ACCEPTED FROM OPEN CALL

Vehicular Sensing Networks in a Smart

City: Principles, Technologies and
Applications
Jingjing Wang, Chunxiao Jiang, Kai Zhang, Tony Q. S. Quek, Yong Ren, and Lajos Hanzo

gent transportation
AbstrAct system (ITS) may be
deemed to be the most
Given the escalating population across the
globe, it has become paramount to crucial function given
construct smart cities, aiming for improving the prolif- eration of
the manage- ment of urban flows relying on vehicles and the
efficient infor- mation and communication constant expansion of
technologies (ICT). Vehicular sensing networks road networks. In recent
(VSNs) play a critical role in maintaining the decades, the total
efficient operation of smart cities. Naturally,
there are numerous challenges to be solved
before the wide-spread introduction of VSNs,
including the conception of an accu- rate
topological analysis method and a beneficial
cooperation mechanism during the process of
city-wide information sharing. Hence, in this
arti- cle, we construct a VSN-aided smart city
model and appraise a range of intelligent
applications in terms of both public services
and urban flow management. Then, the
information source selec- tion algorithm of a
complex network and a rein- forcement
learning based city information sharing
mechanism are considered, complemented by
a range of open challenges.

IntroductIon
The “smart city” concept was proposed both
for resolving bottlenecks during urbanization,
as well as for supporting sustainable urban
development. Relying on efficient information
and communica- tion technologies (ICT), a
smart city is capable of sensing, analyzing, and
integrating critical informa- tion in a city’s
operation and development, includ- ing public
security information, industrial activities, social
public services, and so on. In view of the
emergence and rapid development of the
Internet of Things (IoT) and cloud computing
[1], a smart city is beneficial in terms of
enhancing the quality, performance and
interactivity of urban services, of reducing
costs and resource wastage, and improv- ing
the government’s management capability. For
example, San Diego in the USA was designed
to become a “City of the Future” and Canada
pro- posed the “Smart Capital” concept for
Ottawa. Also, Singapore intended to construct
an “Intelli- gent Island” and constructing
“Smart Shanghai” was targeted by the
government of China.
In the construction of a smart city, an
intelli-

122 1536-1284/18/$25.00 © 2018 IEEE Wireless Communications • February

IEEE 2018
 number of vehicles in cities has increased faster than the of taxi GPS traces in Beijing was proposed,
population [2]. Therefore, ITS is benefi- cial both in terms of including an optimal information source
increasing the efficiency of traffic management, and providing selection technique. We also conceived the
compelling applications for a smart city. Based on the matur- ing theory of shar- ing genuine information in
technologies in vehicular ad hoc networks (VANETs), vehicular VSNs.
sensing networks (VSNs) may be invoked both for vehicle-to-vehicle The remainder of the article is outlined as
(V2V) com- munications and vehicle-to-infrastructure (V2I) fol-
interactions, as well as for sensing, transmitting and integrating lows. VSN architectures, key technologies
important information related to a city’s operation for preventing and applications in a smart city are introduced
traffic jams. in the following section. Optimal information
Relying on cooperation and coordination among vehicles, source selection designed for VSNs and the
diverse sources of information may be sensed, transmitted and associated critical research challenges are then
integrated in the context of VSNs. However, in view of the impact discussed. Then our technique of sharing
of different road conditions and of sudden traffic accidents, VSNs have genuine informa- tion based on reinforcement
irregularly fluctuating dynam- ic topological structures, which gives rise to learning is investigat- ed, followed by a range
serious communication challenges, often leading to a high bit error of future work ideas and our conclusions in
rate (BER) and long delays. Furthermore, a small number of malicious the final section.
vehicle nodes may report false information for the sake of unfair pri-
vate benefits, which threatens the city’s information security in VSNs. VehIculAr sensIng networks In A smArt cIty
These challenges motivate us to conceive this article on the The smart city concept may bring about dramat-
application and key technologies of VSNs in a smart city. Specifical- ic changes in our urban life style, as well as
ly, we list several intelligent smart city applications relying on the basic how smart cities are managed. VSNs play a
architecture of VSNs. More- over, based on the theory of complex critical role both in constructing an intelligent
networks, a new topology analysis method tested on a real dataset transportation

Digital Object Identifier: Jingjing Wang, Chunxiao Jiang (corresponding author), Kai Zhang and Yong Ren are with Tsinghua University; Tony Q. S. Quek is with Sin-
10.1109/MWC.2017.1600275 gapore University of Technology and Design; Lajos Hanzo is with the University of Southampton.

122 1536-1284/18/$25.00 © 2018 IEEE Wireless Communications • February

IEEE 2018
 Datab Multimedia In the V2I
architec- ture, the
Mining Intelligent life
services road-side
Urban
•Life services E-bills
infrastructures, such
Traffic and Internet
management •Remote control
services as the traffic
Smart signals,
•Traffic flow Inter •Multimedia services
•Traffic accident transportation plat-
warning Acc
ess forms, and so on, act
G as the access points
(AP) in cellular
networks, which are
beneficial in terms of
gathering either
global or local
information and then
suggest or even
poten- tially impose
certain behaviors on
a group
V2I of vehicles.
Acc
•Network access ess
•Radio broadcast
•Supervisory
V2V
•P2P file sharing
•Mobile relay
•Information

FIGURE 1. The architecture and applications of VSNs in a smart city. Specifically, the vehicles should
be equipped with sensor units, control and management units, and communication units in order
to fulfil certain tasks. In addition to the wide-spread classic sensors, such as GPS, radar, cameras, and
so on, the vehicles may also be equipped with specific sensors relying on their missions.
Moreover, the control and management units are responsible for the collaboration of each part.
The communication units are composed of multiple modules obeying various protocols, such as
IEEE 802.11, IEEE 802.15, LTE, and so on, in order to support different communication
purposes.

