PreprintPDF Available

Distributed Vehicular Computing at the Dawn of 5G: a Survey

Preprints and early-stage research may not have been peer reviewed yet.
Preprint

Distributed Vehicular Computing at the Dawn of 5G: a Survey

Abstract and Figures

Recent advances in information technology have revolutionized the automotive industry, paving the way for next-generation smart vehicular mobility. Vehicles, roadside units, and other road users can collaborate to deliver novel services and applications. These services and applications require 1) massive volumes of heterogeneous and continuous data to perceive the environment, 2) reliable and low-latency communication networks, 3) real-time data processing that provides decision support under application-specific constraints. Addressing such constraints introduces significant challenges for current communication and computing technologies. Relatedly, the fifth generation of cellular networks (5G) was developed to respond to communication challenges by providing for low-latency, high-reliability, and high bandwidth communications. As a major part of 5G, edge computing allows data offloading and computation at the edge of the network, ensuring low-latency and context-awareness, and 5G efficiency. In this work, we aim at providing a comprehensive overview of the state of research on vehicular computing in the emerging age of 5G and big data.
Content may be subject to copyright.
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2020.DOI
Distributed Vehicular Computing at the
Dawn of 5G: A Survey
AHMAD ALHILAL1, (Member, IEEE), BENJAMIN FINLEY2, TRISTAN BRAUD3, (Member,
IEEE), DONGZHE SU4, and PAN HUI2,5,6 , (Fellow, IEEE).
1Department of Computer Science Engineering, The Hong Kong University of Science and Technology, Hong Kong
2Department of Computer Science, University of Helsinki, Helsinki, Finland
3Division of Integrative Systems and Design, The Hong Kong University of Science and Technology, Hong Kong
4The Hong Kong Applied Science and Technology Research Institute (ASTRI), Hong Kong
5Computational Media and Arts Thrust Area, The Hong Kong University of Science and Technology (Guangzhou), China
6Division of Emerging Interdisciplinary Area, The Hong Kong University of Science and Technology, Hong Kong
Corresponding author: Benjamin Finley (e-mail: benjamin.finley@helsinki.fi).
ABSTRACT Recent advances in information technology have revolutionized the automotive industry,
paving the way for next-generation smart vehicular mobility. Vehicles, roadside units, and other road
users can collaborate to deliver novel services and applications. These services and applications require
1) massive volumes of heterogeneous and continuous data to perceive the environment, 2) reliable and
low-latency communication networks, 3) real-time data processing that provides decision support under
application-specific constraints. Addressing such constraints introduces significant challenges for current
communication and computing technologies. Relatedly, the fifth generation of cellular networks (5G) was
developed to respond to communication challenges by providing for low-latency, high-reliability, and high
bandwidth communications. As a major part of 5G, edge computing allows data offloading and computation
at the edge of the network, ensuring low-latency and context-awareness, and 5G efficiency. In this work, we
aim at providing a comprehensive overview of the state of research on vehicular computing in the emerging
age of 5G and big data.
INDEX TERMS Edge Computing, Cloud Computing, Intelligent Transportation System, Big Data, 5G,
Distributed Computing, Vehicular Networks
I. INTRODUCTION
The automotive industry is on the verge of one of the most
dramatic paradigm shifts in its history. An increasing number
of vehicles embed sensing, computation, and wireless com-
munication capabilities. Such vehicles feature onboard units
(OBU), global positioning system (GPS) units, onboard radio
modules, such as IEEE 802.11p, long-term evolution (LTE),
or 5G modules, and other onboard units. These units perceive
the surrounding environment and perform computation and
communication. Similar to vehicles, the road infrastructure
itself is also embedding more intelligence. Induction loop
detectors can detect vehicles passing or arriving at a certain
location such as approaching a traffic light or in motorway
traffic. Specifically, the pavement is equipped with an insu-
lated, electrically conducting loop which detects the presence
of vehicles and can be connected to roadside units (RSUs).
Roadside units (RSUs) are transceivers mounted along a
road or pedestrian passageway to interact with vehicles and
perform computation, communication, and storage tasks.
These capabilities enable the vehicles and the infrastructure
to form a vehicular ad-hoc network (VANET) spontaneously
and without any additional infrastructure [1]. However, due
to the mobility of the vehicles and dynamic nature of traffic,
links change frequently leading to ever-changing topology
and network partitioning. Resultingly, sparse and heavy traf-
fic frequently alternate leading to intermittent connectivity
and network congestion. These dramatic changes introduce
high latency variability which impacts the quality of service.
These conditions complicate the deployment of vehicular
applications that require real-time interactions and prevent
the deployment of time-critical safety applications.
The development of connected vehicular systems paves
the way for new services and business opportunities. The
deployment of the technologies, infrastructure, and services
relies on an interdisciplinary effort involving not only man-
ufacturers, but also network operators, service providers,
and governmental authorities. Network operators provide
network access, whereas service providers provide access to
VOLUME 4, 2021 1
Distributed Vehicular Computing at the Dawn of 5G
Remote Cloud
Macro
Base Station
Base Station
Edge
Server
RSU
Internet
Data Center
Figure 1: Overview of on-board units [2] and computation
and communication resources for two vehicular applications.
The applications leverage vehicular edge computing (VEC)
and vehicular cloud computing (VCC) for computation re-
sources, mobile network for internet access, and C-V2X
direct communication for critical safety communication.
specific services and bill subscribed users [3]. For example,
service providers can collect real-time traffic data, detect
traffic congestion, and disseminate such information to ve-
hicles, either through RSUs or cellular communication [4].
Finally, government authorities play a critical role in aligning
all actors towards providing safe, reliable, and interoperable
road services. Such a collaboration opens up numerous possi-
bilities for potential life-changing applications in the area of
Intelligent Transportation Systems (ITS). These applications
range from critical safety-related applications and driver
assistance systems to location-based and traffic management
services. The computation and communication resources in-
volved in vehicular applications vary based on the application
requirements, as shown in Figure 1. At one end of the spec-
trum, infotainment applications access the internet through
mobile networks (LTE and 5G) thus allowing multimedia
streaming or web browsing. On the other end, safety-critical
applications rely primarily on vehicles, clouds, and edge
servers for decision-making. These applications also tend to
exploit reliable and lower-latency communication solutions,
for example, cellular vehicle-to-everything (C-V2X). C-V2X
enables vehicle to vehicle (V2V), vehicle to infrastructure
(V2I), and other communication variants to address the real-
time constraints.
Furthermore, road elements (traffic lights, lampposts, in-
duction loop detectors, RSUs) and road users (vehicles,
pedestrians’ smartphones) produce traffic data resulting in
significant data heterogeneity and volume due to sensor
ubiquity and diversity. Thus processing heterogeneous and
big data streams becomes another challenge. Additionally,
as previously stated, the connection may be intermittent,
resulting in massive data bursts that need to be processed in
real-time. Analyzing this data and promptly extracting mean-
ingful and useful information requires the consideration of
specific big data architectures in the deployment of connected
vehicular systems [5], [6].
Cloud computing (CC) enables ubiquitous, convenient,
and on-demand access to a shared pool of configurable
computing resources (e.g., networks, servers, storage, ap-
plications, and services) [7]. However, many applications
have strict latency requirements, which makes edge comput-
ing (EC) a better candidate compared to remote centralized
clouds. EC promises to deliver scalable, highly responsive
cloud services for mobile computing and masks transient
cloud outages. In contrast to CC, EC features proximity
to the subscribing vehicles, context-awareness, dense geo-
graphical distribution, and support for mobility [8], [9]. For
instance, EC-assisted traffic management enables monitoring
the lane occupancy using traffic data and changing traffic
signal phases accordingly.
There is a growing volume of research related to vehicular
communication and computing. However, as far as we are
aware, the domain lacks a systematic survey that studies
the existing literature on distributed vehicular computing,
pinpoints the applications and requirements, highlights the
methodologies, and determines the enabling technologies.
Therefore in this work, we provide a survey that fulfills these
comprehensive goals. We first focus on the communication
methods and evaluate how current and future technologies
can assist vehicular networking in handling significant net-
work load while keeping network latency at a minimum. We
then move on to potential data offloading and computing
architectures, ranging from road to city scale. As a promising
paradigm, we discuss edge computing methods including
fog computing (using RSUs) and multiaccess edge comput-
ing (MEC), a standardized ETSI architecture. Afterwards,
we review different data analytics algorithms to support
decision-making based on large volumes of heterogeneous
data. Finally, we discuss insights and open issues, among
them future communication, background services, and safety
applications for further research and development of novel
applications.
The contributions of this survey are threefold:
1) Challenges and Requirements. We characterize the
vehicular environment including the challenges and re-
quirements of real-world vehicular applications.
2) Bottom-up Overview. We provide a comprehensive
study of existing communication technologies and com-
putation paradigms. We also examine promising com-
munication and computing technologies before further
investigating big data analytics frameworks and ML-
empowered vehicular applications.
3) Integrated Architecture and Case Studies. We study
the requirements and potential architectures of ITS sys-
tems at local, neighborhood, and city scales, and show-
case real-world scenarios.
4) Future direction. We discuss insights and open issues,
which shed light on the development of future novel
vehicular applications and services.
2VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
Table 1: List of acronyms
Acronym Definition
V2X Vehicle-to-Everything
V2I Vehicle-to-Infrastructure
V2V Vehicle-to-Vehicle
C-V2X Cellular (LTE)-based Vehicle-to-Everything
NR C-V2X Cellular (5G)-based Vehicle-to-Everything
ITS Intelligent Transportation System
OBU On-Board Unit
RSU Road Side Unit
CVs Connected Vehicles
AV/D Autonomous Vehicle/Driving
ADAS Advanced Driver Assistance Systems
DSRC Dedicated Short Range Communication
WAVE Wireless Access in Vehicular Environment
CAM Cooperative Awareness Message
DENM Decentralized Environmental Notification Message
WSMP WAVE Short Message Protocol
3GPP The 3rd Generation Partnership Project
LTE Long-Term Evolution
BM-SC Broadcast Multicast Service Center
eMBMS evolved Multimedia Broadcast Multicast Service
EPC Evolved Packet Core
RAN Radio Access Network
RAT Radio Access Technology
UE User Equipment
PDN Packet Data Network
P-GW PDN Gateway
S-GW Serving Gateway
MME Mobility Management Entity
MBSFN Multimedia Broadcast Multicast Service Single Fre-
quency Network
5G Fifth-Generation Mobile Network
MIMO Multiple Input Multiple Output
SDN Software Defined Networking
NFV Network Function Virtualization
uRLLC ultra Reliable Low Latency Communication
eMBB evolved Mobile Broad Band
mMTC massive Machine Type Communication
AMF Access and Mobility Management Function
SMF Session Management Function
UPF User Plan Function
KPIs 5G Key Performance Indicators
CC Cloud Computing
EC Edge Computing
MEC Multiaccess Edge Computing
FC Fog Computing
II. CHALLENGES OF VEHICULAR APPLICATIONS
The deployment of vehicular applications will face multiple
challenges relating to storage, computing capabilities, net-
work, privacy, and security.
Data Volume, Variety, and Velocity (3Vs). Connected
vehicles contain a wide variety of sensors that continuously
produce large amounts of data. RGB cameras alone generate
20 to 40 Mbps and radar sensors produce between 10 and
100 Kbps. The move from simply connected vehicles to
autonomous vehicles will likely further increase sensor data
volume (as autonomous vehicles will have more sensors). An
autonomous vehicle may include 4-6 radar sensors generat-
ing 0.1 - 15 Mbps per sensor, 1-5 lidar sensors generating
20-100 Mbps per sensor, 6-12 RGB camera sensors gener-
ating 500 - 3500 Mbps per sensor, 8-16 ultrasonic sensors
generating less than 0.01 Mbps per sensor, and vehicle mo-
tion, global navigation satellite system (GNSS) and inertial
measurement unit (IMU) with <0.1 Mbps per sensor. The
total sensor bandwidth is thus between 3 Gbps (10.8 Tb/h)
and 40 Gbps (144 Tb/h) [10]. Therefore, disseminating such
massive data to remote servers or processing with near-real-
time constraints are non-trivial issues.
Limited Computing Resources: The addition of thou-
sands of new connected devices stresses not only the net-
works but also the computational power (for example at
RSU) [11]. Advanced driver-assistance systems (ADASs)
and autonomous vehicles (AVs) with numerous onboard sen-
sors will generate large amounts of data to be processed.
Additionally, ensuring a holistic view of the ambient environ-
ment is beyond the capacity of a single vehicle, as it requires
aggregating the point of view of multiple vehicles to recreate
the scene [12], [13]. Therefore, off-board computation is
crucial to cope with such situations [14], [15].
Rapid Topology Change and High mobility: The relative
speed between vehicles ranges from tens of Km/h (vehicles
travelling in the same direction on an urban street) to over 280
Km/h (vehicles travelling in opposite directions on a high-
way). Thus, vehicles may be members of a given VANET for
only a very short time leading to rapid and frequent network
topology changes [16], [17]. Additionally, traffic congestion
can be predictable (e.g., due to rush hours) and unpredictable
(e.g., due to traffic accidents) [18]. These phenomena lead
to large volumes of data (due to frequent updates) in some
urban areas and may run into limitations in VANET-based
application scalability [19].
Detrimental Delay: Propagation and queuing delay are
major sources of delay in ITS. Propagation delay depends
on the communication medium and the physical distance
between the communication source and the destination. For
example, if the source and destination are in the same neigh-
bourhood at a distance of 1 km, the propagation delay will be
5µsec. However, if they are located in different countries
at a distance of 12,000 km the delay can reach 58.2 ms.
Queuing delay denotes the network delay of data while
waiting in network queues to be initially sent by the sender,
forwarded by intermediate nodes, or processed by the desti-
nation. Queuing delay is related to the number of transmitting
vehicles and the volume of data sent by each vehicle (i.e.,
the network traffic), the available number of links between
the source and destination, max queue lengths on the nodes,
and the service policy (e.g., critical safety or non-safety)
which determines the queuing prioritization [20]. Overall,
the networking overhead and latency associated with remote
cloud resources could degrade the overall performance and
prove detrimental to road safety [11], [21], [22]. Moreover,
traffic volume is increasing with user demand and will fur-
ther heavily burden backhaul links and lead to even longer
latencies [23].
Security and Privacy Concern: Some ITS applications
and services require vehicles and RSUs to exchange mes-
sages containing potentially sensitive information such as
VOLUME 4, 2021 3
Distributed Vehicular Computing at the Dawn of 5G
real-time locations. This communication takes place over
networks that are by design somewhat easily accessible 1thus
prompting security and privacy challenges [24]. Furthermore,
the richer the data sharing the more potential exists for tactics
like predatory marketing and user tracking [25]. As the
number of services grows passengers may also use services
over multiple networks, therefore. magnifying the potential
for cyber-security problems or privacy violations.
III. V2X COMMUNICATION OVERVIEW
Vehicles and roadside infrastructure use multiple wireless
technologies to communicate. The most promising wireless
communication technologies can be classified into short-
range communications such as dedicated short-range com-
munication (DSRC) [26] and ITS-G5 [27], [28], and long-
range communications including long term evolution (LTE)
and 5G. These technologies vary according to their range,
capacity, and communication latency. Each technology is
thus suitable for a specific class of applications.
A. SHORT-RANGE COMMUNICATION (DSRC AND
ITS-G5)
Dedicated Short Range Communication (DSRC) is a vehicu-
lar communication technology that operates in 75 MHz of li-
censed spectrum in the 5.9 GHz band in the United States and
supports intelligent transportation systems. DSRC allows ve-
hicles and RSUs to form vehicular ad-hoc networks through
vehicle-to-vehicle (V2V) and infrastructure-to-vehicle (V2I)
communications. DSRC operates based on the interoperabil-
ity between standards that form its protocol stack. IEEE
802.11p [29] is the native technology, a derivative of the
IEEE 802.11 (WiFi) standard, at the physical and medium
access control layers. In data link, network, and transport
layers, DSRC employs a family of IEEE 1609 standards: the
IEEE 1609.2 [30], 1609.3 [31], and 1609.4 [32] standards
for security, network services (including the WAVE short
message protocol - WSMP) and multi-channel operation.
