In this paper, we investigate how self-contained pedestrian navigation can be augmented by the use of foot-to-foot visual observations. The main contribution is a measurement model that uses Zero velocity UpdaTe (ZUPT) and relative position measurements between the two shoes obtained from shoe-mounted feature patterns and cameras. This measurement model provides directly the compensation measurements for the three position states and three velocity states of a pedestrian. The involved features for detection are independent of surrounding environments, thus, the proposed system has a constant computational complexity in any context. The performance of the proposed system was compared to a standalone ZUPT method and a relative-distance-aided ZUPT method. Simulation results showed an improvement in accumulated navigation errors by over 90%. Real-world experiments were conducted, exhibiting a maximum improvement of 85% in accumulated errors, verifying validity of the approach.

Pedestrian Inertial Navigation System Augmented by Vision-Based Foot-to-foot Relative Position Measurements

We present a study on the effects of the inertial measurement unit (IMU) noises on the navigation solution uncertainty during zero-velocity-update (ZUPT)-aided pedestrian inertial navigation. The analytically derived attitude, velocity, and position propagation errors reveal that among many IMU noise terms, the dominant factor affecting the accuracy of ZUPT-aided pedestrian navigation is rate random walk of gyroscopes. A numerical simulation has been conducted, showing the discrepancy of less than 10% between the numerical and analytical results, supporting fidelity of the analytical estimates. Experiments have also been conducted, and the results were on the same level of analytical and numerical predictions. This article offers a closed-form analytical prediction for the errors of ZUPT-aided pedestrian inertial navigation due to IMU noises.

Study on Estimation Errors in ZUPT-Aided Pedestrian Inertial Navigation Due to IMU Noises

Presented a customizable laboratory testbed for the sensor fusion of multiple Inertial Measurement Units (IMU) and SOund Navigation And Ranging sensors (SONAR), for self-contained navigation experiments. We described the architecture and communication interfaces for the simultaneous collection of data from two IMUs and two SONARs, however, parallel acquisition from a larger variety of sensors is also feasible. Representative human gait patterns are experimentally acquired to demonstrate the functionality of the testbed.

A Laboratory Testbed for Self-Contained Navigation

Pedestrian navigation has been of high interest in many fields, such as human health monitoring, personal indoor navigation, and localization systems for first responders. Due to the potentially complicated navigation environment, selfcontained types of navigation such as inertial navigation, which do not depend on external signals, are more desirable. Pure inertial navigation, however, suffers from sensor noise and drifts and therefore is not suitable for long-term pedestrian navigation by itself. Zero-velocity update (ZUPT) aiding technique has been developed to limit the navigation error growth, but adaptivity of algorithms, model fidelity, and system robustness have been major a concern if not properly addressed. In this paper, we attempt to establish a common approach to solve the problem of self-contained pedestrian navigation by identifying the critical parts of the algorithm that have a strong influence on the overall performance. We first review approaches to improve the navigation accuracy in each of the critical part of implementation proposed by other groups. Then, we report our results on analytical estimations and experiments illustrating effects of combining inertial sensor calibration, stance phase detection, adaptive model selection, and sensor fusion.

Scenario-Dependent ZUPT-Aided Pedestrian Inertial Navigation with Sensor Fusion

We present a ranging-based aiding method for pedestrian inertial navigation, utilizing not only ranging readouts, but also orientation of the ranging sensors. Both numerical and experimental results demonstrated improvements in navigation accuracy using the method. Kalman Filter (KF) was implemented to merge inertial navigation with Zero-Velocity-Update (ZUPT) algorithm and foot- to- foot directional ranging information. The improvement of navigation accuracy was achieved purely algorithmically without any increase in complexity of the hardware. We demonstrated experimentally that the navigation errors can be improved by about two times using the method.

Directional Ranging for Enhanced Performance of Aided Pedestrian Inertial Navigation

We present a method to identify and compensate systematic errors in the ZUPT-aided pedestrian inertial navigation. We considered two main categories of systematic errors resulting in an underestimate of the length of the trajectory and a drift in the heading of the trajectory. In this study, we identified the dominant factors resulting in the trajectory length and heading errors to be residual velocity during the stance phase and g-sensitivity error of the gyroscopes, respectively. Magnetic motion tracking system was used to record the velocity of the foot during the stance phase. Rate table, tilt table, and shaker were used to calibrate the IMU g-sensitivity. After compensation, a more than $6\times$ systematic error reduction was demonstrated from 3.24m to 0.50m during a 100m straight line trajectory. To the best of our knowledge, this study is the first attempt to reduce the systematic errors in the ZUPT-aided pedestrian inertial navigation algorithmically, without adding extra sensing modalities.

Chi-Shih Jao

Papers

Pedestrian Inertial Navigation System Augmented by Vision-Based Foot-to-foot Relative Position Measurements

A Laboratory Testbed for Self-Contained Navigation

Scenario-Dependent ZUPT-Aided Pedestrian Inertial Navigation with Sensor Fusion

Directional Ranging for Enhanced Performance of Aided Pedestrian Inertial Navigation

Compensation of Systematic Errors in ZUPT-Aided Pedestrian Inertial Navigation