QACM: Quality Aware Crowd Sensing in Mobile Computing

Abstract

Mobile computing is one of the significant opportunities that can be used for various practical applications in numerous fields in real life. Due to inherent characteristics of ubiquitous computing, devices can gather numerous types of data that led to innovative applications in many fields with a unique emerging prototype known as Crowd sensing. Here, the involvement of people is one of the important features and their mobility provides an exclusive opportunity to collect and transmit the data over a substantial geographical area. Thus, we put forward novel idea about Quality of Information (QOI) with unique parameters with opportunistic uniqueness of people’s mobility in terms of sensing and transmission. Additionally, we propose some of the viable improved ideas about the competent opportunistic data collection through efficient techniques. This work also considered some of the open issues mentioned by previous related works.

1. Introduction

In recent years, Internet of Things (IOT) played an important role in connecting various physical environments from the real world for useful applications through sensing data and interconnecting various objects [1,2]. In addition to this, ubiquitous computing provides more sensible data collection through sensors and transmission technologies. Now it is the time to exploit the capacity of ubiquitous computing through mobile devices such as smart phones which are owned today by most people for their daily routine communications [3].

In fact, the sensors are usually deployed to sense data over the limited areas. However, this technique is not so economically feasible for larger areas of data collection and transmission. Currently, a significant number of people carry smart mobile devices which are inherently pervasive in the real world and can be more effectively used instead of sensors. These mobile devices may be smart phones, wearable devices, or vehicle-mounted devices.

These devices make it possible to sense and monitor a large geographical area in large-scale events or occurrences. Such a sensing prototype is known as mobile crowd sensing (MCS) [4,5] or people/human centric sensing [6]. This prototype is more suitable for dense urban areas to measure/sense various parameters through specific/identified groups of mobile users. In recent mobile technology, all smart phones are equipped with different types of sensors, GPS, receivers, Bluetooth, and Wi-Fi capabilities. These functions made it possible to collect more useful data and send them to major monitoring and processing devices. These devices distribute/analyze the data based on the applications accordingly. There are various applications that have been emerged based on this prototype including traffic monitoring and control, air pollution control and monitoring, different security surveillance systems, and even a possible device to monitor and control the pandemic COVID-19 situation.

MCS mainly requires greater participation of people involvement to attain better efficiency in sensing. Since the people are the better option in sensing coverage and data transmission. Different research in this field have been collectively devised two major models or prototypes of MCS, known as Participatory sensing and opportunistic sensing.

The participatory sensing expects more conscious participant’s involvement in sensing the data based on the demands of different application requirements. These requirements are judged based on the specific attributes such as when. Where, what, and how to sense data. On other hand opportunistic sensing does not requires dedicated/conscious participants to sense the data. here data can be collected just by running the required wireless communication model.

This opportunistic sensing is more feasible for large area deployments with versatile applications. MCS consists of two types of paradigms that influence the different applications based on their demand of sensing and monitoring. These transmission paradigms are known as infrastructure-based transmission and opportunistic transmission. Infrastructure-based transmissions are mainly depending on the internet through which the user can send the sensed data to the sink or processing devices, whereas opportunistic transmission uses wireless communication modules such as Wi-Fi or Bluetooth supported by short range radio communications. These communication modules help the user to send the sensed data through the intermittent connections. The infrastructure-based transmission paradigm is one of the prioritized options in most of the current applications. However, this transmission paradigm showed poor performance due to lack of consistent network coverage areas, whereas opportunistic transmission performs better with higher efficiency and lower costs. More importantly, it does not require a centralized server or extra infrastructure for communication and management. This transmission minimizes the workload of the cellular network with negligible congestion rates.

Mobile crowd sensing is one of the significant paradigms that plays a major role in mobility, especially for urban environments. Present day mobile devices are designed with a variety of sensors that can be used to develop most essential applications for urban environments using crowd sensing. Some of the applications are traffic control and monitoring, pollution control, crime control, and helping to develop a better platform to solve the issues related to pandemic diseases like COVID-19, etc. In our research work, we are proposing a feasible model for mobile crowd sensing with better quality of Information (QOI) for sensing and communication. This work considered significant parameters that effectively improve the QOI. We have developed an efficient mechanism to categorize the mobile users for crowd sensing who can be involved in different applications that demand different levels of QOI. We identified the most relevant and effective parameters such as reliability based on distance (R_bd), transmission range to sink (T_ri), time required to send to sink/destination (T_ri), and quality of sensor devices used in mobile devices (M_qs). These parameters introduce remarkable QOI for various applications of crowd sensing.