system (ITS), and in providing comprehensive

information services for the smart city. traffic management and long-term urban planning
become possible [3].
Vsns: ArchItectures And ApplIcAtIons In A smArt cIty Based on VSN technology, as well as the
Figure 1 shows the architecture of VSNs intelligent wireless information infrastructure and
con- structed for a smart city. Specifically, the sharing, smart and prompt sensing, delivery
vehicles glean information from a multiplicity of and processing of the data becomes feasible.
city opera- tions, and provide broadcast In the following, we focus our attention on a
services with the aid of V2V or V2I number of compelling applications of VSNs,
communication based on a certain routing which are not necessarily limited to smart
strategy. In the V2I architecture, roadside city scenarios.
infrastructures, such as traffic signals, transpor- Traffic Management: Traffic congestion is
tation platforms, etc., act as the access an acute problem for urban administrators,
points (AP) in cellular networks, which are especially in large cities. Traditional traffic
beneficial in terms of gathering global and management pol- icies, such as odd-even day
local information and then suggesting or even vehicle bans, license plate quotas, and so on,
potentially imposing certain behaviors on a which may relieve traf- fic congestion to some
group of vehicles. More- over, relying on extent, fail to take into account the road
access networks or other hot spots, network’s conditions, nor do they consider
important information can also be shared on urban population distribution and the peak/off-
the Internet, while passengers are capable of peak time factor. An effective solu- tion to
surfing the Internet and enjoying rich traffic management relies on information
multimedia services. Based on this information, interaction and coordination in VSNs, which
both efficient is capable of planning and recommending
optimal

IEEE Wireless Communications • February 1

 Key Issues Related study Brief descriptions

Non-geometry based model, I. Sen, 2008 Channel measurement and modeling was proposed for V2V scenarios.

Geometry based stochastic model (GSCM), J. Karedal, 2009 Developed a generic modeling approach for V2V channels for rural areas.

Channel A propagation model based on different link classification, having

Geometry-based efficient model (GEM), M. Boban, 2014
large-scale/small- scale signal variation.

Proposed an empirical dynamic V2V model for time-variant V2V channels relying
Dynamic wide-band directional channel model, R. He, 2015
on dynamic scatterers.

Connectivity dynamics analysis of vehicular mobility, M. Fiore, Revealed the relationship between network topology and vehicular mobility.
2008
Built an urban vehicular mobility model in order to generate accurate urban
Mobility model pattern generator CityMob, F. Martinez, 2008
mobility scenarios.
Mobility
Proposed a network mobility management protocol for VANET, having a
Network mobility (NEMO) management for VANET, Y. Chen,
reduced mobility handoff time and packet loss rate.
2010
Incorporated a more realistic obstacle-based channel model into the analysis of
Obstacle-based vehicle mobility channel mode, N. Akhtar, 2015
VANET topology characteristics.

Intersection-based Geographical Routing Protocol (IGRP), H. Saleet, Relying on a selection of road intersections, Saleet modeled the QoS routing
2011 problem as a constrained optimization.

Proposed the IOLSR for automatically obtaining the configuration that best fits
Intelligent optimized link state routing (IOLSR), J. Toutouh,
the specific characteristics of VANETs.
2012
Routing
A uni-cast and multi-path routing protocol was proposed for urban and rural
Hybrid bio-inspired bee swarm routing (HyBR), S. Bitam,
scenarios having little delay and overheads.
2013
Multihop-authenticated proxy mobile IP (MA-PMIP), S. Studied the secure handover of IP services in an asymmetric VANET relying on
Cpedes, 2013 location and traffic information.

Proposed a transmission control protocol in order to adapt both the

VAdaptive inter-vehicle communication control, C. Huang,
communication rate and the power based on the VANET.
2010
Promptly adapts to fast-changing channel conditions and fully utilizes the precious
Encounter Transfer Protocol (ETP), B. Yu, 2011
link duration as vehicles encounter.
Transmission
A prioritization scheme was proposed to exploit the limited bandwidth for cross-
Priority-Based Congestion Control Algorithm, L. Tung, 2013
traffic assistance.

Defined an adaptive congestion control algorithm applied to the message

Linear message rate integrated control (LIMERIC), G. Bansal,
rate of devices in vehicular environments.
2013
Introduced a communication security architecture for Car2X and a secure
Cryptography-based Key Distribution Protocol, H Schweppe,
transport protocol for in-vehicle communication.
2011
Data flow tracking in automotive on-board network, H. A security framework was proposed for monitoring data flows in order to
Schweppe, 2012 achieve improved security.
Security
Relying on the so-called one-way hash chain and group rekeying scheme, pre-
Signature-based authentication for DoS attacks, L. He, 2012
authenti- cation was conceived for preventing attacks.

Proposed a security authentication method based on trust evaluation, which is

Trust evaluation-based security authentication, A. Zhou, 2015
quanti- fied by recommendation trust vectors.
TABLE 1. Brief descriptions of related key technologies of VSNs. vehicles and transmit
it to the control center.
driving routes, maintaining a reasonable The control
traffic flow, and providing constructive
feedback for future urban road network
design.
As for urban traffic management, we
need global information to facilitate optimal
route deci- sions. Because each vehicle has a
limited sens- ing scope and computational
capability, vehicles equipped with specific
sensors collect information concerning the
surrounding traffic within each subdomain.
Then, traffic information is forward- ed to
gateway vehicles, whose responsibility is to
gather traffic information from the other

1 IEEE Wireless Communications • February

center analyzes the information and then recom- mends the optimal
routes. Afterward, these gate- way vehicles take charge of broadcasting
the final optimal routing messages to other vehicles.
With the aid of smart vehicle sensors and ben- eficial information
processing methods in VSNs, an intelligent urban transportation
monitoring and management mechanism can be constructed [4].
Specifically, the traffic flow prediction mechanism is capable of relieving
roads with heavy traffic and/or predicting the peak-traffic time relying
on historical observation data and real-time GPS loca- tion information.
Moreover, traffic light scheduling

