Research And Design Of Vehicle Channel Modeling Software Based On Ray Tracing

In the vehicle-to-anything wireless communication(V2X)environment,describing vehicle wireless communication channels by measuring data and using stochastic model cannot be widely applied due to the limitations of measurement equipment and costs.The channel modeling software based on ray tracing cannot satisfy the low price,static simulation,dynamic simulation,and complete domestic maps at the same time.An accurate vehicle channel modeling software based on digital map and ray tracing is designed.The main work of the paper is as follows:Firstly,the V2 X channel model and model evaluating indicator are introduced.Then the theory of deterministic channel modeling method and ray tracing including classic algorithm and acceleration algorithm are summarized.After that,the data structure of the KD-Tree used in ray tracing is studied in detail,including the construction algorithm and the traversal algorithm.To solve the problem of slow construction of the KD-Tree,multithreading and caching is used,the experimental results show that the multi-threaded KD-Tree construction algorithm reduces the tree-building time.The traversal algorithm steps and flowchart of KDTee for channel modeling and parallel algorithm running on GPU for acceleration are detailed.The simulation result shows that the average running time of the GPU-based KD-Tree traversal tracking algorithm which have high performance is 1/6 of CPU-based.Secondly,the design ideas and implementation methods of channel modeling software are introduced.Based on C ++ 11 and Qt,the simulation software implements functions such as loading 3D digital map,intercepting 3D digital map,KD-Tree construction and tracking,calculating simulation results,and visualizing simulation results.Based on the object-oriented programming,the software data structure is abstracted.The idea of modularization and separation between user and computation are used in the software.The thread synchronization points are proposed to manage and monitor the state of thread pool and GPU when for CUDAbased parallel tracking algorithm.The design patterns such as the intermediary model,trategy model,factory model,and builder model are used to reduce the coupling between program classes and maintain class scalability.CI and ABG models fitted by minimum mean square error are supported.To simplify the modeling process,machine learning channel modeling using DNN and XGBoost is built into the software.Finally,the software was used to perform static single receiver,static multiple receiver and dynamic simulation of V2 V,and the visualization results and path model fitted by minimum mean square error were given.The comparison of path loss obtained by software simulation and the path loss calculated by the model given in ITU-R P.1411 shows that the software can correctly give the relevant results of channel modeling.