Home › Robotics › ai2thor

ai2thor - AI Robotics Tool

Overview

ai2thor is an open-source simulated environment for embodied and visual AI research, providing high-quality interactive indoor scenes implemented in Unity and exposed via a Python API. The platform supplies richly detailed room layouts (kitchens, living rooms, bedrooms, bathrooms) populated with household objects that have persistent states and affordances, enabling experiments in navigation, manipulation, object search, and common-sense reasoning. Sensory outputs include RGB rendering, depth, instance segmentation, and per-object metadata (positions, rotations, states), and the environment supports fine-grained actions such as move, look, pickup/put, open/close, and toggle. Designed for reproducible research and integration with learning frameworks, ai2thor runs a local Unity executable that the Python controller speaks to over sockets, and it comes with example scripts for agent training and evaluation. The project is actively maintained (see GitHub) and distributed under an Apache-2.0 license, making it suitable for academic and commercial experimentation. Common uses include reinforcement learning baselines, imitation learning datasets, benchmarking embodied perception models, and developing multi-step manipulation pipelines with precise object state tracking.

GitHub Statistics

  • Stars: 1,626
  • Forks: 265
  • Contributors: 24
  • License: Apache-2.0
  • Primary Language: C#
  • Last Updated: 2025-05-29T18:14:02Z
  • Latest Release: 5.0.0

According to the GitHub repository, ai2thor is actively maintained with the most recent commit on 2025-05-29 and is licensed under Apache-2.0. Repository metrics: 1,626 stars, 265 forks, and 24 contributors, indicating a modest but engaged open-source community. The codebase contains Unity scene assets, Python controller bindings, and example notebooks and training scripts; issues and pull requests show ongoing maintenance and feature additions. The contributor count and recent commits suggest steady development, while forks and stars reflect widespread academic use.

Installation

Install via pip:

pip install ai2thor
python -c "import ai2thor.controller as c; controller = c.Controller(); controller.start(); controller.stop()"

Key Features

  • High-fidelity indoor scenes (kitchen, living room, bedroom, bathroom) with household objects
  • Multiple sensory modalities: RGB, depth, instance segmentation, and per-frame metadata
  • Interactive object affordances: pickup, put, open, close, toggle, move, and look actions
  • Python Controller API that communicates with a local Unity executable over sockets
  • Example scripts for RL training and evaluation; suitable for reproducible embodied AI experiments

Community

ai2thor has an active GitHub community (1,626 stars, 265 forks, 24 contributors) and is maintained under Apache-2.0. The repository hosts example code and discussion via issues and pull requests; it is commonly used in embodied AI research for reproducible experiments and benchmarks.

Last Refreshed: 2026-01-09

Key Information

  • Category: Robotics
  • Type: AI Robotics Tool
Visit Official Website

Related Tools

LeRobot

A comprehensive suite of models, datasets, and tools for real-world robotics, providing end-to-end learning resources in PyTorch for robotics AI applications.
FAST: Efficient Action Tokenization for Vision-Language-Action Models

A Hugging Face repository that provides a universal action tokenizer (FAST+) designed to map sequences of robot actions into dense, discrete tokens for training autoregressive vision-language-action models. It supports both usage of a pre-trained tokenizer and easy custom training on users' own data.
Genesis

Genesis is a physics simulation platform and generative data engine for general-purpose robotics and embodied AI learning. It features a universal physics engine, ultra-fast robotics simulation, photorealistic rendering, and a modular generative framework that converts natural language prompts into various data modalities.
NVIDIA Isaac GR00T N1

An open foundation model for generalized humanoid robot reasoning and manipulation. It accepts multimodal inputs (language and images) and uses a vision-language foundation model combined with a diffusion transformer head to denoise continuous actions, enabling robot control and fine-tuning for specific tasks and embodiments.
NVIDIA Isaac GR00T N1

The world’s first open foundation model for generalized humanoid reasoning and skills, aimed at advancing robotics systems and autonomous vehicle technology.
Reachy Mini

Reachy Mini is an expressive, open‐source robot designed for human-robot interaction, creative coding, and AI experimentation. Programmable in Python (with upcoming support for additional languages), it integrates with state-of-the-art Hugging Face models for speech, vision, and personality. Available as a kit in a lite or wireless version, it is intended for AI developers, researchers, educators, and hobbyists to build, test, and share real-world AI applications.