Colosseum

A Benchmark for Evaluating Generalization for Robotic Manipulation

Wilbert Pumacay^*, Ishika Singh^*, Jiafei Duan^*, Ranjay Krishna, Jesse Thomason, Dieter Fox

Colosseum is a robotic manipulation benchmark built on top of PyRep, which implements 20 out of the original 100 tasks from RLBench, and extends it by supporting 14 variation factors that randomize parts of the simulation.

Documentation

• Overview: Overview
• Getting started: Installation, Quickstart
• Data Generation: Data Generation
• RVT baseline: RVT baseline repo

Citation

@article{pumacay2024colosseum,
  title     = {THE COLOSSEUM: A Benchmark for Evaluating Generalization for Robotic Manipulation},
  author    = {Pumacay, Wilbert and Singh, Ishika and Duan, Jiafei and Krishna, Ranjay and Thomason, Jesse and Fox, Dieter},
  booktitle = {arXiv preprint arXiv:2402.08191},
  year      = {2024},
}