A Data-driven Scheduling Strategy for Mobile Air Quality Monitoring Devices <Abstract> Along with the process of urbanization and mechanization, environmental pollution is becoming more and more serious all over the world. To this end, numerous efforts are directed toward evolving effective monitoring operations, including both stationary and mobile systems. Compared to traditional fixed monitoring stations, mobile air quality monitoring systems offer a more flexible and cost-effective way to achieve a fine- grained air quality map. However, as mobile air quality monitoring systems typically rely on lightweight devices with low-capacity batteries, conserving the energy of these devices becomes a significant challenge. A trivial method to reduce devicesf energy is to set a low- frequency sampling rate and allow the device to go to sleep during the idle period. Nonetheless, this method may diminish the temporal granularity of the gathered data. To this end, in this paper, we propose a deep learningbased approach that adaptively regulates the activities of devices to simultaneously accomplish two goals: energy conservation and data quality assurance. The primary idea is to allow devices to go to sleep when the fluctuations of air quality indicators are minimal and to use a predictive model to forecast the air quality during the idle period. We evaluate our proposal on Fi-Mi, an actual bus-based air monitoring system in Hanoi, Vietnam. Experiment results indicate that our proposed method saves approximately 42% of the devicefs energy consumption with an air monitoring error of only 3%. |