PaperlayerView exact revision on arXiv

Source-linked AI summary

Edge-Computing-Enabled Smart Cities: A Comprehensive Survey

Latif U. Khan, Ibrar Yaqoob, Nguyen H. Tran, S. M. Ahsan Kazmi, Tri Nguyen Dang, Choong Seon Hong

arXiv:1909.08747v2cs.NI

TL;DR

Smart cities generate growing volumes of data for compute-intensive applications that require strict latency-aware processing, while conventional cloud computing introduces delay. This survey synthesizes edge computing for smart cities by reviewing its evolution and applications, organizing a taxonomy, identifying requirements and synergies, and discussing open challenges and research directions.

  • Problem

    Smart city applications generate massive data and require instant analytics, but cloud computing’s inherent delay complicates real-time processing.

  • Method

    The survey reviews edge-computing evolution and smart-city applications, evaluates recent advances, develops a taxonomy, and examines requirements, synergies, and open challenges.

  • Results

    The survey organizes the literature across edge analytics, intelligence, resources, caching, resource management, characteristics, sustainability, and security, while presenting requirements and open research challenges.

  • Takeaways & Limitations

    Edge computing is presented as a way to provide smart-city applications with instantaneous computing and storage resources, while implementation still poses significant challenges.

Abstract

from arXiv · show

Recent years have disclosed a remarkable proliferation of compute-intensive applications in smart cities. Such applications continuously generate enormous amounts of data which demand strict latency-aware computational processing capabilities. Although edge computing is an appealing technology to compensate for stringent latency related issues, its deployment engenders new challenges. In this survey, we highlight the role of edge computing in realizing the vision of smart cities. First, we analyze the evolution of edge computing paradigms. Subsequently, we critically review the state-of-the-art literature focusing on edge computing applications in 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.

I. INTRODUCTION

Smart cities combine heterogeneous IoT technologies and services to improve citizens’ quality of life, but population growth and expanding device data create major computational demands. Edge computing addresses latency-sensitive applications by moving computation and storage toward the network edge.

  • Smart cities use ICT infrastructure and IoT-enabled services to support sustainable, secure, reliable, and higher-quality urban life.
  • Population growth and data from smartphones, GPS devices, sensors, and cameras increase smart cities’ computational and infrastructure complexity.
  • IoT-based smart city applications span smart grids, lighting, transportation, healthcare, homes, augmented reality, and agriculture.
  • Edge computing migrates computing power, data, and applications from remote clouds to the network edge, providing low-latency resources for real-time services.
  • Research activity in edge computing and smart cities has increased substantially, alongside urbanization and growth in smart city and edge computing markets.

B. Motivational Scenarios

The survey motivates edge computing through autonomous-vehicle emergency reporting and UAV forest-fire surveillance, while relating these scenarios to broader smart-city applications and prior survey literature. These examples emphasize rapid local processing where cloud-based handling can delay urgent decisions.

  • Autonomous Cars Accident Reporting: Autonomous-car accident reporting requires short emergency response times, making edge-enabled public safety systems relevant for rapid accident evaluation.
  • UAV Smart Forest Surveillance: UAVs can continuously capture forest images and use edge computing to process them locally for timely fire detection and reporting.
  • Smart Parking: Smart parking systems use surveillance cameras, image detection, and machine learning to compute empty parking spaces and improve parking management.
  • Smart Home: AI-enabled smart homes require instant computational resources at the edge to support automated operations using real-time and historical data.
  • Existing Surveys and Tutorials: Existing surveys separately examine smart cities, IoT, or edge-computing paradigms, leaving this survey to jointly organize their environments, taxonomy, requirements, and challenges.

2) Smart Cities Surveys:

The survey positions itself as the first work to jointly examine smart cities, IoT, and edge computing, extending prior surveys with a broader synthesis. It reviews advances, develops a taxonomy, identifies requirements, and discusses synergies and open challenges.

  • The survey claims to be the first to jointly consider smart cities, IoT, and edge computing.
  • The survey evaluates edge-computing-enabled smart-city advances and reports their organization across applications, taxonomies, requirements, synergies, and challenges.The paper structure includes application advances, taxonomy, core requirements, synergies and case studies, and open research challenges.
  • It categorizes the literature through a taxonomy covering characteristics, security, edge analytics, resources, edge intelligence, management, caching, and sustainability.
  • The paper presents requirements for designing smart-city architectures that use edge computing.
  • It discusses open research challenges, their causes, and implementation guidelines for edge computing in smart cities.

A. Metrics

The survey evaluates computing paradigms through six metrics and traces the progression from earlier centralized systems toward mobile cloud computing and edge-oriented approaches. It introduces mobile cloud computing as a response to mobile devices’ resource and energy constraints.

