UPDetector: sensing parking/unparking activities using smartphones

Real-time information about vacant parking spaces is of paramount value in urban environments. One promising approach to obtaining such information is participatory sensing, i.e. detecting parking/unparking activities using smartphones. This paper introduces and describes multiple indicators, each of which provides an inconclusive clue for a parking or an unparking activity. As a result, the paper proposes a probabilistic fusion method which combines the output from different indicators to make more reliable detections. The proposed fusion method can be applied to inferring other similar high-level human activities that involve multiple indicators which output features asynchronously, and that involve concerns about power consumption. The proposed indicators and the fusion method are implemented as an Android App called UPDetector. Via experiments, we show that our App is both effective and energy-efficient in detecting parking/unparking activities.