RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation

Overview

RoboGen is a generative robotic agent that automatically learns diverse robotic skills at scale through generative simulation. The work explores how automated task generation can help expand the data available for robot learning.

Research direction

The project connects vision models and large language models with simulation to generate robot-learning tasks. It investigates a route toward learning a wider range of skills with less manual task setup.

My contribution

I built automated robot-task generation pipelines with vision models and LLMs, and implemented rendering and simulation components.

The work was accepted to ICML 2024; its first preprint appeared in 2023.