BookPDF Available

Analysis and Optimization for Robust Millimeter-Wave Communications

February 2021

February 2021

DOI:10.3384/diss.diva-172553

License
CC BY-NC 4.0

ISBN: 9789179297343

Authors:

Example of direct and reflected beams.

…

Uplink IRS-assisted mm-wave communications. The controller is connected to both the AP and the IRS in order to jointly optimize the beamforming at the AP and the IRS. The terms G, h d , and h r represent the IRS-AP, UE-AP, and UE-IRS channels.

…

Figures - available via license: Creative Commons Attribution-NonCommercial 4.0 International

Content may be subject to copyright.

Available via license: CC BY-NC 4.0

Content may be subject to copyright.

Analysis and Optimization

for Robust Millimeter-Wave

Communications

Linköping Studies in Science and Technology. Dissertations.

No. 2111

Cristian Tatino

Cristian Tatino Analysis and Optimization for Robust Millimeter-Wave Communications 2021

FACULT Y OF SCIENCE AND ENGINEERING

Linköping Studies in Scie nce and Technology, Dissertations. No. 2111, 2021

Department of Science and Technology

Linköping University

SE-581 83 Linköpin g, Sweden

www.liu.se

Link¨oping Studies in Science and Technology. Dissertations. No. 2111

Analysis and Optimization

for Robust Millimeter-Wave

Communications

Cristian Tatino

Department of Science and Technology

Link¨oping University, SE-601 74 Norrk¨oping, Sweden

Norrk¨oping 2021

Analysis and Optimization for Robust Millimeter-Wave Communica-

tions

Cristian Tatino

Link¨oping Studies in Science and Technology. Dissertations. No. 2111

2021 Cristian Tatino, unless otherwise noted

isbn 978-91-7929-734-3

issn 0345–7524

Link¨oping University

Department of Science and Technology

SE-601 74 Norrk¨oping

Printed by LiU Tryck, Link¨oping, Sweden 2021

This work is licensed under a Creative Commons Attribution-

NonCommercial 4.0 International License.

https://creativecommons.org/licenses/by-nc/4.0/

To my family

[IT] ”fatti non foste a viver come bruti, ma per seguir virtute e

canoscenza”.

[EN] ”you were not made to live your lives as brutes, but to be

followers of worth and knowledge”.

Dante Alighieri, La Divina Commedia, Inf.26.118-20.

Abstract

Spectrum scarcity is a longstanding problem in mobile telecommuni-

cations networks. Speciﬁcally, accommodating the ever-growing data

rate and communications demand in the extensively used spectrum

between 800 MHz and 6 GHz is becoming more challenging. For

this reason, in the last years, communications in the millimeter-wave

(mm-wave) frequency range (30-300 GHz) have attracted the interest

of many researchers, who consider mm-wave communications a key

enabler for upcoming generations of mobile communications, i.e., 5G

and 6G. However, the signal propagation in the mm-wave frequency

range is subject to more challenging conditions. High path loss and

penetration loss may lead to short-range communications and fre-

quent transmission interruptions when the signal path between the

transmitter and the receiver is blocked.

In this dissertation, we analyze and optimize techniques that en-

hance the robustness and reliability of mm-wave communications. In

the ﬁrst part, we focus on approaches that allow user equipment (UE)

to establish and maintain connections with multiple access points

(APs) or relays, i.e., multi-connectivity (MC) and relaying techniques,

to increase link failure robustness. In such scenarios, an ineﬃcient

link scheduling, i.e., over or under-provisioning of connections, can

lead to either high interference and energy consumption or unsatis-

ﬁed user’s quality of service (QoS) requirements. In the ﬁrst paper,

we propose a novel link scheduling algorithm for network throughput

maximization with constrained resources and quantify the potential

gain of MC. As a complementary approach, in the second paper, we

solve the problem of minimizing allocated resources while satisfying

users’ QoS requirements for mm-wave MC scenarios. To deal with

the channel uncertainty and abrupt blockages, we propose a learning-

based solution, of which the results highlight the tradeoﬀ between

reliability and allocated resource.

In the third paper, we perform throughput and delay analysis of

a multi-user mm-wave wireless network assisted by a relay. We show

the beneﬁts of cooperative networking and the eﬀects of directional

communications on relay-aided mm-wave communications. These, as

highlighted by the results, are characterized by a tradeoﬀ between

throughput and delay and are highly aﬀected by the beam alignment

duration and transmission strategy (directional or broadcast).

The second part of this dissertation focuses on problems related to

mm-wave communications in industrial scenarios, where robots and

new industrial applications require high data rates, and stringent re-

liability and latency requirements. In the fourth paper, we consider

a multi-AP mm-wave wireless network covering an industrial plant

where multiple moving robots need to be connected. We show how

the joint optimization of robots’ paths and the robot-AP associations

can increase mm-wave robustness by decreasing the number of han-

dovers and avoiding coverage holes. Finally, the ﬁfth paper considers

scenarios where robot-AP communications are assisted by an intel-

ligent reﬂective surface (IRS). We show that the joint optimization

of beamforming and trajectory of the robot can minimize the motion

energy consumption while satisfying time and communication QoS

constraints. Moreover, the proposed solution exploits a radio map to

prevent collisions with obstacles and to increase mm-wave communi-

cation robustness by avoiding poorly covered areas.

Popul¨

arvetenskaplig

sammanfattning

F¨or telekommunikationsoperat¨orer ¨ar det avg¨orande att ha tillg˚ang

till frekvenser f¨or att kunna ¨overf¨ora information, och det ¨ar en st¨andig

konkurrens om att f˚a anv¨anda nya frekvensband. Fram till den fj¨arde

generationens mobilkommunikation (4G) har st¨orre delen av infor-

mations¨overf¨oringen utf¨orts genom att skicka signaler med frekvenser

mellan 800 MHz och 2 GHz. Framtida kommunikationsbehov kr¨aver

att ¨aven andra, tidigare outnyttjade, frekvenser kan anv¨andas. Med

anledning av detta har kommunikation med millimeterv˚ag (mm-v˚ag)

intresserat m˚anga forskare, d˚a dessa mm-v˚agor anses vara av stor be-

tydelse f¨or att kunna f¨orb¨attra och utveckla nya s¨att att kommunicera

- det h¨ar ber¨or t.ex. utvecklingen av 5G och 6G. En utmaning med

mm-v˚agor, som inneb¨ar att relativt h¨oga frekvenser anv¨ands (30 GHz

- 300 GHz), ¨ar dock att kommunikationsf¨oruts¨attningarna blir mer

besv¨arliga och signalerna n˚ar inte lika l˚angt som om l¨agre frekvenser

nyttjas.

Den f¨orsta delen av den h¨ar avhandlingen fokuserar p˚a att f¨orb¨attra

tillf¨orlitligheten och r¨ackvidden f¨or mm-v˚agskommunikation. Detta

¨ar m¨ojligt genom att till¨ampa tekniker som ger f¨oruts¨attningar f¨or

ﬂera samtidiga anslutningar samt att signaler kan vidarebefordras p˚a

ett l¨ampligt s¨att inom ett kommunikationsn¨atverk. Parallella anslut-

ningar och vidarebefordring ¨okar antalet kommunikationsv¨agar som

kan anv¨andas och minskar risken f¨or att information inte n˚ar fram

till avsedd mottagare. F¨or ett tillfredsst¨allande resultat kr¨avs dock

l¨ampliga principer f¨or hur man nyttjar olika kommunikationsv¨agar.

Ol¨amplig hantering kan t.ex. resultera i att resurser anv¨ands p˚a ett

ineﬀektivt s¨att eller att kommunikationen upplevs vara av bristande

kvalitet. I avhandlingen beskrivs den potential som ﬁnns med att

vii

nyttja ﬂera parallella kommunikationsv¨agar och ett tillv¨agag˚angss¨att,

baserat p˚a maskininl¨arning, f¨oresl˚as d¨ar avsikten ¨ar att med ett min-

imalt nyttjande av resurser tillhandah˚alla tillr¨acklig prestanda f¨or ly-

ckad kommunikation. Dessutom presenteras f¨ordelar som kan erh˚allas

om kommunicerande noder i ett n¨atverk samarbetar eﬀektivt, f¨or att

d¨arigenom uppn˚a en prestanda som ej ¨ar m¨ojlig utan samarbete.

Det ﬁnns m˚anga kommunikationsutmaningar med koppling till in-

dustriella till¨ampningar, och detta ¨ar n˚agot som avhandlingen ocks˚a

behandlar. Kommunikation till och fr˚an exempelvis robotar kan st¨alla

h¨oga krav p˚a snabb, omfattande och tillf¨orlitlig informations¨overf¨oring.

I avhandlingen f¨oresl˚as en algoritm som har f¨or avsikt att m¨ojligg¨ora

eﬀektiv kommunikation i en industriell anl¨aggning d¨ar mobila rob-

otar f¨ardas och mm-v˚agor anv¨ands f¨or att ansluta till och kommu-

nicera med ﬂertalet omkringliggande noder. Avhandlingen beskriver

ocks˚a hur robotar p˚a ett eﬀektivt s¨att kan navigera och kommunicera

i milj¨oer med f¨arre noder, d¨ar kommunikationsavst˚andet kan vara

f¨orh˚allandevis l˚angt. Det tillv¨agag˚angss¨att som f¨oresl˚as i detta fall ¨ar

att nyttja intelligenta ytor som mm-v˚agor kan reﬂekteras mot f¨or att

d¨arigenom f¨ardas l¨angre.

Acknowledgments

First and foremost, I would like to thank my supervisor, Prof. Di Yuan

for providing me with the opportunity to undertake my studies at

Link¨oping University. I really appreciate the eﬀort and the time that

he dedicates to my Ph.D. studies. His experience and knowledge have

encouraged me all the time of my academic research and daily life.

I would like to thank also my co-supervisor at Link¨oping University,

Associate Professor Nikolaos Pappas, for his comments and ideas that

are generously shared with me. His expertise, encouragements, and

enthusiasm have been fundamental for my studies and research.

I am deeply grateful to my supervisors at Nokia Bell Labs in

Stuttgart, Dr. Ilaria Malanchini and Dr. Lutz Ewe, for their unwa-

vering support and guidance. Moreover, they provided me the oppor-

tunity to work in one of the most important research laboratories in

the information technology ﬁeld. I would like to thank also Dr. Danish

Aziz, who has been my supervisor at Nokia Bell Labs for a temporary

period, and Prof. Anthony Ephremides, who gave me important sug-

gestions and the opportunity to conduct my research at the University

of Maryland for two months.

I would like to express my gratitude to all the members of the

project ACT5G and the funding sources of my Ph.D. work; my re-

search was supported by the European Union’s Horizon 2020 research

and innovation programme under the Marie Sklodowska-Curie grant

agreement No. 643002 (ACT5G). I would like to extend my sin-

cere thanks to Dr. David Gundleg˚ard and Dr. Erik Bergfeldt for

their insightful comments and suggestions for improving the quality

of this dissertation.

Special thanks to Tania, for always believing in me and her tremen-

dous support, and all the people that made the time that I have spent

in Norrk¨oping and Stuttgart enjoyable. Many thanks in particular to

Antzela, Eleni, Manos, Maria, Nikos, Tobias, Wenjian, Xenos, Yannis

A., Yannis P., Francesco, Arianna, Silvio, Alessandro, and Paolo. I

would like to thank all my friends in Napoli, especially Luca, Lorenzo,

Marco, and Roberto for their lifelong friendship, and Domenico, Luca,

Pasquale, Raﬀaele, and Silvio (Polo) for the inspiring discussions as

well as joyful distractions. Moreover, sincere thanks to Ciro and Anna

for their optimism and aﬀection.

Above all, I would like to express my gratitude to my parents,

Rocco and Rosa, brothers, Dino and Filippo, and my sister, Letizia,

for their unconditional love and constant encouragement. I wouldn’t

be where I am today without them.

Norrk¨oping, February 2021

Cristian Tatino

Abbreviations

3GPP 3rd Generation Partnership Project

4G fourth generation of mobile networks

5G ﬁfth generation of mobile networks

AF amplify and forward

AP access point

BIP binary integer programming

CA carrier aggregation

CG column generation

CoMP coordinated mutlipoint

D2D device-to-device

DC dual connectivity

DF decode and forward

FCC federal communications commission

FD full-duplex

HD half-duplex

IP integer programming

ILP integer linear programming

IRS intelligent reﬂective surface

ITU international telecommunication union

LB-BS low-band base station

LP linear programming

LTE long term evolution

LOS line-of-sight

mm-wave millimeter-wave

mmAP millimeter-wave access point

MC multi-connectivity

ML Machine learning

MIP mixed integer programming

NC network controller

NR 5G new radio

NLOS non line-of-sight

QoS quality of service

Rx receiver

SC single connectivity

SCO successive convex optimization

SINR signal-to-interference-plus-noise ratio

SNR signal-to-noise ratio

Tx transmitter

UAV unmanned aerial vehicle

UE user equipment

VAR virtual and augmented reality

WLAN wireless local area network

WPAN wireless personal area network

xii

Contents

Abstract v

Popul¨

arvetenskaplig sammanfattning vii

Acknowledgments ix

Abbreviations xi

I Introduction and Overview 1

1 Introduction 3

1.1 Millimeter-Wave Communications Background 3

1.2 Purpose and Scope 7

1.3 Thesis Outline and Organization 8

2 Robust Connectivity in Millimeter-Waves 9

2.1 Multi-connectivity 11

2.2 Relaying Technology 13

2.3 Intelligent Reﬂective Surfaces 15

2.4 Mobility Management and Association Opti-

mization 16

3 Mathematical Modeling and Solution Tools 19

3.1 Mathematical Optimization 19

3.2 Machine Learning 23

3.3 Queuing Theory 25

4 Contributions of the Thesis 29

4.1 Papers Included in the Thesis 31

xiii

Contents

4.2 Papers not Included in the Thesis 36

4.3 Conclusions and Future Research 36

Bibliography 39

II Papers 53

Paper I 57

Paper II 77

Paper III 97

Paper IV 147

Paper V 165

xiv

Part I

Introduction and Overview

Chapter 1

Introduction

1.1 Millimeter-Wave Communications Background

Wireless communications technologies have already changed the way

to communicate and interact for billions of people. According to [1],

over 70% of the global population will have mobile connectivity by

2023. Even more signiﬁcant growth is expected by machine-type

connections and autonomous vehicles, i.e., intelligent transportation

systems, drones, and robots that will represent half of the globally

connected devices by 2023. The tremendous increase of connections

will be followed by the development of new data-hungry applica-

tions, e.g., virtual and augmented reality (VAR), that will generate

a huge amount of traﬃc. This is estimated to grow at an annual

rate of around 55 % between 2020 and 2030 reaching the incredible

amount of 5,000 exabytes (EX) of data traﬃc [2]. Mainly three ap-

proaches are driving the research to accommodate this ever-growing

connection and traﬃc demand: i) increase the number of antennas

at the transmitter (Tx) and the receiver (Rx), ii) densiﬁcation of

radio base stations, and iii) increase the spectrum resources by us-

ing new frequency bands. The latter solution identiﬁes millimeter-

wave (mm-wave) communications as a key technology for the next

generations of mobile communications. Speciﬁcally, mm-wave com-

munications exploit the huge amount of frequency bands from 30 GHz

to 300 GHz1to meet the rising demand for connections and traﬃc.

