Weighted Centroid Localization Algorithm Based on MEA-BP Neural Network and DBSCAN Clustering

Abstract

In order to overcome RSSI ranging error and improve the accuracy of positioning results, a weighted localization algorithm based on MEA-BP Neural Network and DBSCAN clustering is proposed in this paper. This algorithm uses MEA-BP Neural Network (MEA-BP NN) model to optimize ranging information firstly, then it uses trilateral measurement method to get multiple initial localization results about unknown node and form a set. After clustering the results by DBSCAN and eliminating noise points, the estimated coordinate of unknown node in each cluster is obtained by using the weighted centroid localization algorithm based on collinearity. Next the number of core points in each cluster is regarded as weight value, the weighted centroid localization algorithm is used again, thus the final coordinates of unknown node can be got. Simulation results show that the localization accuracy of wireless sensor network can be improved significantly by using this algorithm in two-dimensional scene.