Robust train length calculation and monitoring method using GNSS multi-constellation moving-baseline positioning resolution

Train length status reflects whether the carriage is uncoupled or thrown away, it directly affects the safety and efficiency of train operations. At present, satellite positioning technology is used within a train length monitoring system. Such a system can ensure positioning accuracy while reducing the need for trackside equipment. For train length monitoring using GNSS technology, multi-constellation can utilize more visible satellites and achieve better spatial geometric distribution. A robust train length calculation method using multi-constellation GNSS moving-baseline positioning resolution is proposed. The time and space coordinate systems of the Global Positioning System (GPS) and BeiDou Navigation Satellite System (BDS) are unified as the foundation to realize multi-constellation positioning. Double-differenced carrier phase measurements are used to mitigate or eliminate the propagation errors and the satellite and receiver clock errors. Then, the moving-baseline length can be estimated using a Kalman filtering algorithm, and the carrier phase ambiguity terms are fixed using an online ambiguity fix algorithm to further improve the system performance. More measurements are utilized in the multi-constellation train length calculation method, the probability of fault measurement would be increased thereafter, and thus, the fault identification and adaptive filtering algorithm is introduced in the multi-constellation train length calculation to reduce the influence of fault measurement on the accuracy of moving-baseline solution and enhance the robustness of the system. To evaluate the performance of the proposed system, an experiment was conducted on the Beijing-Shenyang high-speed railway line. The results obtained on GPS-FLOAT, GPS-FIX, GPS/BDS-FLOAT, and GPS/BDS-FIX modes were compared. Both the FLOAT and FIX solutions can achieve the train length computation with sub-meter level accuracy, which can prove that the system is able to be adapt to the different GNSS signal conditions. The GPS/BDS-FIX solution can achieve train length accuracy with RMS of 0.155 m, which is better than the other three solutions. In addition, the moving-baseline accuracy is further improved after the adaptive filtering algorithm is added to smooth the noise interference, which had the best performance with RMS of 0.077 m compared with the non-adaptive filtering solutions. The superiority of the adaptive filtering algorithm is verified via using the dataset including two types of random fault measurements. It is proved that the random faults can be tolerated using adaptive filtering algorithm. The effectiveness and superiority of proposed robust train monitoring system are confirmed via comparison of results and simulation of different scenes.