RoboCasa

RoboCasa is a large-scale simulation framework for training generally capable robots to perform everyday tasks. It was originally released in 2024 by UT Austin researchers. The latest iteration, RoboCasa365, builds upon the original release with significant new functionalities to support large-scale training and benchmarking in sim. Four pillars underlie RoboCasa365:

• Diverse tasks: 365 tasks created with the guidance of large language models
• Diverse assets: including 2,500+ kitchen scenes and 3,200+ 3D objects
• High-quality demonstrations: including 600+ hours of human demonstrations in addition to 1,600+ hours of robot datasets created with automated trajectory tools
• Benchmarking support: popular policy learning methods including Diffusion Policy, pi, and GR00T, plus user-submitted models on the leaderboard

This guide contains information about installation and setup. Please refer to the following resources for additional information:

[Home page]   [Documentation]   [RoboCasa365 Paper]   [Original RoboCasa Paper]   [Leaderboard]

---

Updates

• [10/28/2026] The leaderboard now accepts agentic submissions: new models or novel agentic harnesses built around a base policy.
• [10/14/2026] Released the subtask annotation pipeline, with labelers covering all 300 composite tasks. It generates per-frame subtask labels directly from each demo's simulation state.
• [7/7/2026] Our target composite task datasets have been updated to include per-frame subtask annotations. Every timestep is labeled with a subtask index, atomic-skill name, stage (i.e. pick / place / navigate), and a natural-language instruction, to support hierarchical policy learning.
• [5/12/2026] v1.0.1: Updated horizon lengths (1.5x increase) across all tasks for consistency. Please update to the latest version for running evals.
• [4/6/2026] RoboCasa365 Leaderboard published, comparing generalist robot policies on the multi-task learning benchmark and open to community model submissions.
• [2/18/2026] v1.0: RoboCasa365 release, with 365 tasks, 2500+ kitchen scenes, 2200+ hours of robot demonstration data, and benchmarking support.
• [10/31/2024] v0.2: using RoboSuite v1.5 as the backend, with improved support for custom robot composition, composite controllers, more teleoperation devices, photo-realistic rendering.

Table of Contents

• Updates
• Installation
• Basic Usage
• Tasks, datasets, policy learning, and additional use cases
• Questions
• License
• Citation

---

Installation

RoboCasa works across all major computing platforms. The easiest way to set up is through the Anaconda package management system. Follow the instructions below to install:

1. Set up conda environment:

   conda create -c conda-forge -n robocasa python=3.11
   

2. Activate conda environment:

   conda activate robocasa
   

3. Clone and setup robosuite dependency (important: use the master branch!):

   git clone https://github.com/ARISE-Initiative/robosuite
   cd robosuite
   pip install -e .
   

4. Clone and setup this repo:

   cd ..
   git clone https://github.com/robocasa/robocasa
   cd robocasa
   pip install -e .
   pip install pre-commit; pre-commit install           # Optional: set up code formatter.
   
   (optional: if running into issues with numba/numpy, run: conda install -c numba numba=0.56.4 -y)
   

5. Install the package and download assets:

   python -m robocasa.scripts.setup_macros              # Set up system variables.
   python -m robocasa.scripts.download_kitchen_assets   # Caution: Assets to be downloaded are around 10GB.
   

---

Basic Usage

Gym wrapper

You can create environments using gym wrappers and run rollouts:

import gymnasium as gym
import robocasa
from robocasa.utils.env_utils import run_random_rollouts

env = gym.make(
    "robocasa/PickPlaceCounterToCabinet",
    split="pretrain", # use 'pretrain' or 'target' kitchen scenes and objects
    seed=0 # seed environment as needed. set seed=None to run unseeded
)

# run rollouts with random actions and save video
run_random_rollouts(
    env, num_rollouts=3, num_steps=100, video_path="/tmp/test.mp4"
)

Play back sample demonstrations of tasks

(Mac users: for these scripts, prepend the "python" command with "mj": mjpython ...)

Select a task and play back a sample demonstration for the selected task:

python -m robocasa.demos.demo_tasks

Explore kitchen scenes

Explore 2500+ kitchen scenes:

python -m robocasa.demos.demo_kitchen_scenes

Explore library of 2500+ objects

View and interact with both human-designed and AI-generated objects:

python -m robocasa.demos.demo_objects

Note: By default, this demo shows objaverse objects. To view AI-generated objects, add the flag --obj_types aigen.

Teleoperate the robot

Control the robot directly, either through a keyboard controller or spacemouse. This script renders the robot semi-translucent in order to minimize occlusions and enable better visibility.

python -m robocasa.demos.demo_teleop

Note: If using SpaceMouse, you may need to modify the product ID to your appropriate model, setting SPACEMOUSE_PRODUCT_ID in robocasa/macros_private.py.

---

Tasks, datasets, policy learning, and additional use cases

Please refer to the documentation page for information about tasks, datasets, benchmarking, and more.

---

Questions

If you have any questions about RoboCasa, reach out at [email protected].

---

License

Code: MIT License

Assets and Datasets: CC BY 4.0

---

Citation

RoboCasa365:

@inproceedings{robocasa365,
  title={RoboCasa365: A Large-Scale Simulation Framework for Training and Benchmarking Generalist Robots},
  author={Soroush Nasiriany and Sepehr Nasiriany and Abhiram Maddukuri and Yuke Zhu},
  booktitle={International Conference on Learning Representations (ICLR)},
  year={2026}
}

RoboCasa (Original Release):

@inproceedings{robocasa2024,
  title={RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots},
  author={Soroush Nasiriany and Abhiram Maddukuri and Lance Zhang and Adeet Parikh and Aaron Lo and Abhishek Joshi and Ajay Mandlekar and Yuke Zhu},
  booktitle={Robotics: Science and Systems (RSS)},
  year={2024}
}