However, as explained in more detail below, the signal propagation at

1Note that also the 24-30 GHz band is usually considered in the mm-wave

frequency range.

1 Introduction and Overview

!"#$%$"&'()"#$("*+'"$,("**-."&'$(&-(&/'(0-)%*'(1'+2%.'(3456789!(:;<

!"#$%$"&'()"#$(#-&("* *-."&'$(&-(&/'(0-)%*'(1'+2%.'(3456789!(:;<

=>.&%-#'$()"#$(),(?!!@(AB:C

!"#$% &''"#$%('"#$%!'"#$%

)*+,-+./0"

1#$%2

345+6+.789

1:+8+*2

!"!" !"!#

;<66<:+8+*=>45+"?@+/8*-:"A!'=!''B"#$%

!"#!$

ADE;

F:EG

FFEH

HBE;

HAE;

HFE;

"$#$

;BEA

;BEH

;AEI

Figure 1.1: Mm-wave candidate bands to the mobile service

by the international telecommunication union (ITU) [3], and

mm-wave spectrum auctioned by the federal communications

commission (FCC) [4].

mm-wave frequency ranges has some peculiarities that decrease the

reliability of communications, especially for scenarios that are char-

acterized by mobility and numerous physical obstacles.

Millimiter-wave Frequency Bands and Propagation Characteristics

Several frequency bands have been identiﬁed for accommodating mm-

wave mobile communications, e.g., 24.25-27.5 GHz, 37-40.5 GHz,

47.2-50.2 GHz, 50.4-52.6 GHz, 66-76 GHz, and 81-86 GHz [3]. As

shown in Fig. 1.1, some countries have already performed public auc-

tions for allocating part of them [5], whereas the 60 GHz band is

planned to be assigned to unlicensed mm-wave communications [6].

Measurements and tests at these frequency bands have shown that

mm-wave communications can reach several Gbps of data rate [7, 8].

However, free space path loss and atmosphere absorption dramati-

cally increase at such high frequencies, leading to shorter communi-

cation range [9, 10]. Moreover, mm-waves are characterized by high

penetration loss that leads to frequent interruptions when an obsta-

cle blocks the signal paths (blockage) [11, 12]. In general, the line-

of-sight (LOS), or direct path, presents the highest channel gain in

comparison to the reﬂected paths, whose channel gains could be 30

dB worse. However, good reﬂectors, e.g., metal smooth objects and

tinted glasses, can lead to signal paths with higher gains [9].

I Introduction

!"#$%!

!"#$%! " #

!"#$%"

!"#$%$ " #

)(

!"#$%*+,-.$"./

)(

0,1"2/,3.#+%45+,.6%)1#.7$,77,3.

45+,.6%5#26"/

Figure 1.2: Beam scanning for the alignment phase with a pre-

deﬁned set of beams (codebook) followed by directional uplink

transmissions. Each beam is labelled with an ID.

Directional Communications and Beamforming

Directional communications can be used to partially solve the high

path loss for mm-waves. Speciﬁcally, the small wavelength allows con-

structing arrays of antennas with many elements in a limited space

that can focus the energy in narrow beams towards adjustable di-

rections. Although narrow beams provide high beamforming gain

and can decrease the interference level [13], they are more sensitive

to communication interruptions caused by blockages. Moreover, nar-

row beams introduce further complexity and delay when a connection

needs to be established or recovered. Namely, to reach enough beam-

forming gain, both the receiver and the transmitter need to estimate

the channel during the beam alignment phase. The performance of

this procedure in terms of energy consumption, achieved beamforming

gain, complexity, and duration, highly depends on which beamform-

ing technique is used, i.e., digital, analog, or hybrid [14, 15]. Digital

beamforming supports multi-stream transmissions, but each antenna

is equipped with a radio frequency chain and channel between ev-

ery pair of antenna elements at the transmitter and receiver must be

estimated [15]. Although digital beamforming provides higher beam-

forming gains than analog beamforming, the former is characterized

by higher energy consumption and cost than the latter, which sup-

ports single-stream transmissions. Speciﬁcally, in analog beamform-

ing, all the antennas usually share the same radio frequency chain

and the beam alignment phase involves a simpler procedure. As

shown in Fig. 1.2, the transmitter and receiver search for the best

Tx-Rx beam pair that maximizes the gain by sending pilots in a pre-

deﬁned set of directions (codebook) by adjusting the phase shifters for

each antenna element [16, 17]. Finally, hybrid beamforming combines

1 Introduction and Overview

analog and digital beamforming techniques by grouping multiple an-

tenna elements into subarray modules to balance performance and

cost tradeoﬀs. In general, in digital, hybrid, and analog beamform-

ing, a beam alignment and a channel estimation involve transmissions

of a sequence of pilot signals that consume power, time, and frequency

resources.

Initial Access and Mobility Management

Initial access and mobility management are fundamental wireless net-

work functionalities. The former aims to establish an initial connec-

tion between a user equipment (UE) and a cell or access point (AP),

whereas the latter aims to trace UEs while changing their locations to

provide continuous connectivity. At the initial access of a UE, several

procedures must be performed, e.g., cell detection, access request, and

channel estimation [18, 19, 20]. However, the use of narrow beams

and the high path loss increase the complexity of initial access and

mobility management procedures with respect to lower frequencies

communications. Speciﬁcally, an omnidirectional cell detection may

not be an appropriate solution for mm-wave communications, and

the mobility of UEs can rapidly lead to beam misalignment [21] that

dramatically reduce the beamforming gains of communications. Thus,

channel estimations must be frequently repeated due to the UE move-

ment. Moreover, the blockage sensitivity and high path loss cause

a spottier coverage. This requires a dense deployment of mm-wave

cells and APs. In such scenarios, the maximum signal-to-noise ra-

tio (SNR) criteria for selecting the cell and AP association can lead

to frequent handovers and costly beam alignment procedures [22].

Thus, optimized mobility management and association would lead to

a signiﬁcant enhancement of network performance at millimeter-wave

communications.

Applications for Millimiter-wave Communications

For all these reasons, mm-wave wireless networks were initially stud-

ied for scenarios and applications that are characterized by short-

range communications, low mobility, and LOS between the receivers

and the transmitters. This view has led to technologies for provid-

ing access to wireless local area network (WLAN) and wireless per-

sonal area network (WPAN), such as the standards 802.11ad [23]

and 802.15.3 [24], both operating in the 60 GHz band. Mm-waves

I Introduction

are also good candidates to provide access to networks in scenarios

that are characterized by high data and density connection demand,

e.g., data centers, malls, airports, and stadiums, for which intensive

measurement campaigns have been performed [25, 26]. Moreover, the

possibility to transmit a huge amount of data and directional commu-

nications make mm-waves ideal solutions for the deployment of low-

cost wireless backhaulings [27, 28]. Directional communications have

been also exploited to develop localization technologies and radars

that work in mm-wave frequency bands [29, 30, 31].

The huge amount of bandwidth at mm-wave frequency bands

is leading academic and industrial researchers to study and analyze

mm-wave technology solutions also for other scenarios and applica-

tions that are characterized by mobility and non line-of-sight (NLOS)

conditions, e.g., cellular access and industrial scenarios. Speciﬁcally,

5G new radio (NR) operating at mm-waves are under standardization

by 3rd Generation Partnership Project (3GPP) and have been already

tested successfully for the cellular access [32]. Moreover, mm-wave

communications for industrial scenarios have recently attracted the

interest of researchers [33, 34, 35, 36]. More precisely, real-time indus-

trial applications, such as VAR for assisted manufacturing or mining,

may require Gbps for peak data rates [37] that can be accommodated

in the mm-wave frequency bands.

However, both industrial and outdoor cellular access applications

require reliable and robust connections in dynamic environments that

are characterized by high mobility and obstacle density. Indeed, the

analysis and optimization of solutions for increasing the robustness

in outdoor and industrial mm-wave wireless networks are the main

topics of the ﬁve papers presented in this thesis.

1.2 Purpose and Scope

The aim of the thesis is to analyze and optimize techniques that

enhance the robustness and reliability of mm-wave communications.

This would allow the industry to exploit a huge amount of spectrum

also for dynamic scenarios and applications with stringent quality of

service (QoS) requirements. We focus on solutions that optimize mo-

bility management or simultaneously provide multiple connections to

UEs, i.e., multi-connectivity (MC), relaying, and intelligent reﬂec-

tive surface (IRS)s. With the above-mentioned goal, the following

research questions are stated:

2 Robust Connectivity in Millimeter-Waves

Q1 What are the gains of MC in terms of throughput and relia-

bility?

Q2 What are the issues, the main tradeoﬀs, and costs that char-

acterize MC and relaying techniques?

Q3 What are the gains of cooperative relaying in terms of through-

put and reliability for mm-wave wireless networks?

Q4 What are the key system parameters and mm-wave propaga-

tion characteristics that mostly aﬀect the performance of these

techniques?

Q5 How do optimized association and mobility management for

mm-wave communications improve the robustness and aﬀect

robot trajectories in industrial scenarios?

Q6 How do optimized robot trajectories improve the robustness

for mm-wave communications in industrial scenarios?

Q7 Finally, how does improving radio coverage to increase robust-

ness enhance the robot energy eﬃciency in mm-wave industrial

scenarios?

1.3 Thesis Outline and Organization

The thesis is divided into two parts. In Part I, we provide a gen-

eral introduction to the concepts and problems of MC, cooperative

relaying, IRSs, and mobility management for mm-wave communica-

tions. In Chapter 3, we introduce the mathematical tools we have

used to approach the proposed problems. In Chapter 4, we describe

the contributions of the ﬁve research papers that form Part II.

Chapter 2

Robust Connectivity in

Millimeter-Waves

Robustness is the property of a system to withstand stresses, pertur-

bations, and unpredictable variations in its operating environment

without loss of functionality. Robustness is deﬁned with respect to

particular events that can disturb the operative conditions and is re-

lated to the design of the system. Otherwise speciﬁed, in this thesis

we consider the robustness of mm-wave communications with respect

to a connection (or link) failure. Another term that is often used in

this thesis is reliability. This is a quantitative value that measures the

success probability of a system in providing a service under certain

quality constraints. Reliability is usually measured in percentage (%),

e.g., for the ﬁfth generation of mobile networks (5G), ultra reliable

communications require the 99.999 % of successfully received packets.

Reliability and robustness are diﬀerent concepts, however, in many

cases, by increasing the latter it is possible to enhance the former.

As introduced in Chapter 1, the peculiarities of mm-waves make

the communications in this frequency range less reliable than lower

frequencies and less robust with respect to blockages. Even a human

body can block transmission by decreasing the received power by 30

dB [38]. However, the huge amount of frequency resources represent

a big opportunity for telecom operators to provide high throughput

transmissions and increase the number of served UEs. Therefore,

researchers have proposed several solutions to increase robustness and

reliability for mm-wave communications. This thesis analyzes the

2 Robust Connectivity in Millimeter-Waves

performance gains and provides solutions to distinctive optimization

problems that characterize some of these techniques.

Most of these above-mentioned techniques try to increase the num-

ber of available links and alternative signal paths between Txs and

Rxs. This can be combined with mechanisms that allow Tx-Rx pairs

to fast switch link when the one that is used for the communication

is interrupted. Communication interruptions may occur for several

reasons, such as blockages, a beam misalignment that drastically de-

creases the beamforming gain due to mobility, and increasing Tx-Rx

distance that causes a dramatic drop of the received power. Measure-

ments and analysis have shown a dramatic performance gain when

spatial diversity is used to enhance coverage in mm-wave communi-

cations [39, 40, 41]. By combining a proper coverage plan [42] and

mechanisms for providing connections from multiple APs/cells, e.g.,

MC and relaying techniques, we can signiﬁcantly improve throughput,

reliability, and robustness in mm-wave communications [43, 44].

Another solution consists of rapidly re-establishing the communi-

cation in case of blockage by using alternative Tx-Rx signal paths,

as proposed by [45]. More precisely, Tx and Rx nodes store several

beam pairs, which are evaluated during the beam alignment phase.

Then, if the beam pair that is used for the communications experi-

ences a blockage, the Tx and Rx can quickly switch to alternative

beams that exploit other signal paths. Note that the strongest path

is represented by the LOS one, and, when this is blocked, alternative

paths are provided by reﬂections. An example of beams that exploit

direct and reﬂected paths, which are transmitted by a millimeter-

wave access point (mmAP), is shown in Fig. 2.1. Contributions to

the coverage provided by reﬂections and reﬂected beams are analyzed

in [46, 47, 48].

As shown in [46], in some particular scenarios, the contributions

of the reﬂected beams are not negligible and they can be used as

alternative paths when the LOS one is blocked. However, in many

cases, connections that are established mainly using reﬂected paths

are unstable and unreliable depending on the obstacles’ positions,

sizes, orientation, and materials. In this regard, IRSs (also referred

to as reconﬁgurable intelligent surfaces or metasurfaces) are attract-

ing the interest of research as a breakthrough technology. Speciﬁ-

cally, IRSs can eﬃciently control the reﬂections of the incident sig-

nals such that these can be either added coherently or destructively to

Rx nodes for increasing the received power or suppress interference,

I Introduction

!"#$%&'($)*

+$,-$%&$.'($)*

**12

+$,-$%&3#

Figure 2.1: Example of direct and reﬂected beams.

respectively [49]. In the following sections, we will discuss in more

detail the techniques, i.e., MC, relaying technology, and IRSs, that

are related to our work. Finally, in the last section of this chapter,

we will show how mobility management and association optimization

can improve connectivity robustness and reliability by decreasing the

number of handovers and avoiding coverage holes. These solutions are

particularly eﬀective when applied to autonomous vehicles for which

we can control the mobility.

