A Data-Fusion Approach for Speed Estimation and Location Calibration of a Metro Train Based on Low-Cost Sensors in Smartphones

Since the GPS is unavailable in underground environment, it is extremely challenging to measure the speed and location of a metro train. This paper proposes a novel data-fusion approach for speed estimation and location calibration of a metro train in underground environment, simply using the data from the 3-axis accelerometers in smartphones. Firstly, we place multiple smartphones in different cars of a train to measure the longitudinal, lateral and vertical accelerations, then propose a method to transform the measured accelerations from the coordinate systems of smartphones to that of the metro train. In the data fusion model, the initial estimations of train speed and position are obtained by the integral and double integral of the longitudinal accelerations. The lateral and vertical accelerations are used to provide absolute reference for speed estimation, where the local time delay and waveform similarity between the measured accelerations in different smartphones are defined and estimated to obtain the time-delay-based speed. Finally, a more accurate estimation of train speed is obtained by fusing the integral-based speed and the time-delay-based speed. A case study is conducted on Chengdu Metro Line 7 in Chengdu, China. The results show that, taking the interval length between adjacent stations as ground truth, our data-fusion approach achieves higher accuracy than the direct integral method, with the relative errors reduced from 9.5% to 1.6%.