Elsevier

Computer Networks

Volume 142, 4 September 2018, Pages 179-193
Computer Networks

Real-time and energy aware opportunistic mobile crowdsensing framework based on people’s connectivity habits

https://doi.org/10.1016/j.comnet.2018.06.013Get rights and content

Abstract

The trade-off between energy-efficiency and real-time data delivery is seldom considered by the earlier research in mobile crowdsensing paradigm. This paper presents REOPSEK framework designed to satisfy this newly defined compromise while ensuring the required coverage quality. REOPSEK is based on the piggyback approach. In particular, it relies on users’ connectivity sessions, named “Online Episode” (OE), to jointly perform sensing and uploading tasks. To differentiate between these presented opportunities, REOPSEK associates two new parameters to an OE. These parameters serve as condition attributes to determine the availability of a smartphone for immediate detection and upload tasks. Then, based on already experienced OEs, the framework builds a lightweight prediction model to drive tasks allocation process based on an improved Simulated Annealing (SA) metaheuristic method. Simulations on real connectivity contextual data collected from 100 users in Sfax, Tunisia, demonstrate the efficiency of REOPSEK in terms of energy saving, data timeliness and coverage quality.

Introduction

Fuelled by the widespread adoption of sensor-enabled smartphones, mobile crowdsensing came as the next generation of wireless sensor network for the monitoring of large scale phenomena. As millions of smartphones are already deployed in the field and people carry them wherever and whenever they go, by leveraging people mobility and powerful smartphones’ sensors (GPS, microphone, air quality, temperature, humidity), data could be provided to stakeholders without any additional deployments [1]. Local knowledge of communities and special environmental surrounding can be shared, to be aggregated later in a server for large-scale data mining [2], such as transportation (VTrack [4]), noise (Ear-phone [5]), pollution levels (Common Sense [6]) and more.

In particular, human mobility patterns and everyday actions have increased the possibility to participate opportunistically (passively) with a large number of devices. Contrary to participatory (active) sensing, where people are highly engaged in manual detection tasks, passive sensing is fully unconscious. The dedicated systems or applications decide who meets the requirements, and data are collected automatically without the smartphone usage being interrupted [3].

Several studies in opportunistic mobile crowdsensing systems were focused on different issues such as ensuring coverage quality under budget constraint [7], [8], [12], [13]. As energy is consumed in all opportunistic mobile crowdsensing steps, starting from sensing, up to processing and data transmitting, recent years have witnessed an increasing interest in improving users’ experience from an energy-saving perspective [14], [15], [16], [17].

However, while mobile users care more about their consumed energy and their daily usage interruption, urban applications are rather concerned with the coverage quality, especially, data timeliness. In fact, receiving data in real-time is of a great importance in dealing with emergency situations. Instantaneous flows of sensory data are required in many scenarios such as informing drivers of the current traffic situation or helping city planners know the current people density for an accurate assessment of disasters impacts [18]. Early fire detection is also possible through real-time air quality samples, which is typically important for damages minimization. Similarly, asthmatic people would monitor current air pollution before their moves. On the whole, historical data and real-time sensor readings could be merged to produce accurate phenomena observation.

Given the mobile users’ energy-saving constraint, on the one hand, and urban applications’ requirements on coverage quality and data timeliness, on the other hand, one of the most challenging feature that is unfortunately seldom considered by the research community is how to ensure an energy saving, coverage quality and data timeliness triple trade-off.

In fact, from one side, different approaches have been accomplished to reduce the energy placed for sensing, uploading and data processing. In particular, obviating redundant sensing activities and reducing sensing frequency have gained a raising interest during the last years [10], [11], [14], [16], [19]. The key solution of the other proposed systems is to piggyback sensing tasks with smartphones’ usage times, such as playing games or writing a text message [14]. Then data are uploaded later to the server (with the next network access [11], [12]) or transmitted over short distance wireless technologies such as Bluetooth until they encounter an already WiFi accessed node that performs their upload [20].

As they enable a delay tolerant data upload, these proposed strategies are efficient in reducing the burden placed on smartphones, but to the detriment of real-time data acquisition (1). Therefore, they could not provide a good support for emergency situations where real-time data acquisition is imperative.

