upkie/upkie

Open-source wheeled biped robots

What it solves

Upkie provides an open-source platform for building and controlling wheeled biped robots. These robots combine wheels for efficient balance and legs to handle uneven terrain, offering a versatile hardware design that is accessible to developers using off-the-shelf components.

How it works

The project consists of a hardware design and a software library. The software is available as a Python library that integrates with Gymnasium environments, allowing developers to write control code once and deploy it to either a PyBullet simulation or the physical robot (powered by a Raspberry Pi). It supports development in both Python and C++ and includes various control strategies such as Model Predictive Control (MPC) and Proportional-Derivative (PD) feedback.

Who it’s for

It is designed for robotics researchers, hobbyists, and developers interested in legged locomotion, balancing, and reinforcement learning, as it provides a standardized environment for testing behaviors in simulation before deploying to hardware.

Highlights

  • Hybrid Locomotion: Combines wheels and legs for balance and terrain negotiation.
  • Unified API: Use the same Gymnasium-compatible code for both simulation and real-world deployment.
  • Open Hardware: Designed to be built with off-the-shelf tools and mjbots actuators.
  • RL Ready: Compatible with reinforcement learning frameworks like Stable-Baselines3 and GPU-accelerated playgrounds like MjLab.

Related

  • Project
    PetoiCamp/OpenCat-Quadruped-Robot

    An open-source framework for quadruped robots that handles gait coordination and servo control, enabling researchers and makers to build AI-powered robotic pets.

  • Project
    nasa-jpl/open-source-rover

    An open-source, build-it-yourself scaled version of the JPL Mars rover designed for education and research using consumer off-the-shelf parts.

  • Project
    PetoiCamp/OpenCatEsp32-Quadruped-Robot

    An open-source quadruped robotics framework for ESP32-based boards, enabling affordable gait research, reinforcement learning, and physical AI development.

  • Project
    PetoiCamp/OpenCat-Old

    An open-source platform for affordable quadruped mini robots used for physical AI education, gait research, and reinforcement learning.

  • Project
    pink-kinematics/pink

    Pink is a Python library for differential inverse kinematics of articulated robots. Built on the Pinocchio dynamics engine, it lets you define weighted tasks (position, orientation, posture, etc.), automatically builds a quadratic program, and solves for joint‑space velocities while respecting joint and floating‑base limits. Install via conda or PyPI, load any URDF, set task targets, and run a simple integration loop to drive the robot toward the desired pose. The project includes many examples (arms, humanoids, quadrupeds, mobile bases) and is Apache‑2.0 licensed.