See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/349080068

Traffic Urban Control Using an Intelligent PSO Algorithm Based on Integrated

Approach

Article in Algerian Journal of Signals and Systems · March 2020

DOI: 10.51485/ajss.v5i1.89

CITATIONS READS

4 171

3 authors:

Radja Diaf Cherif Tolba

Badji Mokhtar - Annaba University Badji Mokhtar - Annaba University
1 PUBLICATION 4 CITATIONS 38 PUBLICATIONS 355 CITATIONS

SEE PROFILE SEE PROFILE

Ahmed Nait Sidi Moh

Université Jean Monnet - Saint Étienne
104 PUBLICATIONS 947 CITATIONS

SEE PROFILE

All content following this page was uploaded by Radja Diaf on 06 May 2022.

The user has requested enhancement of the downloaded file.

 Traffic Urban Control Using an Intelligent PSO
Algorithm on Integrated Approach

Radja DIAF(1)*, Cherif TOLBA(2), Ahmed Nait Sidi Moh(3)

(1)
Computer Science Department Badji Mokhtar University BP.12 23000, Annaba, Algeria
(2)
Computer Science Department Badji Mokhtar University BP.12 23000, Annaba, Algeria
(3)
Innovative Technologies Laboratory (LTI) University of Picardie Jule Verne (UPJV)

diaf.radja@yahoo.com

Abstract: In this paper, we intend to contribute to the improvement of urban traffic mobility using a learning
method of traffic lights controllers. We proposed a Particle Swarm Optimization (PSO) method in which the
intelligent swarm acts as the cycle time of the traffic signal. The best swarm (solution found) meets the
evaluation criteria selected to describe desired objectives. The main measures of traffic lights efficiency are to
maximize flow-rate at which vehicles can cross a road junction and minimize the additional travel time of the
driver called vehicle delay. Particle Swarm Optimizer was coupled with the traffic flow model based on
Continuous Petri nets (PN). One potential advantage of CPN model is to provide insights regarding a behavior
of the platoon of vehicles on the target road network. The result obtained from this study has tested with
various scenarios related to intersections in different situations. The developed self-scheduling of the optimal
signal timing ensures safety and continuous traffic flow, thus increasing the mobility and reducing fuel
consumption and pollutant emissions.

