Research On Radio Propagation Model Based On ANFIS

With the rapid development of society,people’s demand for wireless communication is more and more big,the cellular mobile communication is one of the most popular in the field of wireless communication field.The most important point in the design of cellular mobile communication system is the need to know by modeling from the transmitting antenna transmission to the spread of mobile antenna of radio signals.The quality of the received signal directly affected by the transmission line.This affects the success of the system.Path loss is caused by signal power attenuation between receiver and transmitter.Improper placement of base station in wireless system will lead to poor network performance and reduced quality of service.Therefore,path loss prediction model can be used to determine the most suitable location for deployment of wireless communication system.Radio propagation models are mainly divided into two categories,namely,the empirical model formed by fitting a large amount of data.The deterministic model obtained by mathematical calculation based on electromagnetic principles.In the past experiments,the empirical model is used frequently but produces large errors,while the deterministic model has high computational complexity and large computation.Therefore,introducing the ideas of the neural network to the path loss prediction.Fuzzy neural network combines the advantages of neural network and fuzzy inference system,not only has the learning ability of neural network,but also has fuzzy inference system reasoning ability.Adaptive neuro fuzzy inference system(ANFIS)is a kind of fuzzy neural network is widely used.It possesses the advantages of high efficiency and convenient.Therefore,ANFIS system is selected for path loss prediction to overcome the shortcomings of empirical model and deterministic model.In this paper,a radio propagation model based on ANFIS is constructed.Longitude,latitude,elevation and distance are taken as input data to determine the triangle membership function.When the number is 3 and the number of epochs is 280,the model constructed has the smallest error.The actual test data were obtained by CW test,and compared with the predicted data,it was found that ANFIS based path loss model produced a low prediction error with Mean Square Error(MSE),Root Mean Square Error(RMSE),Mean Absolute Error(MAE)values of 3.17 d B,17.05 d B,4.13 d B.And the error between the predicted data and the measured data was the least compared with Lee model,Okumura-Hata model,COS231 model and Egli model.It has strong generalization ability,and can also predict path loss well in the case of sharp changes in data.Besides,correlation coefficient(R)values of0.745,it means that the model has strong fitting degree and high explainability.On the other hand,this paper is based on the Chinese topographic region classification standard.The actual test path rendered by Leaflet is used to divide the test path into urban,suburban and mountainous according to building height and density.And the applicability of ANFIS model under these three terrain conditions is studied.The results show that ANFIS model performs well in urban,suburban and mountainous,and the error between predicted data and measured data is smaller than other models.ANFIS also with strong generalization ability and high fitting degree.Therefore,ANFIS model is suitable for path loss prediction of urban,suburban and mountainous.In conclusion,the radio propagation model constructed based on ANFIS has good prediction results and high availability.