A Group-based Recruitment System for Mobile Crowd Sensing

Student thesis: Master's Thesis

Abstract

Mobile Crowd Sensing (MCS) is one of the most recent sensing paradigms that gained its popularity due to the advances in mobile computing. High number of mobile users carrying powerful smart devices and willing to provide the public with updates about various community dynamics and areas of interest (AoI) have contributed immensely to boost the signignicance of this new sensing paradigm. However, there are some challenges that hinder its ongoing advancements, one of them being the participants' recruitment. Ideally, the process of recruiting participants for a given sensing task should determine the optimal set of participants to answer the sensing requests while satisfying a certain level of quality of information (QoI). This selection is in fact sensitive to the type of the sensing task (one time sensing vs. continuous sensing) and to its requirement with regards to the AoI coverage, distribution and budget, to cite a few. To address this challenge, we advocate the use of group-based recruitment strate- gies rather than the individual recruitment strategies proposed in the literature. To this effect, we propose two novel group-based recruitment systems (GRS) for one-time sensing and continuous sensing tasks. Both systems employ genetic algorithms to select groups of participants based on three types of parameters: AoI-related , device-related, and user-related. The GRS proposed for one-time sensing tasks aims at recruiting a group of participants that is capable of achieving the highest QoI with a certain level of guarantee on the confidence level, which re ects participants' commitment to com- pleting the sensing task successfully. Furthermore, a new evaluation metric composed of the QoI and the confidence level is suggested. The GRS proposed for continuous sensing tasks selects the most stable group of participants in the AoI that is capable of achieving a certain level of QoI, where stability re ects the group's temporal and spatial availability. The process of recruitment is i dynamic since it involves adding and removing participants throughout the sensing period. Participants are added to the group if they contribute positively to QoI and are removed otherwise. Simulations are conducted using real-life datasets. Both systems are compared to the corresponding individual-based recruitment systems (IRSs) which employ greedy algorithms to recruit participants. The results show that the proposed one-time sensing and continuous sensing GRSs have proven to outperform IRSs for all key performance metrics, such as the QoI, amount of resources, and costs.
Date of AwardJun 2016
Original languageAmerican English
SupervisorRabeb Mizouni (Supervisor)

Keywords

  • Mobile Crowd System
  • Recruitment
  • Incentives
  • Genetic Algorithm.

Cite this

'