IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 13, No. 2, June 2024, pp. 2374~2385

ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i2.pp2374-2385  2374

Enhancing the smart parking assignment system through

constraints optimization

Nihal Elkhalidi, Faouzia Benabbou, Nawal Sael

Laboratory of Modeling and Information Technology, Faculty of Sciences Ben M’Sick, University Hassan II Casablanca, Casablanca,
Morocco

Article Info ABSTRACT

Article history: Traffic in big cities has become a black spot for drivers. One of the major
concerns is the parking problem that hinders urban mobility, particularly in
Received Jun 2, 2023 big cities and other congested areas. This leads to an increase in accidents, a
Revised Oct 16, 2023 big consumption of fuel, and a spectacular augmentation of pollution. In this
Accepted Nov 15, 2023 paper, we introduce a parking assignment system grounded in constraint
programming to address the growing demand for efficient parking
management in smart cities. Our system is designed to meet the
Keywords: requirements of groups of drivers seeking to reserve parking spaces
simultaneously within the same period and geographical area. This entails
Assignment problem imposing constraints on the desired parking type, including considerations
Constrained optimization such as walking and driving distances, parking costs, and availability.
Management system Within the scope of this study, we propose two formulations: constraint
Multi-agent system satisfaction programming (CSP) with an objective function and
Smart parking mixed-integer linear programming (MILP). Evaluation shows Choco, a CSP
solver, is effective for smaller requests but slower for larger ones, while
MILP excels for larger scenarios. Both solvers produce high-quality
solutions meeting real-time response requirements. Our research offers
innovative solutions for smart city management, considering parking type
preferences, costs, and availability. We contribute significantly to parking
space assignment methodologies, aiming to alleviate the time-consuming
search for parking, reduce accidents, fuel consumption, and pollution.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Nihal Elkhalidi
Laboratory of Modeling and Information Technology, Faculty of Sciences Ben M’Sick
University Hassan II of Casablanca
Driss Harti Sidi othman, Casablanca, BP 7955, Morocco
Email: [email protected]

1. INTRODUCTION
Today, numerous cities are aspiring to transform into smart cities, emphasizing human-centered
development, effective governance, and a heightened commitment to environmental conservation. However,
to achieve this vision, cities must confront significant challenges, foremost among them being the
development of digital technologies and interconnected devices that enhance public services and enable
informed decision-making based on available data. Smart cities draw upon an array of emerging
technologies, including artificial intelligence and the internet of things (IoT), to address various critical
issues. These encompass transportation and mobility management, security enhancements, healthcare
services, optimization of public lighting energy consumption, waste collection efficiency, and environmental
sustainability. Notably, the management of distributed parking within metropolitan areas stands as a pivotal
challenge within the smart city context, given its profound implications for drivers, the environment, and the

Journal homepage: http://ijai.iaescore.com

Int J Artif Intell ISSN: 2252-8938  2375

business landscape [1]. Furthermore, the automotive industry in Morocco has witnessed significant growth in
recent years, contributing to a rise in private car ownership. This surge results in a higher volume of vehicles
circulating in major cities such as Casablanca. The absence of an efficient parking management system has
led to a growing demand for parking slots, particularly in the city center. Owing to extended pick-up times,
the scarcity of parking spaces and the dearth of real-time traffic information, drivers often resort to
inappropriate parking locations, obstructing traffic flow and impeding pedestrian mobility, thereby
exacerbating traffic congestion issues. The concept of smart parking seeks to offer intelligent solutions for
monitoring parking facilities, providing drivers with up-to-date traffic status information, and guiding them
to locate optimal parking spaces. These parking lots are equipped with interconnected sensors that relay
pertinent data to drivers via their smartphones, encompassing the maximum parking capacity, the number of
vacant spaces, and parking lot geo-location. These integrated systems are typically referred to as smart
parking management systems (SPMS). The strategic objective of SPMS is to optimize parking in urban areas
where numerous drivers are searching for parking in various locations. The parking assignment process needs
to be highly efficient in order to accommodate all driver constraints. In a prior study, we introduced SPMS
based on a multi-agent system (MAS) [2], known as the multi-agent smart parking management system
(MASPMS). In the MASPMS model [3], each agent assumes responsibility for one or multiple tasks,
working in coordination to offer a real-time intelligent system for parking slot reservations. This system
comprises eight key agents that collaborate to expedite the search for available parking spaces: i) parking lot,
ii) zone, iii) reservation, iv) synchronization, v) geo-location, vi) classification, vii) orientation, and
viii) security agents.
The reservation agent (RA) relies on the work of what we refer to as a classification agent (CA),
which forms the core of the parking assignment system (PAS). The primary challenge in developing an
intelligent PAS is twofold: ensuring compliance with individual drivers' specific requirements and delivering
real-time responses to a multitude of drivers. Firstly, each driver possesses a unique set of criteria that govern
their parking preferences, encompassing factors such as the type of parking facility (e.g., covered and
open-air), driving distance, walking distance, and parking fees. Secondly, the system must be capable of
offering real-time responses to diverse drivers simultaneously searching for available parking spaces across
different areas within the city. This challenge necessitates the effective allocation of parking facilities that
cater to the needs of all drivers, concurrently. A SPMS disseminates real-time information to drivers about
the current parking scenario, guiding them to the best-suited parking facilities based on their specific
requirements. Three critical factors must be taken into account in the PAS: capacity, distinct assignments
tailored to individual drivers, and adherence to drivers' specific requirements. First and foremost, the
real-time parking capacity and availability determine whether a reservation request is accepted or declined.
Secondly, it is crucial to ensure that different drivers are assigned to separate parking spaces to avoid
overlaps. Lastly, the proposed parking allocation should align with user requirements, including
considerations such as the minimum walking distance to reach the driver's intended destination. The
challenge in parking space assignment stems from the multitude of constraints imposed by users. Each
reservation request for a parking space comprises a diverse array of constraints, reflecting the unique needs
of many drivers. Expressing these rules explicitly can be particularly challenging due to the inherent diversity
in drivers' requirements. Moreover, the task becomes even more complex when a substantial number of
drivers simultaneously request parking spaces on the road, coupled with the continuous fluctuations in
parking slot availability. Thus, the assignment of real-time and optimal solutions within a PAS is a
formidable challenge.
In recent years, several studies have focused on enhancing parking management and assignment
through diverse approaches, including machine learning, as exemplified in [4], [5], as well as optimization
algorithms [6]. However, the traditional first-in-first-out (FIFO) technique becomes inadequate when the
system is confronted with a simultaneous influx of reservation requests across various areas, each containing
differing numbers of available parking spaces. In such scenarios, there arises a necessity to devise a real-time
system capable of efficiently accommodating multiple users' requests, respecting their constraints, and
optimizing the assignment process. This study aims to introduce a system designed for the management of
multiple reservation requests that occur concurrently in a MASPMS responsible for overseeing various
parking facilities. The system takes into careful consideration several constraints pertinent to drivers. The
drivers' requirements under consideration encompass factors such as parking capacity, walking distance from
a designated parking site to their destination, parking fees, the probability of availability, and the driving
distance from the user's present location to the allocated parking space. To achieve this objective, we have
employed two distinct problem assignment formulations. The first utilizes constraint satisfaction problems
(CSP) [7], while the second is based on mixed-integer linear programming (MILP) [8]. In both cases, the
primary goal is to allocate parking spaces to users while adhering to their constraints or optimizing the
proposed solutions.

