RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots

Nasiriany, Maddukuri, Zhang, Parikh, Lo, Joshi, Mandlekar, Zhu · 2024 · arXiv preprint · arXiv:2406.02523 · PDF

Dongyu supplement. Added as a candidate missing/adjacent dataset paper after searching primary arXiv sources. This is a draft summary for triage, not a full paper read.

One-liner. A large-scale simulated kitchen manipulation framework with assets, 100 tasks, LLM-guided composite tasks, demonstrations, and open-source code.

Setup

• Datasets / benchmarks: Introduces RoboCasa: realistic kitchen simulation with thousands of 3D assets, over 150 object categories, dozens of appliances/furniture, 100 evaluation tasks, human demonstrations, and automated trajectory generation. Code/videos are available at the project site.
• Hardware / simulator: Simulation framework for everyday kitchen tasks; intended to scale environments, tasks, and datasets for robot learning.

Method

Uses realistic simulation plus generative AI tools for assets/textures and large language models for composite task generation.

Why it matters for the map

Useful comparator for benchmark design, especially if the project proposes a task x modality x annotation matrix.

Limitations / open questions

Simulation and visual household manipulation; not a non-visual sensor dataset.

Source note

arXiv search result reports assets, 100 tasks, demos, LLM task generation, and open-source code.