ALOHA 2

ALOHA 2 is an open-source, low-cost bimanual teleoperation hardware platform released in February 2024 by an Google DeepMind led team working with the original ALOHA authors at Stanford University. It is the second generation of ALOHA, an acronym for A Low-cost Open-source Hardware system for bimanual teleoperation, and it keeps the leader-follower puppeteering design while adding a redesigned low-friction gripper, passive gravity compensation, and a simplified frame to make large-scale manipulation data collection faster, more comfortable, and more reliable. ALOHA 2 is the data-collection workcell behind several high-profile robot learning efforts, including ALOHA Unleashed and parts of the vision-language-action (VLA) policy lineage, and all of its hardware designs plus a matched MuJoCo model are released openly. ^[1][2]

ALOHA 2 was introduced in a technical report by the "ALOHA 2 Team," a group of 25 authors that includes Tony Z. Zhao, Chelsea Finn, Pete Florence, Jonathan Tompson, Ayzaan Wahid, and Kevin Zakka. ^[1] The hardware, simulation models, and reference code are published on the project site aloha-2.github.io, and commercial kits are produced by Trossen Robotics: as of 2026 the ALOHA Stationary V2 cell (without laptop) lists at $27,999.99, and the single-arm ALOHA Solo starts at $8,999.95. ^[2][7] Across 2024 to 2026, ALOHA 2 has become one of the de facto reference cells for academic and industrial work on bimanual fine manipulation, in part because it costs roughly an order of magnitude less than research-grade dual-arm systems built around Franka Emika Pandas or Universal Robots UR5e arms. ^[1]

What is ALOHA 2?

ALOHA 2 is a bimanual parallel-jaw teleoperation workcell built from two larger "follower" arms and two smaller "leader" arms. A human operator backdrives, or "puppeteers," the two leader arms by hand, and the two follower arms mirror that motion in real time, while the system records joint positions, gripper states, and synchronized multi-view camera streams. ^[1] The follower arms are 6-DoF ViperX 300 arms and the leader arms are 6-DoF WidowX 250 arms, both from Trossen Robotics and both driven by Dynamixel servos from Robotis. ^[1][2] The goal of the platform is to let non-experts collect high-quality demonstration data for imitation learning at a scale and cost that prior bimanual rigs could not reach.

The ALOHA 2 paper frames the problem directly: "Diverse demonstration datasets have powered significant advances in robot learning, but the dexterity and scale of such data can be limited by the hardware cost, the hardware robustness, and the ease of teleoperation. We introduce ALOHA 2, an enhanced version of ALOHA that has greater performance, ergonomics, and robustness compared to the original design." ^[1]

Background

Original ALOHA (2023)

The first ALOHA system was developed at Stanford University by Tony Z. Zhao, Vikash Kumar, Sergey Levine, and Chelsea Finn, and described in the 2023 paper Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware. ^[3] The hardware paired two ViperX 300 follower arms with two WidowX 250 leader arms, all built around Dynamixel servos from Robotis. A human operator grasped the leader arms and moved them through a task; the follower arms mirrored the motion in real time, recording joint angles and RGB video from webcams plus a wrist camera. The full bill of materials came in under about $20,000, roughly 5 to 10 times cheaper than comparable bimanual research setups at the time. ^[3]

Alongside the hardware, the Stanford team released Action Chunking with Transformers (ACT), an imitation learning algorithm that predicts short sequences of future actions rather than a single next action. ACT helped the policies learn fine-grained skills such as opening a Ziploc bag, slotting a battery, and threading a zip tie from only about 50 demonstrations. ^[3]

Mobile ALOHA (2024)

In January 2024, Zhao and collaborators released Mobile ALOHA, which mounted the original ALOHA arms on a wheeled base with a powered torso lift. ^[4] The base let the system collect demonstrations of whole-body tasks such as cooking shrimp, calling an elevator, and rinsing a pan in a sink. Mobile ALOHA used the same teleoperation principle but added a wheeled base that the operator drove with a leader yoke. The Mobile ALOHA paper and viral demo videos drew significant attention on social media and helped trigger the wider 2024 wave of household robot learning work. ^[4]

How does ALOHA 2 improve on the original ALOHA?

ALOHA 2 was announced in a short technical report titled ALOHA 2: An Enhanced Low-Cost Hardware for Bimanual Teleoperation, released on arXiv (2405.02292) with a first version dated February 7, 2024, by the ALOHA 2 Team. ^[1] The stated goal was not to redesign the platform from scratch but to fix the small frictions that had emerged from a year of heavy use across many labs.