WSMP is a bandwidth-efficient protocol used for exchang-
ing single-hop messages and non-routed data. WSMP sends
packets referred to as WAVE short messages (WSMs). IEEE
802.11p and IEEE 1609 standards allow vehicles to operate
in a rapidly varying environment and exchange messages
either without having to join a basic service set (BSS) or
within a WAVE BSS [26], [33]. DSRC seeks to enable
vehicular collision prevention applications that depend on
periodic data exchanges among vehicles and between vehi-
cles and roadside infrastructure with strict round trip latency,
broadcast frequency, and packet error rate requirements [34].
According to ETSI, cooperative collision avoidance requires
a guaranteed maximum latency time of 50 ms and a minimum
frequency of 10 hz to broadcast pre-crash state in cooperative
awareness messages that are associated with direct V2V
communication [35]. DSRC fulfills the requirements of such
1Easy accessibility is important because many ITS applications, such
as cooperative perception, demonstrate local network effects (wherein the
benefit for each user scales with total number of local users).
applications while also providing high security, low latency,
and high-speed direct communication between entities, with-
out involving a centralized network infrastructure [26].
ITS-G5 is an analogous European technology for vehicu-
lar communication that also uses the 5.9 GHz frequency band
but is adapted to European requirements. This standard is
developed by the European Telecommunications Standards
Institute (ETSI) to guarantee interoperability among com-
munication devices from different manufacturers. Similar
to DSRC, it carries V2V and V2I in an ad-hoc fashion.
They are, however, different only in the way they access the
channel [27], [28]. The ITS G5 approach includes a model
consisting of state machines and different tunable parameters
to control the medium access of all nodes. ITS-G5 standard
adds features for decentralized congestion control methods to
control the network load [28], [36].
B. LONG-RANGE COMMUNICATION (LTE AND 5G)
The Long term evolution (LTE) standard is mobile com-
munications standard developed by the 3rd Generation Part-
nership Project (3GPP 2). The LTE system infrastructure
comprises a core network, also known as an evolved packet
core (EPC), and an access network, referred to as an evolved
universal terrestrial radio access network (E-UTRAN). Fur-
ther details on the basics of the LTE architecture can be found
from 3GPP [37].
In the vehicular context, 3GPP has developed Cellular-
V2X (C-V2X) to operate in 5.9 GHz band (similar to DSRC)
in addition to the licensed carriers via network infrastructure.
This enables direct communications in the absence of cellular
infrastructure in a distributed manner [38]. C-V2X works
in two transmission modes to support ITS services: 1) C-
V2X/PC5 which supports V2X direct sidelink communica-
tions, allowing vehicles and RSUs to inter-communicate di-
rectly without the need for infrastructure, and thus providing
lower delay, higher throughput, lower energy consumption,
and better spectrum utilization [39], 2) C-V2X/Uu commu-
nications to connect road users (e.g., vehicles and RSUs)
indirectly through LTE infrastructure. In this mode, since the
V2X transmissions are scheduled, interference and collisions
are lessened [38].
3GPP has been developing C-V2X to base on the fifth-
generation 5G mobile communications standard, leading to
New Radio (NR) C-V2X which is going to be compatible
with the evolution path of 5G. NR C-V2X is gaining mo-
mentum with deployments in many countries. In general, 5G
aims for ultra-reliable low-latency communication. Besides,
5G supports ultra-high throughput which is 10-100×higher
compared to LTE. Similar to LTE C-V2X, sidelink mode
allows direct communication between vehicles, while indi-
rect (via infrastructure) mode works inside the coverage
range of a gNodeB. However, NR C-V2X supports unicast,
group cast, or broadcast transmission modes while LTE C-
V2X only supports broadcast transmission mode. The 5G
2https://www.3gpp.org
4VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
Network Slice
Manager
AMF UDM
PCF
AUSF
SMF
AF
eMBBuRLLC
mMTC
gNodeB
UPF
UPF #1 V2X App AMF #1 SMF #1
UPF #2 MEC
Caching AMF #1 SMF #2
AMF #2 SMF #3 UPF #3
Train Passengers/
Massive Smartphones
Safety/ Autonomous
Driving
Video Streaming/
Augmented Reality
Data Network
mmWave
eNodeB
(F/C)DMA
Figure 2: 5G network slicing using Network Functions (NFs), Access and Mobility Management (AMF), Authentication Server
(AUSF), Unified Data Management (UDM), Session Management (SMF), Application (AF), and Policy Control Functions
(PCF) [40]. The car leverages two slices, a uRLLC slice to support road safety or autonomous driving, and an eMBB slice for
video streaming or augmented reality. While a train leverages an mMTC slice to handle the massive smartphone traffic of the
passengers [41].
basis ensures interoperability with earlier communications
systems such as LTE (i.e., Non-standalone) and provides
good communication performance for vehicular applications.
Additionally, 5G supports integration (beyond simple IP-
based connectivity) with many existing communications and
telecommunication systems including 3G, 4G, WiFi variants,
ZigBee, and Bluetooth. This integration provides vehicular
networks with more flexibility, allowing vehicles, drivers,
passengers, and pedestrians to leverage the most suitable
network for their selected application [42].
In addition, a variety of 5G and 5G-adjacent technologies
including software-defined networks (SDN), network func-
tion virtualization (NFV), and multiaccess edge computing
(MEC) are accelerating the ongoing migration of intelli-
gence closer to the users. These paradigms are the building
blocks of the network softwarization happening in mobile
networks [43]. 5G ecosystem enables vehicle manufactur-
ers, solution integrators, network and service providers, and
small and medium-sized enterprises (SMEs) to efficiently
compete and cooperate. Within the 5G System, end-to-
end network slicing, service-based architecture, software-
defined networking (SDN), and network functions virtual-
isation (NFV) are the fundamental pillars to support the
heterogeneous key performance indicators (KPIs) of the new
use cases in a cost-efficient way. SMEs will be able to provide
technological solutions which comply with the overall sys-
tem design. Manufacturers and solution integrators can offer
rapid deployment enabled by virtualisation and standardised
interfaces to assimilate the advent level of innovation. Mobile
network operators (MNOs) and infrastructure providers will
create tailored slices with specific functionalities and services
to address the requirements of vertical industries [44].
The international telecommunication union (ITU 3) en-
visions the capabilities of future mobile networks in the
international mobile telecommunications-2020 (IMT-2020)
standard. The capabilities entail flexibility, reliability and
3https://www.itu.int
VOLUME 4, 2021 5
Distributed Vehicular Computing at the Dawn of 5G
User experienced data rate
Mobility
Connection density
Area traffic
capacity
(a) Infotainment applications (e.g. Video streaming, AR)
User experienced data rate
Mobility
Connection density
Area traffic
capacity
(b) Critical safety applications.
User experienced data rate
Mobility
Connection density
Area traffic
capacity
(c) Autonomous driving.
User experienced data rate
Mobility
Connection density
Area traffic
capacity
(d) Efficiency and traffic management applications.
Figure 3: The importance levels (High, Medium, Low) of key performance indicators for different classes of applications.
security when providing various services in three intended
usage scenarios, enhanced mobile broadband (eMBB), ultra-
reliable and low-latency communications (uRLLC), and mas-
sive machine-type communications (mMTC). ITU sets the
guidelines for 3GPP to create and maintain the technical
standards for 5G technologies.
The 5G ecosystem and defined use cases (e.g., enhanced
mobile broadband (eMBB) and ultra-reliable low-latency
communication (uRLLC)) are promising enablers of ITS
services and applications. For instance, passengers can watch
an HD movie while the driver is using augmented reality ap-
plications to detect road hazards with real-time and visually
interactive navigation (usage of eMBB). Figure 2 illustrates
the orchestration and architecture to achieve network slicing
and vehicular applications’ use cases. Figure 3 illustrates
the importance level (Low, Medium, or High) for the key
capabilities of 5G to address these use cases. Figure 3a
illustrates the importance level of the key capabilities of
5G to address the infotainment applications such as video
streaming, augmented and virtual reality, and mobile cloud
gaming for passengers during commuting. These applica-
tions belong to the eMBB use case, whereas critical safety
and time-sensitive applications belong to the uRLLC use
case, featuring stringent requirements for reliability, latency,
and continuous, seamless connectivity [40]. Figure 3b illus-
trates the importance level (Low, Medium, or High) of the
core capabilities of 5G to address critical safety applications.
Autonomous driving (AD) requires ultra-high reliability, low
latency, and high bandwidth, a combination of uRLLC and
eMBB use cases. Figure 3c illustrates the importance level
for the key capabilities of 5G to address AD requirements.
Efficiency and traffic management applications are more re-
silient and less dependent on latency and reliability compared
to safety applications. Figure 3d illustrates the importance
level for each 5G KPI to serve these applications.
6VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
C. COMPARISONS OF PERFORMANCE OF
SHORT-RANGE COMMUNICATION STANDARDS
There has been extensive research into comparisons between
DSRC and LTE C-V2X technologies [45]. Such comparisons
have shown that the best technology depends somewhat
on the deployment scenario (e.g., dense urban roads vs.
highways) and the application. For example, 5G Automotive
Association (5GAA) and Papathanassiou et al. have con-
ducted extensive experiments to compare the performance
of DSRC and C-V2X radio technologies for their suitability
to deliver vehicle-to-everything broadcast safety messages.
They confirm that LTE C-V2X significantly outperformed
DSRC in various key areas [46], [47]. Therefore, LTE C-
V2X seems to be the promising candidate to enable these ITS
services and applications. However, DSRC has undergone
several large-scale field trials and is already in production in
the US, Europe, and Japan. In fact, the coexistence of both
DSRC and LTE C-V2X is likely in many regions. Therefore,
Ansari et. al. [45], for example, emphasize the need to enable
V2X communication regardless of the underlying technology
(DSRC or LTE C-V2X). As such, a hybrid V2X system is a
potential comprehensive solution. The hybrid V2X scheme
could apply spectrum sharing techniques such as frequency
division multiplexing with a guard band. Additionally, per-
formance and scalability issues of both IEEE 802.11p DSRC
and LTE C-V2X/PC5 Mode 4 have been driving the future
developments of IEEE 802.11bd DSRC and NR C-V2X [45].
The coexistence of these new future standards is an open
research area.
Summary and Take-Away Message. NR C-V2X is a
promising vehicular technology for transmitting large data
volumes directly or indirectly over short distances. However,
5G-assisted long-distance communication (e.g., vehicle-to-
cloud) will still incur significant latency as the data must
traverse the backhaul (core) network. For instance, when a
vehicle sends sensory data to a manufacturer cloud server
(where the data is processed). Edge computing, thus, is a
useful paradigm to enable data processing in geographic
proximity to the vehicle, thus lowering the latency and en-
abling time-critical applications.
IV. VEHICULAR COMPUTING OVERVIEW
External (to the vehicle) computing resources are important
for an ITS as such resources help with the aggregation and
fusion of heterogeneous data from multiple road users (thus
providing a holistic view) [12], and with enabling of complex
applications so that these applications are accessible regard-
less of the processing power and storage capabilities of the
vehicles. For example, traffic management, emergency man-
agement, fleet management, and Intelligent navigation (e.g.,
through augmented reality overlays on the windshield) are
complex applications that might require relocation of com-
plex tasks to external compute and storage resources [48].
These resources can be remote computing resources (such
as cloud servers) or intermediary nodes such as multiaccess
edge computing (MEC) servers and fog computing nodes.
A. CLOUD COMPUTING
The NIST (American National Institute of Standards and
Technology) defines cloud computing as a model for en-
abling ubiquitous, convenient, and on-demand access to a
shared pool of configurable computing resources (e.g., net-
works, servers, storage, applications, and services). These
resources can be rapidly provisioned and released with min-
imal management effort or service provider interaction [7].
Cloud computing has so far been the dominant paradigm
in terms of offloading vehicular-intensive computation. For
instance, Toyota’s connected car architecture is powered by
Microsoft Azure HDInsight to process millions of events a
day. Toyota provides its vehicles with a data communication
module to transmit the vehicular data to a Toyota smart
center. The latter provides a mobility services platform that
enables public companies to offer Toyota and Lexus vehicles
specific services. In other words, SMEs will be able to
provide technological solutions which will comply with the
overall system.
Beyond traditional cloud services, the huge vehicular
fleets on our roadways, streets, and parking lots can be
seen as massively underutilized computational resources.
Given this framing, Vehicular Cloud Computing (VCC)
has also emerged as a new hybrid technology that incorpo-
rates vehicular ad-hoc networks and cloud computing. In this
paradigm, the underutilized vehicle’s resources, computing
power, internet connectivity, communication resources, and
storage, are shared or rented over the Internet to various
customers [49]. Through seamless and decentralized man-
agement of cyber-physical resources, VCC provides third-
party or community services at low cost and enables efficient
utilization of vehicle resources. Additionally, due to vehicle
mobility, agility and autonomy, VCC can dynamically adapt
its managed vehicular resources allocated to an application
according to the dynamically changing requirements and
system conditions. However, such a paradigm still faces high
relative latency and high communication costs [50], [51]. In
practice, stationary vehicles or mobile vehicles are controlled
by cyber-physical resource management software to form
VCCs. VCCs can thus be categorized into two classes: static
VCCs and dynamic VCCs. These classes are suitable for
different vehicular cloud services or applications [49]. VCC
viability is further enhanced by 5G deployment as 5G pro-
vides capabilities, such as large bandwidth, ultra-reliability,
low-latency, and V2V communication through 5G side-links,
that will support even more VCC use cases.
The proven economic benefits of cloud computing make
it likely to remain a permanent feature of the future comput-
ing landscape. However, the network overhead and latency
of remote cloud computing cannot meet the requirements
of time-critical applications and thus proves detrimental to
overall network performance. Additionally, cloud computing
lacks context-awareness that, for example, captures spatio-
temporal traffic and driving patterns.
VOLUME 4, 2021 7
Distributed Vehicular Computing at the Dawn of 5G
Wide Area Network
Data Center
Upper Tier:
Cloud Computing
Intermediate Tier:
Edge Computing
FCN
RSU
Cloudlet-2 Box
VM#1
VM#2
VM#n
MEC Host
App1
App2
eNodB
DSRC
WAVE 802.11p
DSRC
WAVE 802.11p
Lower Tier:
Vehicles/RSUs
RSU/FCN
MEC Host
App1
App2
gNodB
Figure 4: 3-Tier architecture for an edge computing paradigm
as an extension of cloud computing [52], [53].
B. EDGE COMPUTING
Edge computing (EC) is a distributed computing paradigm
that places computational resources and storage geographi-
cally close to end users (for example, vehicles and RSUs).
Thus service requests typically travel a much shorter physical
distance (and traverse fewer network nodes) for processing
compared to requests to typical centralized cloud servers.
This results in significantly lower latency. Additionally, EC
can complement cloud computing by masking transient cloud
outages and can naturally better capture contextual and sit-
uational information due to the proximity to end users [8].
Overall, EC promises to deliver scalable, reliable, and low
latency cloud services.
Edge computing encompasses three distinct frameworks in
the context of vehicular networks: vehicular fog computing
(VFC), multiaccess edge computing (MEC), and mobile ve-
hicular cloudlets (MVCs). Figure 4 illustrates the architec-
tures of these three frameworks.
Multiaccess Edge Computing (MEC) is an edge archi-
tecture standardized by ETSI 4that brings edge computing to
the mobile network context. Specifically, MEC locates com-
puting resources at the edge of the mobile access network,
typically at the first aggregation level (base stations) [66].
Being an open standard, MEC also creates a standardized and
open environment that enables operators to open their radio
access network (RAN) edge to authorized third-parties, to
flexibly and rapidly deploy innovative applications. This new
ecosystem allows multi-vendor vehicles, manufacturers, and
transportation agents to integrate their applications for more
convenient digital services efficiently. MEC also enables
applications and services to be hosted on top of the mobile
network elements [57], [67]. Different deployment scenarios
address various performance, costs, scalability, and operator
deployment preferences:
4https://www.etsi.org/technologies/multi-access-edge-computing
Deployment at the radio node (eNB or gNB).
Deployment at aggregation points (LTE EPC or 5GC).
Deployment at the edge of the Core Network (e.g. in a
distributed data center, at a gateway).
Figure 4 illustrates MEC deployment where edge servers are
deployed in the cellular base stations - LTE evolved Node B
or 5G NR base station (gNodeB).
