1 Edge Computing Enabled Smart Cities: A Comprehensive …san Kazmi, Tri Nguyen Dang, Choong Seon Hong, Senior Member, IEEE Abstract—Recent years have disclosed a remarkable proliferation

1

Edge Computing Enabled Smart Cities: AComprehensive Survey

Latif U. Khan, Ibrar Yaqoob, Senior Member, IEEE, Nguyen H. Tran, Senior Member, IEEE, S. M. Ah-san Kazmi, Tri Nguyen Dang, Choong Seon Hong, Senior Member, IEEE

Abstract—Recent years have disclosed a remarkable proliferation of compute-intensive applications in smart cities. Such applicationscontinuously generate enormous amounts of data which demand strict latency-aware computational processing capabilities. Althoughedge computing is an appealing technology to compensate for stringent latency related issues, its deployment engenders newchallenges. In this survey, we highlight the role of edge computing in realizing the vision of smart cities. First, we analyze the evolutionof edge computing paradigms. Subsequently, we critically review the state-of-the-art literature focusing on edge computing applicationsin smart cities. Later, we categorize and classify the literature by devising a comprehensive and meticulous taxonomy. Furthermore,we identify and discuss key requirements, and enumerate recently reported synergies of edge computing enabled smart cities. Finally,several indispensable open challenges along with their causes and guidelines are discussed, serving as future research directions.

Index Terms—Smart cities, Internet of Things, Mobile cloud computing, Cloudlet, Fog computing, Mobile edge computing, Micro datacenters.

F

1 INTRODUCTION

The unprecedented plethora of miniaturized sensing tech-nologies has shown a remarkable breadth of smart citiesvision. Smart cities enable citizens to experience sustainable,secure, and reliable developments [1], [2]. Smart cities aredefined formally in numerous ways [3]–[6] and thus, thereis no absolute definition of the smart cities. Generally, thesedefinitions refer to advancements related to information andcommunication technologies (ICT) infrastructure, that en-able citizens to realize high quality of life via smart services.Smart city services are not limited and can refer to manydifferent processes, resulting in more reliable, secure, sus-tainable, and innovative cities with unique entrepreneurialopportunities. The key contributors to smart cities are plat-form developers, research communities, governments, citi-zens, service providers, and application developers [7].

A remarkable increase in city populations is expectedin the foreseeable future, which will impose substantialcomputational complexity regarding advancements in thesmart cities. Moreover, this rapid rise in populations willimpose further challenges on the city infrastructure dueto an exponential increase in data generated by differentdevices, such as smartphones, global positioning systems,smart sensors, and smart cameras. One way to cope with

• L. U. Khan, I. Yaqoob, T. N. Dang and C. S. Hong are with the Departmentof Computer Science & Engineering, Kyung Hee University, Yongin-si17104, South Korea.

• S.M.A. Kazmi is with the Institute of Information Systems, InnopolisUniversity, Innopolis, Tatarstan, Russia, and the Department of ComputerScience and Engineering, Kyung Hee University, South Korea.

• N.H. Tran is with School of Computer Science, The University of Sydney,Sydney, NSW 2006, Australia, and the Department of Computer Scienceand Engineering, Kyung Hee University, South Korea.

• Corresponding Author: Choong Seon Hong ([email protected])

the challenges of massive scale computing and storage isthe use of expensive dedicated computing hardware withsufficient storage capacity. To eliminate the need for dedi-cated expensive computing hardware, cloud computing wasintroduced. This is a powerful technology that is intendedto enhance the Quality of Experience (QoE) by providingon-demand storage and processing capabilities in a cost-effective and elastic manner [8]. The striking features ofcloud computing, such as scalability, elasticity, multitenancy,sufficient storage capacity, and resource pooling, have madeits adoption possible in different areas [9]–[11]. Recently,cloud computing with virtually unlimited resources hasemerged as a promising paradigm to tackle the high compu-tational complexity associated with smart cities [12]. How-ever, its inherent limitations of high latency, non-context-aware behavior, and no support for mobility pose seri-ous limitations on its use in real-time smart environments.Apart from these downsides, cloud computing suffers fromprocessing time inefficiency due to the large overhead ofsmart city device data. On the other hand, edge computingextends the cloud computing resources to the network edgeand offers context-awareness, low latency, mobility support,scalability, to name a few. Hence, to address the limitationsof cloud computing for enabling real-time smart city envi-ronments, edge computing is a viable solution [13]–[16].

Figure 1 elucidates a high-level overview of Internet ofThings (IoT) based smart city environments enabled by edgecomputing. These IoT based smart environments leverageconnected smart IoT devices and sensors to improved lifestandards for citizens. Apart from that, IoT can be definedin different ways because it is associated with a varietyof technologies and concepts in literature [17]–[19]. Mostly,the definitions of IoT are inspired from research projects[20]. In [17], the authors found the common and frequentlyoccurring features among different IoT definitions, which

arX

iv:1

909.

0874

7v1

[cs

.NI]

11

Sep

2019

2

CloudFOG Node

Cloudlet

hospital

hospital

Augmented Reality

glasses

Waste

Recycling

plant

Smart Waste

bin

Smart GridSmart

Agriculture

`

1 2 3 4 5

Smart Parking

Smart Police

Blockchain enabled

Banking

Smart People

Cellular

BS

Smart Airport

Smart

Healthcare

Smart Buildings

Road

Side Unit

Mobile Edge

Computing Server

Fig. 1: An overview of edge computing enabled smart city

include:

• Existence of a global network infrastructure thatenables a unique addressing scheme, seamless inte-gration, and connectivity among heterogeneous IoTnodes.

• IoT nodes must be addressable, readable, locatable,recognizable, and autonomous.

• Existence of heterogeneous technologies that must begiven assiduous attention.

• Association of services to objects to emphasize itsobject dependent nature.

• Intelligent interfaces between things and humans.

We can consider the above-mentioned features (as empha-sized in [17]) in a platform to constitute IoT. However, it maynot always be possible for IoT system to possess all of thesefeatures. IoT is broadly classified into Industrial Internet ofThings (IIoT) and Consumer Internet of Things (CIoT) in[18]. The IoT assisted applications for smart cities includesmart grids, smart lighting systems, smart transportation,smart health-care, smart homes, augmented reality, andsmart agriculture [1], [21]–[23]. Therefore, it is evident fromliterature that IoT is an integral part of smart cities. The nextstep is enabling the resource intensive and strict latency IoTbased smart city applications. Edge computing provides a

promising way of enabling these applications by offeringcomputation and storage resources with low latency.

Edge computing is a cutting-edge computing paradigmcharacterized primarily by its geo-distributed operation,context-awareness, mobility support, and low latency. Itmigrates computing resources, such as computing power,data, and applications, from the remote cloud to the net-work edge, and thus enables numerous real-time smart cityservices. Recently, three different emerging technologies,cloudlet, fog computing, and mobile edge computing, havebeen used in the literature to bring the striking features ofcloud computing to the edge [24]. For example, edge com-puting reduces the network bandwidth usage by alleviatingthe transmission of data from end users to the remote cloud[13].

1.1 Research Trend and StatisticsFigure 2 illustrates edge computing and smart cities researchtrends showing the exponential increases in the numbersof publications in both domains. Apart from the researchpublications trend, according to statistics, the percentagesof people living in cities were 29% and 50% in the years1950 and 2008, respectively, which is expected to reach 65%in 2040 [25]. Approximately 1.3 million people are movingto cities every week. In 1975, there were three megacities,

3

Fig. 2: Number of published articles on edge computingand smart cities

but currently there are 21 mega cities having a populationmore than 10 million. Moreover, the number of megacitiesin 2025 is expected to increase by one city in Africa, two inLatin America, and five in Asia. Furthermore, cities utilizeapproximately 60%-80% of the world energy and generate50% of greenhouse gas emissions.

The considerable rise in the number of smart cities inthe foreseeable future will likely to increase smart citiesmarket share. It has been predicted that the smart citiesmarket share will increase at a Compound Annual GrowthRate (CAGR) of 18.4%. The smart cities market share willreach USD 717.2 billion by 2023 compared to USD 308billion in 2018 [26]. The major players of the smart cityservice providers include Toshiba (Japan), Schneider Elec-tric (France), Siemens (Germany), Hitachi (Japan), Ericsson(Sweden), Huawei (China), Cisco (US), Oracle (US), IBM(US), Microsoft (US), among others. On the other hand, thenumber of smart IoT devices enabling smart city applica-tions is expected to reach 500 billion by 2030.

Grand View Research, Inc. estimates the market shareof edge computing to reach USD 3.24 billion by 2025 ata CAGR of 41.0% [27]. Smart health-care is expected towitness the highest CAGR growth in the period from 2017 to2025. Health-care is expected to exceed the market share ofUSD 326 million. The potential players in edge computingmarket are SAP SE, Microsoft, Intel Corporation, Interna-tional Business Machines (IBM), Huawei Technologies Co.Ltd., Amazon Web Services, Inc. (AWS), Hewlett PackardEnterprise Development LP, General Electric, Cisco Systems,Inc., among others. Edge computing in the market is ex-pected to continually increase. Apart from edge computingmarket, the rapid rise in IoT based smart cities and edgecomputing research is evident from the statistics and re-search trends. These technological advancements in edgecomputing, IoT, and smart cities have revealed captivatingresearch and development opportunities for manufacturersto enhance their market shares.

1.2 Motivational ScenariosWe provide two motivational scenarios to demonstrate theimmense importance of edge computing in developingsmart cities. The first scenario deals with reporting of au-tonomous car accidents, and the second scenario describesthe usage of Unmanned Aerial Vehicles (UAVs) with edgecomputing for smart forest surveillance.

A gigantic increase in the number of autonomous carsis expected in the future. There are six different levelsof autonomy in autonomous cars ranging from level 0 tolevel 5 [28]. The level 0 represents no automation and level5 represents full automation, such as operating withoutdriver assistance.Moreover, autonomous cars can handlelane changing, warning signals, collision avoidance, andin-car infotainment, to name a few. To enable real-timeanalytics for autonomous cars, the use of edge computingenabled Road Side Units (RSUs) is a viable solution. Forinstance, consider autonomous car accidents, where timelyreporting of accidents is required for instant first aid. Thekey stakeholders of the emergency response system aremedical services, smart police, road operators, and edgeenabled public safety answering points [29]. When an au-tonomous car crash occurs, time is the most critical parame-ter. The emergency response time must be short to minimizeany damage. Reporting such accidents can be performedeither by a person or by the edge enabled public answersafety points. Those involved in the crash might sufferfrom severe injuries and cannot inform authorities about theaccident. Therefore, the edge enabled public answer safetypoints must perform accident evaluation which requires theexecution of complex algorithms. To execute such resourceintensive and latency sensitive tasks, edge computing isrequired.

On the other hand, smart forest fire detection cansubstantially reduce damage caused by fires. UAV enablededge computing offers a method of timely reporting of theforest fire. We can use UAVs assisted by edge computingfor surveillance. The UAVs use cameras to continually takeimages of the forest and requires further processing ofthese images to determine the presence of fire. One wayto determine the presence of fire is to use the remote cloudto perform video images analytics. However, this approachsuffers from high latency which is not desirable for im-mediate action against fires. Processing images at the edgecomputing enabled UAVs alleviates this latency issue andallows for quick decision making for emergency services.Furthermore, we can also leverage UAVs assisted by edgecomputing to enable on-demand telecommunication andcomputing infrastructure to assist in rescue activities [30].

1.3 Existing Surveys and TutorialsIn the literature, numerous surveys and tutorials havebeen presented to provide considerable insights into edgecomputing and smart cities [24], [31]–[45]. A summary ofrecently published surveys along with their primary contri-bution areas is given in table 1.

1.3.1 Edge Computing SurveysSeveral surveys pertaining to edge computing have beenconducted in [24], [31]–[41]. In [31], the authors discussed

4

TABLE 1: Summary of existing surveys and tutorials with their primary focus

Reference Smart city Internet ofThings

Mobileedgecomputing

Cloudlet Fog computing Edge computingevolution

Bilal et al. , [31] 7 3 3 3 3 7

Shaukat et al. , [32] 7 7 7 3 7 7

Ai et al.,[33] 7 3 3 3 3 7Mouradian et al. , [34] 7 7 7 7 3 7

Hu et al. , [35] 7 3 7 7 3 7Taleb et al. , [36] 7 7 3 7 7 7

Abbas et al. , [24] 7 3 3 7 7 7

Roman et al. , [37] 7 7 3 3 3 7

Mao et al. , [38] 7 7 3 7 7 7

Khan et al. , [39] 7 7 3 3 3 7

Hassan et al. , [40] 7 3 3 3 3 7

Yu et al. , [41] 7 3 3 3 3 7

Gharaibeh et al. , [42] 3 7 7 7 7 7

Lim et al. , [43] 3 3 7 7 7 7

Alavi et al. , [44] 3 3 7 7 7 7

Silva et al. , [45] 3 3 7 7 7 7

Our Survey 3 3 3 3 3 3

different edge computing paradigms, such as cloudlets, mo-bile edge computing, and fog computing. Additionally, theydiscussed applications and summarized the architecturesof state-of-the-art edge computing paradigms. Finally, theyoutlined open research problems regarding edge comput-ing. A survey of cloudlet architectures, cloudlet solutiontaxonomy, and challenges pertaining to the deployment ofcloudlets in local wireless networks is presented in [32].The taxonomies are devised using three different parame-ters: augmentation models, cloudlet service models, and ar-chitecture. Furthermore, cloudlet deployment requirementsare also discussed. In [33], Ai et al. presented a studyon cloudlets, fog computing, and mobile edge computing.Specifically they presented general architecture, standard-ization efforts, applications, and design principles of theedge computing paradigms. In addition, open research is-sues are also highlighted in their paper. Mouradian et al. in[34] proposed a comprehensive survey on fog computingarchitecture and algorithms. The authors proposed evalua-tion criteria, such as heterogeneity management, scalability,mobility, federation, and interoperability. The evaluation cri-teria are further used to evaluate the literature related to ar-chitecture and algorithms for fog computing. Additionally,the architectural and algorithmic open research problemsare also outlined. A comprehensive survey proposed in[35] discussed fog computing architecture, applications, keytechnologies, and research challenges. The authors proposeda hierarchical architecture of fog computing, followed bya discussion of key technologies. Apart from that, theyalso discussed various smart environments. Finally, openresearch challenges are also presented. In [36], Taleb etal. presented the evolution of edge computing, use casesof mobile edge computing, key mobile edge computing

enablers, reference architecture, deployment challenges, andopen research challenges. Abbas et al. discussed architec-ture, emerging smart applications, security and privacy, andopen research challenges in mobile edge computing[24].The authors in [37] provided a survey on security andprivacy issues in mobile edge computing. In addition, adetailed overview of edge paradigms concepts, challenges,and security threats are also presented. They identifiedthreat models and discussed guidelines for their solutions.Furthermore, a state-of-the-art security mechanism used inedge computing paradigms is also outlined. The study in[38] presented a comprehensive survey on computationand radio resource management in mobile edge computing.They discussed green mobile edge computing, privacy-aware mobile edge computing, mobility management formobile edge computing, and cache enabled mobile edgecomputing. Along with this, open research challenges arealso discussed. In [39], Khan et al. presented survey on edgecomputing. They have discussed different edge comput-ing paradigms, such as cloudlets, mobile edge computing,and fog computing. key characteristics of edge computingare presented. Additionally, fundamental requirements andopen research challenges for realizing edge computing arepresented. In another study in [40], the role of edge com-puting in IoT has been discussed. Taxonomy has been de-vised based on edge computing enabled IoT. Furthermore,numerous smart applications, key requirements, and fewopen research challenges are presented. Yu et al. in [41]conducted a comprehensive survey on the role of edgecomputing in IoT networks. They classified the edge com-puting based on architecture into the far-end, near-end, andfront-end. Furthermore, they discussed the advantages ofedge computing-based IoT infrastructure and reported open

5

research challenges.

1.3.2 Smart Cities SurveysIn the past, several surveys have been conducted on smartcities [42]–[45]. Gharaibeh et al. in [42] surveyed data man-agement techniques used for ensuring reuse-ability, granu-larity, interoperability, and consistency of smart IoT devicedata. Additionally, the authors identified techniques usedfor security and privacy in smart cities. Other than that,emerging technologies enabling smart cities are also dis-cussed. In [43], the authors discussed the use of big dataalong with different projects that enable the development ofsmart cities. The authors have classified the use of urban bigdata into four kind of reference models. Furthermore, thechallenges of transforming smart city data into informationare also presented. The study [44] discussed IoT basedsmart city applications. The authors provided an overviewof current IoT based projects. Additionally, a prototype fora real-time monitoring system using IoT for smart cities hasbeen presented. An overview of smart cities characteristics,architecture, developments, and challenges has been pre-sented in [45].

1.3.3 Our SurveyTo the best of our knowledge, we are the first to jointlyconsider (as can be seen in table 1) smart cities, IoT, andedge computing. Previously published surveys and tutori-als reviewed smart cities with IoT [43]–[45], whereas [42]considered smart city only. However, [24], [31]–[38], [41]reviewed either edge computing alone or edge computingwith IoT. Unlike the existing surveys, we comprehensivelypresent premier smart city environments enabled by edgecomputing, the evolution of edge computing, taxonomy,core requirements, and open research challenges.

1.4 ContributionsThe key contributions of this survey are as follows:

• We present the evolution of edge computing usingscalability, flexibility, cost optimization, automation,mobility, and latency, as metrics.

• We report, critically analyze and evaluate the premieradvances made in edge computing enabled smartcities.

• We categorize and classify the literature by devisinga comprehensive and meticulous taxonomy usingimportant parameters, such as characteristics, se-curity, edge analytics, resources, edge intelligence,resource management strategies, caching, and sus-tainability.

• A few indispensable requirements for the design ofthe smart cities architectures using edge computingare presented.

• Several open research challenges along with theircauses and guidelines regarding the implementationof edge computing in smart cities are discussed.

The rest of the paper is organized (illustrated in figure 3)as follows: Section 2 discusses the evolution of edge com-puting. Recent advances and literature categorization aregiven in section 3. Section 4 presents the devised taxonomy

Sec. 1 Introduction

Sec. 1.1 Research Trend and Statistics

Sec. 1.2 Motivation Scenarios

Sec. 1.3 Existing Surveys and Tutorials

Sec. 1.4 Contributions

Sec. 2 Evolution or

Origin of Edge

Computing

Sec. 2.1Metrics

Sec. 3 Recent

Advances

Sec. 3.1 Smart Transportation System

Sec. 3.2 Augmented Reality

Sec. 3.3 Smart Health-care

Sec. 3.4 Smart Grids

Sec. 3.5 Smart Farming

Sec. 3.6 Smart Buildings

Sec. 3.7 Smart Gaming

Sec. 3.8 Smart Management

Sec. 4 Taxonomy

Sec. 4.1 Characteristics

Sec. 4.2 Edge Analytics

Sec. 4.3 Resources

Sec. 4.4 Resource Management

Strategies

Sec. 4.5 Edge Intelligence

Sec. 4.6 Caching

Sec. 4.7 Sustainability

Sec. 4.8 Security

Sec. 5 Requirements

Sec. 5.1 Scalability and Reliability

Sec. 5.2 Contex-awarenessSec. 5.3 Efficacious Resource

Management

Sec. 5.4 Sustainability

Sec. 5.5 Elasticity

Sec. 5.6 Interoperability

Sec. 5.7 Privacy and Security

Sec. 5.8 Business Model

Sec. 6 Case Studies

Sec. 6.1 Barcelona Smart City

Sec. 6.2 # SmartME

Sec. 6.3 WATCH

Sec. 2.2 Evolution

Sec. 8 Conclusions

and Future Prospects

Sec. 8.1 Conclusions

Sec. 8.2 Future Prospects

Sec. 7 Open

Research Challenges

Sec. 3.9 Lessons Learned: Summary

and Insights


and Insights


and Insights


and Insights

Sec. 7.1 Intelligent Caching

Sec. 7.2 Collaborative Edge Computing

Sec. 7.3 Cooperative and Sustainable

Load Balancing

Sec. 7.5 Intelligent Edge Computing

Sec. 7.6 Cyber and Physical Security

Sec. 7.4 Network Slicing

Fig. 3: Structure of the survey

based on different parameters. The core requirements indesigning edge computing enabled smart cities are given insection 5. The recently reported synergies and case studiesalong with their key objectives, organizations involved, anddeployment year are given in section 6. Section 7 identifies,discusses, and provides guidelines for solutions of the in-dispensable open research challenges. Finally, the paper isconcluded in section 8.

