accessing from the sky: uav challenges from a ... · google, facebook, amazon, huawei , qualcomm,...

Accessing from the sky: UAV challenges from a communication and signal processing perspective

Rui Zhang

IEEE Fellow/Distinguished LecturerClarivate Analytics Highly Cited Researcher

Dean’s Chair Assoc. Professor, National University of Singapore

(e-mail: [email protected])

IEEE SPAWC 2019 2-5 July 2019, Cannes, France

SPAWC 2019 1

mailto:[email protected]

Outline Introduction UAV communication requirement Wireless technologies for UAV The new paradigm: Integrating UAVs into cellular

What’s new over terrestrial communications?

UAV Communication Fundamentals

Challenges and Promising Solutions UAV-assisted wireless communication (UAV as BS/relay) Cellular-connected UAV (UAV as user/terminal)

Conclusions

Rui Zhang, National University of Singapore

SPAWC 2019 2

Rui Zhang, National University of SingaporeIntroduction

UAVs/Drones: Whose Time is Coming

SPAWC 2019 3

Anticipated global consumer UAV market by 2020 16 million UAVs with total market value over 120 billion USD

In June 2016, the U.S. Federal Aviation Administration (FAA) released the UAS operational rules Civilian use of small unmanned aircraft systems (UAS) with aircraft weight

less than 55 pounds (25 kg) In November 2017, FAA further launched “Drone Integration Pilot Program”

Expanded use of drones, including beyond-visual-line-of-sight (BVLoS) flights, night-time operations, flights over people, etc.

In March 2017, 3GPP launched new study item on “enhanced support for aerial vehicles using LTE” Special use cases also considered for 5G NR, V2X, etc.

Numerous pilot projects, prototypes, field measurements by industry giants Google, Facebook, Amazon, Huawei, Qualcomm, AT&T, Nokia, etc.


UAV Applications

SPAWC 2019 4

Aerial photography Drone DeliveryInspection Precision Agriculture

Traffic offloading Mobile relaying IoT Data Harvesting


Notre-Dame de Paris Fire Fight with Drones

SPAWC 2019 5

DJI Drones Assisted Notre Dame Fire-Fighters to Defeat Fire in Paris, France

https://www.droningon.co/2019/04/17/dji-drones-assist-notre-dame-fire-fighters-paris/


Wireless Communications for UAVs: Basic Requirement

Control and Non-Payload Communications (CNPC) Ensure safe, reliable, and effective

flight operation Low data rate, high reliability, high

security, low latency

Payload Communications Application specific data (e.g.,

HD/4K video) Much higher rate than CNPC, less

stringent on reliability/latency

CNPC information flows [ITUReportM.2171]

UAVControl Station

UAV communication requirement

ITU, “Characteristics of unmanned aircraft systems and spectrum requirements to support their safe operation in non-segregated airspace,” Tech. Rep. M.2171, Dec. 2009.

SPAWC 2019 6


3GPP UAV Communication RequirementUAV communication requirement

Data Type Data Rate Reliability Latency

Downlink (DL: BS to UAV)

Command and control

60-100 Kbps

10−3 packet error rate

50 ms

Uplink (UL: UAV to BS)

Command and control

60-100 Kbps

10−3 packet error rate

--

Application data

Up to 50 Mbps

-- Similar toterrestrial user

3GPP TR 36.777: “Technical specification group radio access network: study on enhanced LTE support for aerial vehicles”, Dec. 2017.

SPAWC 2019 7


Existing Wireless Technologies for UAV Communications Wireless Technologies for UAV

Satellite-UAV communicationFlying ad-hoc network of UAVs

Direct UAV-ground communication in unlicensed spectrum (e.g. 2.4 GHz)

SPAWC 2019 8


New Paradigm: Cellular-Connected UAV Wireless Technologies for UAV

SPAWC 2019 9


Comparison of Wireless Technologies for UAV Wireless Technologies for UAV

SPAWC 2019 10

Technology Advantages Disadvantages

Direct link

• Simple• Low cost

• Limited range/data rate• Vulnerable to interference • Non-scalable for massive UAV

deployment

Satellite

• Global coverage • Costly • Heavy/bulky/energy

consuming equipment• High latency

Ad-hoc network• Robust and adaptable• Support for high

mobility

• Intermittent connectivity• Complex routing protocol• Low spectrum efficiency

Cellularnetwork

• Almost ubiquitousaccessibility

• Cost-effective• Superior performance

and scalability

• Unavailable in remote areas• Potential interference with

terrestrial communications


Future UAV Network: An Integrated ArchitectureWireless Technologies for UAV

Y. Zeng, Q. Wu, and R. Zhang, “Access from the Sky: a tutorial on UAV communications for 5G and beyond,’’ submitted to Proceedings of the IEEE, Invited Paper, available on arXiv