This paper is organized as follows. A literature review is presented in Section 2. Section 3 presents the methodology. The detailed performance evaluation is presented in Section 4. The conclusions are given in Section 5.

1.1. Background

Most of the existing works in mobile crowd sensing use different measures to improve the quality of sensing in unpredictable environmental conditions such as spatio-temporal aspects, selection of right participants, selfishness of participants, lack of infrastructure, etc. The different applications require different levels of quality expectations. In order to fulfill these qualities for various applications, we introduced a new concept with more relevant parameters such as such as reliability based on distance R_bd, transmission range to sink T_rt, time that requires to send to sink/destination T_rt, and quality of sensor devices used in mobile devices M_qs. These parameters introduce remarkable QOI in crowd sensing for various applications using the opportunistic transmission paradigm.

1.2. Problem Definition

In a given crowd sensing environment with an opportunistic transmission paradigm, there are three categories of mobile users. M_u1 is set of mobile users who can support very sensitive applications that requires the highest quality level of QOI. M_u2 is the set of mobile users who can support medium range applications, in which middle levels of QOI must be satisfied. M_u3 is the set of mobile users, who can support smaller range applications in which the QOI level is expected to be in the range between medium and lower levels of QOI. In each of these categorizations, it is designed and considered that the more effective parameters are reliability based on distance R_bd, transmission range to sink T_ri, time that requires to send to sink/destination T_rt, and quality of sensor devices used in mobile devices T_rt.

1.2.1. Objectives

• Determine if our method achieves better QOI in crowd sensing based on application requirements.
• To enhance the overall QOI level in crowd sensing for opportunistic transmission paradigm.

1.2.2. Assumptions

• The opportunistic transmission paradigm is the basis for this crowd sensing.
• Identified/specific groups of mobile users are selected based on incentive mechanism (based on category).

2. Literature Review

Andrea Capponi et al. [1] gathered and analyzed huge data sets related to mobile crowd sensing systems and they presented different challenges involved in crowd sensing and provided a detailed survey on current state-of-the-art works. It provides the detailed classifications of applications, methodologies, and architectures. It also gives an idea for future research directions and relationships with other research areas. Daqing Zhang et al. [2] proposed a four-stage life cycle for the MCS process and tried to utilize 4W1H to distinguish the research issues for every stage of the MCS life cycle. This work explained the different possible research issues of MCS in a systematic approach and helped many researchers to find multiple options in different directions. Jose Mauricio et al. [3] presented a framework for MCS that devised a model to characterize the mobile user’s behavior and a novel incentive mechanism. This model describes the behavior of mobile users regarding their resource availability and dissimilar responses. The incentive mechanism proposed different values of incentives based on the user characteristics. However, this idea must be further considered for the real deployment case values to be within the limits of the total budget of the system.

Dragan Stojanovic et al. [4] presented an application of the MCS paradigm for more favorable and competent mobility in urban environments. It introduced the city sensing framework which signified the feasibility of a more general crowd sensing platform that can be applied to different urban mobility areas. This platform helps to design various applications like traffic sensing and pollution sensing with the help of various add-on sensors which are internally and externally attached to mobile devices. This work suggested incorporating people involvement, reliability, security, and privacy in future work.

Susu Xu et al. [5] introduces an algorithm called iLoCuS, that proposes an incentivizing method to realize a target sensing distribution in MCS using non-dedicated vehicle platforms. Because of different commercial aspects, the non-dedicated vehicles may not achieve the required spatio-temporal sensing coverage. iLoCuS gives an idea about incentivizing policies for vehicle agents to divide the sensing distribution into two different levels such as at the time location level and vehicle level so that more consistent results can be achieved with limited budgets. Furthermore, this work needed to consider the specific mobility and acceptance rate pattern of each vehicle agent. Augustin et al. [6] has studied and analyzed the data transfer opportunities among wireless devices carried by humans and it was found that the inter-contact time may be better approximated by a power law. Additionally, it recommended forwarding algorithms for the design of more efficient opportunistic.

Huading Ma et al. [7] investigated opportunistic characteristics of human mobility in terms of sensing and transmission. This work explained some of the possibilities of data collection through opportunistic methods. It also specified some of the research issues related to human involvements in MCS. Francesco et al. [8] presented a survey on QOI in MCS. This work proposed a framework which includes the definition of the concept of QOI. Additionally, it defined and analyzed different research challenges in imposing and approximating QOI. It mentioned and discussed some of the previous related works and analyzed their strengths and weaknesses. Tingting et al. [9] introduced a QOS Sensitive Task Assignment problem for crowd sensing and proposed an idea of rewarding each user based on their level of QOS achievement against the threshold value in the data collection process. This work supported medium-range QOS models and proposed extensions for more complex models.

