USTC-AIS-Lab / QuadrupedAntiSlip

QuadrupedAntiSlip

QuadrupedAntiSlip provides a proprioceptive slip detector and a slip recovery controller for reliable quadruped locomotion on slippery surfaces, which is integrated into an MPC locomotion strategy.

Example

Slip detection on slippery surfaces(ice). Yellow and red dots represent slipping and stationary contacts, respectively.

Locomotion on slippery surfaces(ice) with(left) and without(right) slip-recovery controller.

The video is available at here on YouTube.

Dependencies

The repo has the following dependencies. Make sure you installed them first:

• Eigen
• OSQP
• OSQP

Installation

Gazebo Simulation

The Gazebo Simulation is built as a docker image. You can create the container as follows: Run

docker pull chalkchalk/a1_gazebo

to get the image.

Run

docker run -dit\
    --name a1_gazebo \
    --cap-add=SYS_PTRACE \
    --gpus all \
    --shm-size=4G \
    -v /tmp/.X11-unix:/tmp/.X11-unix:rw \
    -e DISPLAY=unix$DISPLAY \
    -e NVIDIA_VISIBLE_DEVICES=${NVIDIA_VISIBLE_DEVICES:-all} \
    -e NVIDIA_DRIVER_CAPABILITIES=${NVIDIA_DRIVER_CAPABILITIES:+$NVIDIA_DRIVER_CAPABILITIES,}graphics \
    --privileged \
    --network=host\
    chalkchalk/a1_gazebo

to create the container.

Run

docker exec -it a1_gazebo bash

to open the terminal inside container.

Then, run

gazebo_ice

which is actually an alias to run the simulation with the ice map.

QuadrupedAntiSlip Controller

Clone the repository

Clone the repo into a ROS workspace

git clone [e-mail]:USTC-AIS-Lab/QuadrupedAntiSlip.git

Build the project

Enter the workspace and run

catkin build

Run the project

Launch the roslaunch file to start the controller for simulation:

roslaunch doggy_control gazebodog.launch

Then, you can run the joy_node to control the quadruped with a joystick.

rosrun joy joy_node

Acknowledgements

The MPC controller used in this repo is based on A1-QP-MPC-Controller provided by ShuoYangRobotics.