Locomotion Simulator Workspace

A complete simulation environment for the 8DOF quadrupedal robot.

This repository contains the following packages:

• mulinex_description: URDF description of the robot.
• mulinex_gazebo: Gazebo simulation of the robot.
• pi3hat_moteus_int_msgs: Messages for the communication between the controller and the simulation.
• rbt_pd_cnt: PD controller for the robot.
• rlg_quad_controller: Run that perform inference of a trained neural network and send the commands to the robot.
• locomotion_experiment: Launch files and nodes to generate references command for the controller, useful when running experiments.

Installation

Clone the repo:

    git clone --recursive https://github.com/CentroEPiaggio/locosim_ws

This is a project built on devcontainer. To install it, you need to have docker and vscode installed with the extension ms-vscode-remote.remote-containers. Then, you can open the project in vscode and click on the button on the bottom left corner and select "Reopen in container" This will build the devcontainer and install all the dependencies.

WARNING: The project is configured to run with NVIDIA GPU access. If you don't have an NVIDIA GPU, you need to change the devcontainer.json and remove the --gpus=all flag from the run args. By doing this you will not also be able to run the neural network controller.

Host installation

If you have ros humble installed on your host, you can run the simulation without the devcontainer, just open the workspace and build it with colcon.

Usage

Run all the commands from the terminal inside vscode

First, you need to build the project. To do so, you can run the build task of VScode with

    CTRL + SHIFT + B

or, in a terminal:

    colcon build --symlink-install

This will build the packages and create a symlink to the install folder.

To run the simulation, you can use

    ros2 launch mulinex_gazebo launch sim.launch.py

This will launch the simulation and the PD controller. To run the neural network controller, you can use ros2 launch rlg_quad_controller mulinex_simulation.launch.py. This will launch the simulation and the neural network controller.

The controller listens to a reference velocity in the topic /cmd_vel. You can steer and controll the robot using the rqt plugin rqt_robot_steering. To lauch it, run in another terminal

    ros2 run rqt_robot_steering rqt_robot_steering

Real exeriments usage

Troubleshooting

bad_alloc

use a different ROS_DOMAIN_ID

    export ROS_DOMAIN_ID=42

Tips and Tricks:

Simplifying meshes

STL files in the mulinex_description package may be too big for most of the computers. To simplify them, you can use a dedicated script in the mulinex_description/meshes folder. To use it, navigate to the folder with

    cd src/mulinex_description/meshes

and run

    python3 simplify_meshes.py

This will save all the simplified meshes as simpler_[meshname].stl in the same folder. You can then either rename them to replace the original meshes or change the filename in the file src/mulinex_description/urdf/links.xacro