EKF-Based Radar-Inertial Odometry with Online Temporal Calibration (EKF-RIO-TC)

This repository contains the implementation of EKF-Based Radar-Inertial Odometry with Online Temporal Calibration (EKF-RIO-TC).
The corresponding paper has been accepted for publication in IEEE Robotics and Automation Letters (RA-L).

Prerequisites

Make sure the following are installed:

• ROS (e.g., Noetic or Melodic)
• catkin_tools
• catkin_simple:

  cd ~/catkin_ws/src
  git clone https://github.com/catkin/catkin_simple.git

Build and Run

# Clone the repository and its submodules
cd ~/catkin_ws/src
git clone https://github.com/spearwin/EKF-RIO-TC.git
cd EKF-RIO-TC
git submodule update --init --recursive

# Build the workspace
cd ~/catkin_ws
catkin build --cmake-args -DCMAKE_BUILD_TYPE=Release

# Source the workspace
source ~/catkin_ws/devel/setup.bash

To launch the system, choose one of the following based on your use case:

• If you are using rosbag playback:

  roslaunch ekf_rio_tc <dataset_name>_rosbag.launch

• If you are using live ROS topics:

  roslaunch ekf_rio_tc <dataset_name>.launch

  or play the bag file manually:

  rosbag play <dataset_path>/<sequence>.bag

Experimental Setup

The experimental setup includes the following components:

• Radar: Texas Instruments AWR1843BOOST.
• IMU: Xsens MTI-670-DK.
• Ground Truth: Provided using an OptiTrack motion capture system.

Sensor Setup

For the Texas Instruments AWR1843BOOST radar sensor, configuration is performed via the mmWave Demo Visualizer. In our self-collected dataset, the key radar parameters are as follows:

• Frame Rate: 10 fps
• Doppler Velocity Resolution: 0.06 m/s
• Range Resolution: 0.04 m
• CFAR/Doppler Range Threshold: 10 dB (this parameter may need adjustment based on environmental conditions)

Datasets

• Self-collected Dataset
  The self-collected dataset consists of a total of 7 sequences, including data from radar, IMU, and ground truth provided by an OptiTrack motion capture system.

  • Sequence 1–3: Involve smoother motion with smaller gyroscope changes over the time offset interval.
  • Sequence 4–7: Exhibit greater variation, leading to a larger radar ego-velocity discrepancy and clearer impact of the time offset.

  The dataset is included in the repository under the following path:

  EKF-RIO-TC/ekf_rio_tc/datasets
  

• Open Datasets
  Two publicly available datasets were used for testing:

  • ICINS2021 Dataset
  • ColoRadar Dataset

Black: Ground truth, Blue: EKF-RIO, Red: EKF-RIO-TC (proposed).

References

• C. Kim, G. Bae, W. Shin, S. Wang and H. Oh, "EKF-Based Radar-Inertial Odometry With Online Temporal Calibration," in IEEE Robotics and Automation Letters, vol. 10, no. 7, pp. 7230-7237, July 2025, DOI Link.

Contact

• Changseung Kim
  Email: [email protected]

Acknowledgement

• This work builds upon the implementation of EKF-RIO.
• The trajectory evaluation in this project was conducted using the evo tool.

License

• GPLv3 License.