Intelligent Systems 2024

This is the repository for the student projects of the 'Introduction to Intelligent Systems' course at Seoul National University. rccar_gym environment codes are originated from F1TENTH Gym repository.

Original author of F1TENTH Gym: Hongrui Zheng

(Special Thanks to Hyeokjin Kwon, Geunje Cheon, Junseok Kim for editing rccar_gym)

Author of this repo: Minsoo Kim, Yoseph Park, Subin Shin

RCCar Gym Environment Setting

We recommend you install packages inside a virtual environment such as Anaconda (or virtualenv) as you did in keyboard control preproject. Also we recommend creating a new virtual environment instead of reusing the existing one.

conda create -n rccar python=3.8
conda activate rccar

git clone https://github.com/rllab-snu/Intelligent-Systems-2024.git
cd Intelligent-Systems-2024/rccar_gym
pip install -e .

This will install a gym environment for the RC car and its dependencies.

ROS2 Setting

We use ‘ROS2 Foxy’ to run the gym environment and project codes.

First, install ROS2 foxy by following the documentation.

To build the ROS2 packages to use our specific python virtual environment, we should install colcon building tools to our python environment.

Assuming the virtual environment is activated, i.e. conda activate rccar,

pip install colcon-common-extensions

This enables installed files resulting from colcon build to use desired package in our environment.

Now, install dependencies and build the packages.

cd Intelligent-Systems-2024
rosdep update --rosdistro foxy
rosdep install -i --from-path src --rosdistro foxy -y
colcon build --symlink-install

Note that --rosdistro foxy is required for rosdep update since foxy is an end-of-life version.

Note that --symlink-install is required to use modified python files directly without building again.

After building the package, we should use following command in every terminal we want to use our packages.

source install/setup.bash

Running Codes

Project 1

For just project 1, you can run using following command.

ros2 run rccar_bringup RLLAB_project1
# Replace RLLAB with your team name

For project 2, you have to save trajectories from the pure pursuit demonstrations.

You can save trajectory by using --save argument.

Also, you can run without rendering by using --no_render for faster process.

ros2 run rccar_bringup RLLAB_project1 --save --no_render
# Replace RLLAB with your team name

Project 2

When you want to train new model and evaluate, you can run project2 code with --mode train argument.

ros2 run rccar_bringup RLLAB_project2 --mode train
# Replace RLLAB with your team name

When you want to load trained model without training, you can just run without --mode argument since its default value is val.

ros2 run rccar_bringup RLLAB_project2
# Replace RLLAB with your team name

Manually publishing map topic

For each project code, you can publish /query topic manually using following command in another terminal.

ros2 topic pub --once /query message/msg/Query "{id: '0', team: 'RLLAB', map: 'map1', trial: 0, exit: false}"
# Replace RLLAB with your team name and you can use other maps we provide in 'maps' directory

Note that you can publish the topic once with --once argument.

Map generation

You can change the parameters defined in random_trackgen.py and randomly generate your own map with --seed argument. You can run random_trackgen.py using following command.

cd Intelligent-Systems-2024/maps
python random_trackgen.py --seed your_seed --name your_map_name