Received 24 July 2024, accepted 5 August 2024, date of publication 7 August 2024, date of current version 23 September 2024.

Digital Object Identifier 10.1109/ACCESS.2024.3440185

A Fuzzy Ranking Collaborative Multi-Tasks Data

Collection Scheme in Ubiquitous Environments
AHMED A. A. GAD-ELRAB
Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah 21589, Saudi Arabia
Department of Mathematics, Computer Science, Faculty of Science, Al-Azhar University, Cairo 11651, Egypt
e-mail: [email protected]; [email protected]

ABSTRACT In Ubiquitous Computing (UC) and the Internet of Things (IoTs), numerous devices collect and
transmit data for sensing and control. However, connecting these devices effectively to foster collaboration
is crucial for optimizing system performance. With the increasing number of connected sensing devices in
IoT, efficient task completion through collaboration becomes imperative. Therefore, selecting and assigning
sensing tasks to maximize system benefits is a significant challenge requiring resolution. To address this
challenge, this paper proposes a Data Collector Selection Method for Collaborative Multi-Tasks. This
method considers task preferences and uncertainty in data collectors’ contributions. It integrates three key
components: 1) Utilizing Fuzzy Analytical Hierarchy Process (FAHP) to determine optimal weights for task
preferences; 2) Ranking data collectors using Fuzzy Technique for Order Preference by Similarity to Ideal
Solution (FTOPSIS) based on preferences and determined weights; and 3) Introducing Contribution Density
with Ranking as a metric to assess individual data collectors’ contributions to specific tasks. Extensive
experiments validate the proposed strategy’s efficacy, demonstrating superior performance in terms of
Task Completion Rate, Total Profit, Total Reward, and Total Selected data collectors compared to existing
approaches. Overall satisfaction scores improved significantly, surpassing existing models by 31% to 61%.

INDEX TERMS Ubiquitous computing (UC), mobile crowdsourcing, data collector selection, task unit bid,
contribution density, FAHP, FTOPSIS.

I. INTRODUCTION into daily life, becoming intrinsic to our surroundings and

As computing advances, integrating into daily life through reducing reliance on direct user interactions with traditional
a growing array of affordable, interconnected devices, devices.
the demand for essential components like communication Ubiquitous computing, known as the Internet of Things
technologies, wireless networks, RFID tags, middleware, and (IoT), embeds computing power, often microprocessors,
software agents has become indispensable. This transforma- into everyday objects to enable seamless communication
tive progression has led to the emergence of a groundbreak- and autonomous tasks, reducing direct user interaction with
ing computing paradigm known as ubiquitous computing, computers [2]. This shift from standalone computers to inter-
enabling users to effortlessly coexist with computers in every connected devices marks a more sophisticated interaction
aspect of life. Ubiquitous computing [1], often referred to as model [3]. These devices, from sensors in common items to
the ‘‘third wave’’ of computing, marks a fundamental shift in wearable tech, collaborate to deliver personalized services
our approach to interacting with technology. It leads to the across various settings, enhancing digital experiences [4].
emergence of ubiquitous computing, a paradigm shift coined Advances in microprocessor technology have made sensing
by Mark Weiser in the early 1990s. Ubiquitous computing devices smaller yet more powerful, improving computing,
envisions a future where technology seamlessly blends storage, and communication capabilities [5], [6], [7]. This
has led to a proliferation of internet-connected objects,
The associate editor coordinating the review of this manuscript and significantly advancing the IoTs [8], [9], [10], significantly
approving it for publication was Renato Ferrero . contributing to the advancement of the Internet of Things
2024 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
VOLUME 12, 2024 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ 130777
 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

(IoTs) [11], [12] and expanding practical applications [13], negotiations between task publishers and data collectors may
[14], [15], [16]. result in suboptimal collaboration, impeding the development
The evolution of this trend can be primarily attributed to of large-scale IoT deployments [5], [9], [14], [35]. The
two factors: underlying reasons for this situation can be outlined as
• Proliferation of Sensing Devices: A vast number follows:
of objects and things are embedded with sensing • Resource Consumption and Security Concerns: Data
devices, forming a widespread network connected to collectors invest significant resources such as time,
the Internet of Things (IoTs) [5]. These devices are energy, communication resources, and bandwidth. Addi-
strategically deployed across various practical sensing tionally, they encounter security threats related to data
environments [17], [18], [19], serving as a solid collection. To protect their interests, data collectors are
foundation for applications relying on sensing tech- hesitant to participate in data collection without effective
nologies [20], [21], [22]. The spectrum of sensing collaboration and incentive mechanisms in place. [5],
devices includes traditional sensors used in diverse [35].
applications [23], [24], [25], from sensor nodes in • Lack of Collaborative Platforms: Many small and
wireless sensor networks [26], [27], to devices in indus- medium-sized applications lack collaborative platforms
trial production, agriculture, environmental monitoring, essential for their operation. Initiatives like Waze and
ecosystem studies, and traffic management [28], [29], WeatherLah are independently launched and managed
[30], [31], [32]. According to [5], since 2011, the global by their creators [36], [37]. However, within the
count of IoT-connected objects has surpassed the world broader scope of IoTs, numerous applications struggle to
population, reaching 9 billion. Projections suggest that establish their own platforms. Despite their abundance,
by 2020, this network will encompass approximately many applications fail to materialize due to the absence
24 billion connected devices [5]. of platforms, even when the required data volume is
• Facilitation of Compelling Applications by IoTs: The minimal [39], [40]. Furthermore, without a collaborative
proliferation of sensing devices embedded in objects and platform, many data collectors who have the necessary
connected to the Internet of Things (IoTs) enables the resources cannot contribute to sensing tasks [5], [35].
development of various compelling applications. Exam- A recent development in this field is the introduction
ples include smart grids, smart homes, and smart cities, of the first collaborative platform as proposed in [41].
where these devices collect and exchange data, facili- This platform streamlines task selection and data collec-
tating advanced sensing and control functionalities [33], tor assignment, optimizing overall system profitability
[34], [35]. while keeping costs low. However, this platform did
Various applications rely heavily on data collection and not address data collectors’ preferences or achieve a
exchange, where objects and entities equipped with sensing balanced trade-off between total system profit and data
devices contribute data to centralized systems for specific collector rewards.
purposes. For instance, the Waze project [36] utilizes • Neglecting Profit Considerations: Previous studies
real-time vehicle location data from sensors embedded in often prioritize technological feasibility over prof-
vehicles to provide up-to-the-minute traffic information to itability considerations. In reality, for collaborative
travelers. This enables efficient route planning, navigation processes involving tasks, data collectors, and platforms
around congestion, and enhances travel convenience by to succeed long-term, all participants must benefit
saving fuel and time. Another example is the WeatherLah financially [5], [35]. Designing effective collaboration
app [37], which collects weather data through crowd-sourced platforms requires a focus on maximizing profits for all
sensing devices. These devices contribute diverse meteoro- involved.
logical data to the application, empowering meteorological Some proposed strategies for profit maximization in
departments with detailed information for analysis and crowdsourcing networks may not adequately account for
dissemination. market dynamics. For example, using the number of data
However, many current applications are narrowly focused, samples as the sole incentive criterion assumes tasks
requiring dedicated platforms for their implementation. are complete once a specific data volume is reached.
This specificity presents challenges for the widespread However, different tasks require varying amounts and
adoption of crowdsourcing [36], [37]. Moreover, a sig- types of data across different times and locations.
nificant drawback in these applications is the lack of Ensuring a balanced data collection across the network
collaboration [38]. Case studies demonstrate that merely is crucial to meet sensing task requirements uniformly.
connecting objects or things without fostering collaboration Recent research has introduced the concept of data
can compromise performance or escalate system costs within coverage, emphasizing the need for comprehensive data
the Internet of Things (IoTs). In severe cases, this lack of collection across the entire area of interest. However, this
collaboration can lead to application failures [35]. Several approach often relies on single data samples, which may
factors contribute to this issue, with limited resources not align with real-world scenarios where data collection
being particularly critical. Without a structured mechanism, is continuous and mobile [42], [43], [44].

130778 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

• Balance between the System Profit and Data Collec- Additionally, we introduce Task Unit Bid with Ranking
tors’ Rewards: The profitability of the collaborative as a metric to measure the contribution of individual
platform is crucial for its sustainability, serving as a task to the platform. This method strategically selects a
bridge between sensing tasks and data collectors. The task set to maximize the system profits with acceptable
platform generates profit by selecting tasks that secure rewards.
bid income and choosing data collectors that minimize 3) Uncertainty and Task Preferences: To address
costs while offering adequate rewards. Maximizing the user requirements and the uncertainty of available
platform’s profit and balancing overall system profit resources, considering task preferences and the signif-
with data collector rewards require a comprehensive icance of each preference in the selection process is
strategy for task and data collector selection [45]. crucial. Therefore, we propose a weighting strategy for
• Satisfying Tasks’ Preferences: In this collaborative task preferences using the Fuzzy Analytical Hierarchy
platform, selecting the most suitable data collector Process (FAHP). This method determines optimal
that aligns with user requirements and task pref- weights for each preference or criteria, considering
erences across multiple criteria is essential. These their importance in the selection process.
criteria encompass mobility, waiting time, cost, trust, 4) Data Collector Selection Method Based on Fuzzy
data sharing, collaboration capability, workload, data Ranking: We propose a data collector selection
transmission efficiency, and data quality. Given the method that maximizes system platform profit in
dynamic nature of these systems, accurately predicting a greedy manner while satisfying data collectors’
all probabilities can be challenging, highlighting the requirements. The method involves using Fuzzy Tech-
importance of addressing uncertainties in contextual nique for Optimal Solution Similarity (FTOPSIS) to
data [45]. rank data collectors based on required task preferences
In the realm of the Internet of Things (IoTs), a continuous and their weights determined by FAHP. Additionally,
flow of tasks varies in attributes such as data collection time, we introduce Contribution Density with Ranking as a
location, the volume of data, and the bid each task offers. metric to measure the contribution of individual data
However, due to limitations in the number of data collectors collectors to a specific sensing task. This method strate-
and platform capacity within the crowdsourcing network, not gically selects a data collector set to maximize task
all sensing tasks can be completed. Therefore, an intelligent profits with acceptable rewards, continually calculating
approach is crucial for the platform to strategically prioritize and updating Fuzzy Contribution Density values during
tasks that maximize profits. Additionally, the platform must task execution.
select suitable data collectors to minimize costs. Designing
such a selection strategy becomes challenging, particularly II. RELATED WORK
in ensuring that data collectors with available resources In recent years, various schemes have been proposed for tasks
contribute data to the platform to earn rewards. and data management in ubiquitous and IoT environments.
In this paper, we introduce a Fuzzy Ranking Collaborative In the following section, we delve into the details of some of
Multi-Tasks Data Collection Scheme (FRCMDCS) aimed at these schemes.
addressing challenges in multi-task crowdsourcing networks To ensure an adequate number of data collectors for the
through the creation of a Fuzzy Ranking collaborative completion of sensing tasks and maintain a consistent pool
platform for task publishers and data collectors. The primary of participants, the system platform must offer a specified
innovations of this paper are as follows: reward. This monetary incentive is crucial to motivate data
1) Collaboration Platform for Balancing System Profit collectors, considering the time, energy, electricity, and
and Data Collectors’ Rewards: We introduce a financial resources involved in the process of data sensing,
collaboration platform designed to maximize sys- acquisition, and transmission. Without a proper incentive,
tem profits within multi-task crowdsourcing networks the willingness of data collectors to engage in assigned
while ensuring acceptable rewards for data collectors. tasks is likely to be limited [46], [47]. Additionally, data
Building upon this platform, we present the Collabo- collectors face security concerns, particularly related to
rative Multi-Tasks Data Collection method, facilitating privacy breaches during data collection participation [48].
the selection of tasks and data collectors. This method Given that task requirements often involve specific time
aims to optimize overall system profits, taking into and location parameters, reporting data may expose the
account the preferences of each task. privacy of the data collector. Hence, offering rewards within
2) Task Selection Method Based on Fuzzy Ranking: crowdsourcing networks becomes essential to heighten the
We propose a task selection method that maximizes enthusiasm of data collectors and encourage their active
system platform profit in a greedy manner while participation in sensing tasks.
satisfying platform requirements. The method involves In the realm of crowdsourcing, a strategic relationship
using Fuzzy Technique for Optimal Solution Similarity resembling a game exists between sensing task publishers
(FTOPSIS) to rank tasks based on required platform and data collectors. Task publishers aim to accomplish
preferences and their weights determined by FAHP. sensing tasks at a lower cost while ensuring quality, while

