Resource-Aware Crowdsourcing in Wireless Networks

Restricted (Penn State Only)
Wu, Yibo
Graduate Program:
Computer Science and Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
January 26, 2018
Committee Members:
  • Guohong Cao, Dissertation Advisor
  • Guohong Cao, Committee Chair
  • Robert Collins, Committee Member
  • Sencun Zhu, Committee Member
  • Zhenhui Li, Outside Member
  • Crowdsourcing
  • Mobile Devices
  • Photo
  • Video
  • Wireless Networks
The ubiquity of mobile devices has opened up opportunities for a wide range of applications based on photo/video crowdsourcing, where the server collects a large number of photos/videos from the public to obtain desired information. However, transmitting large numbers of photos/videos in a wireless environment with bandwidth constraints is challenging, and it is hard to run computation-intensive image processing techniques on mobile devices with limited energy and computation power to identify the useful photos/videos and remove redundancy. To address these challenges, we propose a framework to quantify the quality of crowdsourced photos/videos based on the accessible geographical and geometrical information (called metadata) including the orientation, position, and all other related parameters of the built-in camera. From metadata, we can infer where and how the photo/video is taken, and then only transmit the most useful photos/videos. The goal of this dissertation is to design and evaluate metadata-based techniques to support resource-aware crowdsourcing in wireless networks. First, we propose a metadata-based framework called SmartPhoto which can select the most useful photos to cover the specified points of interest. Optimization problems regarding the tradeoffs between photo coverage and resource utilization are formalized and efficient algorithms with provable performance bounds are designed and evaluated. Furthermore, SmartPhoto has been evaluated through real experiments with Android smartphones, and various techniques have been designed to improve the accuracy of the collected metadata. Second, for some applications, the target of interest is an area rather than a point. Thus, we extend SmartPhoto to select the most useful photos to cover an area of interest. Since there are infinite number of points in an area and each point can be covered differently, area coverage is hard to analyze. To address this problem, we study how to select photos with the best area coverage under resource constraints and propose an efficient algorithm with provable performance bound. Third, for applications such as disaster recovery and battlefield, the cellular network may be partly damaged or severely overloaded, and thus photos have to be delivered through Disruption Tolerant Networks (DTNs). In DTNs, data can only be transferred when two devices are within the wireless transmission range of each other, and the storage of each device is also limited. Therefore, it is important to prioritize more valuable photos to use the limited resources. We quantify the value of the photos based on their metadata, and propose a distributed photo selection algorithm to maximize the value of the photos delivered to the server considering bandwidth and storage constraints. Finally, we propose VideoMec, a crowdsourcing system which organizes the metadata of all videos taken by mobile devices. VideoMec supports comprehensive queries for any application to find and fetch desired videos from the corresponding mobile devices. For time-critical applications, it may not be possible to upload all desired videos in time due to limited wireless bandwidth and large video files. Thus, we propose efficient algorithms to select the most important videos or video segments to upload under bandwidth constraints. VideoMec has been implemented and experimental results have demonstrated its effectiveness on organizing and retrieving mobile videos from mobile devices with resource constraints.