IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1

Task Scheduling of Real-Time Traffic

Information Processing Based on Digital Twins
Yang Liu a, Qingcheng Zeng a*, Ang Yang a, Yuhui Sun b, Jing Gao b

version of a physical object or process constructed based on

Abstract — The Intelligent Transportation System under Digital real-time two-way data exchange between digital and physical
Twins can provide accurate data sources for traffic control. The present entities [4, 5]. A traffic DTs system can be conceptualized as
work focuses on the real-time information processing and task collecting traffic data in real-time from different physical
scheduling problems of the Internet of Vehicles (IoV) system based on
Virtual Reality. In actual traffic scenarios, mobile users and sensing
systems, such as sensors, traffic signals, and traffic surveillance
tasks in a Mobile Crowd Sensing system are usually distributed in cameras, to create a network replica of the system. Although the
different locations in the area. Therefore, four Polynomial Heuristic concept of DTs replicates the idea of Cyber Physical Systems,
Algorithms are proposed to solve practical task scheduling problems. the main challenge in constructing true DTs is to combine and
They are the Quality/Distance Algorithm (QDA), Task Density link data from heterogeneous data sources of physical systems,
Algorithm, Distance Balance Algorithm (DBA), and Bionic-DBA (B- to create network replicas of real-world traffic operations for
DBA). The simulation experiment analysis suggests that the DBA
algorithm takes the balance of travel distance into account and
real-time traffic management [6-8].
effectively improves task quality. The Utility Function in B-DBA and Facing high car ownership rates and complex and diverse
the Biological Heuristic Search Algorithm in Pareto Ant Colony driving scene changes, improving the adaptive ability of
Optimization play a critically important role in enhancing the overall vehicles to ensure the stability of daily driving has always been
task quality. In addition, a Transmission based on Privacy Protection a pain point in intelligent transportation. DTs technology can
(TPP) algorithm is designed to protect the attribute-based privacy empower smart transportation in various aspects to meet future
information in the traffic information transmission system. This
algorithm ensures that the real-time traffic information processing
travel needs. DTs Traffic, also known as Virtual Traffic,
system resists various attacks from malicious nodes. It has been incorporates the traffic data collected in real-time into the
verified that when the number of selfish nodes accounts for 30%, the established traffic model system to realize virtual digital
transmission efficiency of the TPP algorithm reaches 0.77. The mapping of the traffic system. Chen et al. (2021) generated and
research content has a practical reference value for providing users evaluated traffic optimization schemes through Big Data
with continuous and high-quality IoV network services. Analysis, Artificial Intelligence (AI), and traffic simulation
Index Terms — Digital Twins; Virtual Reality; real-time traffic;
techniques [9]. In the DTs transportation field, video
information processing; task scheduling surveillance, Fused Millimeter-Wave Radar, and Holographic
Perception of traffic elements, including motor vehicles, non-
I. INTRODUCTION motor vehicles, and pedestrians, are used to build a mapping
model of an entire traffic system in digital space for real-time
A s the urban population grows, so does the ratio of
vehicles to population. Every metropolis faces the problem
of effective management of traffic flow. With the increase of
analysis and tracking [10-12]. It can effectively solve traffic
problems such as waste of traffic resources, rigid signal system
functions, unpredictable traffic events, and rapid response.
traffic problems, the modernization and digitization of traffic
The DTs transportation system based on Virtual Reality (VR)
through intelligent systems has become a significant
has made a further leap based on the Internet of Vehicle (IoV).
development trend [1-3]. China's Ministry of Transport of the
The IoV is essentially a dynamic mobile communication system
People’s Republic of China planned to promote the digital
that can collect, process, analyze, and update information in the
transportation system with the digital perception of
network. It has become the cornerstone of the next generation
transportation facilities, extensive coverage of information
of intelligent traffic management systems. The traffic
networks, convenient and intelligent transportation services,
management applications based on IoVs can assist the traffic
online coordination of industrial governance, active
management department in traffic diversion to prevent traffic
technological application innovation, and robust network
congestion. Considering that Mobile Crowd-sensing is a new
security by 2025. Today, the traditional transportation industry
type of vehicle networking sensing mode, Lv et al. (2021) and
slowly transforms into an Intelligent Transportation System
Chen et al. (2021) made full use of the sensing, computing, and
(ITS). In addition, DTs technology can revolutionize traffic
communication capabilities of ubiquitous smart devices to
management and operation mode. DTs refers to a digital
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2

collect multi-dimensional information of physical traffic world B. Smart Traffic Architecture Based on VR
[13, 14]. Task scheduling is a core issue in Mobile Crowd- Taking the transportation field as an example, VR technology
sensing. Besides, designing an efficient task scheduling has been applied to traffic safety education, rail transit
algorithm is significant for improving task scheduling construction, driving training, and in-station navigation. Rail
performance and reducing system operating costs. In the transit simulation uses three-dimensional (3D) virtual and
Crowd-sensing scene represented by intelligent transportation, simulation technology to simulate the 3D environment of each
there are various types of tasks to be completed. Among them, stage and link of rail vehicles, including design, manufacture,
maximizing the ability of each user to complete one or more operation, and maintenance [18]. In this environment, users can
types of jobs has become the primary goal of Intelligent Scene devote themselves to the entire project of rail transit to conduct
Perception. various operations. This technique expands the cognitive means
DTs technology carries out ex-ante simulation evaluation on and fields of relevant practitioners, saves costs and time for the
the scheme in the traffic domain to achieve the expected whole rail transit construction project, and improves Efficiency
delivery goal and actual traffic implementation. If the goal is and Quality. Rozmus et al. (2021) [19] used 3D Scanning,
not reached, the program needs to be revised repeatedly until Computer Simulation, and Instant VR technology to visualize
the desired effect is achieved. Information processing and task the transportation environment, providing a reference for the
scheduling methods for real-time traffic are crucial in this area selection of traffic engineering. They proved that the
process. Based on this, under the background of DTs method is effectively practical for transporting large goods. Lee
technology, this research mainly focuses on the ITS of the VR- et al. (2021) [20] completed the dynamic reconstruction of 2D
based IoVs. Besides, two key issues are analyzed combined planar images through a Model-Free Haptic Rendering
with the mainstream data distribution technology and user Algorithm in an intelligent traffic environment, realizing the
application requirements: the real-time information processing user's real-time feeling of the transportation environment and
method of IoVs and the intelligent perception task scheduling the tactile interaction with the remote environment.
algorithm of traffic group to ensure accurate data sources for
traffic control. The research content reported here has a specific C. Task Scheduling in the IoV
practical reference value for offering users continuously and With the development of high technology and the advent of
high-quality IoV network services. the 5G era, Wireless Communication and IoT technologies have
proliferated. IoVs have become an essential part of the
II. RELATED WORKS information age. The IoV is a specific application of the IoT
extending to intelligent transportation, known as the IoT for
A. DTs Transportation System
vehicles [21]. It receives information from and sends
From the perspective of digital traffic, traffic can be divided information to the car on the road through the sensors, onboard
into real physical traffic and DTs traffic. A DTs intelligent units, communication modules, and other devices installed on
transportation platform can be established using DTs the vehicle. Information is transmitted to the data center
technologies, such as the Internet of Things (IoT), Big Data, and through roadside units, base stations, and other equipment. In
Augmented Reality [15]. It is of great significance to expand this way, vehicle-to-everything communication is realized, that
the existing intelligent transportation platform into a DTs is, network communication between vehicles and vehicles,
platform, use transportation models and AI to drive the virtual vehicles and people, and vehicles and infrastructure, to perceive
transportation world, and analyze and evaluate real the road network, interact data, and intelligently manage
transportation construction projects and improvement plans vehicles [22].
before they are implemented. Martínez -Gutiérrez et al. (2021) Feng et al. (2018) [23] reported that the management of the
[16] proposed a new concept of DTs based on external services vehicle through the IoVs platform could visualize the running
for the transportation system of Automatic Guided Vehicles to status of the car in real-time and monitor the safe state of cars;
meet the material needs of collaborative industrial plants plan. it could also check the location of the vehicle and perform
The authors carried out simulation experiments in two different remote dispatch and other management to improve the
scenarios by defining an industrial Ethernet platform to verify efficiency of enterprise management and improving economic
the results of the DTs traffic system. With the new industrial benefits. Faced with the intricate serial and parallel execution
revolution of digital transformation, manufacturing and scheduling process of Big Data Computing tasks in the
transportation processes can employ more intelligent and IoVs, the rational completion of resource allocation directly
autonomous systems. Almeaibed et al. (2021) [17] adopted new determines the service carrying capacity of the IoVs platform.
safety designs to improve the resilience and safety of the entire Xu et al. (2021) [24] used AI based on the powerful real-time
independent vehicle system. They also put forward a standard information collection capability of the IoVs to optimize edge
DTs framework to facilitate data collection, processing, and services to solve practical task scheduling problems better.
analysis efficiencies to enable safe autonomous systems for According to the above research, the development of
intelligent manufacturing transportation end-to-end. They intelligent transportation has become even more diversified.
found that the DTs showed good early warning performance Besides, applying DTs technology to manage the IoV platforms
when manipulating radar sensor measurements tried to cause a is a general trend. However, there are few studies on the real-
collision. time processing of traffic information and task scheduling in
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 3

DTs scenarios. To fill this research gap, a real-time traffic This layer can also conduct a situation analysis and traffic event
generation model is constructed, and corresponding task prediction according to the target. Fig. 1 presents the DTs
scheduling algorithms are proposed, aiming to promote the architecture in smart transportation scenarios.
operation efficiency of the ITS.

III. TRAFFIC INFORMATION PROCESSING AND TASK

SCHEDULING IN DTS SCENARIOS
A. DTs Construction in Smart Transportation Scenarios
The evolution of computing technology represented by
Cloud Computing, Big Data, and AI, as well as the leap of
Tandem Relay Technique represented by All-Optical Networks,
5G/6G, and Wi-Fi 6, makes people put forward higher
expectations for digital technology. The city-level 3D
reconstruction technology can genuinely restore the elements of
the entire city based on high-precision map data, including the
primary aspects such as buildings and roads and static features
such as tree greening, bus stops, traffic signs, and traffic
markings. The DTs platform collects real-time images in an
actual road monitoring system through the road test camera,
detects and senses various targets (vehicles, driving, etc.), and
then extracts the target data. After modeling, the data is
synchronously integrated into the DTs environment [25-27].
Under the large-scale, lane-level, and real-time simulation of
the entire city, traffic flow data can be visualized by color
(Traffic Heat Map). The digitization of infrastructure and multi- Fig. 1. DTs Architecture in Smart Transportation Scenarios
source heterogeneous physical perception is the base in the
actual construction of ITSs. The data is uploaded to the The traffic scene includes four elements: people, vehicles,
computing platform with twin functions for modeling, roads, and the environment. Each component is related to the
reconstruction, and accurate perception using various low- other. In the construction of DTs, in addition to modeling a
latency network combinations as channels. Thus, the structured single element, it is also necessary to construct its influence on
digital data is applied to multiple traffic businesses to realize different aspects and respond to the impact of other factors on
the functions of traffic state deduction and automatic analysis the DTs in time.
and prediction via Big Data. In innovative traffic scenarios, DTs technology emphasizes
The 3D model collected based on the high-precision map at active operation, i.e., managing the entire expressway. Data
the perception layer digitizes the road network, roads, signs, transmission can be achieved through Electronic Toll
markings, buildings, poles, etc., and matches the corresponding Collection, cameras, IoVs terminals, roadside units, cellular
feature information. Multi-source heterogeneous traffic base stations, and even satellites. Cloud or On-Site Edge
detection sensors can identify the precise location, contour, size, Computing Nodes can calculate and process data. The
speed, heading angle, and other information of various traffic expressway management department can quickly realize active
participants, as well as traffic events and traffic behaviors. safety and fine-grained control of all elements of the
Combined with many IoT sensors, data such as the status and expressway based on the perceived traffic flow, road surface,
life cycle of road surfaces, roadbeds, and equipment are weather, and accident data, such as lane-level positioning and
synchronized to the transmission layer through diverse network guidance, dynamic path scheduling, and transportation
combinations to provide the lowest possible delay to infrastructure regulation. For drivers, the traffic DTs
synchronize time-space between the physical world and the technology has a precise perception capability with an accuracy
virtual world. The data layer distributes, stores, and manages rate of 95% and a communication capability with an end-to-end
the data collected in different ways according to business needs delay of 300ms. It can send abnormal event warnings to the
[28]. The computing layer uses computing resources to fuse and driver's twin terminal in visual and sound, improving the
accurately match the life cycle data of things data with the driver's safety. Especially in bad weather (such as heavy fog or
dynamically collected data, convert physical quantities into rain) and nights, when the driver's line of sight is blocked, the
machine-recognizable language, and build organic DTs. The twin driver terminal can help them understand real-time and
functional layer is oriented to the business system. It accurate surrounding road conditions. Based on the traffic DTs
implements the twin reproduction of traffic scenes, the target technology, the system can also provide accompanying
recognition, tracking and restoration of traffic participants, and Location Based Services for car owners and push the services
intelligent analysis of target behavior and traffic status based on to car owners through WeChat, map App, etc., to improve the
automatic recognition and processing of machine language. driving experience [29, 30].
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 4

