Effective neural network‐based node localisation scheme for wireless sensor networks

A new neural network-based node localisation scheme is introduced that can make the trained network model completely relevant to the topology via online training and correlated topology-trained data and therefore attain more efficient application of the neural networks and more accurate inter-node distance estimation.

Abstract

Wireless sensor networks usually obtain the location of an unknown node by measuring the distance between the unknown node and its neighbouring anchors. To enhance both localisation accuracy and localisation success rates, the authors introduce a new neural network‐based node localisation scheme. The new scheme is distinct because it can make the trained network model completely relevant to the topology via online training and correlated topology‐trained data and therefore attain more efficient application of the neural networks and more accurate inter‐node distance estimation. It is also distinct in adopting both received signal strength indication and hop counts to estimate the inter‐node distances, to improve the distance estimation accuracy as well as localisation accuracy at no additional cost. Experimental evaluation is conducted to measure the performance of the proposed scheme and other artificial intelligent‐based node localisation schemes. The results show that, at reasonable cost, the new scheme constantly produces higher localisation success rates and smaller localisation errors than other schemes.