Enhancing the smart parking assignment system through constraints optimization (Nihal Elkhalidi)
2376  ISSN: 2252-8938

In what follows, in section 2, we propose a review of related work. Section 3 proposes an outline of
some backgrounds. Section 4 presents our materials and used methods. Simulation results based on the three
solvers Gurobi optimizer, Choco, and Google OR-Tools are illustrated in section 5. Section 6 finally
discusses the conclusion and further research.

2. LITERATURE REVIEW
Several researchers in the literature have expressed an interest in the parking assignment problem. In
this context, numerous models and techniques have been presented. For example, Zhao et al. [9] proposed a
method for an intelligent PAS that can provide guidance information to vehicles in real-time. They develop a
linear optimization-based algorithm for a specific assignment problem by keeping parking requests in a
queue for a set period. Kim et al. [10] presented a PAS with two goals: i) minimize parking expenses and
ii) balance parking demand across numerous lots using a MILP. To resolve the assignment problem in this
study, they used a technique known as the alternating direction multipliers method (ADMM). Numerical
results prove that the ADMM outperforms Greedy by 36% in parking utilization. Ratli et al. [11] described a
dynamic PAS to maximize parking occupancy and provide total customer happiness. They proposed a
combination of the estimation of distribution algorithm (EDA) and a local search (LS) procedure based on
MILP. To enhance the approach's performance, the problem was decomposed, and therefore, resolution time
was reduced. The approach adopted in [12] presented a PAS for a group of parking. The price system and
time-of-arrival balance constraints are included in their model. The authors used a standard ant colony
optimization (ACO) algorithm that improved further by employing adaptive techniques, to resolve the issue.
ACO is a nature-inspired algorithm and it is one of the swarm intelligence algorithms.
Hakeem et al. [13] presents a free parking system (FPS). It is an allocation system for assigning free
street parking places to city drivers. It offers a dynamic assignment method that manages a collection of driver
requests that separately reach the system by allocating available spots to each driver individually. The proposed
parking location is the closest to the destination, reducing overall trip time. Alfonsetti et al. [14] proposed PAS
based on the walking distance as constraints. The authors here proposed to model the global social benefit of
users by developing a distributed algorithm based on Lagrange's theory of duality. A parking assignment model
is presented in [15]. The authors suggest a variable neighborhood search-based heuristic to generate
approximations for larger instances as well as a 0-1 programming to compute exact solutions. Sadhukhan [16]
proposed an electronic parking management system (E-parking), based on parking space monitoring according
to the time lag between the sending and the reception of signals which means a reservation request. They have
developed an application on which users send their reservation requests via graphical user interface (GUI). It
enables the user to learn about the availability of different parking facilities across the city.
Kotb et al. [17] proposed a PAS for parking management named i-parking. This approach aims to
achieve three goals: i) increasing parking revenue, ii) increasing the use of parking resources, and iii) minimizing
cost and walking time. Their system is based on mathematical modeling using MILP. They used the IBM ILOG
CPLEX (CPLEX) software to solve the problem. Moreover, they came up with scalability techniques for their
system like grouping the number of resources, splitting the area, and controlling reservations. An intelligent
parking management system is proposed in [18], which exploits detection technologies like ground sensors and
LED devices. The mathematical modeling of the proposed assignment problem is made by MILP, which
guarantees a feasible solution by satisfying the constraints. Rahayu and Mustapa [19] presented a secure
technology global system for mobile communications (GSM)-based parking place reservation system. They
carry out two architectures for two modules: parking reservation and monitoring using GSM. The purpose is
to ensure the security of the vehicles in the parking lot, enhance parking management efficiency, and reduce
the parking search times. The reservation procedure was analyzed in [20] as a resource allocation problem.
Their MILP-based solution has the goal of reducing the tracking cost. The model offers real-time bookings
with fixed prices. An auction-based reservation system is proposed in [21]. To estimate the changes in
parking time following altering parking prices, the authors provided a parking time model based on the study
of real parking data. They compared four parking management strategies: i) the typical no-reservation
parking scheme; ii) a parking reservation system that includes auctions; iii) the increase in parking prices due
to the use of a reservation system; and iv) reduced parking expenses with the use of reservations.
Shiyao et al. [22] carried out a parking management system that measures the occupancy status of
parking spaces using pressure sensors and the ZigBee sensor network to determine the status of parking
spaces' occupancy. Reservations through a website were included to reduce overall parking problems.
Nevertheless, the study focused mostly on technologies or the communication medium between drivers and
systems and did not address a model or algorithm that efficiently handles booking in such a way that
considerable benefits were produced. Abidi et al. [23] presented a hybrid genetic algorithm to solve an
parking slot assignment problem for groups of drivers (PSAPG) respecting the following constraints: the

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2374-2385

Int J Artif Intell ISSN: 2252-8938  2377

walking distance between the parking lot and the destination, expected parking cost, and time restriction.
Their mathematical modeling is based on linear optimization problems. Caliskan et al. [24] focused on
establishing a system to monitor parking occupancy status and providing a predictive model of parking
occupancy to improve the prediction of free space in parking lots. Hanif et al. [25] presented a new
reservation system via short message service (SMS). The authors have only proposed a communication
system by SMS by integrating the micro-remote terminal unit (RTU) and the microcontroller and not a
process of reservation of place for the allocation of parking lots. Errousso et al. [26] created an integrative
method based on fuzzy logic to address the problem of parking allocation for road users, whether they are
professionals or regular citizens. The method is based on processing parking requests from all areas of a city
and redistributing them to balance the occupation load of the different regions. Zhang et al. [27] proposed an
approach for detecting parking locations based on the deep convolutional neural network (DCNN). A
large-scale labeled data set was created to aid in the investigation of vision-based parking place recognition.
This approach has confirmed good effectiveness and efficiency through comprehensive testing.
This related work shows that the parking problem is generally treated as an assignment problem.
The majority of studies were modeled based on the MILP method. However, only a few studies have
considered the execution time of problem-solving despite its usefulness and importance in assignment
problem modeling. On the other hand, the majority of the aforementioned technologies merely attempt to
address the issues of recognizing available parking lots and guiding drivers to these spaces. Furthermore,
certain methods processed reservation requests in a sequential order typically using the FIFO method. When
there are many drivers reserved in numerous regions, the view is limited.

3. BACKGROUND
Constraint programming (CP) is a fascinating field that revolves around the principles of
computational systems grounded in the art of constraint satisfaction. At its core, CP provides a powerful
framework for expressing complex problems in a declarative manner, akin to mathematical modeling. What
sets CP apart is its remarkable ability to tackle a wide range of intricate optimization challenges, often
characterized by combinatorial intricacies. In essence, CP empowers us to focus on defining the rules and
relationships within a problem, allowing the solver to navigate the labyrinth of possibilities and find optimal
solutions. This versatility and problem-solving prowess have made CP an indispensable tool in domains
ranging from logistics and scheduling to artificial intelligence and beyond, offering insights and solutions
that were previously elusive.

3.1. Constraint satisfaction problem

CP deals with the CSP. CSP is a mathematical problem [20]. It is made up of a set of values called
domains to be assigned to variables that satisfy a particular set of constraints. Operations research and
artificial intelligence both conduct extensive studies on CSP. Formally speaking, we have. The triplet
(X, D, and C) defines a CSP where:
- 𝑋 = {𝑋1 , … , 𝑋𝑛 } is the set of variables;
- 𝐷 = {𝐷1 , … , 𝐷𝑛 } is the set of domains, where 𝐷𝑖 is the finite set of possible values for the variable 𝑋𝑖 ; and
- 𝐶 = {𝐶1 , … , 𝐶𝑚 } is a set of constraints. Each 𝐶𝑖 is defined by the set 𝑋𝑖= {𝑋𝑖1 , … , 𝑋𝑖𝑘 } of k variables
involved in 𝐶𝑖 .
Concerning the resolution algorithms, we distinguish two main classes. The first is systematic,
complete, and extends a set of consistent values for a portion of the problem's variables to find a solution,
appending consistent values for other variables continuously until a complete solution is provided, such as
the generate and test, backtracking, and forward checking algorithms. The second class of algorithms are
stochastic, incomplete, and attempt to find a solution by fixing an inconsistent set of values for all variable.
changing an inconsistent value for one variable repeatedly until a comprehensive answer is obtained, as the
case of the tabu search [28], genetic algorithm [29], and swarm intelligent algorithms [30]. However, we can
find hybrid algorithms where the extension and repair methods are merged.

3.2. Constraint optimization problem

A constraint optimization problem (COP) extends a CSP by introducing an objective function that
needs to be optimized. In a CSP, the goal is to find any valid assignment of values to variables that satisfies
the given constraints. In contrast, in a COP, we add an objective function that quantifies solution quality and
can be either maximized or minimized. To solve a COP, the task is twofold:
- Satisfy constraints: as in CSPs, values must be assigned to variables to satisfy all imposed constraints,
ensuring the solution adheres to specified rules.

Enhancing the smart parking assignment system through constraints optimization (Nihal Elkhalidi)
2378  ISSN: 2252-8938

- Optimize the objective function: simultaneously, the aim is to find the best assignment of values to
variables that either maximizes or minimizes the objective function, achieving the desired criteria, such as
maximizing profit or minimizing cost.

3.3. Mixed-integer linear programming

Mixed integer programming (MIP) is used to solve several real-world problems such as scheduling,
supply chain management, and production planning. When the objective function and the constraints are
linear, the problem became MILP. In a MILP, the feasible decision space is described by a set (or some) of
affine constraints and the decision variables are integers. We consider that 𝑓 𝑇 𝑥 is the objective function,
𝐴𝑥 ≤ 𝑏 are the linear constraints, l and u are simple lower and upper bounds on the problem variables x. When
the integer variable bounds are {0, 1}, they called binary variables. Therefore, a MILP can be stated as:

min 𝑓 𝑇 𝑥 (1)

𝐴𝑥 ≤ 𝑏 (2)

𝑙≤𝑥≤𝑢 (3)

𝑥 ∈ ℝ𝑛 (4)

𝑥𝑗 ∈ ℤ ∀𝑗 ∈ 𝐼 (5)

where: 𝐴 ∈ ℝ𝑚×𝑛 , 𝑓 ∈ ℝ𝑛 , 𝑏 ∈ ℝ𝑚 , 𝑙 ∈ (ℝ ∪ {−∞})𝑚 , 𝑙 ∈ (ℝ ∪ {−∞})𝑚 , and 𝐼 ⊆ ℕ = {1, … , 𝑛}.

Solving MIP and CP problems necessitates a deep and potentially exhaustive search. The "classic"
way to solve these problems is called branch and bound. Land and Doig [31] were the first to propose the
branch and bound algorithm in 1960, Gupta and Ravindran [32] suggested a branch-and-bound
implementation for nonlinear convex problems in 1985 and investigated the implications of problem size on
resolution time, as well as various branching and node selection techniques. Branch and bound algorithms
consist of exploring the entire search space of possible solutions and providing an optimal solution. Indeed,
the approach starts by identifying the best answer to the "relaxation" of the problem without the integer
constraints. If the decision variables with integer constraints in this solution take integer values, then no
additional work is required. In the case of incompleteness of the solution, i.e. if one or more of the variables
do not have an integral solution, the branch and bound method chooses such a variable and "branches", to
generate two new subproblems in which the value of this variable is more strictly constrained. These
subproblems are solved, and the procedure is continued until a solution that meets all of the integer
requirements appears.