Fog Computing (FC) in the vehicular context refers to
any intermediary computation, storage, and network services
between vehicles and the cloud [68]. Specifically, there are
rich scenarios of connectivity and interactions in vehicular
networks: vehicle to vehicle, vehicle to access points, smart
traffic lights and roadside units (using Wi-Fi, DSRC), vehicle
to network (using LTE, 5G), and other V2X scenarios. For
instance, a smart traffic light node interacts locally with
many sensors, which detect the presence of pedestrians and
bikers, and measure the distance and speed of approaching
vehicles. The smart traffic light in this context acts as a fog
computing node (FCN). A fog computing node (FCN) can
be any node with communication, computation, and storage
resources. As shown in Figure 4, a FCN can be a moving or
parked vehicle, also referred to as vehicular fog computing
(VFC) [69], a roadside unit, or an edge device installed in a
cellular base station. Vehicular Fog Computing (VFC) ag-
gregates the abundant resources of individual and connected
vehicles and exploits their available computing resources to
enhance the application quality of service. VFC uses moving
and parked vehicles as FCNs to offload computation tasks
and provide networking services [61], [70], [71]. The fog
has several characteristics which make it the ideal platform
and non-trivial extension of the cloud to deliver services in
infotainment, safety, traffic efficiency, and analytics. These
characteristics are 1) low latency, 2) wide-spread and geo-
distributed deployment, 3) location, mobility, and context-
awareness, 4) interoperability, federation, and heterogeneity
(deployable in various environments), and 5) support for real-
time interactions [72], [73].
Cloudlet Computing represents an architecture with aux-
iliary proximate cloud resources for providing highly re-
sponsive services. Specifically, these cloud resources can be
viewed as delegates or proxies of the real cloud and are
located at the middle tier of a three-tier hierarchy, as shown
in Figure 4. A cloudlet can be either a mini data center
in a box [74], [75], or vehicular resources referred to as a
mobile vehicular cloudlet (MVC). As an example use case,
during cloud or backhaul outages, the cloudlet takes over the
responsibilities and masks the outage [76]. Adjacent vehicles
and roadside units can connect via DSRC communication
or 5G sidelink to form MVCs. Thus MVCs harness the
computational resources of the adjacent nodes in a timely
and efficient manner via peer-to-peer communication. An
MVC is a cluster of smart vehicles and RSUs located in
a region. Such Vehicles and RSUs can share resources and
information via V2V communication or indirectly via V2I
communication [49], [77].
Overall, edge computing architectures have similarities such
8VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
Table 2: Summary of Cloud and Edge Computing Frameworks [23], [52], [54]
CC /VCC Fog Computing MEC Cloudlet
Origin Amazon / Olariu et al [55] Cisco [56] ETSI [57] Satyanarayanan et al [58]
Deployment
Location
Data center / stationary and mo-
bile vehicles [49], [59]
At any point between vehicles
and cloud
Radio access network Local or outdoor installation
[58], [60]
Deployed
Nodes
Dedicated servers / underused
vehicle resources
VFC (vehicles) [61],
RSUs [62], or connected
ESs [51], [63]
ESs running in BSs aggregation
or core of RAN
MVC-adjacent smart vehicles
and RSUs [64] Datacenter in
box [58], [60]
Access
Technology
Internet WiFi, mobile networks [52] Mobile networks (LTE, 5G) WiFi [52], DSRC or ITS-G5
Proximity
[52]
Many hops, 10s to 1000s of km One or multiple hops between
vehicles and cloud
One hop, 100s of meters to few
km
One hop, nearby
Context
awareness
[23], [52]
No Medium High Low
Latency High, AWS196±84msa,
Azure176±96msb,
Google172±106msc
Low Low, up to 19.9 ms [65] Low, few ms [58], [60]
aAws ping test (latency): ping.varunagw.com/aws
bMeasure your latency to Google cloud platform(gcp): www.gcping.com/
cAzure latency test: www.azurespeed.com/
as the purpose. However, they have slight differences in
the origin, the deployment location, the involved nodes, the
access technologies, the geographical proximity, the level of
contextual awareness, and the latency. Table 2 summarizes
the differences between edge computing architectures (FC,
MEC, Cloudlet) and compares them with cloud computing
architectures (CC, VCC).
Summary and Take-Away Message. The synergy of NR
C-V2X and edge computing (EC) drops the end-to-end la-
tency significantly, allowing stakeholders to use bandwidth
efficiently, and enabling time-critical applications to run in
real-time. Edge computing reduces the networking overhead
and provides highly responsive services for the users on
wheels. Moving the computation to the edge pushes the
utilization efficiency of the next-generation mobile network
to its limit. VCC increases the utilization efficiency by using
the dispersed underutilized resources of the vehicles. Vehic-
ular networks can be employed to remotely offload latency-
tolerant computation (into moving or parking vehicles) and
storage services (into parking vehicles), or locally offload
latency-sensitive computation (into moving vehicles) and
caching (into moving vehicles).
V. DATA ANALYTICS: TECHNOLOGIES AND
METHODOLOGIES INTEGRATION.
Road users and elements (e.g., Pedestrians’ smartphones, ve-
hicles, lampposts, traffic lights, or other RSUs) generate both
mobility and service-related data which is heterogeneous and
large in volume. Analysing this data and extracting useful
and relevant information in real-time requires an efficient
data analytics architecture. This architecture must support
multiple data sources, allow large data volumes, enable data
streaming (to achieve low latency), and allow developers to
Induction
loop detector
occupancylight duration
speed,
acceleration
road ID, type,
lanes, direction
timestamptimestamp timestamp
location loc, speed limit
Data
Fusion
Data StreamingPre-processingMachine Learning
Car-following
Control
Flow Prediction Incident Detection
Route
Recommendation
accelro/magento
meter, gyroscope
timestamp
location
Data
Analytics
Service
5G
fuel consumed
Trajectories, Traffic and Road Data
Driver Behavior
Recognition
Traffic Light
Control
Figure 5: A general data analytics pipeline illustrating traffic
data collection, aggregation and fusion, streaming into micro-
batches, and processing.
plug-in queries and machine learning algorithms. A variety
of specific technologies and frameworks can be combined
to actually realize such an architecture, we briefly describe
some of the most common technologies and frameworks.
Additionally, we study several specific architectures from
VOLUME 4, 2021 9
Distributed Vehicular Computing at the Dawn of 5G
literature and two case studies.
A. DATA ANALYTICS TECHNOLOGIES
Firstly, in terms of data storage technologies, Hadoop Dis-
tributed File System (HDFS) is a distributed file system
for reliably storing large amounts of unstructured, semi-
structured, and structured data as files (typically on disk).
HDFS was one of the first large-scale distributed file systems
for big data. Several other big data storage systems actually
build on top of HDFS. HBase, for example, is a key-value
pair NoSQL database with master-slave replication that lever-
ages HDFS as underlying storage. Other notable systems
include Cassandra, a popular key-value pair NoSQL database
with asynchronous masterless replication.
In terms of data messaging, collection, and aggregation,
the current dominant system is Apache Kafka5. Apache
Kafka is a distributed event streaming system. Specifically,
Kafka provides a distributed publish-subscribe messaging
system that allows for decoupling of different stages of
data pipelines. Kafka accommodates big heterogeneous data
and Kafka event streaming includes true (event at a time)
streaming with exactly-once semantics.
Finally, in terms of actual distributed computing and ML,
Hadoop MapReduce is a distributed computing framework
for the parallel processing of large datasets often stored
on disk on HDFS (though other storage solutions are also
supported). MapReduce runs on a Hadoop cluster and of-
ten leverages a cluster manager like YARN to schedule
applications and services on the cluster and manage the
cluster resources like memory and CPU. In comparison to
the primarily disk-based MapReduce, Apache Spark 6is a
unified big data analytics engine for distributed in-memory
data processing. Furthermore, Spark provides both batch and
stream processing, libraries for Machine Learning, and an
SQL-like interface. Relatedly, Apache Flink is also a big
data analytics engine for distributed processing. A few of the
major differences between Spark and Flink are that Spark
is more mature with a larger community, while Flink was
designed specifically for stream processing and thus provides
better support for true (event at a time) streaming. In contrast,
Spark primarily supports micro-batch streaming. Kafka also
provides some true (event at a time) stream processing func-
tionality (through the Kafka streams API).
B. INTEGRATION OF TECHNOLOGIES TOWARDS
VEHICULAR COMPUTING
In vehicular environments, the big data analytics architecture
relies on the integration of such technologies into a pipeline
that enables computation offloading.
Figure 6 illustrates a potential pipeline that integrates
various technologies to process the traffic data in real-time.
Firstly, Apache Kafka ingests the live vehicular/traffic data
and partitions the data into distinct topics which enables
5https://kafka.apache.org/
6https://spark.apache.org/
Induction
loop detector
Data
Fusion
5G
Data
Aggregation
Data
Streaming
Flink
Streaming
Data
Processing
MLlib
Machine
Learning
Data Batches
Continous Data Stream
Fused Data
Figure 6: Example open source technologies for a data
analytic pipeline. The stages include online and historic
traffic data collection, aggregation, and fusion (using Kafka,
Hadoop, or Flume), micro-batch streaming (using Flink or
Spark Streaming), and processing including machine learn-
ing (using Flink or Spark MLLib).
multiple readers and writers to operate simultaneously thus
improving scalability. A separate data fusion module facili-
tates fusing and aggregating the topics’ data for richer fea-
tures and better context determination. Spark Streaming then
either consumes data from specific Kafka topics or retrieves
data from HDFS, and splits the data into micro-batches to
feed into SparKMLlib which applies ML algorithms. Apache
Flink could replace spark to achieve a very similar setup. The
ML algorithms output meaningful information (such as pre-
dictions) which is often sent back to applications or services
to assist driving or traffic control, or stored permanently in
stable storage such as Hadoop HDFS or Cassandra for later
use. For instance, Amini et al [6] employ Apache Kafka to
stream traffic data in real-time and control traffic lights in a
distributed manner. Microsoft added Apache Kafka on Azure
HDInsight to run a robust, real-time, big data streaming
pipeline at enterprise scale, natively integrated Kafka with
Azure managed disks, and made it globally available. Since
then, large companies have been using this service in produc-
tion to process millions of events per second and petabytes of
data per day to power scenarios like Toyota’s connected car,
Office 365’s clickstream analytics, fraud detection for large
banks, or log analytics [78], [79].
C. ML-EMPOWERED APPLICATIONS
Traffic-related learning tasks can be primarily sorted into
two main classes, basic safety and advanced efficiency.
10 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
Collision warning and traffic incident detection are examples
of basic safety, whereas traffic flow prediction, car-following,
and driving behavior recognition are examples of advanced
efficiency. To implement these tasks a machine learning algo-
rithm is applied to traffic data in a data analytics pipeline (see
Figure 5). We briefly discuss the details of these different task
classes including example systems from research studies.
Traffic Flow Prediction is the real-time short-term pre-
diction of traffic on the road network that assists in under-
standing the future traffic state. This prediction can leverage
both longer-term historical traffic data (e.g., diurnal patterns)
and up-to-date signals of traffic conditions. Such prediction
plays a significant role in road network traffic planning and
traffic control optimization and lays the foundation for, travel
guidance, navigation, and other mobility services. A variety
of temporal ML algorithms including NN-based and more
traditional algorithms have been applied to the task. The
most sophisticated NN-based models including long short-
term memory (LSTM), stacked LSTM, temporal LSTM,
and spatial-temporal autoencoder LSTM (SpAE-LSTM) out-
perform the more traditional multilayer perceptron (MLP)
model, decision tree model, and support vector machine
(SVM) models for traffic flow (congestion) prediction [80]–
[82]. In terms of details, SpAE-LSTM, for example, is a
hybrid model consisting of a sparse autoencoder and an
LSTM. The sparse autoencoder captures the spatial features
while the LSTM captures the temporal features [82]. In fact,
many of the related models leverage autoencoders as they
can learn generic traffic flow features and obtain the internal
relationship of traffic flow [83]–[85].
In addition to short-term prediction, trend-modelling of
the traffic time series can facilitate longer-term traffic fore-
casting. Such forecasting relies on the implicit temporal
correlations among the time series observed on different
days/locations due to human diurnal patterns. Specifically,
the daily traffic time series at a certain location have similar
M-shapes over consecutive days, in which the morning and
evening rush hours correspond to the two peaks of the M-
shape [86]. Li et al [86] use principle component analysis
(PCA), a well-known mathematical dimensionality reduction
and feature extraction procedure, to project the traffic time
series onto an n-dimensional orthogonal linear space such
that the data with the kth largest variance by projection lies
on the kth dimension. PCA trend can establish a link be-
tween traffic time series collected in different days/locations
because the observed daily data series share the same set
of latent variables. PCA also can assist to predict whether
the traffic is normal or abnormal by comparing the distances
between their projections in the latent space.
Traffic Incident Detection is the detection of real-world
traffic incidents in some given spatiotemporal area. This
detection is essentially a mining task from heterogeneous
traffic data. An interesting data source for incident detection
is social media such as Twitter and Facebook (which are
popular and real-time in nature). Specifically, the public
often immediately posts and shares information when traffic
incidents occur (e.g., road closures, traffic congestion, or
accidents), which spreads rapidly through the network. The
user messages shared in social networks are called status
update messages (SUMs). These messages may contain text
and meta-information such as timestamp, geographic coordi-
nates (latitude and longitude), name of the user, links to other
resources, hashtags, and mentions. Several SUMs referring
to traffic incidents in a limited geographic area may provide
valuable information about abnormal traffic incidents.
Text mining and natural language processing (NLP) are
the specific ML areas relevant for the extraction of useful
information and knowledge from unstructured text (e.g.,
Tweets and Facebook posts) [87], [88]. Thus pre-processing
and algorithms from these areas become important com-
ponents in the data analytics pipeline for such a detection
application. In terms of research implementations, D’Andrea
et al. [89] present a real-time traffic event monitoring system
that leverages Twitter stream analysis. They employ SVM to
classify tweets as traffic event related with very high accuracy
(>95%). Relatedly, Traffic Events Detection and Summary
(TEDS) [90] uses NLP, spatial-temporal mining, and wavelet
analysis techniques to create a traffic incident map with text
summarizations (from multiple same-incident tweets) from
Twitter data.
Vehicle following is the systematic control of vehicles’
velocity to optimize and maintain safe, comfortable, and
convenient traffic flow. To this end, car-following models
determine the velocity of a following vehicle in response
to actions of a lead vehicle [98]. Many works have de-
veloped reinforcement learning (RL) models which control
the vehicle’s velocity to optimize the traffic flow. Meixin
et al [96] develop a deep RL model which uses a reward
function reflecting driving safety, efficiency, and comfort to
fulfil the multi-objectives of the car-following model. They
design the RL reward function to combine driving features
and maximise the cumulative rewards and which leads to
more efficient traffic flow compared to human drivers and
faster running speeds compared to model prediction control.
Collision avoidance strategy is incorporated for safety and
faster convergence.
Relatedly, Wang et al. [95] use deep RL to control lane-
changing behavior for each vehicle with a reward function
defined as a trade-off between the vehicle’s travelling ef-
ficiency (i.e., how efficiently a vehicle maintains a target
speed), traffic flow rate, and level of cooperation between
the vehicles. Specifically, they utilise a deep neural network
named deep Q-network (DQN) as the RL model and the lane-
changing of each vehicle is formulated as a Markov decision
process (MDP). The state space S(t)is the vehicle’s state
(at a given time t) which consists of three sequential frames
of traffic snapshots and the corresponding speed difference
between actual and target speed. The action space A(t)is
the corresponding driving decision such as switch to left or
right lane, speed up by a fixed increment up to a maximum
speed, or maintain the current speed. Walraven et al [99] also
apply MDP, Q-learning, and neural networks to learn policies
VOLUME 4, 2021 11
Distributed Vehicular Computing at the Dawn of 5G
Table 3: Summary of Machine Learning Methods for ITS Applications
ITS Application Phenomenon ML method Data Source Deployment
Distraction Detection Using mobile phone while driv-
ing
Deep Learning -CNN [91] Ceiling-mounted camera Vehicle OBUs
CNN and RNN [92] Inward facing camera and IMU
data
Traffic flow prediction Congestion Stacked LSTM [80], Temporal
LSTM [81], Spatial-Temporal,
LSTM [82]
Traffic Data RSU
Traffic flow, speed prediction Traffic dynamics CNN, RNN, SAE and autoen-
coder [83]
Traffic Data from Infrastruc-
tures, Trajectory AFC Records
& Social media
-
Traffic Accident Detection Traffic anomalies SVM [89] Social networks -
Driving Detection & iDentifica-
tion, D3
Abnormal driving SVM [93] Smartphone sensors Smartphone app
Drowsy Driving Detection
(D3)
Drowsy driving LSTM [94] Embedded Acoustic Sensors in
Smartphones
Cooperative Lane Changing Traffic Competition Deep RL [95] Vehicle on-board sensors Vehicle OBUs
Safe, Efficient and Comfort
Following
Car Following MDP, Deep RL [96] Vehicle on-board sensors Vehicle OBUs
Intelligent Cross-Layer& Co-
operative Offloading
Diverse requirements, Time-
varying of Content Popularity
Deep Deterministic Policy Gra-
dient (DDPG) [97]
Vehicle’s OBUs, RSUs, Envi-
ronment
RSUs, Base sta-
tions
dictating the maximum allowed driving speed on highways
to reduce traffic congestion. The above-mentioned systems
are based on RL with a reward function that improves the
overall traffic efficiency instead of the travel efficiency of
an individual vehicle. In short, cooperation leads to a more
harmonic and efficient traffic system rather than competition.
