DroneTest

For now there are two branches in this repo.

• Main -- the origin agile flight repo (unchanged)
• vitfly -- the vitfly runable code

How to use it

the same with agile flight and in the main branch.

set up workspace

cd     # or wherever you'd like to install this code
export ROS_VERSION=noetic
export CATKIN_WS=./icra22_competition_ws
mkdir -p $CATKIN_WS/src
cd $CATKIN_WS
catkin init
catkin config --extend /opt/ros/$ROS_VERSION
catkin config --merge-devel
catkin config --cmake-args -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_FLAGS=-fdiagnostics-color

cd src
git clone git@github.com:LuNa-shi/DroneSimTest.git agile_flight
cd agile_flight

setup ros

in agile_flight folder

./setup_ros.bash
catkin build

setup pytorch env

1. install anaconda
2. (optional) download cuda
3. create env conda create -n vitfly
4. conda activate vitfly
5. use pip to install requirement packages:matplotlib, numpy, pandas, pytorch, uniplot rospkg pyyaml scipy opencv-python
6. maybe you can use this command to to install everything (EXCEPT pytorch)

pip install matplotlib numpy pandas uniplot rospkg pyyaml scipy opencv-python

How to run vitfly

checkout

git checkout vitfly

what did I changed?

• /envtest/ros/evaluation_node,run_competition,user_code
• /models (with pretrained models)
• launch_evaluation.bash (most important for automatic)

how to run

• open one terminal, enter icra workspace and source devel/setup.bash, recommand write in .bashrc
• run roslaunch envsim visionenv_sim.launch render:=True gui:=False rviz:=True
• this code may fail because unity, try it again when every thing is killed
• when you see two windows(rviz and unity), you succeed

---

• open another terminal
• conda activate vitfly
• in the agile_flight folder, run bash launch_evaluation.bash (N)
• N is optional for running loops
• you can check /envtest/ros/train_set for data gathered in the running process.

Note

If you meet some problem, check out notebook for simExperiment, or ask me.

How to experiment on your own algorithm (recommended)

take astar for an example.

    git checkout -b astar

revise user_code, run_competition ... copy the launch_evaluation.bash in your branch.

Run manually

1. launch the scene roslaunch envsim visionenv_sim.launch render:=True gui:=False rviz:=True
2. enable the drone(4 commands)

rostopic pub /kingfisher/dodgeros_pilot/off std_msgs/Empty "{}" --once
rostopic pub /kingfisher/dodgeros_pilot/reset_sim std_msgs/Empty "{}" --once
rostopic pub /kingfisher/dodgeros_pilot/enable std_msgs/Bool "data: true" --once
rostopic pub /kingfisher/dodgeros_pilot/start std_msgs/Empty "{}" --once

3. use your own algorithm