B. Real-Time Traffic Generation Model based on Virtual points of the urban road network. Here, N pairs of i, j are set in
Online Simulation the traffic road network. The actual travel demand d!i , j and the
VR technology uses computers to generate a realistic 3D
estimated travel demand d i , j of the i, j pair can be expressed as:
virtual environment and interact with it utilizing sensing
devices. Virtual environment modeling aims to obtain the 3D
data of the actual 3D environment and use the acquired 3D data
åå P( ) d
i j
r
i, j i, j (
= v r r Î Lˆ; i, j Î N (1) )
to establish a corresponding virtual environment model
according to the application’s needs. Figure 2 shows the VR-
åå P( ) d!
i j
r
i, j i, j = vr ( r Î Lˆ; i, j Î N ) (2)
based architecture of DTs intelligent transportation system. The The error between actual travel demand and estimated travel
traffic scene DTs system combining virtuality and reality can demand is calculated according to Eq. (3).
solve problems such as analyzing actual traffic problems,
remote operation and maintenance of traffic facilities, and
(
e = åå Pi ,( rj ) d!i , j - di , j r Î Lˆ; i, j Î N (3)
i j
)( )
virtual testing of uncrewed vehicles. Assume that:
In the traffic simulation system based on VR, the traffic
generation model is crucial to initialize the road network to
Zi , j = d!i , j - di , j (4)
describe the process of the random distribution of vehicles then, according to the Maximum Possible Relative Error
reaching the simulated road network. The basis of the vehicle Algorithm, there is:
in the simulation system is primarily the traffic flow e = åå Pi ,( rj ) Zi , j = 0 (5)
information collected in the natural traffic scene. Converting i j

the relevant data can restore the traffic situation at the current If the station cannot observe the flow of any road section
moment in real-time. The road network construction tool ( ) r
between the pair of i, j, then Pi , j = 0 . At this time, the value of
configures the regional road network and adapts and transforms
the road network of other data formats. Traffic flow data is the error Z i , j can be infinite. Let D represent the set di, j = 0.
increasingly available in video, lidar, and Global Position Then, according to Eq. (5), there are:
System (GPS) sensors. This phenomenon has given rise to data-
driven traffic animation techniques. These technologies can
Zi ', j ' / di ', j ' = - å åp
i , j ¹ i ', j '
i, j Zi , j / pi(',r )j ' di ', j ' = 0 (6)
reconstruct spatiotemporal data acquired from existing road Z i ', j ' / di ', j ' = Vi , j (7)
traffic flow sensors, synthesize new traffic flows from limited
The following two equations can be derived from Eq. (6) and
sample trajectories, and learn behavioral patterns and
Eq. (7):
independent features from traffic monitoring datasets to
generate traffic flow [31]. æ ö r
Vi ', j ' = - ç å å pi(, j) Zi , j + å å pi(', )j ' di , jVi , j ÷ / i( ', )j ' di ', j ' (8)
r r

è i , j ¹i ', j ' i , j ¹ i ', j ' ø

( )
Vi , j = d!i , j - di , j / di , j = d!i , j / di , j - 1 ³ -1(9)
If:
C= å å p( ) Z
i , j ¹ i ', j '
r
i, j i, j / pi(',r )j ' di ', j ' (10)

then:
Vi ', j ' £ C + vr / pi(', )j ' di ', j ' - 1 (11)
r

The more significant the proportion of the observed flow in

the total flow of a particular section, the smaller the relative
error Vi ', j ' .
The operation basis of the online traffic simulation system is
to adapt to the actual traffic network conditions. The traffic data
observation points extract traffic distribution information in the
road network. Assume that the time interval between the two
collections of the data observation site is Dt . Then, at the tn
moment, the number of vehicles observed by an edge site is dtn ,
Fig. 2. Architecture of the VR-based DTs intelligent transportation system
the set is v , the speed is v , and the acceleration is !x! . From
The GPS terminal on the vehicle is used to measure the data this, the vehicle generation model at the edge observation site
generated by the road sections with significant traffic in the in the simulation system can be described as:
traffic scene to accurately understand the characteristics of
urban traffic flow [32-34]. The entire traffic network is detected {v '
vehicle - count + i (i = 0,1,..., d )} = {v (i = 0,1,..., d )} (12)
tn i tn

as comprehensively as possible by deploying stations at critical

{x! '
vehicle - count + i (i = 0,1,..., d )} = {x! (i = 0,1,..., d )} (13)
tn i tn
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 5