SPAWC 2019 11


New Paradigm: Integrating UAVs into Cellular Cellular-Connected UAV: UAV as new aerial user/UE in cellular network

UAV-Assisted Wireless Communication: UAV as new aerial platform (BS/AP, relay)

Integrating UAVs into 5G and Beyond

SPAWC 2019 12


Integrating UAVs into 5G: A Win-Win Technology

5G for UAVs: URLLC (with <20ms latency, >99.99% reliability): more secure CNPC eMBB (with 20 Gbps peak rate): real-time UHD video payload for VR/AR mMTC/D2D: UAV swarm communications and networking Cellular positioning (with cm accuracy): UAV localization/detection Massive MIMO: 3D coverage, aerial-terrestrial interference mitigation Edge-computing: UAV computing offloading, autonomous flight

UAVs for 5G: New business opportunities by incorporating new aerial users More robust and cost-effective cellular network with new aerial

communication platforms

Integrating UAVs into 5G and Beyond

SPAWC 2019 13


UAV Communications: What’s New over Terrestrial?Integrating UAVs into 5G and Beyond


SPAWC 2019 14

Characteristic Opportunities Challenges

High altitude • Wide ground coverage as aerial BS/relay

• Require 3D cellular coverage foraerial user

High LoSprobability

• Strong and reliable communication link

• High macro-diversity• Slow communication scheduling

and resource allocation

• Severe aerial-terrestrial interference

• Susceptible to terrestrial jamming/eavesdropping

High 3D mobility

• Traffic-adaptive movement• QoS-aware trajectory design

• Handover management• Wireless backhaul

Size, weight, and power (SWAP) constraint

• Limited payload and endurance• Energy-efficient design, wireless

charging• Compact and lightweight

antenna/RF design

Outline

Introduction

UAV Communication Fundamentals Channel model UAV energy consumption model Performance metric Mathematical formulation

Challenges and Promising Solutions

Conclusions


SPAWC 2019 15

UAV Channel Modelling: An OverviewRui Zhang, National University of SingaporeUAV Channel Model

UAV-UAV UAV-GroundBS-UAV UAV-terminal

Large-scale: Pathloss + shadowing

Free-space path loss model

Customized models for UAV-ground links

Small-scale: multipath fading

Usually negligible

Rician, Nakagami-m

Rayleigh, Rician, Nakagami-m

Generic narrowband (frequency-flat) fading channel model

SPAWC 2019 16

Large-scale gain Small-scale fading

( )[ ] ( ) [ ] [ ]10 010 logd dB d X dB X dBσβ α− = + +

Free-Space Path Loss ModelRui Zhang, National University of Singapore

Channel power inversely proportional to distance square

No shadowing or small-scale fading Applicable scenarios: rural area, sufficiently high UAV altitude Pros:

Simple, channel highly predictable Useful for offline UAV trajectory design

Cons: Oversimplified in urban environment Fails to model change of propagation environment at different UAV

locations/altitude

UAV Channel Model

SPAWC 2019 17

Altitude/Height-Dependent Channel ParametersRui Zhang, National University of Singapore

Channel modelling parameters are functions of UAV altitude/height Altitude-dependent path loss exponent, shadowing variance, Rician

factor,…

Applicable scenarios: urban/suburban area, medium-to-high altitude Pros:

Captures environmental variations as altitude changes Useful for performance analysis, off-line trajectory design

Cons: Fails to model the change of propagation condition with horizontal fly

UAV Channel Model

R. Amorim et al., “Radio channel modeling for UAV communication over cellular networks,” IEEE Wireless Commun. Lett., Aug. 2017.

SPAWC 2019 18

( ) ( )( )1 2 10max log ,2U UH p p Hα = −

Elevation Angle-Dependent Channel ParametersRui Zhang, National University of Singapore

Elevation angle-dependent path loss exponent, Rician factor,… Applicable scenarios: urban/suburban area, low-medium altitude Pros:

Captures environmental variations with 3D flying Useful for performance analysis, off-line trajectory design

UAV Channel Model

Cons: Further experimental

verification is required

M. M. Azari, F. Rosas, K.-C. Chen, and S. Pollin, “Ultra reliable UAV communication using altitude and cooperation diversity,” IEEE Trans. Commun., Jan. 2018.

