Surrogate mobile sensing

The proliferation of smart phones with sensing capabilities motivates exploring the applicability limits of (phone-based) mobile sensing. While a phone can directly measure variables such as location, acceleration, and orientation, other interesting quantities one may want to measure have higher-level semantics that a phone does not directly recognize. For example, one might want to map parking lots that are free after hours, or restaurants that are popular after midnight. How can we measure such higher-level logical quantities using sensors on phones? Techniques that address this question fall in the broad area of surrogate sensing, defined as inferring high-level logical quantities by measuring weaker surrogates. The surrogates in question are variables that can be sensed using a phone, but are only weakly related to the original high-level logical quantities one is really after. The key challenge is to exploit appropriate aggregation techniques that leverage the availability of large numbers of phones to overcome the poor quality of individual surrogates. Recently, significant advances have been made in understanding the quality limits of surrogate sensing. This article overviews the main ideas and insights underlying these advances.