1 IEEE Wireless Communications • February

 Relying on the
Fire warning Source selection region pow- er-law based
information distribu- tion of
betweenness
Selected source F centrality, we can con-
clude that only a few
Smart sensing
vehicle Fire
vehicles of the
network constitute a
bottleneck during a
communica- tion
session. It is critical for
us to pay more
attention to the key
vehicles at crossroads
Energ and transportation
y junctions, or to the
ones playing the role
of bridges than to the
Gas boundary vehicles.
FIGURE 2. Two application scenarios of VSNs in a smart city, that
Source is, fire
selection warning and smart gas meter-
region
read- ing. (The VSN aided fire warning, for example, is processed in three stages. Firstly, the
vehicles equipped with specific sensors collect the fire-warning information within each
detection zone. Sec- ondly, each sensing vehicle transmits the related information to the nearest
access point. Through
the road-side backbone network, the fire brigade control center receives the fire warning and
takes decisive action. Finally, relying on the selected source nodes, the fire warning is broadcast
within each source selection region.)

management may be invoked to control traffic

flow in order to approach the maximum network the requirement of their specific applications.
throughput. The vehicle speed management sys- As for the early warning of a potential fire, for
tem broadcasts the notification of speed exam- ple, it requires a combination of
limits, which is intended to both ensure traffic temperature sen- sors, such as IMP35Q, smoke
safety and exploit the road conditions. Also, sensors, such as MC14468, audio analysis and
parking man- agement may be conducted to essential sensors.
inform drivers of available parking bays and their Multimedia and Internet Services: The
up-to-date prices. Smart Warning: Accidents VSN can be viewed as a mobile ad hoc
are a common occurrence in daily urban life. network. Rely- ing on the emerging device-to-
An early warning can minimize the loss of device (D2D) and device-to-infrastructure (D2I)
lives and damage to property in case of traffic communication modes, as well as on cellular
accidents or other emer- gency situations, such technology, VSNs play a bridging role in the
as fire, flooding, etc. [5]. Also, it is beneficial construction of a metropoli- tan area network
for preventing secondary acci- dents. Relying (MAN), or even of a nationwide super-WiFi
on vehicles and their widely dis- tributed system. The dense yet widely distributed
locations in a city, VSNs provide an ideal vehicles of a VSN significantly increase the
replacement for traditional urban accident- num- ber of wireless relays and access points
warn- ing anytime and anywhere. Given the available for a MAN, which is capable of
information sharing mechanism of VSNs, when providing both stable multimedia services and
and where a traffic accident has happened high-rate Internet services [6]. These mobile
will be spread promptly. Timely and effective vehicles allow the intelligent MAN both to reduce
accident-warn- ing is helpful in assisting its dependence on fixed base sta- tions and
drivers’ decision-mak- ing in order to avoid improve other Internet access.
traffic jams. Furthermore, a potential fire or Reliability of Intelligent Services: In
a water/gas leakage may be perceived by addition to collecting traffic information, road
the intelligent sensor of a vehicle, which conditions and any other sensory information an
immediately raises the alarm to the city’s ITS needs, relying on sharing genuine
command-and-control center, where further information rather than bluffing and spoofing,
mea- sures will be taken. The intelligent sensors VSNs can be used for sensing and handling
rely on genuine “water, electricity and gas billing
information” by a car driving past each
property. Specifically, each vehicle may be

IEEE Wireless Communications • February 1

 a) Taxi GPS distribution in Beijing city b) Clustering coefficient

c) Betweenness centrality d) Source slection probability distribution p(s)

FIGURE 3. Complex network topology-based optimal information source selection for VSNs: a) taxi GPS distribution in Beijing city;
b)clustering coefficient; c) betweenness centrality; d) source selection probability distribution p(s).

assigned a unique ID number that is bound

to specific utility bills. Based on the relevant Environment Monitoring: Environmental pol-
facili- ties located along the road network, the lution has drawn major public attention in
vehicles are capable of reading water, gas and recent years. Moreover, fine-grained and high-
electricity meters and reporting them fidelity space-time environmental monitoring is
automatically. Hence, no one will have to beneficial in terms of both accurately predicting
worry about an extra fine for an overdue and potentially preventing environmental
payment. Another intelligent utility provided pollution and revealing the cause of pollution.
by VSNs in a smart city is the informa- tion Considering high-density of smart vehicles, VSNs
recommendation system, provided that VSNs may be capable of all-day, all- round
are reliably linked to intelligent homes and environmental monitoring [7].
per- sonal information systems. Relying on
advanced data-mining and machine-learning key technologIes of Vsns
aided analy- sis, the driver will receive As mentioned above, VSNs play a critical
recommendations from VSNs based on their role in constructing a smart city in numerous
personal preferences. ways. To make pioneering smart city
Urban Planning: Urban planning is not applications a reality, VSNs face numerous
only for policy-makers and managers, it affects open challenges. Efficient key technologies
the res- idents’ quality of life as well, since and protocols, such as the con- tinuous air-
beneficial urban planning offers convenience interface conceived for long-range and medium-
and comfort for the population. Efficient urban range (CALM) enacted by the International
planning supports the convenient movement of Organization for Standardization (ISO), and the
the population, as well as of goods and other car- to-car network architecture (C2CNet)
resources. In a nutshell, VSNs are capable of designed by the Car-to-Car Communication
collecting relevant urban informa- tion and Consortium (C2C- CC), and the IEEE 1609
feeding it back to policy-makers. Aided by VSNs, protocol, were discussed in [8] for constructing
road network layout planning and infrastruc- ture a large-scale, mobile ad hoc vehicular network.
construction becomes more efficient. Previous studies of the key tech- nologies of
VSNs can be summarized as follows.