2.1 Multi-connectivity

Multi-connectivity (MC) is a technology that allows a UE to establish

and maintain multiple connections at the same time. Together with a

proper coverage planning, MC can increase mm-wave communication

robustness. Multiple connections can be established either by using

several resources of the same node (intra-site), e.g., diﬀerent frequency

bands, or with multiple cells/APs (inter-site). The latter exploits

spatial macro-diversity to provide independent links [50, 39, 40, 43].

We can also distinguish between inter-frequency and intra-frequency

MC, depending on whether the multiple connections use the same fre-

quency band or not. These solutions can be combined, as in Fig. 2.2,

which shows an example of an inter-site and inter-frequency MC sce-

nario where the UE is connected to two mmAPs and one low-band

base station (LB-BS). Moreover, MC can be either throughput or reli-

ability oriented. More precisely, according to [51], MC is characterized

by a tradeoﬀ between reliability and throughput that are improved

2 Robust Connectivity in Millimeter-Waves

!"#$ %&'() *$"*+

,-.*'(*/%0#1/"+2

!!"#$ !!" #%

&'(')

(

!"#$%&%

!"#$%'%

!"#$%(%

% #

% ' % #

!"#$%&%

!"#$%'%

!"#$%(%

!!"#$ !!" #%

&'(')

Figure 2.2: Inter-site and Inter-frequency multi-connectivity

scenario with two mmAPs and one LB-BS. These are connected

to a centralized network controller (NC) co-located with the LB-

BS.

by increasing multiplexing and diversity gain of the communication,

respectively. The main purpose of the throughput oriented MC is to

increase the data rate by transmitting diﬀerent data on several estab-

lished connections. These, in a reliability oriented approach, are used

by Tx nodes to send the same data that are combined at the Rx node

by using several possible combining techniques (e.g., maximum ratio

combining and selection combining).

3GPP has deﬁned for the fourth generation of mobile networks

(4G) several versions of MC at diﬀerent layers, e.g., dual connec-

tivity (DC), carrier aggregation (CA), and coordinated mutlipoint

(CoMP) [52]. For 5G, the technical speciﬁcation [53] deﬁnes several

possible solutions and architectures for both stand-alone and non-

standalone MC scenarios. In the former case, 5G NRs are assisted

by legacy standards, e.g., long term evolution (LTE), whereas, for

stand-alone architectures, MC is performed by using only 5G NRs.

Several studies, e.g., [54, 55, 56], have performed analysis of MC in

mm-wave scenarios and have demonstrated the performance gain in

terms of coverage, throughput, and reliability. Mm-waves can also be

assisted by lower frequencies communications. A similar solution is

proposed in [57] that considers a non-standalone mm-wave MC, with

mmAPs and lower frequency technologies, e.g., LTE, that act as the

data and the control planes, respectively. This is a common approach

I Introduction

for 5G systems and it is used also in [58] to improve and speed up

the initial access and the beamforming alignment. Indeed, the latter

represents a serious challenge for MC in mm-waves, where the UE

needs to establish multiple connections and perform beam tracking

with several mmAPs. The number of connections that can be estab-

lished, the beamforming gain, and the duration of the initial access,

highly depends on the beamforming techniques (i.e., digital, hybrid,

or analog), the number of antennas, and radio frequency chains with

which a node is equipped. These important aspects of the MC for

mm-waves are analyzed in [20, 59]. In this regard, machine learn-

ing techniques can be used to reduce the complexity of the channel

estimation and beam alignment with several mmAPs [60].

Another important aspect, on which we will focus in the ﬁrst two

papers of the thesis, is link scheduling. In an environment where a UE

is in the coverage of several mmAPs, we need to decide which and how

many connections the UE should use for the communications. The

number of links and the selection of them can tremendously impact

the network performance. In general, by serving a UE with multiple

connections, we can increase the throughput and the reliability (by

increasing either the diversity or multiplexing gain), but at the same

time, the interference level and the consumed resources are increased.

Resources here include frequency band, power budget, backhaul ca-

pacity, antennas, and radio frequency chains. Thus, we can observe

that there is a clear tradeoﬀ between the number of served UEs and

the achieved throughput or reliability, or in general, between con-

sumed resources and achieved QoS.

2.2 Relaying Technology

The main purpose of relaying techniques is to provide alternative

paths to the direct one (source-destination) to increase robustness,

reliability, coverage, and throughput of wireless networks. When the

direct path either does not have enough channel gain or it is not avail-

able (e.g., out of coverage or blocked), the UE can transmit to an in-

termediate node called relay. This, in single-hop relaying, transmits

the information to the destination node by either a wired or a wireless

link, whereas, in the multi-hop relaying technique [61], several relays

are used to reach the destination node. Relay nodes usually they do

not have their own traﬃc and can be either stationary nodes, as pow-

erful as a mmAP, or mobile devices with constrained energy capacity

2 Robust Connectivity in Millimeter-Waves

##$%

&'()*

!"#$%&'"()&*+

##$%

&'()*

!"#$%&'"()&*+

Figure 2.3: Uplink directional mm-wave communications as-

sisted by a buﬀer-aided relay. On the left, the UE transmits

a packet directly to the mmAP (destination), whereas, on the

right, the UE transmits a packet to the relay.

and limited computational power as in device-to-device (D2D) relay-

ing techniques [62]. We can distinguish between full-duplex (FD) and

half-duplex (HD) relays [63, 64, 65], depending on whether the relay

node can receive and transmit simultaneously or not. Moreover, we

can diﬀerentiate between amplify and forward (AF) [66] and decode

and forward (DF) relays [67]. The latter decodes, re-modulates, and

retransmits the received signal, whereas, the former simply ampliﬁes

and retransmits the signal without decoding. Although AF relays

are simpler and cheaper, DF relays guarantee higher performance.

Furthermore, relays can be equipped with a queue that allows them

to store received packets and to transmit them when then wireless

conditions are favorable [68, 69].

Given the blockage sensitivity and the high path loss of mm-wave

communications, relays can represent a powerful tool. Speciﬁcally,

relays can be used to decrease communication interruptions by pro-

viding alternative LOS paths [67, 70, 71, 72], or to overcome the severe

penetration loss for outdoor-indoor communications [73]. When the

relay is ﬁxed, a proper relay placement is crucial to provide LOS links

between the relay and the destination and the Tx nodes. However,

the use of narrow beams in mm-waves introduces further complexity.

More precisely, unlike omnidirectional communications, Txs may not

be able to simultaneously transmit to both the destination and the

relay. As shown in Fig. 2.3, the Tx node chooses to send a packet

either to the relay or to the destination. The eﬀect of directional

transmissions in relaying techniques for mm-waves is studied in the

third paper of the thesis.

I Introduction

When multiple relays are available, multi-hop routing and relay

selection are key functionalities that may highly aﬀect network per-

formance. In this regard, several schemes have been proposed and

analyzed both for mm-wave and low-frequency communications [67,

70, 74, 75, 76, 77]. Moreover, given the high blockage sensitivity, it

could be important for mm-wave wireless networks to account for the

spatial correlation between the Tx-destination and the Tx-relay links

during the relay selection procedure. Speciﬁcally, as shown in [78], an

obstacle could block both of these links, and an optimal angular dis-

tance between the relay and the destination to maximize the success

transmission probability can be computed.

2.3 Intelligent Reﬂective Surfaces

In recent years, intelligent reﬂective surfaces (IRSs) have been identi-

ﬁed as a breakthrough technology for the next generations of mobile

communications, e.g., 6G [79, 80]. IRSs consist of arrays of reﬂec-

tive elements that can be electronically controlled to adjust the angle

and the phase of the reﬂected signals to be either added coherently

or destructively to the receiver [49, 81]. The reﬂective elements can

be made of several types of material, e.g., reﬂectarrays, liquid crystal

surfaces, and software-deﬁned meta-surfaces, that have speciﬁc elec-

tromagnetic characteristics [82, 83], These reﬂective elements are elec-

tronically controlled by a controller to change the amplitude, phase,

and even polarization of the incident signal to either increase the

received power at the intended Rx or suppress interference at non-

intended Rx nodes. These functions can be even selected through

a software-deﬁned interface for controlling the electromagnetic en-

vironment [84, 85] to enhance the coverage, throughput, reliability,

and security.

Similarly to AF relays, IRSs are transparent with respect to Tx

and Rx nodes. However, IRSs do not amplify the signals and they

operate in a FD mode without introducing further noise or inter-

ference [82, 86]. In [87], the authors have performed a comparative

performance analysis between IRSs and DF relays and the analysis

has revealed the competitiveness of IRSs with respect to DF relays,

especially when the number of reﬂective elements is suﬃciently large.

Furthermore, IRSs are nearly passive with a negligible energy con-

sumption. All these characteristics make them a low-cost eﬀective

solution as an alternative to AF or DF relays.

2 Robust Connectivity in Millimeter-Waves

𝑮 𝒉𝒓

𝒉𝒅

IRS controller

IRS

Figure 2.4: Uplink IRS-assisted mm-wave communications.

The controller is connected to both the AP and the IRS in or-

der to jointly optimize the beamforming at the AP and the IRS.

The terms G,hd, and hrrepresent the IRS-AP, UE-AP, and

UE-IRS channels.

IRSs have also shown to be incredibly eﬀective at mm-wave fre-

quency bands and, similarly to relays, can provide alternative paths

in NLOS conditions to overcome blockages and increase robustness

of mm-wave communications [49, 88]. Moreover, IRSs have been re-

cently tested successfully in the 28 GHz band by NTT DOCOMO [89],

but research is still at the early stage and ongoing trends include: path

loss modelling [81, 90], beamforming optimization [91, 92], designing

and materials [93, 94], and AI controllable environments [95, 96].

Low-cost deployment and small energy consumption of IRSs make

them attractive for improving throughput and reliability of unmanned

aerial vehicle (UAV)-assisted communications [97, 98]. In this sce-

nario, the possibility to control the position of UAVs requires a joint

optimization of IRS beamforming and UAV’s trajectory. A similar

problem arises in mm-wave wireless networks in industrial scenarios

with wirelessly connected robots that are considered in the ﬁfth paper

of this thesis.

2.4 Mobility Management and Association Opti-

mization

Mobility management is a wireless network functionality that aims

to trace mobile users while changing their locations to continuously

provide connectivity. This is a critical functionality for mm-wave

wireless networks operating in dynamic scenarios, where, beam align-

ments and channel estimations must be performed continuously to

I Introduction

!!"# $ !!"# %

!!"# &

'()(*+*,-./0*(,1

'()(*

Figure 2.5: A robot moving in an area covered by three

mmAPs. The robot is connected to mmAP 1 and, given its

trajectory, it is easy to predict blockages and handovers. For

this scenario, an handover to mmAP 2 will lead to a blockage,

whereas, an handover to mmAP 3 can guarantee a longer con-

nection and more reliable communications.

avoid the degradation of the beamforming gain. More precisely, the

high path loss and penetration loss decrease communication ranges

and cause spottier coverage that requires denser cell deployments.

Moreover, the mobility of the user and obstacles can create tem-

porary blockages and strong reﬂections. In such scenarios, abrupt

blockages make the maximum SNR criterion to perform associations

highly ineﬃcient leading to unnecessary handovers that cause fre-

quent communication interruptions. These signiﬁcantly decrease net-

work performance, e.g., throughput, delay, and reliability, which can

be enhanced with handover and blockage prediction and optimized

associations. These schemes can be combined with methods that

reduce the beam alignment and channel estimation overhead. As

shown in [56, 99], MC can be used to avoid communication interrup-

tions during a handover or a blockage by having multiple connections

that are simultaneously active. Other solutions involve methods that

use context information including SNR, beam ID, radio maps, users’

position, and even camera images [100, 101, 102, 103]. Given such in-

formation, the network can predict possible blockages and select the

optimal mmAP to perform a handover with, or reduce the duration

of the initial access and beam alignment. Moreover, machine learn-

ing methods and context information can be jointly used to improve

the performance of context-aware solutions. These methods can be

3 Mathematical Modeling and Solution Tools

used to predict blockages and optimize handovers [104, 105], reduce

beamforming complexity [60, 106], and both [107, 108].

As shown in Fig. 2.5, handovers and blockages in multi-mmAP sce-

narios are easier to predict for autonomous vehicles (such as robots) [109]

with respect to conventional vehicles. Namely, robots’ trajectory can

be controlled. This introduces an additional degree of freedom to

avoid coverage holes and minimize the number of handovers that has

been fully exploited in the fourth paper of this thesis.

Chapter 3

Mathematical Modeling

and Solution Tools

In this chapter, we brieﬂy introduce the underlying theory of the main

mathematical tools that have been used in the papers.

3.1 Mathematical Optimization

A general optimization problem can be deﬁned as follows:

min

f(x) (3.1a)

s.t. gi(x)≤0, i = 1, ..., p, (3.1b)

hj(x) = 0, j = 1, ..., q, (3.1c)

where, xrepresents the n-dimensional vector of optimization variables

of the problem. Function f(x) is the objective function, gi(x) and

hj(x) represent inequality and equality constraints. The set of all

the possible points xthat satisfy these constraints is called feasible

region. This model describes the problem of ﬁnding x∗in the feasible

region that minimizes the objective function. If f(x∗) = −∞, then

the problem is unbounded. The problem is infeasible if the feasible

region is empty. Optimization problems can be classiﬁed according

to the shape of objective function and constraints (linear, non-linear,

convex, non-convex), and variables (continuous, discrete, binary).