From the other side, some approaches have managed to ensure a real-time data upload. For that, they propose publish/subscribe systems to schedule subscriptions over available publishers that are periodically controlled through exchanged messages [16], [19], [21]. Otherwise, they rely on location triggered applications using excessively the expensive GPS sensor [22]. These proposed systems have emphasized the real-time data acquisition, but to the detriment of the users’ resources energy (2).

Given (1) and (2) observations, we propose in this paper to tackle a new direction of ensuring energy, real-time and coverage quality triple trade-off. Towards meeting this challenge, we introduce a practical framework, the Real-time Energy aware OPportunistic mobile CrowdSEnsing frameworK (REOPSEK), a smartphone and server-based system for real-time crowdsensing of mobile sensor data, that is designed to smartly keep the burden placed on volunteers as low as possible. We mean by volunteers, mobile users that contribute to REOPSEK framework and that are in the same time, data producers and data consumers.

REOPSEK is designed to intelligently leverage the opportunities for collecting and uploading sensor data in real-time. Inspired by works that piggyback sensing tasks jointly with smartphones’ applications, those opportunities occur frequently during everyday connectivity sessions, named “Online Episode” and denoted as (OE). We refer to these situations as smartphone’s “Online Episode opportunities”. These opportunities do not only ensure the real-time data delivery, but also increase the possibility to sense in an energy efficient way. In fact, sensor-based applications (Google Map (GPS), Skype or Viber (Camera, Microphone), Facebook (Location)) have a great chance to be run during OE, which saves a lot of battery as the intentionally-used application itself uses the required sensor. Moreover, it was proved that users’ connectivity follows some behavioural aspects [35], [43], unlike the widely used phone calls opportunities.

As exploiting each smartphone’s connectivity session to sense and upload immediately data would expose volunteers to significant drain on their smartphones’ resources, REOPSEK’s operations are guided by a predictive model that captures the smartphone’s connectivity habits which are specific to each volunteer. Then, by predicting upcoming connectivity sessions, i.e, “Online Episode”, REOPSEK can compare opportunities to sense and upload in real-time and in an energy-efficient way. This prediction model drives tasks allocation process that can balance coverage rate between all sub-regions and obviate redundant sensing activities to reduce the overall consumed energy.

This proposal is consolidated with the design and implementation of two main parts: REOPSEK-M/REOPSEK-C. The REOPSEK-M, is a mobile application designed so as to play a double role. The first is devoted for the historical data gathering phase and the second is for real-time phase. In the first phase, it logs the surrounding context of the smartphone whenever it is Internet accessed, specifically, its location, time, battery level and the current volunteer’s activity. Those data are processed later by the core framework REOPSEK-C, a centralized server that builds a prediction model and allocates sensing tasks jointly with “Online Episode opportunities”. We developed its core algorithms to maximize the sensing revenue while preserving the smartphones’ consumed energy. Specifically, our algorithms (1) accurately identify regular “Online Episode”, (2) predict upcoming “Online Episode”, which in turn enables (3) an intelligent tasks’ allocation based on the well-known metaheuristic, the Simulated Annealing (SA) algorithm [23].

To validate our approach, we deploy the mobile application REOPSEK-M on 100 volunteers in Sfax, Tunisia for the month of January, 2017. Simulations on those real data prove firstly our proposal of regularity in terms of connectivity sessions. Then, results prove, on the one hand, that our proposed algorithms ensure coverage quality and equity under energy and real-time data delivery constraints. On the other hand, they show the benefits gained by piggybacking sensing and uploading tasks jointly with “Online Episodes” in terms of consumed energy and data timeliness.

The remainder of this paper is organized as follows: Section 2 introduces the related work. We present in Section 3 the concept of “Online Episode opportunities” and demonstrate the benefit gained by exploiting these times. A general overview of REOPSEK architecture is given in Section 4. Then, two main steps of its operations are detailed respectively as follows: The first deals with the “Online Episode opportunity” prediction detailed in Section 5, the second is presented in Section 6 and is devoted to the tasks allocation process based on already defined “Online Episodes”. To highlight the effectiveness of our approach, we present in Section 7 evaluation results obtained through simulation based on real data. Finally, we draw conclusion and future work in Section 8.

