ALOHA — Action Chunking with Transformers
ALOHA (A Low-cost Open-source Hardware System for Bimanual Dexterous Manipulation) is a real-robot bimanual manipulation dataset and platform developed at Stanford University by Tony Zhao and collaborators. Released in 2023 under the MIT license, the dataset contains over 50,000 demonstration episodes for the ALOHA bimanual robot platform — a low-cost system built from two ViperX and two WidowX arms mounted on a shared base. Tasks include precise bimanual operations such as slot insertion, cable routing, opening packages, and surgical tool handling. The ACT (Action Chunking with Transformers) policy trained on this data demonstrated state-of-the-art performance on fine manipulation tasks. The dataset is distributed in HDF5 format and compatible with the LeRobot library. ALOHA sparked significant research into affordable bimanual manipulation hardware and remains a primary reference dataset for precise dual-arm manipulation policy learning.
| Year | 2023 |
|---|---|
| Episodes | 50,000 |
| Embodiments | ALOHA bimanual robot, ViperX 300s, WidowX 250s |
| Modalities | rgb, proprioception |
| View type | third-person, wrist-cam |
| Task categories | manipulation, bimanual, dexterous-manipulation, pick-and-place |
| Data format | hdf5, lerobot |
| License | MIT |
| Access | open — commercial use permitted |
| Maintainer | Stanford University, Tony Zhao |
| Origin country | US |
What is it?
ALOHA (A Low-cost Open-source Hardware System for Bimanual Dexterous Manipulation) is a real-robot bimanual manipulation dataset and platform developed at Stanford University by Tony Zhao and collaborators. Released in 2023 under MIT license, the dataset contains over 50,000 demonstration episodes for the ALOHA platform — a low-cost bimanual system built from two ViperX and two WidowX arms. The ACT (Action Chunking with Transformers) policy trained on ALOHA data demonstrated state-of-the-art performance on fine bimanual manipulation including slot insertion, cable routing, and surgical tool handling.
Who is it for?
ALOHA targets researchers working on bimanual dexterous manipulation, low-cost hardware development, and fine manipulation policy learning. It is particularly valuable for teams studying tasks that require two-arm coordination — a capability gap in most single-arm datasets. The low-cost platform design made high-quality bimanual manipulation accessible to academic labs without large hardware budgets.
Key specifications
- Episodes: 50,000+ demonstrations
- Embodiment: ALOHA bimanual robot (2x ViperX 300s + 2x WidowX 250s)
- Tasks: Bimanual coordination, slot insertion, cable routing, packaging, surgical tool handling
- Format: HDF5, LeRobot
- License: MIT — commercial use permitted
- Access: Open — Hugging Face
How it compares
ALOHA pioneered accessible bimanual manipulation at scale. EgoVerse covers more embodiments and tasks but uses single-arm platforms with egocentric cameras. DROID covers more episodes with Franka Panda but is single-arm. ALOHA's specific contribution is precise bimanual coordination tasks that no other major open dataset covers at comparable scale.
Limitations and access notes
The ALOHA platform uses WidowX and ViperX arms rather than full humanoid hands, limiting direct transfer to dexterous hand manipulation tasks. The platform is stationary — no locomotion component. MIT license permits unrestricted commercial use.
Linked professions
Frequently asked questions
What is the ALOHA robot platform?
ALOHA is a low-cost bimanual robot platform designed at Stanford University using four arms: two ViperX 300s and two WidowX 250s mounted on a shared base. The total hardware cost is approximately $20,000, making it one of the most affordable bimanual manipulation platforms available.
What is ACT (Action Chunking with Transformers)?
ACT is an imitation learning algorithm developed alongside the ALOHA platform that predicts action sequences (chunks) rather than individual actions at each timestep. It significantly outperformed previous imitation learning methods on ALOHA's fine manipulation tasks and influenced subsequent robot learning research.
Can ALOHA data be used commercially?
Yes. The ALOHA dataset is MIT licensed, permitting unrestricted commercial use, modification, and redistribution.
How does ALOHA relate to Mobile ALOHA?
Mobile ALOHA is an extended version of the ALOHA platform that adds a mobile base, enabling the bimanual system to move around and perform whole-body manipulation tasks like cooking full meals and doing laundry. The Mobile ALOHA dataset is a separate release from the original ALOHA dataset.
What bimanual tasks does ALOHA cover?
ALOHA covers precise bimanual coordination tasks including slot insertion, cable routing, opening packaging, handling surgical tools, pouring liquids between containers, and tasks requiring simultaneous two-arm control for object stabilisation.