Reinforcement learning for robotic skill acquisition, decision-making, and adaptation under real-world uncertainty.

Transferable, composable, and continually improving robot skills that extend beyond fixed tasks and environments.

VLA models, multimodal models, and embodied agents for robotic learning, generalization, and rigorous evaluation.

ERL develops public research workflows, tools, and reusable artifacts. Our flagship repository is ERL Research Skills, an open-source research operating system for Robotics & AI spanning literature discovery, experiments, paper writing, scientific review, submission, and research engineering.

• Open by default. Reusable skills, workflows, and research infrastructure are developed publicly.
• Evidence before claims. Research outputs should remain traceable to sources, code, experiments, and limitations.
• Reproducibility as practice. Provenance, evaluation, and artifact quality are part of the research process.
• Useful collaboration. Public contributions are welcome under clear scope, review, and attribution rules.

Use GitHub Discussions for research workflow ideas, usage questions, and open community conversation. Concrete bugs and contributions belong in the relevant repository's issues and pull requests.

ERL maintains its public projects and governance while welcoming contributors from any institution or career stage.