Hybrid Prediction Model For Field Strength With Ray Tracing And Artificial Neural Networks

As the number of buildings increases in urban environment, radio propagationenvironment becomes more and more complex. The frequency band of radio is alsohigher. Technology of microcell and pico cell is adopted to improve system capacity.In micro cellular system, traditional statistical methods are inapplicable, due to thedissimilarity between districts. More accurate prediction of field strength is needed inthe planning and optimization of new generation of communication systems. In thiscase, the thesis focuses on the research of new efficient and accurate methods for thefield strength prediction.First, ray tracing is one of the widely used methods for field strength predictionin micro cellular and indoor environment. The reverse ray tracing algorithm based onthree dimensional scene database is implemented in this thesis. Although ray tracingalgorithm can provide accurate results, it is dependable on the accuracy of scenedatabase and requires long computing time.A hybrid scheme with ray tracing and neural networks is proposed to utilize theadvantages of neural networks and avoid the disadvantages of ray tracing. In theproposed method, ray tracing technique is used to coarsely predict the field strengthaccording to the low accuracy geographical databases. Then, the artificial neuralnetworks are trained using measurement data or simulation data from fine models topredict the effects caused by the detail geographical information in the propagationenvironments. On one hand, this proposed hybrid method does not needhigh-accuracy database for the propagation environments and thus shows lowcomputation complexity compared with traditional ray tracing methods. On the otherhand, the neural network method is used to compensate the effects contributed by thescattering objects of small dimensions in the scenarios, which improves the precisionof the predicted field strength. The hybrid model is verified in indoor environment.Simulation results show that using this hybrid scheme, the field strength can bepredicted in higher accuracy with less running time.