3 Mathematical Modeling and Solution Tools

3.1.1 Integer and Combinatorial Optimization

In integer optimization problems, also referred to as integer program-

ming (IP), the variables are constrained to be integer values. If the ob-

jective function is linear, we have an integer linear programming (ILP)

that can be formulated as follows:

IP : min

cTx(3.2a)

s.t. Ax =b,(3.2b)

x≥0 and integer.(3.2c)

This is the same formulation of a linear programming (LP) in stan-

dard form, but with the integrality constraint of the variables. In this

formulation, Ais an mby nmatrix, cTis an n-dimensional row vec-

tor, and bis an m-dimensional column vector. Classes of IP problems

are binary integer programming (BIP) and mixed integer program-

ming (MIP) where the variables are binary or a mix of integer and

continuous variables. Combinatorial problems are a class problems

that are characterized by a discrete set of solutions and often formu-

lated using IP models. Well-known combinatorial problems include

minimum spanning tree, traveling salesman, and cutting stock.

IPs are more challenging to solve than LPs, and global optimality

is hard to achieve in general. Methods that aim to optimally solve

IPs include cutting plane, branch-and-bound, and branch-and-cut al-

gorithms. However, the complexity of these methods grow exponen-

tially with the number of variables. In this regard, computational

complexity theory classiﬁes problems according to their tractability,

and several classes have been identiﬁed. Without going into details

(these can be found in [110]), we can ﬁrst distinguish two classes

of problems: non-deterministic polynomial (NP) and NP-hard prob-

lems. The former is a class of decision problems for which there is

a polynomial-size proof of the answer for any instance for which the

answer is ”yes”, whereas, a problem is NP-hard if any problem in NP

can be reduced to it in polynomial time. If a problem is NP-hard and

its decision problem is NP, then the latter is NP-complete (NP-C).

Finally, polynomial (P) is a class of decision problems in NP that can

be optimally solved in polynomial time. This class of problems is the

easiest to solve, whereas, for NP-hard problems, global optimality is

often hard to achieve, and other methods must be used. These usu-

ally aim to rapidly ﬁnd a suboptimal solution that is close enough to

the optimal one. In this regard, it is crucial to derive good optimality

I Introduction

bound for the solution of the problem. Speciﬁcally, we distinguish

between primal and dual bounds that, for a minimization problem,

provide upper and lower bounds, respectively. Primal bounds can

be obtained by any feasible solution, whereas, dual bounds involve

complex relaxation techniques (e.g., linear, combinatorial, and La-

grangian) [110]. We now present the column generation (CG) method

that has been initially applied to LPs to deal with a large number of

variables. However, as explained in the following, CG can be used for

providing sub-optimal solutions for some IPs that suﬀers from large

dimensionality.

Column Generation

In Papers I and IV, we have used heuristics that are based on a

CG algorithm for solving two combinatorial optimization problems.

CG is a decomposition technique that can be used for solving LPs

that present a large number of variables. At an LP optimum, only

a small subset of variables contribute to the ﬁnal solution, and these

can be iteratively generated starting from an initial feasible solution.

Speciﬁcally, a CG-based algorithm starts from a subset of variables

that correspond to columns of A. This restricted problem is called

master problem that is solved to obtain the optimal dual values for

each constraint. These are used to construct the pricing problem,

of the which solution provides the new variable and corresponding

column of Ato be included in the master problem.

The above process is iteratively repeated until no new improving

column can be generated by the pricing problem. Note that CG re-

quires the variables to be continuous, but we can apply CG to IPs

by considering the continuous relaxation. When the generation of

columns is terminated, we proceed by solving the master problem

that, at this stage, provides an optimal solution for LPs. However,

this does not hold for IPs, for which the variables of the master prob-

lem must be converted to be integer. The resulting problem does not

guarantee optimality and, in some cases, can lead to infeasibility. For

this reason, CG-based algorithms for IP usually include further steps

to generate a ﬁnal integer solution.

For further detail on IP and CG, the reader is referred to [110,

111, 112].

3 Mathematical Modeling and Solution Tools

3.1.2 Convex Optimization

In a convex optimization problem (COP), f(x) and gi(x) are convex

functions, and hj(x) are aﬃne, i.e., hj(x) = aT

jx−bj. Given these

conditions, the feasible set is a convex set. Maximization of a concave

function with convex inequality constraint functions and aﬃne equal-

ity constraints is also a convex optimization problem. A fundamental

property of COPs is that any locally optimal point is also globally

optimal. Moreover, strong duality holds for COPs, and necessary

and suﬃcient conditions to the optimality can be easily found, i.e.,

the Karush-Kuhn-Tucker (KKT) conditions [113]. More precisely, in

any COP with diﬀerentiable objective and constraint functions, any

primal and dual optimal points satisfy the KKT conditions and have

zero duality gap. These properties make COPs particularly easy to

solve and fast algorithms have been developed, e.g., interior point

methods [114].

It is possible to use methods developed to solve COPs to tackle

non-convex problems. However, in these cases, optimality is not guar-

anteed and the optimization can provide a locally optimal solution

that may be far from the global optimum. In this regard, in Paper V,

we have used a successive convex optimization (SCO) algorithm [115].

This technique is particularly eﬀective to solve non-convex problems

by exploiting convex programming.

Successive Convex Optimization

SCO is an heuristic that leverages convex programming to solve non-

convex problems. The goal of SCO techniques is to ﬁnd a local tightly

convex approximation to a given optimization problem with possibly

non-convex objective and constraint functions. Speciﬁcally, consider

problem (3.1), and assume f(x) and gi(x) are non-convex functions.

Then, in a small neighbourhood (trust region) of an initial point x0,

we can ﬁnd convex upper bounds to f(x) and gi(x), i.e., ˆ

f0(x,x0) and

fi(x,x0), respectively. These are used to replace f(x) and gi(x) in

(3.1) to obtain a local convex approximation of the original problem.

This can be easily solved by using techniques for solving COPs and

the solution is used as next local point to construct the successive local

approximation. The above process is repeated until the convergence

is reached. It is possible to prove that, under certain conditions [115],

the algorithm converges to a KKT point of the original problem. For

further details of COP and SCO, the reader is referred to [113, 115].

I Introduction

3.2 Machine Learning

Machine learning (ML) is an application of artiﬁcial intelligence that

aims to automatically derive models by extracting patterns directly

from raw data (dataset) to perform prediction, classiﬁcation and clus-

tering tasks. The main beneﬁts include lower costs for human in-

terventions and the possibility to derive models that describe com-

plex phenomena. Algorithms that can be classiﬁed as ML techniques

have been developed since the 1950s, but nowadays ML is experienc-

ing a golden age thanks also to wireless communications and perva-

sive sensor networks that have dramatically increased the possibil-

ity and decreased the cost of obtaining large datasets. ML applica-

tions are countless and include object detection, image classiﬁcation,

speech recognition, anomaly detection, stock market prediction, and

disease diagnosis. Speciﬁcally, ML represents a key enabler tech-

nology for several ﬁelds, e.g., robotics, automotive, and image and

video processing. Starting from 5G, ML has been gradually used

for developing algorithms to enhance wireless communication perfor-

mance [60, 106, 108] and it is expected that the role of ML will be

even more crucial for next generations of mobile networks [2, 37].

ML algorithms usually consist of a learning (or training) and a

prediction phase. During the learning phase, the model is inferred

from a set of training examples (training dataset) to maximize the

prediction accuracy (minimize the prediction error). Each example

of the training set usually consists of several features. During the

prediction phase, the model is used to perform one of the following

tasks:

•Regression tasks aim to predict quantitative values, e.g., stock

market value and SNR predictions.

•The output of classiﬁcation tasks is a qualitative response and

takes on values in one of a discrete set of classes. Example

of classiﬁcation tasks are disease diagnosis, fault and anomaly

detection, and object classiﬁcation.

•Clustering is the task of ﬁnding subgroups (clusters) in a dataset,

where, the elements belonging to a cluster present certain simi-

larities in the features. Clustering is used in data compression,

pattern recognition, and market research.

Moreover, it is possible to identify two main learning strategies [116]:

•In a supervised learning strategy, the dataset for training the

model includes a set of input training examples and the cor-

3 Mathematical Modeling and Solution Tools

responding true values (label) that must be predicted by the

model. Examples of ML techniques that use supervised learn-

ing strategy are linear regression, logistic regressions, neural

networks (NN), decision trees, and support vector machines

for classiﬁcation and regression tasks. During the learning

phase, the error between the predicted value provided by the

model and the true value is minimized by using several gradient

descent-based algorithms.

•Unsupervised learning consists of inferring unknown patterns

and regularities of the data from an unlabeled training set. Pos-

sible applications for this type of learning techniques include

clustering and segmentation by using NNs, K-means clustering

algorithm, and principle component analysis (PCA).

•Semi-supervised learning strategy combines labeled and unle-

beled training examples.

Although reinforcement learning uses one of the above-mentioned

strategies, it can be considered a fourth approach because the learn-

ing phase is performed online. Speciﬁcally, the training examples

are gradually acquired and, when a new observation is obtained, the

model performs a prediction that is evaluated according to a reward

function and the system learns by experience [117].

Several tradeoﬀs characterize ML techniques e.g., ﬂexibility (or

complexity)/interpretability and bias/variance tradeoﬀs. Speciﬁcally,

more ﬂexible models are able to learn to describe more phenomena,

however, the interpretation of the model results is more complex.

Moreover, higher and lower ﬂexibility increase the variance and the

bias of the model, respectively, which contribute to the prediction

error.

3.2.1 Decision Trees and Random Forests

Decision trees are ML techniques that can be applied to both regres-

sion and classiﬁcation tasks. During the learning phase, a decision

tree divides the features space of the training examples into several

regions. For an observation that belongs to a certain region, the aver-

age value and the most representative class of the region are used to

perform the regression and classiﬁcation tasks, respectively. Despite

the high interpretability of this method, decision trees suﬀer from

high variance. This can be reduced by training several decorrelated

decision trees (random forests) on diﬀerent subsets of the training

I Introduction

dataset. During the prediction phase, the results of the diﬀerent de-

cision trees are averaged. For further detail on ML, decision trees,

and random forests, the reader is referred to [116, 118, 119].

3.3 Queuing Theory

Queuing theory represents a powerful mathematical tool with a large

number of applications, e.g., transportation, ﬁnance, logistic, and

communication networks. Speciﬁcally, queuing analysis can be used

to study dynamic systems and obtain the following key performance

parameters (KPIs): system stability, average queue size, average through-

put, average delay, and outage probability. To derive this information,

we ﬁrst describe the queuing system by deﬁning the following char-

acteristics: arrival statistics (A), service or departure statistics (B),

number of servers (c), queue size (n), and customer population size

(p). These form what is called Kendall’s notation A/B/c/n/p that,

together with the service discipline, e.g., FIFO (ﬁrst-in, ﬁrst-out) or

LIFO (last-in, ﬁrst-out), completely describes the queuing system.

We can obtain the above-mentioned KPIs by studying the evo-

lution of the state, which is usually represented by the queue size,

at discrete time steps or in continuous time. Moreover, arrival and

service processes can be deterministic or represented by random pro-

cesses. We consider the latter case, where arrival and departures are

described either by Poisson or Bernoulli random processes for contin-

uous and discrete time, respectively. In this setup, it is possible to

prove that the evolution of the state is a Markovian process. More

precisely, the future state of the queue depends only on the present

state and not on its past one.

3.3.1 Discrete Time Markov Chain

In this thesis, we focus on discrete time and discrete states of the

queue, of which the evolution can be modeled by a discrete time

Markov chain (DTMC). Speciﬁcally, a DTMC is a discrete-time ran-

dom process that takes values from a discrete set (S). Let deﬁne

s(k) as a vector, of which the n-th element is the probability to be in

state nat timeslot k. This vector is called the state distribution vec-

tor. Then, we can distinguish between periodic and aperiodic Markov

chains. In the former case, s(k) repeats its values at regular inter-

vals of time. Moreover, Markov chains can be either irreducible and

3 Mathematical Modeling and Solution Tools

reducible depending on whether or not it is possible from a state to

reach any other state. For an aperiodic and irreducible Markov chain,

the goal is to compute the steady-state distribution s:

s= lim

k→∞ s(k).(3.3)

This can be obtained by computing the transition probabilities be-

tween all the possible states in Sthat form the transition probability

matrix P. Speciﬁcally, element ij of Prepresents the probability to

transit from state jto state i(pij). For a DMTC, these probabilities

are ﬁxed and do not depend on the previous history, but only on the

current states, and the arrival and departure processes. Given Pand

s(k−1), we can obtain s(k) by using the following relation:

s(k) = P s(k−1),∀k. (3.4)

At the equilibrium, the state distribution assumes values that does

not vary according to kand satisﬁes the following equation:

P s =s.(3.5)

There are several methods to derive sincluding repeated multiplica-

tion of P,Peigenvector, diﬀerence equations, Z-transform, and nu-

merical techniques. For further details of queuing theory and Markov

chains, the reader is referred to [120, 121].

3.3.2 MN/M/1Queue

In Paper III, we have studied the performance of the queue at the relay

that can be modeled as a MN/M/1 queue. Speciﬁcally, the arrival

and departures are Markovian processes (Bernoulli random processes)

and there are Nsource nodes and one server. Given discrete time

and discrete states, the evolution of the queue can be studied by

considering a DMTC whose transition probability matrix is a lower

Hessenberg matrix:







a0b00 0 . . .

a1b1b00. . .

a2b2b1b0. . .

a3b3b2b1. . .

a4b4b3b2. . .

....







.(3.6)

I Introduction

The elements of the matrix can be derived according to the arrival

and departure processes. The steady-state distribution and the main

KPIs of this queue can be obtained by using the Z-transformation

method as explained in detail in [120].

Chapter 4

Contributions of the

Thesis

The thesis analyzes several performance aspects of techniques for in-

creasing the robustness of wireless communications at mm-wave fre-

quencies and deals with speciﬁc optimization problems of these tech-

niques. In the ﬁrst part, we analyze and quantify the beneﬁts of

providing multiple connections to the UEs in mm-wave communica-

tions. We provide rigorous mathematical formulations of two fun-

damental link scheduling problems in MC scenarios that are solved

by using combinatorial optimization and machine learning methods.

These highlight the potential gains of MC in terms of throughput and

reliability and their tradeoﬀs with the allocated resources addressing

research questions Q1 and Q2. In the third paper, we focus on Q2-Q4

and use queueing theory to provide a mathematical analysis of the

performance of relay-aided mm-wave communications. Simulation re-

sults show the beneﬁts and the eﬀect of directional communications

on cooperative networking for mm-waves.

