Luận án tiến sĩ tiếng anh Investigations on key technologies for LTE network optimization

In this dissertation, we investigate the adjustment of antenna tilt angle of base station to optimize the performance for LTE networks, including the coverage and capacity optimization a

分类号: TN929.5

密 级: 技术保护一年（2016 年 1 月 1 日—2017 年 1 月 1 日）

U D C: 621.3

学 号 : 129734

INVESTIGATIONS ON KEY TECHNOLOGIES

FOR LTE NETWORK OPTIMIZATION

研究生姓名： PHAN NHU QUAN

导 师 姓 名： 潘志文 教授

申请学位类别 博士 学位授予单位 东 南 大 学 一级学科名称 信息与通信工程 论文答辩日期 2015 年 12 月 16 日

二级学科名称 通信与信息系统 学位授予日期 20 年 月 日 答辩委员会主席 陈明教授 评 阅 人

2015 年 12 月 25 日

iv

LTE NETWORK OPTIMIZATION

A Dissertation Submitted to Southeast University For the Academic Degree of Doctor of Engineering

BY PHAN NHU QUAN

Supervised by Prof PAN ZHI WEN

School of Information Science and Engineering

Southeast University

Dec 2015

・

本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得的研究成果。尽我所知，除了文中特别加以标注和致谢的地方外，论文中不包含其他人已经发表或撰写过的研究成果，也不包含为获得东南大学或其它教育机构的学位或证书而使用过的材料。与我一同工作的同志对本研究所做的任何贡献均已在论文中作了明确的说明并表示了谢意。

研究生签名： 日期：

东南大学学位论文使用授权声明

东南大学、中国科学技术信息研究所、国家图书馆有权保留本人所送交学位论文的复印件和电子文档，可以采用影印、缩印或其他复制手段保存论文。本人电子文档的内容和纸质论文的内容相一致。除在保密期内的保密论文外，允许论文被查阅和借阅，可以公布（包括以电子信息形式刊登）论文的全部内容或中、英文摘要等部分内容。论文的公布（包括以电子信息形式刊登）授权东南大学研究生院办理。・

研究生签名： 导师签名： 日期：

随着无线互联网的快速发展，无线业务量呈指数倍增长，使得服务提供商必须不断地对无线网络覆盖进行优化，提升网络的系统容量，以满足用户的服务需求。为达到上述目标，可以采用的主要技术方法有：修改系统参数设置、收发站开/关、根据负载状况调整发送功率、优化小区布局、依据地形或用户密度调整天馈单元、安装大规模天线、调整天线参数如倾角等。

本论文主要研究基站天线倾角调整算法，以实现包括网络覆盖优化、网络容量提升和网络负载均衡在内的无线网络性能优化。

本论文主要包括以下研究内容：

第一章介绍论文的研究背景和研究意义。介绍了 LTE 网络中的自组织网络技术及其特性，包括自配置、自优化和自愈特性。详细阐述了自优化中的网络覆盖及容量优化(CCO)和负载均衡(LB)优化及其特征，并指出覆盖及容量优化(CCO)和负载均衡(LB)是本文研究的关键问题。此外，还详细介绍了天线方向图和天线倾角调整的基本原理。 第二章研究 LTE 网络的覆盖问题，提出基于 eNB 天线倾角(ATA)调整的网络覆盖优化算法。覆盖优化算法的性能指标是 eNB 覆盖的移动台 (MS)数目，该数目由 MS 测量到的参考信号接收功率(RSRP)决定。本章通过最大化 eNB 覆盖的 MS 数目优化网络覆盖，提出一种基于改进粒子群优化(MPSO)算法的网络覆盖优化算法。在 MPSO 中存在一群粒子，每个粒子对应一组天线倾角集合，适应度函数由被服务的 MS 数目决定，进化速度为每次迭代中 ATA 的调整尺度。仿真结果表明，与固定倾角相比，得益于提出的天线倾角优化算法，基站服务的 MS 数目增加了 7.2%，接收信号质量提升 20dBm，并且系统吞吐量也得到了 55Mbps 的有效提升。

第三章研究 eNB 负载约束及用户速率需求对网络覆盖的影响，提出考虑网络负载约束的网络覆盖优化算法。虽然按照前一章的方法调整 ATA 能够有效提升整个网络覆盖，但在 eNB 负载约束下，一些 eNB 过载导致一些用户的服务无法得到满足。因此，在第三章中，提出考虑网络负载约束的网络覆盖优化算法。定义无线网络的覆盖能力为综合考虑移动台 RSRP 和 eNB 负载约束的被服务 MS 数目，通过优化网络负载约束下被服务的 MS 数目优化网络覆盖。提出一种基于 MPSO 的覆盖优化算法，该算法考虑网络负载约束，通过调整 eNB 的 ATA 来最大化 eNB 服务的用户数。仿真结果表明，得益于提出的算法，每个 eNB 服务的用户数量显著增加，系统吞吐量得到了显著提升，并且网络平均负载和带宽效率也得到了改善。

第四章研究 LTE 网络的负载均衡问题，通过优化 eNB 的 ATA 来实现 LTE 网络的负载均衡。以简氏公平系数作为评价网络负载均衡的标准，本章提出了基于 MPSO 算法的负载均衡算法，通过联合优化 eNB 的 ATA 获得 LTE 网络负载均衡。仿真表明，所

ii

了网络带宽效率。

第五章研究 LTE 网络覆盖和负载均衡的联合优化问题。如果不考虑负载均衡，仅考虑网络负载约束，优化 eNB 的 ATA 可能会出现以下情况：某信道条件较差的用户接入负载较重的 eNB 并在该 eNB 内占用过多的资源，然而该用户附近还有一个低负载eNB 没有得到利用。这种情况会使得网络资源无法得到有效利用，出现小区间负载不均衡的问题。在第五章中为了在改善 eNB 覆盖的同时保证 LTE 网络负载均衡，通过联合考虑覆盖因子 (CF)和负载均衡指标(LBI)来实现 eNB 覆盖和负载均衡的联合优化，其中覆盖因子反映 eNB 的覆盖能力，负载均衡指标通过简氏公平系数进行评估，反应网络负载均衡能力。将覆盖和负载均衡问题联合建模为一个多目标优化函数，提出一种基于MPSO 的 ATA 调整方案。仿真结果表明，所提方案在有效增加网络覆盖的同时，能显著提升负载均衡和网络带宽效率，并且网络吞吐量也得到了有效改善。