Gurdit Singh et al. [10] devised a prototype to find inattentive driving behavior and authentication on urban roads. The data were sensed and Dynamic Time Wrapping (DWT) was used to detect the events collected from the smart phones with sensors. Here, multiple sensors were used to fuse the patterns for more precise detection. This prototype helped to detect careless driving events using MCS. Key Yan et al. [11] provided a framework to select a task in MCS by combining the facts like location privacy preservation, efficient resource consumption, and high task profits. Here, there was freedom to select the path, altering their privacy requirements according to their convenience without disclosing the location information, and destination inference attack. This work needed further consideration of the task completion rate payment for the users.

Hamed et al. [12] presented a framework to examine context-aware MCS. It specified three different viewpoints of concepts, functionalities, and context awareness. These constraints can be differentiated with diverse parameters that examine and categorize the present related works. This framework provides better insight in distinguishing the domain and cooperation type, context awareness, data sharing, incentive mechanisms, and other aspects of MCS.

Takamasa et al. [13] proposed people centric navigation (PCN), a positioning system that gives a local map of surrounding crowd. It computes the proximity among the people with the help of sensor data through Bluetooth RSS. This research work remarkably improved the position accuracy compared to other methods. Kefuyi et al. [14] presented a Fast-VPR (Vehicle Participant Recruitment) algorithm based on the study of a VPR problem. This work modeled the participant recruitment problem as a constrained maximization problem without an open cost constraint. Here, the recruiter utility is a linear combination of spatio-temporal coverage and recruiter cost. As these two parameters have different dimensional values, it needs to further consider the possible general issues. Bin et al. [15] presented the strengths of large-scale sensing by the mobile users in MCS. It surpassed the participatory sensing by implied and obvious participation and crowd surround data from both mobile social networks and mobile sensing services. This work also gave insight into balancing human and machine intelligence in MCS design.

Ken Yan et al. [16] proposed two task selection mechanisms for workers in mobile crowd sensing, based on the study on task bidding and task assignment for workers and platforms. This work achieved good results on better privacy reservation without much cost compared to earlier works. Weiping Zhu et al. [17] investigated the multitask allocation problem in which the different participants carry different devices and perform different tasks. Additionally, work is then accomplished by introducing a greedy discrete particle swarm optimization with a genetic algorithm. This algorithm assumes that the assigned tasks are static.

Venkat Surya et al. [18] presented some theoretical models for three possible selective data acquisition rules for MCS. These schemes were compared with non-selective but aware reputation data acquisition schemes. Additionally, they suggested that activation and deactivation of built-in sensors through mobile applications decrease platform utility. It was also suggested that no flexibility should be given to activate and deactivate some sensors by subscribing to the published content of the users in the surrounding areas. Yong Feng et al. [19] focused on privacy issues of MCS and carried out a complete literature systematic survey. Additionally, they analyzed the possible threats to the privacy of the people who are involved in crowd sensing. This work categorized the different privacy schemes and collected different possible existing solutions for privacy protection and evaluated them.

Ling Yun et al. [20] proposed a solution to achieve better quality crowd sensing with the least cost. This work formulated the sensing quality assurance problem and proved it to be NP-Hard. It also proposed a polynomial time greedy approximation algorithm. This algorithm selects a suitable number of participants to satisfy the goal of the research and approximates the optimal solution free from the threat of market manipulation. Tie Luo et al. [21] proposed an idea to tackle the problems associated with data credibility in MCS-based IOT applications. It introduced an extra layer of crowd sensing in addition to the basic crowd sensing concept called the cross-validation approach that exploits the power of crowd sensing. This platform integrates four simplicity techniques with the help of a progressive algorithm that is well suited for time-sensitive and critical quality MCS applications.