{!!x
'
vehicle - count + i (i = 0,1,..., d )} = {!!x (i = 0,1,..., d )} (14)
tn i tn
detection.
The future urban traffic system will gradually develop into
Real road network traffic information comes from many automatic control with the support of 5G/6G technology. The
aspects. At the tn moment, the observation station D collects collaborative control of vehicles at traffic intersections
discussed here is vital for realizing multi-vehicle collaborative
the information about the vehicle d t( D) . Meanwhile, the
n decision-making and ensuring vehicle safety at the corner.
observation station { A, B, C} collects the data of vehicles Cellular Vehicle Network (C-V2X) supports longer
communication distance, better non-line-of-sight performance,
{d ( ) , d ( ) , d ( ) } , respectively. At the t
A
tn
B
tn
C
tn n -1 moment, the vehicle more substantial reliability, higher capacity, and better
information in the simulation system at the observation station congestion control than traditional networks [37-39]. Therefore,
cooperative vehicle control at traffic intersections is proposed
{ A, B, C} can be written as Eq. (15). based on C-V2X communication. Fig. 3 displays the prominent
(Vehicle - Set ) ! {å v tn-1
'
vehicle - count + i (i = 0,1,..., d ( tn
A, B , C )
)} (15) architecture. It can be seen that the On Board Unit (OBU), Road
Side Unit (RSU), and the communication equipment together
After the elapsed Dt time, the number of vehicles in the constitute the cooperative vehicle control architecture. RSU
{ A, B, C, D} area in the system can be expressed as: mainly collects the real-time status of vehicles in different lanes;
OBU is responsible for sending the vehicle's data to the RSU to
(Vehicle - Count )t n
= (Vehicle - Count )t
n-1
guide the vehicle to drive safely.
(16)
- å
iÎ( A, B , C ) , j = D
g i , j pi(,i ,j j ) dt( A, B,C )
n-1

count {v ''}
g= (17)
pi(,iD, D ) dt(i )

{
v '' = v ' Î ( position ( v ) Ï area ( A, B, C , D ) ) (18) }
Introducing the error coefficient µ existing in the
simulation process can identify the number of vehicles
generated at the observation site D , as shown in Eq. (9).
d t( D) - ( count {v ''}) µ (19)
n

The final model generation process of the entire real-time

traffic observation point can be regarded as a recursive process
from edge points to interior points.

C. Real-time Information Processing of IoVs via Fig.3. Cooperative Vehicle Control at Traffic Intersection based on C-V2X
Collaborative Perception Communication
Non-contact global perception of roads depends on Video
A suggested speed is given to avoid collisions based on the
Perception, Radar Perception, Laser Perception, Spatial
motion state of the convoy to prevent the convoy from colliding
Positioning (Beidou Navigation Satellite System), Microwave,
with vehicles in other lanes when the car is passing through a
Ultrasonic, and other technologies. Video Perception is based
traffic intersection in the form of a Fleet. Fig. 4 is the schematic
on camera imaging and a series of computer vision algorithms
diagram of the vehicle collision area at the traffic intersection.
to perceive the road. The video capture has many pixels, high
resolution, rich colors, and an extensive dynamic range, which
can best reflect the actual scene of the road. Various computer
vision detection, recognition, tracking, and segmentation
algorithms based on video can accurately identify the road
(pavement, markings, cracks, potholes, roadbed, building
facilities, guardrails, etc.), road participants (people, vehicles,
non-motor vehicles, etc.), road environment (water, snow, fog,
etc.), and road abnormal events. They have the incomparable
advantages of other sensors. Due to the alternation of day and
night in the real world, video imaging will change according to
changes in light. When the light is dim, the recognition effect
will drop sharply [35, 36]. Correspondingly, the visual
perception capability in the infrared or ultraviolet band is
enhanced while detecting visible light to achieve all-weather
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 6

D. Traffic Crowd-sensing Task Scheduling Algorithm

In actual traffic scenarios, mobile users and sensing tasks in
a Mobile Crowd-sensing System are usually distributed in
different locations in the area. Therefore, users need to consider
their travel distance budget to move to a specific target location.
The Crowd-sensing System is composed of three parts: service
platforms, task initiators, and mobile users U = {u1 , u2 ,..., un } .
The task initiator issues a series of dynamic tasks obeying the
Poisson distribution. The task quality function Ap j ( K )
depends on how many times the job is completed. When K is
an odd number, there are:
K
æKö
Ap j ( K ) = å ç ÷ p j r (1 - p j ) (26)
K -r

K +1 è r ø
r=
2
Other cases besides this is:
K
æKö r
ç ÷ p j (1 - p j ) (27)
Fig. 4. Vehicle Collision Area at Traffic Intersection K -r
Ap j ( K ) = å r
r = +1 è ø
K
Take Fleet 1 going straight and Fleet 2 turning left as an 2
example for analysis. The area where the two teams collide is Ap j ( K ) is a non-decreasing function of K . Among them,
AABCD . When Fleet 2 turns left, the length from the intersection
K represents the execution times of the task t j , and p j stands
to the collision area is calculated according to Eq. (20).
2 2 for a task attribute.
æ 1 ö æ1 1 ö If the arrivals of mobile users follow a Poisson Distribution,
LR1 = ç Rleft - Ls ÷ - ç Lcrossing + Ls ÷ (20)
è 2 ø è4 2 ø each user has an initial location and a destination. A user will
In Eq. (20), Ls represents the body width. complete the registration of personal information after arrival.
In the Mobile Crowd-sensing System of Fig. 5, each user ui
The time TPZ 1 of Fleet 1 to reach the collision area at the
has a travel distance Bi ; Ti represents the set of all tasks that
current speed v1 can be expressed as Eq. (21).
are not expired upon ui ’s arrival.
L + LR1
TPZ 1 = 1 (21)
v2
In Eq. (20), L1 refers to the distance from Fleet 1 to the
intersection.
The deflection angle q PZ 2 and time TPZ 2 for Fleet 2 to reach
the collision area at the current speed v2 can be written as:
1 1
Lcrossing - Ls
q PZ 2 = arcsin 4 2 (22)
Rleft
2p Rleft
L2 + q PZ 2
TPZ 2 = 360° (23)
v2
where L2 denotes the distance of Fleet 2 from the intersection.
The time TLK1 and TLK 2 of Fleets 1 and 2 to leave the Fig. 5. Distance-constrained Mobile Crowd-sensing System

collision area can be expressed as:

This paper proposes four Polynomial Heuristic Algorithms:
1 the Quality/Distance Algorithm (QDA), the Task Density
Rleft + Ls + L1 + S1
TLK 1 = 2 (24) Algorithm (TDA), Distance Balance Algorithm (DBA), and
v1 Bionic-DBA, (B-DBA). In the implementation process of QPA,
2p Rleft the core idea is to achieve the local optimum and select high-
L2 + q LK 2 quality and low-cost tasks. The utility function constructed
TLK 2 = 360° (25) accordingly can be expressed as Eq. (28).
v2
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 7

DFp j named Transmission based on Privacy Protection (TPP). Under

Utility ( t j ) = (28) this transmission strategy, the real-time traffic information
dist ( tail , j )
processing system can resist various attack strategies of
In Eq. (28), t j represents a candidate task, and dist (tail, j ) malicious nodes. In addition, the attribute-based encryption
denotes the travel distance from the end of the current path to algorithm is combined with the message transmission process
the task location. to protect the user's attribute privacy.
QPA compares the utility values of different candidate tasks The message processing module uses the Advanced
and selects the task satisfying the distance constraint with the Encryption Standard (AES) symmetric encryption algorithm to
maximum utility value is chosen as the next task. TDA has a encrypt the message after it is generated and before it is sent.
higher task density than QPA to guide the user to the task- The AES encryption routine begins by copying the 16-byte
intensive area to obtain more gains, improving the local input array into a 4x4-byte matrix called State. Next, an
searchability. To achieve this, A new utility function is defined attribute-based encryption algorithm is used to encrypt the
as: encryption algorithm’s key. Then, the node sends the encrypted
key to the local cloud. Finally, the key is encrypted again by the
Utility ( t j ) =
Weighted DFp j ( ) (29) message processing module. The result is returned to the node;
dist ( tail , j ) the node sends the encrypted message and key to the next-hop
node. In this process, three types of keys are generated: the
æ æ d 2j k ö ö
( )
Ti
1 public key, private key, and conversion key. This algorithm
Weighted DFp j = å tk çç - 2s 2 ÷÷ ÷÷ (30)
ç
2ps k =1 çè
DF × exp
uses a random b to define the transfer key. b is only visible to
è øø
the corresponding node. Only the encrypted key can be
where s stands for the standard deviation describing the
processed with the transfer key for the local cloud. The values
radius of the task's impact area.
compared between two nodes are randomly converted to
QPA nor TDA considers the effect of task selection on the
different values in each communication process based on the
distance balance. Therefore, DBA is put forward here. The
secure multi-party computation theory to protect the attribute-
utility function used to measure the benefits of a candidate task
based privacy information. The attribute-based encryption
can be expressed as Eq. (31).
algorithm can record the user's reward. The fake node does not
DFp j × ( Bi - length ( Pi ) - dist ( tail , j ) - ei , j )
a
have a legal public or private key to encrypt the report.
Utility ( t j ) = Under the TPP strategy, nodes can obtain an appropriate
dist ( tail , j )
g

amount of virtual currency by transmitting messages. If any

(31)
node in the ensemble G p discards messages, the report
In Eq. (31), a represents the weight parameter of the
distance balance, and g signifies the distance weight between message mr will not reach the local cloud ck , and all nodes in
the end task position and the candidate task position. the ensemble G p will not be rewarded. Therefore, nodes will
On this basis, to solve the Crowd-sensing task scheduling try to cooperate with the message transmission process as much
problem, the Pareto Ant Colony (PAC) algorithm is chosen as as possible to obtain sufficient virtual currency to send locally
the Object Heuristic Search algorithm, which is integrated with generated messages.
the DBA algorithm. The preference function of the PAC
algorithm can be written as Eq. (32). F. Simulation Experiment Design
DFp j
b The simulation experiment aims to evaluate the performance
æ ö
Pre ( t j ) = t tail
Î
, j ×ç
of the task scheduling algorithm proposed here. The algorithm
ç dist ( tail , j ) ÷÷
(32)
è ø is implemented by JAVA programming, and the Euler distance
is used as the travel distance. All tasks are uniformly distributed
In Eq. (32), t tail , j represents the local pheromone from the
in space; their arrival times and durations follow Poisson and
last task to the task t j ; Î and b are weight parameters. normal distributions. Two algorithms, Optimal Transmission
After all the ants have decided their respective paths, the two Allocation (OTA) and Pareto Ant Colony Optimization (PACO)
ants with the highest sum of task quality increase are selected, are selected as comparisons to evaluate the performance of the
and their corresponding global information elements increase four task scheduling algorithms reported above.
by t 0 and t 0 / 2 . ONE simulation software is adopted for simulation to discuss
the effectiveness of TPP privacy protection policy. The datasets
E. A Privacy Protection Scheme for Real-time Traffic used are INFOCOM06 and SIGCOMM09. The INFOCOM06
Information Processing dataset records the movement trajectory information of 76
Vehicles need to upload accurate event reports based on their participants carrying iMotes. The SIGCOMM09 dataset
perception to enable the traffic management server to respond contains social and encounters information for about 100
to all kinds of emergencies quickly. In this process, protecting participants with mobile phones. The K-means algorithm is
users’ privacy and defending against malicious node attacks used to calculate the location information of the local cloud.
have become vital issues [40, 41]. This study proposes a The transmission range is 10 meters; the transmission speed is
message transmission strategy based on privacy protection, 250KB. The experiment is run 50 times to obtain the average
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 8

result. The TPP scheme proposed here is compared with three A B

12
other representative methods: Prophet Routing Algorithm, B-DBA DBA PACO

Average Number of Users

TDA QDA OTA
Copy Adjustable Incentive Scheme (CAIS) algorithm, and

Average Task Quality

0.92 10

Trust Routing based on Social Similarity (TRSS) algorithm. 8

The Prophet Routing Algorithm estimates transmission 0.88 6

probability through historical encounter information between 4
nodes. The CAIS algorithm motivates selfish nodes to transmit 0.84
2
messages based on the credit incentive mechanism. The TRSS B-DBA
TDA
DBA
QDA
PACO
OTA
algorithm makes routing decisions based on the social 0.80
20 30 40 50 60 70 80 90 100
0
20 30 40 50 60 70 80 90 100
similarity of nodes. The evaluation indicators are transmission Task Duration Task Duration

