Aero Hand Open: A Simulation-Ready Tendon-Driven Anthropomorphic Hand for Dexterous Manipulation

Forum topic · 小凯 · 2026-09-01

Summary

Researchers present Aero Hand Open, a tendon-driven anthropomorphic robotic hand released as simulation-ready for reinforcement learning and dexterous manipulation research. Tendon-driven designs keep costs down by routing force through cables, which allows smaller, cheaper motors placed off-joint, and lets a single motor drive multiple joints. However, underactuated cable transmissions are hard to model in simulators and their coupled joints cannot be commanded independently, making learning harder than on direct-drive hands. The release addresses this with three components: a simulation model that reproduces the cable transmission itself; an identified actuation mapping connecting the model to motor commands bidirectionally, including a three-way coupling for the thumb; and a reinforcement learning training package. Together these enable policies trained fully in simulation to run directly on the hardware with zero-shot transfer—no fine-tuning or state estimation required. The full stack is open-sourced: mechanical design, simulation model, identified mapping, training environments, and deployment code.

Overview

Field: Machine Learning / Robotics Authors: Nan Wang, Mohit Yadav, Jonathan Wulff, Aidan Rosenbaum, Aidan Rosenbaum, Kezhou Chen, Yuvan Sharma, Xu Dong, Yiwei Tao arXiv: 2608.28578

Why tendon-driven hands are affordable

Tendon-driven hands are anthropomorphic, and moving the actuators off the joints is what makes a hand of this capability affordable to build. Two effects produce the saving:

Routing force through a cable removes the requirement that a motor fit inside the joint it drives, so smaller and cheaper motors suffice.
One motor can drive several joints through a single cable, so fewer motors are needed.

The learning challenge

Tendon-driven hands are harder to learn on than a direct-drive hand:

The underactuated transmission that produces the cost saving is itself difficult to represent in a simulator.
Joints driven by a single cable are not independently commandable.

What ships with Aero Hand Open

Aero Hand Open is a tendon-driven anthropomorphic hand released simulation-ready, with three components:

Simulation model — reproduces the cable transmission itself.
Identified actuation mapping — connects the simulation model to motor commands bidirectionally, including a three-way coupling for the thumb.
Reinforcement learning package — for training policies for the hand.

Combined, these allow policies to be trained entirely in simulation and run directly on the hand, with no fine-tuning and no state estimation required.

Open-source release

The authors release the full stack: mechanical design, simulation model, identified mapping, training environments, and deployment stack.

--- *Auto-collected on 2026-09-01.*