Driver Behavior Recognition is the classification of a
driver’s behavior into classes such as normal, aggressive, or
distracted driving. The classifier output is conveyed to the
drivers audibly, visually, or haptically using an infotainment
system to raise the driver’s awareness and allow them to react
in time. Raising the driver’s awareness improves their driving
behaviors and thus promotes safer driving, reduces traffic
accidents, and contributes to social safety [100]. Distracted
driving is an especially relevant and dangerous driving be-
havior. Distracted driving can be defined as any activity that
diverts the driver’s attention from driving including talking
or texting on a mobile phone [101] or using on-board enter-
tainment or navigation systems. In the US, 3,166 people in
2017 [102] and 3,477 people in 2015 died in motor vehicle
crashes involving distracted drivers.
Celaya-Padilla et al [91] leverage a ceiling-mounted wide-
angle camera that feeds data to a convolutional neural net-
work (CNN) to detect distracted drivers. The detection of
distracted driving can then be conveyed to the driver audibly,
visually, or haptically using, for example, the infotainment
system. Relatedly, DarNet [92] is a framework utilizing
CNNs and recurrent neural networks (RNNs) to process im-
ages (from a driver-facing camera) and inertial measurement
unit (IMU) data (from the driver’s mobile device) to detect
distracted driving behavior. These data sources provide rich
contextual information that allows fine-grained. For instance,
an image of a driver sending a text message can be cross-
validated by checking the acceleration of the mobile device
from the embedded accelerometer. Such multi-modal cross-
validation improves the classification accuracy without the
need to deploy additional sensors. More generally, Driving
behaviour Detection and iDentification system (D3) [93] also
detects abnormal driving behaviors using real-time smart-
phone sensors and an SVM-based ML algorithm. These
driving behaviours include, for example, weaving, swerving,
sideslipping, fast u-turn, turning with a wide radius, and
sudden braking.
In addition to distracted driving, drowsy driving is another
problematic behaviour that threatens road safety. Sober-Drive
system [103] is a smartphone-assisted drowsy driving detec-
tion system that uses the smartphone’s front camera feed and
analyses the open/closed states of the driver’s eyes using a
NN model. Thus the system leverages drowsiness indicators
such as the eyelid closure percentage, blink time, and blink
rate. Furthermore, the D3-Guard system [94] detects drowsy
driving using audio recording by smartphones and a long
short term memory (LSTM) network. The system detects
nodding, yawning, and abnormal steering in real-time by
leveraging the Doppler shift of the audio signals to capture
the unique patterns of these drowsy driving actions. In these
systems, model training typically occurs offline whereas the
application uses the real-time smartphone sensory data for
12 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
Figure 7: Integrated V2X and RSU-assisted computing archi-
tecture at local scale
inference in an online phase.
Summary and Take-Away Message. ITS applications
often require accuracy and real-time while handling big and
heterogeneous data. Therefore, the underlying platform must
perform the entire pipeline in real-time, including data inges-
tion, streaming, processing, plugging in the ML model, and
output presentation. Leveraging and combining distributed
data analytics technologies such as Kafka and Spark can
fulfill these requirements. Though the specific setup and
placement of such pipelines will vary significantly given
the diversity in requirements and scope of different ITS
applications.
VI. INTEGRATED COMMUNICATION AND COMPUTING
ARCHITECTURE
Some vehicular applications require offloading computation
tasks to external servers. In this section, we study potential
architectures that support such applications under three dif-
ferent scales: local scale (i.e., road, intersection, or last-mile),
neighborhood scale, and city scale.
A. LOCAL SCALE
At the local level (road/intersection), the first priority is
to provide basic safety functions to prevent accidents. For
instance, active safety applications warn drivers of impending
danger so the driver can take corrective or evasive action. Be-
yond basic safety, advanced efficiency functions also play a
local level role. As an example, an active traffic management
system (an advanced efficiency function) may adapt the local
traffic control system proactively or reactively to improve
local travel flow. Such a system includes consideration of
peak-hour traffic, the detection of and response to incidents,
and the reduction of waiting time due to congestion and
incidents. Thus the system enhances the local transportation
network performance in terms of safety, efficiency, reliability,
scalability, and sustainability [104].
Requirements. Decisions at this scale have to be made
very quickly thus stressing the importance of low overall
system latency (normally less than 100ms).
Basic road safety requires ultra-reliable and low latency
communication. The infrastructure must have the capability
to monitor the traffic situation reliably and make accurate
decisions. For example, the road-side sensors must be capa-
ble of accurately recognizing and localizing various types of
objects (e.g., vehicle, pedestrian, obstacle) with low latency.
In advanced efficiency, the system must observe the situation
and ambient environment and take quick actions on local
scales to ensure smooth traffic flow. For instance, the system
might change the duration of a traffic light phase based on
road occupancy. Though the latency requirement of advanced
efficiency system is on the order of seconds as such delays
only minimally impact performance.
Importantly for both basic safety and advanced efficiency,
local driving patterns vary according to the spatio-temporal
context. For instance, mean driving speed varies fairly pre-
dictably due to time of day (rush hours, night hours, and so
forth), day of the week (weekday or weekend), and road type
(motorway or street road). Moreover, traffic flows change
due to abnormal traffic events such as road incidents. There-
fore, many applications on the local scale require context-
awareness as well as situational-awareness to make accurate
decisions [5], [76]. Additionally, the vehicles must maintain
continuous, uninterrupted, and highly available communica-
tion between each other and with the RSUs. Finally, dynamic
vehicle mobility leads to rapid topology changes in VANETs;
while a variety of events cause transient unbalanced traf-
fic distributions and congestion on roads and intersections.
Therefore, such a category of applications should scale well
with traffic flow and the incurred channel bandwidth us-
age [105].
Architecture and Methodology: To ensure context-
awareness (in advanced efficiency) and low-latency (in basic
safety), a local-scale system should offload computing tasks
to nodes close to road users. Figure 7 illustrates a detailed
example system that includes these architectural consider-
ations and meets the aforementioned local requirements.
Computing nodes co-locate alongside roads (RSUs), or with
cellular base stations (LTE-eNB and 5G-gNB). Relatedly,
offloading computation to edge nodes provides distributed
and parallel computing, allowing the system not only to scale
up with load but also to ensure higher reliability by avoiding
congestion in the back-haul network. Additionally, C-V2X
allows vehicles to communicate with RSUs, thus allowing
the collection and distribution of additional vehicular data
(beyond the data from an RSUs own sensors).
The architecture contains three essential components: a
data streaming module, a data management system, and a
data analysis module to cope with the continuous traffic data
streams. Traffic data includes but is not limited to the vehicle
data, the environment data collected from the area’s installed
VOLUME 4, 2021 13
Distributed Vehicular Computing at the Dawn of 5G
5GC
MEC server
EPC
Cloud
Tier-2
Tier-3
Tier-1
C-V2N
eNB
gNB
gNB
Neighborhood
C-V2N
MEC servers
(a) Communication-aware neighborhood and city scale examples
RSU
RSU
RSU
RSU
Fiber
Fiber
Fiber
Edge
Server
C-V2X
C-V2X
C-V2X
Vehicular
application
Local
application
Server
Central
application
Server
(b) Transport-aware neighborhood scale example.
Figure 8: Scale from communication and transport perspectives. (a) Multiple base stations (local-scale areas) form a neighbor-
hood scale. Tier-2 denotes neighborhood scale, whereas Tier-3 denotes city scale. Vehicles and traffic lights communicate with
base stations (gNB and eNB) using C-V2N. (b) RSU range forms a local-scale area, and multiple RSUs form a neighborhood
scale and collaborate through edge servers. The infrastructure is overlaid on the road network (blue lines) of the Shatin
neighbourhood of Hong Kong.
sensors, and the information obtained from the RSU. The
data streaming module ingests and aggregates the traffic data
streams then fuses both the aggregated data and potentially
stored data from the data management system. Afterwards,
the streaming module divides the continuous data flow into
batches for processing in the data analysis module [5], [6],
[106]. For real-time streaming, existing works [5], [106]
integrate Apache Kafka and Apache Spark to detect unsafe
driving activities. To provide real-time streaming and ana-
lytics of traffic data, they integrate Apache Kafka, to ingest
the continuous stream of traffic data, and Spark Streaming
to divide the stream into periodic micro-batches. The micro-
batches then feed a Spark engine in which they are plugged
into a machine learning algorithm to analyse the data. Amini
et al [6] propose a similar architecture but rather use a
reducer-evaluator instead of Spark to analyse the data. The
architecture is distributed and flexible architecture to control
traffic signals in real-time. It employs Apache Kafka as an
intermediary module between the traffic system with its sen-
sors ( e.g., probe vehicles, loop detectors) and actuators (e.g.,
traffic lights, variable message sign), and the data analysis
module. As data come in, they are being processed via user-
specified reducer functions. After a time interval, a separate
evaluator function is invoked to assess the results and update
the settings of the traffic system accordingly.
B. NEIGHBORHOOD SCALE
Another category of (vehicular) applications involves
decision-making on a larger neighborhood scale. From a
transportation perspective, the neighborhood scale encom-
passes multiple interconnected (via road network) local-scale
areas with a fairly reasonable traffic flow. From a mobile net-
work perspective, it typically encompasses multiple adjacent
macrocells (see Figure 8).
Requirements: Traffic flow (inbound and outbound) and
the underlying road network must drive the decision-making
at neighborhood scale. Decisions are made based on traffic
interactions between local areas and their cascading effects,
and the acceptable latency to take action. Such context-
awareness allows the system to disseminate accident warn-
ings to the relevant road users, balance road occupancy,
alleviate congestion, and thus reduce travel time and CO2
emissions. Importantly, the latency requirements at this scale
are typically less stringent (e.g., on the order of seconds or
minutes) than the local scale. Yang et al [107] consider the
task’s maximum latency (threshold) while minimizing the
required communication and computing resources to imple-
ment a mobility-aware task offloading scheme. The scheme
balances task completion latency and necessary computation
and communication resources.
Architecture and Methodology: To ensure a holistic
neighborhood view, a neighborhood scale architecture of-
ten co-locates the computational units with the aggrega-
tion points of a mobile network, known as communication-
aware neighborhood scale, or the aggregation points of
the RSU system, known as transport-aware neighborhood
scale, as shown in Figure 8. For example, Zhou et al [18],
[108] optimize neighborhood-scale decision making through
14 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
communication-aware MEC servers on two different tiers,
collocated with base stations and aggregation points. In
addition to the three local scale architecture components,
the neighborhood architecture often enables a collaboration
channel between computing nodes to exchange information
(processed data), sometimes through a central computing and
storage node. Some systems [5], [107] enable a neighborhood
or larger view through inter-RSU collaboration whereas Zhou
et al [18], [108] use two-tier MEC computing as shown in
Figure 8a.
CAD3 integrates data streaming and processing compo-
nents for real-time decisions and in-time reactions [5]. At
the core of data analytics, multi-agent reinforcement learn-
ing (MARL) and federated learning (FL) can be utilized to
capture the changing traffic patterns and maintain a smooth
traffic flow [18], [108]. However, dual-mode C-V2X roadside
device, supported by mobility-aware algorithms would allow
the road users to intercommunicate via NR C-V2X/PC5
direct communication channel (5.9GHz band) and allow
them to connect to network infrastructure via the Uu (5G)
communication channel [109], [110]. Dual-mode commu-
nications allow the road users within areas on local scale
to share vehicular movement and traffic status information
with others, leading to a coverage of neighborhood scale (see
Figure 9).
C. CITY OR LARGER SCALE
The last category of applications require city-scale decision
making such as advanced efficiency (e.g., city traffic man-
agement and planning), holiday out-of-city traffic inference,
and studying of big events (i.e., sporting and exhibition
events) [111], [112]. These often involve the think globally,
act locally concept, where data analytics of traffic big data
collected from road users and elements (e.g., vehicles, traffic
lights, pedestrians) covers the entire city.
Requirements: Many city-scale applications require ur-
ban big data that is naturally high volume, high variety, and
high velocity (3Vs). This data can encompass trip and tra-
jectory data, surveillance video, weather, social events, and a
diversity of traffic data. The data must be associated with time
and location stamps, in other words, spatio-temporal data,
to enable many rich spatio-temporal applications including,
for example, finding dynamic dependencies among different
regions at the local and neighborhood-scale [113].
For city traffic planning and management, transport
decision-making involves understanding city-scale human
mobility patterns and discovering traffic problems (traffic
anomaly [111] and accident detection and congestion predic-
tion) [113]. As such, the data analytics engine often resides
in the cloud to handle the heavy computations.
Event prediction and detection often leverages data
streaming including preprocessing and feeding into a ma-
chine learning model to detect events or predict the oc-
currence of events [114]–[116]. For instance, traffic flow
prediction requires city-scale traffic data analytics and urban
dynamics decomposition [111] to identify the traffic flow
5GC
I2N /Uu
gNB
Local Model
Local Model
Global
Model
Global Model Download
Local Model Upload
Agent
Agent
V2I /PC5
Agent
I2N /Uu V2I /PC5
Agent
Neighborhood
Scale
Neighborhood
scale
MEC
Vehicular
application
Local
application
Server
Central
application
Server
Figure 9: Combined transport and communication-aware ar-
chitecture for city-scale deployment. RSU are equipped with
dual-mode connectivity to receive road user data via V2I
direct sidelink (PC5) communication. RSUs then forward
data to the cloud via I2N Uu communication through the
network infrastructure. Road users can also upload local and
download global models (FL) or use local/central agent (RL)
instead of sharing raw data.
patterns and predict the future traffic flow which then helps
to ensure efficient route planning and mobility. While traffic
management requires either a multi-tier communication-
aware architecture [18], [108], [117], or centralized cloud
architecture for collecting and processing the massive traffic
data for monitoring traffic density, throughput, and events
in real time. Though communication-aware systems lack
spatial-temporal correlations and mobility-awareness. Effi-
cient traffic flow requires mobility-aware traffic control so
as to decrease the waiting time of vehicles traveling on
signalized roads.
Architecture and Methodology: A city-level architecture
primarily locates the computing nodes at the cloud given
such centers’ capacity to process large streams of city traffic
data. In such an architecture the road elements (e.g., traffic
lights, lampposts, vehicles, inductive loops) can be equipped
with 5G modules to transmit the large data volumes to
the nearby RSUs via NR C-V2X direct communications
(PC5). The RSUs then transmit data to the cloud through
the network infrastructure, i.e., I2N Uu communications. The
cloud ingests the continuous stream of widespread urban
sensor data and uses data streaming and analytics engines
(e.g., Apache Kafka and Spark) to process them in real-time
including machine learning algorithms for decision making.
Road elements receive data from the cloud to enable vehicu-
lar applications. For instance, traffic lights receive commands
via downlink Uu communications, decision-makers (depart-
ment of transportation) also receive information via downlink
Uu communications, other road users (e.g., vehicles and
pedestrians) receive directions via downlink (Uu) and then
VOLUME 4, 2021 15
Distributed Vehicular Computing at the Dawn of 5G
I2V/PC5 communications (see Figure 9). For traffic manage-
ment, distributed agents can monitor and take actions at the
local scale and neighborhood scale while a city-scale central
agent tunes performance parameters to optimize the overall
traffic flow. However, the transmission of raw traffic data
may still cause issues due to high bandwidth requirements
and privacy restrictions (for example related to General Data
Protection Regulation (GDPR)) [118]. A potential methodol-
ogy for dealing with these issues is federated learning (FL).
