Project: Domestic Robots - LeRobot

Overview of this Project

Motivation

Well then, why not give it a try?

Introduction to the open-source projects

  1. LeRobot - aims to provide models, datasets, and tools for real-world robotics in PyTorch. The goal is to lower the barrier to entry to robotics so that everyone can contribute and benefit from sharing datasets and pretrained models.

    The SO-101 Robot Arms
  2. XLeRobot: Accidentally I came across this project that’s called XLeRobot and was at once fascinated by it. It combines several existing LeRobot projects, such as the SO101-Robot arm and the LeKiWi moving base, with an IKEA RÅSKOG Trolley to create a mobile robot platform that can navigate and manipulate objects in a domestic environment. I sent immediately a direct message to the author of the project, @VectorWang. He answered me very kindly and we found out even some mutual friends. I didn’t hesitate, but ordered the IKEA trolley and started to collect all the necessary parts.

    The LeKiWi The LeKiWi trying to catch a Kiwi.

    The XLeRobot

Plan for the Project

How should it look like?

What should it be able to do?

  1. Movability: The robot should be able to navigate around the house, avoiding obstacles and reaching designated locations.
  2. Remote Control/Teleoperation: The robot base should be controllable via a joypad or a mobile app. The 2 robot arms should be controllable via the 2 corresponding followers. One future development could be a standalone backpack remote control unit, which contains an IMU in detecting the movement and 2 leader arms in controlling the robot arms. These movements would be transmitted to the robot via WiFi or Bluetooth, or even throgh WAN-based communication.
  3. Detection: When not being remotely controlled, the robot should be able to recognize and locate objects in its environment using computer vision. Currently available: 2 cameras on the end-effectors, 1 camera on the moving base, and possibly an additional depth camera.
  4. Manipulation through VLA (Vision-Language-Action) models: The robot should be able to perform simple tasks such as picking up objects, opening doors, and pressing buttons based on natural language commands. This would require the integration of VLA models that can understand and execute instructions.

How am I going to achieve everything? (Timeline)

Here is a timeline I have in mind: Gantt Chart for the Project

Skills and knowledge I would need to acquire: - Robotics: Kinematics, Dynamics, Control Systems - Computer Vision: Object Detection, Image Processing - Machine Learning: VLA Models, Reinforcement Learning - Hardware: RGB-D Camera, Raspberry Pi, Motor Controllers, Optional: Jetson Nano - Software: ROS, Python, PyTorch, OpenCV

Set-ups and Configurations

First Procurement

The first procurement has been done on 2024-06-25 according to the Bill of Materials.

Software Installation

  1. I have installed dual boot Windows 11 and Ubuntu 24.04 on my laptop, following the instructions from here.
  2. For CAD design, I have installed Fusion 360, Shapr3D.
  3. For programming, I have installed VS Code, Anaconda, Git, GitHub Desktop.

Let’s rock’n’roll!