A1 Driver v2.8

A1 is intended mostly for academic community. Unitree Robotics has provided an SDK for their robots in C++ and a demo is also provided in ROS. But this ROS driver lacks several ROS features. To cover them this driver is provided.

Requirements

This driver requires a system setup with ROS. It is recommended to use Ubuntu 18.04 with ROS melodic, however using Ubuntu 16.04 with ROS kinetic should also work.

To make sure that robot control isn’t affected by system latencies, it is highly recommended to connect the robot via lan. The driver should run intially on remote PC but can also be run on Rasberry Pi/Nvidia board on robot.

Building

The ROS driver for Quadruped robot is dependent upon following packages:

Furthermore unitree_legged_sdk and aliengo_sdk have following dependencies.

  • Boost (version 1.5.4 or higher)

  • CMake (version 2.8.3 or higher)

  • LCM (version 1.4.0 or higher)

Attention

It is important to check your CPU type via executing lscpu in the terminal. If you have Architecture:x86_64 then nothing is to be changed, if you have Architecture:aarch64 then in the utils/qre_unitree_envs.bash, you must change it to the cpu from amd64 to arm64.

A script can be found below for automatic installation of all the required dependencies mentioned above and the provided software installation. After downloading and changing directory to the downloaded folder, installation script can be run by the following command:

You first have to give the script root permission via:

sudo chmod +x a1_installation_script.bash
./a1_installation_script.bash

Alternatively user can follow the procesure below to install all the dependencies and required software manually.

Warning

Please do not follow the installation guidelienes below for building the software, if you have executed the above mentioned installation script.

So now let’s start installing the dependencies one by one.

To install Boost run the following command

sudo apt install libboost-all-dev

CMake can be installed by the following command:

sudo apt install cmake

To install LCM, download the latest version here. Unzip the file and change directory to the LCM folder you just downloaded. Follow the steps below:

cd ~/lcm
mkdir build
cd build
cmake ../
make
sudo make install

To install unitree_legged_sdk:

git clone https://github.com/unitreerobotics/unitree_legged_sdk ~/unitree_legged_sdk
cd ~/unitree_legged_sdk
mkdir build
cd build
cmake ../
make

To install aliengo_sdk:

git clone https://github.com/unitreerobotics/aliengo_sdk ~/aliengo_sdk
cd ~/aliengo_sdk
mkdir build
cd build
cmake ../
make

Before installing ROS driver from Quadruped Robotics, we have to install unitree_legged_real(This will be removed in future versions).

source /opt/ros/$ROS_DISTRO/setup.bash
# For ROS melodic Gazebo 9 is supported so here make sure to use your respective Gazebo version.
source /usr/share/gazebo-9/setup.sh
source ~/catkin_ws/devel/setup.bash
export ROS_PACKAGE_PATH=~/catkin_ws:${ROS_PACKAGE_PATH}
export GAZEBO_PLUGIN_PATH=~/catkin_ws/devel/lib:${GAZEBO_PLUGIN_PATH}
export LD_LIBRARY_PATH=~/catkin_ws/devel/lib:${LD_LIBRARY_PATH}
export UNITREE_LEGGED_SDK_PATH=~/unitree_legged_sdk
export ALIENGO_SDK_PATH=~/aliengo_sdk
# amd64, arm32, arm64
export UNITREE_PLATFORM="amd64"
sudo apt-get install ros-$ROS_DISTRO-controller-interface  ros-$ROS_DISTRO-gazebo-ros-control ros-$ROS_DISTRO-joint-state-controller ros-$ROS_DISTRO-effort-controllers ros-$ROS_DISTRO-joint-trajectory-controller
mkdir -p ~/catkin_ws/src
cd ~/catkin_ws
git clone https://github.com/unitreerobotics/unitree_ros src/unitree_ros
catkin build
source devel/setup.bash

Now everything is installed(assuming ROS and Gazebo installed prior to following this tutorial). The next step is to install ROS driver provided by Quadruped Robotics.