1 IEEE Wireless Communications • February

• Wireless Channel Characteristics: Limited Generally, the topo- logical structures of
bandwidth, high Doppler shift and VSNs rely on a determinis- tic road topology conditions causing
building occlusions are the main challenges and on a stochastic vehicular topology. landslides or flooding
facing VSNs in a smart city. Specifically, the road topology is unlike- ly to as well as the
• Mobility and Dynamic Topology: Another change in the short term, once it has been construction of new
characteristic of VSNs is the unpredictable completed in a city. However, extreme roads may influence
mobility of each vehicle, which leads to weather the long-term road
a regionally inhomogeneous but globally topology. Moreover,
stable network topology. Detailed resurfac- ing roads,
analysis of the dynamic VSN topology burst gas or water-
assists the research of tele-traffic routing pipes and traffic jams
strategy. may lead to more
• Efficient Routing Strategy: Providing an effi- short-term changes of
cient tele-traffic routing strategy is beneficial the road network’s
for optimizing the allocation of bandwidth and connectivity. By
reducing the risk of network congestion. contrast, given the mul-
• Congestion Control: An appropriate tiplicity of mobile
transmis- sion congestion control scheme is vehicles in a city, the
capable of reducing communication blocking vehicular topology is
probabilities, hence improving network subject to near-
throughput. instantaneous change.
• Security Mechanism: Relying on VSNs, a large In order to
amount of personal information has to be characterize the
acquired and shared in support of smart city stochastic vehicular
services. In order to prevent malicious topology, we
informa- tion interception, a reliable security discretize the time-
mechanism has to be designed. axis into time- slots.
Table 1 provides a brief summary of the In each time slot, the
most representative previous studies on key vehicular topology is
technolo- gies of VSNs. deemed to be static.
The topology
complex network topology-bAsed optImAl analysis of VSNs
plays a critical role in
InformAtIon source selectIon for Vsns characterizing the
In the subject area of complex network topol- inherent relationship
ogy, we are dealing with a graph-based between each pair
model of realistic systems exhibiting more of communication
random features than lattices, for example. In nodes in order to
order to support the above compelling design beneficial
applications of VSNs in a smart city communication
scenario, we have to con- sider a number of mechanisms,
key technologies, such as VSN topology and because an
the selection mechanisms of the different inaccurate represen-
information sources as well as their tation of a VSN’s
transmission and processing. The nodes topology results in
associ- ated with information replicas are poor per- formance.
considered as information source nodes. For Hence, it is
many networks such as VSNs, some nodes necessary to perform
generate much more traffic or exhibit more an accurate
constraints than others, resulting in an topological analysis
asymmetric tell-traffic distribution. Hence, a in order to con-
uniform/random source selection strat- egy struct an accurate
may not be optimal for network capacity model of VSNs.
improvement. Figure 2 depicts two Moreover, we are
application scenarios in a smart city, i.e., able to model the
fire-warning and smart gas meter reading, vehicular network
where identifying the optimal information topology both by
source selection mechanism is beneficial in relying on theoretical
terms of promoting information broadcast deriva- tions, as well
with the aid of the limited bandwidth as based on
resources of VSNs. statistical analysis
In this section, we commence with an and software
over- simulations.
view of the prevalent vehicular topology. Specifically,
Then, based on this we will introduce our latest Mahmoud et al. [9]
research contributions. conceived a
vehicular traffic
topology AnAlysIs of Vsns: An oVerVIew flow model and the
At the time of this writing, a variety of corresponding link
network topology analysis methods have been reliability model for
proposed in the literature for VSNs. characterizing their

IEEE Wireless Communications • February 1

mobility within a vehicular network. As a
further development in the field, Uppoor et al. Wireless VSNs support a multitude of intelli- gent
[10] designed a realistic synthet- ic applications in a smart city. Information
database trained by recording 24 hours of dissemination/sharing among vehicles is one of the
car traffic in a 400-km 2 region around the fundamental benefits of vehicular networks, which
city of Köln, Germany. Relying on this
sup- ports a diverse range of road traffic services
database, the network’s connectivity was
characterized in comparison to several other and
mobility traces commonly used in vehicular other scenarios.
network simula- tion. In [11], Ferreira et al.
described several sophisticated traffic
simulators that relied either on car-following
models or cellular automata models, such
as the Simulation of Urban MObil- ity
(SUMO), the TRansportation ANalysis and
SIMulation System (TRANSIMS), the
Devel- opment of Inter-VEhicular Reliable
Telematics (DIVERT), and so on. Relying on
realistic road topologies and a rich set of
vehicular parame- ters, these simulators are
capable of generating realistic vehicular
mobility models.
Nevertheless, both the theory-based and the
simulator-based modeling methods only con-
centrate on a basic characterization of the
road network. However, they are typically
oblivious of human factors, such as the
drivers’ habits, or of the dominant direction
of population move- ment at certain periods of
the working day, even though these have a
substantial influence both on the network
topology as well as on its dynam- ic evolution.
In contrast to the above-mentioned theory-
based and simulator-based modeling meth- ods,
vehicular network models relying on a statisti-
cal database may be characteristic of the practical
situation. However, a region-specific database
is not universally suitable for all regions and
cities. Having said this, it is impractical to
construct a separate database for every city.
Accordingly, we have to explore a vehicular
network model- ing method capable of
characterizing both the objective conditions
and the subjective human factors, which is of
pivotal significance during the design of VSNs,
optimization and management in a smart city.

IEEE Wireless Communications • February 1

 a) A nine-block separation of Beijing city b) Outage probability of suburban scenario in Region 7

c)Outage probability of downtown scenario in Region 5 d) Dynamic reputation of all vehicles

FIGURE 4. Simulation results of the mechanism of sharing genuine information: a) A nine-block separation of Beijing city; b)
Outage probability of suburban scenario in Region 7; c) Outage probability of downtown scenario in Region 5; d) Dynamic
reputation of all vehicles.