Nsikak et al. [22] analyzed many Android-based sensing applications and found that these applications are vulnerable to SSL exploitation, eavesdropping, and sensitive information disclosure. To overcome these problems, it introduced a security scheme that provides in-depth security with better encryption and authentication of data from location and motion sensors. Nsikak et al. [23] proposed a sensecript framework that automatically interprets and encrypts sensitive location information of users. This work relies on the K-means algorithm and a certificate-less aggregate signcryption scheme. It includes spatial coding as the data compression technique and message querying telemetry transport as a messaging protocol. It provides better confidentiality, integrity, and nonrepudiation as part of the security services. Federico Montori et al. [24] conducted surveys and analyzed and evaluated the different techniques and challenges in MCS to address the course of sensing problems and presented the impact of sparse and dense data in crowd sensing. Yueyue Chan et al. [25] presented a trajectory segment section algorithm in MCS that selects different trajectory segments of participants rather than random participants for crowd sensing. This objective can avoid wasting of budgets on random participants. We considered both uniform coverage and maximum coverage while designing a trajectory segment selection scheme.

3. Methodology

3.1. System Architecture and Mathematical Model

Figure 1 depicts the system architecture of the proposed model for QACM. This architecture is divided into five phases. The mobile users are identified based on the quality of sensing elements used in their devices in phase 1. Here, incentives are proportional to the quality of sensing elements present in their mobile devices. Phase 2 and phase 3 represent the categorization of mobile users based on QOI parameters. M_u1, M_u2, and M_u3 are the three categories which are ready for deployment in the subsequent phase 4. Phase 5 involves the transmission of data to the sink/destination through crowd sensing.

3.2. Reliability Based on Distance—R_bd

This can be calculated based on the distance between the source mobile device and sink through intermediate devices. Let $n_i$ be the number of mobile devices used to transmit the information to the sink S. Rbd can be calculated as follows.

$$R_{bd}=Distance \left(\left(n_i, n_{i+1}\right)+\left(n_{i+1}, n_{i+2}\right) \dots \left(n_n, n_s\right)\right)$$

(1)

The distance can be calculated using the Euclidian distance formula,

$$Distance\left(S,d\right)=\sqrt{\left(S_1-S_2\right)+{(d_1-d_2)}^2}$$

(2)

where S and d are the source node and destination node or sink, respectively.

3.3. Transmission Rate to Sink–T_rt

T_rt is the time elapsed between the sensing event time to transmission to the sink node. Its value depends on the delay that was incurred at each intermediate node. The T_rt can be calculated as

$$T_{rt}={{\displaystyle \sum }}_{i=1}^{n}di$$

(3)

where the di is delay at each node.

3.4. Transmission Range to Sink—T_ri

T_ri is the average distance between the sensing mobile device to sink node. It is purely based on the bandwidth availability between the source node to sink node. It can be calculated as

$$T_{ri}={{\displaystyle \sum }}_{i=1}^n(n_i\ast bw)$$

(4)

where $n_i$ is the number of nodes participating in crowd sensing and bw is the bandwidth available in each node.

3.5. Quality of Sensing of Device—M_qs

M_qs represents the quality of sensing in each device. The M_qs can be calculated as follows.

$$M_{qs}=N_{sd} \ast Q_s$$

(5)

where N_sd is the number of sensing devices and Q_s is the quality of each sensing element.

Function 1 explains the measurement of Quality of Information (QOI) based on the different threshold values of parameters such as R_bd, T_ri, T_rt, and M_qs. Here, the QOI acquires three different possible values, i.e., high, medium, and low. Function 2 categorizes the mobile users who participated in crowd sensing based on the QOI. Additionally, the QOI is categorize into three category such as M_u1, M_u2, and M_u3 and assigned as high, medium, and low, respectively.

Function 1: QOI_ Measurement ()

Inputs: Rbd, Tri, Trt and Mqs   for node i to n do     calculate Rbd,   calculate Tri,   calculate Trt     calculate Mqs   end for for node i to n do if (Rbd > 1 && Tri > 1 && Trt > 1 && Mqs > 1)       QOI = high;   else if (Rbd == 1 && Tri == 1 && Trt == 1 && Mqs == 1)       QOI = medium;     else if (Rbd < 1 && Tri < 1 && Trt < 1 && Mqs < 1)       QOI = low;     end if   end if end if end for

Function 2: Category_Mobile_users ()

Inputs: QOI for each node i to n do   if (QOI > high)       Insert i to {Mu1}     else if (QOI == medium)       Insert i to {Mu2}     else if (QOI < low)       Insert i to {Mu3}     end if     end if   end if end for

In Algorithm 1, the QACM algorithm describes the entire process of crowd sensing that explains the deployment of the sensing nodes involved in crowd sensing. Once the deployment is over, the QOI measurement function decides the quality of information based on different parameters (mentioned in Function 1). Then, categorization is done using the Categorization function (mentioned in Algorithm 1). There are three categories: M_u1, M_u2, and M_u3 assigned to high-, medium-, and low-range sensitivity requirement applications.

