WikiSensing: Towards a Cloud-based Sensor Informatics Platform for life in a Digital City

Digital Futures (2012) (2013)

This paper provides the methodology and system design of
WikiSensing, which is a cloud-based sensor informatics platform
for sensing applications for digital life. WikiSensing follows a
model-driven and attention-based approach: the sensory target is
presented over the hypotheses or models, a typical instance of
which is a network diffusion model with causality links;
observations are used to train the model in a Bayesian way; when
the model prediction has greater variance than the real
observation, attention is aroused, the event of interest is detected
using compressive sensing-based recovery, and the model is
updated. The platform also considers a system’s objectives and
constraints, thus computing the optimal sensor placement for a
given cost or energy budget. When these objectives and
constraints change, the model is updated, and the attention
mechanism of the sensing system is readjusted. This methodology
forms the basis of WikiSensing, and can be applied to several
sensor applications from body sensing to city traffic monitoring.