D'Claw
D'Claw (pronounced as "The Claw") is a nine degree of freedom platform for exploring learning-based techniques in dextrous manipulation.
Hardware
Simulation
Schematics
D'Lantern
D'Lantern
D'Lantern is lightweight mounting station for D'Claw. D'Lantern is designed with portability in mind to facilitate anywhere operations. It's easy to assemble/ disassemble and can fit in a slim laptop bag. Taking your hardware demos to conferences has never been this easy!
Fully assembled unit
Disassembled unit
Disassembled unit inside a laptop bag
Gallery
Gallery
Flexible Objects
Flexible Objects
On-hardware training with DAPG effectively learns to turn a flexible objects. We observe manipulation targeting the center of the valve where there is more rigidity. D'Claw is robust to sustain on-hardware training facilitating results on hard to simulate tasks.
Disturbance Rejection
Disturbance Rejection
A Sim2Real policy trained via Natural Policy Gradient on MuJoCo simulation with object perturbations (amongst others) being tested on hardware. We observe fingers collaborating to together to resist the external disturbances
Finger Holdout
Finger Holdout
A Sim2Real policy trained via Natural Policy Gradient on MuJoCo simulation with external perturbations (amongst others) being tested on hardware. We observe that free fingers fill in for the missing finger
Multi-Task
Multi-Task
On-hardware training with distributed version of EC-SAC leaning to turn multiple objects to arbitrary angles in conjunction by sharing experiences. Five tasks only need twice the amount of experience of single tasks, thanks to the multi-task formulation. In the video we observe five D'Claws turning different objects to 180 degree (picked for visual effectiveness, actual policy can turn to any angle)