6

Mainframe Computing, Mini

Computing, Client Server

Computing

Desktop Cloud

Computing

Mobile Cloud

ComputingEdge Computing

Ch

arac

teri

stic

sK

ey T

echn

olo

gie

s

1960-1989 Information Management System

Virtual Machines

Customer Information Control System

Multiple Virtual Storage

AS400

Relational Database

1990-2006 .Net

J2EE

Web Sphere

2007-2008 iOS

Android

2009-2019 Cloudlets

Fog Computing

Mobile Edge Computing

Mainframe Computing

- High Processing Power

- High Storage

- Industrial Usage

Mini Computing

- Low Storage

- Low Processing Power

- Easier Management

Client Server Computing

- Centralized Control System

- Increased Security

- High Complexity in Management

Deployed at Network Core

Centralized Architecture

No Support For Mobility

High Latency

Average Scalability

Elastic Services

Deployed at Network Core

Centralized Architecture

Mobility Support

High Latency

Average Scalability

Elastic Services

Deployed at Network Edge

Low Latency

High Scalability

Elastic Services

Context-awareness

Support for Mobility

Fig. 4: Evolution of edge computing

2 EVOLUTION OR ORIGIN OF EDGE COMPUTING

In this section, we consider the chronological developmentand attributes of different computing paradigms, namely,mainframe computing, mini computing, client server com-puting, desktop cloud computing, Mobile cloud Comput-ing (MCC), and edge computing. Similar to [46], twoparadigms, functionalism and structuralism are consideredin this paper to extend our findings. Structuralism enablesidentification of the building blocks and their relationshipsto provide a clearer understanding of the phenomenon,whereas functionalism provides insights into the behaviorand features of a phenomenon through analysis of its rolein present and future with functionalities.

2.1 MetricsWe consider six metrics, namely, scalability, flexibility, costoptimization, automation, latency, and mobility support forevaluation of different computing paradigms:

• Scalability: This parameter reflects the ability of asystem to expand its services in response to increas-ing demands of the users.

• Flexibility: An ability of the system to provide com-puting infrastructure in an elastic way.

• Cost Optimization: This refers to the ability of asystem to provide users on-demand computing re-sources for optimizing the overall cost. In addition,it also refer to the cost associated with hardwareand software in the acquisition of the computingresources.

• Automation: The ability to perform cloud updateswithout any intervention of the end users.

• Latency: This parameter is used to measure the totalexecution time of smart city applications.

• Mobility Support: It aims to provide an ability forenabling seamless execution of IoT-based applica-tions.

2.2 EvolutionFigure 4 presents the gradual evolution of edge computingproceeds by other computing technologies, such as minicomputing, mainframe computing, client-server computing,desktop cloud computing, and Mobile cloud Computing(MCC). Further details are provided below.

2.2.1 Mainframe, Mini, and Client-Server ComputingIn the early era of computing, mainframe computers thatinclude ENIAC, Markl, BINAC, Whirlwind, UNIVAC, IBM701, and IBM 360, were primarily used to enable bulkdata processing. In contrast, minicomputers (developed byIBM) have a smaller size than mainframe computer andare mainly used as mid-size servers to process scientificapplications. Different from mainframe computing and minicomputing, client-server computing is based on providingresources to clients on request. Xerox PARC was the firstclient-server computing model introduced in the 1970s.

2.2.2 Desktop Cloud ComputingThe concept of cloud computing dates back to 1963, whenthe Defense Advanced Research Projects Agency (DARPA)

7

provided $2 million to MIT for project Project on Mathemat-ics and Computation (MAC) [47]. The purpose of projectMAC was to establish a technology that enables simulta-neous use of a computer by multiple people. In 1969, theAdvanced Research Projects Agency Network (ARPANET)was developed by J. C. R. Licklider. The vision “Intergalac-tic Computer Network” (also known as the Internet) waspromoted by Licklider to enables connectivity between allhumans on the earth. Later in 1997, Ramnath Chellappa ofEmory University devised the definition of cloud computingas a ”computing paradigm, where the boundaries of computingwill be determined by economic rationale, rather than technicallimits alone”.

The paradigm of cloud computing began gaining moreinterest when the different organizations started exploitingit to provide services to the end-users. In 1999, Sales-force successful implemented cloud computing to providesoftware services using the Internet. The end-users weregiven the ability to download an application on demandin a cost-effective way from a remote location [48]. Ama-zon in 2002 offered its retail services through the Web.In 2006, Amazon started providing web-based services toclients on-demand. Google also started providing Googledoc services in the same year. Following this, IBM in 2011offered IBM SmartCloud and Apple introduced iCloud toenable storage of personal information, such as videos, mu-sic, photos, among others. In 2012, Oracle started offeringOracle Cloud which enabled three categories of services:SaaS (Software-as-a-Service), PaaS (Platform-as-a-Service),and IaaS (Infrastructure-as-a-Service).

2.2.3 Mobile Cloud ComputingMCC was introduced to enable mobile computing with theability to leverage cloud computing. Mobile devices haveinherent limitations of frequent disconnections, resourcescarcity, and energy constraints. Therefore, to execute highcomplexity tasks, it is preferable to execute a resource-intensive task on a remote cloud. MCC has been definedin three ways in [49]:

• MCC refers to the running of mobile device applica-tions (e.g. Google’s Gmail) on a remote server.

• MCC deals with the use of mobile devices and otherstationary devices resources as a cloud resource. Mo-bile devices form a peer-to-peer network and allowuse of their resources by other devices.

• MCC refers to a cloudlet which is a micro datacenter consisting of several multi-core computersand which provides computation resources to nearbymobile devices [50].

In this paper, we use the term MCC to mean running ofclient applications of mobile devices on remote servers,such as Google servers and Amazon servers. On otherhand, various operating systems, such as iPhone OS (IoS)and android, which were introduced in 2007 and 2008,respectively, provide mobile devices with more capabilitiesby using apps.

2.2.4 Edge ComputingEdge computing refers to practices that move computingand storage resources to the network edge. In the literature,

three different practices, such as cloudlets, fog computing,and mobile edge computing have been used, that extendcloud computing to the network edge. All of these threeterms used in the literature refer to edge computing [24],[33], [51]. Edge computing is distributed architecture thatenables real-time data processing and acceleration of data-streams with low latency. Along with this, edge computingalso minimizes the Internet bandwidth usage, which furtherreduces network congestion and delay.

The concept of cloudlets was introduced in [50] bySatyanarayanan et al. to place a small scale data center nearthe network edge to enable operation of resource-intensiveapplications with low latency. A cloudlet represents a mid-dle layer in an architecture having three tiers: a centralizedcloud, edge cloud platform (cloudlet), and end device [36].CISCO coined the term fog computing, which refers to anarchitecture that allows the use of edge devices to providecomputation services near the network devices [52]. Similarto cloudlets, fog nodes (edge devices) represent the middletier of a three-tier hierarchy that consists of cloud, fog nodes,and end devices. Fog nodes cover a wide range of devices,such as routers, wireless access points, and data centers [53].

The term mobile edge computing was introduced byEuropean Telecommunications Standards Institute (ETSI) toenable the placement of computing resources at the radioaccess network and core network of mobile telecommunica-tion systems [54], [55]. The mobile edge computing serverscan be placed at different positions in the radio access net-work such as a macrocell BS [56], [57], assigned to a group ofBSs [58], or in a hierarchical fashion at different levels [59].Consider table 2, which lists the differences between cloudcomputing, mobile edge computing, cloudlets, and fog com-puting. Mobile edge computing differs from cloudlets andfog computing in its primary scope, which is telecommuni-cation networks. In terms of context-awareness, mobile edgecomputing has the highest value while cloud computing hasno context-awareness. Fog computing and cloudlets havemedium and low context-awareness. Other than context-awareness, mobility is supported by all three categoriesof edge computing, whereas cloud computing has no orlimited mobility support.

3 RECENT ADVANCES

This section critically analyzes the dominant trends towardthe edge computing enabled smart cities. Apart from pro-viding insight into the architectural and algorithmic aspects,we perform a rigorous evaluation of the literature usingvarious assessment parameters. These assessment param-eters are primarily derived from the literature of recentadvances in edge computing enabled smart cities [60]–[79].We consider five different assessment parameters, namely,context-awareness, scalability, sustainability, caching, andsecurity for evaluation of the recent advances:

• Context-awareness: An ability of the system to ob-tain information pertaining to the nodes locationsand surrounding environment. Context-awarenessoffers several advantages including the addition ofmore meaning to smart IoT device data, M2M com-munication facilitation, emergency management as-

8

TABLE 2: Edge computing paradigms comparison

Cloud computing Cloudlets Fog computing Mobile edge computingContext-awareness No Low Medium High

Geo-distribution Centralized Distributed Distributed Distributed

Latency High Low Low Low

Mobility support No/Limited Yes Yes Yes

Distance Multi hop Single hop Single hop/ Multi hop Single hop

sistance, and automatic execution of smart city ser-vices, to name a few.

• Sustainability: This deals with energy efficient de-sign and use of renewable energy sources. In addi-tion, sustainability refers to energy harvesting, whichtakes energy from environmental sources and radiofrequency sources and stores it for further use. Theprimary objective of sustainability is to reduce thecarbon footprint. A significant portion (more than80%) of energy is generated using brown sources[80]. Therefore, sustainability is an important as-pect that must be considered in the design of edgecomputing enabled infrastructure. Additionally, thismetric offers an increase in profit, reduction in carbonfootprint, and hardware reliability [81].

• Scalability: The ability of a system to enable elasticservices as per user demands without losing QoS,resulting in a cost-efficient operation.

• Caching: The temporary storing of popular contentat different locations in a network to enable accesswith low latency. Caching also reduces congestionin a network by avoiding repeated traffic flow.Notethat caching deals with the storage of popular con-tent and edge computing provides computation re-sources, but these two concepts can be leveragedsimultaneously in smart cities to enable a variety ofsmart applications.

• Security: This refers to a devices physical securityand cyber security. More specifically, cyber securitydeals with the protection of the network, computinginfrastructure, and data from attacks.

Table 3 summarizes explanation, key advantages, and di-mensions of the parameters to evaluate recent advances.Every parameter has a set of dimensions. For instance, sus-tainability has architectural, algorithmic, communication,renewable energy, software, and hardware dimensions. Allof these dimensions are key drivers of sustainability inedge enabled smart cities. We use a check mark (3) for aparameter if has one or more dimensions are consideredand cross mark (7) if its no dimension that is considered.Table 4-5 highlight key contributions and categorize the re-cent advances according to different applications. Moreover,these tables present an evaluation of the literature.

3.1 Smart Transportation System

A smart transportation system is envisioned to enable in-novative traffic management schemes, advanced transportmodes, autonomous driving, and in-car infotainment ser-vices.

3.1.1 Smart Transportation ApplicationsGiang et al. in [60] discussed the development of smarttransportation applications leveraging fog computing. Theyhave discussed the requirements, state-of-the-art, and re-search challenges regarding implementation of the fog com-puting enabled smart transportation applications. Althoughthe authors discussed the requirements of scalability andcontext-awareness in fog computing assisted vehicular net-works they did not consider service migration [64] andsecurity [82], which have significant importance in vehicularnetworks. The vehicle might change connectivity from oneroadside unit to another during video streaming. Therefore,service migration must be considered in the design of fogassisted application for vehicles. Similarly, the security as-pect must be addressed to avoid any accidents and physicaldamage. For example, if a hacker accesses the vehicle andalters the lane change information, this might result inaccidents, specifically in autonomous cars.

3.1.2 5G-enabled Software Defined Vehicular NetworksIn [61], the authors proposed a 5G-enabled software definedvehicular network (5G-SDVN) architecture, that consists ofthree planes, such as data plane for the transmission ofdata, social plane for Vehicular Neighbor Groups (VNGs)networking, and control plane with mobile edge computing.The key contributions of the authors include the discoveryof VNGs using real data set and the integration of software-defined networking (SDN) with mobile edge computingto enable flexible architecture. In addition, the proposedarchitecture provides the benefits of efficient data sharing,ubiquitous mobile interaction, and reliable resource coop-eration. Apart from these benefits, SDN offers separationof the control logic from the underlying infrastructure,which provides more freedom for adding functionality(i.e.intelligence) to a network regardless of change in hard-ware. Although, the proposed architecture offers significantadvantages, but it does not consider the case of activeservice migration as in [60].

3.1.3 Cost Optimized Edge Enabled TransportationThe study [62] considered a smart city scenario where ve-hicles ran applications that take data from the environmentand send them to edge computing servers through roadsideunits (RSUs) to enable further processing and analysis. Theedge computing servers are co-located with the RSUs, whichare further connected to the cloud through the Internet.In a vehicular network, maintaining seamless connectivitybetween the vehicles and RSUs is difficult. On the otherhand, the computational capacities of the edge servers alsohave limitations. Therefore, it is imperative to deploy the

9

TABLE 3: Evaluation parameters: explanation, dimension, and advantages

Parameter Explanation Dimension Advantage

P1: Context-awareness It deals with locations of smartdevices and their surroundingenvironment.

• Context-awarearchitecture

• Context-awarealgorithm

• Helps in accuratedecision-making

• Efficient emergencymanagement assistance

• Enables automatedsmart services

P2: Scalability It refers to an ability of a systemto operate effectively with anincrease in demand by addingmore resources elastically.

• Scalable architecture• Scalable hardware• Scalable Software• Scalable algorithm

• Cost optimization• QoS maintenance• Instant service avail-

ability

P3: Sustainability It deals with energy efficient de-sign and use of renewable en-ergy sources to minimize thecarbon footprint.

• Sustainable architec-ture

• Sustainable Software• Sustainable hardware• Sustainable algorithm• Sustainable communi-

cation• Renewable energy

sources

• Carbon footprint re-duction

• Profitable infrastruc-ture

• Hardware reliability

P4: Caching It allows temporary storage ofpopular contents at different lo-cations in a network.

• Caching architecture• Caching algorithm• Caching hardware• Caching software

• Instant access to popu-lar contents

• Back-haul traffic reduc-tion

P5: Security It deals with physical and cybersecurity of smart devices andservers involve in edge comput-ing.

• Security algorithm• Security architecture

• Continuous serviceavailability

• Users privacy preserva-tion

RSUs to enable effective resource management and seamlessconnectivity. The objective of the paper is to minimizethe network deployment cost while fulfilling the minimumrequired QoS for vehicular applications. The problem isformulated as a mixed integer linear programming problem,whose goal is to jointly minimize the total cost associatedwith the deployment of RSUs, the cost associated withassigned power levels to each RSU, and the cost associatedof the distances of RSUs from the cells it serves (the targetarea is divided into cells). A single RSU can serve morethan one cell. Therefore, only the cells adjacent to an RSUcell must be considered as opposed to distant cells, whichincrease the cost.

3.2 Augmented RealityAugmented Reality (AR)enabled smart cities provide userswith smart Industrial Augmented Reality, smart tourism,smart remote live support, smart web-based augmentedreality, as we discuss in the following subsections.

3.2.1 Smart Industrial Augmented RealityThe authors in [63] proposed an industrial ARarchitec-ture based on fog computing, named, Navantia’s indus-

trial AR(IAR). The use of IAR in smart industries leadsto enhancing productivity. The proposed IAR architectureconsists of three layers, such as IAR layer, edge layer, andcloud layer. The IAR layer includes ARdevices that use WiFialong with other wireless technologies for connectivity withthe edge layer. The IAR layer devices interact with the edgelayer through local edge layer gateways that provide low-latency services, such as data caching and sensor fusion.The authors considered both scalability and caching in theirarchitecture. However, they did not consider security whichis required for safe operation. For example, an unauthorizeduser can access the system and interrupt proper operationof the system, which can lead to serious consequences.

3.2.2 Smart TourismIn [64], Taleb et al. envisioned mobile edge computing en-abled architecture for high definition video streaming andconsidered two use cases of smart tourism. One case isinteractive glasses to obtain information about the visitedsite, whereas the other case involves tourists taking videos,commenting on these videos, and sending them to the edgecomputing server. Although migration of such a servicefrom one edge server to another is considered in their

10

TABLE 4: Classification, key contributions, and evaluation of the recent advances

Application Reference Key contributions Evaluation parametersP1 P2 P3 P4 P5

Smart transportationGiang etal. [60] • Studied a fog assisted smart transportation ap-

plications requirements.• Discussed state-of-the-art transportation appli-

cations.• Identified research challenges in fog enabled

smart transportation systems.

3 3 7 7 7

Huang etal. [61] • Proposed a 5G SDN enabled vehicular network

architecture.• Discovered VNG using real data set.• Proposed architecture considered sustainability,

context-awareness, and security.

3 3 3 7 7

Premsankar etal. [62] • Proposed deployment of RSUs to jointly reduce

deployment and transmission power cost.• QoS is considered for vehicular applications

during deployment of RSUs.

3 7 3 7 7

Augmented reality

Fernandez-Carames etal. [63]

• Proposed a three layered edge computing andAR enabled architecture for industries.

• Presented a real time implementation of theproposed architecture.

7 3 7 3 7

Taleb et al.[64] • Proposed an architecture assisted by AR for

high definition video streaming.• Presented a shared memory concept for effec-

tive active service migration.• Validated the proposed architecture using a

practical use case.

3 7 7 3 3

Schneider etal. [65] • Proposed an architecture for remote live sup-

port.• A video compression is adopted to reduce the

bandwidth requirement.

7 7 7 7 7

Qiao et al. [66]• Proposed an edge assisted Web-based aug-

mented reality.• Discussed the deployment challenges of the

proposed system.

7 7 7 3 7

Smart health-careMuhammadet al. [67] • Proposed an edge computing assisted voice dis-

order assessment and treatment framework.• The proposed framework provides interoper-

ability.

3 7 7 3 3

Rahmani etal. [68] • Proposed a three layer fog assisted framework

for smart hospitals.• Presented implementation of the IoT based re-

mote health-care system as proof-of-concept.

3 3 3 3 3

Gia et al. [69]• Proposed a remote Health monitoring System

for ECG along with joint analysis and notifica-tion.

• The fog assisted gate-ways also provides se-curity by enabling restriction to un-authorizedaccess.

3 7 3 7 3

11

work, but they did not consider resource management formultiple users, which is more realistic than a single userscenario. Generally, tourists visit historical places in groups(social communities), therefore device-to-device communi-cation with social awareness must also be considered forperformance improvement. The user may request contentfrom another user that already has it instead of sending arequest from the access point. This will improve the trans-mission rate of the requesting user and reduce congestion ofthe access point.

3.2.3 Smart Remote Live SupportEdge computing enabled AR has been considered in [65] toenable smart management of industrial tasks. The authorsconsidered a scenario of remote live support that uses guid-ance from the remote expert through AR and video trans-mission to tackle complex faults in a machine. The mainparts of the proposed architecture include a client device,edge server, and a remote device. The job of the client devicewith an integrated camera and running an operating systemis to grab video frames for further transport to the edgeserver. The edge server process all the computationally ex-pensive operations. Additionally, the edge server forwardsthe video stream to a remote expert. The remote expert hasthe privilege of pausing the video for addition of hints init. The video contents are sent back to the edge server,which are further sent back to the client after renderingannotations into the camera feed. The author consideredcompression of the video data prior to transmission. Thisadds further delay in addition to transmission delay, whichis not desirable. Apart from that delay, the processing delaymust not exceed the maximum delay limit. The better optionis to use communication technologies with higher data ratesinstead of compression.