VOLUME 12, 2024 130779

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

data collectors aspire to secure higher rewards. Currently, using fuzzy set theory. The approach focuses on improving
incentive mechanisms in crowdsourcing networks can be context understanding and service quality within learning
categorized into two modes [49]: platform-centric and user- scenarios. However, additional considerations such as time
centric. In the platform-centric incentive mechanism, a Stack- constraints, waiting times, and data quality need to be
elberg game design is employed. The platform establishes integrated into the selection process.
a specific reward amount initially, and data collectors then Xue et al. [57] introduced a data management system for
compete for task execution based on this predetermined context-aware computing, featuring a gateway module as a
reward. The Stackelberg equilibrium is ultimately reached standardized interface to collect context data from various
under the platform’s guidance. In the user-centric incentive physical spaces across different domains. This system utilizes
mechanism, users (referred to as data collectors in this paper) semantic P2P networks to organize physical spaces with sim-
exercise more active control over compensation [50]. ilar context attributes, enhancing context query handling and
User-centric incentive mechanisms operate on auction data acquisition methodologies. However, the system has lim-
theory, where users, or data collectors in this context, engage itations concerning time constraints, waiting times, and data
in sensing tasks through auction bids. However, this approach quality, which warrant further consideration and refinement.
may not be suitable for applications with stringent real- Despite the proliferation of IoT devices in crowdsourcing
time requirements. In study [51], the relationship between networks, effectively allocating and scheduling sensing tasks
data collectors and task publishers is examined, integrating remains a significant challenge. In these networks, multiple
a social reputation mechanism in the crowdsourcing network sensing tasks create a competitive environment where not all
to enhance platform efficiency. Another study [52] combines tasks can be completed simultaneously due to the limited
online and offline incentives, initially selecting a candidate number of data collectors available.
data collector statically and determining the final winner To address this challenge, researchers in [58] proposed
through dynamic bidding. However, these models do not a multi-task crowdsourcing model aimed at improving task
account for task execution order. Research [53], introduces completion reliability. The model achieves this by assigning
an online auction model aimed at completing sensing sensing tasks to data collectors multiple times, thereby
tasks within budget constraints and before deadlines. Yet, minimizing the total cost of the sensing system. However,
uncertainties in data collector arrivals may hinder timely task it does not consider the competitive dynamics between
completion. multiple tasks. While multiple assignments can enhance task
Van Bunningen et al. [54] conducted a thorough examina- quality and reliability, they also increase system storage usage
tion of context, exploring its influences and the requirements and introduce significant computational overhead.
for computer systems designed to be context-aware, espe- In [59], a task allocation method is considered for quality
cially those focusing on data management within such assurance, employing F-score for task assignment. However,
frameworks. A platform was developed and applied to this research overlooks task performing order and task bids,
address context implications in data management, with a falling short of maximizing system profits.
specific focus on location information. However, a limitation In [60], a framework is introduced for the adaptive
of this approach is its oversight of factors such as data allocation of sensing tasks, ensuring task quality through
collection time, waiting times, data quality, and energy predefined indicators. Additionally, [61] addresses privacy
consumption. Moreover, the dynamic nature of extensive concerns in sensing task assignments by concealing the
computing environments introduces unpredictability, high- location information of data collectors during the assignment
lighting the need to consider and manage uncertainty in of location-related sensing tasks.
contextual data. Ren et al. [41] introduced a framework aimed at optimizing
Gad-Elrab et al. [55] introduced ASDGA, a novel service system profits using the CMDCS data collection strategy.
selection approach for ubiquitous environments. ASDGA, This framework identifies task publishers and data collectors,
known as Adaptive Service Distribution Genetic Algorithm prioritizes task execution through a task unit display, and
with Multi-Constraints, effectively selects service providers focuses on selecting datasets that maximize profits while
based on user criteria, provider capabilities, and user trust. minimizing costs. However, the framework overlooks factors
Simulations show ASDGA is energy-efficient and distributes like waiting time, data quality, and energy consumption,
service requests equitably among providers by considering which are crucial for comprehensive computing environ-
their current loads. However, ASDGA does not maximize ments. Managing uncertainty in contextual data remains
system profits from service implementation and overlooks essential due to the dynamic nature of these environments.
factors like data collection location and timing. Hassan et al. [62] introduced a node selection scheme
Salah and Saadi [56] devised a context-aware ubiquitous for ubiquitous environments named Integrated Multi-Criteria
learning system that utilizes a fuzzy Analytical Hierarchy Decision-Making for Data Retrieval (IMCDM-NSDR),
Process (AHP) for learning service selection. This fuzzy employing integrated fuzzy Analytical Hierarchy Process
AHP method enhances the accuracy of service selection by (FAHP) and fuzzy Technique for Order Preference by
evaluating factors expressed in ambiguous linguistic terms Similarity to Ideal Solution (TOPSIS) methods to rank the

130780 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

data retrieval node in ubiquitous environments. The main In conclusion, there is currently a deficiency in a compre-
drawbacks of this scheme are that it did not consider the hensive data collection model for multi-task crowdsourcing
location of the necessary data and the timing of its collection. that maximizes overall system benefits and considers the
In [63], the authors introduced Group-Oriented Task balance between the total system profit and the acquired
Allocation (GOTA) for crowdsourcing in Social Networks to rewards of the data collectors. Such a model should consider
address how workers naturally form groups in crowdsourcing factors such as the workload of sensing tasks, associated
systems. Departing from individual-oriented and team forma- bids, tasks’ preferences and the uncertainty of their available
tion approaches, tasks are allocated to worker groups. There resources, and the alignment between data collectors and
are two selection methods: semi-supervised (GOTASemi) sensing tasks. Consequently, the selection and assignment of
and fully-supervised (GOTAFull). In GOTASemi, a set of sensing tasks to optimize system profits represent a crucial
groups with a leader is available, and the requester selects challenge that requires attention and resolution.
the principal group for task execution. The leader of this
principal group determines all assistant groups involved
in the task. In GOTAFull, the requester selects both the III. SYSTEM MODEL AND PROBLEM STATEMENT
principal group and all assistant groups directly. The article A. SYSTEM MODEL
highlighted the NP-hard nature of the problem and proposed In the domain of pervasive computing, applications demand
a heuristic approach with manageable time complexity. the gathering of data at specific times and locations with
This heuristic prioritizes group participation based on a a focus on cost-effectiveness, high-quality data, and energy
crowdsourcing value function, supported by theoretical efficiency. The objectives of these applications vary based on
effectiveness in optimization. Extensive experiments using their unique purposes. For instance, applications dedicated to
real-world crowdsourcing data confirmed the superiority of monitoring air quality and assessing pollution levels across
the group-oriented approach over individual and team-based a city require data collection at various locations. Similarly,
methods across various performance metrics. real-time traffic applications need to gather traffic data at
In [64], the paper introduced a batch crowdsourcing specific times in key road areas.
approach based on distributed team formation. This method In a ubiquitous computing (UC) environment, let’s con-
grouped tasks with similar skill requirements into batches sider that this environment caters to a set of diverse sensing
to lower computational costs, leveraging workers’ abil- tasks, denoted as ST = {st1 , st2 , . . . , sti , . . . , stm } for
ity to self-organize teams through social networks. The each user ui in U . Now, assuming there exists a set of
problem’s optimization objective was formalized, and its criteria or preferences for the platform denoted as PR =
NP-hardness was established. Two heuristic approaches were {pr1 , pr2 , . . . , pry , . . . , prY }, where y ∈ (1, Y ) represents
then proposed to address this challenge. Experimental results multi-constraint parameters used to rank each task. Also,
using real-world data showed that both heuristic approaches assuming there exists a set of criteria or preferences for each
surpassed previous benchmark methods across three key sensing task denoted as CR = {cr1 , cr2 , . . . , crx , . . . , crX },
metrics: team formation costs, success ratio of tasks, and where x ∈ (1, X ) represents multi-constraint parameters
overall efficiency. used to rank each data collector containing the required data
In [65], The authors devised a blockchain-based trust for executing this sensing task. Additionally, specific criteria
model to improve and validate the integrity and security such as cost, completion rate, waiting time, data quality,
of collaborative business processes. This model focused on and energy consumption are associated with each individual
ensuring the credibility of collaborative interactions through requested service.
a trust-based blockchain framework designed to meet specific To fulfill these service requests, a certain number of data
criteria. It was particularly tailored for healthcare data- collectors are required for data collection. Due to limitations
sharing processes, demonstrating its efficacy in creating a in a single data collector’s time, location, and energy, a single
secure and trustworthy environment for data sharing. The service request typically involves the collaboration of multi-
proposed model not only guided the establishment of trust ple data collectors. Consequently, there is a considerable pool
but also provided mechanisms to verify integrity and security of potential data collectors in the ubiquitous environment,
in distributed blockchain environments used in collaborative expressed as: DN = {dn1 , dn2 , . . . , dnj , . . . dnn }, where
business processes. j ∈ (1, n).
In [66], The authors presented a method for application- In smart cities, a myriad of data collectors actively
specific feature selection and clustering using HPC system traverse the urban landscape to gather data for essential
profiling data. This approach utilized PCA-based feature services. These collectors are evaluated based on various
selection to enhance performance analysis techniques. Addi- criteria, including waiting time, data quality, the anticipated
tionally, unsupervised learning was employed to analyze reward for task completion, completion rate, and the efficient
application-specific characteristics derived from profiling utilization of resources such as energy. The ubiquitous
data. In their experiments, they evaluated highly parallel environment has the capability to select the most suitable data
supercomputers using the NAS parallel benchmark and collector from the dispersed pool of candidate data collectors
achieved efficient clustering of applications. across the smart city to collaborate in fulfilling the requested

VOLUME 12, 2024 130781

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

sensing task services. In this ubiquitous environment, the While obtaining a reward can mitigate the expenses for a
regions where applications require data collection are divided data collector and enhance data collection, individual data
into l grids, each assigned a distinct number. The entire city’s collectors may face challenges in submitting data precisely
monitoring area is denoted as: G = {g1 , g2 , . . . , gk , . . . , gp }, at the sampling time. This discrepancy arises from variations
where k ∈ (1, p). in the environments, activity patterns, signal intensity, and
j
Ubiquitous applications need to collect data not only at power resources among data collectors. Using ck,l =
designated locations but also at specific time intervals. Based 1 indicates that data collector cj can collect sensing data at
on the application’s needs, the overall request time T for j
grid location gk and time tl . ck,l = 0 indicates that data
the entire task is segmented into a sequence of time periods, collector cj cannot collect sensing data at grid location gk
denoted as: T = {t1 , t2 , . . . , tl , . . . , tq }, where l ∈ (1, q). and time tl . Then the expected sampling data matrix of data
In the ubiquitous environment, users submit task requests collector dnj can be expressed as follows:
to the platform along with details about the required data,  j j j j 
including location, time, and data type. Various task requests c1,1 c1,2 · · · c1,l · · · c1,q
 j
are published based on diverse application requirements, each  c cj · · · cj · · · cj 

specifying unique criteria for data collection. These requests  2,l 2,2 2,l 2,q 
 . .. . . . 
 . . a1,2 · · · .. 
entail the collection of different data types, occurring at  . . 
specific times and locations, all while minimizing waiting Cj = 
 j j j j 
 (3)
time and maximizing data quality. xk,l,w,d i = 1 denotes that ck,1 ck,2 · · · ck,l · · · ck,q 
 . .. . . .. 
 
the requested task sti needs to collect data dk,l i at location
 .. . . a1,2 · · · . 
i
gk , time tl . xk,l = 0 denotes that the requested task sti
 
j j j j
does not need to collect data dk,l i at location g , time t . The cp,1 cp,2 · · · cp,l · · · cp,q
k l
expected sampling data matrix of the requested task sti can In which j ∈ (1, n); k ∈ (1, p); l ∈ (1, q), and
be expressed by the following matrix as follows: 
 1 dnj can collect data
 i i i · · · xi 

x1,1 x1,2 · · · x1,l 1,q j

 gk at time tl
xi xi · · · xi · · · xi  ck,l = (4)
 2,l 2,2 2,l 2,q  
 0 dnj cannot collect data
 .. .. . . .. 
  

gk at time tl
 . . . a1,2 · · · . 
Di =  i (1)
 
xk,1 xk,2 i i i 
· · · xk,l · · · xk,q  Then a single data collector can be denoted as: dnj =

 .
 {Cj , rj }. Due to resource limitations, a single data collector
 . .. . . . 
 . . . a1,2 · · · ..   can only supply data for a singular requested task, leading to
i i i i competition among different requested tasks in the selection
xp,1 xp,2 · · · xp,l · · · xp,q of data providers. Simultaneously, when choosing a data
In which i ∈ (1, m); k ∈ (1, p); l ∈ (1, q) and collector, a single requested task aims to opt for a collector
 with acceptable costs and higher efficiency to fulfill the

 1 task sti need to service. Therefore, the selection of a data collector for a single
 i requested task involves a certain level of competition among
i
 collect data dk,l
xk,l = (2) data collectors.


 0 task sti do not need to
 i
collect data dk,l
B. PROBLEM STATEMENT
In order to recruit a sufficient number of data collectors, Indeed, the collaboration of multiple data collectors becomes
users are required to make a payment for the requested task. crucial in fulfilling tasks with specific requirements in
Simultaneously, owing to variations in the benefits, workload, terms of time, location, waiting time, and data quality.
and energy consumption associated with different requested However, as you mentioned, limitations in the distribution
tasks, the bids for each task also differ. If the bid paid by and number of data collectors may lead to challenges in
request sti is pi , then the bid set of the requested service sti obtaining the desired data sample matrix for requested
can be expressed as: P = {p1 , p2 , . . . , pi , . . . , pm }, where tasks. This can occur when the selected data collectors,
i ∈ (1, m). Then a single requested service can be denoted although collectively capable, may not cover all the required
as: sti = {Di , Pi }. sampling points. To address this challenge, it becomes
When a data collector engages in data collection, various essential for the platform to carefully select and assign
costs such as energy, electricity, and time are incurred. data collectors, considering their capabilities, coverage, and
Consequently, the data collector anticipates receiving a the specific requirements of the requested task. Optimiz-
specific reward to offset these costs and achieve a certain ing the selection and assignment process can help enhance
level of profit. This anticipated reward for the data collector the chances of obtaining the desired data sample matrix
is denoted as dnj is rj , and the expected reward set of the data for the task. Additionally, introducing mechanisms for
collector is R = {r1 , r2 , . . . , rj , . . . , rn }. collaboration and coordination among data collectors, such

130782 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

as real-time communication or task-specific planning, can these requested tasks with acceptable rewards, ensuring a
further improve the efficiency of data collection for tasks with balance between the system’s profit and the rewards for data
specific and diverse requirements. This collaboration can collectors. Consequently, the overall net profit
P of Pthe system
is articulated as: E = B−Q = zi=1 pi − zi=1 w
P
enable data collectors to work together strategically, ensuring j=1 rj . One
comprehensive coverage and fulfillment of the task’s of the objectives of this paper is to optimize and maximize
For the specified task sti , the platform selects a data the system’s profit to the fullest extent possible.
collector denoted as Wi = {dn1 , dn2 , . . . , dnj , . . . , dnw }, Xz z X
X w
where w ∈ (1, n). Due to resource limitations for each max(E) = max( pi − rj ). (11)
data collector, each collector can only participate in a single i=1 i=1 j=1
requested task. The set of rewards that the platform must
provide to compensate the data collector is represented as D. MAXIMIZING TASK COMPLETION RATE
Ri = {re1 , re2 , . . . , rej , . . . , rer }, where r ∈ (1, n). To fulfill In the pursuit of maintaining the system’s profitability, the
the requested task sti , the total reward that the system needs objective is to comprehensively fulfill the data collector
to disburse is calculated as follows: requirements for various requested tasks. Therefore, a crucial
X w metric for assessing the effectiveness of the overall com-
Fi = rj (5) puting environment is the task completion rate, defined as:
|A|
j=1 Θ = , where || denotes the number of elements in the
|A|
The net profit of the system platform by performing corresponding set.
requested task sti = {Di, Pi} is: Ei = Pi − Fi
|A| z
max(Θ) = max( ) = , z ∈ (1, m) (12)
Ei = Pi − Fi (6) |A| m
The requested task can contribute to the system’s profits E. MAXIMIZING SATISFACTION RATIO OF THE SYSTEM
and prompt the platform to allocate a data collector only when While ensuring the system’s profitability and task completion
Ei > 0. In a UC network, a low bid for a requested task makes rate, the objective is to meet the requirements of all necessary
it challenging for the system platform to secure a sufficient tasks and data collectors with an optimal strategy. Therefore,
number of data collectors for its completion. Additionally, the satisfaction ratio metric is a crucial factor that enhances
if the available number of data collectors in the UC network the overall realism and effectiveness of the computing
is insufficient to fulfill all requested tasks, it may result in the environment. This metric is determined by the number of
failure of these tasks. The system’s capability to ultimately selected data collectors that fulfill the preferences of the
accomplish the requested tasks can be expressed as: tasks and the task completion ratio. Let SDa denote the total
ST = {st1 , st2 , . . . , sti , . . . , stz }, z ∈ (1, m) (7) number of selected data collectors for all tasks, and SDs
represent the number of selected data collectors that satisfy
Then the income set that the system can achieve by the preferences of these tasks, where SDs ⊆ SDa . Thus, the
completing these requested tasks is as follows: satisfaction ratio is defined as follows:
|SDs |
B = {P1 , P2 , . . . , Pi , . . . , Pz }, z ∈ (1, m) (8) max(SS) = max( × Θ) (13)
|SDa |
So, the total income B the system can obtained is: In summary, the main goals of this paper center on
z
X optimizing the ubiquitous computing system. The primary
B= Pi (9) areas of focus involve the selection of requested tasks
i=1 and the scheduling of data collectors, aiming to achieve
In addition, the total reward paid to the data provider for the following objectives: maximizing overall system profit
completing the requested task set sti is: with acceptable rewards, improving the completion rate
z z X
w of requested tasks, and maximizing the satisfaction ratio
of the system. By addressing these objectives, the paper
X X
Q= Fi = rj (10)
i=1 i=1 j=1
aims to contribute to the efficiency, effectiveness, and user
satisfaction within the ubiquitous computing environment.
The main objectives of this paper are as follows: 
 max(E)

C. MAXIMIZING SYSTEM PROFITS WITH ACCEPTABLE Objectives = max(Θ) (14)
REWARDS

max(SS)

The collection of requested tasks completed by the platform
is represented as A. The platform generates corresponding IV. THE PROPOSED FUZZY RANKING COLLABORATIVE
revenue by executing these requested tasks. Simultaneously, MULTI-TASKS DATA COLLECTION SCHEME
costs are incurred by the platform, as it provides rewards In this section, a new multi-tasks data collection scheme in
to the selected data collectors to motivate them to fulfill UC is proposed which is called Fuzzy Ranking Collaborative

VOLUME 12, 2024 130783

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

with Ranking (CDR) for each data collector, selecting the one
with the highest CDR value for integration into the task’s data
collection efforts.
Once a data collector is selected, it contributes to the task
until completion. The algorithm checks if the task is finished
and updates the sampling data matrix accordingly. If the
task is completed, it is removed from the candidate task set
(ST ). The algorithm loops back to prioritize and assign the
next task (Return to Step 4) until all tasks are processed.
Finally, the flowchart concludes when all tasks have been
completed and no additional tasks or data collectors are
available (|ST | = 0 and DN = 0). This structured approach
ensures that tasks are managed efficiently, data collectors
are optimally utilized, and tasks are completed in a manner
that maximizes system profitability and efficiency within the
constraints of a dynamic crowdsourcing environment.

A. BASIC IDEA
To solve the multi-tasks data collection problem in UC,
the basic idea of FRCMDCS is based on the following
phases: (1) Determining the importance of the preferences
or criteria for the platform using the Fuzzy Analytical
Hierarchy Process (FAHP), which aims to determine the
optimal weights for each platform preference or criteria used,
(2) Ranking the tasks using the fuzzy technique for request
FIGURE 1. Flowchart of the proposed FRCMDCS Scheme. preference by Optimal Solution Similarity (FTOPSIS) based
on the platform preferences and their weights which is
determined by FAHP in (1), (3) Selecting a sensing task
Multi-Tasks Data Collection Scheme (FRCMDCS). Firstly, based on the highest task rank in Phase (2), (4) Determining
the basic idea of FRCMDCS will be introduced and then the the importance of the preferences or criteria for tasks using
proposed FRCMDCS will be described in detail. Figure 1 the Fuzzy Analytical Hierarchy Process (FAHP), which aims
shows the flowchart of the proposed FRCMDCS Scheme. to determine the optimal weights for each task preference
Fig. 1 shows the flowchart of the proposed FRCMDCS. or criteria used, (5) Ranking the data collectors using the
This flowchart outlines a systematic approach to managing fuzzy technique for request preference by Optimal Solution
tasks and data collectors in a crowdsourcing environment. Similarity (FTOPSIS) based on the required task preferences
The flowchart begins by checking if there are any tasks and their weights which is determined by FAHP in (2), (6)
remaining in the current set (|ST | ̸ = 0). If tasks are present, Calculating the Contribution Density with Ranking (CDR)
the algorithm proceeds to determine the weights of platform which is a metric to measure the contribution of an individual
selection criteria using the Fuzzy Analytical Hierarchy Pro- data collector to a specific sensing task by consider its rank
cess (FAHP). This step ensures that the platform selects tasks value, its covered ratio of request locations, and its expected
based on predefined criteria that optimize system efficiency reward, and (7) Selecting the set of data collectors with the
and profitability. Afterward, the Fuzzy Technique for Order highest values of CDwR for executing the requested task. For
Preference by Similarity to Ideal Solution (FTOPSIS) is more details about FAHP and FTOPSIS, the reader can find
employed to rank the tasks based on their priority and comprehensive and valuable details for these methodologies
suitability for assignment. The algorithm then selects the in Gad-Elrab [62].
highest-priority task (BST) from the ranked list. It checks In the next subsections these phases will be introduced in
again if there are remaining tasks (|ST | ̸ = 0) to continue details.
the process for subsequent tasks or ends if all tasks have
been processed (|ST | = 0). This structured approach ensures B. PHASE 1: DETERMINING THE IMPORTANCE OF THE
that tasks are assigned in an optimal sequence, maximizing PLATFORM CRITERIA
overall system performance. Simultaneously, the algorithm Each platform may need to satisfy certain own preferences
checks the availability of candidate data collectors (DN ̸= 0). when executing a requested task, so the failure to accomplish
If data collectors are available, it proceeds to determine a requested task can be attributed to the falling in satisfying
their selection criteria weights using FAHP. FTOPSIS is then these required preferences such as the task benefit to the
used to rank the candidate data collectors based on these system which renders it impractical or unnecessary to pursue.
criteria. The algorithm calculates the Contribution Density Only when Ei ≥ 0 can the requested task bring profits to the

130784 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

system. When Ei ≤ 0, the completion of the requested task requested task based on the required platform preferences and
will not increase the profit of the system, which is a waste of their weights Wpr which is determined by FAHP in (1).
system resources. Therefore, the completion of the requested Hwang and Yoon [72] introduced TOPSIS (Technique
task sti has the potential to yield substantial profits for the for Order Performance by Similarity to Ideal Solution),
system. However, not considering the bid of the requested a foundational multi-criteria decision-making methodology.
task and not accounting for the variance in sampling data TOPSIS, developed by Hwang and Yoon [72], aims to
requirements among requested tasks may lead to suboptimal identify solutions from a finite set of alternatives. The
choices that do not maximize system profits. Hence, when methodology revolves around defining the positive ideal
selecting requested tasks, it is imperative to comprehensively solution (PIS) and negative ideal solution (NIS). The key
consider the platform preferences such as task benefit, task idea is that the preferred option should be the closest to the
bid value, task cost (i.e., the volume of data a requested PIS and furthest from the NIS. The PIS maximizes benefit
service necessitates), and others. criteria while minimizing cost criteria, and the NIS follows
In this phase, to take these platform preferences into the reverse logic, maximizing cost while minimizing benefit
account, the proposed FRCMDCS assumes that there is a set criteria [73]. TOPSIS considers distances to both the PIS
of required preferences or criteria for each platform which and NIS simultaneously, ultimately identifying the optimal
is denoted as PR = {pr1 , pr2 , . . . , pry , . . . , prY } y ∈ solution that is closest to the PIS and farthest from the NIS.
(1, Y ). To determine the importance of these criteria, the However, a notable limitation of TOPSIS is its inability
proposed FRCMDCS employs FAHP to find the weights of to account for ambiguity or inaccuracy inherent in group
these platform criteria. FAHP approach is implemented by decision-making situations [74]. To address this, the Fuzzy
integrating fuzzy set theory into the conventional AHP. In this TOPSIS (FTOPSIS) approach was introduced by Chen [75].
context, each pairwise comparison decision is expressed as FTOPSIS enables decision makers to incorporate incomplete
a fuzzy number distinguished by a membership function as and non-quantifiable information by combining fuzzy theory
shown in Table 1. with the TOPSIS method. This approach is particularly useful
for handling uncertain data and utilizing subjective criteria
TABLE 1. Example of a fuzzy number distinguished by a membership and weights to determine the most suitable alternative(s)
function. among a set of feasible options. See more details in [62].
By applying FTOPSIS, we will determine the ranks of all
tasks in DN which is described as follows:

RTSst = {rs1 , rs2 , . . . , rsi , . . . rsm } i ∈ (1, m),

rsi ∈ {1, 2, 3, . . . , m} (16)

where for any two values rsi and rsb in RTSst , if rsi ≤ rsb ,
The incorporation of human heuristics into computer- then the task rsi has higher priority than rsb or it has the same
assisted decision-making, coupled with the ability to priority with respect to the required platform preferences and
represent human thinking and interpretation through the their weights Wpr . Otherwise the task rsb has higher priority
establishment of mathematical rules, particularly in dealing than rsi .
with numerical data and linguistic terms that are more
comprehensible for human reasoning, stands as a primary C. PHASE 3: SELECTING THE SENSING TASK
motivation for employing fuzzy methods. Within the realm In this phase, based on the determined the importance of the
of FAHP, several approaches exist. The commonly employed platform criteria in Phase 1 and the ranked tasks in Phase 2,
methods for calculating the relative weights of criteria include the platform selects the task with highest rank or priority that
the geometric means, as introduced by Buckley [67], and bring higher benefit to the system.
extent analysis methods, proposed by Chang [68]. For this
research, the geometric technique is utilized, known for its D. PHASE 4: DETERMINING THE IMPORTANCE OF THE
efficacy in diverse scenarios [69], [70], [71]. See more details TASK CRITERIA
in [62]. As in our proposed method in [62], for criteria in CR =
By applying FAHP, we will determine the weights of all {cr1 , cr2 , . . . , crx , . . . , crX } x ∈ (1, X ), FAHP is employed
criteria in PR which is defined as follows: to determine the weights of these criteria. For example,
mobility, waiting time, cost, trust, sharing and collaboration,
Wpr = {pr1 , pr2 , . . . , pry , . . . , prY } x ∈ (1, Y ) (15) workload, data transfer, and data quality. FAHP approach
is implemented by integrating fuzzy set theory into the
1) PHASE 2: DETERMINING THE RANKS OF TASKS conventional AHP. In this context, each pairwise comparison
For all tasks in ST = st1 , st2 , . . . , sti , . . . , stm , FTOPSIS is decision is expressed as a fuzzy number distinguished by a
employed for determining the ranks of tasks to execute the membership function.

VOLUME 12, 2024 130785

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

By applying FAHP, we will determine the weights of all process involves not just considering the individual cost of
criteria in CR which is defined as follows: each data collector but also evaluating the contribution each
data collector makes to the task.
Wcr = {cr1 , cr2 , . . . , crx , . . . , crX } x ∈ (1, X ) (17) After determining the importance of the platform criteria
in Phase 1 and ranking the candidate tasks in Phase 2, the
1) PHASE 5: DETERMINING THE RANKS OF DATA platform selects the task with highest rank or priority that
COLLECTORS bring higher benefit to the system in Phase 3. Then after
For all data collectors in DN = {dn1 , dn2 , . . . , dnj , . . . dnw } determining the importance of the task criteria in Phase 4 and
w ∈ (1, n), FTOPSIS is employed for determining the ranks ranking the candidate data collectors in Phase 5, the platform
of data collectors to execute the requested task based on the selects the data collector with highest contribution density
required task preferences and their weights Wcr which is with ranking CDRij to join the data collector set Wi of the
determined by FAHP in (3). By applying FTOPSIS, we will requested task. And each time a data collector is selected, the
determine the ranks of all data collectors in DN which is ranking of the remaining data collectors and their contribution
described as follows: density with ranking are updated. When a data collector dnj is
RDCdn = {rn1 , rn2 , . . . , rnj , . . . rnn } j ∈ (1, n), selected, the data sample of requested collector data covered
by the data provider dnj is deleted from the expected sampling
rnj ∈ {1, 2, 3, . . . , n} (18) data matrix of the requested collector. At the same time,
where for any two values rnj and rnb in RDCdn , if rnj ≤ rnb , the selected data collector is deleted from the candidate data
then the data collector rnj has higher priority than rnb or it has provider set.
the same priority with respect to the required task preferences Based on these phases, the main steps of FRCMDCS
and their weights Wcr . Otherwise the data collector rnb has algorithm can be expressed as:
higher priority than rnj . • step 1: If the current task set |ST| ̸ = 0, use FAHP to
determine the weight of each platform criterion, if the
2) PHASE 6: CALCULATING THE CONTRIBUTION DENSITY current set of tasks |ST| = 0, the algorithm ends.
WITH RANKING • Step 2: Use FTOPSIS to rank all the requested tasks.
• Step 3: If the current requested task set |ST| ̸ = 0,
In this phase, the proposed FRCMDCS calculates the
contribution density of each data collector taking into according to the sorted order, the requested tasks with
account it rank which is determined in Phase 4. Firstly, highest priority BST will be selected. If the current set
the contribution each data collector makes to the task is of tasks |ST| = 0, the algorithm ends.
• step 4: If the current candidate data collector set DN ̸ =
calculated. The greater the coverage of data sampling points
required by the sensing task through the data collected by a 0, use FAHP to determine the weight of each criterion,
specific data collector, the more significant its contribution If the current set of candidate data collectors dn = 0, the
to the task. Let Sj represent the dataset collected by data algorithm ends.
• Step 5: Use FTOPSIS to rank all the candidate data
collector, and then the data collected by data collector Cj =
{Sj , rj }, which can encompass the sampling data necessary for collectors.
• Step 6: Calculate the contribution density with ranking
sensing task sti , is outlined as Di ∩ Sj . Then the contribution
density of the data collector Cj to the sensing task is defined CDRij for each candidate data collector, the data collector
as follows: with largest CDR will be selected.
• Step 7: After selected corresponding data collector,
|Di ∩ Sj |
CDij = , CDij ≥ 0 (19) it added to data collectors Wi of requested task. Update
rj the sampling data matrix of the requested task and
by taking into account the rank of each data collector, determine whether the requested task is completed. If the
we define the Contribution Density with Ranking CDRij for requested task is not completed and Ei > 0, return to
data collector dnj and task sti as follows: Step 4. If Ei < 0, then go to Step 8.
• Step 8: Delete the requested task from candidate task set
CDij |Di ∩ Sj | ST. Return to Step 2 to perform the next requested task.
CDRij = = , rnj ∈ RDCdn (20)
rnj rj × rnj
V. SIMULATION AND RESULTS EVALUATION
3) PHASE 7: SELECTING THE SET OF DATA COLLECTORS To evaluate the effectiveness of the proposed FRCMDCS
The system platform performs the unassigned requested tasks method, diverse sets of task publishers and data collectors
according to the sorted set of tasks which is determined are utilized for testing. A comparative analysis is carried
Phase 1. When the requested task sti is selected by the out with four alternative methods: Random Task selection
system to perform, it is necessary to select the optimal set with Contribution Density data collector random selection
of data collectors to carry out the task. When opting for (Random), A first Task selection with Coverage First
data collectors, the primary objective is to assemble the most data collector selection (FCFS), and the Collaborative
cost-effective set for task sti . Consequently, the selection Multi-Tasks Data Collection Scheme (CMDCS [41]). The

130786 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

CMDCS Scheme employs task unit bids to determine the TABLE 2. the default parameter settings of the simulation model.
execution order of sensing tasks and relies on contribution
density for selecting data collectors. In contrast, Random
employs a random selection approach for sensing tasks
without considering task unit bids and uses random selection
for data collectors associated with those tasks. Similarly,
FCFS does not consider the task unit bids in its selection
process but selects the initial set of sensing tasks in the
sensing tasks list and the initial set of data collectors in the
candidate data collectors list for those tasks. The GOTAFull
[63] scheme divides the data collectors into groups and
selects the principal group and some other groups as assistant
groups for executing the required tasks. GOTAFull is a
fully supervised algorithm in which the assistant groups
are all allocated by the task requester. In addition, with
GOTAFull there are some groups with a leader that takes
the responsibility to determine the workers in its group a versatile, extensible, and modular C++ simulation
to execute the task. While other groups without a leader. library and framework primarily designed for constructing
In our simulations, for GOTAFull there was 5 groups where network simulators [76]. The scope of ‘‘network’’ in this
3 groups with a leader and others without a leader. context encompasses a broad spectrum, including wired
The selection criteria for comparative methods in crowd- and wireless communication networks, on-chip networks,
sourcing, such as Random Task selection with Contribution queueing networks, and more. The framework accommodates
Density data collector random selection (Random), First domain-specific functionalities through model frameworks,
Come First Serve (FCFS), and Collaborative Multi-Tasks developed as stand-alone projects. These functionalities span
Data Collection Scheme (CMDCS), are primarily guided support for sensor networks, wireless ad-hoc networks,
by their suitability to address specific operational needs Internet protocols, performance modeling, photonic net-
and challenges within a crowdsourcing network. Random works, and other specialized areas. OMNeT++ provides
selection is chosen to introduce diversity in data inputs a comprehensive development environment, featuring an
without bias, beneficial for exploratory data analysis. FCFS Eclipse-based Integrated Development Environment (IDE),
prioritizes prompt task execution, crucial in time-sensitive a graphical runtime environment, and various supplementary
applications like real-time monitoring. CMDCS is selected tools. In addition, all devices in conducted simulation are
for its ability to optimize task allocation based on bid values configured as ‘‘cSimpleModule’’ in OMNeT++ which
and contribution density, maximizing system profitability and specifies parameters and settings as the module type,
ensuring efficient use of available resources. Each method’s parameters, and gates for message passing, allowing for clear
selection is tailored to meet varying task characteristics, module structure and communication paths. Parameters can
resource availability, desired outcomes such as data diversity be set to customize module behavior and are accessed during
or profitability, and the overall structure of the crowdsourcing the simulation lifecycle. The configuration file specifies
network, ensuring an effective balance between operational values for these parameters, facilitating easy adjustments for
efficiency, task performance, and economic considerations. different simulation runs.
These selected comparative methods in crowdsourcing In this study, two types of experiments are conducted to
have notable limitations. Random selection can introduce evaluate all methods. The first experiment is executed in the
diversity but may lead to bias in representation and overlook case of an existing small number of data collectors, ranging
important tasks, while also being inefficient in resource from 5 to 45 DCs, while the second is carried out in the
allocation. FCFS can favor faster responders, potentially case of an existing large number of data collectors, ranging
neglecting skilled workers and leading to task overload and from 50 to 800 DCs. For each experiment, the number of tasks
burnout. CMDCS, although optimized for profitability and varies from 50 to 600 tasks. Additionally, five criteria are
efficiency, is complex to implement and relies heavily on employed, namely Data quality, Collaboration, Consumed
accurate contribution density data, which may not always Time Interval, Reward, and Privacy. Table 2 presents all the
be reliable. Overall, these methods may struggle to adapt to default parameter settings of the simulation model.
dynamic environments and ensure consistent quality control,
highlighting the need for careful consideration of operational B. EVALUATION METRICS
context and objectives in their application. In the assessment experiments, several key metrics are
utilized:
A. SIMULATION SETTINGS • Task Compilation Rate (TCR): This metric expresses
The simulation model was implemented using the the proportion of completed tasks to the overall number
OMNeT++ (version 5.6.2) discrete-time simulation package, of tasks.

VOLUME 12, 2024 130787

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

FIGURE 2. Task completion rate vs number of DCs.

FIGURE 3. Total profit vs number of DCs.

• Total Profit (TP): TP quantifies the aggregate profit C. RESULTS AND DISCUSSION
value accrued by the platform from all executed tasks. 1) EFFECTS OF DIFFERENT NUMBER OF DATA COLLECTORS
• Total Reward (TR): TR measures the overall reward Here, the effects of different number of data collectors are
value obtained by all data collectors from the execution analyzed and discussed for all methods when the number of
of all tasks. tasks is 300.
• Total Selected DCs (TSDD): This metric signifies the In Fig. 2, the achieved values of TCR by all methods
overall count of selected data collectors in the execution for both small and large DCs experiments are illustrated.
of all designated tasks. In Fig. 2a, for the scenario with a small number of DCs,
• Satisfaction Score (SS): Originally defined in Eq. 21, the proposed FRCMDCS achieved higher TCR values than
SS is redefined based on the algorithms employed in this other methods. In addition, the TCR values of FRCMDCS,
simulation. CMDCS, Random, and GOTAFull methods increase as the
T
|SDFRCMDCS SDa | number of DCs increases which were up to 100%, 98%,
SS = × TCRa (21) 99%, and 97% for FRCMDCS, CMDCS, and Random, and
|SDa |
GOTAFull methods, respectively. While the TCR values
Here, SDFRCMDCS denotes the set of selected data collec- for FCFS method exhibit instability, fluctuating and being
tors by the proposed method FRCMDCS, SDa signifies the set influenced by varying small numbers of DCs which were
of selected data collectors by algorithm a, TCRa represents between [75%, 100%]. This behavior is attributed to FCFS
the task completion rate achieved by algorithm a, where selecting tasks and their data collectors solely based on their
a ∈ FRCMDCS, CMDCS, FCFS, Random, GOTAFull, and position in the list, without considering task bids or the
| • | denotes the number of elements in a specific set. contribution density of data collectors.