SPAWC 2019 19

H1

d2D

LoS path

ϴ1 ϴ2

Reflected path

H2

Elevation Angle-Dependent Probabilistic LoS ModelRui Zhang, National University of Singapore

Separately model LoS and NLoS propagations LoS probability increases with elevation angle

Applicable scenarios: urban environment with statistical information of building height/distribution

Pros: Useful for performance analysis, off-line trajectory design Cons:

Model in statistical sense only, may fail in actual environment Needs experimental verification

UAV Channel Model

A. Al-Hourani, S. Kandeepan, and S. Lardner, “Optimal LAP altitude for maximum coverage,” IEEE Wireless Commun. Lett., Dec. 2014.

SPAWC 2019 20

( ) 0

0

, LoS enviroment , NLoS enviroment

dd

d

α

α

ββ

κβ

−

−

=

( ) ( )( )1

1 expLoSPa b a

θθ

=+ − −

3GPP BS-UAV ModelRui Zhang, National University of Singapore

Separately model LoS and NLoS propagations LoS probability and channel modelling parameters are functions of

UAV altitude and horizontal distance

Proposed scenarios: BS-UAV for Urban Macro (UMa), Urban Micro (UMi), and Rural Macro (RMa)

Pros: Comprehensive models for path loss, shadowing and small-scale fading Useful for numerical simulation

Cons: Too complicated for performance analysis and trajectory optimization

UAV Channel Model


SPAWC 2019 21

( ), , 1

, 2 1 2

2

1.5 m , , , ,

1, 300 m,

LoS ter U

LoS LoS U D U U

U

P H HP P d H H H H

H H

≤ ≤= ≤ ≤ ≤ ≤

Summary of UAV-Ground Channel Channels Rui Zhang, National University of SingaporeUAV Channel Model

SPAWC 2019 22

Channel model Description Pros and Cons

Free-space channelmodel

Channel power inversely proportionalto distance square, no shadowing orsmall-scale fading

Simple, useful for offline UAV trajectory design; oversimplified in urban environment

Altitude-dependentchannel parameters

Channel modelling parameters suchas path loss exponent and shadowingvariance are functions of UAV altitude

Useful for theoretical analysis and offline trajectory design; fails to model the change of propagation environment when UAV moves horizontally

Elevation angle-dependent channelParameters

Channel modelling parameters such as path loss exponent and Rician factor are functions of UAV elevation angle

Useful for theoretical analysis and offline trajectory design; further experimental verification required

Elevation angle dependent probabilistic LoS model

Separately model LoS and NLoSpropagations; LoS probabilityincreases with elevation angle

Useful for theoretical analysis and offline trajectory design; simplified shadowing; model in statistical sense only, may fail in actual environment

3GPP GBS-UAV channel model

Separately model LoS and NLoSpropagations; LoS probability andchannel modelling parameters areboth functions of altitude andhorizontal distance

Comprehensive models for path loss, shadowing and small-scale fading; useful for numerical simulations but too complicated for analysis or UAV trajectory optimization

UAV Energy Consumption ModelUAV Energy Model

Limited on-board energy: critical issue in UAV communication, for both UAV as user or BS/relay

UAV energy consumption: Propulsion energy >> Communication energy Empirical and Heuristic Models:

Empirical model based on measurement results, e.g., Fuel cost modelled by L1 norm of control force Fuel cost proportional to the square of speed

Analytical Model Closed-form model based on well-established results in aircraft literature Propulsion power as a function of speed and acceleration


Y. Zeng, J. Xu, and R. Zhang, “Energy minimization for wireless communication with rotary-wingUAV,” IEEE Trans. Wireless Commun., Apr. 2019.

Y. Zeng and R. Zhang, "Energy-Efficient UAV Communication with Trajectory Optimization," IEEE Trans. Wireless Commun., June 2017.