3.2.4 Smart Web-based Augmented RealityIn [66], Qiao et al. considered web-based AR enabledby edge computing. The latency sensitive and resource-intensive applications of the web-based AR are hamperedby the weak computational efficiency of web browsers. Toenable resource intensive and latency sensitive web-basedAR applications, web-based AR assisted by edge computingis proposed by the authors. Different from edge computing,cloud computing can be leveraged to assist computation in-tensive web-based AR applications. However, this approachcauses an increase in latency due to the remote location ofthe cloud. The key contributions of the authors include aproposal of edge computing assisted web-based augmentedreality, design details, and deployment challenges.

3.3 Smart Health-careSmart health-care is realized by having low cost and effec-tive real-time health-care facilities ubiquitously.

3.3.1 Smart Voice Disorder Treatment FrameworkA voice disorder assessment and treatment framework en-abled by edge computing and deep learning is proposed in[67]. Initially, smart sensors are used to collect samples ofusers voice, which are then processed on edge nodes andfinally, sent to core cloud. Decision of automatic assessment

is then sent by cloud manager to specialists for issuing aprescription to patients. A database used for training, test-ing and validation is the Saarbruken Voice Disorder (SVD)database. Prime limitation of their work lies in the usage offew disorders, such as vocal fold polyp, vocal fold paralysis,laryngitis, Kontakt pachydermia, hyper-functional dyspho-nia, functional dysphonia, and cordectomy, whose sampleshave sufficient amount in the database.

3.3.2 Smart HospitalsA three-layered fog based health-care architecture forhomes/ smart hospitals was proposed in [68]. The proposedarchitecture comprises of smart devices layer, edge layer,and cloud layer. The devices layer has medical sensors thatcaptures the biomedical signals from the human body. Thesebiomedical signals are further sent to the gateways throughwireless technologies, such as ZigBee, Wi-Fi or Bluetooth.The cloud layer consists of cloud computing that performsdata analytics. The edge layer consists of smart e-healthgateways, that are distributed geographically and capableof supporting protocols for wireless communication andinter-device communication. In addition, the edge layeralso performs local data processing, data filtering, datacompression, data fusion, data analysis, local storage, andinteroperability functions.

3.3.3 Remote Electrocardiography MonitoringIn [69], a smart health-care monitoring system is proposed,that enable continuous remote monitoring of Electrocar-diography (ECG) signals. The monitoring system consistsof a cloud, fog assisted gateways, and sensor nodes. Pur-pose of sensor nodes is to acquire bio-signals (such ashuman temperature, respiration, and ECG) and contextualinformation (such as room temperature and humidity). Fogassisted gateway receives the signals from sensor nodes andprocesses it for further sending it to cloud. For example, pre-processing of the bio-signals are necessary for removal ofnoise and extraction of useful information from it. Further-more, the fog assisted gateways have two storage databases,such as local and external user database. A local databaseis used to store the information that can be updated oredited by the administrators only. On the other hand, anexternal database is used for storage of bio-signals alongwith contextual information, such as room temperature andhumidity. Data stored within the external database hassynchronization with cloud servers. Additionally, the fogassisted gate-ways also provides security to restricts theunauthorized access.

3.4 Smart GridsA smart grid is a cutting edge technology to enable thetransformation of the traditional grids to reduce utility costsand global warming using renewable energy resources,smart meters, and smart appliances.

3.4.1 Smart Grid Data ManagementAuthors in [70] proposed a four-layer architecture for datamanagement based on Vehicular Delay-tolerant Network(VDTN) and edge computing for smart grids. First layerconsists of mobile devices, such as Plug-in Hybrid Electric

12

TABLE 5: Classification, key contributions, and evaluation of the recent advances

Application Reference Key contributions Evaluation parametersP1 P2 P3 P4 P5

Smart gridsKumaretal. [70] • Proposed a generalized architecture for data

management based on VDTN using edge com-puting for smart grids.

• An energy efficient virtual machine migrationutilizing load forecasting is also proposed.

7 7 3 7 3

Okay etal. [71] • Proposed a model for integration of smart grid

with fog computing.• Evaluated the proposed model using qualitative

use case.

3 3 3 7 3

Zahoor etal. [72] • Proposed a three layered cloud-fog-based smart

grid framework.• A hybrid artificial bee ant colony optimization

algorithm is also proposed for virtual machinesallocation.

3 3 3 7 7

Smart farming Ferrandez-Pastor etal. [73]

• Proposed a flexible layered IoT assisted PA ar-chitecture enabled by edge computing.

• Provides interoperability among different subsystems in the proposed PA architecture.

• Experimental validation of the proposed PAarchitecture is also provided.

3 3 3 7 3

Caria etal. [74] • Proposed a programmable Fog computing en-

abled architecture of smart farm for welfare ofanimals.

• Presented a Raspberry Pi assisted implementa-tion of the proposed architecture.

7 3 3 7 3

Zamora-Izquierdo etal. [75]

• Proposed a three planar edge assisted PA archi-tecture.

• NFV is exploited in edge layer to enable flexiblegeneric control modules.

• The proposed architecture is validated usingprototype implementation at real green-housein Spain.

7 7 3 7 7

Smart buildings Vallatietal. [76] • Proposed an edge computing enabled smart

home architecture using LTE.• Discussed the advantages and disadvantages

pertaining to placement of edge severs at dif-ferent positions, such as eNB, home, and UE.

3 3 7 7 3

Stojkoska etal. [77] • Proposed a three tier IoT framework enabled by

Fog computing for smart homes.

7 7 3 7 7

Smart gaming Premsankar etal. [78] • Discussed different edge computing architec-

tures.• Presented the use case of edge computing en-

abled mobile gaming.

7 7 7 7 7

Smart management Jia et al. [79]• Proposed an IMCS enabled by edge computing.• Context-awareness, scalability, and sustainabil-

ity are also exploited in the proposed system.• Narrow-band IoT is used for communication

due to its high energy efficiency, low cost, andsupport for massive number of manhole covers.

3 3 3 7 7

13

Vehicles (PHEV). The gateways and routers in the backbonenetwork make second layer of the proposed architecture.Third layer has different types of servers, such as a certifi-cate authority server, file server, and database server andfourth layer is made by a distributed cloud (infrastructureas a service). In a smart city environment, charging of PHEVbatteries from the nearby charging station may result inmany requests at charging stations. The proposed architec-ture is based on the use of VDTN for delivery of chargingand discharging decisions. The authors compared the delayincurred using the traditional core-cloud infrastructure andedge computing enabled infrastructure and observed lessincurred delay, more throughput, and less response time foredge computing.

3.4.2 Cloud–Fog–Based Smart Grid ModelIn [72], a three-layered framework named cloud-fog-basedsmart grid is proposed. The first layer is the end users layer,followed by the second fog layer, and the third cloud layer.The end layer comprises of buildings which have homes.Energy storage systems with renewable energy generationunits are installed at every home. The information related toenergy generation and consumption is sent to the fog layer.The fog layer that physically exists close to the consumersregion, performs effective management of the network re-source with low latency. Furthermore, the authors proposedhybrid artificial bee ant colony optimization (HABACO)algorithm for optimal allocation of virtual machines in thecloud-fog-based smart grid model. The authors consideredtwo scenarios for performance evaluation of their cloud-fog-based smart grid model but did not consider the commer-cial scenario that has a significantly different nature thanresidential areas.

A reference model for integration of fog computingand smart grids is introduced in [71]. The three-tier modelconsists of smart grid tier, fog tier, and cloud tier. The smartgrid tier has the responsibility of enabling communicationbetween different smart devices, such as smart meters, mo-bile devices, electrical vehicles, and smart appliances. Thefog tier acts as an intermediate layer between the cloud andsmart grid tier. Every fog server communicates with smartmeters for the collection of data of their associated users.Additionally, the fog layer servers take the private data fromthe smart grid layer in encrypted form to ensure the privacyof the users. The fog server does not have the capabilityto decrypt the data and the key is only shared among thecloud and the users. The main purpose of the fog serversis to provide temporary storage and computation capabilityfor the smart grids data. The prime limitation of their worklies in non-consideration of the challenge of interoperability,which is one of a most important issue among differentpower domain areas due to the usage of multiple semanticdata models in a smart grid environment [83].

3.5 Smart FarmingSmart farming is aimed to utilize advanced technologies forincreasing productivity and reduce cost simultaneously [84].

3.5.1 Smart Precision AgricultureIn [73], a distributed IoT-based smart framing infrastructureassisted by edge nodes is proposed to experience farmer

with enhanced benefits. The proposed architecture consistsof four levels, i.e., things layer, edge layer, fog layer, andcommunication layer. In things layer, a thing can be eitheractuators, controller or sensors, which are designed to meetproduction facility. The function of edge computing is toprovide interoperability, storage, smart analysis, and pre-diction near the edge. Reason for using edge computingis due to the low latency requirements of control, anal-ysis response, and monitoring sensors. Although authorsconsidered interoperability but did not consider caching inthe proposed framework. Caching can be utilized to furtherimprove the performance of the smart framing [85]. In [85],a joint latency and energy optimized communication wereachieved through data caching algorithm in Precision Agri-culture (PA) system. The data caching algorithm was basedon the optimization of sleep-up/wake-up time of wirelesssensor network nodes. Data requests of sensor nodes areanalyzed in a cache prior to transmission. This prolongsthe life of the sensor nodes and thus, improves the energyefficiency of the PA system.

In [75], the authors integrated PA with edge computingto enable real-time actions for better productivity. Theyproposed three planar architecture, such as the local plane,edge plane, and cloud plane. In a local plane, cyber-physicalsystems perform a collection of data through interactionwith crop devices to enable real-time control actions. Thesecond edge plane is responsible for performing tasks ofthe PA near the network edge to enhance reliability. Thethird cloud tier exists remotely to enable data analyticsand smart management. The cyber-physical systems areconnected to sensors and actuators, which are installed atthe crop premises for the collection of temperature, humid-ity, electrical conductivity, and pH meter. All the cyber-physical systems have inter-connectivity with the Internetusing multiple access networks technologies, such as digitalsubscriber line, optical fiber, and microwave radio links.In the edge plane, the tasks including energy and alarmmanagement, greenhouse control, and nutrition control areperformed. In the proposed architecture, the authors did notconsider security which must be given proper consideration.Some malicious user might access the system and performwrong activities, such as unnecessary alarms and windowsopening.

3.5.2 Smart Animal Welfare Monitoring SystemIn [74], a smart farm assisted by edge computing for thewelfare of animals is proposed. The proposed architectureconsists of an animal-centric system, environmental system,and farm controller. The farm controller can be implementedeither in a centralized fashion or in a distributed fashion. ARaspberry Pi (R-Pi) which is a small sized micro-computer isused in the implementation of the environmental subsystemand animal-centric system. In system implementation, theenvironment R-Pi monitors the environment, whereas wear-able R-Pi monitors the health of animals. A workstation thatimplements the farm controller performs the central tasksof the system. One of the most prominent advantages of theproposed architecture is its ability to send alarms to mobile.These alarms are use full to detect whether animals are illor not, especially when the cow is pregnant which can bedetected from its body temperature [86]. The authors did

14

not consider the robustness of the sensors network used forthe collection of data which might suffer from the data lossdue to battery depletion or other sources as observed in the[87].

3.6 Smart Buildings

Smart buildings provide the users with advanced auto-mated controls, i.e., smart meters, smart water control, smartplug-in hybrid electric vehicles (PHEV) charging, smartelevators, and smart security.

3.6.1 IoT based Smart HomesA smart home architecture is presented in [76], that lever-ages edge computing and Long-Term Evolution (LTE) net-work. Different alternatives for placement of edge com-puting servers are considered that include:(a) placementof edge computing server at Evolved Node B (eNB), (b)placing the edge computing server at femtocells BS in ahome, and (c) placing the edge computing server at UserEquipment (UE). The positioning of edge computing serverat eNB needs to cover a large number of far users andthus, require high computational power. On the other hand,placement of edge computing server near the UEs requireless capacity due to its small coverage area. Simulationfor two different scenarios, such as device-to-infrastructure(D2I) and D2D communication in IoT was performed usingSimuLTE tool[88].

3.6.2 Sustainable Smart HomesA three-tier framework for smart homes based on IoT andedge computing is proposed in [77]. Three tiers of theframework are a smart homes tier, a nanogrid tier, and amicrogrid tier. The smart home tier consists of householditems with interfaces to enable wireless communication.All of data at smart homes is collected at the sink, whichhas ability of local processing and storage. In the nanogridtier, sinks of different smart homes communicate with eachother. To enable communication between sinks, differenttypologies, such as mesh, cluster or star can be used. In themicrogrid tier, sinks of different smart homes communicatewith utility through gateways. IoT devices have limitedenergy, therefore edge computing enabled gateways can beused to provide the computation services to smart homes.Major limitation of the framework lies in lack of consideringsecurity which poses significant challenges to smart homesoperation. A malicious might access the IoT-based smarthome objects and operates it in an irregular way [89].

3.7 Smart Mobile Gaming

Smart mobile gaming leverages AR/VR, the Internet, edgecomputing, smart devices, among others, to provide immer-sive gaming experiences. In [78], Premsankar et al. discusseddifferent edge computing platforms and architectures. Thekey applications of IoT that leverage edge computing arealso discussed. Along with this, the authors also carriedout an experimental evaluation of edge computing enabledmobile gaming. Such mobile gaming relies on high relia-bility and low latency communication along with mobiledevices sensors data [90]. For example, Pokemon Go [91] is a

game that leverages sensor information (i.e., location of theuser) and AR [92]. The authors selected a resource-intensiveand open-source 3-D arcade game, Neverball, for evaluatingthe performance of edge computing in mobile gaming. Theauthors considered scenario where an input from the mobiledevice is sent to the gaming server. The server, in turn,streams the video and provided the content to the mobiledevice. In such a scenario, edge computing is necessary toenable the users with AR assisted mobile gaming [93], [94].

3.8 Smart Management

In the smart city era, information and communication tech-nologies are incorporated to enable smarter urban infras-tructure management for improving overall user QoE. In[79], an Intelligent Manhole Cover Management System(IMCS) that leverages edge computing is proposed foravoiding accidents due to the manhole cover. The proposedIMCS comprise of three parts: perceived recognition tech-nology, network transmission technology, and informationprocessing technology. In perceived recognition technology,a Radio-Frequency Identification (RFID) tags and sensorsare attached with manhole to facilitate active communica-tion. In the second part, narrow band-IoT and the Internetare used for communication between the manholes and endusers. The third part consists of an edge computing enabledintelligent management system that allows immediate de-cision making. The authors considered context-awareness,scalability, and sustainability in the design of IMCS, but didnot consider security.

3.9 Lessons Learned: Summary and Insights

From this section, we learned that edge computing is aviable for realizing real-time smart cities advancements.We derived five parameters, such as context-awareness,scalability, sustainability, caching, and security, from recentadvances literature for rigorous evaluation. We categorizedthe recent advances into different application areas andsummarized their key contributions in table 4-5. We thenderive that security has considerable importance among theother factors and thus must be implemented in the designof smart applications. For instance, consider autonomousdriving and smart health-care enabled by edge computing;malicious user can access the system and information, whichmight result in accidents. Other than security, the incorpo-ration of context-awareness in edge computing can improvethe performance of emergency response systems.

4 TAXONOMY

Figure 5 elucidates the taxonomy which is devised usingedge intelligence, edge analytics, resources, caching, charac-teristics, sustainability, security, and resource management.Further details are provided in the following subsections.

4.1 Characteristics

Edge computing paradigms generally appear in three differ-ent formats, such as cloudlets, fog computing, and mobileedge computing [37]. All of these formats use the samenotion of migrating the computing resources from the cloud

15

Taxonomy

Edge Intelligence

Micro Services

Virtual Machines

Containers

Edge Server Evolution

End Node Evolution

Advanced Networking

Machine Learning

Edge Analytics

Query Tools

Decision Trees

Statistical Tools

Mathematical

Programming

Optimization

Standard Reports

Adhoc Reports

Data Mining

Forecasting

Resources

Edge Node Storage

Local Storage

Server Computation

Local Computation

Access Network

Core Network

Security

Authentication

Encryption/

Decryption

Service

Availability

Devices Physical

Security

Resource

Management

Resource

Discovery

Resource

Management

Tools

Optimization

Hybrid

Graph Theory

Game Theory

Heuristic

Fairness

QoS

Energy

Throughput

Characteristics

Mobile Edge

Computing

Cloudlets

Fog Computing

Medium Context-

awareness

Partial Internode

Communication

Data Center in a Box

Low Contex-awareness

Internode Communication is

Support

Heterogonous Fog

Nodes

High Context-awareness

Partial Internode

Communication

MEC Servers at BS or

Radio Access Network

Sustainability

Energy Efficient

Design

Renewable

Energy Sources

Energy

Harvesting

Biomass

Geothermal

Wind

Environmental

Radio

Frequency

Energy Efficient

Sotware

Energy Efficient

Resource Management

Energy Efficient

Hardware

Caching

Caching Control

Caching

Mathematical

Tools

Content Caching

Service Caching

Access Point/

Base Station

User Device

Core Network

Distributed

Centralized

Game Theory

Optimization

Resource

Management

Criteria

Cache Placement

Content

Diagnostic

Prescriptive

Descriptive

Predictive

Software

Requirements

Hardware

Requirements

Enablers

Storage

Resources

Computation

Resources

Communication

Resources

Fig. 5: Taxonomy of edge computing enabled smart cities

to the edge of the networks. However, there exist unique dif-ferences among them in terms of context-awareness, inter-node communication, and the position of the computingresources. Mobile edge computing has the highest context-awareness due to their availability of user device location,network load, and network capacity at the edge servers [95].Cloudlets have the lowest context-awareness due to theirstandalone architecture, which is connected to a cloud. Fogcomputing has context-awareness between that of mobileedge computing and cloudlets. The ability of inter-nodecommunication somewhat improves the context-awarenesslevel of fog nodes. Other than context-awareness, inter-nodecommunication has more support in fog computing thanboth cloudlets and mobile edge computing.

4.2 Edge AnalyticsEdge analytics deals with the analysis of data near thenetwork edge, and then sending back these results to the

edge devices. This has a significant impact on reducing thedelay incurred when sending data to a centralized cloudfor analysis. In smart cities many devices, including smartsensors, and other smart devices from a variety of real-time smart applications, including augmented reality, smarthealth-care, smart industrial control, smart transportation,and smart surveillance, generate a huge amount of data. Toenable such real-time applications, it is necessary to performdata analytics at the edge rather than at the centralizedremote cloud. In the future, fog nodes, cloudlets, and mobileedge computing servers will be designed with the ability toperform big data analytics.

Similar to cloud computing, edge analytics can be de-scriptive, diagnostic, predictive, and prescriptive. Descrip-tive analytics is a simple class of analytics that enablesusers to transform big data into small chunks with usefulinformation. It is preferable to use descriptive analyticswhen understanding is required at an aggregate level. This

16

allows users to learn from past events, which can be fromeither not long (e.g., 2 min) or from a long time ago (e.g.,10 weeks). Examples of descriptive analytics are percentagechanges, average, sum, among others. Predictive analyticsallow the use of machine learning and statistical modelingfor analyzing historical data and predicting future trends.Sentimental analysis is an example of predictive analysis.Prescriptive analytics allows performing near optimal oroptimal actions in accordance with the predicted or actualdata. Prescriptive analytics include mathematical program-ming (such as integer programming, non-linear program-ming, and linear programming.) and optimization. On theother hand, diagnostic analytics is based on examining tounderstand ”Why somethings happened?”. For diagnosticanalytics at the edge, query tools, decision trees, and statis-tical tools can be used.

4.3 ResourcesEdge computing converges information technology andtelecommunication services to enable computation ofresource-intensive tasks at the network edge. To carry outthe above process, we need three types of resources, suchas computation resources, communication resources, andstorage resource [24]. Computation resource can be eithera local device or server. A user offloads either all compo-nents of an application (binary offloading) [96]–[98] or somecomponents of an application to the edge server (partialoffloading) [99], [100]. A complication might arise in partialoffloading when the task components are dependent oneach other, and making it difficult to separate the taskcomponents for local execution and execution at server.Apart from computation resource there are two types ofcommunication resources, that are required to enable edgecomputing in smart cities. One is the access network, whichallows connectivity of the end-user devices to the edgeservers, and the other is a core network communicationresource, which allows transmission of data from the edgeserver to cloud server at the time of insufficient computationresources at the edge server. Different from computation andcommunication resources, storage resources can be either ata local device or at an edge.

