LAUSR

Owing to the limited storage capacity, battery-powered wireless sensor nodes often suffer from energy sustainability. To optimize the energy consumption of a multi-parameter sensor hub, a novel edge intelligence-based data-driven priority-aware sensing and transmission framework is proposed in this paper. The proposed framework jointly exploits the cross-correlation among the sensing parameters and temporal correlation of the individual sensing signals to find an optimal active sensor set and optimal sampling instants of the sensors in the next measurement cycle. The length of measurement cycle is dynamically decided based on the change in cross-correlation among the parameters and the system state. A discounted upper confidence bound algorithm-based optimization function is formulated to find the optimal active sensor set by solving the trade-off among cross-correlation, energy consumption, and length of measurement cycle. The proposed framework uses Gaussian process regressor-based prediction models to estimate the temporal and cross-correlated parameters of the active and inactive sensor set, respectively. The sampling interval of each active sensor is dynamically adapted based on the temporal prediction error. Extensive simulations are performed on air pollution monitoring dataset to validate the efficacy of the proposed framework in both real-time and non-real-time applications. The proposed algorithm saves up to 41% energy and 32% bandwidth with 68% data accuracy compared to the existing competitive frameworks for non-real-time systems. The proposed framework also identifies the time-critical sensing scenarios with 98% accuracy.