Finally, in the second part of the thesis, we address research ques-

tions Q5-Q7 and exploit the possibility to control the mobility of

robots in industrial scenarios. Speciﬁcally, in the fourth paper, we

consider a problem that aims to jointly optimize the robots’ paths

and the robot-mmAP associations. Then, in the ﬁfth paper, we con-

sider the problem of minimizing the motion energy consumption of a

robot with QoS-constrained mm-wave communications that are aided

by an IRS. Numerical results highlight the importance of the proposed

4 Contributions of the Thesis

problems and corresponding solutions in order to increase robustness

and satisfy QoS requirements for robots connected using mm-wave

communications.

The main ideas presented in the ﬁve research papers are the result

of discussions among all the authors. The author of this dissertation

has contributed to Papers I-V as the ﬁrst author working on model-

ing, mathematical analysis, problem formulation, implementation of

algorithms, simulation, and numerical results along with the writing

of the papers. Papers I-V and the corresponding main scientiﬁc con-

tributions are summarized in Section 4.1.

The thesis is an extension of the author’s Licentiate thesis:

•C. Tatino, ”Performance Aspects in Millimeter-Wave Wireless

Networks,” Link¨oping: Link¨oping University Electronic Press,

2018. Link¨oping Studies in Science and Technology. Licentiate

Thesis, 1821.

I Introduction

4.1 Papers Included in the Thesis

Paper I: Maximum Throughput Scheduling for Multi- Con-

nectivity in Millimeter-Wave Networks, co-authored with I.

Malanchini, N. Pappas, and D. Yuan. This paper has been pub-

lished in Proceedings of International Symposium on Modeling and

Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt),

May 2018.

In the ﬁrst paper, we deal with the link scheduling problem in

a low frequency assisted mm-wave MC scenario with multiple UEs.

Namely, we consider a downlink inter-frequency reliability oriented

MC scheme, where the control and the data planes are transmitted

by using LB-BS and mmAPs, respectively. Although multiple con-

nections increase the reliability of the communication, at the same

time, they utilize resources in terms of frequency bands, backhaul

capacity, energy, and antennas. For this reason, we consider several

MC schemes that allow a mmAP to be prepared for communication

with a UE at diﬀerent levels (states) without transmitting and con-

suming signiﬁcant resources. A centralized network controller (NC)

has the goal to set the state of several links along with a certain time

window according to the channel conditions and resource constraints

to maximize the network throughput.

This paper addresses research question Q1 stated in Chapter 1.

To do so, we ﬁrst derive a rigorous mathematical formulation of the

constrained throughput maximization problem as a non-linear binary

integer problem. To solve the latter, we propose a column generation

algorithm that can deal with the exponential growth of combinatorial

solutions. The results highlight the potential gain of the proposed

MC architectures that, in some scenario, can increase the network

throughput of almost the 30 % with respect to a single connectivity

(SC) scheme. Moreover, we show that the proposed algorithm can

numerically approach the optimal solution. In conclusion, the results

highlight that a MC scheme with predictive link preparation provides

high throughput without signiﬁcantly increasing resource utilization

with respect to a SC scheme.

4 Contributions of the Thesis

Paper II: Learning-Based Link Scheduling in Millimeter-

wave Multi-connectivity Scenarios, co-authored with N. Pappas,

I. Malanchini, L. Ewe, and D. Yuan. This paper has been published

in Proceedings of IEEE International Conference on Communications

(ICC), June 2020.

In the second paper, we consider a complementary approach to the

ﬁrst paper. Speciﬁcally, we solve the problem of minimizing allocated

resources while satisfying user’s QoS requirements in low frequency

assisted mm-wave MC scenarios. We consider a single user downlink

outdoor scenario with several mmAPs and one LB-BS. These jointly

provide coverage and connectivity to a mobile user and are connected

by high-speed backhauling links to an NC. The user needs to receive a

certain amount of data within a deadline. These can be transmitted

by using several links simultaneously including both mm-wave and

low-frequencies communications. In such scenarios, the network per-

formance is highly aﬀected by the number and the selection of trans-

mitting links that depend on several parameters, e.g., type of user

(vehicle or pedestrian), QoS requirements, and blockage frequency.

In this paper, we focus on research questions Q1 and Q2. To

deal with abrupt blockages and channel uncertainty, we propose a

learning-based solution for link scheduling. The proposed method is

arandom forest classiﬁer that can learn the optimal selection of links

with respect to users’ location and QoS requirements to achieve the

following goals: i) satisfying users’ QoS requirements, ii) minimizing

failed transmissions that represent wasted allocated resources, and

iii) oﬄoading low-frequency communications. We show the perfor-

mance of the learning-based model for several mobility types, i.e.,

pedestrians and vehicles. We compare the obtained results with a

genie-aided solution and two baseline methods. The results highlight

how the proposed learning method for predicting mm-wave channel

conditions and corresponding resource allocation is able to optimize

the tradeoﬀ between allocated resources and reliability. Speciﬁcally,

the performance evaluation shows that the random forest classiﬁer

outperforms all the considered baseline methods and, for some sce-

narios, it approaches the optimum by minimizing failed transmissions

and satisfying QoS requirements.

I Introduction

Paper III: On the Beneﬁts of Network-level Cooperation

in Millimeter-wave Communications, co-authored with N. Pap-

pas, I. Malanchini, L. Ewe, and D. Yuan. This paper has been pub-

lished in IEEE Transactions on Wireless Communications, 2019.

In the third work, we analyze the throughput and delay of mm-wave

wireless networks that are assisted by a relay to increase robustness,

reliability, and communication range. Speciﬁcally, we consider a sym-

metric multi-UE scenario with a FD network cooperative relay and

a mmAP, which represents the destination. The UEs have saturated

queues of inﬁnite size and, in each timeslot, transmit with a certain

probability. We study the eﬀect of directional communications on

cooperative networking by deﬁning two UE transmission strategies:

directional and broadcast. By using the former, the UEs send ei-

ther a packet to the destination or the relay by using narrow beams.

Whereas, by using the broadcast transmission, UEs use wider beams,

with lower beamforming gain, to send a packet to both the mmAP

and the relay in the same timeslot. The relay stores, in an inﬁnite

size queue, the successfully received packets.

We obtain closed-form expressions of the network aggregate and

per UE throughput and packet delay by evaluating the stability con-

dition, the arrival, and service rate of the queue at the relay. These

are numerically evaluated for several scenarios by varying the system

parameters, e.g., the number and position of nodes, transmit prob-

ability, and signal-to-interference-plus-noise ratio (SINR) threshold.

To provide answer to research questions Q2 and Q3, we show that co-

operative networking is characterized by a tradeoﬀ between through-

put and delay and, in some scenarios, can increase the throughput

and the transmission success probability of 25%. Furthermore, we

address research question Q4 by showing how the duration of the

beam alignment phase and directional communications signiﬁcantly

impact the performance aﬀecting the optimal strategy. This depends

also on the network topology. Namely, we can observe that, for short

distances between the UEs, the relay, and the mmAP, it is often ben-

eﬁcial to transmit by using a wider beam. Although this provides a

lower beamforming gain, it gives the possibility to transmit to both

the mmAP and the relay in the same timeslot. However, when the

distances and the SINR threshold increase, a directional transmission

becomes preferable.

4 Contributions of the Thesis

Parts of the work have been published and presented in the fol-

lowing conference:

•C. Tatino, N. Pappas, I. Malanchini, L. Ewe and D. Yuan,

”Throughput Analysis for Relay-Assisted Millimeter-Wave Wire-

less Networks,” in IEEE Globecom Workshops (GC Wkshps),

December 2018.

Paper IV: Multi-Robot Association-Path Planning in Mil-

limeter -Wave Industrial Scenarios, co-authored with N. Pappas,

and D. Yuan. This paper has been published in IEEE Networking

Letters, November 2020.

In this work, we deﬁne and solve multi-robot association-path

planning (MAPP) problems that, in contrast to conventional multi-

robot path planning, aim to jointly optimize the robots’ paths and the

robot-AP associations in mm-wave industrial scenarios. The robots

need to move from their starting positions to their goals within a time

horizon. In this regard, mm-wave can provide high data rates for new

industrial applications, but they are characterized by spotty coverage

requiring dense radio deployments. In such scenarios, coverage holes

and numerous handovers may decrease the communication through-

put and reliability of moving robots. We focus on the type of MAPP

with the goal of i) selecting paths for reaching the destinations in the

shortest possible time, ii) minimizing the number of handovers, and

iii) avoiding AP overloading.

To solve this problem, we use a graph-based method. Speciﬁcally,

the robot can move on a directed graph that is deﬁned by using a radio

map on positions that are free from obstacles and covered by at least

an AP. Then, we deﬁne the cost of the edges to include association

and time information. Thanks to this choice, we can guarantee the

connections of the robot and solve MAPP problems by using path-

based formulations. Although we prove the NP-hardness of the prob-

lem, and we propose a column generation-based algorithm (PGCP)

that can solve MAPP in polynomial time. To provide answers to re-

search questions Q5 and Q6, we show that paths and the associations

resulting from the optimization of MAPP problems avoid coverage

holes, AP overloading, and dramatically reduces the number of han-

dovers without increasing signiﬁcantly the traversal time of the paths.

These results highlight the importance of MAPP problems to reach

fully wireless and automated factories for Industry 4.0.

I Introduction

Paper V: QoS Aware Robot Trajectory Optimization with

IRS-assisted Millimeter -Wave Communications, co-authored

with N. Pappas, and D. Yuan. This paper has been submitted to

IEEE Transactions on Wireless Communications, 2020.

Finally, the ﬁfth paper considers the joint optimization of trajec-

tory and beamforming for a wirelessly connected robot using IRS-

assisted mm-wave communications. The goal is to minimize the mo-

tion energy consumption of a robot moving in an area that is par-

tially occupied by obstacles. The robot needs to transmit uplink data

by satisfying a minimum average data rate requirement to an IRS-

assisted AP using mm-waves. This is a fundamental problem in fully

automated factories that characterize Industry 4.0, where robots may

have to perform tasks with given deadlines while maximizing battery

autonomy and communication eﬃciency. In such scenarios, the low-

cost deployment of IRSs represents an eﬃcient solution for enhancing

robustness and reliability at mm-wave frequencies.

To solve the problem, we account for the mutual dependence be-

tween trajectory and beamforming (at the AP and IRS) optimization.

Speciﬁcally, the latter depends on the robot trajectory that must be

optimized to avoid collisions and satisfy time and QoS constraints.

Thus, we ﬁrst provide a rigorous mathematical formulation as a non-

linear and non-convex problem. Then, we exploit mm-wave channel

characteristics to decouple beamforming and trajectory optimization.

The latter is solved by using a modiﬁed successive-convex optimiza-

tion algorithm (RMAP) that accounts for the obstacles’ positions

and a radio map to avoid collisions and satisfy QoS requirements.

We prove that the algorithm can converge to a solution satisfying the

Karush-Kuhn-Tucker (KKT) conditions, and simulation results show

the fast convergence of RMAP.

The proposed algorithm is able to provide solutions that dramati-

cally reduce the motion energy consumption with respect to methods

that aim to ﬁnd maximum-rate trajectories. Moreover, in this paper,

we focus on research questions Q6 and Q7. More precisely, we ﬁrst

show how RMAP, thanks to the radio map information, can provide

trajectories that avoid poorly covered areas. Then, we show how the

IRS and the beamforming optimization increase the motion energy

eﬃciency of the robot by increasing mm-wave coverage and robust-

ness.

4 Contributions of the Thesis

4.2 Papers not Included in the Thesis

•C. Parera, Q. Liao, I. Malanchini, C. Tatino, A. E. C. Redondi

and M. Cesana, ”Transfer Learning for Tilt-Dependent Radio

Map Prediction,” in IEEE Transactions on Cognitive Commu-

nications and Networking, vol. 6, no. 2, pp. 829-843, June

2020.

•C. Parera, A. E. C. Redondi, M. Cesana, Q. Liao, L. Ewe, and

C. Tatino ”Transferring Knowledge for Tilt-Dependent Radio

Map Prediction,” in Proceedings of IEEE Wireless Communi-

cations and Networking Conference (WCNC), April 2018.

•N. Pappas, E. Fountoulakis, C. Tatino, V. Angelakis, and D.

Yuan, ”Pursuing the Potential of New Mechanisms for Perfor-

mance Engineering of 5G,” in Proceedings of IEEE 22nd Inter-

national Workshop on Computer Aided Modeling and Design

of Communication Links and Networks (CAMAD), June 2017.

•C. Tatino, N. Pappas, and D. Yuan, ”Beam Based Stochas-

tic Model of the Coverage Probability in 5G Millimeter-wave

Systems,” in Proceedings of 15th International Symposium on

Modeling and Optimization in Mobile, Ad Hoc, and Wireless

Networks (WiOpt), May 2017.

4.3 Conclusions and Future Research

Our research has shown that MC can increase robustness, throughput,

and reliability of mm-wave communications at the cost of additional

allocated resources (e.g., number of connections). An optimized allo-

cation of the resources can lead to eﬃcient MC solutions, of which the

performance can dramatically increase if combined with methods aim-

ing to predict channel conditions. These can be implemented by using

machine learning algorithms that show high performance in minimiz-

ing the percentage of failed transmissions and satisfying minimum

QoS requirements. Moreover, we show that the number of available

links, the type of users, and the scattering environment are the main

parameters that aﬀect the performance of MC techniques. Coopera-

tive networking represents another technique to increase throughput

and reliability, but, as shown by the results, at the cost of additional

delay introduced by queue-aided relays. Directional communications

and beam alignments highly aﬀect the performance of this type of

technique, as well as the network topology and the number of UEs.

Additional results show that the possibility to optimize robot tra-

I Introduction

jectory can make robot-AP communications more robust by selecting

paths that avoid coverage holes and poorly covered areas. Moreover,

by jointly optimizing robot-AP associations and robots’ trajectory, it

is possible to minimize handovers without increasing signiﬁcantly the

traversal time. Furthermore, we show that the deployment of IRSs

improves radio coverage that increases the energy eﬃciency of robots

under communication QoS constraints. In this regard, radio maps are

a fundamental tool to achieve the required QoS.