4.4 Resource Management StrategiesIn smart cities enabled by edge computing, the key re-sources are the massive number of smart devices, edgeservers, and communication infrastructure. To enable edgecomputing-based smart environments, it is necessary toeffectively manage the available resources. The main aspectsof resource management include resource discovery, re-source management tools, and resource management crite-ria. Fairness [101], [102], QoS [103], [104], throughput [105],energy [106]–[109], among others, can be utilized as criteriato design resource management schemes. Other than re-source management criteria, we can use optimization-basedschemes [110], game theory-based schemes [111], heuristic[112], graph theory-based schemes [58], and hybrid schemes(which jointly utilize optimization and game theory) [113],for resource allocation. An optimization-based techniquecan be either dynamic programming, linear programming,convex optimization, or integer programming. On the other

hand, game theory-based schemes can be based on eithercooperative games or non-cooperative games.

4.5 Edge IntelligenceEdge intelligence will transform edge computing analyticscapabilities to the next level by enabling self-learning so-lutions. The three fundamental aspects of edge intelligenceare key enablers, hardware requirements, and software re-quirements [114]. Edge intelligence can be enabled throughadvanced network capabilities and machine learning. Otherthan key enablers, the edge server must be evolved to yielda virtualized and truly dynamic micro data center. Addition-ally, end devices must be evolved to handle intelligent oper-ations. Apart from hardware evolution, the key goal of edgecomputing is to isolate the functionality from the hardwarethat is provided by services form. Such isolation has the ad-vantage of being deployable anywhere in edge computingenabled smart cities. The concept of micro-services can beused for isolation, whose deployment requires the same run-time environment at all deployment locations. Both virtualmachines and containers can be used to run micro-services.However, containers offer less overhead than virtual ma-chines. Moreover, containerization is also preferable in thecloud for deployment of micro-services. Therefore, at theedge it is also preferable to use containerization to providea uniform run-time environment.

4.6 CachingCaching is temporarily storing the network contents toavoid repeated transmissions. In a typical communicationnetwork, the back haul-links suffer from congestion dueto repeated requests of the same contents. For instance, inthe Internet traffic, the video stream is the major source oftraffic and as counts for 54% of all traffic, which is estimatedto reach 71% in 2019 [115]. Therefore, it is necessary tostore the video contents temporarily at different locations toavoid repeated transmissions. Edge computing with cachingcan be used to enable resource-intensive applications, suchas smart industries, smart transportation, augmented real-ity, smart tourism, to name a few [116]–[119]. Caching inedge computing enabled smart cities has four main aspectssuch as cache location, cache contents, caching control, andcaching mathematical tools. Caches can also be placed eitherat the core network or at the Access Point (AP)dependingthe on the requirements. Cache placed at the core networkrequires a high capacity due to the fact that multiple accesspoints will be connected to it. Apart from that, a cachecan be placed at the AP or at the UE. Different from cacheplacement, caching control may be either centralized [120]or distributed [121]. Furthermore, control structure andmathematical tools for caching can be based on machinelearning, optimization, and game theory [122], [123].

4.7 SustainabilitySustainability refers to the use of energy efficient designsand renewable energy resources to reduce the overall carbonfootprint. In future smart cities, densification of devices andservers is expected, which in turn will result in energy lim-itations. Therefore, sustainable developments are required.

17

Sustainability in edge computing can be achieved by usingrenewable energy sources [124], energy harvesting [125]–[127], and energy-efficient design [128]. Renewable energysources include biomass, geothermal, and wind energy.Energy efficient design has three important aspects, suchas energy efficient software, energy-efficient hardware, andenergy efficient resource management. Other than energyefficient design and renewable energy sources, energy har-vesting deals with the extraction of energy from externalsources for further use in the operation of small devices,such as sensors and wearable devices. Energy can also beharvested either from the environmental sources, such aswind, solar, and thermal, or from radio frequency sources.There are significant variations in the harvested energy fromboth natural sources and radio frequency sources [125].Therefore, it is imperative to design hybrid systems thatutilize both harvested energy and energy from the grid. Ahybrid system initially must entirely use harvested energy,and then later use energy from the grid in instances whenthe harvested energy level becomes lower than the requiredenergy for operation.

4.8 SecurityEdge computing infrastructure exhibits novel security risksand challenges due to its distributed nature in contrast toa centralized cloud [37], [129]–[131]. In edge computing,numerous players, such as virtualization platforms, dis-tributed and peer-to-peer systems, and wireless networkingtechnologies are involved [132]. To enable security, we mustprotect all of these players from attacks by implementingdiverse security mechanisms, whose orchestration is oftenvery complex [37]. In addition, the security schemes mustbe autonomous due to the decentralized nature of edgecomputing. The prominent benefit of autonomous and dis-tributed security mechanisms is their lower latency com-pared to centralized security mechanisms.

Generally, edge computing security risks can be clas-sified into denial of service attacks, man in the middlesituations, physical damage, and privacy leakage [37]. Anattacker can easily access end user devices or connect tothe edge servers, and then install malicious software onit. Therefore, it is very difficult to secure end-user devicesand edge servers physically due to their geographicallydistributed nature. On the other hand, in man in the mid-dle attacks, a malicious user takes control of the wirelessnetwork, followed by traffic injections, whereas privacyleakage refers to unauthorized access of a users informationstored at the edge data center. Denial of service attacks referto wireless jamming. Both denial of service and privacyleakage have a limited scope than a man in the middleattacks. Therefore, to avoid man in the middle attacks, wemust use effective and lightweight encryption/decryptionschemes [133]–[135]. Furthermore, effective authenticationprotocols must be devised for edge enabled smart cities toensure user privacy [136]–[138].

4.9 Lessons Learned: Summary and InsightsIn this section, we classify the literature and devise a taxon-omy. We found numerous aspects, such as edge analytics,edge intelligence, resources, caching, resource management,

characteristics, sustainability, and security, to classify theliterature. From this section, we learned numerous lessons:

• We derived the open research challenge of Intelli-gent Edge from this section, that is further discussedalong with possible guidelines in section. 7.4. Edgeanalytics can play an important role in future smartcities. Massive numbers of smart devices and sensorsproduce a huge volume of data, out of which onlya small portion of data is useful. This requires datafiltering and intelligent instant processing, whichcan be enabled by intelligent edge-based ArtificialIntelligence (AI).

• The use of hybrid energy sources to enable sus-tainable operation was also derived in this section.Hybrid energy sources make use of both harvestedenergy and grid energy on demand. First, such asystem utilizes the available harvested energy, andthen the grid energy if the harvested energy levelfalls below the required energy level.

• The challenge of Intelligent caching which is dis-cussed in detail along with possible guidelines insection. 7.1, is also derived in this section. Intelli-gent caching based on AI and machine learning isexpected to offer a considerable decrease in latencyfor real-time applications. Moreover, it increases thecore network throughput and offers a solution tomobility-aware caching.

5 REQUIREMENTS

This section outlines and explains core requirements to turnthe vision of edge computing enabled smart cities intoreality. These requirements (illustrated in figure 6) includescalability and reliability, resource management, interoper-ability, sustainability, elasticity, context-awareness, security,and business models. A more detailed description of therequirements is given below.

5.1 Scalability and ReliabilityHow will edge computing enabled smart cities infrastruc-ture simultaneously respond to a gigantic increases in thenumbers of end-user devices and system failures?

In smart cities, we will witness massive increases inthe numbers of actuators, smart sensors, and other smartdevices, subsequently resulting in unprecedented growthin data traffic. To enable the resource-intensive and latencysensitive smart city applications, edge computing provides asolution. To create a scalable edge computing infrastructure,we must consider reliable performance for an increasingnumber of end users, dynamic smart devices, and diversenetworks [13]. Scalability in edge computing infrastructurecan be achieved by the addition of new service providingpoints. Addition of new service points require scalablehardware and software. On the other hand, prominentchallenges of reliability in edge computing infrastructuremight arise because of its dynamic nature, software flaws,hardware flaws, and network failure. To cope with thesystem failures, we must use edge computing based infras-tructures within smart cities which are resilient to hardwareand software flaws. Other way is to use duplicate service

18

Edge Enabled

Smart City

Requirements

Scalability &

Reliability

Resilient to

Hardware Failures

Resilient to

Software Flaws

Scalable Hardware

Scalable SoftwareSustainability

Hybrid Energy

Sources

Energy Harvesting

Renewable Energy

Sources

Energy Efficient

Design

Resources Management

User-defined Utility-

based Resource

Management

Adaptive Resource

Management

Business Model

Pricing Model

Services and

products

Customers

Distribution Context-awareness

Network Load and

Capacity

Smart Devices

Location

Elasticity

Prediction of User

Demands

Elastic Services

Security

Decentralized

Security

Cyber Security

Interoperability

Interoperable

Interfaces

Open Source

Frameworks

Fig. 6: Edge computing enabled smart cities requirements

providing points. However, this will cause a significant in-crease in the cost of a system. Moreover, this approach addscomplexity to the system because of its distributed nature.As the edge computing is mainly intended to enable strictlatency applications. Therefore, the addition of duplicatenodes must be immediate, which is very challenging dueto their distributed nature. Other ways to ensure reliabilityinclude the use of robust hardware, fault-tolerant network-ing protocols, and robust software.

5.2 Context-awarenessHow does one enable smart cities with mission-criticalapplications such as smart health-care services and smarttransportation services using edge computing?

To answer this question, the edge server must becontext-aware (such as it must know the locations of smartdevices) to enable computation with strict latency [139]–[142]. Moreover, the edge server requires knowledge ofnetwork load and capacity to enable resource-efficient com-munication with smart devices. For example, if an accident

occurs in a smart transportation system, then it is necessaryto respond as soon as possible for first aid. The arrival of firstaid services is heavily dependent on the user emergencycall. A severely injured user might not be able to call foremergency service. On the other hand, context-aware edgeservers at RSU can enable reporting of emergencies in atimely manner with very low latency. Therefore, it is evidentthat context-awareness is one of the key requirement in thedesign of edge computing enabled smart cities infrastruc-ture.

5.3 Resource ManagementHow does one effectively manage computation, communi-cation, and storage resources between different players insmart cities to optimize the overall system performance?

To answer this question, we must devise resourcemanagement schemes that consider energy [143]–[146], la-tency [147], traffic handling capacity [58], and user-definedutility [148] to optimize the overall performance of thesystem [149]. However, it is very difficult to simultaneously

19

consider all of these factors when designing a resourcemanagement scheme. There are three main categories of re-sources in edge computing enabled smart cities: computingresources, communication resources, and storage resources[150]. To enable cost-efficient operation with better QoS,resource management algorithms must be adaptive andperform management according to an applications nature.For example, more computing resources must be given toa self-driving car than AR to enable faster execution of theself-driving car functions to avoid accidents [151]. On theother hand, the throughput requirement of AR is larger thanfor self-driving cars [152]. Therefore, a trade-off must bemade between computation and radio resource allocation.

5.4 Sustainability

How does one develop edge computing enabled smart citieswithout losing QoS and adding pollution to the environ-ment?

Sustainability is another important requirement whendesigning edge computing infrastructures for smart cities.In smart cities, the densification of end-user devices andedge computing servers is expected. Consequently, thisputs significant energy limitations on the overall smart cityinfrastructure design. Therefore, it is imperative to tackle thechallenge of high energy consumption without degradingQoS. The three main aspects are related to sustainability ofthe edge computing enabled smart cities: energy efficient de-sign [153]–[156], use of renewable energy sources [80], [157],[158], and energy harvesting [159]–[162]. The edge comput-ing enabled smart cities infrastructure is designed to reduceenergy consumption by employing a novel architecturedesign and communication technologies. Numerous waysto improve the energy efficiency of edge computing sys-tem design include energy-efficient caching [153], energy-constrained offloading and sleeping [154], and energy-aware computation offloading and user association [156].Furthermore, energy-efficient wireless networking technolo-gies, such as IPv6 over low-power WPAN (6LoWPAN) andZigBee can be used to enable sustainable operation of asystem [163]. Other than energy efficient design, an edgecomputing ecosystem that utilizes energy from the fossilfuels-based power plants will add to the carbon footprint.Therefore, an edge computing infrastructure must utilizeenergy from renewable energy sources such as wind, solar,and hydropower. Apart from that, energy can be harvestedeither from environmental sources, such as wind and sun[159], [160] or radio frequency sources, such as from radiofrequency transmitters and interference signals [161], [162],[164]. Energy harvesting offers significant advantages butsuffers from random variations in both environmental andradio frequency energy. Therefore, we must use hybridsources that also utilize harvesting energy and grid energyon demand if the harvested energy level falls below therequired energy.

5.5 Elasticity

How does one enable smart citizens with services in anelastic manner to avoid both under and over-utilization ofresources?

Unprecedented increases in the numbers of sensorsand other smart devices are expected in smart cities. Thiswill impose limitations on the available resources, such ascomputing, communication, and storage resources. There-fore, it is necessary to design systems that permit efficientutilization of the resources. Elasticity in edge computingenables to avoid over-allocation and under-allocation ofcomputing resources. This can be realized in smart cityinfrastructure by providing elastic services based on theallocation of resources with the ability to grow and shrinkdynamically according to demand [165]–[167]. For elasticity,it is necessary to properly predict user demand via edgeservice providers prior to assignment of computational re-sources, such as bandwidth, processing power, and storage.Other than prediction, edge computing-based smart servicesmust be elastic. In [168], a proactive scheme is presentedto enable auto-scaling by predicting future resources basedon past data. Auto-scaling techniques can be either vertical,horizontal, or hybrid [169]. Vertical scaling refers to re-sizingof resource allocation on the same node, whereas horizontalscaling is re-sizing of resource allocation among a clusterof nodes cluster. A hybrid approach combines both verticaland horizontal schemes.

5.6 InteroperabilityHow does one operate the massive number of heteroge-neous devices in edge computing enabled smart cities?

Interoperability is the ability of different systems tounderstand and utilize each others functions. Similar to IoT,interoperability is one of the most important requirementsin the design of the edge computing enabled smart cityinfrastructure. The challenge of interoperability in edgecomputing enabled smart city infrastructure arises due tomassive number of heterogeneous devices running differentprotocols. To enable seamless operation, edge computingarchitectures must be able to provide support for interoper-ability. Interoperability can be implemented in smart citiesby exploiting inter-operable interfaces and open sourceframeworks.

5.7 Privacy and SecurityHow does one preserve user privacy and data from attacksby malicious users?

Security in edge computing infrastructure has twomain aspects: devices cyber security and physical security.Cyber security refers to protection of the network, comput-ing infrastructure, and data from attacks. Specifically, datasecurity is the protection of data from unauthorized access,whereas privacy can be considered as legislation (a set ofrules) that defines the levels of protection. Privacy placesrestrictions on the authorization of users regarding dataaccess and the purpose of data usage. Apart from privacy,security must be properly addressed in smart cities. At thecore of edge computing paradigms, there are many differenttechnology enablers, i.e., virtualization, wireless networks,peer-to-peer, and distributed systems. Therefore, enablingphysical security of edge devices is necessary and very chal-lenging (due to the distributed nature of the edge nodes).An unauthorized user can easily physically access any ofthe edge devices and install malicious software. Other than

20

physical security, effective cryptographic techniques can beused to provide improved security against the attacker. Fur-thermore, security mechanisms in edge computing ecosys-tems can be either centralized or decentralized based oncontinuous dependence on the centralized infrastructure.Centralized security requires continuous availability of thecentralized infrastructure. Although this has the advantageof easier management it is more vulnerable to failure byan attack on the centralized server. On the other hand, adecentralized security mechanism has the advantages of re-duced delay and being less vulnerable to failure by attacks.However, such a mechanism has a higher complexity thancentralized algorithms due to its distributed nature.

5.8 Business ModelHow does one increase revenue of the edge computingservice providers while maintaining user QoS?

Smart cities enable a plethora of horizontal and verticaluse cases that have diverse requirements. Additionally, thecities involve different players, such as end users, cloudproviders, edge computing service providers, and commu-nication infrastructure providers, interplay with each other.The already available business models for cloud computingare not feasible because of the distributed nature of edgecomputing. Therefore, to commercially deploy the edgecomputing infrastructure, the existing service providers willhave to adopt new pricing models to generate more revenuewhile maintaining a better QoS.

5.9 Lessons Learned: Summary and InsightsIn this section, we discussed a few imperative requirementsfor realizing edge computing enabled smart cities. Theserequirements are scalability and reliability, sustainability,resource management, business models, context-awareness,elasticity, security, and interoperability. The lessons learnedfrom this section are:

• The open research challenge of collaborative edgecomputing is derived from the requirements of scal-ability and reliability. References [170] and [171]considered collaborative edge computing to offerscalability and reliability. Edge computing serversare characterized by a low computing capacity thatmust be utilized efficiently. Additionally, there aresignificant variations in smart device traffic. To copewith variations in the massive amount of smartdevices traffic along with scalability and reliabilityconstraints, the use of collaborative edge computingis a viable solution.

• The development of novel business models for edgecomputing enabled smart cities is also derived fromthis section. The already developed business modelsfor cloud computing are insufficient for edge com-puting enabled smart cities due to their distributedarchitecture.

• The adoption of high-performance memory is alsoderived from the requirement of sustainability andreliability. The edge servers are intended for usein multiple diverse applications. Moreover, edgeservers are also expected to enable support for com-putationally expensive machine learning algorithms.

Therefore, high-performance memory is a promisingsolution.

6 CASE STUDIES

This section presents recently reported synergies and casestudies of edge enabled smart cities. A summary of thesecase studies listing the goals, the organization involved,deployment status, and country is given in the section.

6.1 Barcelona smart city

Barcelona smart city is a joint project of Barcelona CityCouncil, Prismtech, i2cat, Technical University of Catalo-nia, Barcelona Supercomputing Center, Schneider Electric,Plat.one, Sensefields, and CISCO to realize smart services[172], [173]. The project partners focused on five use casesthat include power monitoring/element control, access con-trol and cabinet telemetry, event-based video, traffic man-agement, and on-demand connectivity. To realize a smartcity, installation of large number of IoT driven devices isrequired, further creating operational and space problems.In this regard, the number of deployed cabinets is more than3000 for Barcelona smart city realization. On the other hand,fog computing is validated to enable real-time decision mak-ing, power supplies for autonomous operation, guaranteedaccess control, local data analysis, and execution of complexalgorithms for the assistance of vehicles flow.

6.2 #SmartME

#SmartME is a project provisioned to transform the city ofMessina into a smart city [174]. The project was initiatedby a group of researchers from the University of Messina.The key goals of the #SmartME project are establishmentof a smart city infrastructure that allows all citizens tocontribute to infrastructure through hardware sharing. The#SmartME framework consists of three-layer: an applica-tion layer, Stack4Things layer (for fog computing platformimplementation), and city layer. The Stack4Things layer isa framework developed at the University of Messina withthe objective of enabling the administrators to manage IoTdevices without taking care of their physical location. Mainfeatures of the Stack4Things include object virtualization,overlay network of things, remote control and customiza-tion, and fog orchestration. The applications developed onthe top of the #SmartME infrastructure include #SmartMEParking, #SmartME lighting, #SmartME pothole, #SmartMEairport, #SmartME taxi, #SmartME art, and #SmartME trash-can.

6.3 WATCH

The Wide Smart Safe, Robust and Resilient Smart Cities Appli-cation Using Fog Computing (WATCH) project is funded byINNOVATE UK [175]. It was started on Nov 01, 2017 and isexpected to be completed by 2020. The lead participants ofthis project are Future Intelligence Ltd. London, and LondonSouth Bank University, United Kingdom. The goal of thisproject is to exploit novel micro-data center and telecom-munication nodes to provide joint storage resources, local

21

processing, and networking to enable support for user ap-plications in smart cities. Further, they will utilize software-defined networking and Network Function Virtualization(NFV) to enable the formation of interconnected devicesislands. This resulted in fogs, which are actually smallscale clouds at the network edge (i.e. edge computing).The main objective of this project is improvement of smartsurveillance by leveraging edge computing.

