Project topics and materials
Get Complete Project Material File(s) Now! »
Energy Consumption in Processing
This energetic cost is due to the necessary power consumed by mobile phones processors (CPU) during the sensing task. In this context, one of Microsoft Research Projects, Little- Rock [32], developed a tool to reduce the continuous sensing energy overhead by using a dedicated low-power sensing processor (microcontroller) for sampling and low-level processing of sensor data. The dedicated microcontroller minimizes the energy consumed during sensors readings and can transition between sleep and active modes within a very short time. In addition to hardware level implementation, researchers [36,37] propose to refer to ooading techniques where mobile users can delegate the computation to more powerful infrastructure resources generally based on the cloud, thereby extending battery lifetime. Nonetheless, this may come with data uploading energetic cost as detailed hereafter.
Energy Consumption in Uploading
Mobile crowdsensing requires a maintained connectivity between dierent entities to receiv tasks assignment and upload collected data. The latter phase can be conducted via 3G/4G, Wi-Fi networks or short range communications such as Bluetooth, Device to Device, etc. However, the aforementioned communication techniques dier in terms of necessary energy such as detailed in Table 2.2. In order to reduce the incurred energy during uploading, Ma et al. [38] present a crowdsensing system which relies on the relatively low-power communications such as Wi-Fi or Bluetooth to upload data rather than directly using 3G/4G. Yet, this can be adopted only for delay-tolerant sensing applications in order to not decrease the temporal-accuracy of samples. Some other works propose paralleling data uploading with phone calls and show that this can save up to 75-90% of energy [39]. According to this assumption, Lane et al. [40] have proposed the Piggyback CrowdSensing (PCS) solution to upload data during phone calls or jointly with the applications of common use.
Positioning of Dissertation Contributions to the studied Literature
The study of the various MCS opportunities and challenges allows us to position our contributions in this thesis with regard to the literature. We focus, in this dissertation, on both participants and requesters encountered issues in the aim of conquering the existing research work and propose adequate solutions. Therefore, we illustrate rst which challenge(s) each of our developed methods tackles as detailed in Figure 2.3. Besides, we describe brie y our main contributions positioning in the literature.
Our rst work, QEMSS [16], studies the data quality and energy consumption issues in MCS systems. We propose to prevent high energy consumption of participants’ devices while achieving competent data quality levels by developing adequate task assignment methods. This work has been extended to consider a fair task allocation schedule among participants. F-QEMSS [17] targets to jointly maximize data quality and fairness measures while considering energy constraints. In other terms, we propose to answer both requesters’ and participants’ requirements in the same time, which is slightly introduced in literature but not fully addressed.
Table of contents :
1 Introduction
1.1 What is Mobile Crowdsensing?
1.2 Motivations and Contributions
1.3 Organization of the Thesis
2 Crowdsensing: Opportunities and Challenges
2.1 Introduction
2.2 Mobile Sensing: Scale and Paradigms
2.2.1 Mobile Sensing Scale
2.2.2 Crowdsensing Paradigms
2.2.2.1 Participatory Sensing
2.2.2.2 Opportunistic Sensing
2.3 Crowdsensing Opportunities
2.3.1 MCS Applications in Academia
2.3.1.1 Environmental Applications
2.3.1.2 Infrastructure Applications
2.3.1.3 Social Applications
2.3.2 MCS Applications in Industry
2.3.3 MCS Frameworks
2.4 Crowdsensing Challenges: Participants Concerns
2.4.1 Energy Consumption
2.4.1.1 Energy Consumption in Sensing
2.4.1.2 Energy Consumption in Processing
2.4.1.3 Energy Consumption in Uploading
2.4.2 Mobile Data Cost
2.4.3 Incentives
2.4.4 Privacy
2.5 Crowdsensing Challenges: Requesters Concerns
2.5.1 Quality of Information (QoI)
2.5.2 Budget
2.6 Positioning of Dissertation Contributions to the studied Literature
2.7 Conclusion
3 Quality and Energy-aware Task Assignment Schemes
3.1 Introduction
3.2 Related Works
3.2.1 Energy-aware Task Assignment in MCS
