New framework Pegasus translates human videos for robot learning

Researchers have developed Pegasus, a novel framework designed to bridge the embodiment gap in robotics. This system translates human manipulation videos into robot-learnable data by constructing a knowledge graph that includes task, affordance, and constraint graphs. Pegasus utilizes a hierarchical affordance latent space for generalization and a physics verifier to ensure kinematic feasibility and adherence to joint limits. Evaluated on benchmarks like GTEA Gaze+ and EPIC-KITCHENS-100, Pegasus demonstrates reliable cross-embodiment translation, reframing robot data generation as a scalable, low-resource knowledge transfer problem. AI

IMPACT Enables robots to learn from human demonstrations by bridging the embodiment gap, potentially accelerating robot development.

RANK_REASON The cluster contains a research paper detailing a new framework for embodied AI. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• CORE Recommender
• DagsHub
• EPIC-KITCHENS-100
• Gotit.pub
• GTEA Gaze+
• Hugging Face
• ScienceCast

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →