Aero Hand Open: A Simulation-Ready Tendon-Driven Hand for Dexterous Manipulation Learning
Tendon-driven robot hands route motors off the joints via cables, cutting costs by using smaller motors and having one motor drive multiple joints. Aero Hand Open solves a hard problem: these hands are difficult to train because the cable transmission is tricky to simulate and joints aren't independently controlled. The team released the design, simulation model, actuation mapping, and reinforcement learning package together. Training happens entirely in simulation with no fine-tuning needed on the actual hand. Could make dexterous manipulation affordable for factories and beyond.
Researchers released Aero Hand Open, a tendon-driven robot hand designed to train manipulation skills entirely in simulation before deployment on hardware. The key innovation: by routing force through cables instead of placing motors inside joints, the hand uses fewer and cheaper motors while maintaining human-like dexterity. The challenge has always been that this cable transmission system is difficult to simulate accurately, and joints controlled by a single cable can't be commanded independently. The team solved this by shipping three components together: a simulation model that captures the cable mechanics, an actuation map that translates motor commands to joint movements, and a reinforcement learning package for policy training. The result is that a robot can learn complex object manipulation tasks in virtual environments and transfer directly to the physical hand without fine-tuning. They've released the full stack: mechanical design, simulation, mapping, training environment, and deployment code. This could lower barriers to dexterous robotics in manufacturing and domestic settings.