ENERGY EFFICIENT FRAMEWORKS FOR PARTICIPATORY URBAN SENSING
Abstract
Participatory sensing is a powerful paradigm in which users participate in the sensing campaign by collecting and crowdsourcing fine-grained information and opinions about events of interest (such as weather or environment monitoring, traffic conditions or accidents, crime scenes, emergency response, healthcare and wellness management), thus leading to actionable inferences and decisions. Because of the high density of smartphone users in urban population, participatory sensing paradigm can be effectively applied to continuous monitoring of various phenomena in urban scenarios (e.g., fine-grained temperature monitoring, noise or air pollution), leading to what is called urban sensing—the subject of study in this dissertation. However, for creating a fine-grained and real-time map of the monitored area, the data samples need to be collected continuously (at a high frequency) which poses several research challenges. First, how to ensure coverage of the collected data that reflects how well the targeted area is monitored? Second, how to localize the smartphones since continuous usage of the location sensor (e.g., GPS) can drain the battery in few hours? Third, how to provide energy efficiency in the data collection process by collecting minimum number of data samples in each data collection round? Finally, how to store and backup the huge amount of collected data resulting from continuous monitoring? In this dissertation, we first propose a novel framework called PLUS to address three major issues in real-time participatory urban monitoring applications, namely, ensuring coverage of the collected data, localization of the participating smartphones, and overall energy efficiency of the data collection process. Specifically the PLUS framework can guarantee a specified requirement of partial data coverage of the monitored area in an energy efficient manner. Additionally we devised a Markov-Predictor based energy efficient outdoor localization scheme for the mobile devices to participate in the data collection process. Simulation studies and real life experiments exhibit that PLUS can significantly reduce energy consumption of the mobile devices for urban monitoring applications as compared to traditional approaches. We extend the idea of PLUS and further propose another framework called STREET that can ensure k-coverage of the collected data from an urban street network. By simulating an urban monitoring application on a street network, we demonstrate that STREET can achieve k-coverage of the collected data while consuming significantly less amount of energy especially in busy urban area. Next, we propose PeerVault - a reliable online storage and backup service for the collected data based on a peer to peer (P2P) architecture. PeerVault is built on a graph theoretic approach to exploit long term online availability and unused resources of computing devices, and a distributed monitoring algorithm to form an online backup service. Experimental results on real traces confirm that PeerVault can be served as a cheap alternative for online data backup service with high availability and long term reliability.