisition:

Channel estimation uses uplink pilots, with reuse across base-station sites creating pilot contamination. Fractional uplink power control limits transmit power while compensating for only part of path loss.

Channel state information acquisition:

  • Pilot Reuse 3 makes pilots orthogonal across the three sectors at a site but reuses them across sites, generating pilot contamination.Each sector randomly allocates its pilot pool to its served users.Source: Channel state information acquisition · p. 6
  • Fractional uplink power control compensates only for a fraction α of path loss, subject to the maximum user transmit power.Its parameters include the maximum transmit power, a cell-specific parameter, and average channel gain measured from RSRP.Source: Channel state information acquisition · p. 6
  • Each base station correlates the received signal with the known pilot to obtain a least-squares channel estimate, rejecting interference from other orthogonal pilots.Intra-cell pilot contamination is absent because each base station assigns different pilots to users in its own cell.Source: Channel state information acquisition · p. 6

Spatial multiplexing through digital precoding:

Digital zero-forcing precoding serves multiple users per PRB. In the evaluated massive MIMO setup, beamforming improves UAV link performance, and the system supports the 100 kbps C&C target with high reliability across the studied heights.

Spatial multiplexing through digital precoding:

  • Digital zero-forcing precoding attempts to suppress intra-cell interference while serving multiple users simultaneously on each PRB.The precoding vectors are normalized to meet the transmit-power constraint with equal power allocation across users.Source: Spatial multiplexing through digital… · p. 7

A. UAV C&C Channel Performance Improvement

  • Massive MIMO improves UAV coupling loss through increased antenna gain toward the sky and improves SINR through serving-cell beamforming and reduced neighboring-cell interference.Neighboring base stations generally point beams downward because most users are ground users; at around 25 m, more visible neighboring base stations can still act as strong line-of-sight interferers.Source: A. UAV C&C Channel Performance… · p. 7
  • At all considered UAV heights, massive MIMO supports a 100 kbps C&C channel in at least 96% of cases through SINR and spatial-multiplexing gains.The setup assumes perfect channel state information for this theoretical performance comparison; eight UAV and ground users share each PRB.Source: A. UAV C&C Channel Performance… · p. 7

B. Performance Interplay between UAVs and Ground Users

The section examines how supporting UAV C&C affects UAV and ground-user performance under single-user and multi-user operation. In multi-user cases, UAV SINR improves over single-user operation, while pilot contamination particularly harms GUEs and worsens as UAV numbers rise.

B. Performance Interplay between UAVs and Ground Users

  • The evaluation examines UAV and GUE performance in single-user and multi-user settings, including how SRS reuse, contamination, and UL power control affect channel estimation.The multi-user cases include one UAV and 14 GUEs, three UAVs and 12 GUEs, or five UAVs and 10 GUEs per sector.Source: B. Performance Interplay between UAVs… · p. 7, 3GPP case studies · p. 7

3GPP case studies:

  • UAV SINR per PRB greatly improves in multi-user versus single-user mode, owing to beamforming gain and reduced interference from neighboring downward-pointing cells.The comparison uses realistic CSI acquisition with SRS Reuse 3 and UL fractional power control.Source: 3GPP case studies · pp. 7–8
  • More UAVs reduce multi-user UAV SINR from Case 3 to Cases 4 and 5 through greater pilot contamination and more upward-pointing neighboring beams.UAV count does not affect single-user-mode SINR in these scenarios.Source: 3GPP case studies · p. 8
  • GUE SINR does not improve in multi-user mode because severe pilot contamination outweighs beamforming gains, and it degrades further in Cases 4 and 5.GUE SRS signals are likely to collide with those of neighboring UAVs whose strong LoS links reach the GUE-serving BS; more neighboring UAVs intensify this contamination.Source: 3GPP case studies · p. 8
  • The rate analysis compares perfect CSI, SRS Reuse 3 with fractional UL power control, and SRS Reuse 3 with equal UL power allocation.It reports UAV and GUE rates for Case 3, with the 100 kbps C&C target shown as a benchmark in the rate figures.Source: Benefits of massive MIMO and UL power… · p. 8, Figure 9 · p. 8

Benefits of massive MIMO and UL power control:

Massive MIMO supports the UAV C&C rate target in many cases, but pilot contamination reduces achievable rates and can severely affect GUEs. UL fractional power control helps contamination-vulnerable GUEs more than it helps UAV C&C performance.