  • A. Metrics: Six metrics evaluate computing paradigms: scalability, flexibility, cost optimization, automation, latency, and mobility support.
  • A. Metrics: Scalability measures expansion under increasing user demand, while flexibility measures elastic provision of computing infrastructure.
  • A. Metrics: Cost optimization covers on-demand resource provision and hardware and software acquisition costs.
  • A. Metrics: Automation measures cloud updates without end-user intervention, while latency measures total smart-city application execution time.
  • A. Metrics: Mobility support measures the ability to enable seamless execution of IoT-based applications.
  • The evolution proceeds through mainframe, mini, client-server, desktop cloud, and mobile cloud computing before edge computing.
  • Mobile cloud computing uses remote cloud resources for mobile applications, addressing mobile devices’ disconnections, resource scarcity, and energy constraints.

4) Edge Computing:

Edge computing moves computation and storage toward the network edge through cloudlets, fog computing, and mobile edge computing. The survey reviews smart-city applications using assessment parameters spanning context, scalability, sustainability, caching, and security.

  • 4) Edge Computing:: Edge computing is a distributed architecture for real-time data processing and low-latency data-stream acceleration.
  • 4) Edge Computing:: Cloudlets place small-scale data centers near the network edge, while fog computing uses edge devices to provide nearby computation services.
  • 4) Edge Computing:: Mobile edge computing places computing resources at radio access and core networks, with servers deployed at macrocell, grouped, or hierarchical base-station positions.
  • The literature review evaluates recent advances using context-awareness, scalability, sustainability, caching, and security.
  • Context-awareness uses node-location and environmental information to support meaningful IoT data, machine-to-machine communication, emergency management, and automatic services.
  • Sustainability emphasizes energy-efficient design, renewable sources, and energy harvesting to reduce carbon emissions.
  • Caching stores popular content near users to reduce access latency and network congestion, complementing edge computing’s provision of computation resources.
  • The survey records parameter dimensions and uses check marks when at least one dimension is considered, with tables categorizing advances by application.

1) Smart Transportation Applications:

Edge computing architectures support smart-city transportation and augmented-reality applications, but mobility, security, latency, and resource constraints remain central concerns. The reviewed work highlights service migration, socially-aware device-to-device communication, and lightweight security as important directions.

  • Smart Transportation Applications: A 5G-enabled software-defined vehicular network integrates software-defined networking with mobile edge computing across data, social, and control planes.The architecture discovers Vehicular Neighbor Groups using real data and supports flexible data sharing.
  • Smart Transportation Applications: Vehicular edge systems must address seamless connectivity, active service migration, and security as vehicles move between roadside units.Maintaining service continuity during roadside-unit changes is particularly important for moving vehicles and content delivery.
  • Smart Transportation Applications: Roadside-unit deployment should jointly minimize transmission power, the number of roadside units, and interference to improve quality of service.
  • Smart Transportation Applications: Edge-enabled augmented reality spans industrial tasks, smart tourism, remote live support, and web-based applications through layered edge–cloud architectures.These applications process video or interactive content near users, while resource allocation and transmission delay remain practical considerations.
  • Smart Transportation Applications: Socially-aware device-to-device communication can reduce augmented-reality content latency and request overhead at the base station.Users first check nearby members of their social group for requested content before contacting the base station.
  • Smart Transportation Applications: Lightweight security mechanisms are needed for industrial augmented reality because unauthorized access to remote live support can disrupt operations.

C. Smart Health-care

Edge computing supports healthcare architectures that collect, process, and analyze biomedical data across device, edge, and cloud layers. The survey emphasizes federated learning for privacy-preserving intelligent healthcare, while reviewed systems also expose dataset and infrastructure limitations.

  • C. Smart Health-care: Smart healthcare uses edge computing to provide low-cost, effective, and ubiquitous real-time healthcare facilities.
  • C. Smart Health-care: A voice-disorder framework collects voice samples with smart sensors, processes them on edge nodes, and sends results through the cloud to specialists.The framework uses the Saarbruken Voice Disorder database for training, testing, and validation.
  • C. Smart Health-care: The voice-disorder framework is limited by using only a few disorders with sufficient samples in its database.
  • C. Smart Health-care: Fog-based healthcare architectures distribute sensing, local processing, filtering, compression, fusion, analysis, storage, and interoperability across device, edge, and cloud layers.
  • C. Smart Health-care: Fog-assisted gateways enable continuous remote ECG monitoring by receiving biosignals and contextual information from sensor nodes before forwarding processed data to the cloud.
  • C. Smart Health-care: Federated learning is recommended for intelligent edge healthcare because on-device distributed learning preserves user privacy without migrating all data to a centralized location.The survey states that federated learning at the edge requires on-demand computational resources.

2) Cloud–Fog–Based Smart Grid Model:

The surveyed smart-city applications use layered cloud, fog, edge, and IoT architectures to process data near users and support responsive services. The section highlights productivity, reliability, and security requirements across farming, health care, smart grids, homes, and buildings.

