Study On Compensation Method Of NLOS Based On Signal Feature Map

Seamless indoor-outdoor high-precision positioning technology,not only plays an important role in the medical industry,online travel,fire rescue,smart city,internet of things and other aspects but also become an important booster of "Internet +" economy.Compared with positioning method based on Bluetooth,infrared,WiFi,RFID,ultra-wideband and so on,mobile network positioning has several obvious advantages of low maintenance cost,high universality,strong regional applicability and wide user coverage.There are several positioning technologies based on mobile communication network,in which the positioning method based on TOA and AOA is affected by many factors.Among these factors,non-line-of-sight error has the greatest impact on the positioning accuracy.How to compensate non-line-of-sight error has always been a problem for many scholars at home and abroad.There are three major diffculties mitigating the non-line-of-sight error in complex environment:1.signal propagation complexity;2.instability of received signal characteristics;3.uncertainty of Non-line-of-sight error.The main contributions and innovations of this dissertation are as follows:1.Prediction based indoor signal differential feature map.The indoor and outdoor three-dimensional space information,space electromagnetic information and space network element parameters are fully researched and integrated,and indoor and outdoor 3D map construction and channel analysis are completed.A method of constructing indoor signal feature map based on deterministic model is proposed,which use difference method to obtain the finely-available differential signal characteristic map.Then we can calculate the differential non-line-of-sight error(NLOS error)distribution based on the signal characteristic map.This method can reproduce the essential process of signal propagation in the actual positioning system,reduce a large number of repeated acquisition steps,and can construct structure of any environment,which lead to better spatial adaptability and generalization ability for different environments.The algorithms and analysis followed can get high quality and reliable data sources from this method.2.Adaptive updating method of differential signal feature map based on virtual station.Aiming at the instabilities of the receiver signal characteristics,this paper proposes a multi-layer mesh division method based on the quadtree structure,and proposes an adaptive correction method for the differential signal feature map of the sparse virtual station.Through sparse virtual station differential signal feature change values,the signal features and non-line-of-sight errors of each grid point are corrected layer by layer,and an adaptive differential signal feature map under dynamic environment is obtained.Experiments show that in the time-varying environment,the method can update and modify the signal feature map faster and more accurately than the traditional Kriging interpolation method.3.Non-line-of-sight error calculation model based on the characteristics of the differential signal.For the indeterminacy of non-line-of-sight errors,this paper analyzes the characteristics of various types of signals,and obtains the mapping relationship between the characteristics of differential signals and non-line-of-sight errors.In addition,the NLOS error prediction algorithm based on TD-GA-BPNN is proposed.The feature points that cannot be mapped are dimension-enhanced,and the tag library is added.In combination with BP neural network and genetic algorithm,the NLOS error correction performance is improved through learning and improvement.From the calculation method,the minimum root-mean-square error can reach 0.8644×10-16s.Compared with BPNN,this method has improved accuracy by 36%and training time has been shortened by 33.33%.Compared with the improved BPNN algorithm based on PSO and GA,the test errors were improved by 19.68%and 31.51%,and the computing time was improved by 19.05%and 25.4%,respectively.The proposed differential NLOS mitigation method based on signal feature map can achieve high precision,stability and adaptability.The accuracy of 80%positioning accuracy is 1.13m in calibration environment and 1.89m in dynamic environment and the average error in the actual positioning environment is less than 2m.