Research On The Key Problems Of Radio Fingerprints Positioning For Coal Mine WLAN

Coal mine underground personnel positioning is one of the important application fields of coal mine safety management.Radio fingerprints positioning in wireless local area network(WLAN)has gradually become a solution for underground personnel positioning application because of its low cost,controllable accuracy,non-ranging and other characteristics.At present,there are still several major problems in radio fingerprints positioning,including the noise interference in the process of radio fingerprint samples collection,the extremely tedious construction and maintenance of radio fingerprints database,the radio fingerprints positioning problem when underground access points(APs)are sparse distributed,and the low accuracy and efficiency of online radio fingerprints matiching.This issue studies and improves the above key problems one by one.Aiming at the problem of noise interference in fingerprint samples,a τ sampling interval autocorrelation filtering algorithm is proposed to reduce the noise samples.It improves the effectiveness of radio fingerprint samples at each reference point.The τsampling interval autocorrelation filtering algorithm makes use of the volatility of the original received signal strength RSS and takes the sample mean values as the reference of the fluctuation difference between the samples to enlarge the difference characteristics of the noise samples,so as to accurately find and filter out the noise samples.The experimental results show that the positional accuracy of the radio fingerprint samples using τ sampling interval autocorrelation filter was 1m and 0.5m higher than that of the radio fingerprint samples without filtering and using Link Quality Indication(LQI)filter respectivelly.In order to solve the problem of complicated construction and maintenance of the radio fingerprints database,an adaptive construction and update method of the radio fingerprints database based on Quantum Behaved Particle Swarm Optimization-User Location Trajectory Feedback(QPSO-ULTF)algorithm is proposed.The construction process of the radio fingerprints database takes the calibration samples of the candidate reference points of the location track points as the scale,and the QPSO algorithm is used for adaptively adjusting the RSS data feedback by the user location trajectory.The updating process of the radio fingerprints database takes the RSS data of the location track points as the scale,and the QPSO algorithm is used for updating the correction samples of the corresponding reference points.Then the radio fingerprints database completes the updating through the secondary construction of the fingerprint samples.The experimental results show that the QPSO-ULTF algorithm can effectively improves the manual acquisition process of radio fingerprints database and enhances the robustness of the underground personnel positioning system.For the of radio fingerprints positioning when underground APs are sparsely distributed,a positioning method based on multi-association virtual access point(MA-VAP)is proposed.In the MA-VAP method,the multi-association function of virtual access point(VAP)is constructed by least square linear fitting between VAP and off-line sampling data of multiple AP signals.The RSS value of VAP is generated by multi-association function.The generated RSS values by VAP can make up the lack of some information elements of RSS sequence caused by AP sparse distribution.The experimental results show that the MA-VAP method can effectively solves the radio fingerprints positioning problem when underground APs are sparse distributed.The positioning accuracy of the MA-VAP method was improved by 22.22% compared with that of the VAP method.In order to improve the accuracy and efficiency of the radio fingerprint matching process in the online positioning stage,a radio fingerprint positioning method based on region division and AKPCA(Adaptive Kernel Principal Component Analysis)algorithm is proposed.The class relationship K-Means(CRK-Means)algorithm effectively solevs the singularity problem of the sub-regions dividing.And the GA-RF algorithm combining the random forest(RF)algorithm and the genetic algorithm(GA)is used to improving the sub-region positioning accuracy.The AKPCA algorithm,which combines the optimal AP selection algorithm with kernel principal component analysis(KPCA)algorithm,makes the calculation of eigen dimension have a certain sub-region adaptability.AKPCA algorithm effectively improves the problem that the eigen dimension solved by the maximum likelihood estimation method in KPCA algorithm is too single for the radio fingerprints database.The experimental results show that the positioning accuracy of AKPCA algorithm reached 2.5m when the confidence probability is 90%.At the same time,the AKPCA algorithm can effectively reduce the resources consumption during the positioning process.The dissertation contains 76 figures,15 tables and 154 references.