FL addresses privacy and bandwidth issues by training local
models and uploading only the local model parameters to
the cloud for aggregation into a global model (that is then
redistributed to the local agents) [119] (see Figure 9).
Summary and Take-Away Message. A robust vehicular
computing architecture is crucial for satisfying the require-
ments of vehicular applications. These applications vary
with some requiring detailed local spatio-temporal context
with low latency decision making, and others requiring a
city-wide holistic view for optimization beyond the local
or neighborhood scale. Some applications will even require
some combination of these requirements. Incorporating the
V2X and 5G ecosystem, including edge computing, can help
cope with such requirements. Specifically, MEC and NR C-
V2X communication are likely very important. NR C-V2X
is a crucial component for reliably sharing the big traffic
data. While MEC enables processing the data even at the
network edge thus allowing a broader range of potential
latency including real-time or near real-time streaming data
analytics (for example with Apache Kafka or Spark).
Overall, we note the boundary between these scales is
not well defined and any traffic system or application must
still, in some ways, consider the holistic traffic environment
on multiple physical scales. In any case, most systems and
applications will actually target multiple physical scales.
VII. REAL WORLD CASE STUDIES
In this section, we study three real-world scenarios: local-
scale cooperative perception, neighborhood-scale accident
warning, and city-scale event detection (urban planning).
These scenarios highlight the leveraged technologies, the
enabling architecture, and the interactions with road users.
A. CASE STUDY: COOPERATIVE PERCEPTION
Modern vehicles have a variety of sensors to perceive the
nearby environment and warn the driver of potential hazards.
However, external objects (e.g., buildings, other vehicles,
trees) may block the view of these sensors, causing blind
spots and raising road safety concerns. Additionally, each
type of sensor also has inherited limitations in terms of
sensing distance, accuracy, and environmental dependency.
Figure 10a illustrates a classic basic safety example. Ve-
hicles pass through a roundabout, inevitably encountering a
blind spot situation that causes a safety issue. Specifically,
when a new vehicle enters the roundabout, the black vehi-
cle’s driver can not see the gray oncoming vehicle, stopped
vehicles, or pedestrians crossing the road behind the stopped
vehicle. The black vehicle might not have an in-vehicle com-
munication and computation module (OBU) and thus cannot
identify and relay information about the stopped vehicle and
crossing pedestrian. The assumption that all vehicles are
equipped with OBUs does not hold in the real world. As such,
the roadside system becomes essential to recognize the object
type (vehicle or pedestrian), position, and speed, accord-
ingly determine the potential danger and notify the driver.
See-around-the-corner vehicular applications allow the RSU
installed at the intersection to distribute real-time sensor
data or notifications to vehicles in range using NR C-V2I
communications and extend the vehicles’ visibility beyond
the sensors and driver’s visibility [38], [121]. Owing to the
short distance and direct communication between the vehicle
and RSU, the roadside system would be able to recognise the
danger and disseminate corresponding warnings within 100
ms (maximum acceptable latency).
When vehicles are equipped with OBUs, they can co-
operatively perceive a larger area of the environment than
could any single vehicle alone. Using C-V2X, the adjacent
vehicles can communicate directly (PC5) to share raw on-
board sensor data (e.g., camera, LiDAR, radar) and processed
information (e.g., information about identified objects). Thus
they can obtain rich dynamic information in complex traffic
environments with blocked views [122]. NR C-V2X enables
sharing large volumes of sensory data with a peak data
rate of 1 Gbps or more due to the wide bandwidth in the
mmWave region [38]. An example of the benefits of OBU-
enabled cooperative perception is seen in vehicle overtaking
situations. Specifically, in such a situation a see-through
vehicular application could allow the ego vehicle to obtain
the front camera view (using NR C-V2V communication)
of the leading vehicle to help identify the non-line-of-sight
(NLOS) traffic situation ahead [121].
B. CASE STUDY: ACCIDENT WARNING
Figure 10b illustrates a neighborhood-scale accident warning
system that leverages the EPC of an LTE network. A vehicle
involved in an accident or a nearby vehicle that observed
the accident sends a notification to the local evolved NodeB
(eNB). The eNB delivers the notification data packet via the
radio bearer to the serving gateway (S-GW), which, in turn,
forwards the notification to the packet data network (PDN)
gateway (P-GW). The P-GW provides an entry point for
service providers (i.e., dedicated servers to collect informa-
tion or notifications and disseminate them to the subscribed
group). The broadcast multicast service center (BM-SC), part
of the core network, functions as the interface between the
distribution service (MBMS) and service provider (on edge
servers), thus supporting evolved multimedia broadcast mul-
ticast services (eMBMS). The BM-SC transmits the notifica-
tion as broadcast or multicast content through the eMBMS
gateway (MBMS-GW) to the eNBs using IP multicast and
then to the subscribed vehicles in each eNB cell.
In addition to the traditional unicast transmission, the
eMBMS can broadcast to all the users, or multicast to a pre-
16 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
Roundabout Warning
(a) Local-scale roundabout warning of danger ahead using V2X and RSU [120]
eNB
eNB
EPC
Edge Servers
MME BM-SC
MBMS
GW P-GW
S-GW
(b) Broadcast delivery of accident warning using LTE eMBMS.
Cause
Urban Event Urban Dynamic Urban Data
Impact
Record
Indicate
Data Driven Detection
Trajectory
Traffic Speed
Traffic Congestion
(c) City-scale data-driven event (congestion) detection [114]
Figure 10: Real-world case studies at local, neighborhood, and city scales.
determined set of users (drivers, passengers, or other users)
in a cell using a single eNB [123], or in adjacent cells using
multiple eNBs, i.e., a multimedia broadcast multicast service
single frequency network (MBSFN). Alternatively, the road-
side system could also recognize the accident and forward
warnings to the RSUs in affected local areas. Those RSUs
could then disseminate the warnings to vehicles directly
through C-V2X (I2V) or indirectly through LTE, depending
on the vehicle communication module.
C. CASE STUDY: URBAN PLANNING
Smart devices, roadside sensors, and various kinds of road
users collect traffic data city-wide in real-time and form a
large-scale, cross-domain and multi-view data ecosystem.
Such large-scale urban data enables data-driven intelligence
to detect, analyze and predict large urban events. The early
detection or prediction of such large events (e.g., sporting
and entertainment events, protests, weather, or natural phe-
nomenon) allows governments to take timely actions.
Figure 10c illustrates the connection between the physical
space (urban event) and the cyberspace (urban data) and data-
driven detection of congestion. Urban events are a causal
factor in urban dynamics which reflects in urban data. Like-
wise, urban dynamics can be inferred from spatio-temporal
urban data and urban dynamics reveal the underlying urban
events. The collection of urban spatio-temporal trajectory
data in real-time can help calculate the city-wise traffic flow
and localize local areas of congestion [114]. The detection
of congestion allows the intelligent transport system (ITS)
to adjust the phases of traffic light signals dynamically and
change routes on users’ route recommendation applications
(vehicular application in Figure 9) to alleviate the congestion.
For instance, an ITS might arrange more taxis to the area near
a soccer match and recommend unrelated vehicles alternative
routes that bypass the area.
VIII. CONCLUSION AND FUTURE DIRECTIONS
Intelligent transportation systems are set to become a major
part of future human mobility. A crucial part of ITS is vehicu-
lar networks which are spatially distributed and communicate
VOLUME 4, 2021 17
Distributed Vehicular Computing at the Dawn of 5G
directly through short-range or sidelink communications or
indirectly through network infrastructure or roadside units.
These road users (e.g., vehicles) share information and of-
fload computation tasks to nodes close to the vehicles (i.e.,
edge computing), or remote (i.e., vehicular cloud computing
and cloud computing).
In this work, we surveyed the existing literature on dis-
tributed vehicular communication and computation. Specifi-
cally, we highlighted several vehicular network applications
(e.g., basic safety and advanced efficiency ) including their
technical requirements from different viewpoints. We then
detailed the enabling technologies and promising methods
and architecture to support these applications. In detail, we
described briefly the available communication technologies
including DSRC, ITS-G5, and LTE, and comprehensively 5G
and NR C-V2X. Next, we reviewed the computation and data
analytics frameworks applicable to a vehicular network con-
text including general architectures such as cloud and edge
computing and more specific approaches such as vehicular
fog computing and mobile vehicular cloudlets.
Vehicular communications technologies, including V2N/Uu
communication to 5G network infrastructure, and V2X/PC5
direct (sidelink) communications, are maturing and will
jointly support a plethora of vehicular contexts (such as
varying transmission distance, latency, and vehicle density).
Leveraging these communication technologies, edge com-
puting and vehicular cloud computing combined with data
analytics and streaming technologies (e.g., Apache Spark
and Apache Kafka) will enable applications with varying la-
tency windows, ranging from highly latency-sensitive (basic
safety) to latency-tolerant (urban event prediction/detection).
This work provided a broad survey of vehicular commu-
nication and computing. However, many interesting topics in
this area remain active research targets including, for exam-
ple, 1) developing programmable communication links (such
as SDN) that enable on-demand bandwidth and eliminate net-
work bottlenecks; 2) integrating security at the node, domain,
end-to-end, and service levels; and using privacy-preserving
machine learning methods such as federated learning; 3)
developing and applying ML and AI methods that adapt to
dynamic and changing driving patterns; 4) using augmented
reality to improve the quality of experience of ITS, and 5) as-
sessing the interplay between vehicular communications and
future 6G technologies and standards. We briefly describe
each of these aforementioned active research targets.
A. MULTI-CARRIER SELECTION AND AGGREGATION
FOR LATENCY AND CONGESTION ISSUES
The potential for communication network congestion and the
associated higher latency (due to queuing) are significant
issues for latency-sensitive vehicular applications.
Possible methods to help with these issues include multi-
carrier network access and software-defined heterogeneous
vehicular networking architecture (e.g., SDN-based Het-
VNet [124]). Multicarrier network access is a network se-
lection technology that senses and selects the best of several
available networks (e.g., DSRC or 5G) given these networks’
properties and current congestion levels. A study on mobile
latency on multiple operators in two distinct cities [125]
shows that such a carrier selection algorithm drops latencies
10 to 20% compared to single carrier operations in real-
time interactive cloud-based mobile applications such as
augmented reality and cloud gaming. The technology also
provides the potential for aggregation of several networks to
increased bandwidth.
However, such solutions might not solve the congestion
issue at the next hop (edge devices) or in the core network
and might even exacerbate it. An edge-assisted Vehicular
Software Defined Networking (VSDN) architecture could
help solve this issue by installing VSDN controllers on
edge devices at the network edge. The VSDN controllers
can use continuously trained ML models to predict or de-
tect periods of congestion and links with high utilization.
They request more bandwidth upon detection of network
bottlenecks by controlling the multi-carrier algorithm which
either deactivates the current carrier and activates another one
or activates another one and aggregates multi-carriers. The
VSDN controllers also control the selection and aggregation
onboard vehicles, and also find the best route or establish
multiple connections to the destination node (e.g., cloud
server) in the backhaul network to satisfy the bandwidth
demand. Figure 11 illustrates such an AI-assisted VSDN
architecture. Vehicles and other road users can embed various
connectivity modules (DSRC, LTE-A, Wi-Fi, mmWave) and
select the most suitable RAT or combine two or more to
ensure seamless and ubiquitous connectivity. In others words,
the VSDN controller selects the communication channel with
sufficient predicted capacity.
Infrastructure Layer
(Data Plane)
CDPI (OpenFlow)
Application Layer
Control Layer
(Control Plane)
AI-assisted
VSDN
Controller
Spectrum Management
V-SDN
Application
Traffic Safety
Infotainment
API
Traffic Management
Figure 11: AI-assisted VSDN architecture.
18 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
B. PRIVACY AND SECURITY CHALLENGES
Remote vehicle diagnostics and maintenance, anomalous
driving identification, and many other ITS applications are
often based on learning from long-term data. They often re-
quire the transmission of potentially sensitive and private data
(e.g., user identification, computational capacity, and insur-
ance number ) to edge or cloud servers for data aggregation
and computation offloading. To deal with this privacy issue,
the automotive ecosystem should adopt federated learning
(FL) for use cases where privacy is a serious concern and
bandwidth is limited. Using an FL approach, users (e.g.,
vehicles, RSUs) share only locally trained model parameters
which (compared to the raw data) are more difficult to extract
insights from and are often smaller. Although FL copes with
privacy and bandwidth, FL raises several new challenges
such as model poisoning. Specifically, an attacker may poi-
son the model by sending parameters of an anomalous model.
Apart from FL, some vehicles (attackers) might initiate
man-in-the-middle attacks to capture vehicle data (to attempt
to learn about the drivers) or gain illegitimate remote control
of another vehicle. It is thus necessary to define a set of poli-
cies and install tools to provide, confidentiality, reliability,
integrity, and other security services. Vertical integration of
security for data traffic from nearby vehicles and devices is
critical at the node, domain, end-to-end, and service/use case
levels. Besides, the integrity of transmitted and stored data
is a crucial component in security provisioning. Meanwhile,
a verifiable computing scheme (e.g., data attestation [126])
for vehicular users is needed to check the correctness of
any obtained computation results from the edge servers.
Further research is needed to define the authorized users,
vulnerabilities, and potential threats, and to create a trusted
remote computing environment.
C. MODEL LEARNING AND ADAPTATION OVER TIME
Driving patterns (along with other vehicular phenomena)
are naturally dynamic and change over different timescales
according to the road type, weather conditions, vehicle con-
ditions, and driving style [127]. Thus, NN models, for ex-
ample, for the prediction of driving patterns require training
on significant heterogeneous historical data to account for
these dynamics. However, given the possibility of rare novel
events and non-stationarity, such models can benefit from
continuous learning techniques (aka incremental learning).
Specifically, these techniques allow learning from an online
stream of incoming data (without full offline retraining)
while avoiding the serious problem of forgetting previously
learned data (known as catastrophic forgetting) by, for exam-
ple, constraining how the network parameters can be updated
during learning [128]. Continuous learning approaches have
not yet reached the performance levels of offline retraining
and remain a major future research area.
Beyond continual learning, specific models are also de-
signed for dynamic phenomena, for example, Gaussian-
based dynamic probabilistic clustering (GDPC) [129]. GDPC
is a Gaussian mixture model-based unsupervised learning
algorithm processing large amounts of data and coping with
underlying dynamic phenomena (e.g., degradation). GDPC
integrates three well-known algorithms: the expectation-
maximization algorithm to estimate the model parameters,
and the Page-Hinkley test and Chernoff bound [130]. In turn,
they use multiple (heterogeneous) data sources, fuse them,
and based on which they train unsupervised or reinforcement
learning models. If employed, these algorithms provide the
model with the capability to detect the drift in the driving
patterns.
D. AUGMENTED REALITY
Augmented reality (AR) windshields is a novel research area
in both academia and industry that aims to improve driving
safety and experience by augmenting environmental objects
(e.g., roads, vehicles, obstacles, pedestrians) by overlaying
helpful information. For example, a pedestrian image could
be overlaid on the windshield at the location of an out-
of-view pedestrian moving quickly towards the road, thus
allowing the driver to be aware of the danger. However, AR
windshields present significant challenges. Specifically, high
motion-to-photon latency can cause misalignment between
virtual objects and the physical world [131], thus distracting
the driver. Therefore methods to reduce networking and
processing latency [132] including future communication
technologies and computing paradigms as well as HCI meth-
ods to compensate for some degree of misalignment are
important future research topics. Additionally, issues with the
alignment of the driver’s head with the windshield can cause
other HCI issues. Finally, problems such as selecting which
and how many objects to show or emphasize to the driver
remain open.