Section snippets

Related work

Significant research studies have been devoted to the provision of sensory data based on the mobile crowdsensing. Considering the budget [12], [13] and/or the energy saving [14], [20], [24] constraints, the broad objective of ensuring a high coverage quality is common to them.

Unlike previous work on mobile crowdsensing, our paper contributes to the existing literature at a new energy/real-time/coverage quality triple trade-off. As it is a new direction, as we know, we provide separately an

“Online Episode” opportunities

In opportunistic mobile crowdsensing, defining the opportunity to exploit for meeting applications’ requirements is a basic feature for the designer to decide later what must be deployed and implemented.

Applications might aim for different criteria such as high task coverage, data credibility and quality, or low cost. With the design objective of ensuring real-time data delivery, coverage quality and energy efficiency, we present in this section our motivation to exploit people’s connectivity

Overview of REOPSEK framework

The overall architecture of the proposed system is depicted in Fig. 2 and consists of three main parts. REOPSEK-Mobile enabled smartphones, REOPSEK-Core framework and urban applications side.

REOPSEK-Mobile (REOPSEK-M): Is a mobile application residing in volunteers’ smartphones side. The primary target of REOPSEK-M is to extract contextual information specific to volunteers’ connectivity sessions (location, time, battery level and current activity). This information will serve later as a

“Online Episode opportunity” prediction

The prediction of different “Online Episodes opportunities” (Oi) to succeed a sensing task with specific characteristics, is quite important to differentiate and rank providers that can offer similar detected data. It is based on the “Online Episode” probability of happening and adequacy with respect to the monitored cell gk, m.

Thus, to deduce its probability of happening, we were inclined to study people connectivity habits in terms of periods of network access, frequent places of connectivity

Offline tasks allocation

With the observations and research goals elaborated in both the introduction and the related work, the essence of the task allocation problem in this work is to maximize coverage rate and enable coverage equity between regions while satisfying volunteers’ energy saving restrictions. The real-time data delivery is intrinsically ensured as sensing tasks are allocated jointly with “Online Episodes”. The output of allocation process is a set of OE selected for each cell (sub-region gk and sampling

Experimental study

In this section, we first present the experimental settings, including the dataset and the experiment environment. Then, we report different measured parameters in our experiments and discuss our findings.

Dataset: We conduct our experiments on real dataset collected by REOPSEK-M from 100 volunteers in Sfax, Tunisia. Extracted contextual data during connectivity sessions are time, location coordinates, battery level and activity. To meet their privacy requirements, anonymity of volunteers was

Conclusion

This paper presents REOPSEK, a framework for crowdsourcing mobile sensor data. It is designed to intelligently exploit the “Online Episodes opportunities” to jointly piggyback sensing and uploading tasks. The proposed system aims to ensure a triple compromise between energy saving constraint from volunteers’ side, coverage rate and data timeliness from urban applications’ side. We performed a comprehensive set of experiments to evaluate our proposal based on real data of connectivity context of

Salma Bradai is a Ph.D. student at the National School of Engineering of Sfax, Tunisia, and a researcher in distributed applications at the ReDCAD Laboratory at the University of Sfax. Salma obtained her Engineering degree in Computer Science from National School of Engineering of Sfax (ENIS) in 2011 and M.S. degree in Computer Science in 2012 from ENIS. Currently, her research interests are concentrated in Real time and energy aware mobile crowdsensing.

References (45)

  • P. Dutta et al.

    Common sense: participatory urban sensing using a network of handheld air quality monitors

    Proceedings of the Seventh ACM Conference on Embedded Networked Sensor Systems, SenSys 09

    (2009)
  • S. Reddy et al.

    Recruitment framework for participatory sensing data collections

    Proceedings of the Eighth International Conference on Pervasive Computing, Pervasive 10

    (2010)
  • YangD. et al.

    Crowdsourcing to smartphones: incentive mechanism design for mobile phone sensing

    Proceedings of the Eighteenth Annual International Conference on Mobile Computing and Networking, Mobicom 12

    (2012)
  • P. Bellavista et al.

