Search for a command to run...
The presentation of Bluetooth Low Energy (BLE; e.g., Bluetooth 4.0) makes Bluetooth based indoor positioning have extremely broad application prospects. In this paper, we propose a received signal strength indication (RSSI) based Bluetooth positioning method. There are two phases in the procedure of our positioning: offline training and online locating. In the phase of offline training, we use piecewise fitting based on the lognormal distribution model to train the propagation model of RSSI for every BLE reference nodes, respectively, in order to reduce the influence of the positioning accuracy because of different locations of BLE reference nodes. Here we design a Gaussian filter to pre-process the receiving signals in different sampling points. In the phase of online locating, we use weighted sliding window to reduce fluctuations of the real-time signals. In addition, we propose a distance weighted filter based on triangle trilateral relations theorem, which can reduce the influence of positioning accuracy due to abnormal RSSI and improve the location accuracy effectively. Besides, in order to reduce the errors of targets coordinates caused by ordinary least squares method, we propose a collaborative localization algorithm based on Taylor series expansion. Another important feature of our method is the active learning ability of BLE reference nodes. Every reference node adjusts its pre-trained model according to the received signals from detecting nodes actively and periodically, which improve the accuracy of positioning greatly. Experiments show that the probability of locating error less than 1.5 meter is higher than 80% using our positioning method.