SPAWC 2019 23

Energy Model Comparison: Straight and level flightUAV Energy Model

Fixed-Wing Rotary-Wing

Convexity with respect to 𝑉𝑉 Convex Non-convex

Components Induced and parasite Induced, parasite, and blade profile

𝑉𝑉 = 0 Infinity Finite


SPAWC 2019 24

Fixed-Wing Rotary-Wing

Energy Model with General Level Flight (Fixed-Wing)UAV Energy Model

Change in kinetic energyWork required to overcome air resistance

Only depends on speed and centrifugal acceleration (causing heading change)

Independent of actual location or tangential acceleration (causing speed change)

Work-energy principle interpretation

𝒂𝒂(𝑡𝑡)

𝒗𝒗(𝑡𝑡)

𝒂𝒂⊥

(𝑡𝑡)

𝒂𝒂||(𝑡𝑡)

𝒂𝒂⊥𝟐𝟐 (𝑡𝑡)


SPAWC 2019 25

UAV Communication: Performance MetricPerformance Metric

Signal to interference-plus-noise ratio (SINR) Outage/coverage probability Communication throughput/delay Spectral/energy efficiency All dependent on UAV location/trajectory


SPAWC 2019 26

Desired signal

Interference

(a) UAV as a transmitter (b) UAV as a receiver

2ter aer

( )( )( )

kk

k

SI I

γσ−=

+ +qQQ 2

ter aer

( )( )( ) ( )

kk

k

SI I

γσ

=+ +

qq

QQ

Joint Trajectory-Communication Optimization: Generic Formulation

𝒒𝒒(𝑡𝑡): trajectory

𝒓𝒓(𝑡𝑡): commun. resource

U: utility functions, e.g., communication rate, SINR, coverage probability, spectrum/energy efficiency

fi: trajectory constraints, e.g., speed constraint, obstacle/collision avoidance gi: communication resource constraints, e.g., power, bandwidth hi: coupled constraints, e.g., interference temperature, minimum SINR

requirement

Mathematical Formulation Rui Zhang, National University of Singapore

SPAWC 2019 27

Practical Constraints on UAV Trajectory

Maximum/minimum altitude:

Initial/final status:

Maximum/minimum speed:

Maximum acceleration:

Obstacle avoidance:

Collision avoidance:

No-fly zone (cubic):

Mathematical Formulation Rui Zhang, National University of Singapore

SPAWC 2019 28

min 3 max[ ( )] ,H t H t≤ ≤ ∀q

(0) , ( )I FT= =q q q q

min max|| ( )|| , ,V t V t≤ ≤ ∀v

max|| ( )|| , ,t a t≤ ∀a

1|| ( ) || ,t D t− ≥ ∀q r

2|| ( ) ( )|| , ,k kt t D k k t′ ′− ≥ ∀ > ∀q q6

1

( ) ,Ti i

i

t b t=

≥ ∀a qU

Outline Introduction


Challenges and Promising Solutions UAV-assisted wireless communication (UAV as BS/relay) Basic models Recent results Joint trajectory and communication optimization Energy-efficient UAV communication

Cellular-connected UAV (UAV as user/terminal)

Conclusions


SPAWC 2019 29

UAV-Assisted Communications: Basic Models

Downlink

v

u1

s1

v1v2

s2

u2

Multi-UAV Interference Channel

Interference

S1 D1

s1s1

Relaying

u1 u2

s

v

Multicasting

u1 u2

s1s2

v

u1 u2

s1 s2

Uplink

v

UAV-Assisted Wireless Communications Rui Zhang, National University of Singapore

SPAWC 2019 30

Performance analysis Static UAV platform

• Deterministic modelling of UAV location• Stochastic modelling of UAV location: 2D/3D Poisson Point Process; Binomial

Point Process for small-size UAV network Flying UAV platform

• Deterministic modelling of UAV trajectory: circular/straight• Stochastic modelling of UAV trajectory: stochastic trajectory process

UAV placement No user location information (ULI): maximize covered area Perfect ULI: maximize number of covered users, communication throughput,

minimize number of UAVs given per-user requirement Partial ULI

Joint trajectory and communication optimization Exploiting new design DoF of UAV trajectory optimization

Energy-efficient UAV communication

Recent Results: An Overview UAV-Assisted Wireless Communications Rui Zhang, National University of Singapore

SPAWC 2019 31

Probabilistic LoS Channel model Large-scale channel power model for LoS and NLoS conditions

LoS probability:

Expected channel gain:

Exploiting UAV Mobility: How Much Can We Gain?

d2D

HU

V

𝜃𝜃

d

UAV flies towards the ground terminal

Double effects to improve the channel quality: Shorter link distance Less signal obstruction

𝜅𝜅 < 1: additional attenuation for NLoS

Trajectory and Communication Co-Design Rui Zhang, National University of Singapore

SPAWC 2019 32

( ) 0

0

, LoS Link , NLoS Link

dd

d

α

α

ββ

κβ

−

−

=

( ) ( )( )1

1 expLoSPa b a

θθ

=+ − −

( ) ( ) ( )( )0 01LoS LoSE d P d P dα αβ θ β θ κβ− −= + −

0 5 10 15 20 25 30 35 40 45 50

t (s)

-140

-135

-130

-125

-120

-115

-110

-105

-100

-95

Path

loss

(dB)

LoS

NLoS

Exploiting UAV Mobility: How Much Can We Gain?