    Scalable and cost-effective assignment of mobile crowdsensing tasks based on profiling trends and prediction: the participact living lab experience

    Sensors(Basel)

    (2015)
  • ZhaoD. et al.

    Energy-efficient opportunistic coverage for people-centric urban sensing

    Wireless Netw.

    (2014)
  • XiongH. et al.

    Crowdtasker: maximizing coverage quality in piggyback crowdsensing under budget constraint

    Proceedings of the IEEE International Conference on Pervasive Computing and Communications (PerCom)

    (2015)
  • HanK. et al.

    Bliss: Budget Limited Robust Crowdsensing through Online Learning

    Proceedings of the Eleventh Annual IEEE International Conference on Sensing, Communication, and Networking (SECON)

    (2014)
  • N.D. Lane et al.

    Piggyback Crowdsensing (Pcs): energy efficient crowdsourcing of mobile sensor data by exploiting smartphone app opportunities

    Proceedings of the Eleventhth ACM Conference on Embedded Networked Sensor Systems, SenSys 13

    (2013)
  • P. Foremski et al.

    Energy-efficient crowdsensing of human mobility and signal levels in cellular networks

    Sensors’15 (9)

    (2015)
  • M. Marjanović et al.

    Energy-aware and quality-driven sensor management for green mobile crowd sensing

    J. Netw. Comput. Appl.

    (2016)
  • XiongH. et al.

    Eemc: enabling energy-efficient mobile crowdsensing with anonymous participants

    ACM Trans. Intell. Syst. Technol.

    (2015)
  • A. Antonić et al.

    A mobile crowd sensing ecosystem enabled by cupus

    Future Gener. Comput. Syst.

    (2016)
  • Cited by (5)

    • A collaboration-driven mechanism for AI diagnose with multiple requesters under incomplete information

      2023, Computer Networks
      Citation Excerpt :

      Zhan et al. [36] came up with an incentive mechanism with regard to mobile users’ resource demand uncertainties. Bradai et al. [37] presents REOPSEK framework designed to satisfy, the trade-off between energy-efficiency and real-time data delivery, this newly defined compromise while ensuring the required coverage quality. Peng et al. [38] build a data diffusion and transmission paradigm in crowdsensing based on City Public Traffic System (PTS), and thoroughly discuss a paradigm for Multi-Task diffusion and transmission within budget constraints.

    • Industry 4.0 significance and its applications

      2021, Proceedings of the International Conference on Industrial Engineering and Operations Management

    Salma Bradai is a Ph.D. student at the National School of Engineering of Sfax, Tunisia, and a researcher in distributed applications at the ReDCAD Laboratory at the University of Sfax. Salma obtained her Engineering degree in Computer Science from National School of Engineering of Sfax (ENIS) in 2011 and M.S. degree in Computer Science in 2012 from ENIS. Currently, her research interests are concentrated in Real time and energy aware mobile crowdsensing.

    Sofien Khemakhem is an assistant Professor at the National School of Computer Sciences, Tunisia (ENIS). Sofien received his Engineering degree in Computer science from National School of Computer Sciences, Tunisia (ENSI), in 1998 and M.S. degree in Computer Science from National School of Engineers of Sfax (ENIS) in 2001. He obtained his Ph.D. from university of Paul Sabatier(Frensh) and University of Sfax (Tunisia) in 2011. His research interests are in discovery and composition of software components. Sofien is member of the program committees of the Integrated Intelligent Computing Conference.

    Dr. Mohamed Jmaiel is a full Professor of Computer Science (since 2009) at the National School of Engineers of Sfax, Tunisia. He joined National School of Engineers of Sfax as Assistant Professor of Computer Science in 1995. Mohamed obtained his Diploma of Engineer in Computer Science from Kiel University, Germany, in 1992 and Ph.D. from the Technical University of Berlin in 1996. He participated in the initiation of many graduate courses at the University of Sfax, Tunisia. Mohamed's current research interests include software engineering of distributed systems, formal methods in model-driven architecture, component oriented development, self-adaptive and pervasive systems, and autonomic middleware. He has published over 200 regular and invited papers in international conferences and journals and has co-edited ten journal special issues on these subjects.

    Fully documented templates are available in the elsarticle package on CTAN.

    View full text