google-research/ravens
A PyBullet simulation benchmark for training robotic pick-and-place agents using deep learning.
Not currently ranked — collecting fresh signals.
star history
Ravens is a collection of 10 tabletop robotic manipulation tasks designed for learning vision-based pick-and-place policies. It provides a Gym-like API combining scripted oracle demonstrators for imitation learning with reward functions supporting reinforcement learning. The benchmark includes diverse tasks like block insertion, pyramid stacking, palletizing, and deformable rope manipulation, testing generalization to unseen objects and multi-step sequential planning.
Frequently asked
- What is google-research/ravens?
- A PyBullet simulation benchmark for training robotic pick-and-place agents using deep learning.
- Is ravens open source?
- Yes — google-research/ravens is open source, released under the Apache-2.0 license.
- What language is ravens written in?
- google-research/ravens is primarily written in Python.
- How popular is ravens?
- google-research/ravens has 631 stars on GitHub.
- Where can I find ravens?
- google-research/ravens is on GitHub at https://github.com/google-research/ravens.