Research On Indoor Positioning Method With Smartphone Based On WiFi And Inertial Sensors

With the rapid growth of the public's demand for indoor Location-based Services(LBS),indoor positioning technology has become a research hotspot in the field of navigation and LBS,and has been applied in various industries.The smartphone has become an excellent indoor positioning platform due to its built-in large number of sensors that can sense motion and environmental information,as well as its high popularity and portability,among which Wi Fi fingerprint positioning,pedestrian dead reckoning(PDR)positioning and Wi Fi/PDR fusion positioning based on smartphone become the mainstream indoor positioning technology,and have attracted the attention of many scholars and the LBS providers.Due to the complex and variable indoor environment,low-precision sensors,limited computing power of mobile devices and high randomness of pedestrian movements,there are still many technical problems to be solved in achieving high accuracy,high continuity and high stability of Wi Fi fingerprint positioning,PDR positioning and Wi Fi/PDR fusion positioning.In view of the current problems of indoor positioning technologies,relying on the actual positioning environment and application scenarios,combining the current positioning performance requirements for LBS,this thesis researches the indoor pedestrian positioning method with smartphone based on Wi Fi signal and inertial sensors.The main research work of the thesis is as follows:First,to address the problem that received signal strength(RSS)similarity and position similarity are inconsistent in the existing Wi Fi fingerprint positioning,and that the resulting fingerprint clustering results cannot reflect the position relationship between reference points and ensure the accuracy of online cluster matching,as well as the poor positioning accuracy of the RSS Euclidean distance-based fingerprint positioning algorithm,a Wi Fi fingerprint positioning method based on RSS and position similarity has been proposed.The location label information is introduced into the offline fingerprint clustering,forming an RSS fingerprint clustering algorithm with location label constraints,so that the clustering results can reflect the spatial distribution and signal relationships among reference points and effectively improve the accuracy of online cluster matching;The uneven spatial resolution of RSS is analyzed,and the numerical relationship between position distance and RSS similarity between points is deeply explored for the actual positioning environment,and an improved weighted k-nearest neighbor algorithm based on approximate position distance is proposed.The experimental results showed that the proposed Wi Fi fingerprint positioning method effectively improved the offline fingerprint clustering effect and online fingerprint positioning accuracy.Secondly,an improved PDR positioning method in multi-motion mode is proposed to address the problems of poor detection accuracy,large step length estimation error,difficult pedestrian azimuth estimation and poor positioning accuracy due to the large differences in sensor output signals under different pedestrian movement states and phone poses,and the relative motion between smartphone and pedestrian during the positioning process.The time-domain features and frequency-domain features of the sensor data are extracted,the performance of the feature data under different motion modes is analyzed,and a combination of support vector machine and decision tree is used to achieve accurate recognition of the movement state and phone pose in any combination;By analyzing the change in acceleration values under different modes,a step detection algorithm and step length estimation model are proposed;An improved principal component analysis azimuth estimation method based on the gait characteristics is proposed by deeply mining the correlation between acceleration data,device attitude and pedestrian gait,which minimizes the problem of inaccurate extraction of the horizontal component of acceleration that exists in existing azimuth estimation algorithms and the problem of cumulative positioning error caused by time integration of sensor data,and is more suitable for smartphone platforms with low sensor accuracy.The proposed methods of step detection,step length estimation and azimuth estimation together constitute the PDR positioning system in the multi-motion mode.The experimental results show that the proposed PDR positioning method can achieve high positioning accuracy in different motion modes,extending the applicability of the existing PDR technology.Thirdly,a Wi Fi/PDR fusion positioning method based on distance constraint and RSS estimation is proposed to address the problems of positioning accuracy and positioning continuity of the Wi Fi/PDR system,such as position hopping,position aggregation,RSS update delay and RSS inaccuracy under dynamic tracking condition.The filter divergence caused by inaccurate system model and noise statistics is suppressed by using extended Kalman filtering based on asymptotic memory;The constraint relationship between the system position output results is analyzed in conjunction with pedestrian gait characteristics,and a fusion positioning strategy based on distance constraints is proposed;Based on the idea of weighted k-nearest neighbor algorithm,the online RSS is estimated using one-step predicted location and reference point fingerprint information,which solves the problem of inaccurate online RSS measurements caused by too short data acquisition time and improves the validity of the measured information in the filtering process.The experimental results show that the proposed Wi Fi/PDR fusion method has significantly weakened the position point hopping phenomenon of the positioning trajectory,with higher positioning accuracy and positioning continuity.The indoor positioning method researched in this thesis does not rely on additional positioning facilities and high-performance positioning devices,and only uses existing smartphone platforms and Wi Fi access points to achieve indoor pedestrian positioning in different action modes,which can provide important technical support for the promotion and application of indoor LBS.