DVIDIA Skillspace lab
CPU · local execution · benchmark v1

Replay one measured scene.

Install a placement skill, create a scene, then watch the arm ground its motion in the object and target positions.

6 joints + 1 jaw command
objecttargetdrag to orbit
Preparing environment…0.00 s
Task outcome—
Object to target—
Simulation / wall—
Execution time—
Arm and gripper feedback

Create an environment to inspect joint positions and gripper contacts.

Six arm joints rotate in radians. Two jaw slides share one width command in metres. Placement keeps the tool upright; exact object pose and contact feedback come from the simulator.

Recorded native simulation: learned joint-movement mapping with authored task phases, gripper and recovery, plus privileged simulator observations. Jaw contacts move the free box. One exact scene; no hardware qualification or human-video action learning.

DVIDIA research · saved simulation evidence · no live physics