optImAl InformAtIon source selectIon: communication radius of r

= 500m and considered a
bAsed on A complex network perspectIVe reduced-size subregion
In this subsection, we focus our attention on
the topological analysis of VSNs in terms of
a com- plex network theory perspective.
Based on this, an optimal information source
selection algorithm is proposed. First, we
constructed a vehicular net- work model based
on a real-world dataset [12], which was
recorded by storing the GPS posi- tions of
taxies in Beijing (longitude ranging from
116.25 to 116.55 and latitude ranging from
39.8 to 40.05). Figure 3a can be viewed as a
snapshot of the vehicles’ location within
Beijing city at a specific time instant, where the
taxi position distri- bution clearly reflected the
planning of the urban structure and
distinguished the downtown and suburban
areas of Beijing. Second, we included the
edges between the pair of discrete points rep-
resenting the vehicles according to the
commu- nication distances among them.
Explicitly, two vehicles were connected by an
edge, when their distance was confined within
the maximum com- munication radius r. In our
work, we adopted an empirical maximum

1 IEEE Wireless Communications • February

 determined by the longitude ranging from 116.32 to 116.39 and
latitude ranging from 39.96 to 39.985. Below, further complex-
network based topological discussions are presented.
Again, a complex network constituted by a number of agents,
which typically exhibits the small-world phenomenon, indicates
that each agent may be reached by any another agent within six
hops. The VSN is established relying on the physical
communication distance, which is different from the fundamental
characteristics of social networks. Hence, we use the concept of
“local small-world property” proposed in [13] to describe VSNs. In the
following, we first deter- mined several complex network parameters
rely- ing on our unweighed and undirected vehicular network.
Clustering Coefficient: The ability of neighbors to communicate
with each other is characterized by the clustering coefficient, which
reflects the tightness of the network. Figure 3b shows the clus- tering
coefficient of each vehicle in a descending order. The average value
of the clustering coeffi- cients is C = 0.7725, which physically means
that the vehicular network has a close-knit community structure.

1 IEEE Wireless Communications • February

 Betweenness Centrality: Generally, a problem [14]. Figure 3d presents the source
vehicle having a large node degree plays a selection probability distribution of each the information
critical role during communication. However, vehicle rely- ing on our topological model source vehicles are
under some cir- cumstances, a vehicle having a within a source selection region. within a limited
small degree may also act as a bridging node. Accordingly, only a few vehi- cles acted as distance,
Therefore, the associ- ated betweenness information source nodes in the vehicular corresponding to a
centrality, shown in Fig. 3c, is defined as a networks considered. In other words, limited number of
measure of importance of the node in the hops.
network, which is defined as the number of
shortest paths leading from all vertices to all cIty InformAtIon
others that pass through the specific node
consid- ered. Relying on the power-law based shArIng theory for
distribution of betweenness centrality, we can
conclude that only a few vehicles of the Vsns
network constitute a bottleneck during a Wireless VSNs
communication session. It is critical for us to support a multitude of
pay more attention to the key vehicles at intelligent
crossroads and transportation junc- tions, or applications in a
to the ones playing the role of bridges than smart city.
to the boundary vehicles. Information dissemi-
Average Path Length: Average path length nation/sharing
(APL) is proposed as the average number of among vehicles is
hops along the shortest paths for all possible one of the fun-
pairs of the network nodes. The APL of our damental benefits of
vehicular net- work is L = 6.7337. vehicular networks,
Hence, our vehicular network obeys the which supports a
local small-world property, while exhibiting diverse range of road
a high degree of clustering, but a short traffic services and
average path length associated with C = other scenarios.
0.7725 and L = 6.7337. Relying on the Returning to the pair
above characteris- tics of our complex of sce- narios shown
network, we constructed a weight matrix in Fig. 2, the
characterizing each edge in the vehicular cooperation of vehi-
network, which represented the com- cles is capable of
munication performances among vehicles ensuring an efficient
as well as between the vehicles and the spreading of city
infrastruc- ture. Specifically, the weight of information.
each edge was determined by the However, the
communication distance, the channel’s fading provision of
model, by the signal to noise ratio, and by the manipulated
the vehicles’ parameters. The weight matrix information by selfish
was synonymously referred to the vehicles may lead to
communication impedance. Therefore, a potentially
weighed and undirected topological model avoidable traffic
G was formulated as G = <V, E, W>, where jams, for example.
V, E, and W represented the set of vehicle Similarly, the meter-
nodes, the set of communication links, as reading utility may be
well as the weight matrix, respectively. decyphered by
In VSNs, the specific location of the eavesdroppers and
infor- traded during the
mation sources directly influences the com- process of city
munication efficiency achieved, and the information sharing by
communication overhead imposed, malicious vehicles. In
especially in an information broadcast the following, we
scenario. A com- pelling optimization focus our attention on
criterion for an informa- tion source the city information
selection algorithm is to maximize the dissemination/
attainable network capacity of the VSN sharing mechanisms
considered. Relying on our complex-network of VSNs.
based vehicular topology, we characterized
the source vehicles in terms of a cIty
probability distribution p(S), while assuming InformAtIon
that the des- tination vehicles were
uniformly distributed and that they were shArIng In
independently selected. Hence, the total
network capacity Rc can be formulated as a Vsns
function of the communication impedances. Vehicular
Moreover, maximizing the net- work communications and
capacity Rc was shown to be equivalent to their support
a min-max optimization problem, which can networks have
be reduced to a linear programming originally been
proposed for public