130788 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

FIGURE 4. Total Reward vs number of DCs.

FIGURE 5. Total Selected DCs vs number of DCs.

In Fig. 2b, for the scenario with a large number of DCs, This is because, existing more number of selected DCs will
the proposed FRCMDCS achieved higher TCR values than cost more rewards which affects the total system profit.
other methods. In addition, the TCR values for the proposed In addition, the proposed FRCMDCS aachieved the higher
FRCMDCS exhibits minimal sensitivity to changes in the TP values than other methods. FRCMDCS can maximize
number of DCs which was almost 100%, while the TCR the total number of completed tasks, thereby impacting the
values for CMDCS, FCFS, Random, and and GOTAFull overall profit of the platform. Also, CMDCS achieved higher
methods exhibit instability, fluctuating and being influenced TP values than FCFS, Random, and GOTAFull methods, this
by varying large numbers of DCs. These values range is because FCFS and Random methods are influenced by
between [94%, 98%], [58%, 100%], [87%, 99%], and [60%, varying small numbers of DCs. This is attributed to FCFS
99%] for CMDCS, FCFS, Random, and GOTAFull methods, and Random methods selecting tasks and their data collectors
respectively. This consistency stems from their ability to without considering task bids or the contribution density of
effectively choose the most suitable sensing tasks and data data collectors, while GOTAFull method considers the whole
collectors in this scenario and the presence of a large number rewards of each group in the selection process.
of DCs reduces the likelihood of selecting the most suitable In Fig. 3b, for the scenario with a large number of
DCs for executing the chosen sensing tasks, without taking DCs, the TP values for the proposed FRCMDCS, CMDCS,
into account task bids or the contribution density of data and GOTAFull methods are affected by changes in the
collectors. large number of DCs. The presence of a large number of
In Fig. 3, the achieved values of TP by all methods for both DCs increases the paid rewards to these DCs, influencing
small and large DCs experiments are presented. In Fig. 3a, the final TP of the platform. Nevertheless, FRCMDCS,
for the scenario with a small number of DCs, the TP values CMDCS, and GOTAFull achieve larger values than FCFS
of all algorithms decrease as the number of Dcs increases. and Random methods because they can maximize the total

VOLUME 12, 2024 130789

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

FIGURE 6. Satisfaction score vs number of DCs.

FIGURE 7. Task completion rate vs number of tasks.

number of completed tasks, impacting the overall profit of FRCMDCS achieves higher values than the CMDCS method
the platform. Conversely, FCFS and Random methods attain by considering the balance between the total profit and the
lower values of TP, and they are also unstable and influenced total rewards based on task preferences and the ranked DCs
by varying large numbers of DCs. This is because FCFS list.
and Random methods select tasks and their data collectors In Fig. 4b, for the scenario with a large number of
without considering task bids or the contribution density of DCs, the TR values for the proposed FRCMDCS and
data collectors, and the large number of DCs reduces the CMDCS methods decrease as the number of DCs increases,
likelihood of selecting the most suitable DCs for executing with FRCMDCS achieving higher values than the CMDCS
the chosen sensing tasks. method. The large number of DCs provides more choices in
Fig. 4 illustrates the achieved values of TR by all methods selecting the most appropriate data collectors with acceptable
for both small and large DCs experiments. In Fig. 4a, rewards. However, both methods achieve lower values than
for the scenario with a small number of DCs, the TR FCFS, Random, and GOTAFull methods because FRCMDCS
values for the proposed FRCMDCS and CMDCS methods and CMDCS can maximize the total number of completed
remain relatively unaffected by changes in the number of tasks, impacting the overall final TR of the platform.
DCs, and both methods achieve lower values compared to Conversely, FCFS, Random, and GOTAFull methods attain
FCFS, Random, and GOTAFull methods. FRCMDCS and larger values of TR, and they are also unstable and influenced
CMDCS can select the most appropriate data collectors by varying large numbers of DCs. This is because FCFS
with acceptable rewards. Conversely, FCFS and Random and Random methods select tasks and their data collectors
methods attain higher values of TR because they select tasks without considering the expected rewards of the data
and their data collectors without considering task bids or collectors, and the large number of DCs reduces the chance of
the expected rewards of the data collectors. The proposed selecting the most appropriate DCs with acceptable rewards

130790 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

for executing the chosen sensing tasks. While GOTAFull Fig. 6 illustrates the achieved values of SS by all methods
takes into account the grouping collaboration among data col- for both small and large DCs experiments. In Fig. 6a, for
lectors which affects the executing of chosen sensing tasks. the scenario with a small number of DCs, the SS values
As shown in Figs. 2a, 2b, 3a, 3b, 4a, and 4b, the TCR, for the proposed FRCMDCS are less affected by changes
TP, and TR of FRCMDCS are not monotonic with the DC in the number of DCs, ranging between [87%, 100%],
number, respectively. These non-monotonic trends observed and are significantly higher than other methods, which
in TCR and TP for the FRCMDCS with respect to the number range between [31%, 61%], [6%, 47%], [9%, 41%], and
of DCs can be explained by the adaptivity of FRCMDCS [9%, 35%] for CMDCS, FCFS, Random, and GOTAFULL,
and the dynamic nature of crowdsourcing environments. respectively. FRCMDCS considers the ranking contribution
FR CMDCS employs complex decision-making processes, of each data collector in the process of executing the
including fuzzy logic and multi-criteria evaluation, to assign requested task in addition to its contribution density, while
tasks based on their priority and data collector contributions. CMDCS only considers the contribution density without
As the number of data collectors varies, the algorithm adjusts the rank of data collectors to the task. FCFS and Random
its allocation strategy, aiming to optimize resource utilization methods do not consider either of these factors. The SS
and task completion efficiency. Initially, with fewer data values achieved by CMDCS, FCFS, and Random methods
collectors, competition among tasks for resources may be decrease as the number of DCs increases because they do
low, leading to suboptimal TCR, TP, and TR due to inefficient not consider the task’s preferences in selecting data collectors
task assignment. As the number of data collectors increases, that satisfy the needs of each task, unlike FRCMDCS.
competition intensifies, potentially enhancing TCR, TP, and In Fig. 6b, for the scenario with a large number of DCs,
TR by leveraging more available resources. However, beyond the SS values for the proposed FRCMDCS are less affected
a certain point, further increases in data collectors may lead by changes in the number of DCs, ranging between [99%,
to diminishing returns or inefficiencies in task allocation, 100%], and are significantly higher than other methods,
thereby influencing TCR and TP in a non-linear manner. which range between [7%, 35%], [0.5%, 6%], [2%, 8%], and
This dynamic reflects the sensitivity of FRCMDCS to [13%, 28%] for CMDCS, FCFS, Random, and GOTAFull,
resource availability and task characteristics, emphasizing respectively. Similar to the small DCs scenario, FRCMDCS
the need for adaptive algorithms in complex crowdsourcing considers both the ranking contribution and the contribution
environments. density of each data collector, while CMDCS only considers
In Fig. 5, the achieved values of TSDD by all methods the contribution density. FCFS and Random methods do
for both small and large DCs experiments are depicted. not consider either of these factors. Additionally, the SS
In Fig. 5a, for the scenario with a small number of DCs, values achieved by CMDCS, FCFS, and Random methods
the TSDD values for all methods increase as the number of decrease as the number of DCs increases for the same reasons
DCs increases. The limited number of DCs allows methods mentioned earlier.
to select more DCs to contribute to executing tasks. Notably,
the proposed FRCMDCS, CMDCS, and GOTAFUll methods 2) EFFECTS OF DIFFERENT NUMBER OF TASKS
select a lower number of DCs than FCFS and Random In this analysis, the impact of varying numbers of tasks is
methods. FRCMDCS and CMDCS take into account the task examined for all methods when the number of data collectors
bid of the task or the contribution density of the data collectors is set to 25 for small values and 400 for large values in each
in completing the tasks. While GOTAFUll takes into account task.
the task budget. Additionally, the achieved value of TSDD by Fig. 7 displays the achieved values of TCR by all methods
the proposed FRCMDCS is slightly larger than CMDCS and for small and large DCs experiments. In Fig. 7a, for a
GOTAFUll because FRCMDCS aims to balance the profit small number of DCs, the TCR values for the proposed
and reward values of the entire system, incentivizing the FRCMDCS, CMDCS, and Random methods decrease as the
selection of more DCs in executing the required tasks. number of tasks increases due to an insufficient number of
In Fig. 5b, for the scenario with a large number of DCs to execute those tasks. The TCR values for FRCMDCS,
DCs, the proposed FRCMDCS achieves a slightly larger CMDCS, FCFS, Random, GOTAFull methods are almost the
value of TSDD than CMDCS, and the TSDD values for same with small differences ranging between [66% to 100%],
both methods are less affected by changes in the number [65% to 100%], [47% to 100%], [53% to 100%], and [40% to
of DCs. With a greater number of DCs, both methods 100%], respectively. In Fig. 7b, for a large number of DCs, the
approach the upper bound of TSDD values, regardless of TCR values for the proposed FRCMDCS, CMDCS, Random,
the increase in DCs number. Conversely, FCFS Random, and GOTAL methods also decrease as the number of tasks
and GOTAFUll methods achieve high values of TSDD, increases, but this decrease occurs more slowly. This is
and they are also unstable and influenced by varying large because there are enough DCs to execute the tasks, but some
numbers of DCs. FCFS and Random methods select tasks and of them do not satisfy the requirements of the requested tasks.
their data collectors without considering their contribution to The TCR values for the proposed FRCMDCS and CMDCS
executing tasks. While GOTAFull is based on the grouping are better than FCFS and Random methods, ranging between
collaboration among data collectors. [75% to 100%], [73% to 100%] for FRCMDCS and CMDCS,

VOLUME 12, 2024 130791

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

FIGURE 8. Total profit vs number of tasks.

FIGURE 9. Total reward vs number of tasks.

and [63% to 100%], [66% to 100%], and [69% to 100%] for Fig. 9 shows the achieved values of TR by all methods
FCFS, Random, and GOTAFull methods, respectively. FCFS for small and large DCs experiments. In Fig. 9a, for a small
is unstable in both small and large DCs cases. number of DCs, the TR values for all methods increase with
Fig. 8 illustrates the achieved values of TP by all methods certain changes as the number of tasks increases because
for small and large DCs experiments. In Fig. 8a, for a small a larger number of tasks provide a chance to select more
number of DCs, the TP values for all methods increase DCs with additional rewards, affecting the total reward
with some changes as the number of tasks increases because of the system. The proposed FRCMDCS, CMDCS, and
more tasks require more DCs, resulting in more rewards and GOTAFull achieve lower values than FCFS and Random
affecting the total profit. The proposed FRCMDCS, CMDCS, methods because they can select the most appropriate data
and GOTAFull achieve larger values than FCFS and Random collectors with acceptable rewards. In Fig. 9b, for a large
methods. In Fig. 8b, for a large number of DCs, the TP values number of DCs, the TR values for the proposed FRCMDCS
for the proposed FRCMDCS, CMDCS, and GOTAFull also and CMDCS increase as the number of tasks increases. The
increase with some changes as the number of tasks increases. proposed FRCMDCS achieves higher values of TR than
This is because a larger number of DCs require more paid the CMDCS method. This is because a larger number of
rewards, impacting the final TP of the platform. However, the DCs provide more choices in selecting the most appropriate
proposed FRCMDCS and CMDCS achieve larger values than data collectors with acceptable rewards. However, both
FCFS, Random, and GOTAFull methods because they can methods achieve larger values than FCFS and Random
maximize the total number of completed tasks, influencing methods because they can maximize the total number of
the overall final TP. FCFS and Random methods select tasks completed tasks, affecting the overall final TR. FCFS and
and their data collectors without considering task bids or the Random methods achieve low values of TR, and they are
contribution density of data collectors. also unstable and influenced by varying large numbers of

130792 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

FIGURE 10. Total selected dcs vs number of tasks.