4. MATERIALS AND METHODS

4.1. A new parking assignment system using multi-agent system
This section outlines the methodology conducted to propose a parking slot reservation. This service
is based on a MASPMS. MAS are systems that have shown great efficiency in modeling distributed,
coordinated and complex systems [33], and many management systems are based on them. Indeed, agents
can cooperate to provide services to the end user. An agent's actions are determined by a set of rules that
represent the desired control policies.
Figure 1 presents the chart flow of the proposed system based on eight agents. The parking agent
(PA) checks the parking status and transfers the availability data at each change to the RA. RA receives the
reservation request from the drivers agent (DA) and sends the collected information to the CA, which is the
central server responsible for parking slot assignment, it collects and updates the data received using the
geo-location agent (GA). When solving the parking assignment problem, the information gathered is later
used to find the answer. The processing of parking requests is triggered by the driver’s request via a personal
navigation device in a vehicle. A driver enters his destination on the personal navigation system's screen, a
desired maximum walking distance from the parking lots to the destination, and the parking request along
with the necessary information used in selecting a parking space. Furthermore, the GA completes the driver
request by determining the current location of vehicles requesting parking as well as the destination location.
Based on receiving drivers’ information and GA information, the RA execute an algorithm and proposes the
most suitable parking lot meeting the requirement of drivers.

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2374-2385

Int J Artif Intell ISSN: 2252-8938  2379

Figure 1. Diagram of agents exploitation upon receipt of the reservation request

4.2. Problem statement

In this section, we propose a mathematical programming model to assign a parking slot to drivers.
The parking reservation problem is NP-complete restricted combinatorial optimization problem [26], which
can be thought of as a resource assignment problem. One of the most appropriate models for this type of
problem is MIP as proved by the literature review. The MIP model gives a concrete description of the
problem, and once the model is developed, it can easily be adjusted to reflect changes in the problem, such as
removing or adding constraints or objectives. On the other hand, few studies in the literature have used the
CSP approach to the best of our knowledge, despite the demonstrated effectiveness in other domains such as
the airport gate assignment problem [34]. In this paper, we carry out these two parking assignment models
based on CSP and MILP methods. The reservation problem is seen as a problem of multi-variables. Multiple
users’ requests are emitted to the MASPMS, managing multiple parking lots distributed in a big city, and are
processed simultaneously and resolved in real-time while taking the needs of the drivers into consideration.
To present the problem, let us consider the set of reservation requests 𝑅 = {𝑅1 , … , 𝑅𝑛 } received over
a time interval from a group of users 𝑈 = {𝑈1 , … , 𝑈𝑛 }, who book in different destination
𝐷𝑒𝑠(𝑈) = {𝐷𝑒𝑠(𝑈1 ), … , 𝐷𝑒𝑠(𝑈𝑛 )}. The destination 𝐷𝑒𝑠(𝑈𝑖 ) is considered as a parking zone 𝑍𝑖 consisting of
m parking lots 𝑃 = {𝑃1 , … , 𝑃m }. We first list in Table 1, the parameters used in the proposed models. To
illustrate our mathematical programming model, we use many symbols. Table 1 shows the symbols and their
brief descriptions.

Table 1. Notation and symbols

Symbol Description
I 𝐼 = {1, … , 𝑛} the set of the number of user requests
J 𝐽 = {1, … , 𝑚} the set of parking lots in an area
R The set of reservation requests
Z The zone in the city
P The set of parking lots 𝑃 = {𝑝1 , … , 𝑝𝑚 }
𝒑𝒋 The jth parking lots.
𝑼𝒊 The ith user.
𝑪𝒑𝒋 The number of available parking spots in 𝑝𝑗 .
𝒙𝒊𝒋 The assignment matrix. Such as 𝑖 ∈ 𝐼 and 𝑗 ∈ 𝐽.
𝑫𝒆𝒔(𝑼𝒊) The destination of the user.
𝑳𝒐𝒄 − 𝑫𝒆𝒔(𝑼𝒊 ) The location of the users’ destination.
𝑳𝒐𝒄 − 𝒑𝒋 The location of parking lots.
𝒘𝒊 The walking distance between the proposed parking and the user destination.
𝒅𝒊𝒋 The driving distance between the current location of users and the proposed parking.
𝑪𝒖𝒓𝒓𝒆𝒏𝒕 − 𝒍𝒐𝒄(𝑼𝒊 ) The current location of users.
𝒅𝒕𝒊 The stay time of users 𝑈𝑖 in parking lots ∀𝑖 ∈ 𝐼.
𝑷𝒓𝒋 The parking rate of 𝑝𝑗 , ∀𝑗 ∈ 𝐽.

Enhancing the smart parking assignment system through constraints optimization (Nihal Elkhalidi)
2380  ISSN: 2252-8938

4.3. Constraint optimization problem formulation

In this section, we develop the mathematical model of the parking assignment problem using CP.
The main process is to define the variables and their corresponding domains, and then determine the problem
constraints. The objective function must be stated if a criterion is to be optimized. We model our problem as
a COP as follows:
a) The finite set of variables is defined by two sub-sets X and P
− 𝑋 = {𝑋1 , … , 𝑋𝑛 } represents the set of users who send their reservation request at the same time.
− 𝑃 = {𝑃1 , … , 𝑃𝑚 } represents the set of parking lots in the requested zone.
b) The possible values of variables X and P are defined by:

𝐷(𝑋𝑖 ) = {1, … , 𝑚} ; 𝐷(𝑃𝑗 ) = {1,0} ; ∀𝑖 ∈ 𝐼 𝑎𝑛𝑑 𝑗 ∈ 𝐽

c) The problem’s constraints are defined by three constraints

− C1-parking capacity restriction

|𝑋𝑖 ; 𝑖 ∈ 𝐼 ∶ 𝑋𝑖 = 𝑗| ≤ 𝐶𝑝𝑗 ; ∀𝑗 ∈ 𝐽 (6)

The significance of this particular constraint cannot be overstated, as it plays a pivotal role in
guaranteeing the integrity of the parking allocation system. This constraint serves as a safeguard to ensure
that the number of users assigned to a specific parking spot, denoted as 'j,' does not surpass the available
number of free parking spots in that designated area. In essence, it acts as a check and balance mechanism
within the system. Without such a constraint, there would be a risk of overloading a parking area beyond its
capacity, leading to potential chaos and inefficiency. By imposing this constraint, the system optimally
manages the allocation process, preventing any scenario where users are assigned to a parking spot that is
already at full capacity. This careful consideration of user-to-parking spot allocation aligns with the
fundamental goal of ensuring a fair, organized, and efficient parking management system. It minimizes the
likelihood of congestion and frustration while maximizing the utilization of available parking spaces,
ultimately enhancing the overall user experience and the functionality of the parking facility.
− C2-parking slots restriction