6.4 Lessons Learned: Summary and InsightsThis section presented the recently reported case studiesof edge computing enabled smart cities. We discussedthe objectives, organizations involved, and their currentdeployment status. From this section, we concluded thatthe development of high-performance simulation tools forimplementation and validation of edge computing basedsmart cities applications is required to be done. For instance,the IoTIFY is an online cloud based network simulatorthat enables simulation of smart waste management, smartparking spots, smart street lights, smart traffic signals, andsmart transportation [176]. We must develop novel high-performance simulators to test edge computing and otheremerging technologies in smart city environments.

7 OPEN RESEARCH CHALLENGES

This section presents open research challenges in the re-alization of edge computing in smart cities. Furthermore,a summary of these challenges, their causes, and possibleguidelines is given in table 6.

7.1 Intelligent CachingIn smart cities, connectivity of billions of smart IoT devicesis expected. According to CISCO, 507.9 ZB data will be gen-erated by smart IoT devices in 2019 [177]. This rapid rise indata has resulted in a major bottleneck that requires higherback-haul data rates. To tackle the dilemma of back-haulcongestion, caching can be used which deals with the stor-age of popular contents at different locations in a networkto avoid repeated transmissions [178]. Additionally, cachingenable different real-time IoT-based smart city applicationsby reducing latency and energy consumption [179]. Onthe other hand, caching has two main aspects: networktype and caching strategy. Network type deals with cachepositioning, such as cache enabled macro cellular networks[120], cache enabled heterogeneous networks [180], cacheenabled D2D networks [181], and cache enabled cloud radioaccess networks [182]. The caching strategy has two mainphases, such as cache decision phase and content deliveryphase.

The main challenges that exist in intelligent caching atthe edge include the cold-start user problem [183] and secu-rity and privacy [184]. Intelligent caching based on machinelearning uses information and data from all users withinits range. However, the mobility of the users poses manychallenges and it is very difficult to satisfy this requirementfor intelligent caching. To cope with user mobility, mobility-aware hierarchical caching was proposed in [117]. Further-more, we can leverage reinforcement learning for effectivescheduling in mobility-aware caching [185]. Additionally,

intelligent cache decisions are based on the future contentpopularity, whose computation is considerably challenging.The prediction of content popularity using deep learninghas been considered to enable intelligent cache decisionmaking [183], [186]–[188].

7.2 Collaborative Edge Computing

Virtually enabled edge computing servers are characterizedby low computation and storage capacity constraints [189],[190]. They use containers and virtual machines to enabledifferent smart applications. Apart from that, edge comput-ing servers can leverage dimensional parameters, such asserver number, server capacity, and server operation areato optimize the overall performance in terms of QoS andoperator profit. On the other hand, a tremendous increase inthe number of smart devices is expected in the future. Con-sequently, this poses significant challenges pertaining to thescalability and QoS of smart city infrastructure. To enablescalable and resource-optimized operation of edge comput-ing in 5G enabled smart cities, we can leverage collaborativeedge computing. Collaborative edge computing aims toform a collaboration space between edge computing serversin close vicinity. Moreover, this enables the maximizationof resource usage, latency minimization, and reduction ofback-haul traffic. Collaboration in edge computing enabledsmart city infrastructure can be categorized into horizontaland vertical collaboration [170]. Horizontal collaboration in-volves cooperation among different edge computing nodesthat exist simultaneously in the same edge computing layer.Vertical collaboration involves collaboration among differ-ent layers, such as the smart devices layer, edge computinglayer, and remote cloud computing layer.

The major challenges that exist in collaborative edgecomputing include collaboration space formation, socialtrust-based incentive policies, cooperation policies, inter-domain cooperation, intra-domain cooperation, smart col-laborative networking, and mobility management. Collab-oration space formation must be adaptive and based onresource availability. Furthermore, within the collaborationspace, collaboration schemes must be designed that mini-mize back-haul bandwidth usage subject to the constraintsof server resource availability, tasks computation deadlines,and local computation capabilities. Apart from formationof the collaboration space, we must design social trust-based incentive policies and cooperation policies. Within acollaboration space, only edge computing nodes with socialtrust can share resources with each other based on incentivesin the form of payment. A coalition game was used in [191]to design pricing incentive based on trust for collaborationin ultra-dense networks. Additionally, mobility must bemanaged properly within a collaboration space to enableefficient resource usage. The design of collaboration algo-rithms regarding the mobility of resource deficient nodes,such as autonomous vehicles and UAVs, is challenging.

On the other hand, we must design novel algorithmsfor intra-domain and inter-domain collaboration amongedge computing nodes. Inter-node collaboration occurs be-tween mobile edge computing servers, cloudlets, and fognodes, whereas, intra-domain collaboration involves similaredge computing nodes, such as mobile edge computing

22

TABLE 6: Summary of the research challenges and their guidelines

Challenges Causes Guidelines

Intelligent caching• Data sparsity due to mobility• High back-haul congestion• High latency in content delivery

• Deep learning assisted cache content pre-diction

• Reinforcement learning based mobility-aware caching

Collaborative edgecomputing • Edge server inherent low computation and

storage capacity constraints• Profit and QoS dependency on size, num-

ber, and coverage area of edge servers• Absence of high scalability and reliability

• Horizontal edge computing collaboration• Vertical edge computing collaboration• Inter-domain collaboration• Intra-domain collaboration• Smart collaborative networking• Social trust based incentive design for col-

laboration

Cooperative and sus-tainable load balanc-ing

• Under-loaded scenarios• Over-loaded scenarios• Substantial energy wastage• Profit margin reduction

• Novel game theory assisted incentive mech-anisms

• Adaptive cooperation mechanisms• Novel authentications based load balancing

Intelligent edge• Absence of instant big data analytics• High latency in instant and autonomous

decisions• Semantic non-interoperability

• AI based predictive and diagnosis analytics• Machine learning based semantic inter-

operable interfaces• Novel privacy regulation rules for personal

data sets

Network slicing• Diverse requirements of smart city applica-

tions• Lack of effective resource management of

multiple service providers• Slice security challenges

• Separation of control plane from data plane• End-to-end slice management and orches-

tration• SDN-based orchestrator security• Adaptive service function chaining

Cyber and physical se-curity • Privacy leakage

• Denial of service• Service manipulation• Multiple trust domains

• Lightweight authentication schemes• SDN and NFV assisted security• Attribute based access control• Federated capability-based access control

nodes, fog nodes, or cloudlets. Enabling intra-domain col-laboration is easier than intra-node collaboration due tothe homogeneous nature of edge computing nodes. Apartfrom inter-domain and intra-domain collaboration, smartnetworking technologies must be designed to enable latencysensitive and energy optimized collaboration. In [171], thedesign of a smart internet architecture allowing smart col-laboration was proposed. Furthermore, deep learning andgame theoretic approaches can be used to enable smartdevices, edge servers, and other networking devices toperform traffic flow control collaboratively [192].

7.3 Cooperative and Sustainable Load BalancingSmart device data exhibit significant temporal and spa-tial traffic variations resulting in possible congestion atedge computing enabled AP/BS [193]. Consequently, thiscauses performance degradation pertaining to higher la-tency. Moreover, offering edge computing servers with dy-namic loads causes under-loaded and overloaded scenarios.As a result, substantial QoS performance degradation and

profit loss of edge service providers will occur. Additionally,the smart cities infrastructure requirement of high scala-bility is also adversely affected by under-utilization andover-utilization. Hence, it is necessary to perform adequateload balancing. On the other hand, edge computing serversconsist of hardware resources, such as main memory andsecondary storage which are virtualized [194]. Virtual ma-chines are then assigned to requested applications. Everyvirtual machine operation is characterized by two states, theidle state and the active state. The energy consumption ofthe idle state is approximately 60% that of the active state.Therefore, it is necessary to avoid under-utilization of edgecomputing servers to enable sustainable operation.

Numerous challenges exist for cooperative and sus-tainable load balancing, including novel pricing incentivedesign, joint minimization of task dropout and task trans-fer, and secure authentication. In cooperative load balanc-ing, the load balancing between different edge computingservers involves negotiation between the involved players.To enable cooperation between different edge computing

23

servers, we need effective incentive mechanisms. Designof the incentive mechanism depends on the nature of theedge server strategy, such as an open strategy or sealedstrategy. In an open strategy, the edge servers are allowedto share their resource information, whereas the resourceinformation of the edge servers is only available to serviceproviders in a sealed strategy. Game theory can be utilizedin the design of an incentive mechanism to enable cooper-ative load balancing in edge computing [195]. The actionsof the three main players of the edge computing ( i.e.,service providers, users, and edge servers) can be used tomodel various game theory parameters, such as the actionset and preference set. Apart from incentive mechanismdesign, the cooperation mechanism must be adaptive andoffer the ability to trade-off between task dropout and tasktransfer [196]. Furthermore, security must be considered inthe design of cooperative and sustainable load balancing[197].

7.4 Intelligent Edge Computing

A massive amount of data will be generated by smartdevices and sensors for real-time smart city applications thatrequire instant processing. Most of these data contain noiseand only a small portion of it is useful. The challenge is howto separate large useful data sets (such as big data) fromnoisy data. A viable solution to this challenge is the use ofAI in the edge. Intelligent edge computing can perform datafiltering to yield only useful data before sending to a remotecloud. Moreover, it can also offer real-time processing ofdata. Intelligent edge computing offers real-time analyticsthat enable us to answer questions, such as Why are somethings happening and What should be done. Additionally, itenables process analysis and prediction of future trends.More specifically, the use of instant analytics at the edgeis necessary to enable big data analytics. On the other hand,semantic interoperability between different edge computingenabled systems is one of the major bottlenecks regardingthe exchange of information to enable effective trainingof the machine learning algorithms [198]. Intelligent edgecomputing is a way of providing interoperability amongnon-interoperable systems. Additionally, intelligent edgecomputing offers security enhancement by enabling smartapplications to avoid sending data to remote clouds forpredictive analytics.

The challenges relevant to developing intelligent edgecomputing based on AI include its high computationalpower, complex data acquisition and storage, user pri-vacy, and interoperability. The execution of traditional AIalgorithms requires high computational power. However,the computational power at the edge is lower than thatof the remote cloud. Therefore, it is necessary to deviselow-complexity learning algorithms. Apart from its highcomputational power, data acquisition and storage of smartdevices and sensors in various smart environments is alsochallenging. The humongous data sets used for AI algo-rithm validation require high storage capacity. In addition,the data set might contain noisy or missing data, which mustbe handled properly. Another challenge in training AI algo-rithms is the integration of large data sets. Integrating andsynthesizing large data sets from the different players in a

smart environment for effective AI algorithm training has avery high computational complexity. To cope with this issue,an inter-operable system is a solution that enables systemsto interact with each other and share data. In [198], an edgecomputing-based semantic gateway has been proposed toenable inter-operability in health-care systems. To enablesemantic interoperability machine learning-based interop-erable interfaces can be used. On the other hand, maintain-ing user privacy while utilizing data sets for training AIalgorithm pose challenges. For instance, consider an edgecomputing enabled smart health-care system with patientpersonal data. The privacy of users must be maintained viaeffective security mechanisms. Moreover, effective privacyregulation must be devised to ensure the user privacy.

7.5 Network Slicing

In smart cities, a wide range of players includingcloud server providers, edge computing service providers,telecommunication services providers, and IoT serviceproviders, will interplay with one another. The goals ofthese service providers are to increase profit, whereas, userswant to improve their QoS. To enable easier and efficient re-source management among users and service providers, theconcept of network slicing was introduced [199]–[202]. Net-work slicing exploits software-defined networking (SDN)and NFV to separate the control plane from the data plane[200]. The concept of network slicing is based on the creationof logical networks on top of a physical network infrastruc-ture.

To enable network slicing in smart cities, it is impera-tive to tackle the challenges, such as end-to-end slice man-agement and orchestration, slices security, and adaptive ser-vice function chaining. A network slicing architecture con-sists of different layers: a service layer, resource layer, andnetwork slice instance layer [203]. End-to-end slice manage-ment and orchestration spans all three layers and handlesthe creation and management of slices. Moreover, it dealswith translation of all the smart city use cases into slices withits associated network functions. To enable effective net-work slicing, game theoretic, deep reinforcement learning,and auction theory-based schemes can be used [204]–[206].Apart from management and orchestration of slices, slicesecurity is another key challenge in realization of networkslicing in smart cities. The presence of numerous authoritiesand multiple stakeholders in network slicing pose severalnovel security challenges. Network slicing security involveschallenges of multiple operator resource sharing securityand SDN-based orchestrator security. The slices definedfor different use cases have unique security requirements.For instance, a smart health-care slice has stronger securityconcerns than the infotainment slice. Therefore, we mustconsider different security mechanisms for individual slices.Overall, security must be ensured during resource sharingof the infrastructure. However, network wide-security hasdifferent nature than of individual slices security. Therefore,the development of novel and effective security policies fornetwork resource sharing and individual slices is necessary.

On the other hand, SDN-based orchestrator attacksinvolve unauthorized access by a malicious user. An au-thorized user can access only a single network orchestra-

24

tor but still alter the network resources of multiple op-erators. Therefore, meticulous attention must be given toSDN-based orchestrator security. Apart from SDN-basedorchestrator security, another core challenge for networkslicing is adaptive service function chaining. The key driverof network slicing, NFV allows implementation of virtualnetwork functions on virtual machines. To deliver a service,a service function chain can be considered that leverages oneor more virtual network functions in a sequence. The designof an adaptive service function chaining mechanism thatjointly reduces the network deployment cost and maximizesuser QoS is necessary. In [207], traffic-aware and energy-aware virtual network function placements are considered.A solution based on a matching game has been proposed toenable efficient placement of virtual network functions.

7.6 Cyber and Physical SecurityIn edge computing enabled smart cities, the connectivityof a massive number of smart devices is expected, whichcan be prone to serious security attacks. The different areasprone to security attacks include smart devices, networkinfrastructure, edge computing servers, and core infrastruc-ture. These security attacks can be either physical or cyberattacks. Physical security in edge enabled smart cities isdifficult to maintain due to the distributed nature of edgecomputing. A malicious user can easily access the edgenodes and perform malicious activity. Therefore, enablingedge computing with high cybersecurity is of primary im-portance.

The challenges that exist in enabling effective securityof edge computing enabled smart cities includes networksecurity, effective authentication, and access control. Au-thentication offers the ability to determine user validityby sharing identification information between two parties.Traditional cryptography-based authentication schemes canbe used to devise authentication schemes for edge com-puting [208]. However, the high computational complexityassociated with these authentication techniques limits itsapplication to edge computing. Indeed the computationalresources are more limited in edge computing than cloudcomputing. Therefore, we must devise novel light weightauthentication schemes with low complexity. In [209], aLightweight Anonymous Mutual Authentication scheme forntimes Computing Offloading (LAMANCO) was proposedfor anonymous authentication between IoT devices andedge devices. Other than authentication, network securityis another important aspect of smart city infrastructure.Emerging communication technologies such as Sigfox, Lo-RaWAN, ZigBee, 6LowPAN, WAVE, and 5G used in smartcities have their own security protocols. However, there isroom for further research regarding network security inthe context of edge computing. Session keys are negoti-ated using the distribution of credentials and novel andeffective schemes are needed to distribute these credentials.Furthermore, SDN andNFV can be used to enable securenetworking. Specifically, they can be employed for isolationof different types of traffic [37]. They allow for traffic diver-sion from insecure devices towards secure devices.

Similarly, access control in edge computing must begiven considerable attention. Any edge devices after authen-tication, with certain privileges, can access the virtualized

edge servers and misuse them. Additionally, the existenceof multiple trust domains associated with multiple serviceproviders in a single ecosystem pose challenges for de-veloping effective access control schemes. Attribute-basedaccess schemes can be used to restrict access control in edgecomputing [184], [210]. However, the major downside ofattribute-based schemes is the high complexity associatedwith the decryption phase. In [211], an efficient accesscontrol scheme (namely CP-ABE) was proposed for fogcomputing with outsourcing capability. Additionally, a filehierarchy attribute-based encryption scheme was proposedin [212] to enable fine-grained access control for fog comput-ing. On the other hand, [213] proposed federated capability-based access control for IoT to offer effective access controlwith high scalability compared to attribute-based schemes.

8 CONCLUSIONS AND FUTURE PROSPECTS

8.1 Conclusions

Edge computing is a promising computing paradigm toenrich smart cities with instantaneous computing and stor-age resources. Typically, cloud computing is employed tooffer smart devices with computation and storage resources.However, inherent delay of cloud computing has pavedthe way for migrating the computing and storage resourcesfrom a centralized remote location to the edge of the net-work. On the other hand, real-time smart city applicationsrequire instant analytic services. To enable these real-timeapplications, the use of edge computing is required. How-ever, implementation of edge computing in smart citiesposes significant challenges.

In this survey, the adoption of edge computing insmart cities is comprehensively studied. To this end, wefirst presented the evolution of edge computing, discussingthe gradual evolution of computing technologies towardsedge computing. In addition, the technologies involvedand the benefits of different computing paradigms are alsopresented. Second, significant recent advances are presentedand a rigorous evaluation is performed using different as-sessment parameters. Apart from recent advances, we haveclassified the literature and devised a taxonomy accordingto different parameters, such as edge analytics, edge intelli-gence, resources, caching, resource management, character-istics, sustainability, and security. Additionally, numerousrequirements for enabling smart cities via edge computingare proposed. We also reported a few case studies of en-abling edge computing in smart cities. Furthermore, numer-ous open research challenges are discussed in detail, andcauses and possible guidelines to cope with these challengesare summarized.

8.2 Future Prospects

In the future, we expect virtualization to widely be adoptedin developing smart cities. Currently, there are millions ofdevices that are deployed in cities. Further, the additionof billions of smart devices and sensors is expected. Allof these devices generate data traffic that require instantanalytic services. Edge computing will enable such instantanalytics. Apart from edge computing, novel communica-tion technologies with high throughput are required for data

25

TABLE 7: List of abbreviations

Abbreviation Word

AP Access Point

AR Augmented Reality

ARPA Advanced Research Projects Agency Network

AI Artificial Intelligence

BS Base Station

CAGR Compound Annual Growth Rate

CIoT Consumer Internet of Things

DARPA Defense Advanced Research Projects Agency

D2D Device-to-Device

ECG Electrocardiography

eNB Evolved Node B

ETSI European Telecommunications Standards Institute

IaaS Infrastructure-as-a-Service

IoT Internet of Things

IIoT Industrial Internet of Things

IAR Industrial Augmented Reality

IMCS Intelligent Manhole Cover Management System

IoS iPhone Operating System

LTE Long Term Evolution

MCC Mobile cloud Computing

NBIoT Narrow-band Internet of Things

NFV Network Function Virtualization

PHEV Plug-in Hybrid Electric Vehicles

PaaS Platform-as-a-Service

PA Precision Agriculture

QoE Quality of Experience

QoS Quality of Service

R-Pi Raspberry Pi

RSU Road Side Unit

SaaS Software-as-a-Service

SDN Software Defined Networking

RFID Radio-Frequency Identification

UE User Equipment

UAV Unmanned Aerial Vehicle

VDTN Vehicular Delay-tolerant Network

VNG Vehicular Neighbor Group

VR Virtual Reality

transmission. Additionally, the development of the smartcities must be sustainable and reliable. To enable smartcities through novel developments, different players suchas telecommunication network operators, edge computingservice providers, cloud service providers, and IoT serviceproviders will interplay with each other. One issue is howto effectively enable such interplay to maximize profit andimprove QoS? Virtualization-based network slicing is ananswer, which can be leveraged to improve the QoS andincrease profit. This allows to create multiple logical net-works (called slices) on top of a physical infrastructure

of different service providers [214]–[217]. SDN and NFVare key enablers of slicing-based smart cities. SDN offersseparation of the control plane from the data plane, andNFV allows use of a generic hardware for implementationof different network functions.

