An Experimental Study of an Improved BLE Indoor Positioning Scheme Based on Differential Distance Correction

Abstract

Indoor positioning has attracted much attention in recent years due to the trend of Internet of Things (IoT), which is capable of providing numerous applications such as personal tracking, vehicle locator, and Location-Based Service (LBS). Since Global Navigation Satellite System (GNSS) is no longer useful in the indoor environment because of the sheltered satellite signals, Bluetooth Low Energy (BLE) is a good alternative for realizing indoor positioning. BLE, the latest version of Bluetooth in 2010, has the advantages of low cost, low power consumption, and pervasive availability in smartphones. However, BLE still faces some challenges in terms of the fluctuating Received Signal Strength Indicator (RSSI) caused by the environment such as reflection, fading and multipath effect. Those variable RSSI values make it difficult to derive positioning result accurately. To reduce the environmental effect, the concepts of Differential Global Navigation Satellite System (DGNSS) and Network Real Time Kinematic (NRTK) motivate a novel method named Differential Distance Correction proposed in this study. This method utilizes the differential information from the reference station at the known coordinate to correct the distance measurement of the rover station and further enhance the positioning accuracy of it. In this study, power regression model is used to convert RSSI to distance; then Differential Distance Correction is applied to obtain better distance measurements of the rover. In addition, trilateration is selected as the primary positioning technique, which exploits estimated distances to determine the target’s position. This technique can be solved by two main approaches: linear and nonlinear least squares. With corrected distance measurements, it can gain more accurate result through trilateration compared with simply using the original distances. The experimental results presented in this study illustrate that the overall positioning error using the proposed method can be reduced to less than 2.5 m in 90% positioning error, and 2 m in RMSE with both improvement ratios over 30%.