ydsf16/imu_gps_localization
Using error-state Kalman filter to fuse the IMU and GPS data for localization.
📡 IMU & GPS Localization – EKF Fusion
What it does – This repository implements an Error‑State Kalman Filter (EKF) that fuses inertial‑measurement‑unit (IMU) data with Global‑Positioning‑System (GPS) measurements to produce a globally‑consistent pose (position + orientation) for a vehicle or robot. The implementation follows the textbook “Quaternion‑based Kinematics for the Error‑State Kalman Filter”.
Why it matters – Pure GPS gives accurate position but drifts in orientation and suffers from outages; pure IMU gives high‑rate motion estimates but quickly drifts. By combining the two with an EKF you get:
- High‑frequency, smooth pose estimates
- Robustness to GPS drop‑outs
- Accurate 3‑D orientation using quaternion math
Key components
- EKF core – prediction step integrates IMU angular‑velocity and linear‑acceleration, maintaining a quaternion‑based error state.
- Measurement update – GPS fixes the position part of the state when a new fix arrives.
- Dataset support – The code is tested on the UTBM Robocar dataset (link provided).
- Visualization – A sample trajectory plot (
doc/path.png) shows the fused path.
Who it’s for – Robotics or autonomous‑vehicle developers who need a ready‑to‑run EKF implementation for sensor fusion, and who prefer a reference that follows a published academic treatment.
How to get started
- Clone the repo.
- Install any required Python/C++ dependencies (the README does not list them, so check the code or
requirements.txt). - Download the UTBM Robocar dataset.
- Run the main script (likely something like
python run_ekf.py) to see the fused trajectory.
License & contact – GPL‑v3. For commercial questions, contact the author via WeChat or email; issues can be opened on GitHub.
All information above is taken directly from the repository’s README.
Related
- Project
- Project
- Project
- Project