efficiency, transmission delay, and transmission cost. Fig. 7. Effect of Task Duration on Algorithm Performance (A. average task
quality; B. average number of users completing each task)
IV. RESULTS AND DISCUSSION B. Evaluation of Traffic Information Processing and Privacy
A. Performance Evaluation of Task Scheduling Algorithms Protection Schemes
The performance of several proposed task scheduling A comprehensive evaluation is conducted on the
algorithms is compared in the actual scenario. Fig. 6 reveals the performance of different algorithms with the INFOCOM06 and
performance and variation trend of each algorithm after SIGCOMM09 datasets. Fig. 8 to Fig. 10 respectively presents
changing the task arrival rate. It can be seen that the task quality the transmission efficiency, transmission delay, and
of each task scheduling algorithm and the number of users transmission cost of several data transmission algorithms under
completing each task decrease accordingly with the growth of changing message life cycle. The results suggest that in the two
the task arrival rate. The algorithm performance is ranked in datasets, the average transmission efficiency of the TPP
descending order of task quality: B-DBA, DBA, PACO, TDA, algorithm is lower than that of the CAIS algorithm but
QDA, and OTA. This result indicates that DBA takes the significantly better than the Prophet Routing Algorithm and the
balance of travel distance into account, improving the quality TRSS algorithm. The reason is that the TPP algorithm considers
of tasks. The utility function in B-DBA and the Biological the user’s privacy, while the CAIS algorithm does not design a
Heuristic Search Algorithm in the PACO algorithm secure transmission mechanism during the transmission process.
significantly enhance the overall task quality. Fig. 7 provides Therefore, the TPP algorithm is improved in terms of privacy
each algorithm's performance and changing trend after protection compared with the CAIS algorithm. Accordingly, the
changing the task duration. As the task duration increases, the transmission efficiency decreases. With the extension of the
task quality of each task scheduling algorithm and the number message life cycle, the average transmission delay of various
of users completing each task also increase. Except when the data transmission algorithms increases gradually. In the TPP
task duration is less than 40s, the performance of DBA is algorithm, the message cannot be sent if the local cloud does
slightly weaker than that of the PACO algorithm. In other cases, not encrypt the transfer key. There is a slight rise in
the task quality performance of B-DBA and DBA is better than transmission delay compared with the CAIS algorithm.
the other algorithms. The reason is that in the case of low task However, compared with Prophet Routing Algorithm and
duration, there are few tasks at this time, resulting in reduced TRSS algorithm, the transmission delay of the TPP algorithm
search space for task allocation. The spatial search advantage decreases by 35% and 20%, respectively. In addition, the
of the Biological Heuristic Algorithm, PACO, is more evident changing trend of the transmission cost of the TPP algorithm is
in this case. generally consistent with the performance trend of the
transmission delay. The CAIS algorithm only motivates selfish
nodes to transmit messages. Therefore, when the life cycle of
the news is 40 hours, the transmission efficiency of the CAIS
algorithm is 5% higher than that of the TPP algorithm. On the
contrary, the transmission delay and transmission cost are
reduced by 5% and 40% compared with the TPP algorithm.
A B
1.0 1.0
Average Transmission Ratio

Average Transmission Ratio

0.8 0.8

Fig. 6. Influence of Task Arrival Rate on Algorithm Performance (A. average 0.6 0.6
task quality; B. average number of users completing each task)
0.4 0.4

0.2 0.2
TPP CAIS TRSS Prophet TPP CAIS TRSS Prophet
20 30 40 50 60 70 80 90 20 30 40 50 60 70 80 90
Message TTL (Hour) Message TTL (Hour)
Fig. 8. Comparison of the Average Transmission Efficiency of the
Algorithms under Varying Message Life Cycle (A. INFOCOM06; B.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 9

SIGCOMM09) Fig. 11. Comparison of the Average Transmission Efficiency of The

A B Algorithms When the Number of Selfish Nodes Changes (A. INFOCOM06; B.
Average Transmission Delay (Hour)

Average Transmission Delay (Hour)

9 SIGCOMM09)
TPP CAIS TRSS Prophet TPP CAIS TRSS Prophet
9 A B

Average Transmission Delay (Hour)

8

Average Transmission Delay (Hour)

8 8 8
7
7
6 7 7
6
5 6
5 6
4 4
5 5
3 3
2 2 4 4
20 30 40 50 60 70 80 90 20 30 40 50 60 70 80 90
Message TTL (Hour) Message TTL (Hour) 3
TPP CAIS TRSS Prophet
3
TPP CAIS TRSS Prophet
30% 40% 50% 60% 70% 80% 90% 100% 30% 40% 50% 60% 70% 80% 90% 100%
Fig. 9. Comparison of the average transmission delay of the algorithms Precentage of Selfish Nodes Precentage of Selfish Nodes
under changing message life cycle (A. INFOCOM06; B. SIGCOMM09)
Fig. 12. Comparison of the Average Transmission Delay of the Algorithms
A B When the Number of Selfish Nodes Changes (A. INFOCOM06; B.
60 60
TPP CAIS TRSS Prophet TPP CAIS TRSS Prophet SIGCOMM09)
Average Transmission Cost

Average Transmission Cost

50 50 A B
60 60
40 40 TPP CAIS TRSS Prophet TPP CAIS TRSS Prophet

Average Transmission Cost

Average Transmission Cost
50 50
30 30
40 40
20 20
30 30
10 10
20 20
0 0
20 30 40 50 60 70 80 90 20 30 40 50 60 70 80 90
10 10
Message TTL (Hour) Message TTL (Hour)
Fig. 10. Comparison of the Average Transmission Cost of the Algorithms 0 0
30% 40% 50% 60% 70% 80% 90% 100% 30% 40% 50% 60% 70% 80% 90% 100%
under Varying Message Life Cycle (A. INFOCOM06; B. SIGCOMM09) Precentage of Selfish Nodes Precentage of Selfish Nodes
Fig. 13. Comparison of the Average Transmission Cost of the Algorithms
Furthermore, the transmission efficiency, transmission delay, When the Number of Selfish Nodes Changes (A. INFOCOM06; B.
and transmission cost of different data transmission algorithms SIGCOMM09)
are compared under the indefinite quantity of selfish nodes. The
specific results are shown in Fig. 11~ Fig. 13. Specifically, the V. IN CONCLUSION
performance trends of different algorithms are almost the same Digital mapping is conducive to understanding the state of
with the two datasets. The TPP algorithm is better than the the physical entity on the information platform and controlling
TRSS algorithm and Prophet Routing Algorithm in terms of the predefined interface components in the physical entity. DTs
average transmission efficiency. When the number of selfish technology is a crucial technology gradually emerging to
nodes accounts for 30%, the transmission efficiency of the TPP realize the mapping of physical systems to digital models of
algorithm attains 0.77, about 25% higher than that of the TRSS information space. It is suitable for intelligent transportation to
algorithm and the Prophet Routing Algorithm, but 7% lower complete the high-precision restoration of traffic roads,
than the CAIS algorithm. Since the TRSS algorithm and providing technical support to constructing a vehicle-road co-
Prophet Routing Algorithm cannot motivate selfish nodes to simulation platform. The IoT, Cloud Computing, Internet, AI,
transmit messages, they are lower than the other three Automatic Control, Mobile Internet, and other technologies are
algorithms in terms of delay. When the selfish nodes account fully utilized in the transportation field. These high-techs
for 50%, the average transmission delay and transmission cost collect traffic information to optimize practical applications,
of the TPP algorithm is about 15% lower than those of the TRSS such as traffic management, transportation, and public travel.
algorithm. The reason is that the TPP algorithm adopts a game- This paper focuses on the VR-based DTs transportation system
based incentive mechanism, so the virtual currency paid and based on this realistic background.
obtained for data transmission between nodes can reach Nash The DTs system combining virtuality and reality for traffic
Equilibrium. scenarios is applicable to problems such as analyzing actual
A B traffic accidents, remote operation and maintenance of traffic
facilities, and virtual testing of crewless vehicles. A C-V2X
Average Transmission Ratio

Average Transmission Ratio

0.8 0.8
communication-based collaborative vehicle control at a traffic
0.7 0.7
intersection is implemented for the DTs traffic system. Besides,
0.6 0.6 four Polynomial Heuristic algorithms are proposed for task
0.5 0.5
scheduling problems. By testing different task scheduling
algorithms, the utility function in B-DBA and the Biological
0.4 0.4
TPP CAIS TRSS Prophet TPP CAIS TRSS Prophet Heuristic Search Algorithm in PACO play significantly
30% 40% 50% 60% 70% 80% 90% 100% 30% 40% 50% 60% 70% 80% 90% 100% improve the overall task quality. A TPP algorithm is designed
Precentage of Selfish Nodes Precentage of Selfish Nodes
based on the secure multi-party computation theory to protect
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 10