𝑃𝑋𝑖 = 1; ∀𝑖 ∈ 𝐼 (7)

The concept of constraints plays an essential role in the smooth operation of systems involving the
allocation of parking spaces to users. In this context, constraint effectively establishes a vital link between
parking spaces and the people who use them. In the context of this allocation problem, a parking lot is
denoted by "𝑃". Each element of this parking lot, represented by "𝑋𝑖 ", corresponds to a specific user, with "i"
designating his position in the parking lot. The essence of this constraint lies in the fact that these variables
must take the value "one". This singular value means that the parking space has been allocated to the "ith"
user, which sums up a crucial aspect of the allocation process. By respecting this constraint, the parking
space allocation system ensures that each user is precisely associated with the parking space allocated to
them, thus improving the efficiency and organization of the parking space management process.
− C3-walking distance restriction

𝑑𝑖𝑠(𝐿𝑜𝑐 − 𝐷𝑒𝑠(𝑈𝑖 ), 𝑙𝑜𝑐 − 𝑝𝑗 ) ≤ 𝑤𝑖 ; ∀𝑖 ∈ 𝐼 𝑎𝑛𝑑 ∀𝑗 ∈ 𝐽 (8)

d) The objective function

The overarching objective in this context is to achieve a minimization of a specific cost function.
This cost function serves as a quantitative representation of the various factors and considerations at play
within the problem-solving scenario. The fundamental aim is to reduce this cost function to its lowest
possible value, which signifies the optimal outcome or solution:

𝐹(𝑋) = 𝑚𝑖𝑛 ∑𝑛𝑖=1 𝑑𝑖,𝑋𝑖 + 𝑑𝑡𝑖 𝑃𝑟𝑋𝑖 (9)

Recall that,
- |𝐴| of a set A denote the cardinal (the number of element) of the set A.
- The 𝑑𝑖𝑠(𝑥, y) is the distance between x and y. This constraint ensures that the distance from the proposed
parking does not exceed the maximal walking distance for each user.

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2374-2385

Int J Artif Intell ISSN: 2252-8938  2381

- D = (di,j ) = 𝑑𝑖𝑠(𝐶𝑢𝑟𝑟𝑒𝑛𝑡 − 𝑙𝑜𝑐(𝑈𝑖 ), 𝐿𝑜𝑐 − 𝑝𝑗 ) the matrix of the distance of user current location
i,j∈N×M
i to the parking j and 𝑃𝑟𝑋𝑖 is parking prices to be paid by the user.

4.4. Mixed-integer linear programming formulation

As previously highlighted, our analysis revolves around two fundamental sets: the set of users or
requests, denoted as 'U' and represented as 𝑈 = {𝑈1 , … , 𝑈𝑛 }, and the set of available parking lots, denoted as
'P' and represented as 𝑃 = {𝑝1 , … , 𝑝𝑚 }, all of which are conveniently situated within the same geographical
zone. In the pursuit of an effective and well-structured parking allocation system, our model meticulously
incorporates a series of constraints. These constraints are the guiding principles that shape the allocation
process and ensure that it aligns with our overarching goals and requirements:
- Restriction on parking spaces: only one car may be parked in a lot at a time.

∑𝑚
𝑗=1 𝑥𝑖𝑗 = 1 ; ∀𝑖 ∈ 𝐼 (10)

- Parking capacity limitation: the number of vehicles assigned to the parking lot cannot exceed the lot's
available capacity.

∑𝑛𝑖=1 𝑥𝑖𝑗 ≤ 𝐶𝑝𝑗 ; ∀𝑗 ∈ 𝐽 (11)

- Permissible walking distance: the proposed car park's distance from the users' destination must not exceed
the maximum allowable walking distance.

𝑑𝑖𝑠(𝐿𝑜𝑐 − 𝐷𝑒𝑠(𝑈𝑖 ), 𝑝𝑗 ) ≤ 𝑤𝑖 ; ∀𝑖 ∈ 𝐼 (12)

The solution is determined by the assignment matrix:

𝑋 = (𝑥𝑖𝑗 )𝑁×𝑀 ; 𝑥𝑖𝑗 ∈ {0,1} (13)

At the core of our model lies a clear and compelling objective: the minimization of a function
denoted as 𝐹(𝑋). This function serves as the compass that guides our decision-making process, and it is
meticulously composed of two key components that capture the essence of our problem: travel distances and
associated fees imposed on the vehicles involved. By seeking to minimize the function 𝐹(𝑋), we strike a
balance between two critical dimensions: minimizing the physical and logistical costs associated with vehicle
travel while also minimizing the financial costs incurred by users. In doing so, our model aims to provide an
optimal and sustainable solution that enhances the overall efficiency and affordability of the parking
allocation process, ultimately benefiting both users and the broader community.

𝐹(𝑋) = 𝑚𝑖𝑛 ∑𝑖,𝑗(𝑥𝑖𝑗 . 𝑑𝑖𝑗 + 𝑥𝑖𝑗 . 𝑑𝑡𝑖 . 𝑃𝑟𝑗 ) (14)

4.5. Solvers and optimizer

To solve the problem based on MILP, we use the Google OR-Tools solvers and Gurobi optimizer
which have repeatedly been awarded in international competitions of MiniZinc solvers [35]. Google
OR-Tools is an open-source, rapid, and adaptable software package that is used to solve combinatorial
optimization problems. Dedicated to solve problems based on CSP, linear programming (LP), and MILP.
Google OR-Tools [36] is created in C++ but also provides wrappers in Python, C#, and Java.
Gurobi optimizer is an advanced solver for mathematical programming, and the corresponding
solvers were designed from the ground up to exploit modern architectures and multi-core processors, using
the most advanced implementations of the latest algorithms including exhaustive search with the branch and
bound algorithm. Gurobi optimizer [37] is a solver with efficient parallel algorithms for large-scale linear
programs, quadratic programs, MILP, and mixed-integer quadratic CP. To solve the problem based on the
COP approach, we adopt the Choco solver, which is a free and open-source Java library. Users model their
problems declaratively by declaring all the constraints that must be satisfied in each solution by taking into
account a cost function to be optimized. Then, the problem is solved by filtering algorithms based on
constraints alternated with a search mechanism. The first version of Choco-solver in Java was released in
2003. Since then, it has undergone several major revisions, most recently in 2013 [38].