3.2.2 Quality-aware Task Assignment in MCS
3.2.3 Fairness in MCS
3.3 Background
3.3.1 Proposed Utility Function
3.3.2 Tabu Search
3.4 Preliminaries: Proposed Measures
3.4.1 Quality of Information Attributes and Metrics
3.4.1.1 Completeness
3.4.1.2 Timeliness
3.4.1.3 Energy Metric: Aordability
3.4.2 Fairness Metric: Jain Index
3.5 Quality and Energy-aware Problem Denition
3.5.1 Crowdsensing System Overview
3.5.2 Known/Unknown Users Trajectories
3.5.3 Problem Formulation
3.5.3.1 QoI Maximization Problem
3.5.3.2 QoI and Fairness Maximization Problem
3.6 Mobile Sensing Schemes: QEMSS Vs F-QEMSS
3.6.1 Elements of TS-based Schemes
3.6.1.1 Solution X 2
3.6.1.2 Move m
3.6.1.3 Tabu List (TL)
3.6.2 Phases of TS-based Schemes
3.6.2.1 Initialization
3.6.2.2 Neighborhood Formation
3.6.2.3 Neighborhood Selection
3.6.2.4 Update Tabu List
3.7 Performance Evaluation
3.7.1 Simulation Settings
3.7.2 Benchmark
3.7.3 Evaluation Metrics
3.7.4 Evaluation Results
3.7.4.1 Maximum Achieved QoI
3.7.4.2 Spatial and Temporal Accuracy
3.7.4.3 Fairness Metrics
3.8 Discussion
4 Preference and Mobility-aware Task Assignment Schemes
4.1 Introduction
4.2 Motivations and Context
4.3 Existing Work on Distributed Crowdsensing
4.4 Distributed Task Assignment Problem Denition
4.4.1 Semi-Distributed Crowdsensing System Overview
4.4.2 Users’ Arrival Model
4.4.3 Users’ Preferences Model
4.4.4 Problem Denition
4.5 Mobility-aware Task Assignment: MATA
4.5.1 Oine Mode
4.5.1.1 Oine Average Makespan
4.5.1.2 Oine Algorithm: MATAF
4.5.2 Online Mode
4.5.2.1 Online Average Makespan
4.5.2.2 Online Algorithm: MATAN
4.5.3 After Thoughts
4.6 Preference and Mobility-aware Task Assignment: P-MATA
4.6.1 Oine Mode
4.6.1.1 Oine Average Makespan
4.6.1.2 Oine Algorithm: P-MATAF
4.6.2 Online Solution: P-MATAN
4.6.3 After Thoughts
4.7 Performance Evaluation
4.7.1 Real Traces
4.7.2 Requester Selection
4.7.3 Simulation Settings
4.7.4 Evaluation Results
4.7.4.1 Average Number of Assigned Tasks
4.7.4.2 Average Achieved Makespan
4.7.4.3 The Number of Lost Tasks
4.8 Discussion
5 Extended Preference-aware Task Assignment with Incentive Mechanisms
5.1 Introduction
5.2 Motivations and Context
5.3 Preference-aware and Incentivizing MCS
5.3.1 Preference-aware Crowdsensing
5.3.2 Incentivizing Mechanisms in MCS
5.4 Problem Statement
5.4.1 Preliminaries
5.4.1.1 System Overview
5.4.1.2 Users Probabilistic Arrival
5.4.1.3 Discrete Choice Model
5.4.2 Problem Formulation
5.4.2.1 No-incentives-based Assignment
5.4.2.2 Incentives-based Assignment
5.5 Extended Preference-aware Task Assignment: P-MATA+
5.5.1 Oine Mode: P-MATAF+
5.5.2 Online Mode: P-MATAN+
5.5.3 After Thoughts
5.6 Incentives-based Preference-aware Task Assignment: IP-MATA+
5.6.1 Incentive Policies
5.6.1.1 Priority-based Incentives
5.6.1.2 Quality-based Incentives
5.6.2 Oine Mode: IP-MATAF+
5.6.3 Online Mode: IP-MATAN+
5.6.4 After Thoughts
5.7 Performance Evaluation
5.7.1 Simulation Settings
5.7.2 Performance Analysis
5.7.2.1 Average Number of Assigned Tasks
5.7.2.2 Average Makespan
5.7.2.3 Incentives Policies Performance
5.8 Discussion
6 Privacy Preserving Utility-aware Mechanism in Data Uploading
6.1 Introduction
6.2 Motivations and Context
6.3 Research Work on Privacy-preserving Mechanisms
6.3.1 Privacy-preserving Mechanisms in Participatory Sensing
6.3.1.1 Cryptography
6.3.1.2 Anonymization Techniques
6.3.1.3 Data Aggregation
6.3.1.4 Obfuscation/ Perturbation
6.3.2 Privacy-utility Trade-o
6.4 System Model
6.4.1 Participatory Sensing Involved Entities
6.4.1.1 Participants
6.4.1.2 Queriers/Requesters
6.4.1.3 Server/Broker
6.4.2 Adversary Model
6.4.3 Scenarios of Data Reporting Model
6.5 Problem Formulation
6.5.1 Privacy Leakage Metric
6.5.2 Data Utility Metric
6.5.3 Optimization Problem
6.5.3.1 Adversary without Side Information
6.5.3.2 Adversary with Side Information
6.6 Proposed Solution: PRUM
6.6.1 Proposed Algorithm
6.6.2 Mapping of Distorted Data
6.6.3 Data with Large Size Alphabets
6.7 Performance Evaluation
6.7.1 Sensing Datasets
6.7.1.1 Occupancy Detection Data
6.7.1.2 GPS Trajectory Data
6.7.1.3 Crowd Temperature Data
6.7.2 Simulations Settings
6.7.3 Privacy-Utility Trade-o
6.7.3.1 Smart-house Monitoring Scenario
6.7.3.2 Trac Rating Application
6.7.3.3 Crowd-Temperature Application
6.7.4 Side Information Impact on Privacy Leakage
6.7.4.1 Smart-house Monitoring Scenario
6.7.4.2 Trac Rating Application
6.8 Discussion
7 Conclusion and Perspectives
7.1 Contributions
7.2 Perspectives
Bibliography
