DCRML (Doppler compensated Radar Metric Localization)

This repo is an implementation for Accurate Automotive Radar Based Metric Localization with Explicit Doppler Compensation, which will be presented in ICRA’2022 !

@article{gao2021accurate,
  title={Accurate Automotive Radar Based Metric Localization with Explicit Doppler Compensation},
  author={Gao, Pengen and Zhang, Shengkai and Wang, Wei and Lu, Chris Xiaoxuan},
  booktitle={2022 IEEE International Conference on Robotics and Automation (ICRA)},
  year={2022},
  organization={IEEE}  
}

:octocat: Metric localization performance

demo

⛵ Build instructions

Dependencies:

Eigen 3.3
OpenCV 3.4
Ceres 1.14
Sophus 1.0

Note: this code has been build and tested on Ubuntu 18.04, but it should compile for newer versions of Ubuntu.

? Dataset

You may need to download nuScenes dataset (radar) from nutonomy/nuscenes-devkit.

? Examples

  1. preprocess Nuscenes dataset to get loop closure cases and compensate for doppler effect in each radar submap.

cd ./preprocess
python search_for_loopclosure.py PATH_TO_NUSCENES
python list_seq.py
  1. metric localization process

cd ~/DCRML
mkdir build
cd ./build
cmake ../DCRML
make
./DCRML

? video demo

Youtube demo

GitHub

View Github