IEEE UAV Competition 2022 - link1, link2

• IEEE UAV Competition 2022 - Low Power Computer Vision Challenges (LPCVC): Chase
• E. Lee, D. Lee, H. Lim, S. Song, and H. Myung, "Non-uniform motion target tracking system for UAVs," Korea Robotics Society Conference (KRoC), 2022.
• Video - https://youtu.be/zObqq5_M4UA

Team: Hercules

Members: EungChang Lee, Dongkyu Lee, HyungTae Lim, Seungwon Song

Testing setup

• CPU: i9-10900K, GPU: RTX3080
• CUDA version 11.5.119
• cuDNN 8.3.1 for CUDA 11.5
• OpenCV version 4.5.2 with CUDA enabled (not ROS default version) - reference
  • with the options below

$ cmake -D CMAKE_BUILD_TYPE=RELEASE \
      -D CMAKE_C_COMPILER=gcc-6 \
      -D CMAKE_CXX_COMPILER=g++-6 \
      -D CMAKE_INSTALL_PREFIX=/usr/local \
      -D OPENCV_GENERATE_PKGCONFIG=YES \
      -D WITH_CUDA=ON \
      -D OPENCV_DNN_CUDA=ON \
      -D WITH_CUDNN=ON \
      -D CUDA_ARCH_BIN=8.6 \
      -D CUDA_ARCH_PTX="" \
      -D ENABLE_FAST_MATH=ON \
      -D CUDA_FAST_MATH=ON \
      -D WITH_CUBLAS=ON \
      -D WITH_LIBV4L=ON \
      -D WITH_GSTREAMER=ON \
      -D WITH_GSTREAMER_0_10=OFF \
      -D WITH_QT=ON \
      -D WITH_OPENGL=ON \
      -D BUILD_opencv_cudacodec=OFF \
      -D CUDA_NVCC_FLAGS="--expt-relaxed-constexpr" \
      -D WITH_TBB=ON \
      -D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.5.2/modules \
      ../

• cv_bridge manually built with new OpenCV version 4.5.2 (not ROS default version) - reference
• ROS melodic with Ubuntu 18.04 LTS

How to install and setup

• Assuming PX4-SITL simulator is already installed through following instructions to install PX4-Avoidance package
• Clone the repository

$ git clone --recursive https://github.com/engcang/ieee_uav_2022

• Setup UAVCC-simulator

$ cd ieee_uav_2022/uavcc-simulator/trial_1_setup
$ mkdir build && cd build
$ cmake ..
$ make

$ cd ieee_uav_2022/uavcc-simulator/trial_2_setup
$ mkdir build && cd build
$ cmake ..
$ make

$ cd ieee_uav_2022/uavcc-simulator
$ echo "export GAZEBO_MODEL_PATH=$GAZEBO_MODEL_PATH:$(pwd)/trial_1_setup:$(pwd)/trial_2_setup" >> ~/.bashrc
$ source ~/.bashrc

$ cd ieee_uav_2022/uavcc-simulator
$ echo "export GAZEBO_PLUGIN_PATH=$GAZEBO_PLUGIN_PATH:$(pwd)/trial_1_setup/build:$(pwd)/trial_2_setup/build" >> ~/.bashrc
$ source ~/.bashrc

• Setup OOQP

  • Dependencies.

   $ sudo apt-get install gfortran
   $ sudo apt-get install doxygen
   $ sudo apt-get install texlive-latex-base

  • Install ma27 and type below commands in MA27's folder.

   $ cd ieee_uav_2022/ma27-1.0.0
   $ ./configure
   $ make
   $ sudo make install

  • Install OOQP and type below commands in OOQP's folder.

   $ cd ieee_uav_2022/OOQP
   $ ./configure
   $ make 
   $ sudo make install

• Build this Repo

$ cd ieee_uav_2022/
$ echo "export GAZEBO_MODEL_PATH=$GAZEBO_MODEL_PATH:$(pwd)/drone_models" >> ~/.bashrc
$ source ~/.bashrc

$ cd ..
$ catkin build -DCMAKE_BUILD_TYPE=Release

---

How to run

• Run roscore

$ roscore

• Run rviz for debugging

$ roscd ieee_uav
$ rviz -d rviz.rviz

• Run rqt_plot to see distance between rover and the UAV

$ rqt_plot

add /distance/data

• Launch Gazebo world

$ roslaunch ieee_uav trial1.launch

or

$ roslaunch ieee_uav trial2.launch

• Run ieee_uav

$ roslaunch ieee_uav main.launch

Important: launching gazebo and running the main code should be done in a short second (ASAP).

Expected result

• UAV starts to track the rover
• Paths and debugging images should be visualized in rviz
• YOLO detection should work at least at 100Hz, if you setup CUDA, cuDNN, OpenCV, and cv_bridge properly
  • Note, the inference can work upto at 100Hz purely, but it works at fixed 15Hz actually, considering limited computational resource of the UAV.

References

• Targer motion prediction (Bezier curve) - Fast Tracker
• YOLO - ROS code
• MPC - Googling...