Teach and Repeat using Bézier Curves

This repository implements a Teach and Repeat (T&R) navigation system for mobile robots, allowing them to autonomously follow previously demonstrated paths.

Dependencies

• ROS 2 Humble;
• Ubuntu 22.04;
• Colcon Common Extensions;
• Turtlebot3 simulation;

Methods

There are two methods to follow the demonstrated path:

• Dot-to-dot (nodes/repeat_path_coords.py): The robot follows a Pure Pursuit-like approach where it intercepts a look-ahead point. Once it reaches this point, a new point further along the path is set as the next target.
• Bezier curve based (nodes/repeat_bezier_path.py): The robot simulates multiple potential local paths ahead using Bézier curves and selects the optimal path based on predefined criteria.

Installation

To set up the Teach and Repeat system, follow these steps:

First, create a workspace directory and navigate into it:

mkdir -p teach_repeat_ws/src
cd teach_repeat_ws/src

Clone this repository along with its submodules:

git clone --recurse-submodules https://github.com/jardeldyonisio/teach_and_repeat.git

Make the installation script executable and run it with superuser privileges:

cd teach_and_repeat/
chmod +x install.sh
sudo ./install.sh

Navigate back to your workspace and, initialize rosdep in your workspace:

cd ~/teach_repeat_ws
rosdep init

Update rosdep to fetch the latest package information:

rosdep update

Install the required ROS package dependencies:

rosdep install --from-paths src -y --ignore-src --rosdistro humble

Once the dependencies are installed, proceed with building the workspace:

cd ~/teach_repeat_ws
source /opt/ros/humble/setup.bash # Source ROS
source ./install/setup.bash # Source the workspace
colcon build

cd ~/teach_repeat_ws
source /opt/ros/humble/setup.zsh # Source ROS
source ./install/setup.zsh # Source the workspace
colcon build

Run

After setting up the environment, follow these steps to run the system:

First, export the TurtleBot3 model:

export TURTLEBOT3_MODEL=burger

Launch the turtlebot world:

ros2 launch turtlebot3_gazebo turtlebot3_world.launch.py

Open another terminal (CTRL + ALT + T) and launch the navigation system:

export TURTLEBOT3_MODEL=burger
cd ~/teach_repeat_ws
source /opt/ros/humble/setup.zsh # Source ROS
source ./install/setup.zsh # Source the workspace

Run the turtlebot navigation system:

ros2 launch turtlebot3_navigation2 navigation2.launch.py use_sim_time:=True map:=$HOME/teach_repeat_ws/src/teach_and_repeat/map/map.yaml

You must use Pose 2D Estimation to define the initial position before start.

Demonstrate Path

To demonstrate a path, you can either set a nav2 Goal on Rviz or teleoperate the robot. To teleoperate, open another terminal and start the teleoperation node:

ros2 run teleop_twist_keyboard teleop_twist_keyboard

Control the robot using the following keys:

• I to move forward
• K to move backward
• J to turn left
• L to turn right
• Q to increase speed
• Z to decrease speed
• Space to stop

Open another terminal (CTRL + ALT + T) and start the path demonstration:

cd ~/teach_repeat_ws
source /opt/ros/humble/setup.zsh # Source ROS
source ./install/setup.bash # Source the workspace
ros2 run teach_and_repeat teach_path_coords.py --ros-args -p path_name:=path_coords -p reference_frame:=map

To end the demonstration the user have to press CTRL + C.

Parameter	Default Value	Description
path_name	path_coords	The name of the file where the path coordinates are stored, located in the path_saves folder.
reference_frame	map	The reference frame for the path points. If set to map, the points are relative to the map frame. If set to odom, the points are relative to the robot's odometry frame.

Follow Path

To follow a path, first ensure that the coordinates from the path demonstration are stored in the teach_and_repeat/data/path_coords.txt file. This file is also the default file read during the path following process. Make sure you start from the same point where the demonstration began.

If the navigation system is not already running, launch it:

ros2 launch turtlebot3_gazebo turtlebot3_world.launch.py

Then, run the path following node:

ros2 run teach_and_repeat repeat_bezier_path

After starting the repeat node, set a pose using '2DPoseEstimate'. The robot will immediately begin following the path.

Demo Videos

Check out the teach and repeat system in action:

• Following a virtual path (Repeting process): Watch the demo on YouTube
• Teach and Repeat using Odometry and Map (Simulation): Watch the demo on YouTube
• Obstacle Detection (Simulation): Watch the obstacle detection demo on YouTube
• Teach and Repeat using Odometry Only (Simulation): Watch the demo on YouTube

Contributing

Feel free to open issues or submit pull requests to improve this project.

Publication

For more details, refer to our publication Teach and Repeat for Path Planning Using Bézier Curves.

@INPROCEEDINGS{Dyonisio2024,
  author={J. Dos Santos Dyonisio and others},
  title={Teach and Repeat for Path Planning Using Bézier Curves},
  booktitle={2024 Brazilian Symposium on Robotics (SBR) and 2024 Workshop on Robotics in Education (WRE)},
  year={2024},
  pages={79-84},
  doi={10.1109/SBR/WRE63066.2024.10837801}
}

TODO

• Create a .yaml file to configure:
  • The robot model;
  • Future behavior parameters.
• Use multiple Bézier curves (e.g., B-splines) to represent the path.
• Update the teach node to:
  • Accept a string input to name the path;
  • Require a key press to start collecting poses.
• Display poses during the teaching process.
• Show the circular threshold around the closest Bézier reference point.
• Automatically calculate start_num_knots based on the path.