关键词：LTE 网络，网络优化，天线倾角，覆盖优化，负载均衡，改进粒子群算法。

The exponential increase in the traffic volume forces the services providers unavoidably facing with constantly evolving the wireless network system to satisfy the user demands, such

as to optimize the coverage of Evolved Node Base Station (eNB), and increase the capacity of the network By changing the system parameters, or switching on/off the base transceiver stations, or adjusting the transmission power, or suitably rearranging cell layout, or replacing antenna elements according to the topographical or the user density in urban or rural areas, or installing massive MIMO (Multiple Input Multiple Output) or adjusting the antenna parameters such as ATA are important ways to achieve the above goals

In this dissertation, we investigate the adjustment of antenna tilt angle of base station to optimize the performance for LTE networks, including the coverage and capacity optimization and load balancing

The main works of this dissertation are follows:

In Chapter 1, the research background and research significance are introduced The organizing networks technologies and its features including self-configuration, self-optimization and self-healing in LTE network are also introduced Self-optimization is detailed including its use cases such as the Coverage and Capacity Optimization (CCO), and Load Balancing (LB) optimization CCO and LB are two key issues in this study Also, the fundamental of antenna pattern and its tilt angle are introduced

self-In Chapter 2, the coverage problem in LTE networks is investigated and a network coverage optimization algorithm based on the Antenna Tilt Angle (ATA) adjustment of the eNBs is proposed The number of Mobile Stations (MS) under the coverage of eNB is determined by the Reference Signal Received Power (RSRP) measured from MS, and is considered as the performance metrics for coverage optimization algorithm In this chapter, the network coverage is optimized by maximizing the number of MS under the coverage of eNBs and a Modified Particle Swarm Optimization (MPSO) based tilt angle adjusting algorithm for coverage optimization is proposed In MPSO, a swarm of particles known as the set of ATAs is available, the fitness function is defined as the total number of the served MSs, and the evolution velocity corresponds to the tilt angles adjustment scale for each iteration cycle Simulation results show that compared with the fixed tilt angles, the number of served MSs by base stations is significantly increased by 7.2%, the quality of received signal is considerably improved by 20 dBm, and particularly the system throughput is also effectively increased by

55 Mbps benefiting from the proposed algorithm

In Chapter 3, the effect of the load constraint and the requirements of MSs is investigated, and a coverage optimization algorithm considering the load constraint of eNBs is proposed Although adjusting the ATA according to Chapter 2 can efficiently improve the network coverage, but under the load constraint, the service requirements of some MSs might not be met because of the overload of the eNBs Therefore, the network coverage optimization

iv

wireless network is defined as the number of served MSs of eNBs considering both the RSRP measured from the MSs and the load constraint of eNBs, and the network coverage is optimized by optimizing the number of served MSs under the constraint of the network load

An MPSO-based coverage optimization algorithm that adjusts the ATAs of eNBs considering the network load to maximize the number of users served by eNBs is proposed Simulation results show that both of the number of served users by each eNB and the system throughput are significantly increased As well, the average load and the bandwidth efficiency of the net-work are improved benefiting from the proposed algorithm

In Chapter 4, we investigate the problem of load balancing optimization The load balance

of the LTE network is achieved by optimizing the ATAs of the eNBs Jain’s fairness index is used to evaluate the load balance of the network An MPSO-based load balancing algorithm is proposed The load balance of the network is achieved by cooperatively optimizing the ATAs

of the eNBs Simulations show that the proposed approach can efficiently improve load balancing, and significantly improves the call blocking rate, the network bandwidth efficiency

In Chapter 5, the joint coverage and load balancing optimization problem is investigated Without consideration of the load balancing, adjusting the ATAs of the eNBs with the constraints of network load may result in the following problem: some users in the poor channel condition access the heavy load eNB and occupy too many resources in the eNB, however, the light load eNB nearby these users will be under used This results in load imbalance problem between eNBs Therefore, to further improve the coverage of eNB, and simultaneously guarantee the even distribution of load in the LTE networks, in Chapter 5, we jointly optimize the coverage of eNB and load balancing by considering the Coverage Factor (CF) and Load Balancing Index (LBI) The coverage factor represents the coverage ability of eNB, and load balancing is represented by load balancing index such as Jain’s fairness index

We formulate the coverage and load balancing problem as a multi-objective optimization function, and an MPSO algorithm based ATAs adjusting scheme is proposed Simulation results show that our proposed algorithm can efficiently increase the network coverage This significantly improves the load balancing, and appreciably increases the network bandwidth efficiency Also, the system throughput is considerably improved benefiting from the proposed algorithm

Keywords: LTE Networks, Network Optimization, Antenna Tilt Angle, Coverage

Optimization, Load Balancing and Modified Particle Swarm Optimization

After four years of effort and hard work, first and foremost, I would like to express my sincere gratitude and my deepest appreciation to those who have given me their support and love

This dissertation would not have accomplished without my supervisor For that reason, I would like to express my sincere gratitude to my main advisor, Prof Pan Zhiwen, for his support, encouragement, assistance, patience, great and wide knowledge, and personal guidance They all have been of great value to me, so I am proud of to have him as my advisor

I would also like to thank all the committee members of my dissertation: Prof Liu Nan, Dr Jiang HuiLin, Dr Bui ThiOanh and Dr Li Pei, for the time they have spent on reading my Ph.D dissertation and their deeply understanding comments

My thankfulness also goes to the National Mobile Communications Research Laboratory members It was an honor to me to be a part of such a wonderful department and community

I would also like to thank and appreciate my beloved parents: my father and my mother for their teachings, encouragement, assistance, support, patience, continued unconditional love, constant source of motivation and prayers for me to become a successful and better person

I am also thankful to my beloved siblings: my brothers for always being my great role models and setting high standards for me I would also like to express my passion to my sisters, for being always loving, kind and supportive I am very happy with having such wonderful family members

