Skip to content

Latest commit

 

History

History
100 lines (85 loc) · 3.58 KB

README.md

File metadata and controls

100 lines (85 loc) · 3.58 KB
Raw

Quadruped Wrapper

This repository offers a wrapper for a nonlinear Model Predictive Control (MPC) controller tailored for Unitree quadruped robots. It builds upon the legged_control and legged_perceptive_control repositories, which are based on OCS2.

Installation

Prerequisites

The library is written in C++11 and tested on Ubuntu 20.04 with the library versions available in the package sources.

Dependencies

  • A C++ compiler with C++11 support
  • Eigen (version 3.3)
  • Boost C++ (version 1.71)
  • Catkin: sudo apt-get install catkin
  • Catkin-pkg for Python 3: sudo apt install python3-catkin-tools
  • LCM: sudo apt-get install liblcm-dev
  • ROS Noetic
  • Additional packages: sudo apt install liburdfdom-dev liboctomap-dev libassimp-dev libglpk-dev libmpfr-dev
  • ROS-related packages: sudo apt install ros-noetic-pybind11-catkin ros-noetic-octomap-msgs ros-noetic-costmap-2d ros-noetic-realsense2-description

Build the Library

Create a new catkin workspace:

# Create the directories
# Do not forget to change <...> parts
mkdir -p <directory_to_ws>/<catkin_ws_name>/src
cd <directory_to_ws>/<catkin_ws_name>/

# Initialize the catkin workspace
catkin init
catkin config --extend /opt/ros/noetic
catkin config -DCMAKE_BUILD_TYPE=RelWithDebInfo

Clone the code:

# Navigate to the src directory
cd <directory_to_ws>/<catkin_ws_name>/src

# Clone the repo:
git clone --depth 1 --recurse-submodules https://github.com/DRCL-USC/Quadruped_Wrapper.git

Build the package (simulation and hardware):

cd ..
catkin build legged_deployment

Usage

Running Simulation

To start the simulation, first launch the Gazebo world:

roslaunch custom_description gazebo_world.launch wname:=stairs_stepping_stone gui:=true

You can specify the world name with wname and toggle the gui.

Next, load the controller by launching the following file:

roslaunch legged_deployment load_controller.launch

Then, run the controller node:

rosrun legged_deployment controller_client_node <controller_name> start

For <controller_name>, use perceptive_controller for the perceptive controller or legged_controller for the blind controller. If using the perceptive controller, also launch the elevation map and plane segmentation nodes:

roslaunch legged_deployment elevation_mapping.launch rviz:=true

This step is not needed for the blind controller.

Finally, run the keyboard command node to control the robot with the keyboard:

rosrun teleop_twist_keyboard teleop_twist_keyboard.py

Use w and s for forward and backward velocity, a and d for lateral velocity, and g and h for rotational velocity.

Running Hardware

For hardware operation, instead of launching Gazebo, run the hardware stack:

roslaunch legged_unitree_hw legged_unitree_hw.launch

To load cameras such as t265 and d435, use the following launch file or define a custom one:

roslaunch custom_description unitree_hw.launch

The remaining steps, including loading and running the controller, are the same as for the simulation.

Using Tmux Script

Multiple Tmux scripts located in the scripts directory automate the process of running the simulation or hardware, as explained above. To use these scripts, first install:

sudo apt-get install tmux tmuxp

Next, navigate to the scripts directory and load the Tmux session:

cd src/Quadruped_Wrapper/scripts
tmuxp load simulation_with_perception.yaml

You can modify the Tmux scripts based on your specifications.