| With the rapid development of communication network technology and the increasing popularity of mobile intelligent terminals,such as smart phones and smart bracelets,the demand for obtaining indoor and outdoor seamless positions is increasing.Location Based Service(LBS)is gradually changing the way people live and work,such as location and navigation,social networking,precision advertising,public safety and incident handling.Among them,70%~90% of the LBS required by mass users are indoor scene,while the Global Navigation Satellite System(GNSS)in the outdoor environment has basically met the needs of users,so reliable indoor positioning technology has become the main bottleneck for the demand of the seamless location services.This paper proposes an indoor multi-source fusion positioning scheme based on the builtin sensor of the smartphone,the environmental magnetic field,low-power Bluetooth and other auxiliary positioning information.The focus is on enhancing the self-estimation ability of the PDR algorithm,and using PDR to assist in improving the performance of other positioning methods and the final multi-source fusion.The following summarizes the research focus and results from the three aspects of the robust strapdown PDR,the magnetic field matching and the multi-source fusion positioning method:1.For the pedestrian positioning and navigation application,the measurements from the builtin of the smartphone and the various constraints are used to produce the update information of the filter for maintaining the navigation performance of the strapdown PDR,while retaining the autonomous estimation capability of the strapdown inertial navigation algorithm.The experimental results show that the strapdown PDR algorithm can achieve the same positioning accuracy compared with the traditional step-model PDR in the modes of texting,calling and pocket,and has better navigation performance in the swinging mode,such as the heading estimation error and the position estimation error was reduced by 47.5%(RMS)and 37.34%(RMS),respectively.At the same time,the strapdown PDR retains the inherent advantages of the strapdown inertial navigation algorithm,which can maintain positioning accuracy when a small number of steps(such as 1~5 steps)are missing.Such as two steps are missing every ten steps,the average distance error of the traditional footmodel PDR in the four basic modes increases from 1.2% to 18.69%,but the distance average error of the strapdown PDR only increases from 1.35% to 3.07%.In addition,the misalign angle between the heading of mobile phone and the walking direction of users caused by the user's habits seriously affects the positioning performance of the strapdown PDR algorithm.This paper designs a new method to estimate the misalign angle based on the moving speed maintained by the pure inertial navigation algorithm.The results of several tests show that the proposed method can effectively detect the change of the misalign angle and instantly estimate the new misalign angle,thereby basically solving the problem that the misalign angle destroys the availability of the PDR in actual application.More,based on the basic strap-down PDR,the spatial structure feature of typical indoor environment(such as office buildings)— the straight corridors are parallel or perpendicular to each other is used to control the heading error.The heading database is constructed using the current estimated heading information,and is conditionally used for correcting the cumulative error of the heading angle in the future.The results of 12 tests in the four basic modes show that the method can fully utilize the orientation information of the building structure to correct the heading divergence of the PDR without pre-establishing the building heading database,thus effectively improving the navigation performance of the strapdown PDR.2.As the low efficiency of the fingerprint data collection and the poor accuracy of the reference points in the fingerprint database construction phase(such as magnetic field signature library),a higher-efficiency and higher-precision positioning fingerprint collection scheme based on foot-mounted PDR is designed and implemented.In this paper,the self-developed inertial sensors array combined with a small number of calibration points are used for providing the accurate reference coordinate sequences.Which brings the following results that the number of calibration points required is small and without the special requirements for the testers' walking trajectory(e.g.,uniform linear motion).Therefore,the data collection process is more simplified,the probability of testers making mistakes during data collection is reduced,and the efficiency of positioning fingerprint data collection is improved.In addition,for the phenomenon of low spatial resolution of the environmental magnetic field,a strap-down PDR-assisted magnetic field contour feature matching algorithm is designed.The trajectory contour generated by strapdown PDR in real time is used to increase the spatial topological property of the magnetic field sequence,which improves the distinguishability of the basic unit of magnetic field matching.Based on this,a fast magnetic field matching algorithm based on the Gauss-Newton iterative method is derived.The experimental results show that the strapdown PDR-assisted magnetic field matching method can provide stable and continuous positioning results in a variety of environments.The RMS values of the positioning errors in the three typical public scenes of office buildings,library halls and shopping malls are 0.49 m,1.59 m,2.10 m,and the matching success rates are 94.67%,90.24% and 88.01%,respectively.More,the test results of four mobile phones(Huawei P20,Samsung S6,S7 and Google Pixel 2)show that the magnetic field matching algorithm proposed by this paper achieve an average positioning accuracy of 0.77 m and an average matching success rate of 91.96%.3.In order to effectively eliminate the gross error of the observation,a multi-source fusion positioning method based on relative trajectory-assisted optimization is designed.In this paper,the high-precision relative trajectory derived from PDR and the original observations of other localization methods in a period of time are optimized to obtain stable and reliable observation information,and then the observation information is input into Kalman filter to enhance the stability of the filter.The experimental results show that the positioning performance of the proposed method is improved by 19.03% compared with the basic Kalman filter.In addition,for the heavy task of the positioning fingerprint database establishment and maintenance,this paper designs an offline trajectory recovery algorithm based on natural sparse landmarks correction inspired by the idea of updating positioning database based on the crowding source data.The method mainly utilizes the continuity of the robust strapdown PDR combined with a small number of discrete high-precision and high-reliability landmarks' corrections,and adopts a post-processing reverse smoothing algorithm to reliably estimate the user's high-precision position trajectory.The experimental results show that the positioning accuracy of this "FM-PDR+Sparse Landmarks" smoothing algorithm is increased by 67.0%(RMS)compared with the traditional filtering method on average.In addition,the calculation results of the public datasets of two international indoor positioning competitions(organized by the US National Institute of Standards and Technology and the IPIN Conference Organizing Committee)in 2018 are also given,further explaining the calculate precision and advantage of the offline user trajectory recovery algorithm proposed in this paper.Based on the above key technologies,a complete pedestrian indoor positioning solution has been formed.The solution basically meets the need of consumer indoor positioning applications,and has the characteristics of high reliability,strong scalability,good real-time performance and low cost.The research works in this paper provide a feasible reference case for solving the indoor positioning problem of consumer markets. |
PDF Full Text Request