E. 6G VISION
As mentioned, the widespread deployment of 5G vehicu-
lar communications will deliver, for example, higher data
rates, lower latency, and more reliability to help support a
variety of intelligent transportation systems. Research into
5G and 5G-advanced enabled ITS will continue to be a
major research area with actual wide-scale deployment of
such systems still years away. However, some future vehic-
ular applications such as tactile internet use cases (where,
for example, a remote operator would take control of an
autonomous vehicle in an emergency) will require features
5G is lacking. Specifically, tactile internet requires sub 1 ms
end-to-end latency and a combination of the several different
5G modes including ultra-reliable low latency and enhanced
mobile broadband [133]. 6G vehicular communications look
to support such applications that require multiple modes and
overall aim to improve on most 5G KPIs by a factor of ten.
Though, 6G also presents major research challenges such as
maintaining reliability even with the use of high attenuation
(yet large bandwidth) THz or optical wireless communication
technologies that are potential major 6G components. Thus
6G vehicular communication will emerge as a significant
future research topic in the coming years.
VOLUME 4, 2021 19
Distributed Vehicular Computing at the Dawn of 5G
ACKNOWLEDGMENT
This research has been supported in part by project 16214817
from the Research Grants Council of Hong Kong, and the
5GEAR and FIT projects from Academy of Finland.
References
[1] H. Hartenstein and K. Laberteaux, VANET: Vehicular Applications
and Inter-Networking Technologies, ser. Intelligent Transport Systems.
Wiley, 2009. [Online]. Available: https://books.google.com.hk/books?
id=VNbkpbIg1EoC
[2] E. B. Hamida, H. Noura, and W. Znaidi, “Security of cooperative
intelligent transport systems: Standards, threats analysis and
cryptographic countermeasures,” Electronics, vol. 4, no. 3, pp. 380–423,
2015. [Online]. Available: https://www.mdpi.com/2079-9292/4/3/380
[3] H. Moustafa and Y. Zhang, Vehicular networks: techniques, standards,
and applications. Auerbach publications, 2009.
[4] C. Jayapal and S. S. Roy, “Road traffic congestion management using
vanet,” in 2016 International Conference on Advances in Human Ma-
chine Interaction (HMI), 2016, pp. 1–7.
[5] A. Alhilal, T. Braud, X. Su, L. Al Asadi, and P. Hui, “CAD3: Edge-
facilitated Real-time Collaborative Abnormal Driving Distributed De-
tection,” in 2021 IEEE 41st International Conference on Distributed
Computing Systems (ICDCS). IEEE, 2021, pp. 718–728.
[6] S. Amini, I. Gerostathopoulos, and C. Prehofer, “Big data analytics
architecture for real-time traffic control,” in 2017 5th IEEE International
Conference on Models and Technologies for Intelligent Transportation
Systems (MT-ITS), 2017, pp. 710–715.
[7] P. Mell, T. Grance et al., “The nist definition of cloud computing,” 2011.
[8] K. Zhang, Y. Mao, S. Leng, Y. He, and Y. Zhang, “Mobile-edge com-
puting for vehicular networks: A promising network paradigm with
predictive off-loading,IEEE Vehicular Technology Magazine, vol. 12,
no. 2, pp. 36–44, 2017.
[9] M. Satyanarayanan, “The emergence of edge computing,” Computer,
vol. 50, no. 1, pp. 30–39, 2017.
[10] J. Zhang and K. B. Letaief, “Mobile edge intelligence and computing for
the internet of vehicles,” Proceedings of the IEEE, vol. 108, no. 2, pp.
246–261, 2020.
[11] Z. W. Lamb and D. P. Agrawal, “Analysis of mobile edge computing for
vehicular networks,” Sensors, vol. 19, no. 6, p. 1303, 2019.
[12] K. Golestan, R. Soua, F. Karray, and M. S. Kamel, “Situation
awareness within the context of connected cars: A comprehensive
review and recent trends,Information Fusion, vol. 29, pp. 68–83, 2016.
[Online]. Available: https://www.sciencedirect.com/science/article/pii/
S1566253515000743
[13] N. Nwiabu, I. Allison, P. Holt, P. Lowit, and B. Oyeneyin, “Situation
awareness in context-aware case-based decision support,” in 2011 IEEE
International Multi-Disciplinary Conference on Cognitive Methods in
Situation Awareness and Decision Support (CogSIMA). IEEE, 2011,
pp. 9–16.
[14] European Road Safety Observatory, “Advanced driver assistance sys-
tems,” 2016.
[15] V. K. Kukkala, J. Tunnell, S. Pasricha, and T. Bradley, “Advanced
driver-assistance systems: A path toward autonomous vehicles,” IEEE
Consumer Electronics Magazine, vol. 7, no. 5, pp. 18–25, 2018.
[16] F. D. Da Cunha, A. Boukerche, L. Villas, A. C. Viana, and A. A. Loureiro,
“Data communication in vanets: Survey, applications and challenges,” Ad
Hoc Networks, 2016.
[17] N. Lu, N. Cheng, N. Zhang, X. Shen, and J. W. Mark, “Connected
vehicles: Solutions and challenges,” IEEE internet of things journal,
vol. 1, no. 4, pp. 289–299, 2014.
[18] P. Zhou, T. Braud, A. Alhilal, P. Hui, and J. Kangasharju, “Erl: Edge
based reinforcement learning for optimized urban traffic light control,”
in 2019 IEEE International Conference on Pervasive Computing and
Communications Workshops (PerCom Workshops). IEEE, 2019, pp.
849–854.
[19] G. Araniti, C. Campolo, M. Condoluci, A. Iera, and A. Molinaro, “Lte
for vehicular networking: a survey,” IEEE communications magazine,
vol. 51, no. 5, pp. 148–157, 2013.
[20] A. I. Pérez-Neira and M. R. Campalans, “Chapter 7 - different views
of delay in resource allocation for wireless systems11this chapter is
co-authored by dr. nizar zorba.” in Cross-Layer Resource Allocation in
Wireless Communications, A. I. Pérez-Neira and M. R. Campalans, Eds.
Oxford: Academic Press, 2009, pp. 125 – 149. [Online]. Available: http:
//www.sciencedirect.com/science/article/pii/B9780123741417000075
[21] D. A. Popescu, N. Zilberman, and A. W. Moore, “Characterizing the
impact of network latency on cloud-based applications’ performance,”
2017.
[22] S. Khanvilkar, F. Bashir, D. Schonfeld, and A. Khokhar, “7 - multimedia
networks and communication,” in The Electrical Engineering Handbook,
W.-K. CHEN, Ed. Burlington: Academic Press, 2005, pp. 401 –
425. [Online]. Available: http://www.sciencedirect.com/science/article/
pii/B9780121709600500335
[23] S. Wang, X. Zhang, Y. Zhang, L. Wang, J. Yang, and W. Wang, “A
survey on mobile edge networks: Convergence of computing, caching
and communications,” IEEE Access, vol. 5, pp. 6757–6779, 2017.
[24] F. Qu, Z. Wu, F.-Y. Wang, and W. Cho, “A security and privacy review
of vanets,” IEEE Transactions on Intelligent Transportation Systems,
vol. 16, no. 6, pp. 2985–2996, 2015.
[25] B. Acohido. (2019) Get ready for the next great user
interface: Connected cars. [Online]. Available: https://blog.avast.com/
get-ready-for-connected-cars
[26] J. B. Kenney, “Dedicated short-range communications (DSRC) standards
in the United States,” Proceedings of the IEEE, vol. 99, no. 7, pp. 1162–
1182, 2011.
[27] A. Festag, “Cooperative intelligent transport systems standards in Eu-
rope,” IEEE communications magazine, vol. 52, no. 12, pp. 166–172,
2014.
[28] European Telecommunication Standard Institute (ETSI), “Intelligent
Transport Systems (ITS); ITS-G5 Access layer specification
for Intelligent Transport Systems operating in the 5 GHz
frequency band ,” EUROPEAN STANDARD, 2010. [Online].
Available: https://www.etsi.org/deliver/etsi_en/302600_302699/302663/
01.02.00_20/en_302663v010200a.pdf
[29] IEEE WG, “Ieee standard for information technology–
telecommunications and information exchange between systems–
local and metropolitan area networks–specific requirements–part
11: Wireless lan medium access control (mac) and physical
layer (phy) specifications amendment 6: Wireless access in
vehicular environments,IEEE Std, 2010. [Online]. Available:
https://standards.ieee.org/standard/802_11p-2010.html
[30] ——, “Ieee 1609.2-2016 - ieee standard for wireless access in vehicular
environments–security services for applications and management
messages,” IEEE Std, 2016. [Online]. Available: https://standards.ieee.
org/standard/1609_2-2016.html
[31] ——, “Ieee 1609.3-2016 - ieee standard for wireless access in vehicular
environments (wave) – networking services,IEEE Std, 2016. [Online].
Available: https://standards.ieee.org/standard/1609_3-2016.html
[32] ——, “Ieee 1609.4-2016 - ieee standard for wireless access in
vehicular environments (wave) – multi-channel operation,” IEEE
Std, 2016. [Online]. Available: https://standards.ieee.org/standard/1609_
4-2016.html
[33] D. Jiang and L. Delgrossi, “fieee 802.11 p: Towards an international
standard for wireless access in vehicular environments,” in Vehicular
Technology Conference, 2008. VTC Spring 2008. IEEE. IEEE, 2008,
pp. 2036–2040.
[34] Z. MacHardy, A. Khan, K. Obana, and S. Iwashina, “V2X access
technologies: Regulation, research, and remaining challenges,” IEEE
Communications Surveys & Tutorials, vol. 20, no. 3, pp. 1858–1877,
2018.
[35] ETSI, “Intelligent transport systems (ITS); vehicular communications;
basic set of applications; definitions,” Technical Report. ETSI TR 102
6382009, 2009. [Online]. Available: https://www.etsi.org/deliver/etsi_tr/
102600_102699/102638/01.01.01_60/tr_102638v010101p.pdf
[36] D. Eckhoff, N. Sofra, and R. German, “A performance study of coop-
erative awareness in ETSI ITS G5 and IEEE WAVE,” in 2013 10th An-
nual Conference on Wireless On-demand Network Systems and Services
(WONS). IEEE, 2013, pp. 196–200.
[37] 3rd Generation Partnership Project. Long Term Evolution
(LTE) . [Online]. Available: https://www.3gpp.org/technologies/
keywords-acronyms/98-lte
[38] S. A. A. Hakeem, A. A. Hady, and H. Kim, “5G-V2X: Standardization,
architecture, use cases, network-slicing, and edge-computing,” Wireless
Networks, vol. 26, no. 8, pp. 6015–6041, 2020. [Online]. Available:
https://doi.org/10.1007/s11276-020-02419-8
[39] S. Chen, J. Hu, Y. Shi, Y. Peng, J. Fang, R. Zhao, and L. Zhao, “Vehicle-
to-Everything (V2X) Services Supported by LTE-Based Systems and
20 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
5G,” IEEE Communications Standards Magazine, vol. 1, no. 2, pp. 70–
76, 2017.
[40] C. R. Storck and F. Duarte-Figueiredo, “A 5G V2X Ecosystem Providing
Internet of Vehicles,” Sensors, vol. 19, no. 3, p. 550, 2019.
[41] 5G Americas, “White Paper: Cellular V2X Communications Towards
5G,” 5G Americas, Tech. Rep., March 2018. [Online]. Available:
https://www.5gamericas.org/wp-content/uploads/2019/07/2018_5G_
Americas_White_Paper_Cellular_V2X_Communications_Towards_
5G__Final_for_Distribution.pdf
[42] S. A. A. Shah, E. Ahmed, M. Imran, and S. Zeadally, “5G for vehicular
communications,” IEEE Communications Magazine, vol. 56, no. 1, pp.
111–117, 2018.
[43] M. Carugi, “Key features and requirements of 5G/IMT-2020 networks,
2018.
[44] S. Redana, Ö. Bulakci, A. Zafeiropoulos, A. Gavras, A. Tzanakaki,
A. Albanese, A. Kousaridas, A. Weit, B. Sayadi, B. T. Jou et al., “5G PPP
architecture working group: View on 5G architecture,” 5G PPP, Tech.
Rep., 2019.
[45] K. Ansari, “Joint use of dsrc and c-v2x for v2x communications in the
5.9 ghz its band,” IET Intelligent Transport Systems, 2021.
[46] 5G Automotive Association (5GAA), “V2X functional and
performance test report-test procedures and results”,” 5GAA,
Tech. Rep. 5GAA P-190033, 2019. [Online]. Avail-
able: https://5gaa.org/wp-content/uploads/2018/11/5GAA_P-190033_
V2X-Functional-and-Performance-Test-Report_final-1.pdf
[47] A. Papathanassiou and A. Khoryaev, “Cellular V2X as the essential
enabler of superior global connected transportation services,” IEEE
5G Tech Focus, vol. 1, no. 2, pp. 1–2, 2017. [Online]. Available:
https://futurenetworks.ieee.org/tech-focus/june-2017/cellular-v2x
[48] 5G-PPP Association, “5G automotive vision, 5GPPP, white paper,” 10
2015. [Online]. Available: https://5g-ppp.eu/white-papers/
[49] L. Gu, D. Zeng, and S. Guo, “Vehicular cloud computing: A survey,” in
2013 IEEE Globecom Workshops (GC Wkshps). IEEE, 2013, pp. 403–
407.
[50] I. Ahmad, R. M. Noor, I. Ali, M. Imran, and A. Vasilakos, “Character-
izing the role of vehicular cloud computing in road traffic management,
International Journal of Distributed Sensor Networks, vol. 13, no. 5, pp.
1–14, 2017.
[51] J. Grover, A. Jain, S. Singhal, and A. Yadav, “Real-time VANET ap-
plications using fog computing,” in Proceedings of First International
Conference on Smart System, Innovations and Computing. Springer,
2018, pp. 683–691.
[52] K. Dolui and S. K. Datta, “Comparison of edge computing implemen-
tations: Fog computing, cloudlet and mobile edge computing,” in 2017
Global Internet of Things Summit (GIoTS). IEEE, 2017, pp. 1–6.
[53] H. Yao, C. Bai, D. Zeng, Q. Liang, and Y. Fan, “Migrate or not? exploring
virtual machine migration in roadside cloudlet-based vehicular cloud,”
Concurrency and Computation: Practice and Experience, vol. 27, no. 18,
pp. 5780–5792, 2015.
[54] E. Borcoci, “Fog computing, mobile edge computing, cloudlets-which
one?” in SoftNet Conference, 2016.
[55] S. Olariu, I. Khalil, and M. Abuelela, “Taking VANET to the clouds,”
International Journal of Pervasive Computing and Communications,
vol. 7, no. 1, pp. 7–21, 2011.
[56] Cisco, White Paper . (2015) Fog computing and the
internet of things: Extend the cloud to where the things
are. [Online]. Available: https://www.cisco.com/c/dam/en_us/solutions/
trends/iot/docs/computing-overview.pdf
[57] E. I. S. G. (ISG). (2019) Multi-access Edge Comput-
ing (MEC). [Online]. Available: https://www.etsi.org/technologies/
multi-access-edge-computing
[58] M. Satyanarayanan, Z. Chen, K. Ha, W. Hu, W. Richter, and P. Pillai,
“Cloudlets: at the leading edge of mobile-cloud convergence,” in 6th In-
ternational Conference on Mobile Computing, Applications and Services.
IEEE, 2014, pp. 1–9.
[59] S. Basagni, M. Conti, S. Giordano, and I. Stojmenovic, Mobile
Ad Hoc Networking: Cutting Edge Directions, ser. IEEE Series on
Digital & Mobile Communication. Wiley, 2013. [Online]. Available:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/9781118511305
[60] M. Satyanarayanan, V. Bahl, R. Caceres, and N. Davies, “The case for
vm-based cloudlets in mobile computing,” IEEE pervasive Computing,
2009.
[61] X. Hou, Y. Li, M. Chen, D. Wu, D. Jin, and S. Chen, “Vehicular
fog computing: A viewpoint of vehicles as the infrastructures,IEEE
Transactions on Vehicular Technology, vol. 65, no. 6, pp. 3860–3873,
2016.
[62] C. Huang, R. Lu, and K.-K. R. Choo, “Vehicular fog computing: archi-
tecture, use case, and security and forensic challenges,” IEEE Communi-
cations Magazine, vol. 55, no. 11, pp. 105–111, 2017.
[63] E. M. Tordera, X. Masip-Bruin, J. Garcia-Alminana, A. Jukan, G.-J. Ren,
J. Zhu, and J. Farré, “What is a fog node a tutorial on current concepts
towards a common definition,arXiv preprint arXiv:1611.09193, 2016.
