Study on the application of indoor positioning based on low power Bluetooth device combined with Kalman filter and machine learning

EasyChair Preprint 1198

Abstract

In recent years, outdoor positioning technology has approached maturity, but the Global Positioning System is limited by environmental factors and obstacles, and has no effect indoors. There are many research and discussion on indoor positioning, among which the lower cost construction methods are Bluetooth and Wi-Fi. This study uses a device based on the iBeacon protocol proposed by Apple in 2013 as a tool for this research. Due to the Received Signal Strength Indicator (RSSI) value from the Bluetooth is unstable which will affect the positioning results, this research used Kalman Filter Algorithms to improve the RSSI stability of Bluetooth and used machine learning algorithms to improve indoor positioning accuracy. iBeacon and Android smart phones were used as experimental devices to test and compare the differences between K nearest neighbors (KNN), support vector machines (SVM) and random forest algorithms. The experimental results indicate the optimal signal collection density for indoor positioning is about 1 meter and the accuracy can reach to more than 85%. The statistics show that the model which trained by KNN algorithm has the highest accuracy.

Keyphrases: K-nearest neighbor algorithm, Kalman filter, Random Forest Algorithm, indoor positioning, support vector machine algorithm

@booklet{EasyChair:1198,
  author    = {Jia-You Hsieh and Chun-Hung Fan and Jian-Zhi Liao and Jyh-Yih Hsu and Huan Chen},
  title     = {Study on the application of indoor positioning based on low power Bluetooth device combined with Kalman filter and machine learning},
  howpublished = {EasyChair Preprint 1198},
  year      = {EasyChair, 2019}}