| Real-time information of the train location is the fundamental of train control system.An important method to improve the positioning accuracy and reliability is using integrated train positioning system.The positioning precision is mainly affected by data synchronization and noises produced by inertial sensors.The pre-processing method of multiple sensors must be studied in order to assure that the algorithm is correct and effective,to conduct comparison and evaluation of different fusion algorithms.First,the data transmission delay and decoding delay of each sensor(eg.gyroscope, accelerometer)are discussed in this thesis,and these delays are also compensated. Further more,the problem of the data synchronization of different sensors,such as GPS/GPS,GPS/IMU and GPS/ODO,are solved.Different kinds of de-noising methods are compared in the thesis.Depending on the characteristic of train positioning sensors, wavelet analysis is selected to deal with the noises of inertial sensors.The db4 wavelet is selected based on the comparison of different kinds of wavelet de-noising fuctions. With the increasing of decomposing scale,the useful signals will be strengthened and the noises will be weakened.Depending on this trait,an improved Donoho threshold rule,which uses different thresholds at different scales,is brought forward in the thesis. The data reconstruction method can supply more reliable data for the analysis and comparison of different fusion algorithms.An universal data acquisition platform is designed in this thesis,based on ARM7 processor,with multi-sensor,such as GPS,IMU and ODO.In order to validate the method of data synchronization and de-noising presented in the thesis,field tests is carried out in Beijing Railway Bureau San Jiadian marshalling yard,the simulation is also did in Matlab.The results shows that,after data synchronization and de-noising,the precision and reliability of the intergrated train positioning system can be improved by using the same fusion algorithm.
|
PDF Full Text Request