Mixed K-means and GA-based weighted distance fingerprint algorithm for indoor localization system

This paper proposes an application of Wireless Sensor Network (WSN) for indoor localization using IEEE 802.15.4 standard. Proposed algorithm applies K-means clustering and Genetic Algorithm (GA) as engine to prepare offline information which result in increasing accuracy and decreasing computational cost of fingerprint technique for indoor localization. K-means clustering will be applied to cluster received signal strength indicator (RSSI) vector into several classes for coarse positioning estimation. Consequently, GA will be applied to search the optimal weights for each reference sensor and used in order to obtain more accuracy for positioning estimation. Experiments are conducted in indoor environment using zigbee sensor network and the proposed algorithm can be compared with K-Nearest Neighbor (KNN) algorithm and conventional weighted distant fingerprint (WDF) algorithm. Results demonstrate that the proposed algorithm can improve an accuracy increase to 87.56 % for identifying correctly 1.5 m × 1.5 m area of target node and also decrease computational cost of 67.60 %.