Embodied Agentic AI Working Group

Coordinator: Adam Dabrowski (Robotec.ai)

Overview

Develop standards, best practices and reference applications for flexible skill composition, where reasoning models (LLMs, VLMs) orchestrate state of the art (e.g. navigation, manipulation packages) and learned policies through tool use. This is a meta-approach in that it incorporates other methods, triggers skills in the right context, expands upon existing robot stacks, and can serve as a framework for supervised learning.

Scope

Connect to Embodied AI Community

Connect to the Embodied AI Community Group and associated initiatives such as awesome-embodied-ai-ros. The group would welcome presentations about new SIG initiatives.

Open-Source Framework Collaboration

Inform, evaluate, use or contribute to open-source embodied AI frameworks such as RAI or ROSA.

Foundation Model Fine-Tuning

Advance fine-tuning of compact foundation models for embodiment and specific robotic domains.

On-Board Deployment

Deployment of agentic AI fully on board - optimizing for compute.

Roadmap

Long-term Vision

Roadmap content to be added

Current Priorities

Roadmap content to be added

Upcoming Milestones

Roadmap content to be added

Interested Members

Detailed member list to be added

How to Contribute

Contribution opportunities to be added