Finally, and most importantly, I would also like to thank my wife, Vu thiThuTrang for her understanding, unconditional support, always lifting my spirit, and great patience over all these years Specially, I would like to thank my beloved kids, Phan thiThuyDuong and Phan TrungKien for being funny and lighting up my life I could have never successfully completed

my academic dream without their love

D EDICATION

I would like to dedicate this doctoral dissertation to my parents, brothers, sisters, my wife,

my daughter and my son who have given me their love, encouragement and support throughout, for which my mere expression of thanks likewise does not suffice

vi

摘要 iv

Abstract iii

Acknowledgement v

Dedication v

Table of Contents vii

List of Figures ix

List of Tables xi

List of Abbreviations xiii

Chapter 1 Introduction 1

1.1 Background 1

1.2 Introduction To Self-Organizing Networks 1

1.2.1 Architecture of LTE system 1

1.2.2 Overview of SON 2

1.2.3 Base Station Antenna and ATA 8

1.2.3.1 Introduction 8

1.2.3.2 Antenna Parameters 9

1.2.3.3 Antenna Radiation Pattern 11

1.3 Thesis Motivation 13

1.4 Recent Progresses in Related Areas 14

1.4.1 Coverage and Capacity Optimization 14

1.4.2 Load Balancing 16

1.4.3 PSO algorithm 18

1.5 Contributions 19

1.6 Dissertation Outline 20

Chapter 2 An MPSO-Based Antenna Tilt Angle Adjusting Scheme for LTE Coverage Optimization 23

2.1 Introduction 23

2.2 System Model and Problem Formulation 24

2.2.1 Antenna Down Tilt Angle 25

2.2.2 Path-loss 26

2.2.3 Shadow Fading Model 26

2.2.4 The Number of MSs Served by eNB 27

2.3 MPSO Based ATA Adjusting Algorithm 29

2.3.1 Overview of Particle Swarm Optimization (PSO) 29

2.3.2 MPSO Based ATA Adjusting Algorithm 30

viii

2.5 Conclusions 39

Chapter 3 Coverage Optimization of LTE Networks Based on Antenna Tilt Adjusting Considering Network Load 41

3.1 Introduction 41

3.2 System Model and Problem Formulation 43

3.2.1 The Number of Users Served by eNB with the Constraint of Network Load 43

3.3 MPSO-Based ATA Adjusting Algorithm 46

3.4 Simulation Results 49

3.5 Conclusions 55

Chapter 4 Load Balancing Optimization Based on Tilt Adjusting in LTE Networks 57

4.1 Introduction 57

4.2 System Model 58

4.2.1 System Model 58

4.2.2 Link Model 58

4.3 Problem Formulation 60

4.4 Algorithm 60

4.5 Simulation Result and Analysis 63

4.6 Conclusions 68

Chapter 5 Joint Load Balancing and Coverage Optimizing Based on Tilt Adjusting in LTE Networks 69 5.1 Introduction 69

5.1.1 Handover Procedure 70

5.1.2 Load Balancing Mechanism 70

5.2 System Model 72

5.2.1 System Model 72

5.2.2 Link Model 73

5.3 Problem Formulation 75

5.4 Algorithm 75

5.5 Simulation Result and Analysis 79

5.6 Conclusions 85

Chapter 6: Conclusion and Future Work 87

6.1 Conclusion 87

6.2 Future Work 88

Bibliography 89

Publication 99

PATENT 99

Figure 1 1: Architecture of LTE Network 2

Figure 1 2: Self-Organizing Network Features 3

Figure 1 3: Antenna Radiation Patterns 9

Figure 1 4: Mechanical Tilt [20] 10

Figure 1 5: Electrical Tilt [20] 10

Figure 1 6: Illustration of Tilt Angle 10

Figure 1 7: Modeling of Horizontal Pattern 11

Figure 1 8: Modeling of Vertical Pattern 12

Figure 2 1: System Model 24

Figure 2 2: The Relationship between Antenna Main Lobe and Tilt Angle 25

Figure 2 3: Fundamental of antenna tilt angle calculation 25

Figure 2 4: The Simulation System 35

Figure 2 5: Comparison of served MSs number and antennas of eNBs (a) 0; (b) 6; (c) 16; (d) with tilts adjusted by MPSO 36

Figure 2 6: CDF of MSs RSRP 36

Figure 2 7: (a) User’s SINR with Fixed Tilt 6; (b) Users’ SINR with Adjusted Tilt by MPSO; (c) CDF of Users’ SINR 37

Figure 2 8: The Convergence of Solution 37

Figure 2 9: (a) Users’ Throughput; and (b) System Throughput 37

Figure 2 10: RSRP Distribution of Scenario after Adjusting ATA 38

Figure 2 11: SINR Distribution of Scenario after Adjusting ATA 38

Figure 3 1: System Model 43

Figure 3 3: The Simulation System (7 eNBs at the center of 19 wrap-around eNBs) 51

Figure 3 4: The Served User Number Before and After Adjusting ATA without and with Considering the Network Load 51

Figure 3 5: CDF of Users’ RSRP 52

Figure 3 6: CDF of Users’ SINR 52

Figure 3 7: The Convergence of Solution 52

Figure 3 8: Users’ Throughput and System Throughput 53

Figure 3 9: The Average Load of the Network 53

Figure 3 10: The Bandwidth Efficiency of the Network 53

Figure 3 11: RSRP Distribution of Scenario after Adjusting ATA 54

Figure 3 12: SINR Distribution of Scenario after Adjusting ATA 54

Figure 4 1: System Model 58

Figure 4 2: The Simulation System of 19 wrap-around cells 65

Figure 4 3: Effect of Arrival Rate on Average Load 66

Figure 4 4: The Relationship between Arrival Rate and Bandwidth Efficiency 66

Figure 4 5: The Relationship between Arrival Rate and Load Balancing Index 67

Figure 4 6: The Relationship between Arrival Rate and Block Probability 67

Figure 4 7: The Relationship between Arrival Rate and Total Users 68

Figure 5 1: Operational Principle of Adjusting CIO to Handover User 71

Figure 5 2: Operational Principle of Adjusting ATA to Handover User 72

Figure 5 3: System Model 73

Figure 5 4: The Simulation System (7 eNBs at the center of 19 wrap-around eNBs) 82

Figure 5 5: Effect of α on Call Blocking Rate 82

x

Figure 5 7: The Effect of α on LBI 83

Figure 5 8: Users’ Throughput and System Throughput 83

Figure 5 9: Average Load of System 83

Figure 5 10: Bandwidth Efficiency 84

Figure 5 11: Coverage Factor 84

Figure 5 12: Load Balancing Index 84

Table 1 1: Example of Self-Configuration 4

Table 2 1: The pseudo-code of algorithm 32

Table 2 2: Setting of the System Parameters 33

Table 3 1: The Operation of the Algorithm 48

Table 3 2: System Simulation Parameters 49

Table 4 1: The Operation of the Algorithm 62

Table 4 2: Setting of the System Parameters 64

Table 5 1: The Operation of the Algorithm 77

Table 5 2: Setting of the System Parameters 79

xii

2D 2 Dimensional

3D 3 Dimensional

3G the Third Generation (Cellular Systems)