FIGURE 11. Satisfaction score vs number of tasks.

DCs. FCFS and Random methods select tasks and their data certain threshold, further increases in tasks may lead to dimin-
collectors without considering the expected rewards of the ishing returns or inefficiencies in task allocation, affecting
data collectors, and the large number of DCs reduces the TCR and TP in a non-linear fashion. This dynamic illustrates
chance of selecting the most appropriate DCs with acceptable the sensitivity of FRCMDCS to resource availability and task
rewards for executing the chosen sensing tasks. dynamics, highlighting the importance of adaptive algorithms
As shown in Figs. 7a, 7b, 8a, 8b, 9a, and 9b, the TCR, in managing complex crowdsourcing scenarios.
TP, and TR of FRCMDCS do not exhibit monotonic trends Fig. 10 illustrates the achieved values of TSDD by all
with respect to the number of tasks. This non-monotonic methods for small and large DCs experiments. In Fig. 10a,
behavior observed in TCR and TP for FRCMDCS concerning for a small number of DCs, the TSDD values for all methods
dc can be attributed to the algorithm’s adaptive nature and the increase as the number of DCs increases because the limited
dynamic environment of crowdsourcing. FRCMDCS utilizes number of DCs provides a chance for methods to select more
sophisticated decision-making techniques, such as fuzzy DCs to contribute to executing the tasks. However, TSDD
logic and multi-criteria evaluation, to prioritize tasks and values of FCFS and Random methods are more unstable
allocate them based on data collector contributions. As the in their increase because both of them did not consider
number of tasks fluctuates, FRCMDCS adjusts its strategy the contribution density of data collectors in the selection
to optimize resource use and task completion efficiency. process. The proposed FRCMDCS and CMDCS methods
Initially, with fewer tasks, there may be low competition selected a lower number of DCs than FCFS, Random, and
among tasks, resulting in suboptimal TCR, TP, and TR due GOTAFull methods because the proposed FRCMDCS and
to inefficient data collectors. As more tasks are assigned, CMDCS take into account the task bid of the task or the
competition intensifies, potentially improving TCR, TP, and contribution density of the data collectors in completing
TR by leveraging additional resources. However, beyond a the tasks. Simultaneously, the achieved value of TSDD by

VOLUME 12, 2024 130793

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

the proposed FRCMDCS is slightly larger than CMDCS Null Hypothesis (H0): There is no difference in the mean
because the proposed FRCMDCS aims to satisfy the balance system profit between the proposed FRCMDCS method and
between the profit and the reward values of the whole system, any one of the compared approaches.
incentivizing the selection of more DCs in executing the Alternative Hypothesis (H1): There is a significant
required tasks. In Fig. 11b, for a large number of DCs, difference in the mean system profit between the proposed
the TSDD values for all methods increase as the number FRCMDCS method and ny one of the compared approaches.
of DCs increases, and the proposed FRCMDCS achieves a Then we will calculate the differences in the system profit
larger value of TSDD than CMDCS with some differences. Di,j between the proposed FRCMDCS method and each one
While FCFS and Random methods achieve high values of compared approach j for experiment i by using the following
TSDD, and they are also unstable and affected by varying equation:
large numbers of DCs. This is because FCFS and Random
methods select the task and its data collectors without any Di,j = Pi − Ei,j , 1 ≤ i ≤ NE, 1 ≤ j ≤ NA (22)
consideration for their contribution to the executing tasks.
Fig. 11 displays the achieved values of SS by all methods where, Pi represents the system profit by the proposed
for small and large DCs experiments. In Fig. 11a, for a small FRCMDCS method, and Ei,j represents the system profit
number of DCs, the SS value for the proposed FRCMDCS compared approach j for experiment i. NE represents the
decreases as the number of tasks increases, and other methods number of experiments (which was 9) and NA represents the
are affected with a small decreasing manner by this increase number of compared approaches (which was 4 approaches).
in the number of tasks, although it is much higher than by using equation (22), we will compute the Mean and
other methods which range between [66% to 100%], [25% Standard Deviation of Differences by using the following
to 46%], [3% to 20%], [8% to 18%], and [8% to 17%] equations:
for FRCMDCS, CMDCS, FCFS, Random, and GOTAFull NE
1 X
methods, respectively. This is because FRCMDCS takes into D̄j = Di,j , 1 ≤ j ≤ NA (23)
account the ranking contribution of each data collector in NE
i=1
the process of executing the requested task in addition to v
u NE
its contribution density. While CMDCS considers only the u 1 X
SDj = t (Di,j − D̄j )2 , 1 ≤ j ≤ NA (24)
contribution density without the rank of data collectors to NE − 1
i=1
the task, and FCFS and Random methods did not consider
both of them. Additionally, CMDCS, FCFS, and Random by using equations (23) and (24), we can calculate the
methods did not consider the task’s preferences in selecting t-statics as follows:
the data collectors that satisfy the needs of each task, while
D̄j
FRCMDCS did. In Fig. 11b, for a large number of DCs, tj = √ , 1 ≤ j ≤ NA (25)
again the SS values for the proposed FRCMDCS decrease SDj / NE
as the number of tasks increases, and other methods are less
Then we compared the t-statistic tj for each compared
affected by this increase in the number of tasks, and it is
approach to critical value by using a t-distribution table for
much higher than other methods which range between [75%
Degrees of Freedom df = NE − 1, which will be 8, at a
to 100%], [1% to 13%], [1% to 3%], [2% to 4%], and [15% to
chosen significance level (e.g., 0.05).
20%] FRCMDCS, CMDCS, FCFS, Random, and GOTAFull
As shown in Table 3, the values of tj of CMDCS, FCFS,
methods, respectively. This is because FRCMDCS takes into
and Random were 5.09, 11.29, 20.76, repectively which
account the ranking contribution and the contribution density
are much larger than 1.86 that represents the critical value
of each data collector. While CMDCS considers only the
for t0.05,8 in the t-distribution table for degrees of freedom
contribution density, and FCFS and Random methods did not
df = 8. Therefor, the null hypothesis is rejected, indicating
consider both of them.
a significant difference in profits between the proposed
FRCMDCS method and all compared approaches.
D. THE STATISTICAL ANALYSIS VALIDATION FOR FRCMDCS
In this section, to validate the findings of FRCMDCS, based
2) CONFIDENCE INTERVALS:
on the five times conducted experiments to measure the
system profit for the FRCMDCS method and each existing To estimate the mean system profit for the proposed
method, we will determine the statistical significance of FRCMDCS method with a 95%, the confidence interval
FRCMDCS using t-test with p-value and confidence Intervals based on our 9 experimental runs is calculated as follows:
by comparing the proposed FRCMDCS method with each Calculating the Mean and Standard Deviation of Differ-
existing approach to determine if there are significant ences for the proposed FRCMDCS method by using the
differences in system profit, as follows: following equations:
NE
1) THE PAIRED T-TEST: 1 X
P̄ = Pi (26)
Hypothesises: NE
i=1

130794 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

v
u NE • Step 3: This is a simple selection operation after sorting.
u 1 X
SP = t (Pi − P̄)2 (27) So, the complexity is O(1).
NE − 1
i=1 • Step 4: This is similar to step 1 but for data collectors.
determining the t-value for a 95% confidence level and So, the complexity is O(n3 ).
df = 8, the t-value t0.025,8 is approximately 2.306. • Step 5: This is similar to step 2 but for data collectors.
computing the margin of error as follows: So, the complexity is O(|DN | · n).
• Step 6: This involves computing contribution densities
SP
MERR = t0.025,8 × √ (28) for all candidate data collectors. So, the complexity is
NE O(|DN |).
So, the confidence interval CI is calculating as follows • Step 7: Updating and checking a matrix for each data

CI = (P̄ − MERR, P̄ + MERR) (29) collector selection. So, the complexity is O(1).
• Step 8: Simple deletion operation. So, the complexity is
As shown in Table 3, the MERR was ≈ 2571.73, O(1).
by substituting in equation (29), we get Therefor, by considering the looping structure of the
CI = (16611.67 − 2571.73, 16611.67 + 2571.73) algorithm and sum the complexities of each step inside the
loop, the overall complexity will be O(T · n3 + T 2 · n + T ·
= (14039.94, 19183.4) (30)
D · n + T · D), where T represents the number of tasks and D
This means that there is 95% confident that the true represents the number of data collectors.
mean system profit for the proposed FRCMDCS method lies In scenarios where T and D are large, the dominant term
between 14039.94 and 19183.4. depends on their relative sizes and the number of criteria
n. If n, T , and D are all substantial, the complexity will be
TABLE 3. The statistical analysis of FRCMDCS and existing approaches dominated by T 2 · n and T · D · n.
using t-test and p-value in terms of system profit.
Therefore, the final computational complexity can be
summarized as

O(T · n3 + T 2 · n + T · D · n)

VI. IMPLICATIONS AND LIMITATIONS

The proposed FRCMDCS offers a versatile framework that
can be adapted and extended to various domains beyond its
original context as follows:
• Adaptation to Different Platforms and Tasks: The
weighting strategy using the Fuzzy Analytical Hierarchy
Process (FAHP) can be adjusted and applied to different
platforms and types of tasks, allowing for customization
based on specific domain requirements. For instance,
in healthcare, preferences may revolve around patient
E. THE COMPUTATIONAL COMPLEXITY FOR FRCMDCS privacy and data accuracy, while in manufacturing, they
To compute the computational complexity of the FRCMDCS may focus on production efficiency and quality control.
algorithm, we need to analyze the complexity of each step • Extension to Different Ranking Methods: While the
and sum them up. We will denote the following variables: paper employs the Fuzzy Technique for Order Pref-
|ST | is the number of requested tasks, |DN | is the number of erence by Similarity to Ideal Solution (FTOPSIS) for
candidate data collectors, n is the number of criteria used in ranking tasks and data collectors, other ranking methods
FAHP and FTOPSIS, and m is the number of requested tasks could be explored and integrated based on the domain’s
processed in the loop. needs and characteristics. For example, in financial ser-
Now, we will analysis the computational complexity step vices, techniques like utility theory or Bayesian ranking
by step as follows: could be considered for decision-making processes.
• Step 1: FAHP typically involves pairwise comparisons • Integration of Additional Metrics: The introduction
and eigenvalue computations. The complexity for n of Contribution Density with Ranking as a metric to
criteria is O(n3 ) due to eigenvalue computations. So, the measure individual data collectors’ contributions can
complexity is O(n3 ). be expanded by incorporating other relevant metrics or
• Step 2: FTOPSIS involves normalizing the decision performance indicators specific to different domains.
matrix, calculating the weighted normalized decision For example, in environmental monitoring, metrics
matrix, determining the ideal and negative-ideal solu- related to data accuracy, coverage, and timeliness
tions, and calculating the distance of each alternative could be included to assess collectors’ contributions
from these solutions. So, the complexity is O(m · n). comprehensively.

