Context-Driven and Resource-Efficient Crowdsensing

Mobile data through crowdsensing has become a prime requirement in various research fields due to its ubiquity, ease of collection, and portability. A typical crowdsensing application collects various sensor data and information from the participant's devices (e.g., smartphone), often continuously at higher frequencies. However, the limited resources available on mobile devices (e.g., energy) pose a significant challenge for many mobile crowdsensing efforts. Therefore, an efficient sensing system maintains reasonable data quality while being resource conscious. Context plays a critical role in balancing the trade-off between the quality of opportunistic sensing data and the energy of the participating device. Moreover, context information can be utilized in determining the right moment to trigger user-centric data collection. This dissertation investigates several aspects of resource efficiency and leveraging context in crowdsensing applications. First, it identifies the challenges of continuous data collection in longitudinal cohort studies and addresses some of them. Secondly, it designs a flexible crowdsensing platform where both opportunistic and participatory sensing activities can be configured in a context-driven way. In addition, it facilitates the data collector to configure the collection target and generates efficient sensing rules to achieve the goal. Finally, it studies the relationship between mobility and phone usage patterns and proposes an efficient location sensing mechanism.