0 5 10 15 20 25 30 35 40 45 50

t (s)

-140

-135

-130

-125

-120

-115

-110

-105

-100

-95

Aver

age

chan

nel p

ower

gai

n (d

B)

Channel gain for LoS and NLoSLoS probability

Expected channel gain

Initial distance d2D 1000 m

UAV altitude Hu 100 m

Flying speed v 20 m/s

Path loss exponent α 2.3

Reference channel gain β0

-50 dB

Probabilistic LoSmodel parameters

𝑎𝑎 = 10, 𝑏𝑏 = 0.6,𝜅𝜅 = 0.01

40 dB

23 dB


SPAWC 2019 33

0 5 10 15 20 25 30 35 40 45 50

t (s)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

LoS

prob

abilit

y

Exploiting UAV Mobility for Communication: Key Points

Moving UAV closer to ground terminals brings significant performance gain, beyond the conventional communication design

Main considerations: communication delay and UAV energy consumption

Useful techniques for trajectory and communication co-design Graph theory (e.g., travelling salesman problem, shortest-path

problem) Trajectory quantization (time/path discretization) Optimization techniques (block-coordinate descent, successive

convex approximation, machine learning, etc.)


SPAWC 2019 34

Path Planning: Travelling Salesman Problem

Travelling salesman problem (TSP): Given 𝐾𝐾 cities and the distances between each pair of cities, find the shortest route that visits each city and returns to the origin city

Complexity by exhaustive search: 𝐾𝐾! (NP hard) Many heuristic and optimal algorithms (up to tens of thousands of cities)

have been proposed

u1 u2

s1 s2

Uplink/downlink

v UAV path planning: For UAV-enabled

communications with ground users, determine the optimal flying path to serve them sequentially

Intuition: fly to each ground user as close as possible

s3

u3


SPAWC 2019 35

Variations of Travelling Salesman Problem

The standard TSP requires the traveler return to the origin city For UAV communications, the UAV may not necessarily return to the

location where it starts the mission, and the initial and/or final locations may be pre-specified

TSP Variation 1: No return TSP Variation 2: No return, specified initial and final locations 𝒒𝒒0 and 𝒒𝒒𝐹𝐹 TSP Variation 3: No return, specified initial location 𝒒𝒒0, any final location

Trajectory and Communication Co-Design

Y. Zeng, X. Xu, and R. Zhang, “Trajectory design for completion time minimization in UAV-enabled multicasting”, IEEE Trans. Wireless Commun., April 2018.


SPAWC 2019 36

TSP Variation: No Return

Standard TSP No Return


SPAWC 2019 37

No Return, any initial/final No Return, specified initial and final

TSP Variation: Specified Initial and Final Locations Trajectory and Communication Co-Design Rui Zhang, National University of Singapore

SPAWC 2019 38

Travelling Salesman Problem with Neighborhood When the total operation time 𝑇𝑇 is small, the UAV may not be able to visit

all users TSP with neighborhood (TSPN): Given 𝐾𝐾 cities and the neighborhoods of

each city, find the shortest route that visits each neighborhood once A generalization of TSP, also NP-hard

𝒫𝒫: set of all 𝐾𝐾! possible permutations


SPAWC 2019 39

( ) ( )

( ) ( )

1{ },{ ( )}min

s.t. 1 ,...,

,

kk kk k

k k k

k

r k

π ππ

π π

+ −

∈ − ≤ ∀

∑qq q

q w

P

Pickup-and-Delivery Problem (PDP) For UAV-enabled multi-pair relaying, determine the UAV flying path subject to Information-causality constraint: UAV needs to first receive data from a

source before forwarding to its destination Pickup-and-Delivery Problem (PDP): A generalization of TSP with precedence

constraints: for each source-destination pair, visit source before destination NP-hard, while algorithms for high-quality solutions exist PDP with neighborhood (PDPN)

S1 S2

s1

s2

Multipair relaying

vs1

D1

D2

s2


SPAWC 2019 40

UAV Path Planning with TSPN and PDPN

TSPN PDPN


J. Zhang, Y. Zeng, and R. Zhang, “UAV-enabled radio access network: multi-mode communication and trajectory design,” IEEE Trans. Signal Process., Oct. 2018.