  • Cloud–Fog–Based Smart Grid Model: Cloud-fog smart-grid models place end users or smart-grid devices below a fog layer that intermediates with the cloud.Fog components collect or process information near consumers before cloud-level services.
  • Smart Grids: Distributed edge architectures require strong security and privacy protections because their geographically dispersed components are more vulnerable than centralized cloud systems.The section identifies blockchain for edge energy trading but notes significant implementation challenges.
  • Smart Farming: Edge nodes in smart farming provide interoperability, storage, analysis, and prediction near production facilities to satisfy low-latency control requirements.The architecture includes things, edge, fog, and communication layers.
  • Smart Farming: Intelligent edge caching keeps frequently requested instructions near sensors, reducing remote-cloud transmissions and improving precision-agriculture throughput.The cache analyzes sensor requests at the edge before forwarding unresolved requests to the cloud.
  • Smart Buildings: Smart-home edge designs vary server placement across eNBs, femtocells, and user equipment, with simulations covering D2I and D2D communication.Placing servers nearer UEs requires less capacity because the coverage area is smaller.

2) Sustainable Smart Homes:

This section connects edge-enabled smart-city services with interoperability, security, context-aware analytics, robust architecture, and trade-offs between reliability and cost. It also frames edge analytics as a way to reduce delay for real-time applications.

  • Sustainable Smart Homes: Smart-home systems require interoperability so heterogeneous IoT devices can interact seamlessly with one another and edge servers.The section also calls for forensic tools because devices and edge servers are susceptible to attacks.
  • Smart Management: Edge-enabled smart management can act as a platform for compute-intensive algorithms, while context-aware machine learning accounts for differences among cities.Sensor-generated data motivates machine-learning techniques for smarter management.
  • Computing Paradigms: Edge computing paradigms share cloud-to-edge resource migration but differ in context-awareness, inter-node communication, and resource placement.Mobile edge computing is described as having the highest context-awareness among the listed formats.
  • Characteristics: Robust architectures integrate edge servers, IoT devices, remote clouds, and communication infrastructure using fail-over and redundancy techniques.Backup servers can replace fault-affected edge servers, but this increases system cost.
  • Characteristics: Critical applications require a trade-off between robustness and cost because redundancy improves continuity while increasing system expense.Examples include autonomous driving, delivery UAV control, and smart industrial manufacturing.
  • Edge Analytics: Edge analytics analyzes data near the network edge and returns results to devices, reducing delay relative to sending data to a centralized cloud.The surveyed applications include augmented reality, health care, industrial control, transportation, and surveillance.

C. Resources

The resources section identifies computation, communication, and storage as the core resources of edge-enabled smart cities and surveys mechanisms for managing them. It also links resource use with replication, intelligence, caching, sustainability, and security.

  • Resources: Edge computing uses computation, communication, and storage resources, with applications offloading complete or partial components to edge servers.Computation resources may reside on local devices or servers.
  • Resources: Resource management must coordinate massive smart-device populations, edge servers, and communication infrastructure using criteria such as fairness, QoS, throughput, and energy.Resource discovery, management tools, and management criteria are identified as main aspects.
  • Resource Management Strategies: Computation replication sends copies of tasks or subtasks to multiple servers, improving robustness and download time through result diversity.The approach increases upload time because uplink bandwidth is generally limited.
  • Resource Management Strategies: Mobility-management schemes are needed to maintain seamless connectivity when IoT nodes move across smart environments.The section specifically associates mobility with applications such as smart transportation.
  • Edge Intelligence: Edge intelligence combines advanced networking and machine learning with virtualized, dynamic micro-data-center capabilities.The section identifies key enablers, hardware requirements, and software requirements as its fundamental aspects.
  • Caching: Caching temporarily stores contents near the network edge to avoid repeated transmissions and support resource-intensive smart-city applications.Its main aspects are cache location, cache contents, caching control, and mathematical tools.
  • Sustainability: Sustainability addresses energy limitations from device and server densification through renewable energy, energy harvesting, and energy-efficient design.The stated objective is reducing overall carbon footprint.
  • Security: Distributed edge infrastructure creates security risks involving denial of service, man-in-the-middle attacks, physical damage, and privacy leakage.The survey calls for lightweight encryption and effective authentication protocols.

A. Scalability and Reliability

The requirements section treats scalability and reliability as joint infrastructure challenges under growing device populations, failures, and traffic variation. It also identifies context-awareness, resource optimization, sustainability, elasticity, and network slicing as related design priorities.

