Reference Platform & Applications Working Group

Coordinator: Yadunund Vijay (Intrinsic)

Overview

This group makes the initiative’s work tangible by integrating all technologies into a physical platform with accessible examples and community-facing activities.

Scope

Reference Hardware Platform (T(AI)rtle Bot 5)

Build and maintain a low-cost, open-source robot that serves as the primary integration and demonstration target for the entire initiative. The platform includes simulation demonstrations for accessibility.

Showcase Applications & MCP Introspection

Develop easy-to-use, Python-centric examples on the platform that also demonstrate advanced features like MCP introspection for runtime system analysis and debugging.

Community Engagement & Outreach

Use the reference platform and its applications as a tool to grow the user base through hackathons, tutorials, and community events.

Simulation Resources

Provide examples using teleoperation and deployment with various simulators including O3DE, Mujoco, and Isaac Sim. Note: RL training in simulation is not prioritized as people favor lock-step approaches.

Roadmap

Long-term Vision

Create a low-cost, open-source reference platform (T(AI)rtle Bot 5) that serves as the primary integration and demonstration target for the entire Physical AI initiative. The platform will enable researchers and developers to:

• Test and validate Physical AI standards and interfaces
• Develop and share applications accessible to the broader community
• Engage through hackathons, tutorials, and community events
• Access simulation resources for accessibility and development without hardware

Current Priorities

• Platform Design & Specifications: Define hardware requirements and component selection for T(AI)rtle Bot 5
• Simulation Environment: Develop simulation demonstrations for accessibility and testing
• Example Applications: Create Python-centric examples demonstrating Physical AI capabilities
• MCP Introspection Tools: Implement runtime system analysis and debugging capabilities
• Multi-Simulator Support: Establish examples using O3DE, Mujoco, and Isaac Sim for teleoperation and deployment

Upcoming Milestones

• Q1 2025: Complete T(AI)rtle Bot 5 hardware design and initial prototype
• Q2 2025: Release simulation environment with basic teleoperation examples
• Q2 2025: Publish first set of showcase applications with MCP introspection
• Q3 2025: Launch community engagement program with tutorials and documentation
• Q4 2025: Host first hackathon event with reference platform

Interested Members

• Yadunund Vijay (Intrinsic) – Coordinator
• Leonardo Rossetti (Red Hat)
• Peng Xu (Google DeepMind)
• Zeming Zhao (Sharpa)
• Hemal Shah (NVIDIA)
• Adam Dabrowski (Robotec.ai)
• Anh Vu Vu (Huawei)
• Sebastian Castro (RAI)
• Julia Jia (OSS contributor)

How to Contribute

Interested in contributing to the Reference Platform & Applications working group? Here’s how to get started:

1. Review the roadmap and current priorities above
2. Check the Getting Involved page for onboarding information
3. Reach out to the coordinator or attend bi-weekly SIG meetings
4. Contribute to open-source development of the reference platform