The most concrete gains are in the gripper and the operator interface. The redesigned leader gripper replaces the original scissor mechanism with a low-friction rail design and lower-gear-ratio motors (swapping the XL430-W250-T for the XC430-W150-T), which cuts the force needed to open and close the leader gripper by roughly 10 times, from about 14.68 N to about 0.84 N. ^[1] On the follower side, the new gripper can exert more than double the closing force of the old design, about 27.9 N versus 12.8 N, improving grasps on small or thin objects. ^[1] ALOHA 2 also adds passive gravity compensation using off-the-shelf adjustable hanging retractors rather than active software compensation; in a user study this passive approach let operators complete more shapes per minute (about 1.38 versus 0.97 for the software-based system). ^[1]

The main changes compared with the original ALOHA are summarised below. ^[1]

Scroll sideways for more →

The MuJoCo model is calibrated through system identification on 11 real trajectories, tuning proportional gain, damping, armature, joint friction, and torque limits so that the simulator matches the physical arms. ^[1] The paper emphasises that ALOHA 2 is intentionally not a step toward a productised robot; it remains a research data collection cell, and the total bill of materials stays in the same low-five-figure range as the original. ^[1]

What is ALOHA 2 used for?

ALOHA 2 became the data collection backbone for a string of high-profile robot learning papers in 2024 and 2025.

ALOHA Unleashed, released by Google DeepMind in October 2024, used a fleet of ALOHA 2 cells to gather data at a scale not previously reported for any bimanual platform: over 26,000 demonstrations across 5 real-world tasks, plus more than 2,000 demonstrations on 3 simulated tasks. ^[5] The tasks included dexterous, contact-rich skills such as tying shoelaces, hanging a shirt on a hanger, and inserting a gear into a recess. The policies were trained with a diffusion policy head on top of a transformer trunk, and the paper argues that the combination of large-scale ALOHA 2 data collection and expressive diffusion models is a "simple recipe" for learning challenging bimanual manipulation involving deformable objects. ^[5]

ALOHA 2 data also fed into the broader Open X-Embodiment effort and RT-X follow-up work. ^[9] Several demonstrations of dexterous bimanual skills, including cloth folding and small parts assembly, were collected on ALOHA 2 hardware before being co-trained with mobile manipulator data from other labs. The Octo generalist policy from Berkeley was also evaluated on ALOHA-style bimanual tasks, although Octo was trained primarily on the Open X-Embodiment corpus rather than on ALOHA 2 data alone.

Outside Google, Physical Intelligence used ALOHA-style teleoperation as one of several data sources during training of the π0 generalist policy and its successor π0.5. ^[6] The π0 technical report lists ALOHA-style data among the sources for the bimanual portion of its training mix, alongside data from other dual-arm systems. ^[6]

The combination of low cost, open hardware, and a growing body of pre-trained checkpoints means that a researcher can buy or build an ALOHA 2 cell, plug in a published policy, and reproduce headline tasks within days rather than months. That feedback loop is widely credited with accelerating the 2024 to 2025 wave of bimanual manipulation research.

Which labs use ALOHA 2?

ALOHA 2 is used by academic and industrial groups beyond the original Stanford and DeepMind teams. Stanford's IRIS lab, Berkeley's RAIL lab, MIT CSAIL, CMU, ETH Zurich, Tsinghua, Seoul National University, and the University of Tokyo all have ALOHA 2 cells in use as of 2025, according to the project page and published papers using the platform. ^[2] Trossen Robotics, which sells the hardware kits, reports that the platform is also used in industry research groups, although exact deployment numbers are not public. ^[7]

The Hugging Face LeRobot project ships official drivers and example policies for ALOHA 2, which has made it easier for hobbyists and smaller labs to onboard. ^[8] LeRobot's example notebooks reproduce ACT, diffusion policies, and a handful of small VLA fine-tunes on ALOHA 2 data. ^[8]

How does ALOHA 2 compare to other teleoperation platforms?

