ENERGY-AWARE DISCRETE PROBABILISTIC LOCALIZATION OF WIRELESS SENSOR NETWORKS

Localizing sensor nodes is critical in the context of wireless sensor network applications. It has been shown that, for some applications, low-overhead discrete localization achieves results comparable to costly fine localization. This research presents a hybrid energy-aware discrete localization method that requires no transmission overhead from the sensor nodes. The proposed method, E-KalmaNN, is a combination of a Kalman-inspired localization and Artificial Neural Networks estimation that updates the position of a node with respect to a mobile reference. E-KalmaNN runs on the sensor nodes and supports different listening/wakeup frequencies for different nodes to balance power requirements with localization accuracy for each node. Simulation results show that the method converges to the correct position of the node in a relatively short time with high average location accuracy. Compared to the localization methods found in the literature, E-KalmaNN localizes with comparable accuracy, lower transmission costs and/or fewer motion restrictions.