VOLUME 12, 2024 130795

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

• Application in Various IoT Environments: The conflicting priorities and uncertainties inherent in col-
proposed scheme’s applicability in IoT environments laborative environments. One solution to this limitation
suggests its potential for adaptation and deployment is using advanced optimization algorithms or machine
in diverse settings, such as smart cities, agriculture, learning models to automate decision-making processes.
transportation, and energy management. By tailoring Conduct thorough sensitivity analyses and scenario
the scheme’s parameters and methodologies to suit planning to anticipate and manage conflicting priorities
specific IoT applications, its effectiveness can be further effectively.
enhanced.
So, the flexibility and effectiveness demonstrated by the
proposed scheme make it a promising candidate for adapta- VII. CONCLUSION
tion and extension across a wide range of domains and IoT This paper presented a fuzzy ranking collaborative multi-task
environments, offering valuable insights and improvements data collection scheme that considers the preferences of the
in collaborative multi-task data collection scenarios. platform and the task and the uncertainty in data collectors’
However, while the fuzzy ranking collaborative multi-task contributions. The primary objective is to maximize the
data collection scheme demonstrates versatility and effective- total system profit, ensuring acceptable rewards for all data
ness, it also carries some limitations as follows: collectors while satisfying these preferences. The proposed
• Complexity of Implementation: The incorporation of method comprises three key elements: (1) A weighting
fuzzy logic and weighting strategies may introduce strategy for the preferences of the platform and the task
complexity in implementation and interpretation, neces- using the Fuzzy Analytical Hierarchy Process (FAHP)
sitating careful calibration and validation processes. to determine optimal weights for each criterion; (2) A
One solution to this limitation is simplifying the fuzzy ranking method employing the Fuzzy Technique for Order
logic and weighting strategies by adopting predefined Preference by Similarity to Ideal Solution (FTOPSIS) to
algorithms or frameworks that require less calibration. rank the tasks and the data collectors based on their
Utilize clear documentation and validation tools to required preferences and their determined weights from
streamline implementation. FAHP; and (3) Introduction of Contribution Density with
• Dependence on Input Data Accuracy: The scheme’s Ranking as a metric to measure the contribution of an
performance heavily relies on the accuracy of input data, individual data collector to a specific sensing task. Extensive
including platform and task preferences, which can be experiments were conducted to validate the effectiveness
challenging to ascertain, particularly in dynamic envi- of the proposed strategy. The results demonstrate that
ronments. One solution to this limitation is implement- the proposed scheme outperforms existing approaches in
ing robust data validation techniques and error handling terms of achieving a balance metric between profit and
mechanisms to mitigate inaccuracies. Incorporate data rewards, as well as a higher satisfaction score for the
quality checks and real-time feedback mechanisms to system.
ensure accuracy in dynamic environments. In the future work, the proposed scheme will be applied in a
• Scalability Challenges: In large-scale data collection real scenario of IoTs environments and it could focus on sev-
scenarios, the computational complexity of ranking eral specific areas to enhance the effectiveness and applica-
and weighting processes may increase substantially, bility of the proposed fuzzy ranking collaborative multi-task
potentially leading to scalability issues. One solution data collection scheme. Firstly, exploring dynamic adaptation
to this limitation is employing distributed comput- mechanisms to adjust weighting and ranking strategies based
ing techniques or cloud-based solutions to handle on evolving platform and task preferences would enhance
large-scale data efficiently. Optimize algorithms for adaptability in dynamic environments. Secondly, integrat-
parallel processing and leverage scalable infrastructure ing real-time feedback from data collectors and platform
to manage computational complexity. stakeholders could improve input data accuracy and scheme
• Need for Customization: The scheme’s applicability responsiveness. Addressing scalability challenges through
across diverse domains may require significant cus- the development of scalable algorithms and distributed
tomization and domain-specific adjustments, potentially computing techniques would enable broader applicability.
limiting its generalizability and ease of adoption. Additionally, conducting comparative studies across domains
One solution to this limitation is developing modular could inform the development of more generalized versions
components that allow easy customization for dif- of the scheme. Investigating multi-objective optimization
ferent domains. Provide configurable parameters and techniques and ethical considerations would ensure a balance
domain-specific templates to facilitate adaptation while between competing objectives and address ethical impli-
maintaining core functionality. cations. Understanding human-computer interaction aspects
• Difficulty in Achieving Optimal Trade-offs: While and conducting real-world deployment studies would provide
aiming to balance profit, rewards, and system satis- insights into user behavior and scheme performance in
faction, achieving an optimal trade-off among these practical settings, guiding further refinement and deployment
objectives may prove challenging in practice due to efforts.

130796 VOLUME 12, 2024

A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

ACKNOWLEDGMENT [19] S. Zhao and A. Liu, ‘‘High-performance target tracking scheme with low
The authors would like to thank Al-Azhar University and prediction precision requirement in WSNs,’’ Int. J. Ad Hoc Ubiquitous
Comput., vol. 29, no. 4, p. 270, 2018.
King Abdulaziz University, for its support for doing this [20] W. Qi, W. Liu, X. Liu, A. Liu, T. Wang, N. N. Xiong, and Z. Cai,
research. The completion of this research would not have ‘‘Minimizing delay and transmission times with long lifetime in code
been possible without the contributions and support of many dissemination scheme for high loss ratio and low duty cycle wireless sensor
networks,’’ Sensors, vol. 18, no. 10, p. 3516, Oct. 2018.
individuals and organizations. They are deeply grateful to [21] X. Xiang, W. Liu, N. N. Xiong, H. Song, A. Liu, and T. Wang, ‘‘Duty cycle
all those who played a role in the success of this research. adaptive adjustment based device to device (D2D) communication scheme
In addition, they wish to thank the Journal for helping to for WSNs,’’ IEEE Access, vol. 6, pp. 76339–76373, 2018.
[22] X. Wang, Z. Ning, and L. Wang, ‘‘Offloading in Internet of Vehicles: A fog-
publish our research. enabled real-time traffic management system,’’ IEEE Trans. Ind. Informat.,
vol. 14, no. 10, pp. 4568–4578, Oct. 2018.
REFERENCES [23] C. Yang, Z. Shi, K. Han, J. J. Zhang, Y. Gu, and Z. Qin, ‘‘Optimization
of particle CBMeMBer filters for hardware implementation,’’ IEEE Trans.
[1] M. Weiser, ‘‘The computer for the 21 st century,’’ Sci. Amer., vol. 265, no. 3, Veh. Technol., vol. 67, no. 9, pp. 9027–9031, Sep. 2018.
pp. 94–105, 1991.
[24] T. Li, N. N. Xiong, J. Gao, H. Song, A. Liu, and C. Zhiping, ‘‘Reliable code
[2] B. Johanson, A. Fox, and T. Winograd, ‘‘The interactive workspaces disseminations through opportunistic communication in vehicular wireless
project: Experiences with ubiquitous computing rooms,’’ IEEE Pervasive networks,’’ IEEE Access, vol. 6, pp. 55509–55527, 2018.
Comput., vol. 1, no. 2, pp. 67–74, Aug. 2002.
[25] M. Huang, A. Liu, N. N. Xiong, T. Wang, and A. V. Vasilakos, ‘‘A
[3] S. Voida, E. D. Mynatt, B. MacIntyre, and G. M. Corso, ‘‘Integrating virtual low-latency communication scheme for mobile wireless sensor control
and physical context to support knowledge workers,’’ IEEE Pervasive systems,’’ IEEE Trans. Syst., Man, Cybern., Syst., vol. 49, no. 2,
Comput., vol. 1, no. 3, pp. 73–79, Jul. 2002. pp. 317–332, Feb. 2019.
[4] J. O’Donoghue and J. Herbert, ‘‘Data management within mHealth [26] X. Li, W. Liu, M. Xie, A. Liu, M. Zhao, N. N. Xiong, M. Zhao, and W. Dai,
environments: Patient sensors, mobile devices, and databases,’’ J. Data Inf. ‘‘Differentiated data aggregation routing scheme for energy conserving and
Qual., vol. 4, no. 1, pp. 1–20, Oct. 2012. delay sensitive wireless sensor networks,’’ Sensors, vol. 18, no. 7, p. 2349,
[5] S. Sarkar, S. Chatterjee, and S. Misra, ‘‘Assessment of the suitability of Jul. 2018.
fog computing in the context of Internet of Things,’’ IEEE Trans. Cloud [27] X. Ju, W. Liu, C. Zhang, A. Liu, T. Wang, N. N. Xiong, and Z. Cai,
Comput., vol. 6, no. 1, pp. 46–59, Jan. 2018. ‘‘An energy conserving and transmission radius adaptive scheme to
[6] X. Liu, Y. Liu, N. N. Xiong, N. Zhang, A. Liu, H. Shen, and optimize performance of energy harvesting sensor networks,’’ Sensors,
C. Huang, ‘‘Construction of large-scale low-cost delivery infrastructure vol. 18, no. 9, p. 2885, Aug. 2018.
using vehicular networks,’’ IEEE Access, vol. 6, pp. 21482–21497, 2018. [28] J. Tan, W. Liu, M. Xie, H. Song, A. Liu, M. Zhao, and G. Zhang,
[7] T. Li, Y. Liu, N. N. Xiong, A. Liu, Z. Cai, and H. Song, ‘‘Privacy-preserving ‘‘A low redundancy data collection scheme to maximize lifetime using
protocol for sink node location in telemedicine networks,’’ IEEE Access, matrix completion technique,’’ EURASIP J. Wireless Commun. Netw.,
vol. 6, pp. 42886–42903, 2018. vol. 2019, no. 1, p. 5, 2019.
[8] Z. Ding, K. Ota, Y. Liu, N. Zhang, M. Zhao, H. Song, A. Liu, and [29] X. Liu, Q. Yang, J. Luo, B. Ding, and S. Zhang, ‘‘An energy-aware
C. Zhiping, ‘‘Orchestrating data as a services-based computing and offloading framework for edge-augmented mobile RFID systems,’’ IEEE
communication model for information-centric Internet of Things,’’ IEEE Internet Things J., vol. 6, no. 3, pp. 3994–4004, Jun. 2019.
Access, vol. 6, pp. 38900–38920, 2018. [30] S. Zhang, G. Wang, M. Z. A. Bhuiyan, and Q. Liu, ‘‘A dual privacy
[9] H. Zhou, H. Wang, X. Li, and V. C. M. Leung, ‘‘A survey on mobile data preserving scheme in continuous location-based services,’’ IEEE Internet
offloading technologies,’’ IEEE Access, vol. 6, pp. 5101–5111, 2018. Things J., vol. 5, no. 5, pp. 4191–4200, Oct. 2018.
[10] M. Huang, A. Liu, M. Zhao, and T. Wang, ‘‘Multi working sets alternate [31] X. Liu, J. Yin, S. Zhang, B. Ding, S. Guo, and K. Wang, ‘‘Range-based
covering scheme for continuous partial coverage in WSNs,’’ Peer–Peer localization for sparse 3-D sensor networks,’’ IEEE Internet Things J.,
Netw. Appl., vol. 12, no. 3, pp. 553–567, May 2019. vol. 6, no. 1, pp. 753–764, Feb. 2019.
[11] Y. Liu, M. Ma, X. Liu, N. N. Xiong, A. Liu, and Y. Zhu, ‘‘Design and [32] Q. Deng, H. Zeng, J. Zhang, S. Tian, J. Cao, Z. Li, and A. Liu, ‘‘Com-
analysis of probing route to defense sink-hole attacks for Internet of pressed sensing for image reconstruction via back-off and rectification of
Things Security,’’ IEEE Trans. Netw. Sci. Eng., vol. 7, no. 1, pp. 356–372, greedy algorithm,’’ Signal Process., vol. 157, pp. 280–287, Apr. 2019.
Jan. 2020. [33] B. Huang, A. Liu, C. Zhang, N. Xiong, Z. Zeng, and Z. Cai, ‘‘Caching
[12] H. Teng, Y. Liu, A. Liu, N. N. Xiong, Z. Cai, T. Wang, and X. Liu, ‘‘A novel joint shortcut routing to improve quality of service for information-centric
code data dissemination scheme for Internet of Things through mobile networking,’’ Sensors, vol. 18, no. 6, p. 1750, May 2018.
vehicle of smart cities,’’ Future Gener. Comput. Syst., vol. 94, pp. 351–367, [34] F. Xiao, X. Xie, Z. Li, Q. Deng, A. Liu, and L. Sun, ‘‘Wireless network
May 2019. optimization via physical layer information for smart cities,’’ IEEE Netw.,
[13] X. Wang, Z. Ning, M. Zhou, X. Hu, L. Wang, B. Hu, R. Y. K. Kwok, vol. 32, no. 4, pp. 88–93, Jul. 2018.
and Y. Guo, ‘‘A privacy-preserving message forwarding framework for [35] Y. Ren, W. Liu, Y. Liu, N. N. Xiong, A. Liu, and X. Liu, ‘‘An effective
opportunistic cloud of things,’’ IEEE Internet Things J., vol. 5, no. 6, crowdsourcing data reporting scheme to compose cloud-based services in
pp. 5281–5295, Dec. 2018. mobile robotic systems,’’ IEEE Access, vol. 6, pp. 54683–54700, 2018.
[14] J. Zhang, X. Hu, Z. Ning, E. C.-H. Ngai, L. Zhou, J. Wei, J. Cheng, [36] (2013). Waze-Outsmarting Traffic Together. [Online]. Available:
and B. Hu, ‘‘Energy-latency tradeoff for energy-aware offloading in http://www.waze.com/
mobile edge computing networks,’’ IEEE Internet Things J., vol. 5, no. 4, [37] Buuuk. (2012). Weatherlah Iphone Application. [Online]. Available:
pp. 2633–2645, Aug. 2018. http://itunes.apple.com/us/app/weatherlah/id411646329?mt=8
[15] M. Wu, Y. Wu, C. Liu, Z. Cai, N. N. Xiong, A. Liu, and M. Ma, [38] T. Li, K. Ota, T. Wang, X. Li, Z. Cai, and A. Liu, ‘‘Optimizing the coverage
‘‘An effective delay reduction approach through a portion of nodes with via the UAVs with lower costs for information-centric Internet of Things,’’
a larger duty cycle for industrial WSNs,’’ Sensors, vol. 18, no. 5, p. 1535, IEEE Access, vol. 7, pp. 15292–15309, 2019.
May 2018. [39] Y. Liu, A. Liu, X. Liu, and X. Huang, ‘‘A statistical approach to participant
[16] C.-M. Huang, Y.-F. Chen, S. Xu, and H. Zhou, ‘‘The vehicular social selection in location-based social networks for offline event marketing,’’
network (VSN)-based sharing of downloaded geo data using the credit- Inf. Sci., vol. 480, pp. 90–108, Apr. 2019.
based clustering scheme,’’ IEEE Access, vol. 6, pp. 58254–58271, 2018. [40] J. Tan, W. Liu, T. Wang, N. N. Xiong, H. Song, A. Liu, and Z. Zeng,
[17] Y. Ren, Y. Liu, N. Zhang, A. Liu, N. N. Xiong, and Z. Cai, ‘‘Minimum- ‘‘An adaptive collection scheme-based matrix completion for data
cost mobile crowdsourcing with QoS guarantee using matrix completion gathering in energy-harvesting wireless sensor networks,’’ IEEE Access,
technique,’’ Pervas. Mobile Comput., vol. 49, pp. 23–44, Sep. 2018. vol. 7, pp. 6703–6723, 2019.
[18] H. Zhou, H. Wang, X. Chen, X. Li, and S. Xu, ‘‘Data offloading techniques [41] Y. Ren, W. Liu, T. Wang, X. Li, N. N. Xiong, and A. Liu, ‘‘A collaboration
through vehicular ad hoc networks: A survey,’’ IEEE Access, vol. 6, platform for effective task and data reporter selection in crowdsourcing
pp. 65250–65259, 2018. network,’’ IEEE Access, vol. 7, pp. 19238–19257, 2019.