Keywords: PSO, large scale-optimization, prediction, traffic flow modelling, traffic lights, Petri nets
of socio-economic activities (transportation
1. INTRODUCTION of humans, goods and products), traffic
In the vehicular traffic flow field, the vehicle
dynamics is enough complex. This networks have been extended more and
complexity is due to the nonlinear more so that they represent almost half of
interaction between travel decision surface area of urban cities. However this
behaviour, routing of vehicles and traffic development of traffic networks is not
congestion in networks [4]. In practice it is enough to eliminate traffic congestion
hard to take into account all interacting phenomena. Under some circumstances
elements when processing a given (rush hours, road accident) traffic flow
engineering transportation problem. The exceeds the road and crossroads capacity
traditional approach of transportation causing large queuing of vehicles. Traffic
engineering is to isolate the elements most management systems are intended to
relevant for separate handling and to bring reduce congestion and the resulting
these elements and the relationships negative effects: vehicle delay, fuel
between them together into a unified consumption, pollution, etc. The significant
modular system. From viewpoint of traffic increase in computing power allowed
flow modeling, the behaviour of vehicles in implementing online optimization algorithms
road is often seen as in two main levels: that improve the efficiency of controllers in
macro-mobility and micro-mobility. Micro- real time.
mobility level focuses on the individual The main purpose of this paper is to
vehicle behaviours in road. Vehicles are present a new methodology for traffic lights
individually represented by considering the control using Particle Swarm Optimization
interactions between them taking into (PSO) and an integrated approach based on
account (headway, speed, acceleration). Rakha model [10] and a Macroscopic Traffic
Although, the macro-mobility level has been Flow (MTF) model commonly known
inspired from the kinematic wave model of Lighthill, Witham and Richards (LWR) model
Lighthill and Whitham and Richards (LWR) [16][19]. This latter is used for estimating the
[16][19]. In this case, the traffic flow is arrival profiles of vehicles. PSO algorithm
described by global variables as the flow- performed the cycle time optimization of
rate, the flow density and the flow average traffic lights using the cumulative platoon
speed. Traffic congestion occurs frequently arrival profile provided by MTF model. The
within traffic networks. Indeed, in order to prediction of the platoon movement and
absorb the streams induced by the growing arrivals in intersections are reproduced both
using the VCPN (Continuous timed Petri Net (Time is discretized in steps, empty places in
with Variable speeds) and Rakha vehicle finite time). Each time step can be seen as
dynamics model. The meta-heuristic the travelling time (delay) of vehicles between
algorithm (PSO) is used to compute the two places and in order to avoid a negative
optimal traffic signals with aim to minimizing marking (vehicles) of place, its input flow
the traffic jam. during each time step is bounded. Moreover,
The traffic flow macroscopic parameters in order to determine urban road traffic
such as: density, flow-rate and speed are behaviors, the continuous PN are combined
determined by LWR-VCPN model. The with discrete part that represents traffic lights.
deterministic parameters of LWR-VCPN Consequently, the resulting model is
model namely the free speed of each road characterized by hybrid dynamic. The
segment is obtained from the Rakha model. management of road traffic by hybrid
Thus, the advent cumulative platoon at each formalism have been developed based on
intersection is provides by LWR-VCPN minimizing the total delay of cars.
model. The estimated cumulative platoon
Reference [1] use the Hybrid Petri Nets
arrival is used as an input data of PSO
algorithm [11] for minimizing the traffic jam (HPNs) model who is modeling the flow of
vehicles as a fluid, to present an urban traffic
by finding the optimal signal time plan.
network of signalized intersections, and to
A continuous timed Petri net with variable
speeds (VCPN) traffic model [5][6] is used to solve the problem of synchronizing many
traffic signals, in the goal of developing the
describe in macroscopic manner of platoon
performance of some special vehicles such
evolutions in time and [15].
as public and emergency vehicles. The
This paper is arranged as follow: In section 2 proposed model is validated employed real
review of related work is presented, in traffic data of Italian. A microscopic model is
Section 3 exhibits the architecture of the proposed by [2] based on deterministic timed
proposed traffic light control system, in PN (DTPN) for representing the traffic and
section 4 describes the system modeling, in signalized intersections, to proceed the aim of
section 5 presents the system optimization, traffic management, this proposed model is
and in section 6 the main contribution known tested on real traffic data. The model of urban
an integrated approach and its application, in traffic system considered by [3] and its
the last the conclusion and future works. representation with stochastic timed PN for
optimizing the duration of green signal in each
2. RELATED WORK stage in order to augment the performance of
Various researches on modeling of traffic flow the system by minimizing the queue delay.
progression use different approaches to solve The work elaborate by [18] used the
problems of traffic light scheduling, and deterministic and stochastic Petri net model
optimization methods to reach the optimal [DSPN] to represent the urban traffic system
traffic lights sequence utilize specified criteria in intersection, this model permits easy
such as the minimization of waiting times, modification of vehicle flow which has been
delays etc…., the maximization of the number validated with real traffic data, the aim of this
of vehicles outflow, etc. In this regard different work is to augment the performance of
works use Petri Net (PN) tool to modeling the vehicle flows. References [5][6] proposed a
urban traffic network, and several studies new approach to model macroscopic traffic
employ the Particle Swarm Optimization flow and representing its variables through
Algorithm (PSO) combined with other sections of road using Continuous PN with
techniques or algorithms in order to find the Variable speed (VCPN) in which the VCPN
optimal traffic light timings. For the sake of parameters are described by road traffic
simplicity, the unfamiliar reader with PN can ones. The work in [23] proposed a new signal
refer to [21]. control method based on the Petri net model.
When the urban signalized intersection is
Reference [17] offer a macroscopic model for modeled by a hybrid colored Petri net to
urban traffic network based on First Order optimize the traffic signal control of isolated
Hybrid PN (FOHPN) formalism, where the intersections, which divided on two parts the
traffic flows is represented as a fluid traffic signal control module and a traffic flow
approximation, and the traffic signal is module.
considered as a discrete part and modeled by
timed Petri nets. The dynamic of FOHPN Reference [12] Propose an optimization
model is governed by time-driven and event- strategy based in particle swarm optimization
driven dynamics. Obtained model is validated method to find optimum signal timings and
and compared with a microscopic model then the results obtained are evaluated with a
based on colored timed PN (CTPN). Jùlvez microscopic simulator in the context of two
and Boel [9] use a several features of areas, in the aim to maximize the number of
continuous PN with infinite server semantics vehicles that reach arrive in their destination.
A new method to optimize the traffic-jam one system: modeling, optimization and
probability is introduced by [7] where the performance metrics assessment. The
probabilistic distribution of the ready vehicles topology of the proposed system is
is predicted from The Bayesian Network (BN) illustrated in the following fig. 1.
model and the Cellular Automaton traffic
model (CA) is used to update this
probabilities, and then used the Particle
Swarm Optimization Algorithm (PSO) to
obtain the optimal traffic signal scheduling,
and determine the efficiency of the offered
method by a micro-simulator. a system was
also proposed by [22], (IOCA-PSO) which is a
method applied to attain an urban traffic
signalization schedule dynamic and optimum,
combined two approaches the inner and outer
cellular automaton model (ICM –OCM) with
particle swarm optimization algorithm (PSO)
who calculate the fitness value. In [13]
Particle Swarm Optimization Algorithm (PSO)
is used like a method of optimization to find
an optimal traffic light programs, and a
microscopic traffic simulator evaluate the Fig.1 System architecture
solution obtained, in this work two areas
American, and European city are tested, in 4. SYSTEM MODELISATION
which objectives to maximize the number of
A. VCPN Model of Urban traffic network
vehicles that reach the downstream line and
In this section we present a continuous timed
to minimize the total delay. Reference [14]
Petri net with variable speeds (VCPN) traffic
designed a real time traffic signal control
model [5][6] used for macroscopic modeling
approach at isolated junctions and bridges
of urban traffic flow. One of the major features
based on Fuzzy Artificial Neural Network
of macroscopic traffic flow approach is the
(FANN), the aim of this work is to minimize
fundamental diagram (FD) that gives the
delays and maximize the number of vehicles
relation between the three traffic flow average
outflow, the authors used two optimization
parameters: density (vehicles/km), speed
algorithms to compute the performance of
(km/hour) and flow-rate (vehicles/hour).
(FANN) such as Genetic Algorithm (GA) and
Particle Swarm Optimization (PSO). A new Definition 1 VCPN is a marked graph depicted
method using Particle Swarm Optimization by quintuplet G = (P, T, Pre, Post, θmax)
Algorithm (PSO) proposed by [20], aims to where P and T are respectively the finite-set
optimize traffic signal time, cycle length, and of places and the finite-set of transitions, Pre
phase splitting in intersections by minimizing and Post are two incidence functions where
the total delay, to improve the performance of Pre(Pi, Tj) is the weight of the input arc
road network. directed from place Pi (Pi  P) to transition Tj
According to the synthesis of the different (Tj  T) and Post(Pi, Tj) is the weight of the
works concerning the Petri nets and the PSO output arc directed from place T j to transition
algorithm, we note that there is no work Pi. θmax is the real positive vector of
integrating these two approaches to transitions maximal firing speeds, θmax 
determine the optimal cycle time of traffic (IR+)q. where the real positive firing speeds
lights, which is our proposed work. vector at time t, Q(t) (IR+)q and the real
positive marking vector M(t) (IR+)n . The
3. INTEGRATED ROAD TRAFFIC marking evolution of the VCPN is given by the
MANAGEMENT ARCHITECTURE following differential system:
The main objective of all contributions in this
area is the search for more fluid and efficient
circulation levels. A new method of Where M0 is the initial marking vector of
controlling traffic lights is proposed, the places at time t0 and Γ is the incidence matrix
Rakha model of vehicle dynamics integrated of VCPN defined as Γ = Post – Pre. The
with the LWR-VCPN model. Then, the components of the firing speeds vector Q(t)
estimated cumulative platoon arrival is used depend continuously on the marking of the
as an input data of PSO algorithm for VCPN according to the following equation:
minimizing the traffic jam by finding the
optimal signal time plan.
Our contribution focus on the bringing
together three component sub-systems into
 For all , max j max and mk(t)M(t) segments of length i (i = 1, ..., N). Each
segment of road is modeled by the average
traffic parameters such as density ρ(t), flow-
rate q(t) and speed S(t).