Enhancing the smart parking assignment system through constraints optimization (Nihal Elkhalidi)
2382  ISSN: 2252-8938

5. EXPERIMENT AND RESULT

5.1. Simulation parameters
We analyze the performance of the proposed models in this section by using solvers of COP and
MILP approaches. The study was carried out in order to assess the effectiveness of the COP and MILP
solvers using the city of Casablanca as a real use case. We performed a test of statistical significance in order
to choose the best solver that was significantly efficient and ensure the scalability of the assignment system.
Based on three basic settings, we built three datasets of three instances for each set, with varying degrees of
difficulty: 30, 50, and 100 for the number of drivers looking for available parking in an area of 40 car parks.
We found that the three solvers offer almost the same total cost of the objective function. This prompted us to
do a second experiment to assess the scalability of the approach and see which solution is optimized by
scaling to 200, 500, and 1000 users. The area considered in the test is assumed to consist of 100 parking lots.
The experiments presented here focus on how long it takes to solve the problem by using the
computation times of our approach parameter and the scalability of the approach. In fact, in constrained
programming solvers, optimization is done by computing improved solutions until an optimum is reached. So
this can be thought of as solving the model multiple times while adding constraints along the way to prevent
the solver from computing a dominant solution. The optimization proceeds as follows: each time a solution is
found, store the value of the target variable and issue the cut. The cut is an additional constraint stating that
the following solution must be (strictly) better than the previous one. The simulation data was taken from
Casablanca city of Morocco, which is considered the largest and crowded city in Morocco and even in the
Maghreb Region. The City of Casablanca shares a similar context with many cities in the world, which
makes it possible to effectively draw important results.

5.2. Results and analysis

The experiments were conducted according to the number of requests that created three instances
(30, 50, and 100). In each instance, a random number is used to determine the users' destination. The drivers'
locations are produced within a 300-meter radius of the destination. The driving distance (between the
driver's position and the parking lot) and the walking distance (between the parking lot and the destination)
can then be determined. The performance of the proposed models is assessed using four parameters:
- Total cost of the objective function: this gives the overall cost values of distance plus the execution time.
The lower value of this parameter is preferable since it can eliminate unnecessary time and energy
consumption and detrimental effects like air pollution.
- Execution time by seconds.
- Rate cost: the cost of parking according to the stay time of users
- Distance cost: the sum of the driven distance of users from their current positions to the proposed parking
lots.
In Table 2, the significance test and performance measure results are provided and succinctly described.

Table 2. Solvers comparison of the first test with 30, 50 and 100 users
Solvers Instance Total cost Time (s) Rate cost (dhs) Distance cost (m)
Google OR-Tools 1=30 7,091 0.033 548 6,543
2=50 13,231 0.054 822 12,461
3=100 31,231 0.127 1,542 26,689
Gurobi optimizer 1=30 7,091 0.014 548 6,543
2=50 13,283 0.017 822 12,461
3=100 31,231 0.032 1,542 29,689
Choco 1=30 7,091 0.3 548 6,543
2=50 13,283 0.4 822 12,461
3=100 31,536 0.5 1,561 29,975

Table 3 shows that the results achieved by the solvers are close. Indeed the three solvers give similar
results in the two instances (30 and 50 users) regarding the total cost and the sum of each component of the
objective function. The third instance of 100 users shows a nuance regarding the total cost. The Choco solver
offers a higher total cost than the Gurobi and OR tools solvers. Furthermore, the best running time for the
Gurobi solver, which is presented in bold, is readily apparent. The results of the first experiment do not show
a significant difference between the solvers, which prompted us to verify if all approaches meet the
scalability. This is why we proposed to increase the number of requests set by fixing the number of parking
lots in the target zone at 100 parking lots. In the second simulation, we created three new instances based on

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2374-2385

Int J Artif Intell ISSN: 2252-8938  2383

three different setups: 200, 500, and 1000 parking-seeking drivers. The results obtained out of 30
independent runs are shown in Table 3. The comparison is based on the following parameters:
- Min-max total cost/run time: the minimum-maximum total cost of the objective function (the driving
distance + the rate cost) of the 30 execution of the problem. The minimum execution time obtained from
30 runs.
- Average of total cost/run time: the sum of the total costs obtained during the 30 runs divided by the
number of executions (30)/the sum of the run times obtained during the 30 runs divided by the number of
executions.
- Type of solution: there are two types of solutions: i) optimal, which offers the best solution while
respecting constraints and ii) satisfiability (often abbreviated SAT) that respects constraints without
searching for the best solution.
The result of the second experiment showed that the Choco solver reached only satisfiability
solutions, contrary to Gurobi and OR-tools solvers that offer optimal solutions in a reasonable execution
time. Moreover, in the last instance (1000 users); the Choco solver did not even manage to find a satisfiable
solution. In terms of execution time, we can plainly see that Gurobi optimizer is the best in all three cases. In
addition, the results of the average of the total cost obtained by Gurobi and OR-tools are close and much
better compared to those of Choco. The results exhibit that the Choco solver is poor in solving assignment
problems, when it comes to large datasets, it does not even give a response when the number of users exceeds
a thousand. This means that Gurobi and OR tools solvers are much better in terms of scalability.
Additionally, among these, we can see that Gurobi is faster to run and find the optimal solution which is an
important criterion for real-time parking lot assignment.

Table 3. Solvers comparison of the second test with 200, 500, and 1000 users
Minimum Maximum Average of Average of Type of
Solvers Instance
Total cost Run time (s) Total cost Run time (s) total cost run time (s) solution
Google 1=200 212,423 4.059 222,494 5.218 216,396.61 4.248 Optimal
OR-Tools 2=500 537,141 13.64 458,551 14.798 512,215.66 14.132
3=1000 1,074,556 28.8 1,091,819 28.810 1,080,809 26.809
Gurobi 1=200 211,773 0.166 225,213 0.156 216,949.83 0.154 Optimal
optimizer 2=500 531,710 0.421 547,389 0.374 538,944 0.451
3=1000 1,073,680 0.859 1,093,034 0.890 1,082,698 0.874
Choco 1=200 443,166 67.55 500,903 68.45 465,805.04 67.63 SAT
2=500 1,146,909 376.5 1,205,056 374.41 1,169,615.15 375.18
3=1000 - - - - - -

6. CONCLUSION
In this work, we have proposed a complete description of our reservation system based on the MAS
for intelligent parking management for groups of drivers reserving at the same time. A modeling of the smart
parking assignment problem based on CP and MILP is proposed in this paper. In the resolution of CP, we
used a solver called Choco solver. It was chosen because it is an open-source library. It can manipulate a
large number of variables. In this library, more than 70 constraints are defined. Moreover, for the resolution
of the MILPs we have chosen two best solvers according to the MiniZinc 2022 competition named Gurobi
optimizer and Google OR-Tools solvers. The results of the two-simulation test showed that the Gurobi
optimizer performs well in terms of scalability. As a future direction, we intend to add more constraints to the
problem such as the preferred parking places. We also plan to propose an algorithm for solving constrained
assignment problems based on other optimization algorithms.

