Research On Indoor Positioning Algorithm Based On Crowdsourcing Correction And Data Fusion

With the growth of population and economic development,it is often necessary to obtain accurate location information of users in large public indoor environments such as exhibition halls,shopping malls,and underground parking lots.One of the key issues for location-based services is how to quickly and inexpensively obtain accurate location information of mobile users.However,existing indoor positioning technologies are still imperfect,with problems such as low positioning accuracy.To address these issues,this thesis conducts in-depth research on problems such as low accuracy of single sensor positioning,difficulty in determining data fusion weighting coefficients,and insufficient location feature mining in indoor positioning based on crowdsourcing calibration and data fusion.The experiments are conducted on UJIIndoor Loc,IPIN2017,and real-world environments.Firstly,a multi-source fusion indoor positioning algorithm based on crowdsourcing correction is proposed to address the problems of low positioning accuracy of a single sensor and difficult determination of weighting coefficients in multi-sensor positioning.The algorithm first uses the received signal strength indication of Wi-Fi for coarse positioning of the user,constructs user groups based on the received signal strength indication of Bluetooth,and then distinguishes between high and low error users based on the deviation distance from the user data center.Low error users are used to correct high error users.Finally,to weaken environmental differences,a virtual space unrelated to the real space is constructed,and the spatial distribution characteristics of user groups are used to perform secondary correction on high-error users,ultimately obtaining highly accurate corrected positions.Secondly,to address the problems of difficult determination of data fusion weight coefficients,insufficient mining of positioning features,and low individual user positioning accuracy in non-corrected states in multi-source fusion indoor positioning algorithms based on crowdsourcing correction,this thesis proposes a Wi-Fi-based spatial-temporal fusion dead reckoning indoor positioning algorithm,WDR-Net.The algorithm first uses a convolutional network to extract spatial features from the Wi-Fi signal sequence to obtain the user’s current and historical positions,and then uses a long short-term memory network to extract temporal features to obtain the user’s historical displacement.Finally,to avoid calculating the fusion weight coefficients,the spatial-temporal features are fused using a long short-term memory network to obtain the high-precision fused position.Finally,in order to validate the improvement of the algorithms in terms of positioning accuracy,this thesis designs and implements multiple sets of experiments in two datasets and real environments,and conducts comprehensive analysis on the positioning accuracy and positioning time in the results.In addition,this thesis also proposes a positioning algorithm dataset with multiple distance limitations between reference points based on Wi-Fi received signal strength indicator.