Beyond 1.1 — more value.
We refined every rough edge of 1.1 and added a variable-controlled Cell environment. A next-generation humanoid arm built to run reinforcement and reward learning the way they should.
The full evaluation setup, with OpenArm 2.0 installed inside the Cell.
A zero-position jig that physically locks the gripper to its exact CAD-defined angles, calibrating out assembly errors and tolerances for a consistent dataset.
Research-grade performance at a fraction of a high-end humanoid’s cost.
Leader and follower trade force back and forth, conveying the feel of contact and pressure as-is.
Everything is open — CAD, firmware, and control code.
Freely test reinforcement learning, imitation learning, and other AI approaches.
A simpler actuation makes the gripper noticeably smaller. The in-hand camera captures the exact view at the moment of grasping, and the fingers can be swapped to fit each task.
Add the recommended top-down ZED stereo camera to the gripper’s built-in in-hand camera, and you capture depth-rich training data.
Pair the variable-controlled OpenArm Cell with the motorless KER leader arm and you get a sturdier learning setup that anyone can reproduce identically.
An evaluation cell that keeps background, lighting, cameras, and arm position identical every time. It creates a standard environment for fair, automated model comparison.
A motorless leader arm (KER) with the exact same joint structure as 2.0. No motors means it’s light, so you can operate it for hours without fatigue.
Check the configurations yourself, add the setup you need, and leave a pre-order. Pricing is provided as an individual quote based on your configuration.
Browse pre-order options →