Research On Highly Available Localization Method Based On Inertial MEMS/Wi-Fi Intelligent Fusion

With the rapid development of the mobile Internet and the rise of the Internet of Things technology,the development of location-based services technology and industry has given rise to such industry user needs as intelligent recreation,elderly guardianship,security monitoring,emergency rescue,smart cities,and such popular application markets as urban express,intelligent travel,super shopping navigation,intelligent car search,large airport station navigation,and so on.With the popularity of mobile intelligent terminals,especially smart phones,the demand for location services around intelligent terminal devices is also expanding.In addition,along with the increasing innovation of sensor technology,the performance of abundant built-in sensors in smart terminal devices is constantly improved,and the rapid development of technologies such as Internet of Things,cloud computing and artificial intelligence,the highly available and high-precision positioning system based on smart terminal devices has ushered in new development opportunities and application prospects.In this paper,the research objective is to achieve a highly available positioning system without any additional sensor hardware facilities by using the highly available positioning technology of inertial MEMS and Wi-Fi intelligent fusion.Based on the research status of indoor positioning technology at home and abroad,we investigate the problems and technical difficulties faced by the current inertial MEMS and Wi-Fi based highly available positioning technology,and achieve research results with certain theoretical significance and practical value.Specifically,the main work of this paper includes the following:(1)Heading error is an important error source affecting the inertial MEMS-based pedestrian dead-reckoning(PDR)localization technique,and accurate heading estimation is the key to improve the accuracy of PDR estimation.Therefore,this paper proposes an improved heading estimation method based on cubic Kalman filtering to improve the usability of the PDR algorithm in practical use by addressing the problems that the heading of the traditional attitude estimation method is susceptible to magnetic field interference and unconventional pose interference of the device.The conventional PDR outputs the position in steps as frequency,which has a low degree of continuity in time and fails to effectively restore the continuously changing process of pedestrian walking.In this paper,we propose a fine-grained PDR gait segmentation algorithm based on the sine and cosine function approximation to achieve a highly fine-grained PDR location output as a way to extend the usability of the PDR algorithm under highfrequency location update requirements.The effectiveness of the proposed algorithm is verified by the measured data.(2)Wi-Fi fingerprint positioning is an important technical means of positioning based on Wi-Fi signals.Crowdsourcing database construction is an effective means to alleviate the problem that traditional fingerprint database construction methods are very time-consuming and laborious,and to improve the usability of fingerprint localization.In this paper,we investigate the automatic construction of crowdsourced inertial MEMS and Wi-Fi fingerprint databases and their fusion localization methods.For the trajectory optimization and smoothing algorithms in the automatic construction of crowdsourced databases,we propose an improved trajectory optimization algorithm based on LM and a trajectory smoothing algorithm based on fast forward-backward smoothing,which achieve significant improvements in stability and computational efficiency,respectively.On this basis,three offline crowdsourcing databases are constructed in this paper:inertial MEMS database,Wi-Fi fingerprint database and corner trajectory geomagnetic database for indoor fusion localization.Further,this paper proposes a fusion localization method based on PDR,Wi-Fi fingerprint and corner trajectory geomagnetic fusion,using the corner trajectory geomagnetic matching algorithm to achieve further improvement of the fusion localization results of PDR and Wi-Fi fingerprints.The traditional model matching-based Wi-Fi fingerprint localization method has problems such as low accuracy,poor stability and low matching efficiency,while the traditional PDR algorithm has problems such as the heading and walking direction are difficult to keep consistent and the granularity of position update is low.To this end,this paper proposes a data-model jointly driven inertial MEMS velocity estimation network model and a data-driven Wi-Fi fingerprint positioning network model,respectively.Then a data-model jointly driven positioning system based on the inertial MEMS/Wi-Fi fingerprint intelligent fusion is proposed,which achieves an effective improvement of the positioning accuracy and thus provides a technical foundation for highly available indoor positioning solution.The experimental results on three publicly available datasets(IPIN2018,IPIN2019 and IPIN2020)show that the proposed positioning system can achieve highly available,highly continuous and highly accurate indoor localization in large indoor scenarios.(3)Crowdsourced inertial MEMS/Wi-Fi fingerprint data fusion localization solves the indoor localization availability problem in large-scale scenarios;however,its localization accuracy is very limited and cannot meet the localization demand in some complex scenarios.In this paper,we study the fusion localization method of inertial MEMS and single Wi-Fi FTM ranging base station to realize the expansion of crowdsourced Wi-Fi fingerprint localization without adding additional communication base stations,so as to achieve a more highly accurate indoor localization system.To this end,this paper first proposes an EKF-based loosely coupled positioning method of inertial MEMS,Wi-Fi RSSI and Wi-Fi FTM ranging,which achieves higher positioning accuracy than Wi-Fi fingerprint positioning.This paper further proposes an EKF-based tightly coupled positioning method of inertial MEMS,Wi-Fi RSSI and Wi-Fi FTM ranging,which achieves higher modeling accuracy and significantly higher positioning accuracy compared to the loosely coupled method.The traditional Wi-Fi ranging-based positioning method generally requires more than four base stations at the same time,however,in practice,Wi-Fi FTM base stations with known locations are very rare,and sometimes the initial position error provided by Wi-Fi fingerprint positioning fluctuates greatly or is unusable.In this situation,this paper proposes an absolute starting position estimation method based on the fusion of inertial MEMS and single Wi-Fi FTM ranging base station,thus further enhancing the usability of the positioning system.Experimental results show that the proposed single-base station fusion positioning algorithm can effectively achieve high-precision absolute position estimation.