Research On Coal Mine Underground Location Method Based On CSI

Coal is the main energy source in China,and ensuring coal mine safety production has always been the focus of the country.The underground personnel location system plays an important role in improving the safety generation level of coal mines.Due to its easy realization,fingerprint location method based on received signal strength has become the main direction of positioning technology.However,the received signal strength is susceptible to environmental interference,and the location accuracy is not high enough to meet the requirements of future applications.Therefore,it is of great theoretical and practical significance to study the method of high-precision fingerprint location in underground mine.At present,there are three main problems in downhole fingerprint positioning:(1)lack of features to generate high-precision fingerprints;(2)lack of fingerprints suitable for banded sparse distribution of wireless access points;(3)inaccurate description of location by a single fingerprint.The research work is carried out on the above issues.(1)Analyze the factors affecting the positioning and verify that the received signal strength is not suitable for high-precision fingerprint positioning.By studying the relationship between channel state information and transmission path,a path transmission model based on channel state information is proposed to provide theoretical support for constructing fingerprint state information.Experiments show that compared with the received signal strength,the channel state information has finegrained characteristics,which can describe the difference of different positions from the two dimensions of amplitude and phase,and is more suitable for high-precision fingerprint positioning in underground.(2)Channel state information consists of amplitude and phase.Aiming at the construction of amplitude fingerprint,firstly,the sources of amplitude noise are analyzed,and a variety of filters are proposed to suppress the interference of noise to amplitude.Then,the amplitude fingerprint is generated by combining the characteristics of MIMO.Aiming at the phase fingerprint construction,this paper firstly analyses the factors causing the phase measurement error,and proposes a linear transformation algorithm to process the phase error.Then,by studying the relationship between the transmission path and the channel state information phase,a Hankel matrix constructed from the subcarrier phase is established.Finally,the phase of the path is obtained by decomposing the Hankel matrix by using the Vandermonde matrix decomposition method,and the phase of the obtained path is generated.Experiments are carried out in downhole in line of sight and Non-line of sight scenes respectively.The experimental results show that the average positioning error of fingerprint location method based on channel state information is about 55% lower than that based on signal strength fingerprint location method.(3)In order to further improve the accuracy of fingerprint positioning,the influence of off-line training stage meshing on the accuracy of fingerprint positioning is studied.Combining the advantages of Fuzzy c-means(FCM)algorithm and Linear Discriminant Analysis(LDA)algorithm,and utilizing the fast and accurate optimization characteristics of quantum genetic algorithm,a method of fuzzy LDA fingerprint fusion based on quantum genetic algorithm is proposed.Quantum genetic algorithm is used to find the optimal fuzzy factor to suppress the fluctuation of fingerprint characteristics.The experimental results show that this method can help refine the mesh,and the average positioning error is reduced by about 20% compared with the pre-processing fingerprint.(4)In the traditional fingerprint location method,the location is described by using a single fingerprint.Since the fingerprint is time-varying,a single fingerprint cannot accurately represent the relationship between the location and the fingerprint feature.To solve the problem of inaccurate position description of a single fingerprint,this paper proposes to transform a single fingerprint into a sequence fingerprint to describe the position and a temporal difference LSTM sequential fingerprint matching method for variable length sequences is proposed.The experimental results show that in line of sight scenarios,the average positioning error of sequence fingerprint is 1.48 meters,which is about 25% lower than that of single fingerprint positioning method and 72% lower than that of RSSI-based fingerprint positioning method.In Non-line of sight scenarios,the average positioning error of sequential fingerprints is 1.71 meters,which is about 28% lower than that of single fingerprint positioning method and 71% lower than that of RSSI-based fingerprint positioning method.The paper contains 89 figures,24 tables and 147 references.