Centaurus : A Client-Based Scripting Framework for Mobile Crowdsensing

Smartphones can digitally interleave and merge shared observations, thoughts, and feelings with automatically sensed behaviors, environmental conditions, and patterns. This mix of shared and sensed information can be aggregated in the cloud over communities and geographies to track, study, and react to practically any phenomena in human proximity. We refer to this capability as mobile crowdsensing, and it can be applied to supply apps and services with highly relevant, real-time information, to guide business and marketing strategy, or to revolutionize research on societal issues. The chief obstacle to crowdsensing is in managing the tremendous scale and diversity of private data and application needs associated with smartphones. We present Centaurus, a clientbased scripting framework that addresses critical challenges in diversity, scalability, and privacy. Through the design, prototype, and evaluation, we show that it can express most crowdsensing tasks as scripts, achieves orders of magnitude more scalability that cloud-centric approaches, and offers a design that respects privacy and complies with regulations. We also use a novel, low-cost data collection methodology based on crowdsourcing.