B. LWR Model
VCPN is used for modeling the LWR i
macroscopic model to estimate the
cumulative platoon arrival profile, this model qi-1(t) qi(t)
is based on hydrodynamical theory and vi(t) , pi(t)
consider the flow as a compressible vi(t) , pi(t)
fluid[16][19], where vehicles are not
represented individually but as a continuous
flow, and on the theory of waiting line [8]. The Fig. 3 Section [xi-1,xi]
LWR approach is based on three principle
laws such as (4), (5), and then (6). References [5][6] presented the possibility to
describe the LWR model based a VCPN
- The conservation law formalism that used for macroscopic
modeling of urban traffic flow. One of the
principal laws of macroscopic traffic flow
approach is the fundamental diagram (FD)
- The continuous temporal traffic variables that gives the relation between the three
(5) traffic flow average parameters such as:
density, speed and flow-rate.
- The fundamental diagram
In VCPN the place Pi represent a segment
(6) road i, and the transition Ti correspond to the
junction between two segments i and i+1.
q
Each segment i has the marking mi(t) which
vfree is the available number of vehicles in place
Pi. The flow-rate qi(t) is formulated according
qmax the firing speed vi(t) associated to the
transition Ti. As consequent, the relationship
between the average density ρi(t) and the
average velocity Si(t) for each segment i is
given by (7) and (8).
F C

