Explore / Case study / SayCan — Grounded Robot Skill Selection

SayCan — Grounded Robot Skill Selection

Demo only

Combine task relevance and physical feasibility to choose a robot skill.

by admin · listed 5 days ago

⚡ What System1 does here

A language model scores how useful each candidate skill is; skill value functions estimate feasibility. The system combines these scores to select the next skill, which a robot policy executes. No Jev integration is established.

About

A research system that turns everyday requests into sequences of available robot skills. The project demonstrates mobile manipulation in kitchen environments and provides an open-source simulated tabletop implementation. Evidence and limits: Research demonstrations depend on the available skills, value functions and hardware. The simulated release is not a complete physical-robot deployment or a general safety guarantee. We reviewed sources and media URLs; we did not run the project. Original project images and demonstrations (with attribution):

SayCan method diagram
SayCan method diagram ↗

Official PaLM-SayCan demonstration ↗

Official robot task demonstration ↗

Sources: https://say-can.github.io/ https://arxiv.org/abs/2204.01691 https://github.com/google-research/google-research/tree/master/saycan https://youtu.be/ysFav0b472w

🧪 Community test reports

0 reports · 0 passed on the real model
No one has tested this yet. Be the first.

💬 Discussions

New post →
No discussions yet.