Network slicing-based smart cities will offer a widevariety of slices for different smart services. Different refer-ences, such as [218]–[221] have considered network slicing-based smart city advancements. These slices have diverserequirements and must share the same physical infras-tructure resulting in a cost-efficient operation. Allocationof an end-to-end network for a typical slice would beexpensive. Therefore, it is necessary for multiple slices tocoexist simultaneously. However, resource management ofresources between co-existing slices is challenging, thusrequiring novel resource management schemes. Differentorganizations are also attempting to enable smart applica-tions via network slicing. For instance, consider reference[218], which introduces a network slicing-based testbedusing edge computing at Hamburg port. Additionally, manyorganizations are working on the AutoAir project in the UKto produce a testbed based on edge computing and networkslicing for testing connected and autonomous vehicles. Tosum up, edge computing and network slicing will be thedominant solution for enabling future smart cities.

APPENDIX

Abbreviations

See table 7.

REFERENCES

[1] E. Ahmed, I. Yaqoob, A. Gani, M. Imran, and M. Guizani,“Internet-of-things-based smart environments: state of the art,taxonomy, and open research challenges,” IEEE Wireless Commu-nications, vol. 23, no. 5, pp. 10–16, 2016.

[2] J. Xie, H. Tang, T. Huang, F. R. Yu, R. Xie, J. Liu, and Y. Liu, “Asurvey of blockchain technology applied to smart cities: Researchissues and challenges, in press,” IEEE Communications SurveysTutorials, 2019.

[3] D. Eckhoff and I. Wagner, “Privacy in the smartcity—applications, technologies, challenges, and solutions,”IEEE Communications Surveys Tutorials, vol. 20, no. 1, pp. 489–516,2018.

[4] N. Ianuale, D. Schiavon, and E. Capobianco, “Smart cities, bigdata, and communities: Reasoning from the viewpoint of attrac-tors,” IEEE Access, vol. 4, pp. 41–47, 2016.

[5] G. Rostirolla, R. d. R. Righi, J. L. V. Barbosa, and C. A. da Costa,“Elcity: An elastic multilevel energy saving model for smartcities,” IEEE Transactions on Sustainable Computing, vol. 3, no. 1,pp. 30–43, 2018.

[6] A. Ramaprasad, A. Sanchez-Ortiz, and T. Syn, “A unified def-inition of a smart city,” in Springer International Conference onElectronic Government, St. Petersburg, Russia, 2017, pp. 13–24.

[7] Y. Mehmood, F. Ahmad, I. Yaqoob, A. Adnane, M. Imran, andS. Guizani, “Internet-of-things-based smart cities: Recent ad-vances and challenges,” IEEE Communications Magazine, vol. 55,no. 9, pp. 16–24, 2017.

[8] R. Buyya, C. S. Yeo, S. Venugopal, J. Broberg, and I. Brandic,“Cloud computing and emerging it platforms: Vision, hype,and reality for delivering computing as the 5th utility,” FutureGeneration computer systems, vol. 25, no. 6, pp. 599–616, 2009.

[9] K. Bilal, S. U. R. Malik, S. U. Khan, and A. Y. Zomaya, “Trendsand challenges in cloud datacenters,” IEEE Cloud Computing,vol. 1, no. 1, pp. 10–20, 2014.

26

[10] S. Lehrig, H. Eikerling, and S. Becker, “Scalability, elasticity, andefficiency in cloud computing: A systematic literature review ofdefinitions and metrics,” in Proceedings of the 11th ACM Interna-tional Conference on Quality of Software Architectures, Montreal, QC,Canada, 2015, pp. 83–92.

[11] Y. Al-Dhuraibi, F. Paraiso, N. Djarallah, and P. Merle, “Elasticityin cloud computing: State of the art and research challenges,”IEEE Transactions on Services Computing, vol. 11, no. 2, pp. 430–447, 2018.

[12] V. Delgado, “Exploring the limits of cloud computing,” Master’sthesis, KTH Information andCommunication Technology, Stock-holm, Sweden, 2010.

[13] E. Ahmed, A. Ahmed, I. Yaqoob, J. Shuja, A. Gani, M. Imran,and M. Shoaib, “Bringing computation closer toward the usernetwork: Is edge computing the solution?” IEEE CommunicationsMagazine, vol. 55, no. 11, pp. 138–144, 2017.

[14] W. Shi, J. Cao, Q. Zhang, Y. Li, and L. Xu, “Edge computing:Vision and challenges,” IEEE Internet of Things Journal, vol. 3,no. 5, pp. 637–646, 2016.

[15] W. Ji, J. Xu, H. Qiao, M. Zhou, and B. Liang, “Visual iot: Enablinginternet of things visualization in smart cities, in press,” IEEENetwork, 2019.

[16] M. A. Rahman, M. M. Rashid, M. S. Hossain, E. Hassanain, M. F.Alhamid, and M. Guizani, “Blockchain and iot-based cognitiveedge framework for sharing economy services in a smart city,”IEEE Access, vol. 7, pp. 18 611–18 621, 2019.

[17] L. Atzori, A. Iera, and G. Morabito, “Understanding the internetof things: definition, potentials, and societal role of a fast evolvingparadigm,” Ad Hoc Networks, vol. 56, pp. 122–140, 2017.

[18] M. Miraz, M. Ali, P. Excell, and R. Picking, “Internet of nano-things, things and everything: future growth trends,” FutureInternet, vol. 10, no. 8, p. 68, 2018.

[19] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, andM. Ayyash, “Internet of things: A survey on enabling technolo-gies, protocols, and applications,” IEEE Communications Surveys& Tutorials, vol. 17, no. 4, pp. 2347–2376, 2015.

[20] Internet of things definition. [Online; Accessed: 019,April 1]. [Online]. Available: https://www.postscapes.com/internet-of-things-definition/

[21] A. K. Sikder, A. Acar, H. Aksu, A. S. Uluagac, K. Akkaya, andM. Conti, “Iot-enabled smart lighting systems for smart cities,”in 8th IEEE Annual Computing and Communication Workshop andConference, Las Vegas, USA, 2018, pp. 639–645.

[22] A. A. Jaafar, K. H. Sharif, M. I. Ghareb, and D. N. Jawawi, “Inter-net of thing and smart city: State of the art and future trends,” inAdvances in Computer Communication and Computational Sciences.Springer, 2019, pp. 3–28.

[23] E. Park, A. P. del Pobil, and S. J. Kwon, “The role of internetof things (iot) in smart cities: Technology roadmap-orientedapproaches,” Sustainability, vol. 10, no. 5, p. 1388, 2018.

[24] N. Abbas, Y. Zhang, A. Taherkordi, and T. Skeie, “Mobile edgecomputing: A survey,” IEEE Internet of Things Journal, vol. 5, no. 1,pp. 450–465, 2018.

[25] Smart city infographic. [Online, Accessed Nov. 10,2018]. [Online]. Available: https://www.postscapes.com/anatomy-of-a-smart-city/

[26] MARKETSANDMARKETS. Smart cities market. [On-line, Accessed Nov. 13, 2018]. [Online]. Avail-able: https://www.marketsandmarkets.com/Market-Reports/smart-cities-market-542.html

[27] G. view research. Market research report. [Online; Accessed 2019,April 1]. [Online]. Available: https://www.grandviewresearch.com/industry-analysis/edge-computing-market

[28] Q. Yuan, H. Zhou, J. Li, Z. Liu, F. Yang, and X. S. Shen, “Towardefficient content delivery for automated driving services: An edgecomputing solution,” IEEE Network, vol. 32, no. 1, pp. 80–86, 2018.

[29] L. Chen and C. Englund, “Every second counts: Integratingedge computing and service oriented architecture for automaticemergency management,” Journal of Advanced Transportation, vol.2018, 2018.

[30] M. Narang, S. Xiang, W. Liu, J. Gutierrez, L. Chiaraviglio,A. Sathiaseelan, and A. Merwaday, “Uav-assisted edge infras-tructure for challenged networks,” in IEEE Conference on Com-puter Communications Workshops, Atlanta, GA, USA, 2017, pp. 60–65.

[31] K. Bilal, O. Khalid, A. Erbad, and S. U. Khan, “Potentials, trends,and prospects in edge technologies: Fog, cloudlet, mobile edge,

and micro data centers,” Computer Networks, vol. 130, pp. 94–120,2018.

[32] U. Shaukat, E. Ahmed, Z. Anwar, and F. Xia, “Cloudlet deploy-ment in local wireless networks: Motivation, architectures, appli-cations, and open challenges,” Journal of Network and ComputerApplications, vol. 62, pp. 18–40, 2016.

[33] Y. Ai, M. Peng, and K. Zhang, “Edge computing technologies forinternet of things: a primer,” Digital Communications and Networks,vol. 4, no. 2, pp. 77–86, 2018.

[34] C. Mouradian, D. Naboulsi, S. Yangui, R. H. Glitho, M. J. Morrow,and P. A. Polakos, “A comprehensive survey on fog computing:State-of-the-art and research challenges,” IEEE CommunicationsSurveys Tutorials, vol. 20, no. 1, pp. 416–464, 2018.

[35] P. Hu, S. Dhelim, H. Ning, and T. Qiu, “Survey on fog computing:architecture, key technologies, applications and open issues,”Journal of Network and Computer Applications, vol. 98, pp. 27–42,2017.

[36] T. Taleb, K. Samdanis, B. Mada, H. Flinck, S. Dutta, andD. Sabella, “On multi-access edge computing: A survey of theemerging 5g network edge cloud architecture and orchestration,”IEEE Communications Surveys Tutorials, vol. 19, no. 3, pp. 1657–1681, 2017.

[37] R. Roman, J. Lopez, and M. Mambo, “Mobile edge computing,fog et al.: A survey and analysis of security threats and chal-lenges,” Future Generation Computer Systems, vol. 78, pp. 680–698,2018.

[38] Y. Mao, C. You, J. Zhang, K. Huang, and K. B. Letaief, “A surveyon mobile edge computing: The communication perspective,”IEEE Communications Surveys Tutorials, vol. 19, no. 4, pp. 2322–2358, 2017.

[39] W. Z. Khan, E. Ahmed, S. Hakak, I. Yaqoob, and A. Ahmed,“Edge computing: A survey,” Future Generation Computer Systems,vol. 97, pp. 219 – 235, 2019.

[40] N. Hassan, S. Gillani, E. Ahmed, I. Yaqoob, and M. Imran, “Therole of edge computing in internet of things,” IEEE Communica-tions Magazine, vol. 56, no. 11, pp. 110–115, 2018.

[41] W. Yu, F. Liang, X. He, W. G. Hatcher, C. Lu, J. Lin, and X. Yang,“A survey on the edge computing for the internet of things,”IEEE Access, vol. 6, pp. 6900–6919, 2018.

[42] A. Gharaibeh, M. A. Salahuddin, S. J. Hussini, A. Khreishah,I. Khalil, M. Guizani, and A. Al-Fuqaha, “Smart cities: A surveyon data management, security, and enabling technologies,” IEEECommunications Surveys Tutorials, vol. 19, no. 4, pp. 2456–2501,2017.

[43] C. Lim, K.-J. Kim, and P. P. Maglio, “Smart cities with big data:Reference models, challenges, and considerations,” Cities, vol. 82,pp. 86–89, 2018.

[44] A. H. Alavi, P. Jiao, W. G. Buttlar, and N. Lajnef, “Internet ofthings-enabled smart cities: State-of-the-art and future trends,”Measurement, vol. 129, pp. 589–606, 2018.

[45] B. N. Silva, M. Khan, and K. Han, “Towards sustainable smartcities: A review of trends, architectures, components, and openchallenges in smart cities,” Sustainable Cities and Society, vol. 38,pp. 697–713, 2018.

[46] I. Yaqoob, I. A. T. Hashem, A. Gani, S. Mokhtar, E. Ahmed, N. B.Anuar, and A. V. Vasilakos, “Big data: From beginning to future,”International Journal of Information Management, vol. 36, no. 6, pp.1231–1247, 2016.

[47] Dataversity. A brief history of cloud computing. [Online;Aceessed 2019, April 1]. [Online]. Available: https://www.dataversity.net/brief-history-cloud-computing

[48] Virtualization and cloud computing. [Online; Accessed 2019,April 5]. [Online]. Available: https://sites.google.com/a/lpu.co.in/cse423/home

[49] N. Fernando, S. W. Loke, and W. Rahayu, “Mobile cloud comput-ing: A survey,” Future generation computer systems, vol. 29, no. 1,pp. 84–106, 2013.

[50] M. Satyanarayanan, P. Bahl, R. Caceres, and N. Davies, “Thecase for vm-based cloudlets in mobile computing,” IEEE PervasiveComputing, vol. 8, no. 4, pp. 14–23, Oct 2009.

[51] M. Satyanarayanan, “The emergence of edge computing,” Com-puter, vol. 50, no. 1, pp. 30–39, 2017.

[52] F. Computing, “the internet of things: Extend the cloud to wherethe things are,” Cisco White Paper, 2015.

[53] S. Yi, C. Li, and Q. Li, “A survey of fog computing: concepts,applications and issues,” in Proceedings of the 2015 ACM workshopon mobile big data, Santa Clara, CA, USA, 2015, pp. 37–42.

https://www.postscapes.com/internet-of-things-definition/

https://www.postscapes.com/internet-of-things-definition/

https://www.postscapes.com/anatomy-of-a-smart-city/

https://www.postscapes.com/anatomy-of-a-smart-city/

https://www.marketsandmarkets.com/Market-Reports/smart-cities-market-542.html


https://www.grandviewresearch.com/industry-analysis/edge-computing-market

https://www.grandviewresearch.com/industry-analysis/edge-computing-market

https://www.dataversity.net/brief-history-cloud-computing

https://www.dataversity.net/brief-history-cloud-computing

https://sites.google.com/a/lpu.co.in/cse423/home

https://sites.google.com/a/lpu.co.in/cse423/home

27

[54] M. T. Beck, M. Werner, S. Feld, and S. Schimper, “Mobile edgecomputing: A taxonomy,” in Proc. of the Sixth Citeseer InternationalConference on Advances in Future Internet, Lisbon, Portugal, 2014,pp. 48–55.

[55] L. Gupta, R. Jain, and H. A. Chan, “Mobile edge computing-an important ingredient of 5g networks,” IEEE SoftwarizationNewsletter, 2016.

[56] A. Al-Shuwaili and O. Simeone, “Energy-efficient resource al-location for mobile edge computing-based augmented realityapplications,” IEEE Wireless Communications Letters, vol. 6, no. 3,pp. 398–401, 2017.

[57] X. Zhang, Y. Mao, J. Zhang, and K. B. Letaief, “Multi-objectiveresource allocation for mobile edge computing systems,” in 28thAnnual International Symposium on Personal, Indoor, and MobileRadio Communications, Montreal, QC, Canada, 2017, pp. 1–5.

[58] M. Bouet and V. Conan, “Mobile edge computing resourcesoptimization: A geo-clustering approach,” IEEE Transactions onNetwork and Service Management, vol. 15, no. 2, pp. 787–796, 2018.

[59] Q. Fan and N. Ansari, “Workload allocation in hierarchicalcloudlet networks,” IEEE Communications Letters, vol. 22, no. 4,pp. 820–823, 2018.

[60] N. K. Giang, V. Leung, and R. Lea, “On developing smarttransportation applications in fog computing paradigm,” in Pro-ceedings of the 6th ACM Symposium on Development and Analysisof Intelligent Vehicular Networks and Applications, Malta, 2016, pp.91–98.

[61] X. Huang, R. Yu, J. Kang, Y. He, and Y. Zhang, “Exploring mobileedge computing for 5g-enabled software defined vehicular net-works,” IEEE Wireless Communications, vol. 24, no. 6, pp. 55–63,2017.

[62] G. Premsankar, B. Ghaddar, M. Di Francesco, and R. Verago,“Efficient placement of edge computing devices for vehicularapplications in smart cities,” in IEEE/IFIP Network Operations andManagement Symposium, Taipei, Taiwan, 2018, pp. 1–9.

[63] T. M. Fernandez-Carames, P. Fraga-Lamas, M. Suarez-Albela,and M. Vilar-Montesinos, “A fog computing and cloudlet basedaugmented reality system for the industry 4.0 shipyard.” Sensors(Basel, Switzerland), vol. 18, no. 6, 2018.

[64] T. Taleb, S. Dutta, A. Ksentini, M. Iqbal, and H. Flinck, “Mobileedge computing potential in making cities smarter,” IEEE Com-munications Magazine, vol. 55, no. 3, pp. 38–43, 2017.

[65] M. Schneider, J. Rambach, and D. Stricker, “Augmented realitybased on edge computing using the example of remote livesupport,” in IEEE International Conference on Industrial Technology,Ontario, Canada, 2017, pp. 1277–1282.

[66] X. Qiao, P. Ren, S. Dustdar, and J. Chen, “A new era for web arwith mobile edge computing,” IEEE Internet Computing, vol. 22,no. 4, pp. 46–55, 2018.

[67] G. Muhammad, M. F. Alhamid, M. Alsulaiman, and B. Gupta,“Edge computing with cloud for voice disorder assessment andtreatment,” IEEE Communications Magazine, vol. 56, no. 4, pp. 60–65, 2018.

[68] A. M. Rahmani, T. N. Gia, B. Negash, A. Anzanpour, I. Azimi,M. Jiang, and P. Liljeberg, “Exploiting smart e-health gatewaysat the edge of healthcare internet-of-things: A fog computingapproach,” Future Generation Computer Systems, vol. 78, pp. 641–658, 2018.

[69] T. N. Gia, M. Jiang, V. K. Sarker, A. M. Rahmani, T. Westerlund,P. Liljeberg, and H. Tenhunen, “Low-cost fog-assisted health-careiot system with energy-efficient sensor nodes,” in 13th Interna-tional Wireless Communications and Mobile Computing Conference,Valencia, Spain, 2017, pp. 1765–1770.

[70] N. Kumar, S. Zeadally, and J. J. Rodrigues, “Vehicular delay-tolerant networks for smart grid data management using mobileedge computing,” IEEE Communications Magazine, vol. 54, no. 10,pp. 60–66, 2016.

[71] F. Y. Okay and S. Ozdemir, “A fog computing based smart gridmodel,” in IEEE international symposium on networks, computersand communications, Yasmine Hammamet, Tunisia, 2016, pp. 1–6.

[72] S. Zahoor, S. Javaid, N. Javaid, M. Ashraf, F. Ishmanov, andM. Afzal, “Cloud–fog–based smart grid model for efficient re-source management,” Sustainability, vol. 10, no. 6, p. 2079, 2018.

[73] F. J. Ferrandez-Pastor, J. M. Garcıa-Chamizo, M. Nieto-Hidalgo,and J. Mora-Martınez, “Precision agriculture design method us-ing a distributed computing architecture on internet of thingscontext,” Sensors, vol. 18, no. 6, p. 1731, 2018.

[74] M. Caria, J. Schudrowitz, A. Jukan, and N. Kemper, “Smartfarm computing systems for animal welfare monitoring,” in 40thIEEE International Convention on Information and CommunicationTechnology, Electronics and Microelectronics, Opatija, Croatia, 2017,pp. 152–157.

[75] M. A. Zamora-Izquierdo, J. Santa, J. A. Martınez, V. Martınez,and A. F. Skarmeta, “Smart farming iot platform based on edgeand cloud computing,” Biosystems Engineering, 2018.

[76] C. Vallati, A. Virdis, E. Mingozzi, and G. Stea, “Mobile-edgecomputing come home connecting things in future smart homesusing lte device-to-device communications,” IEEE Consumer Elec-tronics Magazine, vol. 5, no. 4, pp. 77–83, 2016.

[77] B. R. Stojkoska and K. Trivodaliev, “Enabling internet of thingsfor smart homes through fog computing,” in 25th IEEE Telecom-munication Forum, Belgrade, Serbia, 2017, pp. 1–4.

[78] G. Premsankar, M. D. Francesco, and T. Taleb, “Edge computingfor the internet of things: A case study,” IEEE Internet of ThingsJournal, vol. 5, no. 2, pp. 1275–1284, 2018.

[79] G. Jia, G. Han, H. Rao, and L. Shu, “Edge computing-basedintelligent manhole cover management system for smart cities,”IEEE Internet of Things Journal, vol. 5, no. 3, pp. 1648–1656, 2018.