3GPP the 3rd Generation Partnership Project

4G the Fourth Generation

AAS Adaptive Antenna System

ATA Antenna Down-Tilt Angle

BSC Base Station Controller

BTS Base Transceiver Station

CAPEX Capital Expenditure

CBR Call Blocking Rate

CCO Coverage and Capacity Optimization

CDF Cumulative Distribution Function

CDR Call Dropping Ratio

CF Coverage Factor

CP Cyclic Prefix

DAS Distributed Antennas Systems

DL Down-Link

eNB Evolved Node Base Station

EPC Evolved Packet System

E-UTRAN Evolved UMTS Terrestrial Radio Access Network FDD Frequency Division Duplexing

FFT Fast Fourier Transform

GBR Guaranteed Bit Rate

GERAN GSM EDGE Radio Access Network

GSM Global System for Mobile Communications HARQ Hybrid Automatic Retransmission Request

xiv

HO Handover

HPBW Half Power Beam-Width

HSS Home Subscriber Server

IP Internet Protocol

ISD Inter-Site Distance

ISI Inter-Symbol Interference

LB Load Balancing

LHCP Left Hand Circular Polarization

LBI Load Balancing Index

LMR Land Mobile Radio

LTE Long Term Evolution

LTE-A Long Term Evolution-Advanced

MDT Mechanical Down-Tilt

MIMO Multiple Input Multiple Output

MLB Mobility Load Balancing

MME Mobility Management Entity

MRO Mobility Robustness Optimization

MPSO Modified Particle Swarm Optimization

NMS Network Management System

OAM Operation Administration and Maintenance OFDM Orthogonal Frequency Division Multiple

OFDMA Orthogonal Frequency Division Multiple Access OMC Operation and Maintenance Center

OPEX Operation Expenditure

QAM Quadrature Amplitude Modulation

QPSK Quadrature Phase-Shift Keying

PCRF Policy charging and Rules Function

P GW Packet Data Network Gateway

PRB Physical Resource Block

QoS Quality of Service

RACHO Random Access Channel Optimization

RAT Radio Access Technology

RET Remote Electrical Tilt

RNC Radio Network Controller

RHCP Right Hand Circular Polarization

RSRP Reference Signal Received Power

SAE System Architecture Evolution

SC-FDMA Single Carrier Frequency Division Multiple Access S-GW Serving Gateway

SGSN Serving GPRS Support Node

SINR Signal to Interference plus Noise Ratio

SON Self-Organizing Network

TDD Time Division Duplexing

TNL Transport Network Load

xvi

C HAPTER 1 I NTRODUCTION

1.1 Background

LTE is the next step in the GSM evolutionary path beyond 3G technology, and it is strongly positioned to be the dominant global standard for 4G cellular networks LTE also represents the first generation of cellular networks to be based on a flat IP architecture and is designed to seamlessly support a variety of different services, such as broadband data, voice, and multicast video Its design incorporates many of the key innovations of digital communication, such as MIMO and OFDMA, that mandate new skills to plan, build, and deploy an LTE network [1]

LTE has key features as follows [2]:

(1) Flexible bandwidth: 1.4/3/5/10/15/20 MHz;

(2) Support FDD and TDD

(3) 4x4 MIMO, multi-user collaborative MIMO;

(4) Beam forming in the downlink;

(5) Modulation: OFDM with QPSK, 16 QAM, 64 QAM;

(6) OFDMA downlink, SC-FDMA uplink;

(7) HARQ transmission;

(8) Short frame sizes of 10ms and 1ms, faster feedback and better efficiency at high speed; (9) Persistent scheduling to reduce control channel overhead for low bit rate voice transmission;

(10) IP based flat network architecture

1.2 Introduction To Self-Organizing Networks

1.2.1 Architecture of LTE system

The great challenge of the limits of current 3G networks are the trend of ever increasing transmission bandwidths, hence, in 2005 the Third Generation Partnership Project (3GPP) standardization body decided to start work on next generation wireless network designs that are only based on packet-switched data transmission In order to guarantee the competitiveness of 3G for the next 10 years and beyond, LTE is the latest version in the mobile network technolo-

gy tree that is being implemented within the 3GPP Time Division Duplex (TDD) and quency Division Duplex (FDD) both are supported by LTE Nevertheless, LTE also supports a flexible and scalable bandwidth e.g., 1.25, 5, 10 and 20MHz LTE also has a very flexible radio interface [3, 4]

Fre-2

In order to differentiate NodeB from Universal Mobile Telecommunication System (UMTS), the base station of LTE is designated to as Evolved Node Base station (eNB) per 3GPP standard eNBs are designed more to be more intelligent than NodeB by removing Radio Network Controller (RNC), transferring the functionality to eNB, and partly to the core net-work gateway Through X2 interface, eNB can also carry out handovers as well they can com-municate with each other directly Through S1 interface, eNB connects to the gateway nodes i.e., connection between radio network and core network based on Internet Protocol (IP) Serv-ing Gateway (S-GW) and the Mobility Management Entity (MME) are two entities of the gateway between core network and radio access networks MME is the Control Plane (C-plane) entity which is mainly in charge of session management signaling, mobility of User Equipment (UE), selection of a gateway to the internet when mobile requests the IP address from the net-work and the location tracking of mobile devices S-GW is in charge of User Plane (U-plane) Both components can be implemented on the same hardware or separated S11 interface is used

to communicate between MME and S-GW when we implement components separately LTE network architecture is illustrated in Fig.1.1

Figure 1 1: Architecture of LTE Network

1.2.2 Overview of SON

With the continuous development of wireless access to the internet, it needs to have a competitive advantage in the mobile operations A very promising approach is to maximize the performance of the whole system, that not only provides better wireless access performance, but also responds more effectively to network operation and maintenance [5] Self-Organizing Network (SON) is a long-term process of improving 3GPP (Third Generation Partnership Project) network operation and a key factor to maintain its main objectives such as effectively reducing operating expenses, improving network coverage, resource utilization and quality of service As every mobile network, LTE system also needs to manage However, LTE network itself will inevitably lead to the complexity of the operation and maintenance the network and

of new requirements: SON will open up a new path in the effective cost-saving issue for operator