Algorithm 1: QACM Algorithm

Input: Mobile users participate in crowd sensing no-     N nodes   Step 1: Random Node Deployment   Step 2: Sensing the random events by deployed       Nodes N   Step 3: QOI_ Measurement ()   Step 4: Category_Mobile_users ()   Step 5: (Assign Different categories of mobile sensing group based on data sensitivity level requirements of applications.)   for Category Mui i to 3 do     if (i = 1)       Assign Category Mu1 for high Sensitive Applications:         else if (i = 2)       Assign Category Mu1 for Medium         Sensitive range Applications:         else if (i = 3)       Assign Category Mu1 for Low range         Sensitive Applications:         end if         end if     end if   end for

4. Simulation and Performance Analysis

Performance Metrics

The following performance metrics are considered in our QACM algorithm.

Signal Reliability (SR): It is a measure of reachability signal transmission in crowd sensing between source node/s to the sink node. This metric is compared with previous works and was found to be significantly increased as shown in Figure 2. Additionally, the signal strength increased linearly from the initiation phase to completion phase that involved the available wireless infrastructure at that instance based on the requirements.

Transmission Time (TT): It is defined as the time taken to send the sensed event signal information to the sink through different intermediate sensing nodes as shown in Figure 3.

Crowd Sensing Efficiency (CSE): It is the level of crowd sensing efficiency based on different application requirements.

Table 1. Simulation parameters.

Parameter	Values
Network size/range	1000 m ∗ 1000 m
Number of nodes	300
Node distribution	Random
Node types	Involves different. level of sensing elements
Simulation time	14,000 s

shows the TT comparison values of QACM with QICM and QOSA. There were remarkable changes in TT because of the two parameters of the QACM algorithm T_ri and T_rt. There was a markedly reduced delay in transmission due to application-based transmission.

Table 2. Comparative values of SR and TT of QACM with QIMC and QOSA.

Simulation Time	SR			TT
	QIMC	QOSA	QACM	QIMC	QOSA	QACM
2000	0.41	0.51	0.59	0.89	0.91	0.76
4000	0.47	0.50	0.55	0.86	0.88	0.74
6000	0.49	0.52	0.59	0.84	0.86	0.72
8000	0.51	0.54	0.61	0.83	0.84	0.70
10,000	0.53	0.57	0.65	0.81	0.83	0.69
12,000	0.56	0.63	0.71	0.78	0.81	0.68
14,000	0.58	0.65	0.74	0.76	0.79	0.67

depicts the comparison of SR and TT of QACM with QIMC and QOSA. The comparison values are plotted in Figure 3. The simulation results indicate that QACM demonstrated greater signal reliability overall, but specifically between 6000 and 14,000 s, where its reliability was significantly higher than that of QIMC and QOSA. This is a clear indication of enhanced reliability due to the consideration of proper measures such as categorizing the mobile sensing nodes based on the parameter R_bd.

Table 2 shows the comparative values of TT of QACM with those of QICM and QOSA. Thus, there was an approximately 33% improvement in transmission delay.

Figure 4 illustrates the crowd sensing efficiency (CSE). This graph clearly indicates very significant improvement in the crowd sensing process because of very efficient parameters such as R_bd, T_ri, T_rt, and M_qs that collectively increased the efficiency 32% when comparing QIMC with QOSA and CACM.

5. Conclusions

Most of the previous works in mobile crowd sensing utilized more infrastructure-based transmission available at the respective locality or geographical limits during the sensing time. QACM exploits the opportunistic transmission by hiring the mobile users that provide greater flexibility in providing the necessary hardware resources like advanced sensors bandwidth and software support due to advanced devices with the fastest growing technologies. However, the reliability, transmission time, and crowd sensing efficiency are most important factors in improving QOI in MCS. Many previous works attempt to bring the QOI to the required level. In QACM, very effective parameters have been devised to increase the reliability, reduce the transmission time, and increase the efficiency of crowd sensing. The parameters R_bd, T_ri, T_rt, and M_qs strengthened the ability of QACM to increase efficiency in MCS. Here, categorization of sensing devices based on their sensing capacity brought great changes in sensing efficiency and bandwidth allocations based on high-, medium-, and low-range sensitivity requirement applications. This work can be extended to highly sensitive MCS applications like defense surveillance systems, medical applications, forest surveillance system, agricultural crops security and monitoring, etc.