Research On Pedestrian Navigation Algorithm Based On Multi-sensor Fusion

In recent years,pedestrian navigation and positioning systems in the global market has a huge market demand,China is also doing its best to promote the development of indoor and outdoor navigation and positioning technology,but most navigation and positioning systems rely on professional equipment support,expensive,and in complex environments(such as in-door,urban canyons)positioning accuracy is poor.So how to find low-cost,mass-applicable technology while ensuring system positioning performance is the question we are now facing.Therefore,the basic starting point of this study is to implement pedestrian navigation positioning using low-cost hardware and to improve its positioning accuracy in harsh environments.At present,intelligent terminals have been popularized and become an important part of people's daily life,its built-in MEMS sensor has the advantages of light weight,small size,low power consumption,low cost,easy integration,etc.,which makes the MEMS sensor-based navigation and positioning technology become an ideal means of pedestrian navigation and po-sitioning.However,when the MEMS sensor-based navigation and positioning system works alone,the positioning error increases rapidly over time and eventually the system does not work properly.The iBeacon has the advantages of low cost,easy installation and anti-interference.Based on these,this study uses a smartphone as a platform to study the fusion positioning algo-rithm in combination with MEMS sensor,GNSS,iBeacon,and map.Around this core objective,this paper has been studied in depth and has produced results in the following areas:(1)Smart terminal MEMS sensor is a consumer-grade product,its price is low,but the noise,low accuracy,poor stability,then how to effectively use MEMS measurements for pedes-trian navigation and positioning has become the primary problem.In this study,a noise reduc-tion algorithm and a simple IMU calibration algorithm were designed based on the charac-teristics of the MEMS sensor,which improved the data quality of the MEMS sensor to some extent(the heading standard deviation from 5.96 degrees to 2.18 degrees).The relative eleva-tion and floor changes were estimated by combining air pressure and temperature,and the effect of pedestrian movement on the relative elevation was analyzed,and the relative elevation and floor determination algorithm was finally optimized.Direction,step counter,and step length are several key elements of pedestrian heading estimation,especially heading,and in this study,we design the step counter and step length algorithm based on accelerometer,magnetometer,and gyroscope measurements,and optimize the MEMS heading algorithm with Kalman fil-tering and complementary filtering algorithm,and the results show that the heading STD of the improved algorithm is reduced by about 30%,confirming the superiority of the improved algorithm.(2)There is a general problem of error accumulation in navigation positioning algorithms based on MEMS sensor observations,especially in the direction of the error will lead to mul-tiplication of the positioning error,therefore,the error needs to be corrected in combination with other information.In this study,iBeacons are classified into strong,medium,and weak types based on their transmitting power;integrated positioning,heading,and step improvement algorithms are constructed based on RSSI of different types of iBeacons,and MEMS heading and iBeacon heading fusion algorithms are designed.Indoor positioning experiments showed that with the addition of iBeacon positioning correction and heading correction,the overall po-sitioning accuracy improved to within 3 meters,effectively correcting the cumulative error of the MEMS sensor algorithm and enhancing the stability and sustainability of the navigation positioning system.(3)Currently,pedestrian navigation and positioning-related research mostly assumes that pedestrians walk in a certain posture or fixed smart terminal posture.And in reality,the way pedestrians carry their smart terminals and their own movement patterns are diversified,and positioning algorithms are closely related to pedestrian activity.The effective,real-time,and accurate determination of pedestrian movement patterns determines the design of precise navi-gation and positioning algorithms.In this study,the recognition model is trained by combining the depth learning technique with the measurements of the smart terminal MEMS sensor,and comparing the traditional machine learning algorithms,it is found that the recognition of pedes-trian activity based on depth learning is not only simple to achieve,but also accurate and can assist the real-time navigation positioning of the mobile terminal.Experiments show that the average positioning error of the PDR+GNSS+iBeacon+AR fusion algorithm is reduced by about 1.1 m,indicating that real-time identification of pedestrian activity improves the naviga-tion positioning effect.(4)Usually the smart terminal GNSS can obtain better positioning results under headroom conditions,but in indoor or urban canyons,the signal will lose its lock or the positioning will be abnormal,which will eventually lead to ineffective positioning or excessive error of the GNSS-dependent pedestrian navigation positioning system.This study integrates the advantages of GNSS positioning,PDR positioning,iBeacon positioning,and indoor maps,and combines the designed EKF algorithm and PF algorithm to achieve fusion positioning.The Hong Kong city canyon positioning experiment showed that the percentage of positioning errors below 10 me-ters of GNSS+PDR+iBeacon fusion positioning result increased from 38%to 60%.The percentage of positioning errors of less than 20 meters has been increased from 55%to 80%.The fusion of GNSS+PDR increased the percentage below 20 meter positioning error to 60%.This shows that fusion positioning is significantly better than single positioning in both indoor and outdoor mixed areas and urban canyons,especially in urban canyons regions,the position-ing accuracy is significantly better than smart terminal's own GNSS positioning accuracy,and it has better continuity.Therefore,it can be considered that the fusion algorithm improves the navigation and positioning capability of the system.