Bootcamp for Legged Robots

Introduction

This is a learning bootcamp for legged robotics newbies. Nevertheless, fundamentals in robotics are grossly omitted; for those who have some basic knowledge, you are in the right place. In which, we have 2 main camps:

1. 3D Walker: Naive simulator from scratch, which consists the development of:

   • Hopper
   • Hopper Control
   • Passive Walker
   • Walker Control
   • 3D Walker Control

   Goal:

   

   via:

   • dynamics modelling (Euler-Lagrange Equations),
   • fixed points with Poicare maps,
   • forward and inverse kinematics,
   • trajectory optimization, and
   • feedback linearization.
   
   

2. Quadruped: Basic controller design for quadruped with unitree Go1 in ROS/Gazebo simulation:

   • Joints Control
   • Base Control
   • Force Control
   • Gait Design
   • Trot/Crawl/Pronk

   Goal:

   via:

   • forward and inverse kinematics,
   • dynamics modelling (Euler-Lagrange Equations),
   • quadratic programming, and
   • cycloid trajectories.
   
   

Usage

For biped scripts:

1. Please first setup conda environment.

   git clone https://github.com/Dynamics-Learning-Workshop/Legged-Robots-Bootcamp.git
   
   conda create --name leg_bootcamp
   conda activate leg_bootcamp
   
   cd ./Legged-Robots-Bootcamp/biped_ctrl_scripts
   conda install --file setup.txt
   

2. For 3D modelling, please do the following to wrap up python scripts in C, which will take around 1 minute (tested on M1 pro):

   cd ./biped_ctrl_scripts/5_walker_3D_control/f_walker_3D/dynamics/compiled_funcs
   python gen_lib.py
   

3. Execute the python files.

For quadruped codes:

1. Please follow this and that to setup the simulation environment.
   • please note that ROS is a prerequisite; we tested all the codes on ubuntu 20.04 + ROS Noetic.
   • also note that lcm, osqp and osqp-eigen should be installed.
   • I use tmux a lot, and this repo also used tmux. Please install tmux by sudo apt install tmux.
2. Prepare a gaming controller that is compatible with /joy. Setup the controller via this instruction.
3. My practice is to install unitree_ros and unitree_guide in one workspace, and this repo in another workspace.

# do sim_env install as instructed above

cd && mkdir -p leg_boot_ws/src && cd ~/leg_boot_ws_src
git clone https://github.com/Dynamics-Learning-Workshop/Legged-Robots-Bootcamp.git

1. Say the workspaces are respectively leg_sim_ws and leg_boot_ws, first compile them with catkin_make, and then:

   cd ~ && ./leg_sim_ws/src/unitree_guide/sim.sh
   

2. Then

   source ~/leg_boot_ws/devel/setup.bash 
   roslaunch quadruped_ros_ctrl ctrl.launch
   

3. With the terminal, you will see some texts telling you the current FSM state. The control inputs via the controller are:

   

Reference

Based on

• UIC course (legged robotics) by Pranav Bhounsule,
  • I did my own implementation in Python
• Open project by Boston Cleek and Unitree Guide.,
  • I took reference from the code structure
  • and designed my own controller that can
    • stand
    • walk
    • pronk
• Simulation platform developed by UniTree Gazebo.
  • my implementation is based in Gazebo

@misc{pranav,
  title        = {Legged Robotics},
  author       = {Pranav Bhounsule},
  year         = 2021,
  note         = {\url{https://pab47.github.io/legs.html} [Accessed: 17/Sep/2024]}
}

@misc{qcontrol,
  title        = {Quadruped Control},
  author       = {Boston Cleek},
  year         = 2021,
  note         = {\url{https://github.com/bostoncleek/quadruped_control} [Accessed: 17/Sep/2024]}
}

@misc{unitree,
  title        = {Unitree Ros},
  author       = {Unitree},
  year         = 2024,
  note         = {\url{https://github.com/unitreerobotics/unitree_ros} [Accessed: 17/Sep/2024]}
}

@inproceedings{gehring2013control,
  title={Control of dynamic gaits for a quadrupedal robot},
  author={Gehring, Christian and Coros, Stelian and Hutter, Marco and Bloesch, Michael and Hoepflinger, Markus A and Siegwart, Roland},
  booktitle={2013 IEEE international conference on Robotics and automation},
  pages={3287--3292},
  year={2013},
  organization={IEEE}
}