Low-Power Sensor Polling for Context-Aware Services on Smartphones

The growing availability of sensors integrated in smartphones provides much more opportunities for context-aware services, such as location-based and profile-based applications. Power consumption of the sensors contributes a significant part of the overall power consumption on current smartphones. Furthermore, smartphone sensors have to be activated in a stable period to match to the request frequency of those context-aware applications, known as full polling-based detection, which wastes a large amount of energy in unnecessary detecting. In this paper, we propose a low-power sensor polling for context-aware applications, which can dynamically shrink the extra sensor activities so that the unrelated sensors can keep in sleeping status for a longer time. We also provide an algorithm to find the relationship of application invoking and those sensor activities, which is always hidden in the context middleware. With this method, the polling scheduler is able to calculate and match the detecting frequency of various application combinations aroused by user. We evaluate this framework with different kinds of context-aware applications. The results show that our new low-power polling spends a tiny responding delay (97ms) in the middleware to save 70% sensor energy consumption, comparing with the traditional exhausting polling operation.