Can GPTs generate infinite and diverse data for robotics? Introducing RoboGen, a generative robotic agent that keeps proposing new tasks, creating corresponding environments and acquiring novel skills autonomously! code: 👇🧵 (better with audio)

RoboGen ( is an automated pipeline for endless skill acquisition. It first proposes new tasks to solve, then generates required environments, configurations, and skill learning supervisions. Lastly, it teaches itself skills to solve the self-proposed task.

We first proposed this paradigm, termed as Generative Simulation, in a white paper released back in May: This new paradigm aims at auto-generating all the stages required for automated robot learning, leading to an infinite-data era for robotics!

Foundation models such as LLMs have encoded massive and general knowledge. Our work attempts to transfer these extensive and versatile knowledge to the dynamic world of robotics, coupled with physics-grounded simulations.

RoboGen is powered by Genesis, a multi-material multi-solver physics simulation platform that will natively support generative simulation: It will be made publicly available soon!

There's still a long way to go, but this is a first crucial step towards this ambitious goal! Kudos to amazing @YufeiWang25 leading the work and other collaborators @johnsonwang0810, @WangYian01, @KaterinaFragiad, Zackory Erickson, @davheld and @gan_chuang

Everyone is getting ahead with AI. You should be too. Summarize documents, craft emails, and generate custom content instantly with this powerful tool. It's like having ChatGPT tailored for your job. Try it for free.

Congratulations! Looking forward for more application based on RoboGen.

Cool!!!!!

Looks great

@fdellaert ... with a network that starts to catastrophically forget immediately! Lol, who could have thought that, it turns out what we need is just more data. Great!

Such a great concept!