| When mobile operators plan and deploy mobile communication networks,they need to ensure that the wireless network coverage and received signal strength of radio signals meet the needs of network users.At the same time,with the large-scale application of mobile communication networks,data organization and query for wireless network management and maintenance become more important and complex.The location of the base station is very important for the simulation and analysis of the spatial distribution and receiving strength of the radio signal,and an accurate radio propagation model is needed.The radio propagation model is key to simulation and analysis of radio signal spatial distribution,and accurate prediction of the coverage and received signal strength of radio signal;and effective organization of network trajectory and regional data can help to find and query weak coverage,over coverage and overlapping coverage areas,judge possible network interference areas,and help to optimize and adjust network deployment.In this regard,the research work described in this thesis is as follows.(1)Radio propagation model with parameter tuning and data-driving modeling.The propagation mode can predict the spatial distribution and received strength of radio signal by simulating radio propagation in network coverage areas.Due to the low prediction accuracy,empirical propagation models with fixed structure are not suitable for complex environment;while the coefficients of ray tracing models are relatively fixed and difficult to precisely determine.According to the application scenarios of ray propagation,this thesis designs a ray tracing-based propagation model with coefficient.This model uses driving test data and the genetic algorithm with appropriate crossover and mutation operators to calibrate the model coefficients.The root mean square error of model prediction is less than 8.10db.In addition,considering the high cost of geographic scene modeling for ray tracing models,this thesis also proposes a data-driven radio propagation model based on machine learning.The input features of the model are extracted by using the NLOS propagation characteristics of radio signals.The cross-scenario generation model structure is designed based on XGBoost,and the model is trained by using the actual driving test data.The model is evaluated and verified by the data collected in the urban areas.The results show that the root mean square error of the model prediction is less than 10.33db.The prediction accuracy of the above-mentioned propagation models is better than that of empirical models.The scheme of coefficient calibration provides higher accuracy.The prediction performance of XGBoost based cross scenario model is close to that of ray tracing models,and its modeling cost is lower than that of ray tracing models.Therefore,the XGBoost based model is an effective method for prediction of radio signal distribution in complex urban environment.(2)Area-oriented query of driving test trajectory.Driving test is an important means for LTE network coverage and interference investigation.In practical network maintenance,the operation and maintenance technicians often need to plan and search the driving test paths near a certain interference area in which more users report network quality problems,so as to conduct driving test to locate and solve the network problems.We model this real-life requirement as the problem of area-oriented trajectory query,and define the concept of spatial density correlation to measure the distance between the track and the query area,and the density of the interference spots,over coverage spots and bad-quality spots around the trajectory.A trajectory search algorithm for the query area is designed and implemented,which can search the trajectories from the existing driving test paths with most correlation.Finally,the effectiveness of the proposed method is verified to be effective for driving test trajectory query for network operation and management.(3)Spatial organization of coverage area in network.With the large-scale application of mobile communication networks,organization and query of spatial data in networks become more important and complex.In addition,operation and maintenance technicians may need to query the area near certain spots that meet certain interference condition or coverage type,or the track of a spot with certain received signal strengths,so as to analyze network quality.Due to the large scale and many dimensions of radio network signal data,the density-based clustering algorithm is used to generate coverage and interference area,and R-tree is then taken as the index structure for organizing interference and coverage spatial data.The inverted index R-tree is employed as the index organization structure of track data to support the organization and retrieval of certain scale area data and trajectory data.Based on the above research work,we design and implement the analysis software for radio signal spatial distribution in mobile communication networks,and verifies its effectiveness by using the LTE electric power network data of a city in southern China.The results show that the work described in the thesis can provide strong support for the spatial distribution analysis of radio signals in mobile communication networks. |