[64] C. Wang, Y. Li, D. Jin, and S. Chen, “On the serviceability of mobile ve-
hicular cloudlets in a large-scale urban environment,IEEE Transactions
on Intelligent Transportation Systems, vol. 17, no. 10, pp. 2960–2970,
2016.
[65] P. Zhou, W. Zhang, T. Braud, P. Hui, and J. Kangasharju, “ARVE:
Augmented Reality Applications in Vehicle to Edge Networks,”
in Proceedings of the 2018 Workshop on Mobile Edge Communications,
ser. MECOMM’18. New York, NY, USA: Association for Computing
Machinery, 2018, p. 25–30. [Online]. Available: https://doi.org/10.1145/
3229556.3229564
[66] I. Hadži´
c, Y. Abe, and H. C. Woithe, “Server placement and selection for
edge computing in the epc,” IEEE Transactions on Services Computing,
vol. 12, no. 5, pp. 671–684, 2019.
[67] E. I. S. G. (ISG). (2019) Multi-access Edge Computing
(MEC); Study on MEC Support for V2X Use Cases.
[Online]. Available: https://www.etsi.org/deliver/etsi_gr/MEC/001_099/
022/02.01.01_60/gr_MEC022v020101p.pdf
[68] R. Buyya and A. Vahid Dastjerdi, “Chapter 4 - fog computing:
principles, architectures, and applications,” in Internet of Things.
Morgan Kaufmann, 2016, pp. 61–75. [Online]. Available: https:
//www.sciencedirect.com/science/article/pii/B9780128053959000046
[69] H. M. Birhanie, M. A. Messous, S.-M. Senouci, E.-H. Aglzim, and A. M.
Ahmed, “Mdp-based resource allocation scheme towards a vehicular fog
computing with energy constraints,” in 2018 IEEE Global Communica-
tions Conference (GLOBECOM), 2018, pp. 1–6.
[70] X. Xiao, X. Hou, X. Chen, C. Liu, and Y. Li, “Quantitative analysis for
capabilities of vehicular fog computing,” Information Sciences, vol. 501,
pp. 742–760, 2019.
[71] Z. Ning, J. Huang, and X. Wang, “Vehicular fog computing: Enabling
real-time traffic management for smart cities,” IEEE Wireless Communi-
cations, vol. 26, no. 1, pp. 87–93, 2019.
[72] G. I. Klas, “Fog computing and mobile edge cloud gain momentum open
fog consortium, etsi mec and cloudlets,” Google Scholar, vol. 1, no. 1,
pp. 1–13, 2015.
[73] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli, “Fog computing and
its role in the internet of things,” in Proceedings of the First Edition of
the MCC Workshop on Mobile Cloud Computing, ser. MCC ’12. New
York, NY, USA: Association for Computing Machinery, 2012, p. 13–16.
[Online]. Available: https://doi.org/10.1145/2342509.2342513
[74] J. Grover, P. Vinod, and C. Lal, Vehicular Cloud Computing for Traffic
Management and Systems. IGI Global, 2018.
[75] J. Park, V. Loia, K. Choo, and G. Yi, Advanced Multimedia and
Ubiquitous Engineering: MUE/FutureTech 2018, ser. Lecture Notes in
Electrical Engineering. Springer Singapore, 2018. [Online]. Available:
https://books.google.com.hk/books?id=DoJ8DwAAQBAJ
[76] M. Satyanarayanan, G. Lewis, E. Morris, S. Simanta, J. Boleng, and
K. Ha, “The role of cloudlets in hostile environments,IEEE Pervasive
Computing, vol. 12, no. 4, pp. 40–49, 2013.
[77] E. Koukoumidis, D. Lymberopoulos, K. Strauss, J. Liu, and D. Burger,
“Pocket cloudlets,” in ACM SIGARCH Computer Architecture News,
vol. 39, no. 1. ACM, 2011, pp. 171–184.
[78] S. Hanselman and R. Caron. (2017) Apache Kafka on HDInsight.
[Online]. Available: https://channel9.msdn.com/Shows/Azure-Friday/
Apache-Kafka-on-HDInsight
[79] Raghav Mohan . (2017) Announcing Apache
Kafka for Azure HDInsight general availabil-
ity. [Online]. Available: https://azure.microsoft.com/en-us/blog/
announcing-apache-kafka-for-azure-hdinsight-general- availability/
[80] Y.-y. Chen, Y. Lv, Z. Li, and F.-Y. Wang, “Long short-term memory
model for traffic congestion prediction with online open data,” in 2016
IEEE 19th International Conference on Intelligent Transportation Sys-
tems (ITSC). IEEE, 2016, pp. 132–137.
[81] L. MOU, P. ZHAO, H. XIE, and Y. CHEN, “T-lstm: A long short-term
memory neural network enhanced by temporal information for traffic
flow prediction,” IEEE Access, p. 99, 2019.
VOLUME 4, 2021 21
Distributed Vehicular Computing at the Dawn of 5G
[82] F. Lin, Y. Xu, Y. Yang, and H. Ma, “A spatial-temporal hybrid model
for short-term traffic prediction,” Mathematical Problems in Engineering,
vol. 2019, 2019.
[83] Z. Liu, Z. Li, K. Wu, and M. Li, “Urban traffic prediction from mobility
data using deep learning,” IEEE Network, vol. 32, no. 4, pp. 40–46, 2018.
[84] W. Wei, H. Wu, and H. Ma, “An autoencoder and lstm-based traffic flow
prediction method,” Sensors, vol. 19, no. 13, p. 2946, 2019.
[85] Y. Lv, Y. Duan, W. Kang, Z. Li, and F.-Y. Wang, “Traffic flow prediction
with big data: a deep learning approach,” IEEE Transactions on Intelli-
gent Transportation Systems, vol. 16, no. 2, pp. 865–873, 2014.
[86] L. Li, X. Su, Y. Zhang, Y. Lin, and Z. Li, “Trend modeling for traffic time
series analysis: An integrated study,IEEE Transactions on Intelligent
Transportation Systems, vol. 16, no. 6, pp. 3430–3439, 2015.
[87] A. Salas, P. Georgakis, and Y. Petalas, “Incident detection using data from
social media,” in 2017 IEEE 20th International Conference on Intelligent
Transportation Systems (ITSC), 2017, pp. 751–755.
[88] M. Allahyari, S. Pouriyeh, M. Assefi, S. Safaei, E. D. Trippe, J. B.
Gutierrez, and K. Kochut, “A brief survey of text mining: Classification,
clustering and extraction techniques,” arXiv preprint arXiv:1707.02919,
2017.
[89] E. D’Andrea, P. Ducange, B. Lazzerini, and F. Marcelloni, “Real-time
detection of traffic from twitter stream analysis,” IEEE Transactions on
Intelligent Transportation Systems, vol. 16, no. 4, pp. 2269–2283, 2015.
[90] M. Liu, K. Fu, C.-T. Lu, G. Chen, and H. Wang, “A search and sum-
mary application for traffic events detection based on twitter data,” in
Proceedings of the 22nd ACM SIGSPATIAL International Conference on
Advances in Geographic Information Systems. ACM, 2014, pp. 549–
552.
[91] J. M. Celaya-Padilla, C. E. Galván-Tejada, J. S. A. Lozano-Aguilar,
L. A. Zanella-Calzada, H. Luna-García, J. I. Galván-Tejada, N. K.
Gamboa-Rosales, A. Velez Rodriguez, and H. Gamboa-Rosales,
“"texting & driving" detection using deep convolutional neural
networks,” Applied Sciences, vol. 9, no. 15, 2019. [Online]. Available:
https://www.mdpi.com/2076-3417/9/15/2962
[92] C. Streiffer, R. Raghavendra, T. Benson, and M. Srivatsa, “Darnet: A deep
learning solution for distracted driving detection,” in Proceedings of the
18th ACM/IFIP/USENIX Middleware Conference: Industrial Track, ser.
Middleware ’17. New York, NY, USA: Association for Computing
Machinery, 2017, p. 22–28. [Online]. Available: https://doi.org/10.1145/
3154448.3154452
[93] Z. Chen, J. Yu, Y. Zhu, Y. Chen, and M. Li, “D 3: Abnormal driving
behaviors detection and identification using smartphone sensors,” in 2015
12th Annual IEEE International Conference on Sensing, Communication,
and Networking (SECON). IEEE, 2015, pp. 524–532.
[94] Y. Xie, F. Li, Y. Wu, S. Yang, and Y. Wang, “D3-guard: Acoustic-based
drowsy driving detection using smartphones,” in IEEE INFOCOM 2019 -
IEEE Conference on Computer Communications, 2019, pp. 1225–1233.
[95] G. Wang, J. Hu, Z. Li, and L. Li, “Cooperative lane changing via deep
reinforcement learning,” arXiv preprint arXiv:1906.08662, 2019.
[96] M. Zhu, Y. Wang, Z. Pu, J. Hu, X. Wang, and R. Ke, “Safe,
efficient, and comfortable velocity control based on reinforcement
learning for autonomous driving,” Transportation Research Part C:
Emerging Technologies, vol. 117, p. 102662, 2020. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S0968090X20305775
[97] Y. Dai, D. Xu, S. Maharjan, G. Qiao, and Y. Zhang, “Artificial intelligence
empowered edge computing and caching for internet of vehicles,IEEE
Wireless Communications, vol. 26, no. 3, pp. 12–18, 2019.
[98] M. Zhu, X. Wang, A. Tarko, and S. Fang, “Modeling car-following
behavior on urban expressways in shanghai: A naturalistic driving
study,Transportation Research Part C: Emerging Technologies, vol. 93,
pp. 425–445, 2018. [Online]. Available: https://www.sciencedirect.com/
science/article/pii/S0968090X18308635
[99] E. Walraven, M. T. Spaan, and B. Bakker, “Traffic flow optimization: A
reinforcement learning approach,” Engineering Applications of Artificial
Intelligence, vol. 52, pp. 203–212, 2016.
[100] M. Wu, S. Zhang, and Y. Dong, “A novel model-based driving behavior
recognition system using motion sensors,” Sensors, vol. 16, no.10, 2016.
[Online]. Available: https://www.mdpi.com/1424-8220/16/10/1746
[101] L. Li, R. A. Shults, R. R. Andridge, M. A. Yellman, H. Xiang, and
M. Zhu, “Texting/emailing while driving among high school students
in 35 states, united states, 2015,” Journal of Adolescent Health, vol. 63,
no. 6, pp. 701–708, 2018.
[102] National Highway Traffic Safety Administration. (2017) Distracted
driving. [Online]. Available: https://www.nhtsa.gov/risky- driving/
distracted-driving
[103] L. Xu, S. Li, K. Bian, T. Zhao, and W. Yan, “Sober-drive: A smartphone-
assisted drowsy driving detection system,” in 2014 International Confer-
ence on Computing, Networking and Communications (ICNC), 2014, pp.
398–402.
[104] C. Xiong, X. T. Yang, L. Zhang, M. Lee, W. Zhou, and M. Raqib,
“An integrated modeling framework for active traffic management and
its applications in the Washington, DC area,Journal of Intelligent
Transportation Systems, vol. 25, no. 6, pp. 609–625, 2021. [Online].
Available: https://doi.org/10.1080/15472450.2021.1878891
[105] H. Hartenstein and K. Laberteaux, VANET: vehicular applications and
inter-networking technologies. Wiley Online Library, 2010, vol. 1.
[106] N. Peppes, T. Alexakis, E. Adamopoulou, and K. Demestichas, “Driving
behaviour analysis using machine and deep learning methods for
continuous streams of vehicular data,” Sensors, vol. 21, no. 14, 2021.
[Online]. Available: https://www.mdpi.com/1424-8220/21/14/4704
[107] C. Yang, Y. Liu, X. Chen, W. Zhong, and S. Xie, “Efficient mobility-
aware task offloading for vehicular edge computing networks,IEEE
Access, vol. 7, pp. 26 652–26 664, 2019.
[108] P. Zhou, X. Chen, Z. Liu, T. Braud, P. Hui, and J. Kangasharju, “DRLE:
Decentralized Reinforcement Learning at the Edge for Traffic Light
Control in the IoV,” IEEE Transactions on Intelligent Transportation
Systems, vol. 22, no. 4, pp. 2262–2273, 2021.
[109] Hong Kong Applied Science and Technology Research Institute,
“ASTRI C-V2X Technology,” ASTRI, Tech. Rep. astri2016.11, 9 2018.
[Online]. Available: https://www.astri.org/wp-content/uploads/2016/11/
C-V2X-Technology.pdf
[110] HKT Limited, “C-V2X Technology Trial,” HKT Limited, Tech.
Rep. hkt20180401, 2018. [Online]. Available: https://www.ofca.gov.hk/
filemanager/ofca/en/content_669/tr201804_01.pdf
[111] M. Zhang, T. Li, H. Shi, Y. Li, P. Hui et al., “A decomposition approach
for urban anomaly detection across spatiotemporal data,” in IJCAI Inter-
national Joint Conference on Artificial Intelligence. International Joint
Conferences on Artificial Intelligence, 2019.
[112] M. Zhang, T. Li, Y. Yu, Y. Li, P. Hui, and Y. Zheng, “Urban anomaly
analytics: Description, detection and prediction,” IEEE Transactions on
Big Data, pp. 1–1, 2020.
[113] S. Chawla, Y. Zheng, and J. Hu, “Inferring the root cause in road
traffic anomalies,” in 2012 IEEE 12th International Conference on Data
Mining, 2012, pp. 141–150.
[114] C. Zhang, K. Zhang, Q. Yuan, H. Peng, Y. Zheng, T. Hanratty,
S. Wang, and J. Han, “Regions, periods, activities: Uncovering urban
dynamics via cross-modal representation learning,” in Proceedings of
the 26th International Conference on World Wide Web, ser. WWW ’17.
Republic and Canton of Geneva, CHE: International World Wide Web
Conferences Steering Committee, 2017, p. 361–370. [Online]. Available:
https://doi.org/10.1145/3038912.3052601
[115] Y. Zheng, Q. Li, Y. Chen, X. Xie, and W.-Y. Ma, “Understanding
mobility based on gps data,” in Proceedings of the 10th International
Conference on Ubiquitous Computing, ser. UbiComp ’08. New York,
NY, USA: Association for Computing Machinery, 2008, p. 312–321.
[Online]. Available: https://doi.org/10.1145/1409635.1409677
[116] M. C. Gonzalez, C. A. Hidalgo, and A.-L. Barabasi, “Understanding
individual human mobility patterns,” nature, vol. 453, no. 7196, pp. 779–
782, 2008.
[117] J. Liu, J. Wan, D. Jia, B. Zeng, D. Li, C.-H. Hsu, and H. Chen, “High-
efficiency urban traffic management in context-aware computing and 5g
communication,” IEEE Communications Magazine, vol. 55, no. 1, pp.
34–40, 2017.
[118] P. Voigt and A. Von dem Bussche, “The eu general data protection regula-
tion (GDPR),” A Practical Guide, 1st Ed., Cham: Springer International
Publishing, vol. 10, p. 3152676, 2017.
[119] Z. Du, C. Wu, T. Yoshinaga, K.-L. A. Yau, Y. Ji, and J. Li, “Federated
learning for vehicular internet of things: Recent advances and open
issues,” IEEE Open Journal of the Computer Society, vol. 1, pp. 45–61,
2020.
[120] Hong Kong Applied Science and Technology Research Institute
(ASTRI). (2021) Roundabout warning. [Online]. Available: https:
//www.astri.org/tag/press-release/page/2/
[121] K. Sakaguchi, R. Fukatsu, T. Yu, E. Fukuda, K. Mahler, R. Heath, T. Fujii,
K. Takahashi, A. Khoryaev, S. Nagata et al., “Towards mmwave V2X in
22 VOLUME 4, 2021
A. Alhilal et al.: Distributed Vehicular Computing
5G and beyond to support automated driving,IEICE Transactions on
Communications, 2020.
[122] R. Fukatsu and K. Sakaguchi, “Automated Driving with Cooperative
Perception Using Millimeter-wave V2I Communications for Safe and
Efficient Passing Through Intersections,” in 2021 IEEE 93rd Vehicular
Technology Conference (VTC2021-Spring), 2021, pp. 1–5.
[123] F. Gabin, T. Stockhammer, and E. Guttman, “Enhanced Television Ser-
vices over 3GPP eMBMS,” 2010.
