Accuracy Enhancement Technology of BLE Device Indoor Positioning System Based on Kalman Filter and Machine Learning

• Main Authors: Jian-Zhi Liao, 廖健智
• Other Authors: 陳煥
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/2k7gy6

碩士 === 國立中興大學 === 資訊科學與工程學系所 === 107 === In recent years, outdoor positioning technology has been quite mature, but subject to the indoor environment for the obstacles such as GPS satellite signals, many research and discussion on indoor positioning technology has been launched, and the technology with lower cost and convenient construction method is more convenient, such as Bluetooth, Wi-Fi, etc. This paper is based on the 2013 iBeacon proposed by Apple Inc. The Bluetooth signal RSSI value drifting seriously, which will result in poor positioning. At present, most of the literature on indoor positioning mainly proposes how to apply technology to achieve and improve positioning. There are few discussions about the relationship between density of signals collection and accuracy of positioning[1]. According to the experimental results of this study, under the same conditions, the higher the positioning signal collection density, the lower the positioning accuracy. This paper explores the inaccuracy of the Beacon positioning device built in the Art Center of National Chung Hsing University, and attempts to propose a set of algorithms that can be used to improve the accuracy of existing indoor positioning. In this study, the Kalman Filter (KF) is used to improve the stability of the Bluetooth signal RSSI, combined with machine learning algorithms to improve the accuracy of the indoor location. The instant positioning method proposed in this paper uses iBeacon and Android smart phone as experimental devices to test and compare the differences between K nearest neighbor (KNN), support vector machine (SVM) and random forest (Random Forest) algorithms. The experimental results show that when the indoor positioning signal is collected at a density of about 1 meter, the positioning accuracy is optimal.