VOLUME 12, 2024 130797

 A. A. A. Gad-Elrab: Fuzzy Ranking Collaborative Multi-Tasks Data Collection Scheme

[42] H. Teng, W. Liu, T. Wang, A. Liu, X. Liu, and S. Zhang, ‘‘A cost- [63] J. Jiang, B. An, Y. Jiang, C. Zhang, Z. Bu, and J. Cao, ‘‘Group-oriented
efficient greedy code dissemination scheme through vehicle to sensing task allocation for crowdsourcing in social networks,’’ IEEE Trans. Syst.,
devices (V2SD) communication in smart city,’’ IEEE Access, vol. 7, Man, Cybern., Syst., vol. 51, no. 7, pp. 4417–4432, Jul. 2021.
pp. 16675–16694, 2019. [64] J. Jiang, K. Di, B. An, Y. Jiang, Z. Bu, and J. Cao, ‘‘Batch crowdsourcing
[43] J. Gui and K. Zhou, ‘‘Cellular throughput optimization by game-based for complex tasks based on distributed team formation in E-markets,’’
power adjustment and outband D2D communication,’’ EURASIP J. IEEE Trans. Parallel Distrib. Syst., vol. 33, no. 12, pp. 3600–3615,
Wireless Commun. Netw., vol. 2018, no. 1, p. 254, 2018. Dec. 2022.
[44] S. Zhang, K.-K.-R. Choo, Q. Liu, and G. Wang, ‘‘Enhancing privacy [65] J. Jeon, J. Kim, M. Shin, and M. Kim, ‘‘A blockchain-based trust model
through uniform grid and caching in location-based services,’’ Future for supporting collaborative healthcare data management,’’ Comput. Syst.
Gener. Comput. Syst., vol. 86, pp. 881–892, Sep. 2018. Sci. Eng., vol. 46, no. 3, pp. 3403–3421, 2023.
[66] M. Shin, G. Park, C. Y. Park, J. Lee, and M. Kim, ‘‘Application-specific
[45] Z. Liu, T. Tsuda, H. Watanabe, S. Ryuo, and N. Iwasawa, ‘‘Data
feature selection and clustering approach with HPC system profiling data,’’
driven cyber-physical system for landslide detection,’’ Mobile Netw. Appl.,
J. Supercomput., vol. 77, no. 7, pp. 6817–6831, Jan. 2021.
vol. 24, no. 3, pp. 991–1002, Jun. 2019.
[67] J. J. Buckley, ‘‘Fuzzy hierarchical analysis,’’ Fuzzy Sets Syst., vol. 17, no. 3,
[46] M. Allahbakhsh, B. Benatallah, A. Ignjatovic, H. R. Motahari-Nezhad, pp. 233–247, Dec. 1985.
E. Bertino, and S. Dustdar, ‘‘Quality control in crowdsourcing systems:
[68] D.-Y. Chang, ‘‘Applications of the extent analysis method on fuzzy AHP,’’
Issues and directions,’’ IEEE Internet Comput., vol. 17, no. 2, pp. 76–81,
Eur. J. Oper. Res., vol. 95, no. 3, pp. 649–655, Dec. 1996.
Mar. 2013.
[69] Z. Ge and Y. Liu, ‘‘Analytic hierarchy process based fuzzy decision fusion
[47] D. Yang, G. Xue, X. Fang, and J. Tang, ‘‘Incentive mechanisms for system for model prioritization and process monitoring application,’’ IEEE
crowdsensing: Crowdsourcing with smartphones,’’ IEEE/ACM Trans. Trans. Ind. Informat., vol. 15, no. 1, pp. 357–365, Jan. 2019.
Netw., vol. 24, no. 3, pp. 1732–1744, Jun. 2016. [70] S. Kubler, J. Robert, W. Derigent, A. Voisin, and Y. Le Traon, ‘‘A state-of
[48] J. Lin, S. Amini, J. I. Hong, N. Sadeh, J. Lindqvist, and J. Zhang, the-art survey & testbed of fuzzy AHP (FAHP) applications,’’ Expert Syst.
‘‘Expectation and purpose: Understanding users’ mental models of mobile Appl., vol. 65, pp. 398–422, Dec. 2016.
app privacy through crowdsourcing,’’ in Proc. ACM Conf. Ubiquitous [71] P. T. Nguyen, N. B. Vu, L. Van Nguyen, L. P. Le, and K. D. Vo, ‘‘The
Comput., 2012, pp. 501–510. application of fuzzy analytic hierarchy process (F-AHP) in engineering
[49] D. C. Brabham, ‘‘Moving the crowd at threadless: Motivations for project management,’’ in Proc. IEEE 5th Int. Conf. Eng. Technol. Appl.
participation in a crowdsourcing application,’’ Inf., Commun. Soc., vol. 13, Sci. (ICETAS), Nov. 2018, pp. 1–4.
no. 8, pp. 1122–1145, 2010. [72] C.-L. Hwang and K. Yoon, ‘‘Methods for multiple attribute decision
[50] C. Zhu, V. C. M. Leung, J. J. P. C. Rodrigues, L. Shu, L. Wang, and making,’’ in Multiple Attribute Decision Making, vol. 186. Berlin,
H. Zhou, ‘‘Social sensor cloud: Framework, greenness, issues, and Germany: Springer, 1981, pp. 58–191.
outlook,’’ IEEE Netw., vol. 32, no. 5, pp. 100–105, Sep. 2018. [73] J. M. Benítez, J. C. Martín, and C. Román, ‘‘Using fuzzy number for
[51] Y. Zhang and M. van der Schaar, ‘‘Reputation-based incentive protocols measuring quality of service in the hotel industry,’’ Tourism Manage.,
in crowdsourcing applications,’’ in Proc. IEEE INFOCOM, Mar. 2012, vol. 28, no. 2, pp. 544–555, Apr. 2007.
pp. 2140–2148. [74] A. Lekbich, A. Belfqih, T. Ouaderhman, C. Zedak, J. Boukherouaa,
[52] Y. Wang, Z. Cai, G. Yin, Y. Gao, X. Tong, and G. Wu, ‘‘An incentive and F. El Mariami, ‘‘A multi-objective optimization-based model for the
mechanism with privacy protection in mobile crowdsourcing systems,’’ deployment of reclosers and remote-controlled switches using NSGA2 and
Comput. Netw., vol. 102, pp. 157–171, Jun. 2016. entropy weighted TOPSIS method,’’ Indonesian J. Electr. Eng. Comput.
Sci., vol. 20, no. 3, p. 1128, Dec. 2020.
[53] D. Zhao, X.-Y. Li, and H. Ma, ‘‘Budget-feasible online incentive
[75] C.-T. Chen, ‘‘Extensions of the TOPSIS for group decision-making under
mechanisms for crowdsourcing tasks truthfully,’’ IEEE/ACM Trans. Netw.,
fuzzy environment,’’ Fuzzy Sets Syst., vol. 114, no. 1, pp. 1–9, 2000.
vol. 24, no. 2, pp. 647–661, Apr. 2016.
[76] Andras and Admin. (2023). Andras. [Online]. Available: http://www.
[54] A. H. van Bunningen, L. Feng, and P. M. G. Apers, ‘‘Context for ubiquitous omnetpp.org/
data management,’’ in Proc. Int. Workshop Ubiquitous Data Manage.,
Apr. 2005, pp. 17–24.
[55] A. A. Gad-Elrab, T. Alzohairy, and B. Hassan, ‘‘An adaptive service
distribution genetic algorithm with multi-constraints in ubiquitous envi-
ronments,’’ Int. J. Comput. Sci. Inf. Secur., vol. 14, no. 12, p. 156, 2016.
[56] N. B. Salah and I. B. Saadi, ‘‘Fuzzy AHP for learning service
selection in context-aware ubiquitous learning systems,’’ in Proc. Int.
IEEE Conf. Ubiquitous Intell. Comput., Adv. Trusted Comput., Scalable
Comput. Commun., Cloud Big Data Comput., Internet People, Smart
World Congr. (UIC/ATC/ScalCom/CBDCom/IoP/SmartWorld), Jul. 2016, AHMED A. A. GAD-ELRAB received the B.S.
pp. 171–179. degree in computer science from the Faculty of
Science, Alexandria University, Egypt, in 1999,
[57] W. Xue, H. Pung, W. Ng, and T. Gu, ‘‘Data management for context-
aware computing,’’ in Proc. IEEE/IFIP Int. Conf. Embedded Ubiquitous the M.S. degree in computer science from the Fac-
Comput., vol. 1, Dec. 2008, pp. 492–498. ulty of Science, Cairo University, Egypt, in 2008,
[58] D. R. Karger, S. Oh, and D. Shah, ‘‘Budget-optimal task allocation for and the Ph.D. degree from Nara Institute of
reliable crowdsourcing systems,’’ Oper. Res., vol. 62, no. 1, pp. 1–24, Science and Technology (NAIST), Japan, in 2012.
Feb. 2014. He is currently an Associate Professor with the
[59] Y. Zheng, J. Wang, G. Li, R. Cheng, and J. Feng, ‘‘QASCA: A quality- Department of Computer Science, Faculty of
aware task assignment system for crowdsourcing applications,’’ in Proc. Computing and Information Technology, King
ACM SIGMOD Int. Conf. Manag. Data, 2015, pp. 1031–1046. Abdulaziz University, Jeddah, Saudi Arabia. He is also an Associate
[60] S. Basu Roy, I. Lykourentzou, S. Thirumuruganathan, S. Amer-Yahia, Professor of ubiquitous and mobile computing with the Department of
and G. Das, ‘‘Task assignment optimization in knowledge- Mathematics, Faculty of Science, Al-Azhar University, Cairo, Egypt. His
intensive crowdsourcing,’’ VLDB J., vol. 24, no. 4, pp. 467–491, current research interests include software engineering, cloud computing,
Aug. 2015. mobile computing, the Internet of Things applications, smart homes, data
[61] L. Pournajaf, L. Xiong, V. Sunderam, and S. Goryczka, ‘‘Spatial task science, sensor networks, dynamic distributed systems, big data, and mobile
assignment for crowd sensing with cloaked locations,’’ in Proc. IEEE 15th crowd sensing. He received the NAIST Best Ph.D. Student Award, in March
Int. Conf. Mobile Data Manage., vol. 1, Jul. 2014, pp. 73–82. 2012; the Outperformance Award from the Graduate School of Information
[62] B. Z. Hassan, A. A. Gad-Elrab, M. S. Farag, and S. E. A. Youssef, Science, NAIST, in March 2012; and the 2011 IPSJ Yamashita Memorial
‘‘An integrated node selection model using FAHP and FTOPSIS for data Award (given to only one or two papers among all papers presented in one
retrieval in ubiquitous computing,’’ Appl. Comput. Intell. Soft Comput., year in each IPSJ, SIG, and Japan).
vol. 2022, no. 1, 2022, Art. no. 8092432.

130798 VOLUME 12, 2024