Succesful Build procedure

Unsigned-Long / RIs-Calib

a continuous-time-based multi-radar multi-imu spatiotemporal calibrator

MIT License

27 stars 5 forks source link

Succesful Build procedure #1

Closed TitouanTyack closed 2 weeks ago

TitouanTyack commented 2 weeks ago

Hi,

I succesfully built RIs-Calib, after encountering several issues. So, to share the information of the build procedure that worked in my case on a fresh 20.04 ubuntu install was:

Install ros1 noetic
build from source ceres version 2.2.0
build from source Pangolin version 0.9.2: In ./scripts/install_prerequisites.sh, remove all ‘catch2’ (cf https://github.com/stevenlovegrove/Pangolin/issues/913) i don’t run tests ./scripts/install_prerequisites.sh -m apt all Built with cmake (not ninja), did not run python make
build from source Sohpus version 1.22.10 with SOPHUS_USE_BASIC_LOGGING=ON
build from source magic-enum
build from source spdlog version 1.14.1 (So i did not do sudo apt-get install libspdlog-dev)
sudo apt-get install libcereal-dev
sudo apt-get install libyaml-cpp-dev
clone repos
rosdep install --from-paths src --ignore-src -r -y
catkin_make -DCATKIN_WHITELIST_PACKAGES="ainstein_radar;ti_mmwave_rospkg;serial;sbg_driver"
catkin_make -DCATKIN_WHITELIST_PACKAGES="ainstein_radar_msgs"
catkin_make -DCATKIN_WHITELIST_PACKAGES="ris_calib"

Unsigned-Long commented 2 weeks ago

Thanks for sharing! 👍 As a multi-IMU multi-radar spatiotemporal calibrator, RIs-Calib's existing modules have been fully integrated into iKalibr. So if possible, using iKalibr is a better choice (some small tricks have been added for radar initialization to better recover the initial values of parameters).

iKalibr: https://github.com/Unsigned-Long/iKalibr.git

Tips: Maybe iKalibr is not easy to compile, hahaha 😃 , but I feel that the environment that ikalibr depends on is similar to RIs-Calib. So if RIS-Calib is compiled successfully, you can probably try to compile iKalibr again :muscle:.

TitouanTyack commented 2 weeks ago

Thank you for your answer.

Yes, I'll try again to compile iKalibr at some point, but for now I only need the radar-IMU calibration. I think it will work by compiling colmap from source.

It could be great to propose your work with a Pangolin free implementation, it would makes it way easier to dockerize. Also, I look forward to see a ros2 release ;)

I tried RIs-calib with real data using an ellipse-N sbg IMU and an AWR1843 and it looks like it works great, I have similar results usings several records.

Thanks again

Unsigned-Long commented 2 weeks ago

Thank you for your valuable comments :smiley: . If we have time later, we will remove the dependency on pangolin (replace it with a visualizer like rviz or add an option to enable the gui). Since we are currently extending iKalibr to support event cameras, support for ROS2 may have to be postponed.