the attribute-based privacy information in the traffic [18] Yang Z, Xiang D, Cheng Y. VR Panoramic Technology in Urban Rail
Transit Vehicle Engineering Simulation System,. Ieee Access, 2020, 8:
information transmission system. This algorithm shows 140673-140681.
excellent performance on both the INFOCOM06 and [19] Rozmus M, Tokarczyk J, Michalak D, et al. Application of 3D Scanning,
SIGCOMM09 datasets. The content discussed here is of great Computer Simulations and Virtual Reality in the Redesigning Process of
Selected Areas of Underground Transportation Routes in Coal Mining
value for advancing DTs transportation systems and dealing
Industry,. Energies, 2021, 14(9): 2589.
with real traffic problems. However, in many Crowd-sensing [20] Lee J, Zhang X, Park CH, et al. Real-Time Teleoperation of Magnetic
scenarios, the perception data shows a strong correlation in both Force-Driven Microrobots With 3D Haptic Force Feedback for Micro-
the time and space dimensions. This factor’s impact is ignored Navigation and Micro-Transportation,. IEEE Robotics and Automation
Letters, 2021, 6(2): 1769-1776.
in the task scheduling reported here, which may influence the [21] Manias DM, Shami A. Making a Case for Federated Learning in the
experiment precision. The follow-up study will discuss this Internet of Vehicles and Intelligent Transportation Systems,. IEEE
issue in detail. Network, 2021, 35(3): 88-94.
[22] Bagga P, Das AK, Wazid M, et al. On the design of mutual authentication
and key agreement protocol in internet of vehicles-enabled intelligent
VI. REFERENCES transportation system,. IEEE Transactions on Vehicular Technology,
2021, 70(2): 1736- 1751.
[23] Feng J, Liu Z, Wu C, et al. Mobile edge computing for the internet of
[1] Elsisi M, Tran MQ, Mahmoud K, et al. Towards secured online vehicles: Offloading framework and job scheduling,. IEEE vehicular
monitoring for digitalized GIS against cyber-attacks based on IoT and technology magazine, 2018, 14(1): 28-36.
machine learning,. IEEE Access, 2021, 9: 78415-78427. [24] Xu X, Li H, Xu W, et al. Artificial intelligence for edge service
[2] Yuan Q, Peng Y, Xu X, et al. Key points of investigation and analysis on optimization in internet of vehicles: A survey,. Tsinghua Science and
traffic accidents involving intelligent vehicles,. Transportation safety and Technology, 2021, 27(2): 270-287.
environment, 2021, 3(4): tdab020. [25] Deng T, Zhang K, Shen ZJ M. A systematic review of a digital twin city:
[3] Yusupbekov NR, Tarassov V B. Integrated intelligent systems for A new pattern of urban governance toward smart cities,. Journal of
industrial automation: the challenges of Industry 4.0, information Management Science and Engineering, 2021, 6(2): 125-134.
granulation and understanding agents,. Chemical Technology, Control, [26] Saroj AJ, Roy S, Guin A, et al. Development of a Connected Corridor
and Management, 2018, 2018(4): 62-75. Real-Time Data-Driven Traffic Digital Twin Simulation Model,. Journal
[4] Fedorko G, Molnár V, Vasiľ M, et al. Proposal of digital twin for testing of Transportation Engineering, Part A: Systems, 2021, 147(12):
and measuring of transport belts for pipe conveyors within the concept 04021096.
Industry 4.0,. Measurement, 2021, 174: 108978. [27] Yang D, Karimi HR, Kaynak O, et al. Developments of digital twin
[5] Lu Y, Liu C, Kevin I, et al. Digital Twin-driven smart manufacturing: technologies in industrial, smart city and healthcare sectors: a survey,.
Connotation, reference model, applications and research issues,. Complex Engineering Systems, 2021, 1(1): 3.
Robotics and Computer-Integrated Manufacturing, 2020, 61: 101837. [28] Novák P, Vyskočil J, Wally B. The digital twin as a core component for
[6] Ding K, Chan FTS, Zhang X, et al. Defining a digital twin-based cyber- Industry 4.0 smart production planning,. IFAC- PapersOnLine, 2020,
physical production system for autonomous manufacturing in smart shop 53(2): 10803-10809.
floors,. International Journal of Production Research, 2019, 57(20): [29] Zhang J, Deng T, Jiang H, et al. Bi-level dynamic scheduling architecture
6315-6334. based on service unit digital twin agents,. Journal of Manufacturing
[7] Fuller A, Fan Z, Day C, et al. Digital twin: Enabling technologies, Systems, 2021, 60: 59-79.
challenges and open research,. IEEE access, 2020, 8: 108952-108971. [30] Kupriyanovsky V, Namiot D, Klimov A, et al. The ontologies of cyber-
[8] Bellavista P, Giannielli C, Mamei M, et al. Application-Driven Network- physical systems of the national digital twin of Great Britain and BIM on
Aware Digital Twin Management in Industrial Edge Environments,. the examples of smart cities, railways, and other projects,. International
IEEE Transactions on Industrial Informatics, 2021, 17(11): 7791-7801. Journal of Open Information Technologies, 2021, 9(3): 91-130.
[9] Chen D, Wawrzynski P, Lv Z. Cyber security in smart cities: a review of [31] Li M, Li Z, Huang X, et al. Blockchain -based digital twin sharing
deep learning-based applications and case studies,. Sustainable Cities platform for reconfigurable socialized manufacturing resource
and Society, 2021, 66: 102655. integration,. International Journal of Production Economics, 2021, 240:
[10] Wang FY, Li Y, Zhang W, et al. Digital Twin and Parallel Intelligence 108223.
Based on Location and Transportation: A Vision for New Synergy [32] Feng H, Bai F, Xu Y. Identification of critical roads in urban
Between the IEEE CRFID and ITSS in Cyberphysical Social Systems transportation network based on GPS trajectory data,. Physica A:
[Society News],. IEEE Intelligent Transportation Systems Magazine, Statistical Mechanics and its Applications, 2019, 535: 122337.
2021, 13(1): 249-252. [33] Barmpounakis E, Geroliminis N. On the new era of urban traffic
[11] Qian C, Liu X, Ripley C, et al. Digital Twin—Cyber Replica of Physical monitoring with massive drone data: The pNEUMA large-scale field
Things: Architecture, Applications and Future Research Directions,. experiment,. Transportation research part C: emerging technologies,
Future Internet, 2022, 14(2): 64. 2020, 111: 50-71.
[12] Pan YH, Wu NQ, Qu T, et al. Digital-twin-driven production logistics [34] Wang S, Yu D, Ma X, et al. Analyzing urban traffic demand distribution
synchronization system for vehicle routing problems with pick-up and and the correlation between traffic flow and the built environment based
delivery in industrial park,. International Journal of Computer Integrated on detector data and POIs,. European Transport Research Review, 2018,
Manufacturing, 2021, 34(7 -8): 814-828. 10(2): 1 -17.
[13] Lv Z, Chen D, Lou R, et al. Intelligent edge computing based on machine [35] Wang X, Ning Z, Hu X, et al. A city-wide real-time traffic management
learning for smart city,. Future Generation Computer Systems, 2021, 115: system: Enabling crowdsensing in social Internet of vehicles,. IEEE
90-99. Communications Magazine, 2018, 56(9): 19-25.
[14] Chen Z, Fiandrino C, Kantarci B. On blockchain integration into mobile [36] Darwish TSJ, Bakar K A. Fog based intelligent transportation big data
crowdsensing via smart embedded devices: A comprehensive survey,. analytics in the internet of vehicles environment: motivations,
Journal of Systems Architecture, 2021, 115: 102011. architecture, challenges, and critical issues,. IEEE Access, 2018, 6:
[15] González M, Salgado O, Croes J, et al. A digital twin for operational 15679-15701.
evaluation of vertical transportation systems,. Ieee Access, 2020, 8: [37] Zhou H, Xu W, Chen J, et al. Evolutionary V2X technologies toward the
114389-114400. Internet of vehicles: Challenges and opportunities,. Proceedings of the
[16] Martínez -Gutiérrez A, Díez -González J, Ferrero- Guillén R, et al. IEEE, 2020, 108(2): 308-323.
Digital twin for automatic transportation in industry 4.0,. Sensors, 2021, [38] Wang Y, Ning W, Zhang S, et al. Architecture and key terminal
21(10): 3344. technologies of 5G-based internet of vehicles,. Computers & Electrical
[17] Almeaibed S, Al- Rubaye S, Tsourdos A, et al. Digital twin analysis to Engineering, 2021, 95: 107430.
promote safety and security in autonomous vehicles,. IEEE [39] Ji B, Zhang X, Mumtaz S, et al. Survey on the internet of vehicles:
Communications Standards Magazine, 2021, 5(1): 40-46. Network architectures and applications,. IEEE Communications
Standards Magazine, 2020, 4(1): 34-41.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 11

[40] Wu Z, Li G, Shen S, et al. Constructing dummy query sequences to

protect location privacy and query privacy in location-based services,.
World Wide Web, 2021, 24(1): 25-49.
[41] Mousavi R, Chen R, Kim DJ, et al. Effectiveness of privacy assurance
mechanisms in users' privacy protection on social networking sites from
the perspective of protection motivation theory,. Decision Support
Systems, 2020, 135: 113323.