Benefits of massive MIMO and UL power control:

  • Median UAV rates with imperfect CSI fall to 40% of rates without channel-estimation errors.Pilot contamination can significantly degrade the rates of both UAVs and GUEs.Source: Benefits of massive MIMO and UL power… · p. 8
  • UL fractional power control does not significantly protect UAV C&C, but helps GUEs in the lower tail by reducing UAV-generated SRS power.UAVs’ strong LoS links to many BSs make them a main source of pilot contamination, while the affected GUEs benefit from the larger power reduction of UAV SRSs.Source: Benefits of massive MIMO and UL power… · p. 8
  • Massive MIMO boosts GUE data rates through multiplexing gain rather than SINR gain.The cited SINR comparison is shown in Fig. 10.Source: Benefits of massive MIMO and UL power… · p. 8, Figure 10 · p. 8
  • With perfect CSI, massive MIMO meets the UAV C&C channel target rate of 100 kbps in 96% of cases, versus 16% under single-user setups; under pilot contamination, it achieves 74% (“MU, R3 PC”).To address the pilot-contamination performance gap, the text proposes better pilot assignment and more advanced channel estimation and precoding based on multi-cell processing.Source: Benefits of massive MIMO and UL power… · p. 8

V. FUTURE OUTLOOK

The future-outlook section surveys complementary ways to improve UAV C&C support and UAV–GUE coexistence. It discusses interference blanking and opportunistic scheduling, with blanking constrained by coordination and resource costs as UAV density rises.

V. FUTURE OUTLOOK

  • The outlook identifies complementary techniques for UAV performance and UAV–GUE interplay, many of which rely on identifying UAVs through mobility or handover history or UAV-provided reports and altitude information.Table III summarizes the proposed solution categories for future research.Source: §V · p. 9, Table III · p. 9

Interference blanking:

  • Interference blanking uses almost blank subframes from the strongest interfering cells on resources assigned to a UAV C&C channel.The number of cells needing protection and the coordination cluster and sacrificed time-frequency resources grow with UAV height, making the approach potentially suitable only for low-density UAV networks.Source: Interference blanking · p. 9

Opportunistic scheduling:

  • Reducing UAV density fivefold from Case 5 to Case 3 increases by 50% the number of UAVs achieving the minimum MCS SINR.This motivates opportunistically scheduling nearby UAV C&C channels on different PRBs to avoid nearby upward-pointing beams.Source: Opportunistic scheduling · p. 9

Fractional pilot reuse:

The section reviews pilot-reuse and power-control options alongside cooperative processing and UAV-side antennas. The conclusion frames these as complementary procedures for further investigation across time, frequency, power, and spatial domains.

Fractional pilot reuse:

  • Fractional pilot reuse could reserve a dedicated, sporadically reused pilot pool for UAVs while allowing more aggressive reuse for GUE pilots, limiting contamination without large overhead [38].Fully orthogonal network-wide pilots provide higher-quality CSI but their overhead is infeasible in practical systems.Source: Fractional pilot reuse · pp. 9–10
  • Power control could use UAV-specific P0 and α values or account for neighboring-cell RSRP, since serving-cell RSRP alone may prompt a high UAV to increase interference toward multiple LoS-connected BSs.The proposed adaptations include reducing UAV power when serving and neighboring RSRPs are both low and similar.Source: Uplink power control · p. 10

Cooperative MIMO:

  • CoMP can turn interference into useful signal through coherent BS transmission, but requires UAV-data sharing over X2 and tight symbol-level synchronization.CEA precoding instead uses inter-cell CSI to steer interference toward neighboring UAV channel nullspaces; both approaches face larger coordination clusters and frequent updates under high UAV mobility,, –.Source: Cooperative MIMO · p. 10

UAV directional antennas:

  • Directional antennas without beamsteering provide no significant DL benefit in preliminary studies and may harm UL performance, while UAV-side antenna selection or beamforming may improve DL reception and mitigate UL interference,.The latter capabilities increase UAV hardware and computational complexity, and manufacturer uptake is uncertain.Source: UAV directional antennas · p. 10

VI. CONCLUSIONS

  • The article compares traditional single-user cellular service with multi-user massive MIMO using realistic 3GPP models, then examines UAV height, cell selection, interference, and pilot contamination.Its conclusion presents time-, frequency-, power-, and spatial-domain procedures as topics meriting further investigation.Source: §VI · p. 10
Loading 1804.08489v1…