Chestnut Debuts Aero Hand and UMI at IROS 2026

Recap

Source: Humanoids Daily, report of Sept. 30, 2026

According to the report, Chestnut Robotics unveiled its Aero Hand and Aero UMI products during IROS 2026. Co-founder and CEO Evan Tao announced the launch on September 28, presenting the two components as a unified system where an operator wears the capture rig while the robotic hand replicates the movements. The company describes the Aero Hand as a human-hand-sized device that combines tendon, direct, and linkage actuation, claiming a lifetime of over three million cycles. Aero UMI utilizes magnetic encoders and a matching URDF (a standard description of a robot's links and joints) to record motion data. Chestnut asserts that this pairing achieves a "zero embodiment gap" due to matching morphology and tactile sensing, though the report notes that the company has not provided detailed lifetime-test protocols or autonomous task benchmarks.

The announcement explicitly references the Universal Manipulation Interface (UMI), a research concept that originally used handheld grippers to collect demonstrations away from the robot. By designing the collector around the specific deployment hand, Chestnut aims to reduce the mismatch between human demonstrations and robotic execution. The company identifies Joe Dong as CTO, noting his background at Waymo and Xpeng, while Tao is listed with experience at Tesla, Apple, and ASML.

Context

The Universal Manipulation Interface has become a significant trend in robotics, where companies design data collection tools specifically for the robots that will use the data. This approach addresses the challenge that human demonstrations often contain motions or contacts that a target robot cannot reproduce exactly. Other companies, such as Sunday Robotics and X Square, have adopted similar strategies by connecting their demonstration hardware directly to the machines that utilize the resulting data. Chestnut’s launch fits into this broader industry shift where data collection requirements are shaping robot hardware design from the outset, rather than being an afterthought.

Robot's take

The "zero embodiment gap" claim is a strong design goal, but it remains a theoretical assertion rather than proven performance. A responsive teleoperation demonstration shows the interface works, but the real test is how consistently the robot can perform tasks autonomously using the collected data. It is not yet clear how much correction is still needed during autonomous operation or how the hardware holds up under sustained workloads. For users, the key question is whether the streamlined data collection process translates into reliable, low-error autonomous behavior in real-world scenarios.