[80] W. Li, T. Yang, F. C. Delicato, P. F. Pires, Z. Tari, S. U. Khan, andA. Y. Zomaya, “On enabling sustainable edge computing withrenewable energy resources,” IEEE Communications Magazine,vol. 56, no. 5, pp. 94–101, 2018.

[81] S. S. Gill and R. Buyya, “A taxonomy and future directions forsustainable cloud computing: 360 degree view,” ACM ComputingSurveys (CSUR), vol. 51, no. 5, p. 104, 2018.

[82] C. Huang, R. Lu, and K. R. Choo, “Vehicular fog computing:Architecture, use case, and security and forensic challenges,”IEEE Communications Magazine, vol. 55, no. 11, pp. 105–111, 2017.

[83] D. Kim, A. Alaerjan, L. Lu, H. Yang, and H. Jang, “Towardinteroperability of smart grids,” IEEE Communications Magazine,vol. 55, no. 8, pp. 204–210, 2017.

[84] S. Wolfert, L. Ge, C. Verdouw, and M.-J. Bogaardt, “Big data insmart farming–a review,” Agricultural Systems, vol. 153, pp. 69–80, 2017.

[85] K. P. Musaazi, T. Bulega, and S. M. Lubega, “Energy efficientdata caching in wireless sensor networks: A case of precisionagriculture,” in Springer International Conference on e-Infrastructureand e-Services for Developing Countries, KAMPALA, UGANDA,2014, pp. 154–163.

[86] A. H. H. Nograles and F. S. Caluyo, “Wireless system for preg-nancy detection in cows by monitoring temperature changes inbody,” in IEEE 9th International Colloquium on Signal Processingand its Applications, Kuala Lumpur, Malaysia, 2013, pp. 11–16.

[87] L. Kong, M. Xia, X. Liu, G. Chen, Y. Gu, M. Wu, and X. Liu,“Data loss and reconstruction in wireless sensor networks,” IEEETransactions on Parallel and Distributed Systems, vol. 25, no. 11, pp.2818–2828, 2014.

[88] A. Virdis, G. Stea, and G. Nardini, “Simulating lte/lte-advancednetworks with simulte,” in Simulation and Modeling Methodologies,Technologies and Applications. Springer, 2015, pp. 83–105.

[89] I. Yaqoob, I. A. T. Hashem, A. Ahmed, S. A. Kazmi, and C. S.Hong, “Internet of things forensics: Recent advances, taxonomy,requirements, and open challenges,” Future Generation ComputerSystems, vol. 92, pp. 265–275, 2019.

[90] S. Liszio and M. Masuch, “Designing shared virtual reality gam-ing experiences in local multi-platform games,” in InternationalConference on Entertainment Computing. Springer, 2016, pp. 235–240.

[91] Pokemongo. [Online; Accessed Sept. 20, 2018;. [Online].Available: https://www.pokemongo.com/en-us/

[92] E. Graells-Garrido, L. Ferres, D. Caro, and L. Bravo, “The effectof pokemon go on the pulse of the city: a natural experiment,”EPJ Data Science, vol. 6, no. 1, p. 23, 2017.

[93] M. Jia and W. Liang, “Delay-sensitive multiplayer augmentedreality game planning in mobile edge computing,” in Proceedingsof the 21st ACM International Conference on Modeling, Analysis andSimulation of Wireless and Mobile Systems, Montreal, QC, Canada,2018, pp. 147–154.

[94] M. Erol-Kantarci and S. Sukhmani, “Caching and computing atthe edge for mobile augmented reality and virtual reality (ar/vr)in 5g,” in Ad Hoc Networks. Springer, 2018, pp. 169–177.

[95] K. Dolui and S. K. Datta, “Comparison of edge computingimplementations: Fog computing, cloudlet and mobile edge

https://www.pokemongo.com/en-us/

28

computing,” in IEEE Global Internet of Things Summit, Geneva,Switzerland, 2017, pp. 1–6.

[96] S. Bi and Y. J. Zhang, “Computation rate maximization for wire-less powered mobile-edge computing with binary computationoffloading,” IEEE Transactions on Wireless Communications, vol. 17,no. 6, pp. 4177–4190, 2018.

[97] M. Chen and Y. Hao, “Task offloading for mobile edge computingin software defined ultra-dense network,” IEEE Journal on SelectedAreas in Communications, vol. 36, no. 3, pp. 587–597, 2018.

[98] H. Guo and J. Liu, “Collaborative computation offloading formultiaccess edge computing over fiber–wireless networks,” IEEETransactions on Vehicular Technology, vol. 67, no. 5, pp. 4514–4526,May 2018.

[99] S. Yu, R. Langar, X. Fu, L. Wang, and Z. Han, “Computationoffloading with data caching enhancement for mobile edge com-puting,” IEEE Transactions on Vehicular Technology, vol. 67, no. 11,pp. 11 098–11 112, 2018.

[100] J. Liu and Q. Zhang, “Offloading schemes in mobile edgecomputing for ultra-reliable low latency communications,” IeeeAccess, vol. 6, pp. 12 825–12 837, 2018.

[101] Y. Dong, C. Han, and S. Guo, “Joint optimization of energyand qoe with fairness in cooperative fog computing system,” in2018 IEEE International Conference on Networking, Architecture andStorage (NAS). IEEE, 2018, pp. 1–4.

[102] Z. Zhu, J. Peng, X. Gu, H. Li, K. Liu, Z. Zhou, and W. Liu,“Fair resource allocation for system throughput maximization inmobile edge computing,” IEEE Access, vol. 6, pp. 5332–5340, 2018.

[103] C.-H. Hong, K. Lee, M. Kang, and C. Yoo, “qcon: Qos-aware net-work resource management for fog computing,” Sensors, vol. 18,no. 10, p. 3444, 2018.

[104] Y. Peng, D.-K. Kang, F. Al-Hazemi, and C.-H. Youn, “Energyand qos aware resource allocation for heterogeneous sustainablecloud datacenters,” Optical Switching and Networking, vol. 23, pp.225–240, 2017.

[105] J. Xue, H. Shao, and Q. Ma, “Resource allocation for systemthroughput maximization based on mobile edge computing,” inProceedings of the IEEE International Conference on Electronics andElectrical Engineering Technology. Tianjin, China: ACM, 2018, pp.177–181.

[106] C. You, K. Huang, H. Chae, and B.-H. Kim, “Energy-efficientresource allocation for mobile-edge computation offloading,”IEEE Transactions on Wireless Communications, vol. 16, no. 3, pp.1397–1411, 2017.

[107] Y. Cui, W. He, C. Ni, C. Guo, and Z. Liu, “Energy-efficientresource allocation for cache-assisted mobile edge computing,”in IEEE 42nd Conference on Local Computer Networks, Singapore,2017, pp. 640–648.

[108] Y. Wang, M. Sheng, X. Wang, and J. Li, “Cooperative dynamicvoltage scaling and radio resource allocation for energy-efficientmultiuser mobile edge computing,” in IEEE International Confer-ence on Communications, Kansas City, MO, USA, 2018, pp. 1–6.

[109] K. Cheng, Y. Teng, W. Sun, A. Liu, and X. Wang, “Energy-efficient joint offloading and wireless resource allocation strategyin multi-mec server systems,” in IEEE International Conference onCommunications, Kansas City, MO, USA, 2018, pp. 1–6.

[110] Y. Chen, Y. Zhang, and X. Chen, “Dynamic service requestscheduling for mobile edge computing systems,” Wireless Com-munications and Mobile Computing, vol. 2018, 2018.

[111] F. Zafari, J. Li, K. K. Leung, D. Towsley, and A. Swami, “Agame-theoretic approach to multi-objective resource sharing andallocation in mobile edge,” in Proceedings of the ACM 2018 onTechnologies for the Wireless Edge Workshop, ser. WirelessEdge ’18,New York, NY, USA, 2018, pp. 9–13.

[112] A. B. Alam, M. Zulkernine, and A. Haque, “A reliability-based re-source allocation approach for cloud computing,” in IEEE 7th In-ternational Symposium on Cloud and Service Computing, Kanazawa,Japan, 2017, pp. 249–252.

[113] X. Chen, L. Jiao, W. Li, and X. Fu, “Efficient multi-user compu-tation offloading for mobile-edge cloud computing,” IEEE/ACMTransactions on Networking, vol. 24, no. 5, pp. 2795–2808, 2016.

[114] I. E. Commission. Edge intelligence. [Online; Accessed Oct. 20,2018]. [Online]. Available: https://basecamp.iec.ch/download/iec-white-paper-edge-intelligence-en/

[115] V. N. I. Cisco, “Global mobile data traffic forecast update, 2015–2020 white paper,” Document ID, vol. 958959758, 2016.

[116] X. Li and J. Wan, “Proactive caching for edge computing-enabledindustrial mobile wireless networks,” Future Generation ComputerSystems, vol. 89, pp. 89–97, 2018.

[117] K. Zhang, S. Leng, Y. He, S. Maharjan, and Y. Zhang, “Co-operative content caching in 5g networks with mobile edgecomputing,” IEEE Wireless Communications, vol. 25, no. 3, pp. 80–87, 2018.

[118] M. Li, F. R. Yu, P. Si, H. Yao, and Y. Zhang, “Software-defined ve-hicular networks with caching and computing for delay-tolerantdata traffic,” in IEEE International Conference on Communications,Kansas City, MO, USA, 2018, pp. 1–6.

[119] J. Chakareski, “Vr/ar immersive communication: Caching, edgecomputing, and transmission trade-offs,” in Proceedings of theACM Workshop on Virtual Reality and Augmented Reality Network,Los Angeles, CA, USA, 2017, pp. 36–41.

[120] A. Khreishah and J. Chakareski, “Collaborative caching formulticell-coordinated systems,” in IEEE Conference on ComputerCommunications Workshops, Hong Kong, China, 2015, pp. 257–262.

[121] J. Ji, K. Zhu, R. Wang, B. Chen, and C. Dai, “Energy efficientcaching in backhaul-aware cellular networks with dynamic con-tent popularity,” Wireless Communications and Mobile Computing,vol. 2018, 2018.

[122] F. Shen, K. Hamidouche, E. Bastug, and M. Debbah, “The busi-ness side of wireless edge caching: A game theoretical view.”

[123] F. De Pellegrini, A. Massaro, L. Goratti, and R. El-Azouzi, “Com-petitive caching of contents in 5g edge cloud networks,” in 15thIEEE International Symposium on Modeling and Optimization inMobile, Ad Hoc, and Wireless Networks, Paris, France, 2017, pp.1–8.

[124] K. Zhang, S. Leng, Y. He, S. Maharjan, and Y. Zhang, “Mobileedge computing and networking for green and low-latency in-ternet of things,” IEEE Communications Magazine, vol. 56, no. 5,pp. 39–45, 2018.

[125] J. Xu, L. Chen, and S. Ren, “Online learning for offloading andautoscaling in energy harvesting mobile edge computing,” IEEETransactions on Cognitive Communications and Networking, vol. 3,no. 3, pp. 361–373, 2017.

[126] Y. Mao, J. Zhang, and K. B. Letaief, “Dynamic computationoffloading for mobile-edge computing with energy harvestingdevices,” IEEE Journal on Selected Areas in Communications, vol. 34,no. 12, pp. 3590–3605, 2016.

[127] H. Zhang, Z. Chen, J. Wu, Y. Deng, Y. Xiao, K. Liu, and M. Li,“Energy-efficient online resource management and allocationoptimization in multi-user multi-task mobile-edge computingsystems with hybrid energy harvesting,” Sensors, vol. 18, no. 9, p.3140, 2018.

[128] X. Lyu, H. Tian, L. Jiang, A. Vinel, S. Maharjan, S. Gjessing, andY. Zhang, “Selective offloading in mobile edge computing for thegreen internet of things,” IEEE Network, vol. 32, no. 1, pp. 54–60,2018.

[129] P. Zhang, M. Zhou, and G. Fortino, “Security and trust issues infog computing: A survey,” Future Generation Computer Systems,vol. 88, pp. 16–27, 2018.

[130] S. Yi, Z. Qin, and Q. Li, “Security and privacy issues of fog com-puting: A survey,” in International conference on wireless algorithms,systems, and applications. Guilin, China: Springer, 2015, pp. 685–695.

[131] F. Manco, J. Martins, K. Yasukata, J. Mendes, S. Kuenzer, andF. Huici, “The case for the superfluid cloud.” in HotCloud, 2015.

[132] L. M. Vaquero and L. Rodero-Merino, “Finding your way inthe fog: Towards a comprehensive definition of fog computing,”ACM SIGCOMM Computer Communication Review, vol. 44, no. 5,pp. 27–32, 2014.

[133] T. K. Goyal and V. Sahula, “Lightweight security algorithmfor low power iot devices,” in IEEE International Conference onAdvances in Computing, Communications and Informatics, Jaipur,India, 2016, pp. 1725–1729.

[134] F. A. Alaba, M. Othman, I. A. T. Hashem, and F. Alotaibi,“Internet of things security: A survey,” Journal of Network andComputer Applications, vol. 88, pp. 10–28, 2017.

[135] C. Wang, J. Shen, Q. Liu, Y. Ren, and T. Li, “A novel securityscheme based on instant encrypted transmission for internet ofthings,” Security and Communication Networks, vol. 2018, 2018.

[136] N. Saxena, S. Grijalva, and N. S. Chaudhari, “Authenticationprotocol for an iot-enabled lte network,” ACM Transactions onInternet Technology (TOIT), vol. 16, no. 4, p. 25, 2016.

https://basecamp.iec.ch/download/iec-white-paper-edge-intelligence-en/

https://basecamp.iec.ch/download/iec-white-paper-edge-intelligence-en/

29

[137] M. N. Aman, K. C. Chua, and B. Sikdar, “A light-weight mutualauthentication protocol for iot systems,” in IEEE Global Commu-nications Conference, Singapore, 2017, pp. 1–6.

[138] B. L. Parne, S. Gupta, and N. S. Chaudhari, “Segb: Securityenhanced group based aka protocol for m2m communication inan iot enabled lte/lte-a network,” IEEE Access, vol. 6, pp. 3668–3684, 2018.

[139] E. Saurez, H. Gupta, R. Mayer, and U. Ramachandran, “Demo ab-stract: Fog computing for improving user application interactionand context awareness,” in IEEE/ACM Second International Con-ference on Internet-of-Things Design and Implementation, Pittsburgh,USA, 2017, pp. 281–282.

[140] H. Rahman, R. Rahmani, and T. Kanter, “Multi-modal context-aware reasoner (can) at the edge of iot,” Procedia Computer Science,vol. 109, pp. 335–342, 2017.

[141] N. B. Truong, G. M. Lee, and Y. Ghamri-Doudane, “Softwaredefined networking-based vehicular adhoc network with fogcomputing,” in IFIP/IEEE International Symposium on IntegratedNetwork Management, Ottawa, Canada, 2015, pp. 1202–1207.

[142] J. Cho, K. Sundaresan, R. Mahindra, J. Van der Merwe, andS. Rangarajan, “Acacia: context-aware edge computing for con-tinuous interactive applications over mobile networks,” in Pro-ceedings of the 12th ACM International on Conference on emergingNetworking EXperiments and Technologies, Irvine, California, USA,2016, pp. 375–389.

[143] C. You, K. Huang, and H. Chae, “Energy efficient mobile cloudcomputing powered by wireless energy transfer,” IEEE Journalon Selected Areas in Communications, vol. 34, no. 5, pp. 1757–1771,2016.

[144] W. Zhang, Y. Wen, and D. O. Wu, “Collaborative task executionin mobile cloud computing under a stochastic wireless channel,”IEEE Transactions on Wireless Communications, vol. 14, no. 1, pp.81–93, 2015.

[145] S. E. Mahmoodi, K. Subbalakshmi, and V. Sagar, “Cloud offload-ing for multi-radio enabled mobile devices,” in IEEE InternationalConference on Communications, London, UK, 2015, pp. 5473–5478.

[146] J. Kwak, Y. Kim, J. Lee, and S. Chong, “Dream: Dynamic resourceand task allocation for energy minimization in mobile cloud sys-tems,” IEEE Journal on Selected Areas in Communications, vol. 33,no. 12, pp. 2510–2523, 2015.

[147] J. Ren, G. Yu, Y. Cai, and Y. He, “Latency optimization for re-source allocation in mobile-edge computation offloading,” IEEETransactions on Wireless Communications, vol. 17, no. 8, pp. 5506–5519, 2018.

[148] X. Lyu, H. Tian, C. Sengul, and P. Zhang, “Multiuser joint taskoffloading and resource optimization in proximate clouds,” IEEETransactions on Vehicular Technology, vol. 66, no. 4, pp. 3435–3447,2017.

[149] C. Anglano, M. Canonico, and M. Guazzone, “Profit-aware re-source management for edge computing systems,” in Proceedingsof the 1st ACM International Workshop on Edge Systems, Analyticsand Networking, Munich, Germany, 2018, pp. 25–30.

[150] K. Tocze and S. Nadjm-Tehrani, “A taxonomy for managementand optimization of multiple resources in edge computing,”Wireless Communications and Mobile Computing, vol. 2018, 2018.

[151] S.-C. Lin, Y. Zhang, C.-H. Hsu, M. Skach, M. E. Haque, L. Tang,and J. Mars, “The architectural implications of autonomous driv-ing: Constraints and acceleration,” in Proceedings of the Twenty-Third ACM International Conference on Architectural Support forProgramming Languages and Operating Systems, Williamsburg, VA,USA, 2018, pp. 751–766.

[152] M. S. Elbamby, C. Perfecto, M. Bennis, and K. Doppler, “To-ward low-latency and ultra-reliable virtual reality,” IEEE Network,vol. 32, no. 2, pp. 78–84, 2018.

[153] Z. Luo, M. LiWang, Z. Lin, L. Huang, X. Du, and M. Guizani,“Energy-efficient caching for mobile edge computing in 5g net-works,” Applied Sciences, vol. 7, no. 6, p. 557, 2017.

[154] L. Chen, S. Zhou, and J. Xu, “Energy efficient mobile edgecomputing in dense cellular networks,” in IEEE InternationalConference on Communications, Porte Maillot, Paris, France, 2017,pp. 1–6.

[155] S. Kim, S. Park, M. Chen, and C. Youn, “An optimal pricingscheme for the energy-efficient mobile edge computation offload-ing with ofdma,” IEEE Communications Letters, vol. 22, no. 9, pp.1922–1925, 2018.

[156] Y. Dai, D. Xu, S. Maharjan, and Y. Zhang, “Joint computationoffloading and user association in multi-task mobile edge com-

puting,” IEEE Transactions on Vehicular Technology, vol. 67, no. 12,pp. 12 313–12 325, Dec 2018.

[157] M. N. H. Nguyen, N. H. Tran, M. A. Islam, C. Pham, S. Ren,and C. S. Hong, “Fair sharing of backup power supply in multi-operator wireless cellular towers,” IEEE Transactions on WirelessCommunications, vol. 17, no. 3, pp. 2080–2093, March 2018.

[158] F.-J. Ferrandez-Pastor, H. Mora, A. Jimeno-Morenilla, and B. Vol-ckaert, “Deployment of iot edge and fog computing technologiesto develop smart building services,” Sustainability, vol. 10, no. 11,p. 3832, 2018.

[159] G. Qiao, S. Leng, K. Zhang, and K. Yang, “Joint deploymentand mobility management of energy harvesting small cells inheterogeneous networks,” IEEE Access, vol. 5, pp. 183–196, 2017.

[160] N. A. Chughtai, M. Ali, S. Qaisar, M. Imran, M. Naeem, and F. Qa-mar, “Energy efficient resource allocation for energy harvestingaided h-cran,” IEEE Access, vol. 6, pp. 43 990–44 001, 2018.

[161] W. Lee, M. J. W. Schubert, B. Ooi, and S. J. Ho, “Multi-sourceenergy harvesting and storage for floating wireless sensor net-work nodes with long range communication capability,” IEEETransactions on Industry Applications, vol. 54, no. 3, pp. 2606–2615,May 2018.

[162] L. Jiang, H. Tian, Z. Xing, K. Wang, K. Zhang, S. Maharjan,S. Gjessing, and Y. Zhang, “Social-aware energy harvestingdevice-to-device communications in 5g networks,” IEEE WirelessCommunications, vol. 23, no. 4, pp. 20–27, August 2016.