[124] A. Mahmood, “Towards software defined heterogeneous vehicular net-
works for intelligent transportation systems,” in 2019 IEEE International
Conference on Pervasive Computing and Communications Workshops
(PerCom Workshops), 2019, pp. 441–442.
[125] T. Braud, T. Kämäräinen, M. Siekkinen, and P. Hui, “Multi-carrier
measurement study of mobile network latency: The tale of hong kong
and helsinki,” in 2019 15th International Conference on Mobile Ad-Hoc
and Sensor Networks (MSN), 2019, pp. 1–6.
[126] H. Chen, C. Fu, B. D. Rouhani, J. Zhao, and F. Koushanfar, “DeepAttest:
An End-to-End Attestation Framework for Deep Neural Networks,” in
2019 ACM/IEEE 46th Annual International Symposium on Computer
Architecture (ISCA), 2019, pp. 487–498.
[127] C. Marina Martinez, M. Heucke, F.-Y. Wang, B. Gao, and D. Cao, “Driv-
ing style recognition for intelligent vehicle control and advanced driver
assistance: A survey,IEEE Transactions on Intelligent Transportation
Systems, vol. 19, no. 3, pp. 666–676, 2018.
[128] Z. Mai, R. Li, J. Jeong, D. Quispe, H. Kim, and S. Sanner, “Online
continual learning in image classification: An empirical survey,Neuro-
computing, vol. 469, pp. 28–51, 2022.
[129] J. Diaz-Rozo, C. Bielza, and P. Larrañaga, “Clustering of Data Streams
With Dynamic Gaussian Mixture Models: An IoT Application in Indus-
trial Processes,” IEEE Internet of Things Journal, vol. 5, no. 5, pp. 3533–
3547, 2018.
[130] J. a. Gama, I. Žliobaitundefined, A. Bifet, M. Pechenizkiy, and
A. Bouchachia, “A survey on concept drift adaptation,” ACM
Comput. Surv., vol. 46, no. 4, mar 2014. [Online]. Available:
https://doi.org/10.1145/2523813
[131] L. Liu, H. Li, and M. Gruteser, “Edge assisted real-time object detection
for mobile augmented reality,” in The 25th Annual International
Conference on Mobile Computing and Networking, ser. MobiCom ’19.
New York, NY, USA: Association for Computing Machinery, 2019.
[Online]. Available: https://doi.org/10.1145/3300061.3300116
[132] T. Braud, F. H. Bijarbooneh, D. Chatzopoulos, and P. Hui, “Future
networking challenges: The case of mobile augmented reality,” in 2017
IEEE 37th International Conference on Distributed Computing Systems
(ICDCS), 2017, pp. 1796–1807.
[133] W. Jiang, B. Han, M. A. Habibi, and H. D. Schotten, “The road towards
6g: A comprehensive survey,” IEEE Open Journal of the Communica-
tions Society, vol. 2, pp. 334–366, 2021.
AHMAD YOUSEF ALHILAL is currently pur-
suing Ph.D. degree at HKUST-DT Systems and
Media Lab (SyMLab) in Computer Science De-
partment, Hong Kong University of Science and
Technology, Hong Kong. He received the bach-
elor’s and M.S. degree from Damascus Univer-
sity, Syria. His research interests include vehicular
communication and networking, edge computing,
mobile cloud gaming, space communication and
networking.
BENJAMIN FINLEY received the B.S. degree in
software engineering from the Milwaukee School
of Engineering, Milwaukee, WI, USA, and the
M.S. and D.S. degrees in telecommunication engi-
neering from Aalto University, Helsinki, Finland.
He is currently a postdoctoral researcher with the
Department of Computer Science, University of
Helsinki. His current research interests include big
telecom data analysis and user quality of experi-
ence.
TRISTAN BRAUD is an assistant professor at
Division of Integrative Systems and Design, Hong
Kong University of Science and Technology. He
was a postdoctoral fellow at HKUST-DT Systems
and Media Lab (SyMLab) in Computer Science
Department, Hong Kong University of Science
and Technology, Hong Kong. He got his Ph.D.
degree from Université Grenoble Alpes, France in
2016. Before that, he was an engineering student at
Grenoble INP Phelma, France, and received both
a MSC from the Politecnico di Torino, Italy, and Grenoble INP, France.
His major research interests include Augmented and Virtual Reality, with
interests in pervasive and cloud computing as well as human centered system
designs.
DONGZHE SU is a Chief engineer in Commu-
nication Technologies Group of The Hong Kong
Applied Science and Technology Research Insti-
tute. He has been leading the system architecture
in R&D of Vehicle-to-Everything (V2X) commu-
nication and application system, Connected Au-
tonomous Vehicles (CAV) System, and Internet-
of-Things (IoT). His role has been to define the
technical scope and overall system design for
ASTRI’s V2X Networking system. He received
MPhil degree in Computer Science from the Hong Kong University of
Science and Technology. He received bachelor’s degree from Huazhong
University of Science and Technology. In 2021, ASTRI launched one of the
world’s largest C-V2X public road tests in Hong Kong, covering a 14 km
route with various road environment of Hong Kong.
VOLUME 4, 2021 23
Distributed Vehicular Computing at the Dawn of 5G
PAN HUI (F’18) is a Professor of Computational
Media and Arts and Director of the Center for
Metaverse and Computational Creativity at the
Hong Kong University of Science and Technology.
He is also the Nokia Chair in Data Science at
the University of Helsinki. He received his PhD
from the Computer Laboratory at University of
Cambridge, and both his Bachelor and MPhil
degrees from the University of Hong Kong. He
was an adjunct Professor of social computing and
networking at Aalto University, Finland, and a Distinguished Scientist at
the Deutsche Telekom Laboratories (T-labs), Germany. His industrial profile
also includes his research at Intel Research Cambridge, UK and Thomson
Research Paris, France. Pan Hui is a world-leading expert in Augmented
Reality and Mobile Computing, with more than 400 research papers, 30
patents, and over 22,000 citations. He is an International Fellow of the
Royal Academy of Engineering, a member of Academia Europaea, an IEEE
Fellow, and an ACM Distinguished Scientist. He has founded and chaired
many IEEE/ACM conferences/workshops, and has served as track chair,
senior program committee member, organising committee member, and
program committee member of numerous top conferences including ACM
WWW, ACM SIGCOMM, ACM Mobisys, ACM MobiCom, ACM CoNext,
IEEE Infocom, IEEE PerCom, IEEE ICNP, IEEE ICDCS, IJCAI, AAAI,
UAI, and ICWSM. He served as an Associate Editor for IEEE Transactions
on Mobile Computing and IEEE Transactions on Cloud Computing, and
as guest editor for various journals including IEEE Journal on Selected
Areas in Communications (JSAC) and IEEE Transactions on Secure and
Dependable Computing. He also served on the IEEE Computer Society
Fellow Evaluation Committee.
24 VOLUME 4, 2021
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
Real time interactive cloud-based mobile applications such as augmented reality and cloud gaming require low and stable latency, especially in urban areas. These conditions are difficult to meet with the traditional single carrier LTE network access and consolidated server deployment in a cloud. Yet, with multiple SIM/multiple radio devices, latency can be kept under a given threshold through dynamic selection among multiple carriers and server deployment at network edge. To this end, it is necessary to understand how mobile network latency changes over time during a session with different carriers and how the server placement affects the latencies. In this paper, we present results from a measurement study of mobile network latency and jitter in 4G networks of Hong Kong and Helsinki, two very different cities in terms of population density and mobile infrastructure. Based on the results, we introduce a lightweight carrier selection algorithm that displays latencies 10 to 20% lower than single carrier operation.
Article
Full-text available
Even though vehicular ad-hoc networks (VANETs) bring tremendous benefits to society, yet they raise many challenges where the security and privacy concerns are the most critical ones. In this paper, we provide a detailed overview of the state-of-the-art security and privacy requirements in VANET. Also, a brief of the approachesthat are proposed in the literature to fulfil these requirements is given in this paper. Besides that, a classification of the various VANET attacks based on the communication system layersisprovided in this paper. In addition, the different types of VANET adversaries and attackers arepresented here.In general, this paper aims to provide a good piece of information about VANET security and privacy, in order to be used as a tool to help researchers in this field in developing secure privacy-preserving approaches for VANET.
Article
Full-text available
The effects of distracted driving are one of the main causes of deaths and injuries on U.S. roads. According to the National Highway Traffic Safety Administration (NHTSA), among the different types of distractions, the use of cellphones is highly related to car accidents, commonly known as “texting and driving”, with around 481,000 drivers distracted by their cellphones while driving, about 3450 people killed and 391,000 injured in car accidents involving distracted drivers in 2016 alone. Therefore, in this research, a novel methodology to detect distracted drivers using their cellphone is proposed. For this, a ceiling mounted wide angle camera coupled to a deep learning–convolutional neural network (CNN) are implemented to detect such distracted drivers. The CNN is constructed by the Inception V3 deep neural network, being trained to detect “texting and driving” subjects. The final CNN was trained and validated on a dataset of 85,401 images, achieving an area under the curve (AUC) of 0.891 in the training set, an AUC of 0.86 on a blind test and a sensitivity value of 0.97 on the blind test. In this research, for the first time, a CNN is used to detect the problem of texting and driving, achieving a significant performance. The proposed methodology can be incorporated into a smart infotainment car, thus helping raise drivers’ awareness of their driving habits and associated risks, thus helping to reduce careless driving and promoting safe driving practices to reduce the accident rate.
Article
Full-text available
Short-term traffic flow prediction is one of the most important issues in the field of Intelligent Transportation Systems. It plays an important role in traffic information service and traffic guidance. However, complex traffic systems are highly nonlinear and stochastic, making short-term traffic flow prediction a challenging issue. Although Long Short-Term Memory (LSTM) has a good performance in traffic flow prediction, the impact of temporal features on prediction has not been exploited by existing studies. In this paper, a temporal information enhancing LSTM (T-LSTM) is proposed to predict traffic flow of a single road section. In view of the similar characteristics of traffic flow at the same time each day, the model can improve prediction accuracy by capturing the intrinsic correlation between traffic flow and temporal information. Experimental results demonstrate that our method can effectively improve the prediction performance and obtain higher accuracy compared with other state-of-the-art methods. Furthermore, we propose a novel missing data processing technique based on T-LSTM. According to the experimental results, this technique can well restore the characteristics of original data and improve the accuracy of traffic flow prediction.
Article
Full-text available
Traffic jam has been and will remain a major problem in most cities around the world. We view this situation as a computation opportunity and propose to build the cloud-computing facilities on the top of jammed cars and other vehicles to turn the energy and other resources that otherwise would be wasted into computing power. Specifically, we define the vehicular mobile cloudlet as a group of nearby vehicular mobile devices congested in the traffic jams while connected by short-range communications. Based on local mobile cloudlets of congested vehicles and available remote cloud-computing resources, we propose and evaluate the JamCloud, a system to collect and aggregate the computation capacities of congested vehicles in the city. For this newly-conceived novel cloud system, the fundamental problems are how much computation capacity the mobile cloudlets have and what is the overall achievable performance of the whole JamCloud system. Based on the three realistic large-scale urban vehicular mobility traces, we analyze and model the vehicular mobility patterns as well as the computation capacity in both the mobile cloudlet and systemwide. Specifically, by analyzing the patterns of staying time, resident number, and incoming and outgoing of vehicles in the regions with traffic jams, we model the mobile cloudlet as a periodic non-homogeneous immigration-death process, which predicts its computational capability with accuracy above 90%. Based on the observed strong Poisson features of mobile cloudlets, we further propose a queueing network model to characterize the overall performance of JamCloud with the computing resources of multiple mobile cloudlets and remote clouds. Our study thus reveals the microscopic computation capability of local cloudlets as well as the overall and asymptotic performance of the JamCloud, which provides foundational understanding to design, such systems in practice. With the inevitably growing trend of making vehicles electric, and in particular with the forthcoming 5th generation (5G) mobile communication technology, the time has finally come to turn JamCloud into reality.
Article
Full-text available
Smart cities can effectively improve the quality of urban life. Intelligent Transportation System (ITS) is an important part of smart cities. The accurate and real-time prediction of traffic flow plays an important role in ITSs. To improve the prediction accuracy, we propose a novel traffic flow prediction method, called AutoEncoder Long Short-Term Memory (AE-LSTM) prediction method. In our method, the AutoEncoder is used to obtain the internal relationship of traffic flow by extracting the characteristics of upstream and downstream traffic flow data. Moreover, the Long Short-Term Memory (LSTM) network utilizes the acquired characteristic data and the historical data to predict complex linear traffic flow data. The experimental results show that the AE-LSTM method had higher prediction accuracy. Specifically, the Mean Relative Error (MRE) of the AE-LSTM was reduced by 0.01 compared with the previous prediction methods. In addition, AE-LSTM method also had good stability. For different stations and different dates, the prediction error and fluctuation of the AE-LSTM method was small. Furthermore, the average MRE of AE-LSTM prediction results was 0.06 for six different days.
Article
Full-text available
Recent advances in edge computing and caching have significant impacts on the developments of vehicular networks. Nevertheless, the heterogeneous requirements of on-vehicle applications and the time variability on popularity of contents bring great challenges for edge servers to efficiently utilize their resources. Moreover, the high mobility of smart vehicles adds substantial complexity in jointly optimizing edge computing and caching. Artificial intelligence (AI) can greatly enhance the cognition and intelligence of vehicular networks and thus assist in optimally allocating resources for problems with diverse, time-variant, and complex features. In this article, we propose a new architecture that can dynamically orchestrate edge computing and caching resources to improve system utility by making full use of AI-based algorithms. Then we formulate a joint edge computing and caching scheme to maximize system utility and develop a novel resource management scheme by exploiting deep reinforcement learning. Numerical results demonstrate the effectiveness of the proposed scheme.
Article
In recent years, mobile devices are equipped with increasingly advanced sensing and computing capabilities. Coupled with advancements in Deep Learning (DL), this opens up countless possibilities for meaningful applications, e.g., for medical purposes and in vehicular networks. Traditional cloud-based Machine Learning (ML) approaches require the data to be centralized in a cloud server or data center. However, this results in critical issues related to unacceptable latency and communication inefficiency. To this end, Mobile Edge Computing (MEC) has been proposed to bring intelligence closer to the edge, where data is produced. However, conventional enabling technologies for ML at mobile edge networks still require personal data to be shared with external parties, e.g., edge servers. Recently, in light of increasingly stringent data privacy legislations and growing privacy concerns, the concept of Federated Learning (FL) has been introduced. In FL, end devices use their local data to train an ML model required by the server. The end devices then send the model updates rather than raw data to the server for aggregation. FL can serve as an enabling technology in mobile edge networks since it enables the collaborative training of an ML model and also enables DL for mobile edge network optimization. However, in a large-scale and complex mobile edge network, heterogeneous devices with varying constraints are involved. This raises challenges of communication costs, resource allocation, and privacy and security in the implementation of FL at scale. In this survey, we begin with an introduction to the background and fundamentals of FL. Then, we highlight the aforementioned challenges of FL implementation and review existing solutions. Furthermore, we present the applications of FL for mobile edge network optimization. Finally, we discuss the important challenges and future research directions in FL.
Article
The Internet of Vehicles (IoV) is an emerging paradigm that is driven by recent advancements in vehicular communications and networking. Meanwhile, the capability and intelligence of vehicles are being rapidly enhanced, and this will have the potential of supporting a plethora of new exciting applications that will integrate fully autonomous vehicles, the Internet of Things (IoT), and the environment. These trends will bring about an era of intelligent IoV, which will heavily depend on communications, computing, and data analytics technologies. To store and process the massive amount of data generated by intelligent IoV, onboard processing and cloud computing will not be sufficient due to resource/power constraints and communication overhead/latency, respectively. By deploying storage and computing resources at the wireless network edge, e.g., radio access points, the edge information system (EIS), including edge caching, edge computing, and edge AI, will play a key role in the future intelligent IoV. EIS will provide not only low-latency content delivery and computation services but also localized data acquisition, aggregation, and processing. This article surveys the latest development in EIS for intelligent IoV. Key design issues, methodologies, and hardware platforms are introduced. In particular, typical use cases for intelligent vehicles are illustrated, including edge-assisted perception, mapping, and localization. In addition, various open-research problems are identified.