Self-organizing network technology means to sense the surrounding environment changes

at any time and to make the appropriate action technology Self-organizing network technologies include three areas: self-configuration, self-optimization and self-healing [6]

Figure 1 2: Self-Organizing Network Features

We can see from the Fig.1.2, self-organizing network technology works throughout the entire network process, starting from the base station power has been on to the network stable operation and abnormal situation

In Self-configuration process, when a new base station is added to an existing network, the system automatically processes to configure a series of parameters during the installation, so that it can adapt to run in the current network environment This is the preparation stage, i.e

4

from the base station power on and the backbone network with a basic connection start, direct

to the RF wireless transceiver The self-configuration process consists of two components: the basic settings and initial wireless parameter settings

When the system is stable, in order to adapt to the changes in network conditions, some system parameters have to be necessarily optimized This is part of the task mainly done by self-optimization process Optimization process uses the measurement report of the UE and eNB and some network performance metrics of the actual operation to automatically adjust the parameters

The self-healing process is performed during the operation of the network It automatically detects and locates the fault, and then uses the self-healing optimization algorithms to solve these problems, such as increasing the transmission power to solve the coverage gaps, capacity not enough problems

Through self-configuration process, the majority of the above work can be done automatically (see Table 1.1)

Table 1 1: Example of Self-Configuration

The self-configuration process of base stationWhen the network is placed a new eNB

(1) Assign an IP address to the newly installed eNB, assign an DNS (Domain Name System) server address etc., and OAM (Operation Administration and Maintenance) configuration subsystem information is sent to eNB;

(2) give this new eNB configure a gateway GW, so that it can interact with other IP packet between network nodes;

(3) This new eNB requires own the hardware configuration, the cell type etc., information is sent to the OAM control center to authenticate certification; download from the OAM some

of the necessary software and configuration data;

(4) Data transmission and wireless configuration follows the base station configuration;(5) The new eNB connection to OAM control center in all areas, and contact with other management functions;

(6) Establishing a connection between S1 and X2 interface

In this configuration process, network engineers only need the cable connection of eNB then turn on power

in an outage state, this problem is called cell outage [9] This problem may occur in the radio layer, or it may also occur in the transport layer Cell outage management mainly includes three aspects: cell outage prediction, cell outage detection and cell outage compensation Cell outage prediction is mainly by observing some of the early warning signal to shorten the time

to monitor cell outage, and begin the preparatory work for cell outage compensation Cell outage detection is major to confirm and trigger cell outage compensation for the happened cell outage with an appropriate compensation method Cell outage prediction and cell outage compensation is through analysis and correlation of the network measurement reports (including eNB measurements, user measurements and OAM measurements) to make decision Cell outage is compensated by selecting a number of neighboring cells to reduce the burden of cell outage, or by coordinating with a number of other self-optimizing algorithms (such as load balancing, handover optimization and coverage optimization, etc.) to reduce the quality of the user experience For example, in the area covered by multiple cells, we can adjust the coverage

Self-optimization means to automatically sense the changes in the surrounding environment, automatically adjust the radio relevant parameters to make the network work in the best condition Self-optimization includes Coverage and Capacity Optimization (CCO), Random Access Channel Optimization (RACHO), Energy Saving Optimization, Mobile Robustness Optimization (MRO) and Mobility Load Balancing Optimization (MLBO), etc Following are specific description some use cases

1.2.2.3.1 Coverage and Capacity Optimization

An important task of network operation phase is network coverage and capacity optimization [11] In the network planning phase, operators mainly use planning tool to plan the coverage and capacity In the practical operation of the network, it is necessary to constantly measure the network performance, and analyze the demands of users Among them, the call dropping rate can be used to indicate whether an area has sufficient network coverage, while the traffic counter is mainly used to detect whether the network has a capacity problem

In the LTE system, according to the operator's requirements, the users can at least establish and maintain acceptable, or at least to be connected and receive the default quality of service For network coverage and capacity issues, the traditional method is through the link test to collect data, and then analyze to find solutions with the planning tools In SON, this use case automatically discovers the coverage and capacity issues and makes optimization, which can reduce the workload of network engineers to improve efficiency of the system through measuring and reporting user information to base station

The main purposes of coverage and capacity optimization include: (1) optimize network coverage; (2) maximize system capacity, improve the spectrum resource utilization; (3) provide

a continuous cell coverage and improve the quality of user experience; (4) reduce the

interference between cells; (5) monitor the cell edge performance; (6) reduce the number of drive test; (7) co-work with self-healing, such as when a base station detect fails, through surrounding base stations can be reconfigured and the fault automatically recover

The coverage and capacity optimization can be done based on: (1) the user reports through the signal measurement from the serving cell and neighbors cells; (2) signaling and reporting burden on users; (3) network timing advance (TA) parameter; (4) radio link failure counter; (5) coverage trigger-based mobile counter; (6) monitoring of the load distribution of the network

by analysis of these statistics; (7) downlink transmission power adjust; (8) downlink reference signal power offset; (9) antenna tilt angle, etc

1.2.2.3.2 Load Balancing Optimization

With the rapid development of mobile communications, and because of the spatial and temporal distribution of traffic, load imbalance between cells occurs Load balancing is to try

to balance the load of the entire network as much as possible Hence, user of congested cells will be assigned or transferred to neighboring light load cells

Load balancing consists of three parts: (1) monitoring and reporting the load information; (2) adaptive handover or reselection parameter configuration; (3) handover-based load balancing

1.2.2.3.2.1 Monitoring and Reporting the Load Information

This is mainly to detect the load of cell and the interaction between relevant information through the X2 interface (intra-LTE scene) or Sl interfaces (inter-RAT scene) Load information includes [4]:

• Intra-LTE scene

 Radio resource utilization rate (UL / DL GBR (Guaranteed Bit Rate) PRB utilization rate, UL / DL non-GBR PRB utilization rate, UL / DL all PRB utilization rate);

 Hardware load indicator (UL / DL hardware load: low, medium, high, overload);

 TNL (Transport Network Load) load indication (UL / DL TNL load: low, medium, high, overload);

 (Optional) cell capacity level values [UL / DL relative capacity indication, when the cell capacity is mapped to this value, E-UTRAN, UTRAN and GERAN (GSM EDGE Radio Access Network) should adopt the same standard];

 Capacity value (to perform load balancing UL / DL capacity are obtainable, to occupy as percentage of the total cell capacity)