  • A. Scalability and Reliability: Smart-city infrastructure must handle increasing end-user devices and system failures while preserving scalable and reliable performance.Proposed responses include scalable service points and resilience to hardware, software, and network failures.
  • A. Scalability and Reliability: Collaborative edge computing is proposed to address traffic variation and limited edge-server capacity while supporting scalability and reliability.The cited work considers collaboration among edge servers as a response to changing smart-device traffic.
  • B. Context-awareness: Context-aware edge servers use device locations, network load, and capacity information to support strict-latency computation and timely emergency responses.Roadside units can report transportation emergencies even when an injured user cannot place a call.
  • C. Resource Management: Resource-management schemes must jointly consider energy, latency, traffic-handling capacity, and user-defined utility, although simultaneous optimization is difficult.The section distinguishes computing, communication, and storage resources.
  • C. Resource Management: AI-based schemes, including deep reinforcement learning and federated learning, are recommended for managing communication, caching, and computation at the edge.The motivation includes the large number of configurable parameters in a typical 5G node.
  • D. Sustainability: Sustainability requires reducing high energy consumption without degrading QoS as smart devices and edge servers become denser.The section presents sustainability as a core infrastructure requirement.
  • E. Elasticity: Elasticity avoids both over-allocation and under-allocation of computing resources as sensor and smart-device populations increase.The requirement targets efficient utilization of constrained computing, communication, and storage resources.
  • Business Models: Network slicing is identified as an open challenge for coordinating multiple service providers while jointly optimizing user QoS and provider profit.The setting includes edge, cloud, and telecommunications infrastructure providers.

F. Interoperability

Edge-enabled smart cities must coordinate heterogeneous devices, protect data, and align distributed services with provider revenue and user QoS. Reported projects illustrate these requirements across practical deployments.

  • Interoperability: Interoperability is essential because edge infrastructures combine massive numbers of heterogeneous devices running different protocols.
  • Privacy and Security: Blockchain supports decentralized, transparent, secure, and immutable data sharing, but its transaction scalability remains a barrier.Edge computing can provide on-demand computation and storage, although their integration introduces further challenges.
  • Business Model: Existing cloud business models are insufficient for distributed edge architectures, requiring new pricing models that increase revenue while maintaining QoS.
  • Case Studies: Barcelona’s project targets power monitoring, access control, event-based video, traffic management, and on-demand connectivity.
  • Case Studies: #SmartME combines application, Stack4Things fog-platform, and city layers while enabling citizens to contribute infrastructure through hardware sharing.

D. Lessons Learned: Summary and Insights

The survey identifies simulation, caching, collaboration, load balancing, and intelligent edge processing as important directions for managing smart-city scale and complexity. These directions address congestion, constrained resources, mobility, dynamic loads, noisy data, and interoperability.

  • Lessons Learned: Summary and Insights: High-performance simulation tools are needed to implement and validate edge-enabled smart-city applications.IoTIFY is cited as an online simulator for smart waste management, parking, street lights, traffic signals, and transportation.
  • Intelligent Caching: Caching can mitigate backhaul congestion by storing popular content at multiple network locations and avoiding repeated transmissions.
  • Intelligent Caching: Intelligent caching remains challenged by cold-start users, security and privacy, user mobility, and difficult content-popularity prediction.
  • Collaborative Edge Computing: Collaborative edge computing addresses constrained servers and growing device numbers by maximizing resource usage, minimizing latency, and reducing backhaul traffic.
  • Collaborative Edge Computing: Collaborative operation requires adaptive collaboration spaces, incentive and cooperation policies, inter-domain and intra-domain coordination, networking, and mobility management.
  • Cooperative Load Balancing: Dynamic and spatially varying device traffic can overload edge access points or base stations, causing higher latency, QoS degradation, and provider profit loss.
  • Intelligent Edge Computing: Intelligent edge computing can filter noisy device data before cloud transmission and improve security by avoiding remote-cloud predictive analytics.
  • Intelligent Edge Computing: AI-enabled edge systems require low-complexity learning because edge computational power is lower than remote-cloud capacity, alongside solutions for data, privacy, and interoperability challenges.

E. Network Slicing

Network slicing is presented as a way to coordinate diverse smart-city services and stakeholders over shared infrastructure while addressing management, security, orchestration, and resource-allocation challenges.

  • Network Slicing: Realizing network slicing requires end-to-end slice management and orchestration, slice security, and adaptive service function chaining.
  • Network Slicing: Game theory, deep reinforcement learning, and auction theory are proposed for network-slice management and orchestration.
  • Security: Network slicing faces security risks from multiple authorities, operator resource sharing, and attacks on SDN-based orchestrators.
  • Motivation: Edge computing supports real-time smart-city analytics by moving computation and storage toward the network edge to reduce cloud-induced delay.
  • Network Slicing: Network slicing creates multiple logical networks over shared physical infrastructure to support diverse smart-city services and stakeholder goals.
  • Resource Management: Coexisting slices improve infrastructure cost efficiency but make inter-slice resource management challenging.
Loading 1909.08747v2…