Research On The Key Technologies Of Pedestrian Autonomous Positioning Using Wearable Inertial Sensors

The requirements for accurate and reliable pedestrian positioning in specific applications,such as emergency rescue and special operations,are becoming increasingly urgent.Wearable inertial sensor-based autonomous reckoning positioning(WIS-ARP)is an irreplaceable core technology in pedestrian positioning,because it does not rely on infrastructure and has the advantages of being environment-independent and fully selfcontained.However,the traditional WIS-ARP method faces the issues of insufficient accuracy and low robustness.This thesis enhances pedestrian autonomous positioning performance by using constraint information such as pedestrian motion patterns,human joint constraints,and positioning environment characteristics.The key study in this thesis follows the track from theory to practice.As a result,this thesis has developed relevant significant techniques for pedestrian positioning using single-foot inertial sensors(SFIS),dual-foot inertial sensors(DFIS),and lower-limb multi-inertial sensors(LLMIS).The main research works and contributions of this thesis include the following:(1)In order to solve the problem of poor stability of the SFIS method,this thesis proposes a more robust real-time SFIS positioning method consisting of improved zerovelocity detection and elevator(escalator)scenario constraint algorithms.The method uses a ”coarse-fine”two-step criterion to determine the zero-velocity interval accurately.First,the two key moments corresponding to heel-strike and heel-off the ground are roughly determined by increasing the absolute zero-velocity detection threshold.Then,the interval with low fluctuations is filtered out using a relatively small threshold.Furthermore,the proposed method separates the positioning applications into the elevator(escalator)and normal walking scenarios through inertial sensor signals.Then a uniform velocity constraint model method is applied to enhance pedestrian positioning performance according to the elevator and escalator operation features.The experimental results show that the proposed method effectively improves the stability of SFIS pedestrian positioning in different walking modes.The proposed method is also validated with the public dataset of the IPIN-2020 international indoor positioning competition(large shopping mall scenario)and obtains the best positioning accuracy in all the competition schemes.In order to meet the demand for post-processed accurate estimation of user trajectories in SFIS applications(e.g.,indoor map acquisition),we propose a macroscopic building orientation-assisted SFIS positioning method in post-processing mode.In the single-floor user trajectory estimation,the position constraint of a natural position revisit point(e.g.,elevator and fire stair)and smoothing algorithms are used to improve the trajectory estimation accuracy.Furthermore,a trajectory orientation rectangular fitting method is used to extract the macroscopic multistory building orientation.Then the method matches the pedestrian trajectories of all floors with the fitted building orientation rectangle.The proposed method solves the trajectory orientation drift problem in the SFIS system effectively.Thus,the proposed method can obtain an accurate pedestrian trajectory without error divergence in multistory building scenarios.(2)In order to solve the unreasonable distance constraint problem in a DFIS system without the range sensor,this thesis proposes a minimum distance nonlinear constraint method for the DFIS system.The minimum distance constraint problem is solved using nonlinear iterative optimization.First,the distance constraint is converted to a linear constraint model by iterative update.Then,an accurate state and covariance under the distance constraint condition are estimated using the estimating projection algorithm and iterative update process.Furthermore,the detection mechanism for the dual-foot minimum distance moment is optimized in this thesis.Based on a large-scale statistical dataset,60% point of the zero-velocity interval is determined as the dual-foot minimum distance moment.Such a mechanism is not affected by the positioning state estimation error and therefore has good stability.Simulation results show that the proposed nonlinear constrained method has higher state estimation performance and lower computational complexity than the conventional methods.We have conducted experimental tests under rigorous scenarios with a walking trajectory of about 1000 m without turn-around and close-loop.The experimental results show that the positioning error of the proposed method is less than 1.5% of the walking distance,which is83.5%,70.0%,and 62.9% less than that of the conventional SFIS,the maximum distance constraint-based DFIS,and the minimum distance constraint-based DFIS methods,respectively.Furthermore,compared with the traditional methods,the proposed method makes fuller use of the minimum distance constraint information and can achieve higher accuracy and more robust autonomous positioning.(3)In order to reduce the dependence of WIS pedestrian positioning on motion assumptions and enhance its positioning accuracy and feasibility,this thesis proposes a human lower limb-based multi-inertial cooperative positioning method.The vector position and velocity constraints are constructed using the human lower limb structural model and an accurate vector projection in the proposed method.A generic multiinertial cooperative algorithm architecture is designed,and the mathematical models with two configurations are built,i.e.,using three inertial sensors(dual-foot and midwaist)and seven inertial sensors(dual-foot,dual-shank,dual-thigh,and mid-waist).The three-inertial-sensor positioning method uses the law that the foot periodically coincides with the mid-waist in the horizontal position.The seven-inertial-sensor positioning method constructs strict mathematical constraints using the human lower limb structural model,which does not rely on usual walking motion assumptions.We have conducted experimental tests under rigorous scenarios with a walking trajectory of about2000 m without turn-around and close-loop.The experimental results show that the positioning errors of the designed two modes of multi-inertial cooperative methods reduce by 25.5% and 48.4%,respectively,compared to the DFIS positioning method.More importantly,the cooperative positioning method with seven-inertial-sensor mode does not rely on any constraint or assumption on the pedestrian’s motion,which therefore has the advantages of natural robustness and applicability and can significantly improve pedestrian autonomous positioning performance.In summary,this thesis presents a comprehensive and in-depth study of the key technologies of wearable inertial sensor-based pedestrian autonomous positioning.The proposed key technologies and methods are comprehensively evaluated and analyzed through simulations and experimental tests.The outcomes of this thesis provide a reliable theoretical basis and feasible reference solution for promoting the applications of wearable pedestrian autonomous positioning solutions in complex scenarios.