Outlier-RObust RAdar odometry (ORORA)

:bookmark_tabs: About ORORA (IRCA'23)

• A robust radar odometry method in urban environments

NOTE Code & youtube video will be updated until the end of Feb!

• Please refer our paper for detailed explanantions and experimental results!

• Validated on MulRan dataset.

• :bulb: Contents: YouTube

Requirement Version & Test Environment

• CMake version > 3.13

• gcc/g++ > 9.0

• In Ubuntu 18.04, we use

  • Eigen 3.3
  • Boost 1.5.8
  • OpenCV 3.3 (or 3.4)

• In Ubuntu 20.04, we use

  • Eigen 3.3
  • Boost 1.71
  • OpenCV 4.2

How to Build

Just follow the below command:

$ mkdir -p \~/catkin_ws/src/ && cd \~/catkin_ws/src
$ git clone https://github.com/url-kaist/outlier-robust-radar-odometry.git
$ cd ..
$ catkin build orora

TMI: Following the philosophy of target-oriented CMake, PMC is automatically installed when you run catkin build orora

How to Run and Evaluate Radar Odometry in MulRan dataset

1. Set file tree of the MulRan dataset as follows:

${MULRAN_DIR}
_____/KAIST03 
     |___global_pose.csv
     |___/gt (Synchonized GT poses are saved)
         |___... 
     |___polar_oxford_form
         |___1567410201812840928.png
         |___1567410202062436262.png
         |___1567410202312110509.png
         |___...    
_____${OTHER SEQ}
     |...
_____...
   

2. Generate synchronized ground truth poses to the radar data as follows:

$ rosrun orora mulran_generate_gt ${MULRAN_DIR} ${SEQ1} ${SEQ2}...
// e.g.
$ rosrun orora mulran_generate_gt /media/hyungtaelim/UX960NVMe/mulran KAIST03

3. Then, set right seq_dir in launch/run_orora.launch & run the below command

$ roslaunch orora run_orora.launch

4. Run script/evaluate_odometry.py as follows:

// E.g.
$ python evaluate_odometry.py -f /media/hyungtaelim/UX960NVMe/mulran/KAIST03/outputs/mulran_ORORA_cen2018_0.6_0.75_0.1_0.15708eval_odom.txt

Acknowledgement

Many thanks to Keenan Burnett to provide outstanding radar odometry codes!

Please refer to Yeti-Radar-Odometry for more information

:mailbox: Contact Information

If you have any questions, please do not hesitate to contact us

• Hyungtae Lim :envelope: shapelim at kaist.ac.kr