This repository provides OpenAI Gym environments for the SpotMicro, an open-source 3D-printed quadruped robot. It enables researchers and engineers to train control policies using Reinforcement Learning, with the goal of transferring learned behaviors from simulation to the physical robot for tasks like locomotion and manipulation.

How It Works

The project utilizes Proximal Policy Optimization (PPO) with a hybrid policy approach, allowing for a mix of user-specified control and learned behavior. Environments are implemented using pyBullet for simulation, supporting both a Bezier controller (inverse kinematics-based gait generation) and an Open Loop mode (learning from scratch or simple trajectories). This dual approach facilitates comparison of different control strategies.

Quick Start & Requirements

• Install: pip install rex_gym or install from source.
• Prerequisites: Python 3.7 virtual environment (e.g., using Anaconda).
• Documentation: https://github.com/nicrusso7/rex-gym

Highlighted Details

• Supports various locomotion tasks: static poses, galloping, walking, turning, and standing up.
• Includes terrain variations for robustness testing (random heightfield, hills, mounts, maze).
• Offers a "playground" mode for interactive simulation and manual pose adjustment.
• Can integrate a 6DOF robotic arm (Poppy Ergo Jr) for more complex manipulation tasks.

Maintenance & Community

The project is primarily maintained by nicrusso7. Related repositories include rexctl for robot control and rex-models for URDF visualization.

Licensing & Compatibility

The repository does not explicitly state a license in the README. Users should verify licensing for commercial or closed-source use.

Limitations & Caveats

The project targets Python 3.7, which is nearing end-of-life. While it mentions transferring policies to a real robot, the README focuses on simulation and does not detail the hardware interface or real-world deployment challenges.