Fingerprint-based fusion of magnetic field data with multiple wireless technologies for indoor mobile robot positioning

Abstract

Mobile robot pose information is getting utilized in more and more indoor applications. In this paper, a novel fingerprinting-based positioning method is proposed. The method utilizes magnetic field data together with measurements collected using anchors equipped with multiple wireless technologies. The collected fingerprints are used to train MultiLayer Perceptron (MLP) neural networks, which can later be utilized to estimate the 2D position based on sensor readings in an unknown position. Real measurements collected in a laboratory are used to validate the performance of the method. The measurement database consists of 3D magnetometer measurements and four different parameter types measured between the mobile node and the anchor nodes using multiple wireless technologies. The parameters include Time of Flight (ToF) measurements using Ultra-WideBand (UWB) and Received Signal Strength Indicator (RSSI) measured using WiFi, UWB and 433 MHz frequency band. Various combinations of data were tested in the evaluation process. The obtained results showed that significant improvement, even more than 35%, can be achieved in the positioning performance for most combinations by fusing magnetometer measurements and wireless communication data.