Pedestrian Tracking with Single/Multi IMU Based on Shoe-mounted Inertial Sensors

Abstract

The main objective of this dissertation is to improve the performance of single/multi IMU based PDR (Pedestrian Dead Reckoning) system based on shoe-mounted IMU. Three algorithms are proposed in this dissertation such as [pedestrian dead reckoning using shoe-mounted IMU], and [kinematic model based PDR for heading correction and lower body motion tracking]. First, algorithms for PDR using foot mounted IMU are proposed. The proposed algorithm is consisted with three algorithms. An advanced stance-phased detection and a step length estimation based on a linearly calibrated Zero velocity Update (ZUPT) and map assisted PDR are proposed. The proposed algorithm works with various movements such as walking, crawling, sideways stepping, and climbing up and down stairs. A single inertial measurement unit imbedded at the subject`s right foot is used. Modified signals to detect stance phase in various motions are proposed and results are provided. Also, linearly calibrated ZUPT algorithm is proposed for increasing step length estimation accuracy in various movement. The algorithm reduces effect of accelerometer bias due to sensor and unstable movement in stance phase. Link-node based map information helps correcting pedestrian`s heading. An Extended Kalman Filter (EKF) is used for fusing the information and estimating pedestrian position and sensor errors. Furthermore, to help complicacy in hall map structures, we propose a Virtual Link (VL) algorithm combined with Virtual Track (VT) algorithm. Experimental results show that the proposed algorithm increase the accuracy of the step length estimation in various movement. Finally, we present a method for finding enhanced heading and position of pedestrian by fusing the ZUPT based PDR and kinematic constraint of lower human body. For integrating these information, the kinematic model of lower human body, which are calculated by using orientation sensors mounted on both thighs and calves, is adopted. Notice that the position of left and right foot cannot be apart because of kinematic constraints of body, the kinematic model generates new measurements for the waist position. The EKF (Extended Kalman Filter) on the waist which estimates and corrects error states uses these measurements and magnetic heading measurements, which enhances heading accuracy. The updated position information is fed into the foot mounted sensors, and reupdate processes are performed to correct the position error of each foot. The proposed update-reupdate technique ensures improved observability of error states and position accuracy consequently. Moreover, the proposed method provides all the information of lower human body, so that it can be applied to motion tracking area more effectively. The effectiveness of the proposed algorithm is verified via experimental results, which shows 1.25% of RPE (Return Position Error) with respect to walking distance is achieved.