REFERENCES
[1] S. Charmonman and P. Mongkhonvanit, “Internet of things in e-business,” in The 10th International Conference on e-Business
(iNCEB2015), 2015, pp. 1–9.
[2] S. Kumar and P. R. Cohen, “Towards a fault-tolerant multi-agent system architecture,” in Proceedings of the International
Conference on Autonomous Agents, ACM, 2000, pp. 459–466, doi: 10.1145/336595.337570.
[3] N. El Khalidi, F. Benabbou, and N. Sael, “Distributed parking management architecture based on multi-agent systems,” IAES
International Journal of Artificial Intelligence, vol. 10, no. 4, pp. 801–809, 2021, doi: 10.11591/ijai.v10.i4.pp801-809.
[4] E. Simhon, C. Liao, and D. Starobinski, “Smart parking pricing: A machine learning approach,” in 2017 IEEE Conference on
Computer Communications Workshops, INFOCOM WKSHPS 2017, IEEE, 2017, pp. 641–646, doi:
10.1109/INFCOMW.2017.8116452.
[5] X. Zhang, C. Zhao, F. Liao, X. Li, and Y. Du, “Online parking assignment in an environment of partially connected vehicles: a
multi-agent deep reinforcement learning approach,” Transportation Research Part C: Emerging Technologies, vol. 138, pp. 1–22,
2022, doi: 10.1016/j.trc.2022.103624.

Enhancing the smart parking assignment system through constraints optimization (Nihal Elkhalidi)
2384  ISSN: 2252-8938