IEEE Wireless Communications • February 1

safety applications and traffic efficiency
enhancements, where the efficiency of In order to support big data based applica-
informa- tion sharing critically relies on the tions,cloud computing and storage should
cooperation and coordination of the be incorporated into our proposed VSNs. How
vehicles and on that between vehicles and to efficiently and dynamically allocate cloud
the infrastructure [15]. Cenerario et al. [16]
constructed an event-relat- ed information resources, such as computational resources,
exchange/sharing protocol for VSNs. Relying storage
on the short interactions among vehicles to resources, bandwidth resources, and
exchange/share information about relevant so on, becomes an
events, they proposed the concept of open challenge.
encounter probability, in order to decide when
an information re-diffusion becomes feasible.
Given the re-diffusion mechanism, this
informa- tion exchange/sharing protocol can be
applied in a range of vehicular network
scenarios, including the quest for parking
spaces, the broadcasting of accidents or
obstacles along the road ahead, announcing
the routing of emergency vehicles, and so on.
Further applications in a smart city
mentioned above can be implemented in terms
of the in-depth city information sharing
mecha- nisms.
Although most vehicles act in a cooperative
manner during information sharing, a small num-
ber of selfish or malicious vehicles may delib-
erately provide either random or manipulated
information for the sake of attaining an unfair
pri- ority. In order to make this behavior
more unat- tractive during information
sharing, Fadul et al.
[17]created a trust-management toolkit for
iden- tifying and mitigating the compromised
actions of the nodes. Relying on Fadul’s
work [17], a four-module information sharing
framework was conceived for constructing a
genuine information sharing mechanism, which
relies on a monitor- ing module, a reputation-
management module, a path-management
module and a trust-manage- ment module.
Monitoring Module: Each vehicle is
equipped with a monitoring module in order
to both detect misbehavior and provide
responses to cooperat- ing vehicles. When
detecting a given malicious behavior, the
monitoring module activates the reputation-
management module;

IEEE Wireless Communications • February 1

 Reputation-Management Module: The reputa- charac-
tion-management module evaluates the terized by the
vehicle’s integrity according to historical cumulative distribution
records, while relying on online learning function
algorithms. The informa- tion received from (c.d.f) of this vehicle’s
malicious vehicles whose repu- tation scores SINR. Relying on the out-
cannot reach a certain threshold will be age probability, we
forwarded to the path-management module; modeled the vehicles’
Path-Management Module: The path-man- future utility trend for
agement module is in charge of redirecting the predicting their future
information dissemination paths, relying on the information sharing
attributes of the vehicles derived from the report decisions. Furthermore, a
provided by the reputation-management module; reinforcement
Trust-Management Module: The trust-manage-
ment module relies on a threshold. Moreover, it
includes a list of both genuine and malicious
nodes.
As a further development, Khabazian et al.
[18]presented an analytical model for
character- izing the performance measures of
information sharing in VSNs under the
assumption of two pri- ority classes of traffic.
Specifically, the distribution of the number of
concurrent transmissions was discussed, and
numerical results were provided along with
simulation results in order to confirm the
accuracy of the proposed analysis. Moreover,
several other performance indicators, such as
the packet loss rate, packet transmission
distance, end-to-end reliability, throughput, and
delay were proposed for characterizing the
performance of information sharing in VSNs.

shArIng genuIne InformAtIon wIth the

AId of reInforcement leArnIng
In this subsection, we focus our attention
on designing a specific mechanism relying on
rein- forcement learning in order to ensure
that all vehicles share genuine information.
We assume that each vehicle is capable of
acquiring city infor- mation, and deciding
whether to share genuine information. It is
plausible that this leads to a gam- ing strategy,
where individual income depends on the
strategy of both sides. All vehicles aim for
maximizing their own utility. However, a
typical Prisoners’ Dilemma based game may not
benefi- cially contribute to cooperation in the
end. Also, considering that the drivers may infer
experience from previous games, we adopt
reinforcement learning for modeling this
process. Relying both on their own and on
their opponent’s reputation, the driver
makes a decision about whether to share to
cooperate genuine information. For a low-
reputation vehicle, a decision to cooperate
may be needed in order to enhance their
own reputation for the sake of gleaning an
increased utility. After each game, the vehicle
updates its estimate of the other vehicles’
reputation. Relying on the above assumption,
we invoke the outage probability of sharing
genuine information. We consider both a
Nakagami-m fading and a more specific
highway scenario of a Rayleigh fading
channel. Specifically, the outage probability
(OP) of a specific vehicle vi is defined as the
probability of its signal to interference plus
noise ratio (SINR)
dipping below a threshold of Q, which is

1 IEEE Wireless Communications • February

 learning based reputation evaluation mechanism was designed.
Similar to the well established human social networks, each vehicle
of a VSN has a reputation value depending on its past behavior, which
is also likely to govern its future behavior, when sharing information
with others. Given the above basic assumptions, the information
shar- ing process was modeled as a game, where the payoff
attained is a function both of the outage probability and the joint
strategies. Each vehicle adjusted its information sharing behavior
based on the estimation of the other vehicles’ reputa- tions and on
its accumulated perception.
Simulations were conducted to verify our theoretical analysis
and to characterize the pro- posed schemes, which are based on a
real-world dataset consisting of the spatial distribution of Beijing
taxis. However, in order to distinguish the downtown and suburban
areas, we partitioned Beijing city into nine regions, as shown in
Fig. 4a. Figures 4b and 4c portray the outage proba- bility of the
suburban and downtown scenarios,
respectively, where a is the path-loss coefficient.
The simulation results were all consistent with the
theoretical results. The outage probability of the downtown scenario
was lower than that in the suburban scenario due to the reduced
distance between a pair of vehicles as well as owing to the
benign Nakagami-m fading channels encoun-
tered. Moreover, increasing the path loss coef- ficient a from 2 to 4
can lead to an increase of the outage probability due to the
higher power
attenuation of the channel, while increasing the transmission power
reduced the outage proba- bility. Figure 4d presents the dynamic
reputation of vehicles during their learning and interaction process,
which characterized the city informa- tion sharing strategy of each
vehicle. Although the vehicles were initially configured to have low
reputations below 0.5, our regime encouraged them to share
genuine information, hence grad- ually achieving a near-unity
reputation, which corroborated the high efficiency of our city infor-
mation sharing mechanism.

future work And conclusIons

The networked cooperation within VSNs has a vast array of
beneficial applications in a smart city. A range of future research
ideas related to VSNs designed for a smart city can be summarized
as follows:
• It is necessary to construct intelligent standard- ized protocols for
future smart cities in order to integrate heterogeneous VSNs with
cellular mobile communication networks, with satellite networks
and wireline based networks.
• The network’s capacity has to be quantified and techniques of
approaching the VSN’s capacity have to be found. Given the
dynamically chang- ing network topology, agile network-capaci-
ty-approaching resource-allocation and source selection algorithms
have to be conceived for these demanding operating conditions.
• The foundation for accurately evaluating and designing VSNs
hinges on establishing more realistic mobility models for
vehicles. In com- parison to the random movement, the mobil-
ity of vehicles should follow some clear rules. Therefore, it is
essential to accurately capture the mobility attributes of VSNs.

