The case for platform-integrated sensors and people-centric sensing

We investigate the feasibility of leveraging platform-integrated ambient sensors (PISs), which are sensors being built into personal computing devices (hosts), to enhance people-centric sensing. PISs can utilize resources (e.g., energy, computation, and communication) from their hosts (e.g., cellphones and laptops). Hence, the sensors have many advantages including low manufacturing cost, ease of deployment, and battery-free operation compared to standalone sensors (e.g., Telosb [1]). PISs, however, have a constraint that they must be attached to their hosts, which may not be their sensing targets (e.g., users). Therefore, it is unclear if PISs are actually useful for sensing applications. We conducted an empirical study of platform-integrated ambient sensor data collected from various locations on a wide range of personal computing devices and locations surrounding a user, who is using the devices. We found that (i) ambient data collected from PISs are close and highly correlated to ambient conditions at the user (ii) sensor placement, host operation mode (e.g., active or power-saving), host computational load, host mobility, and user activities impact PIS readings, (iii) simple models can be developed to estimate ambient conditions at the user from PISs data accurately, (iv) the estimated ambient conditions can be useful for sensing applications such as building management system, and (v) for laptops, PISs and sensing services have negligible impacts on the host and experience of the user.