The First Steps of Solo - Sim2Real for a Quadruped Robot

This repository provides a ROS2 package including a MuJoCo simulation environment, tasks and an interface to the physical hardware of the Solo8 quadruped robot by Open Dynamic Robot Initiative.

To build and source the package run:

colcon build --packages-select solo_mujoco
source install/setup.bash

Simulation Environment

The simulation environment is powered by the MuJoCo physics engine and connected to ROS2-control via a system interface. This system interface accepts control targets from ROS2 topics and publishes state information.

Start the simulation with

ros2 launch solo_mujoco simulate.launch.py

Options:

Option	Type	Description
plot	bool	Whether to start rqt alongside with the simulation

Sending Commands to the Simulation

The simulation accepts commands to the following topics:

Topic	Description
/position_controller/commands	Accepts joint position targets as Float64MultiArray in the order hip_front_left, knee_front_left, ..., hip_back_right, knee_back_right
/velocity_controller/commands	Accepts joint velocity targets as Float64MultiArray in the order hip_front_left, knee_front_left, ..., hip_back_right, knee_back_right

State Published by the Simulation

The system interface relays the following data to the ROS2 system:

Topic	Frequency	Description
/imu	60Hz	Publishes an Imu object containing acceleration, angular velocity and quaternion orientation
/foot_contacts	Publishes a Float64MultiArray of length 4 with the vertical force. Order: [FL, FR, HL, HR]

Real Robot Interface

The real robot system interface connects to the real robot using the Open Dynamic Robot Initiative SDK. The built SDK files are places in the master_biard_sdk_libs folder.

It accepts the same control targets as the simulation.

Launch the real robot interface with:

ros2 launch solo_mujoco real_solo.launch.py

Options:

Option	Type	Description
eth_interface	string	The ethernet interface name of the robot

Task Controller

The repository comes along with a task controller which manages the execution of several tasks in the simulation and on the real robot, e.g., walking or sitting.

To change the executed task, use the following services. All services accept a std_srvs/srv/Trigger message:

Service	Description
/tasks/trigger/TROT	Triggers a walk of the robot with a trot pattern
/tasks/trigger/SIT	Triggers the robot to sit
/tasks/trigger/STEER	The robot walks a curve to reach a specified steering angle and then maintains a stable stand. The steering angle can be specified by publishing the angle in degrees as std_msgs/msg/Float64 to the /steering_angle topic. For the simulation holds: East=0°, South=-90°, North=90°

Installation

1. Make sure to place a MuJoCo release somewhere in the ROS2 workspace and reference it with the MUJOCO_DIR environment variable.
2. Make sure to have the rclcpp, pluginlib, hardware_interface, glfw3 and GLEW packages installed. See the CMakeLists.txt file for details.