Study On Indoor Positioning Of Smartphone Based On Wi-Fi And PDR

Since most indoor environments are currently equipped with Access Points(APs),Wi-Fi-based indoor positioning technology eliminates the need to install specific positioning devices to provide indoor positioning,making it cost-effective and suitable for commercial use.In addition,Pedestrian Dead Reckoning(PDR),which is based on the built-in accelerometer,magnetometer,and gyroscope of cell phones,can achieve autonomous navigation and positioning by avoiding the problem of installing equipment support,and is therefore a cost-effective indoor positioning technology.However,Wi-Fi signals are highly fluctuating,which can have a large impact on the positioning accuracy;while PDR suffers from error accumulation,and the positioning accuracy of a single positioning technique is lower compared to that of multi-source fusion positioning techniques;in order to deeply investigate the indoor positioning techniques of Wi-Fi,PDR,and Wi-Fi/PDR fusion,the main research contributions of this thesis are as follows:(1)The fluctuation of Received Signal Strength Indication(RSSI)is revealed,and the effective information in the RSSI vector is extracted using Z-Score normalized translation and scaling transformations.Then,based on the transformed RSSI vectors,an RSSI ranging model is constructed by a Genetic Algorithm(GA)to optimize the Back Propagation(BP)neural network,and the proposed RSSI ranging model is named GTBPD ranging model for the convenience of the full thesis.The experimental results conducted in three different experimental environments show that the mean range accuracy of the proposed GTBPD ranging model is about 2m,which has a good range accuracy and can be well adapted to different indoor environments for ranging.(2)Based on the proposed GTBPD ranging model,this thesis designs a new WiFi optimized positioning algorithm named GTBPD-LSQP positioning algorithm by combining linear least squares(LS)and Sequential Quadratic Programming(SQP)algorithms.The experimental results in three different indoor scenes show that the localization accuracy of the proposed GTBPD-LSQP algorithm is 0.921 m,1.128 m and1.038 m higher than that of the traditional Weighted K-Nearest Neighbor(WKNN)algorithm,indicating that our proposed localization algorithm has high localization accuracy and good robustness.(3)A heading estimation method that integrates map corners information is proposed;the method integrates four data sources: accelerometer data,magnetometer data,gyroscope data and data from map corners information.Since there is an obvious error accretion in the heading estimation part of PDR,and the irregular magnetic field in the indoor environment affects the output of magnetometer data,thus affecting the accuracy of heading estimation,this thesis defines some corners according to the structure of the indoor building in advance,and uses the defined corners information to constrain the heading estimation to improve the accuracy of heading estimation.The experimental results in two floors of a building show that the heading estimation method proposed in this thesis can effectively eliminate the effect of error cumulation and improve the accuracy of heading estimation by more than 15% compared with the three existing heading estimation methods.(4)An integrated Wi-Fi/PDR positioning algorithm based on map information constraint is proposed,and the proposed GTBPD-LSQP positioning results are used as the results of the integrated positioning Wi-Fi part,and the heading estimation results of integrated map corners information are used to constrain the PDR.In addition,this thesis not only considers the constraints of map wall line information when performing integrated Wi-Fi/PDR positioning,but also the defined corners information is also introduced for constraint,which further optimizes the integrated positioning results.The experimental results at two experimental fields show that the mean positioning accuracy of the integrated positioning algorithm proposed in this thesis is 1.095 m and1.348 m,and the probability of positioning error within 3m is as high as 97.9% and94.7%,which is significantly better than the existing fusion positioning algorithm.Our thesis contains 42 figures,16 tables,and 102 references.