Research On Indoor Positioning Technology Based On Machine Learning

With the rapid development of network optimization,ubiquitous computing and the Internet of Things,indoor positioning technology has been applied more and more in fire fighting and disaster relief,underground parking and car search,shopping malls and other fields.In recent years,machine learning technology has been widely used in many fields and achieved good results.Therefore,more and more researchers begin to use machine learning method to study indoor positioning,which promotes the development of machine learning in indoor positioning.This paper studies the indoor positioning technology based on machine learning.Aiming at the problems existing in the existing research,three effective indoor tracking and positioning algorithms are proposed.Firstly,in order to improve the accuracy of location,this paper proposes a location algorithm based on access node selection,and presents two access node selection algorithms to determine the location of part of the access node participating in the target,namely,the access node selection algorithm based on greedy heuristic and the access node selection algorithm based on neural network.Secondly,in the localization process,the overlapping triangle region formed by reference points is used to determine the target area,and several iterations are used to gradually reduce the size of the target area.Then,the exact location of the target node is estimated using the triangle's center of mass.Finally,the advantages and effectiveness of the proposed algorithm are verified by experiments.The effects of the number of reference points and mesh size on the positioning results were analyzed,and the positioning accuracy of the target under different trajectories was investigated.Secondly,this paper studies the fingerprint location technology method based on probability,and proposes a target location method MLP-Bay based on multi-layer perception classification model and Bayesian probability.MLP-Bay uses a combination of multi-layer perceptrons and Bayesian algorithm to estimate the location of the target.Firstly,in the offline stage,the MLP classification algorithm learns the fingerprint database to train the MLP model;in the online stage,we put the collected target RSSI vector into the trained MLP model,and use the trained MLP model fingerprint database to find the fingerprint vector that is most similar to the target's RSSI vector.Secondly,we will use MLP to select the most similar fingerprint vector and the target RSSI vector to calculate their similarity probability value using the Naive Bayes probability method.Then,we multiply the similarity probability value with the coordinate corresponding to the most similar fingerprint vector in the fingerprint database,so as to estimate the target position.Finally,we test the proposed algorithm in a simulation environment,and the experimental results show the effectiveness and superiority of the MLP-Bay method.Finally,this paper proposes an indoor location algorithm based on principal component analysis and long and short time memory.Firstly,the algorithm collects RSSI measurement values from different access points according to the movement trajectory of the target,and then forms RSSSI vectors and stores them in the fingerprint database.Secondly,the principal component analysis method is used to extract the main features of each group of RSSI vectors in the database,and dimensionality reduction processing is carried out on the data,so as to generate a new dimensionality reduction fingerprint data database.Thirdly,we execute the localization algorithm based on BILSTM on the newly generated fingerprint database.Finally,the effect of the proposed algorithm on the target accuracy is analyzed through experiments,and the superiority and effectiveness of the proposed algorithm are verified.