ALOHA 2 sits in a distinct part of the design space from the industrial dual-arm cells that preceded it. The table below compares it with several commonly cited alternatives. Figures are taken from vendor pages, peer-reviewed papers, and project websites, and reflect publicly reported numbers rather than internal estimates. ^[1][2][7]

Scroll sideways for more →

The ALOHA family trades absolute precision and payload for cost and openness. Franka and UR5 cells offer sub-millimetre repeatability and force-torque sensing out of the box, which matters for industrial tasks. ALOHA 2 relies on hobby-grade Dynamixel servos that are noticeably less stiff, but the puppeteering interface lets non-experts collect data with very little training, which is the bottleneck for imitation learning at scale. ^[1]

Is ALOHA 2 open source?

Yes. The ALOHA 2 paper states that the team open sources all hardware designs of ALOHA 2 with a detailed tutorial, together with a MuJoCo model of ALOHA 2 with system identification. ^[1] The hardware bill of materials, CAD, assembly instructions, and software stack are published on the project site, and the simulation model is distributed through MuJoCo Menagerie. ^[2] Because the designs are open and the parts are off-the-shelf, third parties such as Trossen Robotics sell pre-assembled kits, and the Hugging Face LeRobot library distributes drivers and reference policies. ^[7][8]

Reception

ALOHA 2 was received warmly by the robot learning community. Reviews on the Robot Learning Workshop at NeurIPS 2024 and at CoRL 2024 frequently cited ALOHA 2 as a baseline platform for new bimanual algorithms. Several authors of competing systems, including those working on mobile humanoids and on lower-cost single-arm rigs, have publicly credited ALOHA 2 with raising expectations for what a low-cost teleoperation cell should provide.

Criticism has focused on three areas. The Dynamixel-based arms still have limited payload of roughly one kilogram per arm, which rules out heavier household tasks. The cameras and timing stack, while improved, remain less precise than industrial vision systems, which makes some millimetre-scale insertion tasks difficult. And the open hardware design is harder to procure outside the United States because of customs and shipping issues with the servos and aluminium extrusions. The ALOHA 2 authors have acknowledged these limits in talks at CoRL and ICRA, and have framed the platform as deliberately optimised for breadth of skills rather than for absolute task difficulty.

In his 2024 talks at Stanford and at Google DeepMind, Tony Z. Zhao described ALOHA 2 as a deliberate step toward making bimanual manipulation feel like a commodity research substrate, similar to how MuJoCo became a commodity for simulated control. As of 2026, that framing appears to have largely held: papers that report new bimanual policies routinely run a baseline on ALOHA 2 hardware even when their main contribution is on a different platform.

See also

• ALOHA (robot system)
• Mobile ALOHA
• Imitation learning
• Diffusion policy
• Robot teleoperation
• Robot learning
• Vision-language-action model
• RT-2
• π0.5
• Google DeepMind
• Stanford University
• Chelsea Finn
• Sergey Levine

References

1. ALOHA 2 Team, Google DeepMind. *ALOHA 2: An Enhanced Low-Cost Hardware for Bimanual Teleoperation*. arXiv:2405.02292, first version 7 February 2024. <https://arxiv.org/abs/2405.02292> ↩
2. ALOHA 2 project page. <https://aloha-2.github.io/> ↩
3. Zhao, T. Z., Kumar, V., Levine, S., Finn, C. *Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware*. Robotics: Science and Systems, 2023. <https://tonyzhaozh.github.io/aloha/> ↩
4. Fu, Z., Zhao, T. Z., Finn, C. *Mobile ALOHA: Learning Bimanual Mobile Manipulation with Low-Cost Whole-Body Teleoperation*. CoRL 2024 / arXiv:2401.02117. <https://mobile-aloha.github.io/> ↩
5. ALOHA 2 Team / Google DeepMind. *ALOHA Unleashed: A Simple Recipe for Robot Dexterity*. CoRL 2024 / arXiv:2410.13126. <https://aloha-unleashed.github.io/> ↩
6. Physical Intelligence. *π0: A Vision-Language-Action Flow Model for General Robot Control*. 2024. <https://www.physicalintelligence.company/blog/pi0> ↩
7. Trossen Robotics, ALOHA 2 / ALOHA Stationary product page. <https://www.trossenrobotics.com/aloha-stationary> ↩
8. Hugging Face LeRobot, ALOHA 2 examples. <https://github.com/huggingface/lerobot> ↩
9. Open X-Embodiment Collaboration. *Open X-Embodiment: Robotic Learning Datasets and RT-X Models*. ICRA 2024. <https://arxiv.org/abs/2310.08864> ↩