• Inter-RAT scene

 Cell capacity level value (UL / DL relative capacity indication, when the cell capacity is mapped to this value, E-UTRAN, UTRAN and GERAN should adopt

8

the same standard);

 Capacity value (to perform load balancing UL / DL capacity are obtainable, as percentage of the total cell capacity) Here capacity value represented the obtainable E-UTRAN capacity; according to availability monitoring, a cell can obtain load capacity

When a node monitor cell load exceeding the load threshold, it will trigger an event-based inter-RAT load report

1.2.2.3.2.2 Adaptive handover or reselection parameter configuration

The main function is to change handover parameters and reselection parameters in the target cell After initializing load balancing, the source cell needs to determine whether adjustments are needed to change parameters in the source cell or a target cell If parameter adjustment values of the source cell and the target cell do not match, the source cell will start the mobile parameter negotiation process

The source cell notifies to the target cell a new mobility parameter setting and tells the reason parameter changes (load balancing) Handover trigger parameter is the control parameters of the handover preparation process, is also a cell-specific offset Reselection parameter settings should be consistent with the change of handover parameters When the target cell receives the information sent by the source cell, it executes analysis If it is within the acceptable range, the target cell adjusts corresponding to handover parameter, and transmits

a completion message to the source cell; if the handover parameters excess the scope of the target cell, coordinated failure information is sent to the source cell The source cell executes a new parameters adjusting process based on the return message

1.2.2.3.2.3 Handover-Based Load Balancing

When the source cell receives load information, it will initialize the handover, some users will be transferred to neighboring cells, the target cell will initiate admission control for a load balancing handover

1.2.3 Base Station Antenna and ATA

1.2.3.1.Introduction

Antenna parameters selection and optimization play an important role in achieving maximum capacity, coverage performance and LB in LTE and LTE-A Dipole and monopole antennas are commonly used for wireless mobile communication systems [12-16] Due to the broadband characteristics and simple construction of dipole antenna, an array of dipole elements is extensively used at the eNB of a cellular radio network

The antennas of eNB which have adaptable parameters such as RET (Remote Electrical Tilt) and MIMO mode, called AAS (Adaptive Antenna System), will become an integral part

of LTE eNB for providing better system performance and radio network capacity

(c) Circular Array (Uniform) Antenna (d) Linear Array (Uniform) Antenna

Figure 1 3: Antenna Radiation Patterns

1.2.3.2.Antenna Parameters

Although there are variety of antenna types and geometries, all antennas can be described

by a small set of main parameters, such as antenna azimuth and tilt [17-19]

• Antenna Azimuth: Antenna azimuth is defined as the direction, in degrees referenced to true north, that an antenna must be pointed The angular distance is measured in a clockwise direction

• Antenna Tilt: Antenna tilt is defined as the angle of the main beam of the antenna below the horizontal plane Negative and positive angles are also referred to as up-tilt and down-tilt respectively [20] Antenna down-tilt can be adjusted mechanically and/or electrically as shown in Fig.1.4 and Fig.1.5 respectively

10

 Horizon

Main beam eNB

Antenna tilt angle



Figure 1 4: Mechanical Tilt [20]

There are different existing types for electrical tilt such as RET (Remote Electrical Tilt), VET (Variable Electrical-Tilt) and fixed electrical tilt Usage of RET antennas removes the need for tower climbing and base station site visits by controlling electrical tilt angle by NMS (Network Management System) so that operational cost is saved On the other hand, MDT (Mechanical Down-Tilt) is also needed because the mechanical tilt range is larger than the electrical tilt range

Horizon

Main beameNB

Antenna tilt angle



Figure 1 5: Electrical Tilt [20]

Adjusting antenna tilt by changing the characteristics of signal phase of each element of the antenna is briefly described as follows [21] (see Fig 1.6)

Figure 1 6: Illustration of Tilt Angle

A phased array is an example of N-slit diffraction It may also be viewed as the coherent addition of N line sources Since each individual antenna acts as a slit, emitting radio waves, their diffraction pattern can be calculated by adding the phase shift φ to the fringing term Define the length of antenna element:

sin s

where, d and  s are the distance between antenna elements and tilt angle, respectively

The relationship between the length of antenna element and the operation wavelength can

where,  and are the deviation of phase and the operation wavelength, respectively

The deviation of phase is then calculated as:

1.2.3.3.Antenna Radiation Pattern

Figure 1 7: Modeling of Horizontal Pattern

Figure 1 8: Modeling of Vertical Pattern

Extrapolation of the 3 dimensional (3D) pattern from two perpendicular cross-sections, azimuth and elevation patterns, is defined in [22] as shown below:

The term Self-Organizing Network (SON) is generally taken to mean the tasks of configuring, operating, and optimizing in wireless networks are largely automated [29, 30], and

is an important feature of LTE and beyond The SON aims to leapfrog to a higher level of automated operation in mobile network Network operators expect SON to reliably reduce their operational and capital expenses while improving the network’s QoS (Quality of Service) The SON use cases of major interest include the Coverage and Capacity Optimization (CCO) [31-34] and Mobility Load Balancing (MLB) [35-42] Optimal coverage requires that in an area where LTE system is provided, users can establish and maintain connections with acceptable or default service quality, according to operator’s requirements Therefore, it implies that the coverage is continuous and users have no knowledge of cell borders The coverage must be guaranteed in both idle and active mode for both Up-Link (UL) and Down-Link (DL) While coverage optimization has higher priority than capacity optimization in Rel-9, the coverage optimization algorithms must take the impact on capacity into consideration Since coverage and capacity are linked, a trade-off between the two of them may also be a subject of optimization The objective of MLB is to deal effectively with the unequal traffic load between cells and to minimize the number of handovers and redirections needed to achieve the load balancing [43]

Antenna Tilt Angle (ATA) is investigated because of its effectiveness in CCO and Load Balancing (LB) [44-47], and adjusting ATA is one of ways to effectively perform CCO and LB However, the effects of tilt angle on CCO and LB have not been comprehensively investigated The aforementioned challenges give rise to a motivation to design feasible CCO and LB schemes for wireless networks The design specifications to overcome the challenges may be different from each other In this dissertation, we will highlight our recently developed schemes for CCO and LB

