Proliferation of mobile smartphones has opened up possibilities of leveraging the people and devices in a crowd, (i.e., crowd-sourcing) to gather data from and monitor large crowds. However, current solutions either put unpredictable stress on the wireless or cellular infrastructure to a cloud and on energy-constrained smartphones or do not accurately capture crowd behavior. In response, our CrowdWatch project will investigate monitoring crowds from the ?inside-out? via a scalable, distributed and energy-efficient in-network crowd-sourcing framework. Local energy-efficient coordination and processing will enable the off-loading of some of the processing to the devices by establising a hierarchy of participants? multi-radio devices (i.e., Wi-Fi and Bluetooth). Through probabilistic monitoring of a crowd, CrowdWatch will reduce the demand on the bandwidth to the cloud and enhance traditional crowd-sourcing by enabling information to be delivered back to and among people within the crowd.
The main challenges of this project are (1) deployment and validation of the hierarchical crowd-sourcing system architecture, (2) design and validation of adaptive sampling algorithms to enable distributed sensing and control using Bluetooth and Wi-Fi sensing, (3) role selection algorithms based on resource availability (i.e., energy, bandwidth). Validation of CrowdWatch will entail experimentation and measurements of performance metrics such as resource usage, crowd density and user mobility, effectiveness of information distribution and monitoring of interaction frequency among users during two crowed events on our campus, the Engineering Open House and a University Basketball Game.