MC solutions present further tradeoﬀs that have not been deeply

analyzed in this thesis and in literature. Speciﬁcally, MC increases

backhauling traﬃc and needs accurate coordination between the APs

and the NC, and proper coverage planning, which introduce addi-

tional delay, complexity, and deployment cost, respectively. More-

over, emerging technologies for 6G will introduce new possible MC

schemes e.g., with terahertz and satellite communications. The anal-

ysis in Paper III has shown whether to transmit to the relay or to

the AP or both. However, the optimal strategy for a UE depends on

the network topology and selected strategies of possible interferers.

Thus, joint optimization of the transmission strategy at multiple UEs

can improve throughput and reliability. Moreover, IRSs are eﬃcient

alternatives to relays, of which the pros and cons are still not fully ana-

lyzed at mm-wave frequencies. Finally, new problems arise for robotic

systems that will represent important use cases for 6G and beyond.

Joint coverage planning and trajectory optimization and beam track-

ing for mobile robots can represent promising research directions for

mm-wave communications.

Bibliography

[1] Cisco, “Cisco annual internet report (2018-2023),” Tech. Rep.,

Mar. 2020.

[2] F. Tariq, M. R. A. Khandaker, K. K. Wong, M. A. Imran,

M. Bennis, and M. Debbah, “A speculative study on 6G,” IEEE

Wireless Communications, vol. 27, no. 4, pp. 118–125, 2020.

[3] ITU, “ITU radio regulations,” Tech. Rep., 2020.

[4] FCC, “Use of spectrum bands above 24 GHz for mobile radio

services,” Tech. Rep., 2019.

[5] G. Kang, S. Jiang, and Z. Zhao, “Research on millimeter wave

spectrum planning for 5G,” Journal of Physics: Conference Se-

ries, vol. 1345, p. 032040, Nov. 2019.

[6] X. Lu et al., “Integrated use of licensed- and unlicensed-band

mmwave radio technology in 5G and beyond,” IEEE Access,

vol. 7, pp. 24 376–24 391, 2019.

[7] S. Rangan, T. S. Rappaport, and E. Erkip, “Millimeter-wave

cellular wireless networks: Potentials and challenges,” Proceed-

ings of the IEEE, vol. 102, no. 3, pp. 366–385, Mar. 2014.

[8] T. S. Rappaport, Y. Xing, G. R. MacCartney, A. F. Molisch,

E. Mellios, and J. Zhang, “Overview of millimeter wave com-

munications for ﬁfth-generation (5G) wireless networks-with a

focus on propagation models,” IEEE Transactions on Antennas

and Propagation, vol. 65, no. 12, pp. 6213–6230, Dec. 2017.

[9] H. Zhao et al., “28 GHz millimeter wave cellular communication

measurements for reﬂection and penetration loss in and around

buildings in new york city,” in IEEE International Conference

on Communications (ICC), Jun. 2013, pp. 5163–5167.

[10] G. R. MacCartney and T. S. Rappaport, “73 GHz millimeter

wave propagation measurements for outdoor urban mobile and

backhaul communications in new york city,” in IEEE Interna-

tional Conference on Communications (ICC), Jun. 2014, pp.

4862–4867.

[11] T. Bai, R. Vaze, and R. W. Heath, “Analysis of blockage eﬀects

on urban cellular networks,” IEEE Transactions on Wireless

Communications, vol. 13, no. 9, pp. 5070–5083, Sept. 2014.

[12] M. Gapeyenko et al., “On the temporal eﬀects of mobile block-

ers in urban millimeter-wave cellular scenarios,” IEEE Trans-

actions on Vehicular Technology, vol. 66, no. 11, pp. 10 124–

10 138, Nov. 2017.

[13] H. Shokri-Ghadikolaei and C. Fischione, “The transitional be-

havior of interference in millimeter wave networks and its im-

pact on medium access control,” IEEE Transactions on Com-

munications, vol. 64, no. 2, pp. 723–740, Feb. 2016.

[14] Y. Niu, Y. Li, D. Jin, L. Su, and A. V. Vasilakos, “A survey

of millimeter wave communications (mmwave) for 5G: Oppor-

tunities and challenges,” Wireless Networks, vol. 21, no. 8, pp.

2657–2676, Nov. 2015.

[15] M. Xiao et al., “Millimeter wave communications for future mo-

bile networks,” IEEE Journal on Selected Areas in Communi-

cations, vol. 35, no. 9, pp. 1909–1935, Sept. 2017.

[16] J. Wang et al., “Beam codebook based beamforming protocol

for multi-Gbps millimeter-wave wpan systems,” IEEE Journal

on Selected Areas in Communications, vol. 27, no. 8, pp. 1390–

1399, 2009.

[17] H. Shokri-Ghadikolaei, L. Gkatzikis, and C. Fischione, “Beam-

searching and transmission scheduling in millimeter wave com-

munications,” in IEEE International Conference on Communi-

cations (ICC), 2015, pp. 1292–1297.

[18] M. Giordani, M. Mezzavilla, and M. Zorzi, “Initial access in 5G

mmwave cellular networks,” IEEE Communications Magazine,

vol. 54, no. 11, pp. 40–47, Nov. 2016.

[19] C. N. Barati, S. A. Hosseini, M. Mezzavilla, T. Korakis, S. S.

Panwar, S. Rangan, and M. Zorzi, “Initial access in millimeter

wave cellular systems,” IEEE Transactions on Wireless Com-

munications, vol. 15, no. 12, pp. 7926–7940, Dec. 2016.

[20] M. Giordani, M. Polese, A. Roy, D. Castor, and M. Zorzi, “A

tutorial on beam management for 3GPP NR at mmwave fre-

quencies,” IEEE Communications Surveys Tutorials, vol. 21,

no. 1, pp. 173–196, 2019.

[21] S. Hur, T. Kim, D. J. Love, J. V. Krogmeier, T. A. Thomas,

and A. Ghosh, “Millimeter wave beamforming for wireless back-

haul and access in small cell networks,” IEEE Transactions on

Communications, vol. 61, no. 10, pp. 4391–4403, Oct. 2013.

[22] H. Shokri-Ghadikolaei, C. Fischione, G. Fodor, P. Popovski,

and M. Zorzi, “Millimeter wave cellular networks: A MAC layer

perspective,” IEEE Transactions on Communications, vol. 63,

no. 10, pp. 3437–3458, 2015.

[23] IEEE, “IEEE standard for information technology–

telecommunications and information exchange between

systems–local and metropolitan area networks–speciﬁc

requirements-part 11: Wireless LAN medium access control

(MAC) and physical layer (PHY) speciﬁcations amendment

3: Enhancements for very high throughput in the 60 GHz

band,” IEEE Std 802.11ad-2012 (Amendment to IEEE Std

802.11-2012, as amended by IEEE Std 802.11ae-2012 and

IEEE Std 802.11aa-2012), pp. 1–628, 2012.

[24] T. Baykas, C. S. Sum, Z. Lan, J. Wang, M. A. Rahman,

H. Harada, and S. Kato, “IEEE 802.15.3c: the ﬁrst IEEE wire-

less standard for data rates over 1 Gb/s,” IEEE Communica-

tions Magazine, vol. 49, no. 7, pp. 114–121, Jul. 2011.

[25] M. Z. Zaaimia, R. Touhami, L. Talbi, M. Nedil, and M. C. E.

Yagoub, “60-GHz statistical channel characterization for wire-

less data centers,” IEEE Antennas and Wireless Propagation

Letters, vol. 15, pp. 976–979, 2016.

[26] mmMAGIC, “Deliverable d2.1: Measurement campaigns and

initial channel models for preferred suitable frequency ranges,”

Tech. Rep., 2016.

[27] C. Dehos, J. L. Gonzalez, A. D. Domenico, D. Ktenas, and

L. Dussopt, “Millimeter-wave access and backhauling: the so-

lution to the exponential data traﬃc increase in 5G mobile

communications systems?” IEEE Communications Magazine,

vol. 52, no. 9, pp. 88–95, 2014.

[28] W. Feng, Y. Li, D. Jin, L. Su, and S. Chen, “Millimetre-wave

backhaul for 5G networks: Challenges and solutions,” Sensors,

vol. 16, pp. 892–908, Jun. 2016.

[29] Y. Golovachev, A. Etinger, G. Pinhasi, and Y. Pinhasi, “Mil-

limeter wave high resolution radar accuracy in fog conditions-

theory and experimental veriﬁcation,” Sensors, vol. 18, pp.

2148–2159, Jul. 2018.

[30] S. Sugimoto, H. Tateda, H. Takahashi, and M. Okutomi, “Ob-

stacle detection using millimeter-wave radar and its visualiza-

tion on image sequence,” in Proceedings of the 17th Interna-

tional Conference on Pattern Recognition (ICPR)., vol. 3, 2004,

pp. 342–345.

[31] F. Lemic et al., “Localization as a feature of mmwave commu-

nication,” in Proceedings of the 12th IEEE International Con-

ference on Wireless Communications and Mobile Computing

(IWCMC’16), Sept. 2016.

[32] Ericsson, “Ericsson, Qualcomm and U.S. cellular achieve

extended-range 5G data call over mmwave,” Tech. Rep.,

Sept. 2020. [Online]. Available: https://www.ericsson.

com/en/press-releases/2020/9/ericsson-qualcomm-and-u.s.

-cellular-achieve-extended-range-5g-data-call-over-mmwave

[33] M. Cheﬀena, “Industrial wireless communications over the mil-

limeter wave spectrum: opportunities and challenges,” IEEE

Communications Magazine, vol. 54, no. 9, pp. 66–72, Sept. 2016.

[34] J. Sachs, K. Wallstedt, F. Alriksson, and G. Eneroth, “Boosting

smart manufacturing with 5G wireless connectivity,” Ericsson

Technology Review, 2019.

[35] C. Pielli, T. Ropitault, and M. Zorzi, “The potential of

mmwaves in smart industry: Manufacturing at 60 GHz,” in

17th International Conference on Ad Hoc Networks and Wire-

less (AD HOC-NOW), 2018, pp. 64–76.

[36] S. Saponara, F. Giannetti, B. Neri, and G. Anastasi, “Exploit-

ing mm-wave communications to boost the performance of in-

dustrial wireless networks,” IEEE Transactions on Industrial

Informatics, vol. 13, no. 3, pp. 1460–1470, 2017.

[37] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and

M. Zorzi, “Toward 6G networks: Use cases and technologies,”

IEEE Communications Magazine, vol. 58, no. 3, pp. 55–61,

2020.

[38] G. R. MacCartney, S. Deng, S. Sun, and T. S. Rappaport,

“Millimeter-wave human blockage at 73 GHz with a simple dou-

ble knife-edge diﬀraction model and extension for directional

antennas,” in the 84th IEEE Vehicular Technology Conference

(VTC-Fall), Sept. 2016, pp. 1–6.

[39] G. R. MacCartney and T. S. Rappaport, “Millimeter-wave base

station diversity for 5G coordinated multipoint (CoMP) ap-

plications,” IEEE Transactions on Wireless Communications,

vol. 18, no. 7, pp. 3395–3410, 2019.

[40] J. Choi, “On the macro diversity with multiple BSs to mitigate

blockage in millimeter-wave communications,” IEEE Commu-

nications Letters, vol. 18, no. 9, pp. 1653–1656, Sept. 2014.

[41] E. Hriba and M. Valenti, “Correlated blocking in mmwave cel-

lular networks: Macrodiversity, outage, and interference,” Elec-

tronics, vol. 8, p. 1187, Oct. 2019.

[42] F. Devoti and I. Filippini, “Planning mm-wave access net-

works under obstacle blockages: A reliability-aware approach,”

IEEE/ACM Transactions on Networking, vol. 28, no. 5, pp.

2203–2214, 2020.

[43] M. Gerasimenko, D. Moltchanov, M. Gapeyenko, S. Andreev,

and Y. Koucheryavy, “Capacity of multiconnectivity mmwave

systems with dynamic blockage and directional antennas,”

IEEE Transactions on Vehicular Technology, vol. 68, no. 4, pp.

3534–3549, 2019.

[44] G. Yang and M. Xiao, “Performance analysis of millimeter-wave

relaying: Impacts of beamwidth and self-interference,” IEEE

Transactions on Communications, vol. 66, no. 2, pp. 589–600,

2018.

[45] X. An et al., “Beam switching support to resolve link-blockage

problem in 60 GHz WPANs,” in IEEE 20th International Sym-

posium on Personal, Indoor and Mobile Radio Communica-

tions, Sept. 2009, pp. 390–394.

[46] C. Tatino, I. Malanchini, D. Aziz, and D. Yuan, “Beam based

stochastic model of the coverage probability in 5G millimeter

wave systems,” in the 15th International Symposium on Model-

ing and Optimization in Mobile, Ad Hoc, and Wireless Networks

(WiOpt), May 2017, pp. 1–6.

[47] N. A. Muhammad, P. Wang, Y. Li, and B. Vucetic, “Analytical

model for outdoor millimeter wave channels using geometry-

based stochastic approach,” IEEE Transactions on Vehicular

Technology, vol. 66, no. 2, pp. 912–926, Feb. 2017.

[48] A. Narayanan, T. V. Sreejith, and R. K. Ganti, “Coverage

analysis in millimeter wave cellular networks with reﬂections,”

in IEEE Global Communications Conference (GLOBECOM),

2017, pp. 1–6.

[49] P. Wang, J. Fang, X. Yuan, Z. Chen, and H. Li, “Intelligent re-

ﬂecting surface-assisted millimeter wave communications: Joint

active and passive precoding design,” IEEE Transactions on

Vehicular Technology, pp. 1–1, 2020.

[50] A. Wolf, P. Schulz, M. D¨orpinghaus, J. C. S. Santos Filho, and

G. Fettweis, “How reliable and capable is multi-connectivity?”

IEEE Transactions on Communications, vol. 67, no. 2, pp.

1506–1520, 2019.

[51] A. Wolf, P. Schulz, D. ¨

Ohmann, M. D¨orpinghaus, and G. Fet-

tweis, “Rate-reliability tradeoﬀ for multi-connectivity,” in

IEEE Wireless Communications and Networking Conference

(WCNC), Apr. 2018, pp. 1–6.

[52] 3GPP, “Evolved universal terrestrial radio access (E-UTRA)

and evolved universal terrestrial radio access network (E-

UTRAN); overall description (release 16), 3GPP TS 36.300,”

Tech. Rep., Jan. 2020.

