| In recent years,the full commercialization of the 4G LTE mobile communication networks and the issuance of 5G commercial licenses by the Ministry of Industry and Information Technology in China have announced the arrival of the 5G era.The demand for mobile communications has grown by leaps and bounds,and the requirements for high-quality communication networks are also increasing.Therefore,how to quickly and efficiently analyze,maintain and optimize the wireless network quality is an urgent problem.Ray tracing is a key technique for network coverage and interference analysis in wireless network optimization.In order to deal with the difficulties of high computation volume and slow calculation speed in ray tracing models,this thesis proposes a parallel ray racing method by using NVIDIA’s Optix ray tracing engine and CUDA programming paradigm,to analyze the spatial distribution of radio signals in wireless networks.The study described here is organized as follows.(1)To quickly and accurately analyze the spatial distribution of radio signals in complex urban areas where buildings are densely deployed,a GPU-based parallel ray tracing algorithm is proposed to simulate and track the propagation trajectory of radio signals in networks.Firstly,the terrain and buildings in areas are modeled by the triangulation algorithm;the primary-level direct,reflection and diffraction rays are generated from signal sources such as eNodes in network.The redundant primary-level rays are removed by occlusion test and ray merging;then,the Optix engine is used to parallely detect collision,generate and track subsequent-level rays,and calculate path propagation attenuation;finally,the method of main path de-duplication is employed to solve the double counting problem,and the received field strengths of the signals arriving at locations in network are calculated by accumulating the signals from multiple propagation trajectories,so as to obtain the signals’ spatial distribution in network.Compared with the CPU-based ray tracing approaches,the proposed GPU-based scheme achieves a speedup ratio of 12 or more.(2)A practical LTE network deployed in a southern city in China is simulated and taken as the testbed to evaluate the feasibility and effectiveness of the proposed parallel ray tracing approach and its applications in coverage analysis,intra-network interference identification and extra-interference localization.The evaluation results are compared with that of CPU-based schemes and demonstrate the superiority of the GPU-based approach.The metrics of coverage analysis include the accuracy of cell-level and region-level analysis results and computation time;interference identification calculates the percentage of interference defect points in coverage areas,and identifies the possible causes;interference localization takes parallel ray tracing to quickly generate possible propagation trajectories of interference signals for determining the locations of interference sources.(3)A software system for analyzing the spatial distribution of wireless signals is developed to meet practical application requirements.The system adopts Vue framework for front-end,thrift-based RPC communication for back-end and Restful API for architecture.It has the properties of front and back-end separated architecture,high response and low coupling among the software modules.It is applied to optimization of a practical LTE power wireless network in a southern city,and provides rapid and precise results of network coverage and interference analysis.In summary,the study proposes a GPU-based parallel ray tracing approach with a high GPU acceleration ratio,which deals with the difficulty of slow calculation speed caused by intensive computation in ray tracing in large-scale network deployed in complex and dense urban areas.The method and the developed software system can provide accurate and quick network coverage and interference analysis results for practical large-scale networks,and improve the efficiency of network management and optimization. |
PDF Full Text Request