To build the driver download the source code from here and copy/paste it inside catkin_ws/src and run the following commands:

source /opt/ros/$ROS_DISTRO/setup.bash
cd ~/catkin_ws && catkin build
source devel/setup.bash
sudo apt update -qq
rosdep update
rosdep install --from-paths src --ignore-src -y
catkin build
source devel/setup.bash

Software Overview

Quadruped robots work in two modes:

  1. High level mode

  2. Low level mode

User can be in one of these modes and they can not be switched after running.

High level mode

High level mode is mainly for walking and running. This mode too has two modes normal and motion mode which can be distinguished from their ips(For more details see Quadruped software guide). At the moment this driver is working with default configuration.

In high level mode the robot can walk/run. To run the robot in high level mode run the following command as root user(after turning on the robot and connecting through lan, the robot will be standing):

sudo su
source ~/catkin_ws/devel/setup.bash
roslaunch qre_ros high_level_mode.launch

With high_level_mode.launch unitree_legged_real/real.launch launches a communication channel between the robot and remote PC. Then later in qre_unitree_ros/high_level_mode a ROS node is run.

Node [high_level_driver]

Publications: state [unitree_legged_msgs/HighState]

Subscriptions: cmd_vel [geometry_msgs/Twist]

Services: set_body_pose [qre_msgs/SetBodyPose]

The robot pose can be set using the robot set_body_pose service. Robot state is published in state topic. This message contains a lot of information. For more information see high state message message.

Using the driver

The node subscribes to cmd_vel topic. Here special attention needs to be paid. As the robot is holonimic so different Twist message fields determine different movement. A nice tool here would be teleop_twist_keyboard. Here in the teleop holonomic and non holonomic modes can be used(Be aware key strokes and their usage).

Body height and walking

Body height and walking demo

Low level mode

In this mode all the motors can be controlled directly. For that one needs to first change the operation mode either to Servo mode or Electronic brake mode. Low level mode has again three control modes:

  1. Position

  2. Velocity

  3. Torque

In low level mode joint level control can be achieved. In Quadruped A1 this low level mode can be in three different levels i.e. Position, Velocity, Torque. All three of them can be used with this driver by publishing appropriate messages. Unlike high level mode robot communication is achieved inside the driver node so no seperate node should run.

After turning on the robot and connecting through lan run the following command:

sudo su
source ~/catkin_ws/devel/setup.bash
roslaunch qre_ros low_level_mode.launch

A node with follwoing information will be run:

Node [low_level_driver]

Publications: state [unitree_legged_msgs/LowState]
              joint_states [sensor_msgs/JointState]

Subscriptions: joint_cmd [qre_msgs/JointCMD]

Services: set_body_pose [qre_msgs/SetBodyPose] -> Not implemented yet
          set_control [qre_msgs/SetControl]
PositionControl VelocityControlFast TorqueControl
             Position control              Velocity control              Torque control

The node subscribes to joint_cmd topic. Again here special attention to be paid as well for the joints. For joint sequence see qre_msgs README. The driver publishes two topics state topic pubslihes the information from the robot. For more details see low state message. Joint state messages are also published. set_control service is can change the robot three modes namely i.e. Position, Velocity, Torque(Case sensitive). For every control level here similar joint messages need to be filled in joint_cmd topic other fields are not considered.

Position -> JointCMD.q
            JointCMD.Kd
            JointCMD.Kd
Velocity -> JointCMD.dq
            JointCMD.Kp
            JointCMD.Kd
Torque   -> JointCMD.tau

Using the driver

For velocity and torque control communication can done easily by just publishing the respected values e.g. by using rqt topic publisher, in terminal:

rqt

For position control mode the same control can be used through rqt, but to get smoother motions and communication a demo is provided for some predefined joint values. After running the low_level_driver.launch, execute the following command:

rosrun qre_ros llm_demo.py
PositionControlExample

Low level position control demo with ROS