[6] S. Abidi, S. Krichen, E. Alba, and J. M. Molina, “A new heuristic for solving the parking assignment problem,” Procedia
Computer Science, vol. 60, no. 1, pp. 312–321, 2015, doi: 10.1016/j.procs.2015.08.132.
[7] K. Marriott and P. J. Stuckey, Programming with constraints: an introduction. Cambridge, Massachusetts: The MIT Press, 1998,
doi: 10.7551/mitpress/5625.001.0001.
[8] T. Kleinert, M. Labbé, I. Ljubić, and M. Schmidt, “A survey on mixed-integer programming techniques in bilevel optimization,”
EURO Journal on Computational Optimization, vol. 9, pp. 1–21, 2021, doi: 10.1016/j.ejco.2021.100007.
[9] X. Zhao, K. Zhao, and F. Hai, “An algorithm of parking planning for smart parking system,” in Proceedings of the World
Congress on Intelligent Control and Automation (WCICA), IEEE, 2015, pp. 4965–4969, doi: 10.1109/WCICA.2014.7053556.
[10] O. T. T. Kim, N. H. Tran, C. Pham, T. Leanh, M. T. Thai, and C. S. Hong, “Parking assignment: minimizing parking expenses
and balancing parking demand among multiple parking lots,” IEEE Transactions on Automation Science and Engineering, vol.
17, no. 3, pp. 1320–1331, 2020, doi: 10.1109/TASE.2019.2948200.
[11] M. Ratli, A. A. El Cadi, B. Jarboui, and A. Artiba, “Dynamic assignment problem of parking slots,” in 2019 International
Conference on Industrial Engineering and Systems Management (IESM), Shanghai, China, 2019, pp. 1-6, doi:
10.1109/IESM45758.2019.8948174.
[12] Y. Zhang, Z. Zhang, and S. Teng, “Adaptive ant colony optimization for solving the parking lot assignment problem,” in ICNC-
FSKD 2018 - 14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, IEEE, 2018, pp.
91–96, doi: 10.1109/FSKD.2018.8687175.
[13] A. Hakeem, N. Gehani, X. Ding, R. Curtmola, and C. Borcea, “On-the-fly curbside parking assignment,” in MobiCASE 2016 - 8th
EAI International Conference on Mobile Computing, Applications and Services, ACM, 2016, pp. 1–10, doi: 10.4108/eai.30-11-
2016.2267101.
[14] E. Alfonsetti, P. C. Weeraddana, and C. Fischione, “A semi distributed approach for min-max fair car-parking slot assignment
problem,” arXiv-Mathematics, pp. 1–15, 2014.
[15] M. Mladenović, T. Delot, G. Laporte, and C. Wilbaut, “The parking allocation problem for connected vehicles,” Journal of
Heuristics, vol. 26, no. 3, pp. 377–399, 2020, doi: 10.1007/s10732-017-9364-7.
[16] P. Sadhukhan, “An IoT-based e-parking system for smart cities,” in 2017 International Conference on Advances in Computing,
Communications and Informatics, ICACCI 2017, IEEE, Sep. 2017, pp. 1062–1066, doi: 10.1109/ICACCI.2017.8125982.
[17] A. O. Kotb, Y. C. Shen, X. Zhu, and Y. Huang, “iParker-a new smart car-parking system based on dynamic resource allocation
and pricing,” IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 9, pp. 2637–2647, Sep. 2016, doi:
10.1109/TITS.2016.2531636.
[18] Y. Geng and C. G. Cassandras, “A new smart parking system infrastructure and implementation,” Procedia - Social and
Behavioral Sciences, vol. 54, pp. 1278–1287, 2012, doi: 10.1016/j.sbspro.2012.09.842.
[19] Y. Rahayu and F. N. Mustapa, “A secure parking reservation system using GSM technology,” International Journal of Computer
and Communication Engineering, pp. 518–520, 2013, doi: 10.7763/ijcce.2013.v2.239.
[20] K. C. Mouskos, J. Tavantzis, D. Bernstein, and A. Sansil, “Mathematical formulation of a deterministic parking reservation
system (PRS) with fixed costs,” in 2000 10th Mediterranean Electrotechnical Conference. Information Technology and
Electrotechnology for the Mediterranean Countries, Lemesos, Cyprus, 2000, pp. 648-651, doi: 10.1109/MELCON.2000.880017.
[21] S. Hashimoto, R. Kanamori, and T. Ito, “Auction-based parking reservation system with electricity trading,” in Proceedings -
2013 IEEE International Conference on Business Informatics, IEEE CBI 2013, IEEE, 2013, pp. 33–40, doi:
10.1109/CBI.2013.14.
[22] C. Shiyao, W. Ming, L. Chen, and R. Na, “The research and implement of the intelligent parking reservation management system
based on ZigBee technology,” in Proceedings - 2014 6th International Conference on Measuring Technology and Mechatronics
Automation, ICMTMA 2014, IEEE, 2014, pp. 741–744, doi: 10.1109/ICMTMA.2014.182.
[23] S. Abidi, S. Krichen, E. Alba, and J. M. M. Bravo, “A hybrid heuristic for solving a parking slot assignment problem for groups
of drivers,” International Journal of Intelligent Transportation Systems Research, vol. 15, no. 2, pp. 85–97, 2017, doi:
10.1007/s13177-016-0123-1.
[24] M. Caliskan, A. Barthels, B. Scheuermann, and M. Mauve, “Predicting parking lot occupancy in vehicular ad hoc networks,” in
IEEE Vehicular Technology Conference, IEEE, Apr. 2007, pp. 277–281, doi: 10.1109/VETECS.2007.69.
[25] N. H. H. M. Hanif, M. H. Badiozaman, and H. Daud, “Smart parking reservation system using short message services (SMS),” in
2010 International Conference on Intelligent and Advanced Systems, IEEE, Jun. 2010, pp. 1–5, doi:
10.1109/ICIAS.2010.5716179.
[26] H. Errousso, J. El Ouadi, E. A. A. Alaoui, S. Benhadou, and H. Medromi, “A hybrid modeling approach for parking assignment in
urban areas,” Journal of King Saud University - Computer and Information Sciences, vol. 34, no. 6, pp. 2405–2418, 2022, doi:
10.1016/j.jksuci.2020.11.006.
[27] L. Zhang, X. Li, J. Huang, Y. Shen, and D. Wang, “Vision-based parking-slot detection: A benchmark and a learning-based
approach,” Symmetry, vol. 10, no. 3, pp. 1–18, 2018, doi: 10.3390/SYM10030064.
[28] G. Barbarosoglu and D. Ozgur, “A tabu search algorithm for the vehicle routing problem,” Computers and Operations Research,
vol. 26, no. 3, pp. 255–270, 1999, doi: 10.1016/S0305-0548(98)00047-1.
[29] M. Srinivas and L. M. Patnaik, “Genetic algorithms: a survey,” Computer, vol. 27, no. 6, pp. 17–26, 1994, doi: 10.1109/2.294849.
[30] A. Chakraborty and A. K. Kar, “Swarm intelligence: a review of algorithms,” in Modeling and Optimization in Science and
Technologies, Springer International Publishing, 2017, pp. 475–494, doi: 10.1007/978-3-319-50920-4_19.
[31] A. H. Land and A. G. Doig, “An automatic method of solving discrete programming problems,” Econometrica, vol. 28, no. 3, pp.
497-520, 1960, doi: 10.2307/1910129.
[32] O. K. Gupta and A. Ravindran, “Branch and bound experiments in convex nonlinear integer programming.,” Management
Science, vol. 31, no. 12, pp. 1533–1546, 1985, doi: 10.1287/mnsc.31.12.1533.
[33] M. Wooldridge and N. R. Jennings, “Intelligent agents: Theory and practice,” The Knowledge Engineering Review, vol. 10, no. 2,
pp. 115–152, 1995, doi: 10.1017/S0269888900008122.
[34] B. Liang, Y. Li, J. Bi, C. Ding, and X. Zhao, “An improved adaptive parallel genetic algorithm for the airport gate assignment
problem,” Journal of Advanced Transportation, vol. 2020, pp. 1–17, 2020, doi: 10.1155/2020/8880390.
[35] MiniZinc Team, “MiniZinc–challenge,” MiniZinc Team, 2014. [Online]. Available:
https://www.minizinc.org/challenge2022/results2022.html (accessed May 20, 2023).
[36] Mizux, “OR-tools-Google optimization tools,” Github. [Online]. Available: https://github.com/google/or-tools (accessed May 26,
2023).

Int J Artif Intell, Vol. 13, No. 2, June 2024: 2374-2385

Int J Artif Intell ISSN: 2252-8938  2385

[37] Gurobi, “Gurobi/optimizer - Docker image,” Hub Docker. [Online]. Available: https://hub.docker.com/r/gurobi/optimizer
(accessed May 26, 2023).
[38] C. P. Homme and J. -G. Fages, “Choco-solver: A Java library for constraint programming,” Journal of Open Source Software,
vol. 7, no. 78, pp. 1–6, 2022, doi: 10.21105/joss.04708.

BIOGRAPHIES OF AUTHORS

Nihal Elkhalidi received the Master in Mathematical Sciences from the Ben
M'Sick Faculty of Sciences, Hassan II University of Casablanca, Morocco, in 2016, where she
is currently pursuing the Ph.D. in Computer Science. Her research interests include smart
cities, intelligent transportation systems, multi-agent systems, constraint programming, and
combinatorial optimization. She can be contacted at email: [email protected].

Faouzia Benabbou is a professor of Computer Science and member of Compute

Science and Information Processing Laboratory. She is head of the team “Cloud Computing,
Network and Systems Engineering (CCNSE)”. She received his Ph.D. in Computer Science
from the Faculty of Sciences, University Mohamed V, Morocco, 1997. His research areas
include cloud computing, data mining, machine learning, and natural language processing. She
can be contacted at email: [email protected].

Nawal Sael is a professor of Computer Science and member of Computer Science

and Information Processing Laboratory at Faculty of Science Ben M’sik (Casablanca,
Morocco). She received her Ph.D. in Computer Science from the Faculty of Sciences,
University Hassan II Casablanca, Morocco, 2013 and her engineer degree in software
engineering from ENSIAS, Morocco, in 2002. Here research interests include data mining,
educational data mining, machine learning, deep learning, and internet of things. She can be
contacted at email: [email protected].

Enhancing the smart parking assignment system through constraints optimization (Nihal Elkhalidi)