SPAWC 2019 41

Limitations of TSP/PDP For Trajectory Optimization

Suboptimal trajectory in general: Straight flight between waypoints only, while optimal trajectory for

communication are curved in general Ignores various communication/trajectory constraints:

Rate requirement, interference, obstacle avoidance, maximum/minimum speed, no-fly zone….

Only gives UAV flying path, but trajectory optimization includes both path planning and speed optimization

A general framework: joint UAV trajectory and communication resource allocation optimization, by employing TSP/PDP-based path as initial trajectory Time/path discretization Optimization (block coordinate descent, successive convex

approximation, etc.)


SPAWC 2019 42

Joint Trajectory-Communication Optimization: Generic Formulation

𝒒𝒒(𝑡𝑡): trajectory

𝒓𝒓(𝑡𝑡): commun. resource

The continuous-time representation of trajectory involves infinite number of variables

Discretization is necessary for optimization and computation purposes Two discretization methods: time discretization and path discretization


SPAWC 2019 43

Time vs. Path Discretization Path discretization: generalized time discretization with variable slot length

Time Discretization Path Discretization

Pros • Equal time slot length• Linear state-space representation• Incorporate maximum

acceleration constraint easily

• Fewer variables if UAV hovers or flies slowly

• No need to know T a priori

Cons • Excessively large number of time slots when UAV moves slowly

• Needs to know T a priori

• More variables if UAV flies with high/maximum speed most of the time

0 𝑇𝑇 = 𝑁𝑁𝛿𝛿𝑡𝑡Time discretization: 𝑇𝑇 must known

𝛿𝛿𝑡𝑡 2𝛿𝛿𝑡𝑡

𝒒𝒒[1] 𝒒𝒒[2] 𝒒𝒒[𝑁𝑁]……

Path discretization:𝑇𝑇 can be unknown𝒒𝒒1 ……𝑇𝑇1 𝑇𝑇2

𝒒𝒒2 𝒒𝒒𝑀𝑀


𝑇𝑇 =∑𝑇𝑇𝑚𝑚


SPAWC 2019 44

Block Coordinate Descent

Time or path discretization converts the problem into a discrete form The (discrete) joint trajectory and resource optimization problems are

generally com-convex and difficult to solve Block coordinate descent: alternately update one block of variables (say,

trajectory) with the other (resource allocation) fixed. Monotonically converge to a locally optimal solution

Optimize 𝒒𝒒[𝑛𝑛]Optimize 𝒓𝒓[𝑛𝑛]

𝑙𝑙 = 𝑙𝑙 + 1

𝒒𝒒(𝑙𝑙)[𝑛𝑛] 𝒒𝒒(𝑙𝑙+1)[𝑛𝑛]𝒓𝒓(𝑙𝑙+1)[𝑛𝑛]


SPAWC 2019 45

Successive Convex Approximation Even with given resource allocation, UAV trajectory optimization is usually non-

convex, and thus difficult to solve Non-concave objective functions: e.g., rate maximization Non-convex constraints: e.g., obstacle/collision avoidance, minimum speed

Successive convex approximation (SCA): local optimization via solving a sequence of convex problems converges to a KKT solution if appropriate local bounds are found

• Convex optimization problem• Solution is feasible to the original

non-convex problem

Non-convex optimization problem

Global concave lower bound


SPAWC 2019 46

Successive Convex Approximation

Communication rate maximization:

Minimum speed constraint:

Convex optimization based

on lower bounds

Find global concave lower

bounds

𝑙𝑙 = 𝑙𝑙 + 1

𝐴𝐴𝑘𝑘,𝐵𝐵𝑘𝑘: poisitive coefficients depending on 𝒒𝒒(𝑙𝑙)[𝑛𝑛]

𝒒𝒒(𝑙𝑙)[𝑛𝑛] 𝒒𝒒(𝑙𝑙+1)[𝑛𝑛]


SPAWC 2019 47

( )0 ( )2log 1 || [ ] || [ ]

|| [ ]|| ||

||l

k k k kk

A B n nn α

γ + ≥ − − − − −

q w q wq w

( )2 ( ) 2 ( ) ( ) 2min|| [ ]|| || [ ]|| 2 [ ] [ ] [ ]l l T ln n n n n V≥ + − ≥v v v v v

Case Studies

Example 1: Multi-UAV Enabled Wireless Network

Example 2: Energy-Efficient UAV Communication

Q. Wu, Y. Zeng, and R. Zhang, “Joint trajectory and communication design for multi-UAV enabled wireless networks,” IEEE Trans. Wireless Commun., Mar. 2018.


