Study Of GSM-R Wireless Propagation Model Based On BPNN And SVR

GSM-R (Global System of Mobile for Railway) is a wireless communication system. It bases on GSM, the public wireless communication platform, and designs specifically to fulfill the needs of dispatch and control and to support the working of high-speed railway. To use the GSM-R system better, railway wireless communication channels should be studied for designing the network better and for predicting the input voltage of receiver more precisely. Studying wireless communication propagation model is very important. A good propagation model should be able to describe the propagation between sender and receiver under the influence of nature environment, man-made structures and engineering settings. In the real applications, by choosing the proper propagation model, mid-value of path lose can be obtained. Combining with the budget of wireless communication link, the receiving voltage and cell radius can be calculated, which satisfy different level of railway and acceptance criteria of different signal standards. Therefore, the design accuracy can be improved, various resources can be utilized more profitably, meanwhile time and construction cost may be reduced.Data mining can find hiding information from mass data using different algorithms. It can also uncover and show factors and their inner relations. These excellent features are used to study the propagation models of GSM-R system in this paper.In this paper, the properties of wireless propagation and the structure of GSM-R are introduced; the propagation models that are commonly used in GSM-R wireless communication system are summarized. Some brief facts of selected High-speed Railway are also introduced. And this selected High-speed Railway are classified into different scenes for further studies. After collecting and optimizing the actual propagation data, BP neural network algorithm and SVR algorithm are applied to predict the receiver voltage of wireless communication system of selected High-speed Railway, aiming at finding a more accurate propagation model related to frequency, geography conditions, propagation distance,imputation of base station and remote unit, sender power, receiver sensibility and the height of tower. Using some examples, the questions of how to plan and optimize communication network depending on data-mined propagation models suitable for various situations are discussed at the last part of this paper.