Control Quadrupeds Soft Contacts

Control of quadrupedal robots with soft contact constraints.

Table of Contents

• Control Quadrupeds Soft Contacts
  • Table of Contents
  • Installation with Docker
  • Dependencies
  • Installation
  • Usage
    • Simulations
    • Plot
    • Add a new robot model
  • Troubleshooting
  • Known Bugs
  • Author
  • Citation

Installation with Docker

Install Docker Community Edition (ex Docker Engine) and Manage Docker as a non-root user.

Install NVIDIA proprietary drivers if the NVIDIA graphics card should be used.

Install NVIDIA Container Toolkit (nvidia-docker2) for NVIDIA support in the container.
If you do not want to use NVIDIA, edit the Docker image to remove the NVIDIA section and the run.bash script, removing the --gpus all flag in the docker run command. In addition, remove the additional_env from the Gazebo process gzserver in robot_launch/launch/robot.launch.py.

If you do not have access to all the submodules, including the private ones, clone the repo with

git clone https://github.com/ddebenedittis/control_quadrupeds_soft_contacts
cd control_quadrupeds_soft_contacts
git submodule update --init --recursive --remote src/external/quadprog/ src/external/rapidyaml/ src/rviz_legged/ src/robot/robots/anymal_c_simple_description/ src/robot/robots/solo_description/

ONLY IF you have access to all the submodules, including the private ones, clone the repo with

git clone --recursive https://github.com/ddebenedittis/control_quadrupeds_soft_contacts

If not already done, navigate to the workspace with

cd Quadruped_Control_Soft_Contacts

Build the docker image (-d to install additional packages for development) (-h for printing the help) (-p to install all the dependencies to use plot) (-r to rebuild the underlying images) (-t for installing bly to generate the terrain meshes from some heightmaps) (--t for installing tracing tools) (-a to install all the optional dependencies):

./build.bash [-a] [-d] [-h] [-p] [-r] [-t] [--tracing]

Run the container:

./run.bash

Build the ROS workspace:

colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release -DCMAKE_EXPORT_COMPILE_COMMANDS=ON && source install/setup.bash

Dependencies

• git
• Eigen
• xterm
• ROS2 Humble, and the following ROS2 packages: ros2-control, ros2-controllers, gazebo-ros-pkgs, gazebo-ros2-control, xacro, joint-state-publisher, joint-state-publisher-gui
• Pinocchio
• numpy, scipy, numpy_quaternion, quadprog
• latex only for plotting some figures.

Installation

If you do not have access to all the submodules, including the private ones, clone the repo with

git clone https://github.com/ddebenedittis/control_quadrupeds_soft_contacts
cd control_quadrupeds_soft_contacts
git submodule update --init --recursive --remote src/external/quadprog/ src/external/rapidyaml/ src/rviz_legged/ src/robot/robots/anymal_c_simple_description/ src/robot/robots/solo_description/

ONLY IF you have access to all the submodules, including the private ones, clone the repo with

git clone --recursive https://github.com/ddebenedittis/control_quadrupeds_soft_contacts

Build the workspace with

cd control_quadrupeds_soft_contacts
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release -DCMAKE_EXPORT_COMPILE_COMMANDS=ON && source install/setup.bash

Usage

Simulations

• ANYmal C simulation:

ros2 launch robot_gazebo anymal_c.launch.py [terrain:={rigid,soft,very_soft,multi_terrains,heightmap}] [gait:={teleop_base,static_walk,walking_trot,teleop_walking_trot}] [velocity_cmd:="(0.0, 0.0, 0.0)"] [use_rviz:={False,True}] [save_csv:={False,True}] [reset:={False,True}]

• SOLO12 static walk simulation:

ros2 launch robot_gazebo solo12.launch.py [terrain:={rigid,soft,very_soft,multi_terrains,heightmap}] [gait:={teleop_base,static_walk,walking_trot,teleop_walking_trot}] [velocity_cmd:="(0.0, 0.0, 0.0)"] [use_rviz:={False,True}] [save_csv:={False,True}] [reset:={False,True}]

• ANYmal C with SoftFoot-Qs static walk simulation:

ros2 launch robot_gazebo anymal_c_softfoot_q.launch.py [terrain:={rigid,soft,very_soft,multi_terrains,heightmap}] [gait:={teleop_base,static_walk,walking_trot,teleop_walking_trot}] [velocity_cmd:="(0.0, 0.0, 0.0)"] [use_rviz:={False,True}] [save_csv:={False,True}] [reset:={False,True}]

Optional arguments:

• terrain: select the terrain type: a rigid terrain (the gray one) or a soft terrain (the grass-like one).
• gait: select the gait type. With teleop_base the base pose can be teleoperated without moving the feet, static_walk implements a static walk, and walking_trot implements a dynamic gait. The speed of the walking_trot gait is commanded with the following parameter: velocity_cmd:="(0.0, 0.0, 0.0)". With teleop_walking_trot the robot is teleoperated by giving it a velocity command.
• use_rviz: when True, RViz is used to display the contact forces of the robot (still in Alpha version).
• save_csv: when True, some data is logged and saved in several .csv files in the log/csv/ folder. This data can be plotted with plot.py (in the logger_gazebo package).
• reset: must be used in another terminal when there is an already running simulation. The simulation will be restarted. The time is not reset to avoid problems with the controllers.

• contact_constraint_type: not working. Must be changed directly in the specific robot config .yaml file.

In some simulations (e.g. the simulation of ANYmal with SoftFoot-Q or the simulation with a trotting gait), due to the relatively high computational cost, it may be helpful to start or to run the simulation with a reduced real-time factor.

Plot

• Plot and save the figures in log/svg/ or log/pdf/ (must be executed while in the workspace folder):

ros2 run logger_gazebo plot

To save the plots, it is necessary to have a Docker image with latex installed. The image must have been built with the -l option.

Add a new robot model

Add a new robot model in the all_robots.yaml file located in src/robot/robot_model/robots/. Pay attention to the order of the feet' names (should be Left Front, Right Front, Left Hind, Right Hind).

Add a new robot description package. It is recommended to place it in src/robot/robots/. The package should have both a .urdf file and a .xacro file, which needs to be augmented with the necessary ros2 plugins etc (see, for example, anymal_c_simple_description/urdf/anymal_gazebo.xacro).

Generate a new launch file similar to the ones already present in src/robot/robot_gazebo/.

Create a new effort_controller.yaml file, similar to the ones already present in robot_control/config. Edit at least the robot_name field and the joints fields (according to the names of the joints of your robot).

Troubleshooting

• If you do not have an NVIDIA graphics card, or you do not have the propietary drivers (you can check this by using the command nvidia-smi), you should remove the additional_env from the Gazebo process gzserver in robot_launch/launch/robot.launch.py.

Known Bugs

• None

Author

Davide De Benedittis

Citation

If you find this project useful in your research, please consider citing my related work (available here):

@ARTICLE{debenedittis2024soft,
  author={De Benedittis, Davide and Angelini, Franco and Garabini, Manolo},
  journal={IEEE Transactions on Systems, Man, and Cybernetics: Systems}, 
  title={Soft Bilinear Inverted Pendulum: A Model to Enable Locomotion With Soft Contacts}, 
  year={2024},
  volume={},
  number={},
  pages={1-14},
  keywords={Legged locomotion;Quadrupedal robots;Foot;Vectors;Optimization;Computational modeling;Trajectory;Tracking;Planning;Jacobian matrices;Contacts;legged locomotion;optimal control;predictive control;quadratic programming},
  doi={10.1109/TSMC.2024.3504342}}