Y. Zeng and R. Zhang, “Energy-Efficient UAV Communication with Trajectory Optimization,” IEEE Trans. Wireless Commun., June 2017.


SPAWC 2019 48

Multi-UAV Enabled Wireless Network Multi-UAV Enabled Wireless Network

Multi-UAV downlink communications with ground users TDMA for user communication scheduling Problem: maximize the minimum average rate of all users via joint

communication (scheduling, power control) and UAV trajectories optimization


SPAWC 2019 49

Problem FormulationMulti-UAV Enabled Wireless Network Rui Zhang, National University of Singapore

SPAWC 2019 50

Minimum rate requirement

UAV mobility constraint

TDMA constraints

power constraint

Initial/final location constraint

collision avoidance constraint

Nonconvex, solved by time-discretization and block coordinate descent

Simulation ResultsMulti-UAV Enabled Wireless Network

New Interference-mitigation approach: coordinated multi-UAV trajectory design


SPAWC 2019 51

Simulation Results: Throughput-Delay Tradeoff

Multi-UAV Enabled Wireless Network

Longer flight period achieves higher max-min throughput, but incurs larger user delay on average


SPAWC 2019 52

Energy-Efficient UAV CommunicationEnergy-Efficient UAV Communication

UAV energy consumption (fixed-wing):

Aggregate throughput as a function of UAV trajectory

Energy efficiency in bits/Joule:


SPAWC 2019 53

Energy Efficiency MaximizationEnergy-Efficient UAV Communication

Maximize energy efficiency in bits/Joule via trajectory optimization

Non-convex, solved by time discretization and successive convex approximation (SCA)

Initial/final location constraint

Min./Max. speed constraint

Initial/final velocity constraint

Max. acceleration constraint


SPAWC 2019 54

Simulation Results: Throughput-Energy Tradeoff Energy-Efficient UAV Communication

Rate-max trajectory: stay as close as possible with the ground terminal Energy-min trajectory: less acute turning EE-max trajectory: balance the two, “8” shape trajectory


SPAWC 2019 55

Fundamental Tradeoffs in UAV Trajectory and Communication Design

Throughput-Delay Tradeoff Throughput-Energy Tradeoff Delay-Energy Tradeoff

Q. Wu, L. Liu, and R. Zhang, “Fundamental tradeoffs in communication and trajectory design for UAV-enabled wireless network,” IEEE Wireless Communications, Feb. 2019.

Rui Zhang, National University of SingaporeTrajectory and Communication Co-Design

SPAWC 2019 56

Thr

ough

put

EnergyDelay

Thr

ough

put

Del

ay

Energy

Trajectory Optimization: More Comments

Trajectory optimization generally comprises offline and online phases: Offline: Based on given information on user location/channel

statistics/communication requirement, design UAV trajectory based on graph theory and optimization techniques

Online: Adjust UAV trajectory based on real-time measured channel by applying techniques such as radio mapping, reinforcement learning…

Useful techniques for complexity reduction Alternating Direction Method of Multipliers (ADMM) for

distributed/parallel computing Receding horizon (sliding-window) based optimization Machine learning techniques (such as deep learning)


SPAWC 2019 57

Outline Introduction


Challenges and Promising Solutions UAV-assisted wireless communication (UAV as BS/relay) Cellular-connected UAV (UAV as user/terminal) Main challenges 3GPP study Aerial-ground interference mitigation

Conclusions


SPAWC 2019 58

Rui Zhang, National University of SingaporeCellular-Connected UAV

Cellular-Connected UAV: Main Challenges High altitude

3D coverage is challenging: existing BS antennas tilted downwards High 3D mobility

Frequent handovers, cell selection Asymmetric downlink/uplink: ultra-reliable CNPC versus high-rate payload data Strong air-ground LoS dominant channel

