Crowdsourced location-based sensing
Bulut, Muhammed Fatih
MetadataShow full item record
In recent years, along with market penetration of smartphones, location-based services (such as proximity-based services and nearby-search services) have been rapidly gaining popularity. In addition, social networks such as Twitter, Facebook and Foursquare, provide a platform (with their open-access APIs) to sense and share large volume of location data and provide value for implicit location-based sensing. These trends enable more fine-grained and larger-scale location-based sensing opportunities which we call crowdsourced location-based sensing. Crowdsourced location-based sensing uses the crowds as a sensor to collect raw data and interprets it by using machine learning and statistical analysis techniques in the cloud. Although the progress in recent years is encouraging, the lack of tools, energy-efficient sensing models and algorithms to analyze raw data remain as challenges. This dissertation explores two key areas to enable crowdsourced location-based sensing: energy-efficient explicit location-based sensing using smartphones and implicit location-based sensing at large using social networks. First part of the dissertation focuses on the explicit crowdsourced location-based sensing. We propose a novel crowdsourced line wait-time monitoring service for coffee shops, called LineKing. LineKing is the first automated crowdsourced line wait-time monitoring service and is a prototype application for explicit crowdsourced location-based sensing. It consists of a smartphone component that provides automatic, energy-efficient and accurate wait-time detection, and a cloud backend that uses the collected data from smartphones to provide accurate wait-time estimations. To achieve energy-efficiency in LineKing, we develop a novel proximity alert service on Android. This service enables applications that use proximity alert to detect entrance and exit from points of interest in an energy-efficient way. Experimental results indicate that, our service saves considerable amount of energy compared to the baseline implementation of proximity alert service on Android. The second part of the dissertation explores the use of social networks for implicit crowdsourced location-based sensing. This approach extends the first approach by implicitly inferring the location using social networks' data. We show how to employ passive sensing of city-level information via Twitter to detect trends and important events using a corpus that is generated from Foursquare venue categories. We also present a crowdsourced location-based question-answering system over Twitter. We demonstrate the effectiveness of employing location-based services (such as Foursquare) for finding appropriate people to answer a given location-based query.