[163] I. Froiz-Mıguez, T. Fernandez-Carames, P. Fraga-Lamas, andL. Castedo, “Design, implementation and practical evaluation ofan iot home automation system for fog computing applicationsbased on mqtt and zigbee-wifi sensor nodes,” Sensors, vol. 18,no. 8, p. 2660, 2018.

[164] M. Maso, “Energy harvesting oriented transceiver design for 5gnetworks,” in Energy Management in Wireless Cellular and Ad-hocNetworks. Springer, 2016, pp. 67–95.

[165] A. Zanni, S. Forsstrom, U. Jennehag, and P. Bellavista, “Elasticprovisioning of internet of things services using fog comput-ing: An experience report,” in 6th IEEE International Conferenceon Mobile Cloud Computing, Services, and Engineering, Bamberg,Germany, 2018, pp. 17–22.

[166] M. D. de Assuncao, A. da Silva Veith, and R. Buyya, “Distributeddata stream processing and edge computing: A survey on re-source elasticity and future directions,” Journal of Network andComputer Applications, vol. 103, pp. 1–17, 2018.

[167] J. Xu and S. Ren, “Online learning for offloading and au-toscaling in renewable-powered mobile edge computing,” inIEEE Global Communications Conference, Washington, D.C., UnitedStates, 2016, pp. 1–6.

[168] M. Ghobaei-Arani, S. Jabbehdari, and M. A. Pourmina, “An au-tonomic resource provisioning approach for service-based cloudapplications: A hybrid approach,” Future Generation ComputerSystems, vol. 78, pp. 191–210, 2018.

[169] S. Taherizadeh and V. Stankovski, “Auto-scaling applications inedge computing: Taxonomy and challenges,” in Proceedings of theInternational Conference on Big Data and Internet of Thing. ACM,2017, pp. 158–163.

[170] K. Wang, H. Yin, W. Quan, and G. Min, “Enabling collaborativeedge computing for software defined vehicular networks,” IEEENetwork, vol. 32, no. 5, pp. 112–117, Sep. 2018.

[171] H. Zhang, P. Dong, W. Quan, and B. Hu, “Promoting efficientcommunications for high-speed railway using smart collabora-tive networking,” IEEE wireless communications, vol. 22, no. 6, pp.92–97, 2015.

[172] M. Yannuzzi, F. van Lingen, A. Jain, O. L. Parellada, M. M. Flores,D. Carrera, J. L. Perez, D. Montero, P. Chacin, A. Corsaro et al., “Anew era for cities with fog computing,” IEEE Internet Computing,vol. 21, no. 2, pp. 54–67, 2017.

[173] M. Sellebraten. Fog computing in barcelona smartcity. [Online; Accessed Oct. 16, 2018]. [Online]. Avail-able: https://www.marketsandmarkets.com/Market-Reports/smart-cities-market-542.html

[174] smartme. [Online; Accessed Oct. 20, 2018]. [Online]. Available:http://smartme.unime.it/

[175] Watch. [Online; Accessed Oct. 20, 2018]. [Online]. Available:https://gtr.ukri.org/projects?ref=103845

[176] Smart city simulation. [Online; Accessed Jan. 4, 2019].[Online]. Available: https://iotify.help/network/smart-city/simulation.html



http://smartme.unime.it/

https://gtr.ukri.org/projects?ref=103845

https://iotify.help/network/smart-city/simulation.html

https://iotify.help/network/smart-city/simulation.html

30

[177] N. Hassan, S. Gillani, E. Ahmed, I. Yaqoob, and M. Imran, “Therole of edge computing in internet of things,” IEEE Communica-tions Magazine, vol. 56, no. 11, pp. 110–115, November 2018.

[178] L. Li, G. Zhao, and R. S. Blum, “A survey of caching techniquesin cellular networks: Research issues and challenges in contentplacement and delivery strategies,” IEEE Communications SurveysTutorials, vol. 20, no. 3, pp. 1710–1732, thirdquarter 2018.

[179] X. Sun and N. Ansari, “Dynamic resource caching in the iotapplication layer for smart cities,” IEEE Internet of Things Journal,vol. 5, no. 2, pp. 606–613, 2018.

[180] W. Yi, Y. Liu, and A. Nallanathan, “Cache-enabled hetnets withmillimeter wave small cells,” IEEE Transactions on Communica-tions, vol. 66, no. 11, pp. 5497–5511, 2018.

[181] S. Soleimani and X. Tao, “Cooperative crossing cache placementin cache-enabled device to device-aided cellular networks,” Ap-plied Sciences, vol. 8, no. 9, p. 1578, 2018.

[182] J. Yao and N. Ansari, “Joint caching in fronthaul and backhaulconstrained c-ran,” in GLOBECOM 2017 - 2017 IEEE Global Com-munications Conference, Dec 2017, pp. 1–6.

[183] Z. Chang, L. Lei, Z. Zhou, S. Mao, and T. Ristaniemi, “Learn tocache: Machine learning for network edge caching in the big dataera,” IEEE Wireless Communications, vol. 25, no. 3, pp. 28–35, 2018.

[184] J. Zhang, B. Chen, Y. Zhao, X. Cheng, and F. Hu, “Data securityand privacy-preserving in edge computing paradigm: Surveyand open issues,” IEEE Access, vol. 6, pp. 18 209–18 237, 2018.

[185] C. Wu, T. Yoshinaga, Y. Ji, T. Murase, and Y. Zhang, “A rein-forcement learning-based data storage scheme for vehicular adhoc networks,” IEEE Transactions on Vehicular Technology, vol. 66,no. 7, pp. 6336–6348, 2017.

[186] W. Liu, J. Zhang, Z. Liang, L. Peng, and J. Cai, “Content popu-larity prediction and caching for icn: A deep learning approachwith sdn,” IEEE Access, vol. 6, pp. 5075–5089, 2018.

[187] K. C. Tsai, L. Wang, and Z. Han, “Mobile social media networkscaching with convolutional neural network,” in Wireless Com-munications and Networking Conference Workshops (WCNCW), 2018IEEE. IEEE, 2018, pp. 83–88.

[188] K. Thar, N. H. Tran, T. Z. Oo, and C. S. Hong, “Deepmec:Mobile edge caching using deep learning,” IEEE Access, vol. 6,pp. 78 260–78 275, 2018.

[189] P. L. Ventre, C. Pisa, S. Salsano, G. Siracusano, F. Schmidt,P. Lungaroni, and N. Blefari-Melazzi, “Performance evaluationand tuning of virtual infrastructure managers for (micro) virtualnetwork functions,” in IEEE Conference on Network Function Virtu-alization and Software Defined Networks, Palo Alto, California, USA,2016, pp. 141–147.

[190] E. Ahmed and M. H. Rehmani, “Mobile edge computing: oppor-tunities, solutions, and challenges,” pp. 59–63, 2017.

[191] L. Chen and J. Xu, “Socially trusted collaborative edge computingin ultra dense networks,” in Proceedings of the Second ACM/IEEESymposium on Edge Computing, San Jose/Fremont, CA, USA, 2017,p. 9.

[192] N. Kato, Z. M. Fadlullah, B. Mao, F. Tang, O. Akashi, T. Inoue,and K. Mizutani, “The deep learning vision for heterogeneousnetwork traffic control: Proposal, challenges, and future perspec-tive,” IEEE wireless communications, vol. 24, no. 3, pp. 146–153,2017.

[193] Q. Fan and N. Ansari, “Towards workload balancing in fog com-puting empowered iot, in press,” IEEE Transactions on NetworkScience and Engineering, 2018.

[194] S. K. Mishra, B. Sahoo, and P. P. Parida, “Load balancing incloud computing: A big picture, in press,” Journal of King SaudUniversity-Computer and Information Sciences, 2018.

[195] P. Li and S. Guo, “Incentive mechanisms for device-to-devicecommunications,” IEEE Network, vol. 29, no. 4, pp. 75–79, 2015.

[196] R. Beraldi, A. Mtibaa, and H. Alnuweiri, “Cooperative loadbalancing scheme for edge computing resources,” in Second IEEEInternational Conference on Fog and Mobile Edge Computing, Valen-cia, Spain, 2017, pp. 94–100.

[197] D. Puthal, M. S. Obaidat, P. Nanda, M. Prasad, S. P. Mohanty, andA. Y. Zomaya, “Secure and sustainable load balancing of edgedata centers in fog computing,” IEEE Communications Magazine,vol. 56, no. 5, pp. 60–65, 2018.

[198] T. Sigwele, Y. F. Hu, M. Ali, J. Hou, M. Susanto, and H. Fitriawan,“An intelligent edge computing based semantic gateway forhealthcare systems interoperability and collaboration,” in IEEE6th International Conference on Future Internet of Things and Cloud(FiCloud), Barcelona, Spain, 2018, pp. 370–376.

[199] M. R. Raza, M. Fiorani, A. Rostami, P. Ohlen, L. Wosinska, andP. Monti, “Dynamic slicing approach for multi-tenant 5g trans-port networks,” Journal of Optical Communications and Networking,vol. 10, no. 1, pp. A77–A90, 2018.

[200] I. Afolabi, T. Taleb, K. Samdanis, A. Ksentini, and H. Flinck,“Network slicing and softwarization: A survey on principles, en-abling technologies, and solutions,” IEEE Communications SurveysTutorials, vol. 20, no. 3, pp. 2429–2453, 2018.

[201] A. U. Akgul, I. Malanchini, and A. Capone, “Dynamic resourcetrading in sliced mobile networks,” IEEE Transactions on Networkand Service Management, vol. 16, no. 1, pp. 220–233, March 2019.

[202] T. Guo and A. Suarez, “Enabling 5g ran slicing with edf slicescheduling,” IEEE Transactions on Vehicular Technology, vol. 68,no. 3, pp. 2865–2877, March 2019.

[203] A. Devlic, A. Hamidian, D. Liang, M. Eriksson, A. Consoli,and J. Lundstedt, “Nesmo: Network slicing management andorchestration framework,” in IEEE International Conference onCommunications Workshops, Cambridge, UK, 2017, pp. 1202–1208.

[204] T. M. Ho, N. H. Tran, S. M. A. Kazmi, Z. Han, and C. S. Hong,“Wireless network virtualization with non-orthogonal multipleaccess,” in IEEE/IFIP Network Operations and Management Sympo-sium, Taipei, Taiwan, 2018, pp. 1–9.

[205] S. A. Kazmi, N. H. Tran, T. M. Ho, and C. S. Hong, “Hierarchicalmatching game for service selection and resource purchasingin wireless network virtualization,” IEEE Communications Letters,vol. 22, no. 1, pp. 121–124, 2018.

[206] M. Jiang, M. Condoluci, and T. Mahmoodi, “Network slicing in5g: An auction-based model,” in IEEE International Conference onCommunications, Paris, France, 2017, pp. 1–6.

[207] C. Pham, N. H. Tran, S. Ren, W. Saad, and C. S. Hong, “Traffic-aware and energy-efficient vnf placement for service chaining:Joint sampling and matching approach, in press,” IEEE Transac-tions on Services Computing, 2017.

[208] O. Salman, I. Elhajji, A. Chehab, and A. Kayssi, “Iot survey: Ansdn and fog computing perspective,” Computer Networks, vol. 143,pp. 221–246, 2018.

[209] F. Wang, Y. Xu, L. Zhu, X. Du, and M. Guizani, “Lamanco:A lightweight anonymous mutual authentication scheme for n-times computing offloading in iot, in press,” IEEE Internet ofThings Journal, 2018.

[210] M. Sookhak, F. R. Yu, M. K. Khan, Y. Xiang, and R. Buyya,“Attribute-based data access control in mobile cloud computing:Taxonomy and open issues,” Future Generation Computer Systems,vol. 72, pp. 273–287, 2017.

[211] K. Xue, Y. Xue, J. Hong, W. Li, H. Yue, D. S. L. Wei, and P. Hong,“Raac: Robust and auditable access control with multiple at-tribute authorities for public cloud storage,” IEEE Transactionson Information Forensics and Security, vol. 12, no. 4, pp. 953–967,April 2017.

[212] C. Zuo, J. Shao, G. Wei, M. Xie, and M. Ji, “Cca-secure abe withoutsourced decryption for fog computing,” Future GenerationComputer Systems, vol. 78, pp. 730–738, 2018.

[213] R. Xu, Y. Chen, E. Blasch, and G. Chen, “A federated capability-based access control mechanism for internet of things (iots),” inSensors and Systems for Space Applications XI, vol. 10641. Interna-tional Society for Optics and Photonics, 2018, p. 106410U.

[214] I. Afolabi, T. Taleb, K. Samdanis, A. Ksentini, and H. Flinck,“Network slicing and softwarization: A survey on principles, en-abling technologies, and solutions,” IEEE Communications SurveysTutorials, vol. 20, no. 3, pp. 2429–2453, thirdquarter 2018.

[215] X. Foukas, G. Patounas, A. Elmokashfi, and M. K. Marina, “Net-work slicing in 5g: Survey and challenges,” IEEE CommunicationsMagazine, vol. 55, no. 5, pp. 94–100, May 2017.

[216] A. Kaloxylos, “A survey and an analysis of network slicing in 5gnetworks,” IEEE Communications Standards Magazine, vol. 2, no. 1,pp. 60–65, MARCH 2018.

[217] S. Zhang, W. Quan, J. Li, W. Shi, P. Yang, and X. Shen, “Air-ground integrated vehicular network slicing with content push-ing and caching,” IEEE Journal on Selected Areas in Communica-tions, vol. 36, no. 9, pp. 2114–2127, Sep. 2018.

[218] P. Rost, M. Breitbach, H. Roreger, B. Erman, C. Mannweiler,R. Miller, and I. Viering, “Customized industrial networks: Net-work slicing trial at hamburg seaport,” IEEE Wireless Communica-tions, vol. 25, no. 5, pp. 48–55, October 2018.

[219] S. Chaabnia and A. Meddeb, “Slicing aware qos/qoe in softwaredefined smart home network,” in IEEE/IFIP Network Operationsand Management Symposium, Taipei, Taiwan, 2018, pp. 1–5.

31

[220] M. Boussard, D. Thai Bui, R. Douville, P. Justen, N. Le Sauze,P. Peloso, F. Vandeputte, and V. Verdot, “Future spaces: Rein-venting the home network for better security and automation inthe iot era,” Sensors, vol. 18, no. 9, p. 2986, 2018.

[221] A. E. Kalør, R. Guillaume, J. J. Nielsen, A. Mueller, andP. Popovski, “Network slicing in industry 4.0 applications: Ab-straction methods and end-to-end analysis,” IEEE Transactions onIndustrial Informatics, vol. 14, no. 12, pp. 5419–5427, Dec 2018.

Latif U. Khan is currently pursuing his Ph.D.degree in Computer Engineering at Kyung HeeUniversity (KHU), South Korea. He is working asa leading researcher in the intelligent NetworkingLaboratory under a project jointly funded by theprestigious Brain Korea 21st Century Plus andMinistry of Science and ICT, South Korea. Hereceived his MS (Electrical Engineering) degreewith distinction from University of Engineeringand Technology (UET), Peshawar, Pakistan in2017. Prior to joining the KHU, he has served

as a faculty member and research associate in the UET, Peshawar,Pakistan. He has published his works in highly reputable conferencesand journals. He has received the best paper award in 15th IEEEInternational Conference on Advanced Communications Technology,PyeongChang, South Korea, in 2013. His research interests include an-alytical techniques of optimization and game theory to edge computingand end-to-end network slicing.

Ibrar Yaqoob (S’16-M’18-SM’19) is a researchprofessor with the Department of Computer Sci-ence and Engineering, Kyung Hee University,South Korea, where he completed his postdoc-toral fellowship under the prestigious grant ofBrain Korea 21st Century Plus. Prior to that,he received his Ph.D. (Computer Science) fromthe University of Malaya, Malaysia, in 2017. Heworked as a researcher and developer at theCentre for Mobile Cloud Computing Research(C4MCCR), University of Malaya. His numerous

research articles are very famous and among the most downloaded intop journals. He has reviewed over 200 times for the top ISI- Indexedjournals and conferences. He has been listed among top researchersby Thomson Reuters (Web of Science) based on the number of citationsearned in last three years in six categories of Computer Science. He iscurrently serving/served as a guest/associate editor in various Journals.He has been involved in a number of conferences and workshops in var-ious capacities. His research interests include big data, edge computing,mobile cloud computing, the Internet of Things, and computer networks.

Nguyen H. Tran (S’10-M’11-SM’18) is currentlyworking as a senior lecturer in the School ofComputer Science, The University of Sydney.He received his BS degree from Hochiminh CityUniversity of Technology and Ph.D. degree fromKyung Hee University, in electrical and computerengineering, in 2005 and 2011, respectively. Hewas an Assistant Professor with Department ofComputer Science and Engineering, Kyung HeeUniversity, Korea from 2012 to 2017. His re-search interest is to applying analytic techniques

of optimization, game theory, and stochastic modeling to cutting-edgeapplications such as cloud and mobile edge computing, data centers,heterogeneous wireless networks, and big data for networks. He re-ceived the best KHU thesis award in engineering in 2011 and best paperaward at IEEE ICC 2016. He has been the Editor of IEEE Transactionson Green Communications and Networking since 2016, and served asthe Editor of the 2017 Newsletter of Technical Committee on CognitiveNetworks on Internet of Things.

S. M. Ahsan Kazmi is with the Institute ofInformation Systems (IIS), Innopolis University,Innopolis, Tatarstan, Russia, where he is cur-rently working as an Assistant Professor. Hereceived his Master degree in CommunicationSystem Engineering from the National Universityof Sciences and Technology (NUST), Pakistan,in 2012, and the Ph.D. degree in Computer Sci-ence and Engineering from Kyung Hee Univer-sity (KHU), South Korea. His research interestsinclude applying analytical techniques of opti-

mization and game theory to radio resource management for future cel-lular networks. He received the best KHU Thesis Award in engineeringin 2017 and several best paper awards from prestigious conferences.

Tri Nguyen Dang is currently pursuing his Ph.D.in Computer Science and Engineering underfully funded scholarship at Kyung Hee University,South Korea. He received his B.S. in InformationTeaching from Hue University’s College of Ed-ucation, Vietnam in 2014. His professional ex-perience include Mobile Application Developer,middle-ware programming. He had won consola-tion prize of Olympiad in Informatics ACM ICPCVietnam in 2013. His research interest includesNetwork Optimization, and Mobile Cloud com-

puting, Mobile Edge Computing, Internet of Things.

Choong Seon Hong (S’95-M’97-SM’11) re-ceived the B.S. and M.S. degrees in electronicengineering from Kyung Hee University, Seoul,South Korea, in 1983 and 1985, respectively,and the Ph.D. degree from Keio University, Mi-nato, Japan, in 1997. In 1988, he joined Ko-rea Telecom, where he worked on broadbandnetworks as a Member of Technical Staff. InSeptember 1993, he joined Keio University. Heworked for the Telecommunications NetworkLaboratory, Korea Telecom, as a Senior Member

of Technical Staff and the Director of the Networking Research Teamuntil August 1999. Since September 1999, he has been a Professor withthe Department of Computer Science and Engineering, Kyung Hee Uni-versity. His research interests include future Internet, ad hoc networks,network management, and network security. He is a member of ACM,IEICE, IPSJ, KIISE, KICS, KIPS, and OSIA. He has served as the Gen-eral Chair, a TPC Chair/Member, or an Organizing Committee Memberfor international conferences such as NOMS, IM, APNOMS, E2EMON,CCNC, ADSN, ICPP, DIM, WISA, BcN, TINA, SAINT, and ICOIN. Inaddition, he is currently an Associate Editor of the IEEE Transactionson Network and Service Management, International Journal of NetworkManagement, and Journal of Communications and Networks and anAssociate Technical Editor of the IEEE Communications Magazine.

Documents

1 Edge Computing Enabled Smart Cities: A Comprehensive …san Kazmi, Tri Nguyen Dang, Choong Seon Hong, Senior Member, IEEE Abstract—Recent years have disclosed a remarkable proliferation