This is the codebase we released for our ICRA’22 F1tenth Competition. Check an exciting video in which we competed with the ETH Zurich team!

Overview

We are using pure-pursuit controller and a lane switcher for overtaking/obstacle avoidance.

Installation

Download the package

To use the stack in your workspace, you need to put them inside the src folder so the structure of your workspace should be like this:

├── build
├── install
├── log
└── src
    ├── config
    ├── csv
    ├── dummy_car
    ├── trajectory_generator
    ├── lane_follow
    ├── maps
    ├── opponent_predictor
    ├── scripts
    └── <YOUR OTHER PACKAGES>

Create virtual environment and install the dependencies

Note: This package use the TUM global raceline optimization which requires specific version of numpy and sklearn packages which can be incompatible with other application. So, it is better to use Anaconda3/venv to create a virtual environment specific for using the raceline optimization. Here, we use the venv package in python.

• Go inside the src folder: cd ./src

• Install venv: sudo apt install python3.8-venv

• Create virtual environment, exclude it from the colcon build and source the virtual environment: python3 -m venv ./venv && touch ./venv/COLCON_IGNORE && source ./venv/bin/activate

• Install the dependencies: pip install -r trajectory_generator/requirements.txt

Process the map and generate the optimal raceline

Before you execute the following steps, make sure you are inside the ./src folder and the virtual environment is activated with: source ./venv/bin/activate.

• Put maps(.png, .pgm, etc) and its configuration(.yaml) under maps
• Edit config/params.yaml to include correct map name
• Run ./scripts/populate.sh to populate data files into subdirectories
• Run python3 trajectory_generator/lane_generator.py to generate track data from image(Click the pop-up image, Press any keys( Enter, for example) to continue).
• Run python3 trajectory_generator/main_globaltraj.py to generate race line file.
• Run python3 trajectory_generator/raceline_scripts.py to visualize the curvature of the generated optimal raceline.
• Run ./scripts/populate.sh again to populate generated raceline into subdirectories

Use the generated raceline for lane_follow(single agent)

You should do colcon build and launch the node from the workspace folder ./ as normal, not inside the src.

• Bring up simulation RVIZ(see f1tenth_gym_ros): ros2 launch f1tenth_gym_ros gym_bridge_launch.py

• Run ros2 launch lane_follow lane_follow_launch.py

You can load the config file ./config/race.rviz for rviz2. The global raceline is published under topic/global_path/optimal_trajectory, and the target of pure pursuit is published under topic /waypoint.

ThirdParty Library

raceline optimization: TUM global raceline optimization

Attendees for ICRA’22 F1tenth Competition

Zhijun Zhuang, Jiatong Sun, Pankti Hitesh Parekh