[53] ——, “Evolved universal terrestrial radio access (E-UTRA) and

NR; multi-connectivity (release 16), 3GPP TS 37.340 V16.3.0,”

Tech. Rep., Sept. 2020.

[54] D. Moltchanov, A. Ometov, S. Andreev, and Y. Koucheryavy,

“Upper bound on capacity of 5G mmwave cellular with multi-

connectivity capabilities,” Electronics Letters, vol. 54, no. 11,

pp. 724–726, 2018.

[55] M. Gapeyenko, V. Petrov, D. Moltchanov, M. R. Akdeniz,

S. Andreev, N. Himayat, and Y. Koucheryavy, “On the degree of

multi-connectivity in 5G millimeter-wave cellular urban deploy-

ments,” IEEE Transactions on Vehicular Technology, vol. 68,

no. 2, pp. 1973–1978, 2019.

[56] V. Petrov et al., “Dynamic multi-connectivity performance in

ultra-dense urban mmwave deployments,” IEEE Journal on Se-

lected Areas in Communications, vol. 35, no. 9, pp. 2038–2055,

2017.

[57] D. Aziz, J. Gebert, A. Ambrosy, H. Bakker, and H. Halbauer,

“Architecture approaches for 5G millimetre wave access assisted

by 5G low-band using multi-connectivity,” in IEEE Globecom

Workshops (GC Wkshps), Dec. 2016, pp. 1–6.

[58] M. Polese, M. Giordani, M. Mezzavilla, S. Rangan, and

M. Zorzi, “Improved handover through dual connectivity in 5G

mmwave mobile networks,” IEEE Journal on Selected Areas in

Communications, vol. 35, no. 9, pp. 2069–2084, Sept. 2017.

[59] M. Giordani, M. Mezzavilla, S. Rangan, and M. Zorzi, “Multi-

connectivity in 5G mmwave cellular networks,” in Mediter-

ranean Ad Hoc Networking Workshop (Med-Hoc-Net), Jun.

2016, pp. 1–7.

[60] A. Alkhateeb, S. Alex, P. Varkey, Y. Li, Q. Qu, and D. Tujkovic,

“Deep learning coordinated beamforming for highly-mobile mil-

limeter wave systems,” IEEE Access, vol. 6, pp. 37 328–37 348,

2018.

[61] X. Lin and J. G. Andrews, “Connectivity of millimeter wave

networks with multi-hop relaying,” IEEE Wireless Communi-

cations Letters, vol. 4, no. 2, pp. 209–212, 2015.

[62] M. N. Tehrani, M. Uysal, and H. Yanikomeroglu, “Device-to-

device communication in 5G cellular networks: challenges, solu-

tions, and future directions,” IEEE Communications Magazine,

vol. 52, no. 5, pp. 86–92, May 2014.

[63] D. Kim, H. Lee, and D. Hong, “A survey of in-band full-duplex

transmission: From the perspective of PHY and MAC layers,”

IEEE Communications Surveys Tutorials, vol. 17, no. 4, pp.

2017–2046, Fourthquarter 2015.

[64] N. Shende, O. Gurbuz, and E. Erkip, “Half-duplex or full-

duplex relaying: A capacity analysis under self-interference,”

in the 47th Annual Conference on Information Sciences and

Systems (CISS), Mar. 2013, pp. 1–6.

[65] Z. Wei, X. Zhu, S. Sun, Y. Huang, L. Dong, and Y. Jiang, “Full-

duplex versus half-duplex amplify-and-forward relaying: Which

is more energy eﬃcient in 60-GHz dual-hop indoor wireless sys-

tems?” IEEE Journal on Selected Areas in Communications,

vol. 33, no. 12, pp. 2936–2947, Dec. 2015.

[66] L. Jiang and H. Jafarkhani, “Mmwave amplify-and-forward

mimo relay networks with hybrid precoding/combining design,”

IEEE Transactions on Wireless Communications, vol. 19, no. 2,

pp. 1333–1346, 2020.

[67] K. Belbase, Z. Zhang, H. Jiang, and C. Tellambura, “Coverage

analysis of millimeter wave decode-and-forward networks with

best relay selection,” IEEE Access, vol. 6, pp. 22 670–22 683,

2018.

[68] N. Nomikos, T. Charalambous, I. Krikidis, D. N. Skoutas,

D. Vouyioukas, M. Johansson, and C. Skianis, “A survey on

buﬀer-aided relay selection,” IEEE Communications Surveys &

Tutorials, vol. 18, no. 2, pp. 1073–1097, 2016.

[69] N. Pappas, M. Kountouris, A. Ephremides, and A. Traganitis,

“Relay-assisted multiple access with full-duplex multi-packet

reception,” IEEE Transactions on Wireless Communications,

vol. 14, no. 7, pp. 3544–3558, 2015.

[70] S. Biswas, S. Vuppala, J. Xue, and T. Ratnarajah, “On the

performance of relay aided millimeter wave networks,” IEEE

Journal of Selected Topics in Signal Processing, vol. 10, no. 3,

pp. 576–588, Apr. 2016.

[71] Z. Lin, X. Peng, F. Chin, and W. Feng, “Outage performance of

relaying with directional antennas in the presence of co-channel

interferences at relays,” IEEE Wireless Communications Let-

ters, vol. 1, no. 4, pp. 288–291, Aug. 2012.

[72] S. Singh, F. Ziliotto, U. Madhow, E. Belding, and M. Rodwell,

“Blockage and directivity in 60 GHz wireless personal area net-

works: from cross-layer model to multihop MAC design,” IEEE

Journal on Selected Areas in Communications, vol. 27, no. 8,

pp. 1400–1413, Oct. 2009.

[73] K. Ntontin and C. Verikoukis, “Relay-aided outdoor-to-indoor

communication in millimeter-wave cellular networks,” IEEE

Systems Journal, vol. 14, no. 2, pp. 2473–2484, 2020.

[74] Z. Zhang, X. Chai, K. Long, A. V. Vasilakos, and L. Hanzo,

“Full duplex techniques for 5G networks: self-interference can-

cellation, protocol design, and relay selection,” IEEE Commu-

nications Magazine, vol. 53, no. 5, pp. 128–137, May 2015.

[75] H. Cui, M. Ma, L. Song, and B. Jiao, “Relay selection for two-

way full duplex relay networks with amplify-and-forward proto-

col,” IEEE Transactions on Wireless Communications, vol. 13,

no. 7, pp. 3768–3777, Jul. 2014.

[76] Y. Xu, H. Shokri-Ghadikolaei, and C. Fischione, “Distributed

association and relaying with fairness in millimeter wave

networks,” IEEE Transactions on Wireless Communications,

vol. 15, no. 12, pp. 7955–7970, Dec. 2016.

[77] K. Belbase, C. Tellambura, and H. Jiang, “Two-way relay se-

lection for millimeter wave networks,” IEEE Communications

Letters, vol. 22, no. 1, pp. 201–204, 2018.

[78] S. Kwon and J. Widmer, “Relay selection for mmwave com-

munications,” in the 28th IEEE Annual International Sympo-

sium on Personal, Indoor, and Mobile Radio Communications

(PIMRC), Oct. 2017, pp. 1–6.

[79] Z. Zhang et al., “6G wireless networks: Vision, requirements,

architecture, and key technologies,” IEEE Vehicular Technology

Magazine, vol. 14, no. 3, pp. 28–41, 2019.

[80] Q. Wu and R. Zhang, “Towards smart and reconﬁgurable envi-

ronment: Intelligent reﬂecting surface aided wireless network,”

IEEE Communications Magazine, vol. 58, no. 1, pp. 106–112,

2020.

[81] W. Tang et al., “Wireless communications with reconﬁgurable

intelligent surface: Path loss modeling and experimental mea-

surement,” IEEE Transactions on Wireless Communications,

2020.

[82] E. Basar, M. Di Renzo, J. De Rosny, M. Debbah, M. Alouini,

and R. Zhang, “Wireless communications through reconﬁg-

urable intelligent surfaces,” IEEE Access, vol. 7, pp. 116 753–

116 773, 2019.

[83] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioanni-

dis, and I. Akyildiz, “A new wireless communication paradigm

through software-controlled metasurfaces,” IEEE Communica-

tions Magazine, vol. 56, no. 9, pp. 162–169, 2018.

[84] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioannidis,

and I. Akyildiz, “A novel communication paradigm for high ca-

pacity and security via programmable indoor wireless environ-

ments in next generation wireless systems,” Ad Hoc Networks,

vol. 87, pp. 1 – 16, 2019.

[85] A. C. Tasolamprou et al., “Exploration of intercell wireless

millimeter-wave communication in the landscape of intelligent

metasurfaces,” IEEE Access, vol. 7, pp. 122931–122 948, 2019.

[86] E. Bj¨ornson, ¨

O. ¨

Ozdogan, and E. G. Larsson, “Reconﬁgurable

intelligent surfaces: Three myths and two critical questions,”

2020. [Online]. Available: https://arxiv.org/abs/2006.03377

[87] E. Bj¨ornson, ¨

O. ¨

Ozdogan, and E. G. Larsson, “Intelligent re-

ﬂecting surface versus decode-and-forward: How large surfaces

are needed to beat relaying?” IEEE Wireless Communications

Letters, vol. 9, no. 2, pp. 244–248, 2020.

[88] W. Khawaja, O. Ozdemir, Y. Yapici, I. Guvenc, and Y. Kak-

ishima, “Coverage enhancement for mm wave communications

using passive reﬂectors,” in 11th Global Symposium on Millime-

ter Waves (GSMM), 2018, pp. 1–6.

[89] NTT DOCOMO, “Docomo conducts world’s ﬁrst successful

trial of transparent dynamic metasurface,” Tech. Rep.,

Jan. 2020. [Online]. Available: https://www.nttdocomo.co.jp/

english/info/media center/pr/2020/0117 00.html

[90] ¨

O. ¨

Ozdogan, E. Bj¨ornson, and E. G. Larsson, “Intelligent re-

ﬂecting surfaces: Physics, propagation, and pathloss modeling,”

IEEE Wireless Communications Letters, vol. 9, no. 5, pp. 581–

585, 2020.

[91] Q. Wu and R. Zhang, “Intelligent reﬂecting surface enhanced

wireless network via joint active and passive beamforming,”

IEEE Transactions on Wireless Communications, vol. 18,

no. 11, pp. 5394–5409, 2019.

[92] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: The

potential of data transmission with large intelligent surfaces,”

IEEE Transactions on Signal Processing, vol. 66, no. 10, pp.

2746–2758, 2018.

[93] A. Pitilakis et al., “A multi-functional reconﬁgurable meta-

surface: Electromagnetic design accounting for fabrication as-

pects,” IEEE Transactions on Antennas and Propagation, pp.

1–1, 2020.

[94] A. Zappone, M. Di Renzo, X. Xi, and M. Debbah, “On the

optimal number of reﬂecting elements for reconﬁgurable intel-

ligent surfaces,” IEEE Wireless Communications Letters, pp.

1–1, 2020.

[95] Q. Zhang, W. Saad, and M. Bennis, “Millimeter wave

communications with an intelligent reﬂector: Performance

optimization and distributional reinforcement learning,” 2020.

[Online]. Available: https://arxiv.org/abs/2002.10572

[96] M. Di Renzo et al., “Smart radio environments empowered by

ai reconﬁgurable meta-surfaces: An idea whose time has come,”

EURASIP Journal on Wireless Communications and Network-

ing, p. 129, May 2019.

[97] L. Ge, P. Dong, H. Zhang, J. Wang, and X. You, “Joint beam-

forming and trajectory optimization for intelligent reﬂecting

surfaces-assisted UAV communications,” IEEE Access, vol. 8,

pp. 78 702–78 712, 2020.

[98] S. Li, B. Duo, X. Yuan, Y. Liang, and M. Di Renzo, “Re-

conﬁgurable intelligent surface assisted UAV communication:

Joint trajectory design and passive beamforming,” IEEE Wire-

less Communications Letters, vol. 9, no. 5, pp. 716–720, 2020.

[99] M. Giordani, M. Mezzavilla, S. Rangan, and M. Zorzi, “An

eﬃcient uplink multi-connectivity scheme for 5G millimeter-

wave control plane applications,” IEEE Transactions on Wire-

less Communications, vol. 17, no. 10, pp. 6806–6821, 2018.

[100] R. Parada and M. Zorzi, “Context-aware handover in mmwave

5G using ue’s direction of pass,” in 24th European Wireless

Conference, 2018, pp. 1–6.

[101] I. Filippini, V. Sciancalepore, F. Devoti, and A. Capone, “Fast

cell discovery in mm-wave 5G networks with context informa-

tion,” IEEE Transactions on Mobile Computing, vol. 17, no. 7,

pp. 1538–1552, 2018.

[102] Y. Koda, K. Nakashima, K. Yamamoto, T. Nishio, and

M. Morikura, “Handover management for mmwave networks

with proactive performance prediction using camera images and

deep reinforcement learning,” IEEE Transactions on Cognitive

Communications and Networking, vol. 6, no. 2, pp. 802–816,

2020.

[103] M. Alrabeiah, A. Hredzak, and A. Alkhateeb, “Millimeter wave

base stations with cameras: Vision-aided beam and blockage

prediction,” in IEEE 91st Vehicular Technology Conference

(VTC2020-Spring), 2020, pp. 1–5.

[104] Y. Sun, G. Feng, S. Qin, Y. Liang, and T. P. Yum, “Reinforce-

ment learning based handoﬀ for millimeter wave heterogeneous

cellular networks,” in IEEE Global Communications Conference

(GLOBECOM), 2017, pp. 1–6.

[105] A. Alkhateeb, I. Beltagy, and S. Alex, “Machine learning for

reliable mmwave systems: Blockage prediction and proactive

handoﬀ,” in IEEE Global Conference on Signal and Information

Processing (GlobalSIP), 2018, pp. 1055–1059.

[106] Y. Wang, A. Klautau, M. Ribero, A. C. K. Soong, and R. W.

Heath, “Mmwave vehicular beam selection with situational

awareness using machine learning,” IEEE Access, vol. 7, pp.

87 479–87 493, 2019.

[107] M. Alrabeiah and A. Alkhateeb, “Deep learning for mmwave

beam and blockage prediction using sub-6 GHz channels,” IEEE

Transactions on Communications, vol. 68, no. 9, pp. 5504–5518,

2020.

