A Study on a High-Precision 3D Position Estimation Technique Using Only an IMU in a GNSS Shadow Zone

Area of Science:

• Robotics and Autonomous Systems
• Sensor Fusion and Navigation
• Human Motion Analysis

Background:

• Global Navigation Satellite System (GNSS) limitations in indoor or obstructed environments necessitate alternative positioning methods.
• Inertial Measurement Units (IMUs) offer a self-contained solution but suffer from cumulative errors like heading drift and stride inaccuracies.
• Accurate 3D trajectory reconstruction requires robust gait recognition and precise displacement estimation.

Purpose of the Study:

• To develop and validate a continuous, accurate 3D path estimation method using a single nine-axis inertial sensor.
• To address challenges of heading drift, stride error, and gait uncertainty in GNSS-denied environments.
• To achieve low computational cost trajectory reconstruction without external aiding.

Main Methods:

• Implementing stationary state detection to halt updates and suppress error propagation.
• Classifying gait modes (walking, stair ascent/descent) using vertical acceleration and dynamic thresholds.
• Estimating vertical displacement via gait pattern and posture, and planar displacement via adaptive stride length.
• Deriving heading from an attitude matrix aligned with magnetic north for unified frame projection.

Main Results:

• Achieved planar path errors below 3% for 100-meter trajectories.
• Demonstrated vertical error below 2% in stair environments up to ten stories.
• Maintained stable heading estimation throughout experiments.
• Showcased reliable gait recognition and continuous 3D trajectory reconstruction.

Conclusions:

• The proposed method provides a robust and low-cost solution for 3D trajectory reconstruction using only an inertial sensor.
• It effectively mitigates common error sources in IMU-based navigation, particularly in challenging terrains like stairs.
• The system offers reliable performance without external support, suitable for GNSS-denied applications.