LAUSR

Wireless sensor nodes equipped with multiple sensors often have limited energy availability. To optimize the energy sustainability of such sensor hubs, in this paper a novel adaptive sensor selection framework is proposed. Multiple sensors monitoring different parameters in the same environment often possess cross-correlation, which makes the system predictive. To this end, a learning-based optimization strategy is developed using Upper Confidence Bound algorithm to select an optimum active sensor set in a measurement cycle based on the cross-correlations among the parameters, energy consumed by the sensors, and the energy available at the node. Further, a Gaussian process regressor-based prediction model is used to predict the parameter values of inactive sensors from the cross-correlated parameters of active sensors. To evaluate the performance of the proposed framework in real-life applications, an air pollution monitoring sensor node consisting of seven sensors is deployed in the campus that collects data at a default high sampling rate. Simulation results validate the efficiency and efficacy of the proposed framework. Compared to the current state-of-the-art the proposed algorithm is 54% more energy efficient, with complexity $\mathcal {O}(2^{P})$ for $P$ sensors in the node, while maintaining an acceptable range of sensing error.