Pro: High macro-diversity (beneficial for BS association) Con: Severe aerial-ground interference (in both uplink and downlink)

Mainly served by antenna side-lobe with current LTE BS

SPAWC 2019 59

Recent results by 3GPPCellular-Connected UAV

Interference detection Uplink interference mitigation

Uplink power control with altitude-dependent compensation factors 𝛼𝛼UE

Full-dimensional MIMO (3D beamforming) Directional antenna at UE

Downlink interference mitigation Intra-site joint transmission (CoMP) Enhanced initial access Coordinated data and control transmission

Mobility Refining handover parameters based on airborne status, path information…



SPAWC 2019 60

tx max 10 RB 0 UEmin{ ,10log ( ) TPL}P P M P α= + + ⋅

Rui Zhang, National University of SingaporeCellular-Connected UAV

Reshaping Cellular Networks to Cover the Sky

Possible spectrum allocation: Protected aviation spectrum licensed to cellular operators for UAV CNPC Cellular spectrum shared by aerial and terrestrial users for payload data

Interference management techniques: Aerial-ground NOMA 3D beamforming QoS-aware trajectory design

SPAWC 2019 61


Inter-Cell Interference Coordination (ICIC) with UAVs Cellular-Connected UAV

ICIC region is much larger than for terrestrial users due to LoS channels Dedicated orthogonal channels for UAVs: spectrum inefficient, especially when

terrestrial user density is high


SPAWC 2019 62

Aerial-Ground NOMA

W. Mei and R. Zhang, “Uplink cooperative NOMA for cellular-connected UAV,” IEEE J. Sel. Topics Sig. Process, 2019 (to appear).

Non-Orthogonal Multiple Access (NOMA) Non-cooperative NOMA: each co-channel BS employs local interference

cancelation only, inefficient due to large number of interfered BSs Cooperative NOMA: idle BSs decode UAV message first and forward to

adjacent co-channel BSs for interference cancelation

Cellular-Connected UAV Rui Zhang, National University of Singapore

SPAWC 2019 63

M: BS Cooperation Size

3D Beamforming BS array configuration: fixed pattern (downtilted) vs 3D beamforming

Y. Zeng, J. Lyu, and R. Zhang, “Cellular-connected UAV: potentials, challenges and promising technologies,” IEEE Wireless Communications, Feb. 2019.

Cellular-Connected UAV Rui Zhang, National University of Singapore

3GPP urban macro (UMa) scenario Total # of terrestrial and UAV users: 20

SPAWC 2019 64

QoS-Aware Trajectory DesignCellular-Connected UAV

Exploit UAV trajectory to improve communication performance (e.g., SINR), based on (statistical) knowledge of channel/interference spatial distribution

Equivalent to constrained shortest path problems in graph theory

S. Zhang, Y. Zeng, and R. Zhang, “Cellular-enabled UAV communication: a connectivity-constrained trajectory optimization perspective,” IEEE Trans. Commun. Mar. 2019 (Invited Paper)


BS coverage region 𝛾𝛾1 < 𝛾𝛾2

SPAWC 2019 65

Outline

Introduction


Challenges and Promising Solutions: UAV-assisted wireless communication (UAV as BS/relay) Cellular-connected UAV (UAV as user/terminal)

Conclusions


SPAWC 2019 66

ConclusionsRui Zhang, National University of Singapore

Integrating UAVs into 5G and beyond: a promising paradigm to embrace the new era of Internet-of-drones (IoD)

Cellular-Connected UAV: UAV as new aerial user/terminal

UAV-Assisted Wireless Communication: UAV as new mobile BS/relay/data collector

Main challenges: Air-ground channel modelling 3D coverage Joint trajectory and communication design Energy efficiency Aerial-ground interference management

Much more to be done…

Conclusions

SPAWC 2019 67

Rui Zhang, National University of SingaporeExtensions/Future Directions

Extensions/Future Directions

UAV-BS/UE channel modelling and experimental verification 3D network modelling and performance analysis General UAV energy model and energy-efficient communication Security issues in UAV communications Massive MIMO/mmWave for UAV communications Joint offline-online UAV trajectory design with QoS constraints Cellular-enabled UAV swarm communications UAV communications with limited wireless backhaul UAV meets wireless power/energy harvesting/caching/edge

computingMachine learning/AI for UAV communications and networking ……

SPAWC 2019 68

accessing from the sky: uav challenges from a ... · google, facebook, amazon, huawei , qualcomm,...

Documents