14

The research described in this thesis focuses on CCO and LB technologies for LTE networks The LTE network standard provides a very flexible self-organizing including self-healing, self-configuring and self-optimizing, but all of them present a problem of how to ensure the fairness between users when they receive the services from service providers with the ever increasing quantity of user population Also, the antenna tilt plays a key role in optimizing the CCO and LB in the LTE networks Since adjusting the tilt angle of the antenna

is not standardized, there are always presented with the question of, “How good is a particular adjustment of tilt angle in a given set of circumstances” The research proposed in these documents aim at addressing this question

1.4 Recent Progresses in Related Areas

1.4.1 Coverage and Capacity Optimization

Service providers expect SON to undoubtedly reduce their CAPEX and OPEX while improving the network’s QoS and the end users’ quality of experience Fundamental work for SON has been done within the framework of the Socrates project [48] They proposed frameworks, requirements, and algorithms for SON

One of the main interests of SON use cases is the CCO Researchers in academia and industry have already proposed a larger number of solutions and concepts for solving the CCO use case [33, 49] Conventional manual procedures for CCO are intricate and time consuming due to the increasing complexity of wireless cellular networks

Recently, a few people have been investigating the CCO In [50], a framework about CCO optimization problem for single tier model is introduced Authors considered a pixel-based model for an LTE network consisting of cells where each pixel represents a potential user location A mathematical model for user traffic in coverage and capacity optimization was used, including relative user density map, general definitions (such as the average number of users per pixel, the area covered by cell, the number of physical resource blocks), optimization functions based on cell-specific measures and optimization functions based on network-wide measures For the optimization objective is a well-known non-convex function and a typical nondeterministic polynomial time problem, people usually concentrate on an intelligent algorithm to obtain a local optimal solution Some promising artificial intelligence techniques such as genetic programming [51], fuzzy q-learning [52] and reinforcement learning [53] is used to solving CCO problem in single tier model In [51], the fitness function is the measurement traces, such as statistics of load, rejected calls, handovers collected whilst the algorithm is being run on the network When having specified the functions and terminals and the fitness function, an initial population of algorithms is firstly created through random combinations of functions and terminals These programs are then tested individually in a network scenario, and the fitness of the algorithms is calculated using the fitness function Then, “parent” programs are selected based on their calculated fitness and genetic operations

such as mutation and crossover are performed on them These genetic operations create new

“child” programs that retain some characteristics of the parents that they were derived from This process of parent selection and reproduction is then repeated for a set number of generations or computation time, or when a program that is good enough has been found, it performs the evolution in an online manner Since q-learning cannot handle problems where the state or action space is continuous, a fuzzy q-learning combining the fuzzy logic with q-learning is proposed in [52] to overcome the disadvantage of q-learning and solve CCO problem in single tier model The states are defined as including the current antenna down-tilt, the mean spectral efficiency and the edge spectral efficiency The actions are defined as the change to be applied to the current tilt value The policy is defined as the mapping the state to action, fuzzy q-learning controller try to learn this policy The membership functions are defined as the degree of membership of the continuous variable to a specific label, which is the state vector component within [0, 1] The cooperative rewards, which used to combine the center spectral efficiency with the edge spectral efficiency Finally, a rule-based Fuzzy Inference System (FIS) gives the rules based inference In each iteration cycle, the learning agent identifies its current state based on the degree of truth of each FIS rule, and update q-values To speed up the learning process in fuzzy q-learning, [53] proposed a reinforcement learning scheme, this method is totally based on fuzzy q-learning and the learning have strategies [54] extended previous work [34] on effective techniques to address the CCO SON use case In a scenario with uniform user density, they predefined 9 basic beams, such that one beam points to the upper left, one to the upper middle, one to the upper right, and one to the middle left, and so on To learn which is the best subset of the 9 basic beams that should be used, they used the modified q-learning, i.e., learn the optimized beam configuration vector that minimizes the joint coverage and capacity cost, a basic beam either to be on or off (0 or 1) which corresponds to an action in classical q-learning The antenna array is the player in the proposed modified q-learning, the antenna array decides to switch on or off the basic beams which lead to the minimum cumulative cost As bacterial foraging algorithm, a new evolutionary computation technique is introduced in [55] The foraging behavior of bacterial foraging algorithm can be explained by four processes, which are named chemo taxis, swarming, reproduction, and elimination and dispersal To balance coverage and capacity and maximize the throughput, [56] proposed a joint optimization model for a random user distribution in heterogeneous cellular networks The optimization objective is to meet coverage required to choose antenna electronic down-tilt, transmission power and bandwidth as adjusting variable, and consider the probability distribution of users by dividing girds

To optimize the system throughput and coverage Berger et al [57] proposed the antenna tilt-based SON algorithm A general concept for the self-organization of multiple Key Performance Indicators (KPIs) while having only very sparse system knowledge is introduced

An effective tilt-based algorithm that manages to jointly optimize coverage and capacity in downlink and uplink, and using Simulated Annealing (SA) to obtain an upper bound of the KPI performance (KPIs represent the network coverage and network capacity) Based on the

16

proposed method of [57], the authors of [58] proposed an efficacious modification of the antenna tilt-based SON algorithm, by using a PDF (Probability Distribution Function) of throughput measurements and an estimate of the number of covered and uncovered users of each cell considered for optimization, and constructed an objective function that is better suited

to jointly maximize throughput and coverage The CCO is also handled by using reinforcement learning strategies of one-cell-per-snapshot and all-cell-per-snapshot (cluster strategy) in [47] The CCO problem is also solved by employing multi-level random Taguchi’s method through adjusting the tilt angle of antenna [31] The authors of [33] considered cells that are permanently switched off for CCO evaluation, their work might also be related to the context

of self-healing Particularly, by tuning transmission power and antenna down-tilt, they considered a subset of optimization parameters that is supposed to be the most effective not only for CCO [10] but also for self-healing purposes as well [11] To reduce the complexity of using 2-dimensional antenna arrays for concurrent CCO, [54] proposed a Nelder-Mead and a Q-Learning-based approach to adjust the remaining parameters, i e vertical and horizontal angles of every basic beam, and considered only a limited set of basic beams which flexibly combine with each other in order to obtain the overall beam pattern of the actual transmission The usage of 2D antenna arrays can considerably improve the performance of joint CCO However, this work did not facilitate a multi-user MIMO approach [33] presented a traffic-light-related scheme for autonomous self-optimization of tradeoff performance indicators in LTE multi-tier networks A traffic-light-related control mechanism automatically triggers reconfiguration of eNB transmission parameters to improve system performance, if necessary But the related computational complexity is too high