[108] S. Khosravi, H. S. Ghadikolaei, and M. Petrova, “Learning-

based handover in mobile millimeter-wave networks,” IEEE

Transactions on Cognitive Communications and Networking,

pp. 1–1, 2020.

[109] Y. Lu, M. Gerasimenko, R. Kovalchukov, M. ˇ

St˚usek, J. Urama,

J. Hosek, M. Valkama, and E. S. Lohan, “Feasibility of location-

aware handover for autonomous vehicles in industrial multi-

radio environments,” Sensors, vol. 20, Nov. 2020.

[110] G. L. Nemhauser and L. A. Wolsey, Integer and Combinatorial

Optimization. Wiley-Interscience, 1988.

[111] D.-S. Chen, R. Batson, and Y. Dang, Applied Integer Program-

ming: Modeling and Solution. John Wiley & Sons, 2011.

[112] D. Bertsimas and J. Tsitsiklis, Introduction to Linear Optimiza-

tion. Athena Scientiﬁc, 1998.

[113] S. Boyd and L. Vandenberghe, Convex Optimization. Cam-

bridge University Press, 2004.

[114] Y. Nesterov and A. Nemirovski, Interior-point Polynomial Al-

gorithms in Convex Programming. Siam studies in applied

mathematics, 1994.

[115] B. R. Marks and G. P. Wright, “A general inner approximation

algorithm for nonconvex mathematical programs,” Operations

Research, vol. 26, no. 4, pp. 681–683, 1978. [Online]. Available:

http://www.jstor.org/stable/169728

[116] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of

Statistical Learning – Data Mining, Inference, and Prediction.

Springer–Verlag New York, 2009.

[117] R. S. Sutton and A. G. Barto, Introduction to Reinforcement

Learning, 1st ed. MIT Press, 1998.

[118] K. P. Murphy, Machine Learning: a Probabilistic Perspective.

MIT Press, 2012.

[119] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning.

MIT Press, 2016, http://www.deeplearningbook.org.

[120] F. Gebali, Analysis of Computer and Communication Networks.

Springer-Verlag, 2010.

[121] D. Bertsekas and R. Gallager, Data Networks (2nd Ed.), 1992.

Part II

Papers

The papers associated with this thesis have been removed for

http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-172553

Analysis and Optimization

for Robust Millimeter-Wave

Communications

Linköping Studies in Science and Technology. Dissertations.

No. 2111

Cristian Tatino

Cristian Tatino Analysis and Optimization for Robust Millimeter-Wave Communications 2021

FACULT Y OF SCIENCE AND ENGINEERING

Linköping Studies in Scie nce and Technology, Dissertations. No. 2111, 2021

Department of Science and Technology

Linköping University

SE-581 83 Linköpin g, Sweden

www.liu.se

Channel Modeling Techniques for Multiband Vehicle to Everything Communications: Challenges and Opportunities

Article

Jan 2024

Wireless Communications With Reconfigurable Intelligent Surface: Path Loss Modeling and Experimental Measurement

Article

Full-text available

Sep 2020

Reconfigurable intelligent surfaces (RISs) comprised of tunable unit cells have recently drawn significant attention due to their superior capability in manipulating electromagnetic waves. In particular, RIS-assisted wireless communications have the great potential to achieve significant performance improvement and coverage enhancement in a cost-effective and energy-efficient manner, by properly programming the reflection coefficients of the unit cells of RISs. In this paper, free-space path loss models for RIS-assisted wireless communications are developed for different scenarios by studying the physics and electromagnetic nature of RISs. The proposed models, which are first validated through extensive simulation results, reveal the relationships between the free-space path loss of RIS-assisted wireless communications and the distances from the transmitter/receiver to the RIS, the size of the RIS, the near-field/far-field effects of the RIS, and the radiation patterns of antennas and unit cells. In addition, three fabricated RISs (metasurfaces) are utilized to further corroborate the theoretical findings through experimental measurements conducted in a microwave anechoic chamber. The measurement results match well with the modeling results, thus validating the proposed free-space path loss models for RIS, which may pave the way for further theoretical studies and practical applications in this field.

MmWave Vehicular Beam Selection With Situational Awareness Using Machine Learning

Article

Full-text available

Jan 2019

Establishing and tracking beams in millimeter-wave (mmWave) vehicular communication is a challenging task. Large antenna arrays and narrow beams introduce significant system overhead configuring the beams using exhaustive beam search. In this paper, we propose to learn the optimal beam pair index by exploiting the locations and types of the receiver vehicle and its neighboring vehicles (situational awareness), leveraging machine learning classification and past beam training data. We formulate the mmWave beam selection as a multi-class classification problem based on hand-crafted features that capture the situational awareness in different coordinates. We then provide a comprehensive comparison of the different classification models and various levels of situational awareness. Furthermore, we examine several practical issues in the implementation: localization is susceptible to inaccuracy; situational awareness at the base station (BS) can be outdated due to vehicle mobility and limited location reporting frequencies; the situational awareness may be incomplete since vehicles could be invisible to the BS if they are not connected. To demonstrate the scalability of the proposed beam selection solution in the large antenna array regime, we propose two solutions to recommend multiple beams and exploit an extra phase of beam sweeping among the recommended beams. The numerical results show that situational awareness-assisted beam selection using machine learning is able to provide beam prediction, with accuracy that increases with more complete knowledge of the environment.

Feasibility of Location-Aware Handover for Autonomous Vehicles in Industrial Multi-Radio Environments

Article

Full-text available

Nov 2020
SENSORS-BASEL

The integration of millimeter wave (mmWave) and low frequency interfaces brings an unique opportunity to unify the communications and positioning technologies in the future wireless heterogeneous networks (HetNets), which offer great potential for efficient handover using location awareness, hence a location-aware handover (LHO). Targeting a self-organized communication system with autonomous vehicles, we conduct and describe an experimental and analytical study on the LHO using a mmWave-enabled robotic platform in a multi-radio environment. Compared to the conventional received signal strength indicator (RSSI)-based handover, the studied LHO not only improves the achievable throughput, but also enhances the wireless link robustness for the industrial Internet-of-things (IIoT)-oriented applications. In terms of acquiring location awareness, a geometry-based positioning (GBP) algorithm is proposed and implemented in both simulation and experiments, where its achievable accuracy is assessed and tested. Based on the performed experiments, the location-related measurements acquired by the robot are not accurate enough for the standalone-GBP algorithm to provide an accurate location awareness to perform a reliable handover. Nevertheless, we demonstrate that by combining the GBP with the dead reckoning, more accurate location awareness becomes achievable, the LHO can therefore be performed in a more optimized manner compared to the conventional RSSI-based handover scheme, and is therefore able to achieve approximately twice as high average throughput in certain scenarios. Our study confirms that the achieved location awareness, if accurate enough, could enable an efficient handover scheme, further enhancing the autonomous features in the HetNets.

Millimeter Wave Communications With an Intelligent Reflector: Performance Optimization and Distributional Reinforcement Learning

Article

Sep 2021

In this paper, a novel framework is proposed to optimize the downlink multi-user communication of a millimeter wave base station, which is assisted by a reconfigurable intelligent reflector (IR). In particular, a channel estimation approach is developed to measure the channel state information (CSI) in real-time. First, for a perfect CSI scenario, the precoding transmission of the BS and the reflection coefficient of the IR are jointly optimized, via an iterative approach, so as to maximize the sum of downlink rates towards multiple users. Next, in the imperfect CSI scenario, a distributional reinforcement learning (DRL) approach is proposed to learn the optimal IR reflection and maximize the expectation of downlink capacity. In order to model the transmission rate’s probability distribution, a learning algorithm, based on quantile regression (QR), is developed, and the proposed QR-DRL method is proved to converge to a stable distribution of downlink transmission rate. Simulation results show that, in the error-free CSI scenario, the proposed approach yields over 30% and 2-fold increase in the downlink sum-rate, compared with a fixed IR reflection scheme and direct transmission scheme, respectively. Simulation results also show that by deploying more IR elements, the downlink sum-rate can be significantly improved. However, as the number of IR components increases, more time is required for channel estimation, and the slope of increase in the IR-aided transmission rate will become smaller. Furthermore, under limited knowledge of CSI, simulation results show that the proposed QR-DRL method, which learns a full distribution of the downlink rate, yields a better prediction accuracy and improves the downlink rate by 10% for online deployments, compared with a Q-learning baseline.

On The Optimal Number of Reflecting Elements for Reconfigurable Intelligent Surfaces

Article

Oct 2020

This work considers a point-to-point link where a reconfigurable intelligent surface assists the communication between a transmitter and a receiver. The system rate, energy efficiency, and their trade-off are optimized by tuning the number of reflecting elements to be activated and the phase shifts that they apply. Unlike most previous works, the considered resource allocation problem explicitly accounts, at the design stage, for the time and energy that are necessary for channel estimation and for reporting the optimal configuration of the phase shifts to the reconfigurable intelligent surface. Numerical results confirm the optimality of the proposed methods and show the potential gains of reconfigurable intelligent surfaces.

Intelligent Reflecting Surface-Assisted Millimeter Wave Communications: Joint Active and Passive Precoding Design

Article

Oct 2020

Millimeter wave (MmWave) communications is capable of supporting multi-gigabit wireless access thanks to its abundant spectrum resource. However, severe path loss and high directivity make it vulnerable to blockage events, which can be frequent in indoor and dense urban environments. To address this issue, in this paper, we introduce intelligent reflecting surface (IRS) as a new technology to provide effective reflected paths to enhance the coverage of mmWave signals. In this framework, we study joint active and passive precoding design for IRS-assisted mmWave systems, where multiple IRSs are deployed to assist the data transmission from a base station (BS) to a single-antenna receiver. Our objective is to maximize the received signal power by jointly optimizing the transmit precoding vector at the BS and the phase shift parameters used by IRSs for passive beamforming. Although such an optimization problem is generally non-convex, we show that, by exploiting some important characteristics of mmWave channels, an optimal closed- orm solution can be derived for the single IRS case and a near-optimal analytical solution can be obtained for the multi-IRS case. Our analysis reveals that the received signal power increases quadratically with the number of reflecting elements for both the single IRS and multi-IRS cases. Simulation results are included to verify the optimality and near-optimality of our proposed solutions. Results also show that IRSs can help create effective virtual LOS paths and thus substantially improve robustness against blockages in mmWave communications.

Learning-Based Handover in Mobile Millimeter-Wave Networks

Article

Oct 2020

Millimeter-wave (mmWave) communication is considered as a key enabler of ultra-high data rates in the future cellular and wireless networks. The need for directional communication between base stations (BSs) and users in mmWave systems, that is achieved through beamforming, increases the complexity of the channel estimation. Moreover, in order to provide better coverage, dense deployment of BSs is required which causes frequent handovers and increased association overhead. In this paper, we present an approach that jointly addresses the beamforming and handover problems. Our solution entails an efficient beamforming method with a few number of pilots and a learning-based handover method supporting mobile scenarios. We use reinforcement learning algorithm to learn the optimal choices of the backup BSs in different locations of a mobile user. We show that our method provides an almost constant rate and reliability in all locations of the user’s trajectory with a small number of handovers. Simulation results in an outdoor environment based on narrow band cluster mmWave channel modeling and real building map data show the superior performance of our proposed solution in achievable instantaneous rate and trajectory rate.

A Multi-Functional Reconfigurable Metasurface: Electromagnetic Design Accounting for Fabrication Aspects

Article

Aug 2020

In this paper we present the theoretical considerations and the design evolution of a proof-of-concept reconfigurable metasurface, primarily used as a tunable microwave absorber, but also as a wavefront manipulation and polarization conversion device in reflection. We outline the design evolution and all considerations taken into account, from the selection of patch shape, unit cell size, and substrate, to the topology of the structure that realizes the desired tunability. The presented design conforms to fabrication restrictions and is co-designed to work with an integrated circuit chip for providing tunable complex loads to the metasurface, using a commercially available semiconductor process. The proposed structure can perform multiple tunable functionalities by appropriately biasing the integrated circuit: Perfect absorption for a wide range of incidence angles of both linear polarization states, accommodating a spectral range in the vicinity of 5 GHz, with potential also for wavefront control, exemplified via anomalous reflection and polarization conversion. The end vision is for such a design to be scalable and deployable as a practical HyperSurface, i.e., an intelligent multi-functional metasurface capable of concurrent reconfigurable functionalities: absorption, beam steering, polarization conversion, wavefront shaping, holography, and sensing.

Planning mm-Wave Access Networks Under Obstacle Blockages: A Reliability-Aware Approach

Article

Jul 2020

Millimeter-wave (mm-wave) technologies are the main driver to deliver the multiple-Gbps promise in next-generation wireless access networks. However, the GHz-bandwidth potential must coexist with a harsh propagation environment. While strong attenuations can be compensated by directional antenna arrays, the severe impact of obstacle blockages can only be mitigated by smart resource allocation techniques. Multi-connectivity, as multiple mm-wave links from a mobile device to different base stations, is one of them. However, the higher reliability provided by several access alternatives can be fully exploited only if uncorrelated link statuses are guaranteed. Therefore, spatial diversity must be enforced. Moreover, since interposing obstacles can block a link, short access links allow reducing the link unavailability probability. Smart base-station selections can be made once the network is deployed, however, our results show that much better results are achievable if spatial diversity and link-length aspects are directly included in the network planning phase. In this article, we propose an mm-wave access network planning framework that considers base-station spatial diversity, link lengths, and achievable user throughput, according to channel conditions and network congestion. The comparison against traditional k-coverage approaches shows that our approach can obtain much better access reliability, thus providing higher robustness to random obstacles and self-blockage phenomena.

Millimeter Wave Base Stations with Cameras: Vision-Aided Beam and Blockage Prediction

Conference Paper

May 2020

Analysis and Optimization for Robust Millimeter-Wave Communications

Figures

Recommended publications

Robust Secrecy Rate Optimizations for Healthy Monitoring System MISO Channel with D2D Communications

QoS Aware Robot Trajectory Optimization with IRS-Assisted Millimeter-Wave Communications

Performance Analysis of Wireless Networks with Intelligent Reflecting Surfaces

Indoor and Outdoor Physical Channel Modeling and Efficient Positioning for Reconfigurable Intelligen...

Beyond Max-SNR: Joint Encoding for Reconfigurable Intelligent Surfaces