1 IEEE Wireless Communications • February

• Given the ubiquitous restriction on energy Internetworking, Beijing, China, Sept. 2009, pp. 53–62.
con- sumption, using less energy to provide [12]J. Yuan et al., “T-Drive: Driving Directions Based on Taxi Trajectories,” Proc. 18th
SIGSPATIAL Int’l. Conf. Advances Geographic Information Systems, New York, NY, Nov. 2010,
the same service in VSNs becomes a critical pp. 99–108.
issue. It is important to consider energy-
efficient network- ing, when intelligent
vehicles cooperate with each other or with
the roadside infrastructure.
• In order to support big data based
applica- tions, cloud computing and storage
should be incorporated into our proposed
VSNs. How to efficiently and dynamically
allocate cloud resources, such as
computational resources, storage resources,
bandwidth resources, and so on, becomes an
open challenge.
• The network-security and privacy-
protection mechanisms of VSNs, such as
fingerprint authentication, data encryption,
remote attes- tation, secure payment, and
so on, have to be carefully considered, in
order to support high-integrity personal
services in smart cities. In a nutshell, several
application scenarios of
VSNs operating in a smart city have been
dis- cussed, followed by an overview of a
range of associated technologies. We also
highlighted the advantages of constructing an
accurate topolog- ical structure for VSNs.
Based on this, we pro- posed an information
source selection model in order to improve the
attainable network capacity. Finally, the pros
and cons of various information sharing solutions
were investigated. Indeed, VSNs constitute a
promising subject-area of smart city research!

Acknowledgment
This work was supported by the NSFC China
under projects 61371079 and 92338203.

references
[1]J. Jin et al., “An Information Framework for Creating a
Smart City Through Internet of Things,” IEEE Internet
Things J., vol. 1, no. 2, Apr. 2014, pp. 112–21.
[2] J. Dargay, D. Gately, and M. Sommer, “Vehicle
Ownership and Income Growth, Worldwide: 1960–
2030,” The Energy J., vol. 28, no. 4, 2007, pp. 143–70.
[3] S. Djahel et al., “A Communications-Oriented Perspective
on Traffic Management Systems for Smart Cities:
Challenges and Innovative Approaches,” IEEE Commun.
Surveys Tutori- als, vol. 17, no. 1, Mar. 2015, pp.
125–51.
[4]R. Du et al., “Effective Urban Traffic Monitoring by Vehicular
Sensor Networks,” IEEE Trans. Vehic. Technol., vol. 64,
no. 1, Apr. 2015, pp. 273–86.
[5] J. Liu, Z. Yang, and I. Stojmenovic, “Receiver
Consensus: On-Time Warning Delivery for Vehicular
Ad-Hoc Net- works,” IEEE Trans. Emerging Topics
Computing, vol. 1, no. 1, July 2013, pp. 57–68.
[6] L. Sarakis et al., “Providing Entertainment Applications
in VANET Environments,” IEEE Wireless Commun., vol.
23, no. 1, Feb. 2016, pp. 30–37.
[7]S.-C. Hu et al., “Measuring Air Quality in City Areas by
Vehic- ular Wireless Sensor Networks,” J. Systems Software,
vol. 84, no. 11, June 2011, pp. 2005–12.
[8] G. Karagiannis et al., “Vehicular Networking: A Survey
and Tutorial on Requirements, Architectures, Challenges,
Stan- dards and Solutions,” IEEE Commun. Surveys
Tutorials, vol. 13, no. 4, July 2011, pp. 584–616.
[9] M. H. Eiza and Q. Ni, “An Evolving Graph-Based
Reliable Routing Scheme for VANETs,” IEEE Trans.
Vehic. Technol., vol. 62, no. 4, Feb. 2013, pp. 1493–
1504.
[10] S. Uppoor et al., “Generation and Analysis of a Large-
Scale Urban Vehicular Mobility Dataset,” IEEE Trans.
Mobile Com- puting, vol. 13, no. 5, Feb. 2014, pp.
1061–75.
[11]M. Ferreira et al., “Stereoscopic Aerial Photography:
An Alternative to Model-Based Urban Mobility
Approaches,” Proc. 6th ACM Int’l. Workshop Vehicular

