SeriesFusion
Science, curated & edited by AI
First Ever  /  AI

Robots are learning how to give you a sponge bath or scratch an itch by 'dreaming' about it after reading descriptions of how it feels.

This bypasses the need for massive, dangerous, or impractical real-world datasets for physical human-robot interaction. By using generative models to build 3D training grounds, robots can learn complex social and physical skills entirely in virtual space.

Original Paper

Generative Simulation for Policy Learning in Physical Human-Robot Interaction

Junxiang Wang, Xinwen Xu, Tiancheng Wu, Julian Millan, Nir Pechuk, Zackory Erickson

arXiv  ·  2604.08664

Developing autonomous physical human-robot interaction (pHRI) systems is limited by the scarcity of large-scale training data to learn robust robot behaviors for real-world applications. In this paper, we introduce a zero-shot "text2sim2real" generative simulation framework that automatically synthesizes diverse pHRI scenarios from high-level natural-language prompts. Leveraging Large Language Models (LLMs) and Vision-Language Models (VLMs), our pipeline procedurally generates soft-body human mo