cr j

Fig. 2 fundamental diagram

The parameters are defined as: Figure 4 illustrates the VCPN traffic model
v: free flow speed. [5][6] in which the road segment i is described
qmax: maximum flow allowed. by the place Pi and the two transitions Ti-1 and
j: jam density.
Ti.
cr: critical density (the density that
characterizes the appearance of congestion). m`i+1
In the case when the number of vehicles in a
segment is bigger than what is exiged, a
congestion is defined. If a segment i is
congested (C), the density i in this last is
greater than the critical density cr. And if a
segment i is not congested (or free (F)), the mi mi+1
density i is less than or equal the critical
density cr.
So,The two modes are defined as follows:
free mode (0 ≤ i ≤ cr) and congested mode m``i
( i > cr).
The model requires spatial and temporal Fig. 4 VCPN traffic model of segment i
discretization, the spatial discretization
consists of decomposing the road into a set of
 These programs coordinate traffic signals at
adjacent intersections to maximize the
number of vehicles leaving the junction and
minimize the traffic jam.
The computational population based meta-
The marking m'i+1(t) of the place P'i+1 stands heuristic PSO was introduced by the
for the available vehicles emplacement pioneers Kennedy and Eberhart [11]. PSO is
segment [xi, xi+1]. In addition, the place P''i
associated to the transition Ti represents its Algorithm: The procedure of PSO
maximum vehicles number that pass 1: initializeSwarm();
simultaneously. 2: ComputeGlobalBestParticle (b);
The flow rate qi(t) of the segment i is defined 3: while g < MaxNumbIterations do
as (9) 4: for ( i=1 to size of swarm) do
5: Calculation of new velocity by (17)
6: Calculation of new position by (16)
7: end for
vfree is the free speed in segment i and Xmax i 8: updateComputeGlobalBestParticle (bg)
represents the maximal firing speed of 9: end while
transition Ti. based on the exploring the search-space of
any problem in order to find the set of
C. Rakha Model particles that maximizes/minimizes a defined
Based on the free speed found by the Rakha objective function and each potential solution
model which allows the segmentation of the to the problem is called particle position. PSO
road, the authors in [10] have proposed the algorithm is presented as follow:
vehicle dynamic model with constant power.
The maximum acceleration of vehicle is To apply the PSO algorithm in the traffic
expressed by the different forces acting on signal control, it is enough to consider each
vehicle divided by the vehicle total mass M particle as a cycle of the traffic signal. Each
(11). traffic light dimension is determined by the
number of phases allowed at the considered
(11) intersection
In this algorithm at each iteration, the particle i
The instantaneous acceleration a(ti) can be will move between the iterations i and i+1 in
deduced according to the net force F(ti) at terms of its velocity and its two known best
time t and the total resistance R(ti) at time t. positions (its own and that of the swarm)
Thus, the relationship in (12) allows according to two equations (16, 17).
calculating the dynamic equations system
constituted by (14) and (15). (16)

Where:
(13) pgi: the best personal solution of the particle i;
bg: the best global solution of all particles;
w: the mass of inertia of the
particle;
φ1 and φ2 are coefficients that accelerate and
control the relative components namely the
best personal and the global particles;
Where: U[0 , φk ] is a random value U[0 , φk ] où k ϵ
a: acceleration [m/s]; 1,2.
v: velocity [m/s];
As mentioned above, each particle describes
x: distance travelled [m] during t; a cycle of traffic lights and its dimension is
ti = t0 + it for i = 1,2,….,N. determined by the number of phases allowed
at the intersection. So, each phase is known
5. SYSTEM OPTIMIZATION by duration and its state which is a sequence
A.The PSO algorithm for the optimization of possible traffic lights (green, yellow, red).
of traffic signals In order to estimate the quality of the particle
Based on the arrival profiles of vehicles found i, it is necessary to calculate its fitness
bya LWR model, The PSO algorithm using to function.
find optimal traffic signal cycle program.
This last is calculated using a special function  The inputs in the different LWR model
for the problem addressed. In order to update segments are the optimum speed and vehicle
the values of xi, pbest and gbest, their fitness head time characteristics.
which is defined by the following equation are  To determine the maximum speed that a
computed at each iteration of the algorithm. vehicle can reach in each segment. The
Thereby the initial values of the fitness Rakha model generates the vehicle speed /
function are extracted from the Rakha-LWR time curve.
model.  Before the leading vehicle reaches the
free flow speed, the segments are 100
meters each, as the vehicles move at free
speed the length of the segment is residual
distance of road to the stop point.
Where:
B. Simulation and Implementation Model
z1 = EW - h*qEW
The considered system is a motorway section
z2 = NS - h*qNS (Fig. 5), which has a length of 700m, with one
EW : Density in east and west direction. intersection contained four traffic lights (east-
NS : Density in north and south direction. weast and north-south). If we consider that the
qEW: Flow rate in east and west direction. length of the vehicle is 6.6 m, then this motorway
qNS : Flow rate in north and south direction. can contain of 106 vehicles.
EWend : Output Density in east and west
direction.
NSend : Output Density in north and south
direction.
h =1: The time step.
Fitness function algorithm is presented as
follow:

Algorithm: The procedure of Fitness Function

1: Simulation_parameters();
2: IntializeSauvegaredVariables();
3: ComputeFlowRateInitialTime_EW();
4: ComputeFlowRateInitialTime_NS(); Fig.5 Test Road
5: while h < CycleTime do
6: ComputeDensitySegments_EW();
7: ComputeDensitySegments_NS(); 1. Vehicle Arrival Model Using Poisson
8: UpdateFlowRateSegments_EW(); Distribution Model
9: UpdateFlowRateSegments_NS(); In this proposed work the inputs of the
10: Compute(z1); system follow Poisson distribution model
11: Compute(z2); which is defined as follows:
12: updateComputeGlobalBestParticle (bg) When the number of vehicles arriving in a
13: end while given period of time is random, The Poisson
14: Compute(F(x)); Distribution is proposed as early as the
1930s is commonly used to describe such a
random process. For this distribution, the
6. APPLICATION AND INTEGRATED probability that a number x of vehicles will
APPROACH arrive during a period of time t is given as:
The studied system is composed of two parts
as follows: (19)
 Part 1: use the Rakha-LWR model to
determine the traffic data values;
 Part 2: use the optimization algorithm  the average number of vehicles arriving
PSO to determine the optimal traffic signal during a period t
cycle program; The probability that some number of vehicles
A. Integration of the LWR-VCPN and the (n) arriving in a period can be calculated as:
Rakha models
The notion of integration of the two (20)
approaches is founded on the next points:
2. Simulation of Rakha and LWR-VCPN
 The behavior of the leading vehicle as the
speed and the acceleration is simulated by models
the Rakha model. To find speed and time characteristics we
implement the Rakha model, in the
implementation of this model the input is
based on the parameters of the Snare In figure. 8, Marking m0 represents the
database [13], using t =0.1s, the speed and number of vehicles ready to depart; and the
distance travelled by the vehicle are obtained number of vehicles vehicles in segment 1, 2,
by the equations from (11) to (15) 3 and 4 respectively are represented by: m1,
implemented via MATLAB. m2, m3, and m4, the number of vehicles that
reached the downstream is represented by
m5, markings mi-ci represent the number of
Fig.6. Speed versus distance curve available site in each segment.
The incidence matrix of VCPN is defined as
follow: (P,T)
where:
P: places (P1, P2, P3, P4, P1`, P2`, P3`, P4`)
T: Transitions (T0, T1, T2, T3, T4)
[1 -1 0 0 0
0 1 -1 0 0
0 0 1 -1 0
0 0 0 1 -1
-1 1 0 0 0
0 -1 1 0 0
0 0 -1 1 0
Speed simulation of the leading vehicle 0 0 0 -1 1]
according to the distance is showing in fig. 6. The following table 1 present the VCPN
parameters
Table 1 VCPN parameters
According to Fig. 6, it is observed that a
vehicle attained speed of 50.58 km/h after it
travels 100 m from the upstream stop line; i: the segments longer
speeds of 70.04 km/h, 82.59 km/h and 90.02 Vfree i : the free flow-rate obtained with the
km/h are attained at 200 m, 300 m and 400 implementation of Rakha model.
m respectively. Ci-mi: the number of available sites in each
It is observed that the leading vehicle arrives segment, calculated based on the segment
at the free flow-rate vfree i which is equal to Segment 01 02 03 04
82 km/h at 300 m. From this moment the i 100 100 100 400
vehicle is supposed to keep on going to pass
Vfree i (m/s) 14.05 19.45 22.94 25.00
over 300 m at the same speed.
So, the road of the motorway studied can be Ci-mi (veh) 15.15 15.15 15.15 60.61
segmented into four segments as shown in αi 3.55 2.57 2.17 8
figure 8. By using the speed profile, the first length () and the maximum occupancy
three segments are 100m and the fourth one length of 6.6 m per vehicle.
is 400m. αi: the average number of vehicles that can
cross the junction simultaneously, is
calculated based on the maximal flow rate
( , segment length ( and the segment
limited maximal firing speed (
according to equation (21)

Fig. 7 Road segments

Considering saturation flow rate of 1800
The VCPN model of the road is presented as
follow: veh/h/lane, which is 0.5 veh/s/lane
Thus, the VCPN model is ready to be
implemented to estimate the arrival profile at
the road:

Fig.8. VCPN model of road

 Table 3 Best cost values of PSO

Swarm Max Cost Position Cycle

size Iteration time

100 2000 [30 6 32 6] 74 s

50 300 1915 [30 5 31 5] 71s
500 1926 [30 6 31 5] 72s
100 1904 [30 5 30 6] 71s
100 300 1902 [30 7 30 6] 73s
500 1865 [31 5 31 5] 72s
Fig. 9 cumulative number of vehicles versus time curve 100 1889 [30 5 30 7] 72s
200 300 1889 [30 7 30 6] 73s
3. Simulation of PSO algorithm 500 1865 [30 6 30 5] 71s
Particle Swarm Optimization Algorithm
adopted to find the optimal cycle programs of
traffic lights, it works as follow:
- The initial swarm initialized with a set of a
random values representing the phase
durations, this values are within the Lower
Bound VarMin = [30 5 30 5] , and the Upper
Bound VarMax = [40 10 40 10], constitute the
range of possible time spans (Green Red,
Yellow Red, Red Green, Red Yellow).
- The values of velocity is specified according
to a defined limits as follows:
VelocityMax=0.1*(Vmax-Vmin), VelocityMin=
-VelMax;
and Vmin = -4; Vmax = 4; Fig.10 Trace progress of the best cost values of PSO

