Research On WIFI Underground Personnel Location Method Based On Partition And Location Fingerprint

The underground environment is complex and prone to various disasters and accidents,so it is necessary to conduct in-depth research and development on underground personnel location technology.At present,the underground location fingerprint positioning technology has problems such as positioning accuracy and stability not meeting the requirements,low positioning efficiency and heavy pre-data preparation,etc.The research is conducted around the above problem through the idea of region division combined with improved algorithm model,and the main work is as follows:In order to solve the problems of poor localization accuracy and low localization efficiency of large-scale scenes due to the fluctuation of signals in the underground environment,the paper proposes an algorithm for locating fingerprints of underground personnel based on improved partition matching and multitask Beetle Antennae swarm optimized random forest regression(P-MBAS-RFR),which introduces an improved partition matching model,divides the physical sub-regions of the roadway and then uses the improved partition matching model to locate the points to be measured The algorithm introduces an improved partition matching model,which divides the physical sub-areas of the roadway and then uses the improved partition matching model to partition the points to be measured to improve the location efficiency.The MBAS algorithm is used to optimize the parameters of the RFR positioning model,which solves the problems that the positioning model tends to fall into local optimum and the unsatisfactory positioning accuracy due to signal fluctuation,and the superiority of the P-MBAS-RFR algorithm is verified through simulation experiments.Physical partitioning can cause regional signal differences resulting in large errors at individual points to be measured and MBAS-RFR fingerprint localization is timeconsuming and laborious in the early stage of signal data collection,in addition to the algorithm optimization when the amount of data increases the parameter optimization calculation will gradually become complex.To address the above problems,the paper proposes an Improved K-means and Inexact Augmented Lagrangian Multiplier Method and Improved Weighted Nearest Neighbor fingerprint Localization for Local Distance-Based Outlier Factor(IK-means-IALM-LWKNN)based on P-MBAS-RFR algorithm.Firstly,the non-complete fingerprint data are clustered and partitioned by the improved K-means clustering algorithm,and the correlation coefficient combined with two types of Euclidean distance indexes are used to divide the signal fingerprints with large similarity into the same region,which effectively solves the problem of large errors at individual points to be measured due to the difference of physical partitioned signals.A non-exact Lagrange multiplier method is used to reconstruct the regional fingerprint data to improve the overall density of fingerprint data and reduce the manual workload.Subsequently,the local distance-based outlier factor algorithm is used to eliminate reference points with large outliers,and finally the specific localization is performed by mutual information(MI)combined with inverse distance weighted nearest neighbor algorithm,which solves the complex problem of parameter optimization calculation while ensuring the localization accuracy.The algorithm has good localization accuracy and stability and saves 40% of the manual workload through experimental comparison analysis of real scenes.The paper designs the beacon node hardware device and downhole beacon node distribution,followed by a comparative analysis of the algorithms proposed in Chapters 3and 4 from two directions respectively,through which it is concluded that the IK-meansIALM-LWKNN algorithm is more practical and more suitable for the application of largescale downhole localization scenarios.