IEEE Wireless Communications • February 1

[13] X. Li and G. Chen, Chunxiao Jiang [S’09, M’13, SM’15] received his B.S. degree
“A Local-World in information engineering from Beijing University of Aeronau-
Evolving Network tics and Astronautics (Beihang University) in 2008, and the
Model,” Physica A: Ph.D. degree from Tsinghua University, Beijing in 2013, both
Statistical Mechanics with the highest honors. From 2011 to 2012 he visited
and its Applica- tions, the Signals and Information Group at the Department of
vol. 328, no. 1, Oct. Electrical & Computer Engineering, University of Maryland.
2003, pp. 274–86. From 2013 to 2016 he was a postdoc researcher with the
[14] J. Wang et al., Department of Electronic Engineering, Tsinghua University.
“Complex Network He is currently an assistant research fellow at Tsinghua
Theoretical Analysis Space Center. His research interests include the
on Information applications of game theory and queuing theory in wireless
Dissemination over communication and networking. He received the Best
Vehicular Networks,” Paper Award from IEEE GLOBECOM in 2013, the Best
Proc. 2016 IEEE Int’l. Student Paper Award from IEEE GlobalSIP in 2015, the
Conf. Commun. (ICC), Tsinghua Outstanding Postdoc Award in 2015, the
Kuala Lumpur, Beijing Distinguished Graduated Student Award, Chinese
Malaysia, May 2016. National Fellowship and Tsinghua Outstanding
[15] Y. Cao, T. Jiang, Distinguished Doctoral Dissertation in 2013.
and C. Wang,
“Cooperative Device- Kai Zhang received his B.S. degree in communication engineer-
to-De- vice ing from Beijing Jiaotong University (BJTU), Beijing, in 2016,
Communications in with the honors of excellent graduate of both Beijing city
Cellular Networks,” and BJTU. He is currently a master degree candidate in
IEEE Wireless Commun., wireless communication and networking at the Department
vol. 22, no. 3, June of Electronic Engineering, Tsinghua University. His research
2015, pp. 124–29. interests include wireless communications and networking,
[16] N. Cenerario, T. and big data analysis over social networks.
Delot, and S. Ilarri, “A
Content-Based Dis- Tony Q.S. QueK [S’98, M’08, SM’12] received the B.E. and
semination Protocol M.E. degrees in electrical and electronics engineering from
for VANETs: Exploiting the Tokyo Institute of Technology, Tokyo, Japan. At the
the Encounter Massachusetts Insti- tute of Technology, he earned the Ph.D. in
Probability,” IEEE electrical engineering and computer science. Currently, he is a
Trans. Intell. Transport. tenured associate profes- sor with the Information Systems
Syst., vol. 12, no. 3, Technology and Design Pillar at Singapore University of
July 2011, pp. 771– Technology and Design (SUTD). He is also a deputy director of
82. SUTD-ZJU IDEA. His current research areas include
[17]J. E. Fadul et al., “A heterogeneous networks, green communications, wire- less
Trust-Management security, Internet of Things, big data processing, and cognitive
Toolkit for Smart- radio. He is currently an editor for the IEEE Transactions on
Grid Protection Com- munications and an elected member of the IEEE Signal
Systems,” IEEE Trans. Processing Society SPCOM Technical Committee. He was
Power Delivery, vol. 29, honored with the 2008 Philip Yeo Prize for Outstanding
no. 4, Dec. 2014, pp. Achievement in Research, the IEEE Globecom 2010 Best Paper
1768–79. Award, the 2012 IEEE William
[18] M. Khabazian, S. R. Bennett Prize, the IEEE SPAWC 2013 Best Student Paper
Aissa, and M. Mehmet- Award, the IEEE WCSP 2014 Best Paper Award, and the
Ali, “Performance 2015 SUTD Out- standing Education Awards — Excellence in
Modeling of Message Research.
Dissemination in
Vehicular Ad Hoc
yong Ren received his B.S, M.S and Ph.D. degrees in
Networks with elec- tronic engineering from Harbin Institute of
Priority,” IEEE J. Sel. Technology, China, in 1984, 1987, and 1994,
Areas Commun., vol. 29, respectively. He worked as a post doctor in the
no. 1, Jan. 2011, pp. Department of Electronics Engineering,
61–71.

bIogrAphIes
JingJing Wang [S’14]
received a B.S. degree in
electronic infor- mation
engineering from Dalian
University of Technology
(DUT) in 2014 with the
highest honors. He is
currently working toward
his Ph.D. degree at
Complex Engineered
Systems Lab (CESL),
Tsinghua University,
Beijing. From January
2016 to March 2016 he
worked as a visiting
student with WNDS Group
at Singapore University of
Technology and Design
(SUTD) supervised by
Prof. Tony Q. S. Quek.
His research interests
include complex network
based modeling and game
theory in wireless commu-
nication and networking.
He received the Liaoning
Province Distinguished
Graduated Student Award
in 2014 and a Chinese
National Fellowship in
2010, 2011 and 2013.

IEEE Wireless Communications • February 1

 Tsinghua University, China from 1995 to 1997. He is
cur- rently a professor in the Department of Electronics Germany and the UK. Since 1986 he has been with the
Engineer- ing and the director of the Complexity School of Electronics and Computer Science, University of
Engineered Systems Lab at Tsinghua University. He Southamp- ton, UK, where he holds the chair in
holds 12 patents, and has authored or co-authored telecommunications. He has successfully supervised 111
more than 100 technical papers on the behavior of Ph.D. students, co-authored 18 John Wiley/IEEE Press
computer network, P2P networks and cognitive books on mobile radio communi- cations totalling in
networks. He has served as a reviewer for IEICE excess of 10 000 pages, published 1600+ research
Transactions on Communications, Digital Signal Processing, contributions in IEEE Xplore, acted both as TPC and General
Chinese Physics Letters, Chinese Journal of Electronics, Chi- Chair of IEEE conferences, presented keynote lectures and
nese Journal of Computer Science and Technology, and the has been awarded a number of distinctions. Currently he
Chinese Journal of Aeronautics, among others. His is directing a 60-strong academic research team, working on
current research interests include complex systems a range of research projects in the field of wireless
theory and its applications to the optimization and multimedia communications sponsored by industry, the
information sharing of the Internet, Internet of Things and Engineering and Physical Sciences Research Council
ubiquitous network, cog- nitive networks and cyber- (EPSRC) UK, the Europe- an Research Council’s Advanced
physical systems. Fellow Grant, and the Royal Society’s Wolfson Research
Merit Award. He is an enthusiastic supporter of industrial
LaJoS hanZo, FREng, FIEEE, FIET, Fellow of EURASIP, D.Sc., and academic liaison and he offers a range of industrial
received his degree in electronics in 1976 and his doctorate courses. He is also a Governor of the IEEE Vehicular
in 1983. In 2009 he was awarded an honorary doctorate by Technology Society. From 2008-2012 he was the editor-in-
the Technical University of Budapest and in 2015 by the chief of the IEEE Press and a Chaired Professor also at
University of Edinburgh. In 2016 he was admitted to the Tsinghua University, Beijing. For further information on
Hungarian Acade- my of Science. During his 40-year career in research in progress and associated publications please
telecommunications he has held various research and refer to http:// www-mobile.ecs.soton.ac.uk. Lajos has 26
academic posts in Hungary, 000+ citations and an H-index of 62.

1 IEEE Wireless Communications • February