ComP

cyber/, ros/ and ros2/ contain code for different environment(middleware). It is recommended that the benchmark be run with root priviledge inside docker container. Detail guidances are provided in these directories. apps contains detect algorithms for tracking and line detecting. You need to recompile them in cyber & ros docker container. It is managed by cmake.

Libraries needed:

• opencv3.4.10, opencv_contrib 3.4.10, and cv_bridge is needed in ROS
• darknet(Yolo)
  • clone darknet and compile
  • cp libdarknet.so to usr/lib or /lib

Tips for users

• opencv_contrib install commands: cmake -D CMAKE_BUILD_TYPE=Release -D CMAKE_INSTALL_PREFIX=/usr/local -D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib/modules -DBUILD_opencv_ximgproc=OFF -DBUILD_opencv_xfeatures2d=OFF -DBUILD_opencv_waldboost_detector=OFF ..
• Understand Cyber in Chinese language
• For Yolo, check the darknet we provide. We only add two functions and you can find them in include/darknet.h (init_result, detect_result).
• All three middleware containers can be downloaded from their official websites
• Please open an issue if you encounter any trouble

ComP was our first try to build an usable benchmark pack for autonomous driving vehicles and the test scenario is simple. We are working on a future version of ComP, any advice is welcomed. The paper was published in RA-L in 2021.

Bibtex:

@article{wu2021oops,
  title={Oops! It's Too Late. Your Autonomous Driving System Needs a Faster Middleware},
  author={Wu, Tianze and Wu, Baofu and Wang, Sa and Liu, Liangkai and Liu, Shaoshan and Bao, Yungang and Shi, Weisong},
  journal={IEEE Robotics and Automation Letters},
  volume={6},
  number={4},
  pages={7301--7308},
  year={2021},
  publisher={IEEE}
}