Skip to content
Open-SourceSprint

real-to-sim-to-real

🌍 Mission

The mission of the Real-to-Sim-to-Real Framework is to create tools and libraries that empower users to train their robots in custom environments tailored to their specific needs. The framework also enables users to create their own environments easily using just their smartphone. This mission aims to democratize access to robotics development, making it accessible to everyone, regardless of their expertise or resources.

🔑 Why It Matters

This framework is significant because it allows everyone—from hobbyists to professionals—to train their robots for custom tasks, specific workflows, and various real-world applications. The ability to create simulations on a smartphone gives users the flexibility to design environments that closely match the tasks their robots need to perform.

By making it easy to create and deploy robots for real-world work without requiring expensive hardware or specialized tools, we are opening up new possibilities in fields such as:

  • Robotics education
  • Automated manufacturing
  • Home automation
  • Healthcare robots

This framework ensures that robotics development is no longer limited to those with large budgets or advanced technical teams, but available to anyone with an idea.

🚀 Key Features

Environment Setup

  • Simplified process to load and configure Isaac Gym environments tailored to your simulation needs.

Data Collection

  • Effortlessly collect and store demonstration data from the simulated environment for reinforcement learning tasks.

Reinforcement Learning

  • Use GPU-accelerated reinforcement learning methods to train policies that can be transferred to real-world robots.

Teacher-Student Distillation

  • Transfer skills from complex teacher models to simpler student models, making it easier to deploy advanced policies on real robots.

Real-World Deployment

  • Deploy your trained policies on real robotic setups, ensuring a smooth transition from simulated to real environments.

🤝 How to Contribute

We invite the community to contribute to the Real-to-Sim-to-Real Framework! You can help by:

  • Offering New Features: If you have ideas for new functionalities, submit a feature request or pull request.
  • Fixing Bugs: Report bugs or help improve the existing codebase.
  • Experimenting with the Framework: Test the framework with different environments and share your insights or results.

To contribute, submit your pull requests on the GitHub Repository.

If you’re unsure where to begin, feel free to open an issue for any questions or suggestions.

🎉 Benefits of Contributing

  • Collaboration Opportunities: Work with robotics innovators globally.
  • Access to Resources: Get access to high-quality datasets and tools.
  • Guidance from Experts: Receive insights and mentorship from experienced professionals.
  • Recognition:
    • Be featured on the Contributor Wall.
    • Potential job opportunities if your work is selected for further collaboration.
    • Earn the SmilingRobo Contributor Tag to add to your resume.
    • Shoutouts on SmilingRobo’s social media to recognize your contributions.

Start contributing and help us advance robotic policy training using the Real-to-Sim-to-Real Framework!