1.4.2 Load Balancing

Load balancing provides a cost-effective, efficient method to increase the flexibility and availability of networks [59, 60] Load vector minimization based LB method based on load vector (load vector is a vector whose elements is the load values of cells and sorted in descending order) is used for LB problem for LTE downlink network [35] Call blocking rate defined as the QoS metric to be guaranteed during LB process, the capacity of LB and call blocking rate with different LB patterns based on CoMP (Coordinated Multiple Points) are evaluated and analyzed in an area with a moderate dense distribution of high-loaded base stations, and a LB strategy based on CoMP transmission/reception among the base stations were proposed in [37] In order to solve the problems of local network congestion and the waste of resource due to the inhomogeneity of traffic distribution in LTE network, a mobility

LB algorithm based on handover optimization was proposed in [42]

Since both of MLB and MRO (Mobility Robustness Optimization) adjust the handover parameters at the same time to achieve their objectives independently, conflicts may occur when applying MLB and MRO simultaneously To solve the conflict problem between LB and handover optimization mentioned in [61], a novel coordination algorithm for LB and handover optimization functions has been proposed in [62] In [62], since the conflict between LB and

HOO (Handover Optimization) functions caused by a lack of coordination in the adjustment of the HOM (Handover Margin) parameter, the coordination scheme was proposed When simultaneously operates LB and HOO, by considering the call dropping rate to achieve the high values of HOM, the conflict problem is solved In order to improve user’s QoS while solving the conflict problem between MLB and MRO (Mobility Robustness Optimization), [38] proposed a scheme to adjust the handover parameters based on the load distribution of the serving cell and the target cell When some cells are overloaded, the proposed scheme is triggered to reduce the overloaded cell’s load level rapidly through adjusting the handover parameter The handover parameter should be adjusted to a reasonable value to not cause the handover problems

In [63], LB problems are formulated as game models through down-link power modification Using Game theory, the LB problem is also studied in [64], where each cell independently makes decision on the volume of load to maximize its individual utility in an uncoordinated way In [65], the traffic load was balanced by changing handover parameters considering the capacity available in the neighboring cells of the heavy load cell A new relay-assisted scheme is presented in [66], which uses relay node power tuning as a tool to address the problem of load balancing By using a proposed guided local search heuristic algorithm to find the best relay node power configuration, the load balancing with minimal overall power consumptionwhile meeting the users’ data rate requirements is obtained A method for load estimation based on Signal-to-Interference-plus-Noise Ratio (SINR) prediction and UE measurements after handover occurs is presented in [67] by optimizing the offset value

to make the users be handed over to the target cell An autonomic flowing water balancing method is proposed in [68], and new modules are added in eNBs to detect the overload conditions and trigger handover actions

Some works formulate LB behaviors of rational users as game-theoretic models [63, 72] In [69], a downlink power modification-based LB is formulated as a Fisher game model and a linear pricing technique is used to specify the behaviors of the eNBs to achieve a more desirable equilibrium point The LB issue is studied using a PRB game-theoretic approach in [63], where each eNB independently decides on the amount of load to maximize its individual utility in an uncoordinated way In [72] and [71], two game-theoretic models were used to analyze the behaviors of point-to-point LB To handle the Ping-Pong and slow-convergence problems of the conventional MLB, [73] presented a game-theoretic solution to the LTE SON

69-LB problem The proposed solution is referred to as zone-based mobility load balancing The solution carries out the LB action from the perspective of the multi-cell region, which means that the participants of the LB process are extended from a pair of cells to a zone of cells The average blocking probability and the number of unsatisfied users of the system is reduced, and the cell throughput is guaranteed

As stated in [74, 75], the antenna tilt of cell (cell and eNB are interchangeable) has an important role in reducing or expanding the coverage ratio of the cell, and also has a potential

18

impact on LB [57] proposed a joint down-link and up-link tilt-based coverage optimization scheme based on the sparse system knowledge to increase the cell edge user throughput while simultaneously decreasing the number of uncovered users To control the LB process based on the impact on network performance, a weighted KPI is used consisting of three performance metrics: the call blocking ratio (CBR), call dropping ratio (CDR) and the user bit rate The KPI can be calculated after each tilt change and will reflect the variations in network performance

in order to improve the GoS (Grade of Service) in a congested cell The value of the KPI can

be compared to a pre-defined threshold in order to determine if the tilt change is accepted or not [75] Kazuhide Toda at el of [39] compared between LB-based tilt and LB-based handover parameter, they simulated and evaluated the LB under conditions, such as one source and target cells exist, user distribution is uniform, and UEs have the same amount of traffic load In order

to improve cell edge user throughput performance, the baseline system with fixed down-tilt and the cell-specific antenna down-tilt were introduced in [76], load balancing methods based on cell association algorithms as well as antenna down-tilt control schemes were proposed Both round robin and proportional fairness are evaluated for the scheduling policy The authors of [77] evaluated the potential gain that can be achieved from tilt adaptation based on the traffic situation in a network By proposing a traffic driven tilt adaptation optimization scheme, in which the user traffic is modeled in a pixel based approach, where a pixel is defined as a potential location of users In that way, the network area is divided into a pixel grid of 5m resolution, where a pixel has an area of 25m2 Taguchi’s optimization method based on nearly orthogonal array is employed to find the best tilt combination of number of antennas found around the Hot Spot traffic situations in an LTE-A network All the above works focused either

on LB or on maximizing coverage ratio How to jointly optimize the LB and coverage ratio through adjusting Antenna Tilt Angle (ATA) is still an open issue

In summary, LB and coverage optimization are two essential techniques in LTE networks

to boost the user experience and improve the system performance [57, 74, 78] Under traditional antenna configuration scheme without LB and coverage optimization, each cell is assigned a fixed antenna tilt and each user selects the cell with the highest received power as its serving cell [74] On one hand, this may lead to unbalanced traffic load among the cells and hence congestion to the cells with large amount of users while the PRB in the cells with few users are not efficiently used; on the other hand, this may cause low coverage ratio due to the fixed antenna tilt Furthermore, the LB optimization always force users to access the cells with large amounts of spare PRB, which may further harm the received signal quality of the users and degrades the coverage ratio of the network Therefore, to efficiently use the network resource and guarantee the basic service of users, it is indispensable to jointly consider the LB and coverage optimization in the LTE networks In the aforementioned literatures, coverage problem caused by the LB was not taken into consideration, and only few people focus on jointly solving LB and coverage problem

1.4.3 PSO algorithm