The Research On Propagation Model Calibartion In Wireless Network Optimization

In recent years, mobile communications are at a stage of rapid development. With the increasing number of application and users, there are growing concerns about the radio network optimization. High-quality network optimization can hardly be achieved without an accurate radio wave propagation prediction model, and a relatively precise propagation model could contribute to an accurate assessment of network coverage status, which provides a strong theoretical basis for improving the network performance. The environment of mobile communication is very diverse, and varies from places to places. Therefore, in order to gain a propagation model reflecting the local wireless environment with high accuracy, a calibration of radio propagation model is required significantly.Firstly, this paper briefly introduces the network optimization technology of mobile communication, including the objectives, contents and process of the network optimization. Meanwhile, based on the research of the characteristics of the radio wave propagation, this paper analyzed several commonly-used radio wave propagation model. The empirical models are simple, and with a good applicability, but the accuracy of prediction is not ideal. Recently, research shows that artificial neural network presents a good performance in propagation prediction, and relevant algorithms reveal a strong ability of nonlinear fitting.After comprehensive study of traditional and current methods of propagation model calibration, by combining the strengths of both Okumura-Hata propagation model and neural network algorithm, this paper comes up with a calibration algorithm based on back propagation neural network, with adaptive classification according to Hata propagation model. Take the project of CDMA network optimization in some province of south China as an example, and verify propagation model calibration in several different areas. This paper also did some performance analysis and compassion between the improved algorithm and other algorithms. Experiments results shows that, compared to other algorithms, the improved algorithm has higher prediction accuracy, a better ability of nonlinear fitting, and a stronger applicability to wireless environment.In the end, a simple radio network optimization system is designed, which evaluated the coverage status based on the proposed propagation model calibration method, and conducted network performance improvement on the prediction results. The experiments provide a valuable reference for the local CDMA network optimization, which has a practical significance.