Other parameters of simulation are defined in • From viewpoint, Fig 10 plots the trace
the following table: progress of the best cost values Vs
Iterations, we can observe that our algorithm
Table 2 PSO algorithm parameters practically converged after the first 200
iterations
Parameter Value • Our optimization solver is able to report
- Swarm Size 20, 50, 100, 150, 200 successful cycle programs for the different
- Acceleration coefficient 2.05 instances
(φ1) • As shown in previous figure.10 for almost
- Acceleration coefficient 2.05 all the combinations (of Swram Size and
(φ2) MaxIt) our PSO got to converge on the
- The inertia weight (w) 0.5 interval of 300 and 500 iterations
- Inertia Weight 0.99
Damping Ratio (wdamp) 7. CONCLUSION & FUTURE
WORKS
We show a representation view of the In this paper, a new methodology for traffic
behavior of PSO algorithm under different signal control using a Particle Swarm
conditions of swarm size, maximum number Optimization (PSO) based on integrated
of iterations. In the following table present approach is presented. PSO is used to
the best cost values and the traffic logics compute the optimal sequence of traffic
(positions) correspond to different signal using the arrival profile of vehicles.
configurations of swarm size with: 50, 100, This latter is provided by the Macroscopic
and 200 particles, and maximum number of Traffic Flow model. The parameters of MTF
iteration 100, 300,500. model are determined according the speeds
profile obtained by Rakha model. The
combination between these models
highlighted the integrated approach. And as
a future works this contribution may be large
scale optimization with enhancement for
adjacent intersections.
 References [16] M.J. Lighthill, and G.B. Whitham, " On
kinematic waves I: Flood movement in long
[1] A. Di Febbraro, D. Giglio, and N. Sacco, rivers. II: A theory of traffic flow on long
"Urban Traffic Control Structure Based on crowded roads", Proceedings Royal Society
Hybrid Petri Nets", IEEE TRANSACTIONS (London) 229A: 281–345, 1955.
ON INTELLIGENT TRANSPORTATION
SYSTEMS,2004, VOL. 5, NO. 4 [17] M.Dotoli, M.P. Fanti, and G. Iacobellis, "An
Urban Traffic Network Model by First Order
[2] A. Di Febbraro,and D. Giglio,"On representing Hybrid Petri Nets", IEEE International
signalized urban areas by means of Conference on Systems, Man and
deterministic-timed Petri nets", IEEE Cybernetics, 2008.
Intelligent Transportation Systems
Conference Washington, D.C., USA, 2004 [18] M. A. Badamchizadeh, and M.Joroughi,
"Deterministic and stochastic Petri net for
[3] A. Di Febbraro,N. Sacco, and D. Giglio, "On urban traffic systems", The 2nd International
using Petri nets for representing and Conference on Computer and Automation
controlling signalized urban areas: new model Engineering (ICCAE), 2010.
and results", Proceedings of the 12th
International IEEE Conference on Intelligent [19] P.I. Richards, "Shock waves on the highway",
Transportation Systems, St. Louis, MO, USA, Operation Research 4: 42–51, 1956.
2009. [20] P. Hirulkar, R. Deshpande, and P.Bajaj,
[4] Boris S.Kerner, " Introduction to Modern "Optimization of Traffic Flow through
Traffic Flow Theory and Control ", Springer- Signalized Intersections using PSO",
Verlag Berlin Heidelberg, 2009. International Journal of Computer Science
and its Applications,2011, ISSN 2250 – 3765
[5] C.Tolba, D. Lefebvre, P. Thomas, and A.
ELMoudni, " Continuous Petri nets models for [21] T. Murata, N.Komoda, K. Matsumoto, and K.
the analysis of urban traffic networks ", Haruna, " A Petri net-based controller for
Proceedings of 2001 IEEE-SMC Int. Conf. on flexible and maintainable sequence control
Systems, Man and Cybernetics, 2001, pp. and its application in factory automation ",
1323–1328 IEEE Trans. Ind. Electron, 33 (1) ,1986, 1–8.
[6] C. Tolba, D. Lefebvre, P. Thomas, and A. [22] W.Hu, H.Wang, L.Yan, and B.Du, " A swarm
ELMoudni, "Continuous and timed Petri nets intelligent method for traffic light scheduling:
for the macroscopic and microscopic traffic application to real urban traffic networks ",
flow modeling", Simulation, Modeling Practice The International Journal of Research on
and Theory 13: 407–436, 2005 Intelligent Systems for Real Life Complex
Problems 44:208–231, 2016.
[7] C.Y.Cui, and H.H. Lee, "Distributed Traffic
Signal Control Using PSO Based on [23] Y. Zhang, W. Qiang, and Z.Yang, " A New
Probability Model for Traffic Jam", Intelligent Traffic Signal Control Method Based on
Autonomous Systems 12, AISC 193, 2013, Hybrid Colored Petri Net in Isolated
pp. 629–639. Intersections", Int. J. Intelligent
Transportation Systems Research, 2017.
[8] D.Heidemann,"A Queueing Theory Approach
to Speed-Flow-Density Relationships", Proc.
13th Symp. Int. Conf. Transp. and Traffic
Theory,1996, pp. 103-118.
[9] G.Júlvez, and R. Boel, " Modelling and
controlling traffic behaviour with continuous
petri nets", 16th Triennial World Congress,
Prague, Czech Republic, 2005.
[10] H. Rakha, I. Lucic, S. Demarchi, and J. Setti,
"vehicle dynamics model for predicting
maximum truck acceleration levels", Journal
of Transportation Engineering 3500(5):231-
1505, 2001.
[11] J. Kennedy, and R. Eberhart, " Particle swarm
optimization ", in Proc. of IEEE International
Conference on Neural Networks, 1995.
[12] J.Garcìa-Nieto, E. Alba, and A. Carolina
Olivera, " Swarm intelligence for traffic light
scheduling: Application to real urban areas ",
Engineering Applications of Artificial
Intelligence 25 ,2012, 274–283
[13] J. Garcìa-Nieto, A. Carolina Olivera, and E.
Alba, "Optimal Cycle Program of Traffic Lights
with Particle Swarm Optimization", IEEE
Transactions on Evolutionary Computation,
2013, Volume: 17 , Issue: 6
[14] J. Venkatesh, and B. Chiranjeevulu, "Traffic
Signal Control Based On Fuzzy Artificial
Neural Networks With Particle Swarm
Optimization ", International Journal of New
Technologies in Science and Engineering Vol.
3, Issue.7, July 2016, ISSN 2349-0780
[15] K.M. Ng, M.B.I. Reaz, and M.A.M.A. Ali,
"review on the applications of Petri nets in
modeling ", analysis and control of urban
traffic, 2013.

View publication stats