A novel Tunnel Positioning Approach via Long Term Evolution Cellular Signal

Abstract Currently, Radio Frequency Identification System (RFIS), ZigBee and Ultra-Wide Band (UWB) methods are mainly used to positioning in enclosed space. But they require complex hardware layout and high hardware costs, resulting in the inability to meet the positioning needs of complex environments. Therefore, we designed a novel tunnel positioning approach via Long Term Evolution (LTE) cellular signal. This approach includes: signal acquisition, data preprocessing, feature database construction, model training and real-time positioning. In the data preprocessing stage, we adopt 3sigma and Kalman filtering to filter outliers and noise, and use information gain and information gain rate to select effective features. In the real-time positioning phase, a combination of K-Weighted Nearest Neighbor (KWNN) and Support Vector Regression (SVR) is used for positioning in the tunnel. To verify the designed approach, we did an experiment using data from the actual tunnel. The experimental results show that this approach has better positioning accuracy than FK-NN [1] and TSVR [2].