The Design And Implement Of Adaptive WKNN Algorithm For Bluetooth-based Indoor Localisation

The RSS-based method is one of the indoor localisation technologies.It measures the Bluetooth signal continuously transmitted by Bluetooth Low Energy(BLE)beacons,thereby estimating one's location.Unlike the unreliability of GPS in indoor scenarios,the Bluetooth signal is accessible via mobiles in such circumstances.Furthermore,the RSS-based method not only feature the simplicity of deployment and low cost,but also achieve desirable localisation accuracy in comparison with other indoor localisation methods.In this paper,we investigate the topology and number of BLE beacon deployment for indoor location determination based on Received Signal Strength(RSS).This investigation starts from above two factors that affect the localisation accuracy and accomplishes the following work: First of all,the locations of 377 samples are determined by manual site survey in the Mile End Library of Queen Mary University of London.RSS data are collected at each sample.On the basis of RSS data collection,a radio map which contains 377 samples is built.Then,with 3 topologies employed and 10 BLE beacons virtually removed one by one,we utilise this radio map and another 33 test samples to compute the estimation error at different beacon densities.In terms of the localisation algorithm,an adaptive WKNN(Weighted K Nearest Neighbours)algorithm is also proposed to address the problem of traditional KNN algorithm for the error computation.Owing to the selection of constant K values,the traditional KNN algorithm and WKNN algorithm fail to obtain the optimal estimated location.The results suggest that a topology called “Square” with 10 beacons removed could acquire the lowest estimation error.In addition,the proposed adaptive WKNN algorithm dramatically outperforms the KNN as well as WKNN algorithms.This helps us to optimise the indoor localisation system.