Anyone using imitation learning purely from human video for grasp policies?

[sarah.santos3]

@carlossanchez Just to be precise about one thing: Wet, oily, or otherwise low-friction objects are still a genuine edge case for most current hands, since tactile sensing and grasp-force controllers are typically tuned and validated on dry, higher-friction test objects. Figure's fourth-generation Dexterous Hand (on Figure 02/03) reportedly offers 16 degrees of freedom per hand with sensors integrated into each finger, aimed at fine force control tasks like handling small electronic components without crushing them.

[james15]

@sarah.santos3 Minor factual note: Bimanual manipulation - two arms coordinating on one task - is harder than it looks mostly because of the added degrees of freedom and the timing/force coordination required; a lot of 'two-handed' demos are actually closer to two independent single-hand tasks done in sequence.

[emily.kumar]

@james15 Minor factual note: Palm sensing gets less attention than fingertip sensing, but a lot of power grasps (holding a box, a tool handle) rely more on palm and lateral finger contact than fingertip contact, so under-sensing the palm can leave a real blind spot in grasp confidence.

[matthew43]

@emily.kumar Worth being a little skeptical of the marketing angle here. Grasp planning for deformable or non-rigid objects (bags, cables, cloth) remains one of the genuinely unsolved problems in manipulation - rigid-body grasp models simply don't capture how the object will behave once contact starts.

[johnrossi]

@matthew43 Speaking from personal experience here, Vision-based grasp confidence estimation (predicting success before attempting a grasp) and tactile-based confirmation (confirming after contact) are complementary rather than competing - vision helps you choose a grasp, tactile tells you if it actually worked.

[jchen]

Can I ask a dumb follow-up - Wet, oily, or otherwise low-friction objects are still a genuine edge case for most current hands, since tactile sensing and grasp-force controllers are typically tuned and validated on dry, higher-friction test objects.

[matthew43]

@jchen Minor factual note: Imitation learning from human demonstration video (without robot teleoperation data) is an appealing way to scale up training data cheaply, but it runs into the embodiment gap - human hand kinematics and force profiles don't map directly onto a robot hand's very different mechanism.

[kim37]

Agreed, and I'd add: Object occlusion by the robot's own hand during the final approach to a grasp is a common and annoying perception problem - the closer the hand gets to a good grasp position, the more it blocks the camera's view of exactly what it's about to grab.

[young58]

@kim37 One nitpick - Figure's fourth-generation Dexterous Hand (on Figure 02/03) reportedly offers 16 degrees of freedom per hand with sensors integrated into each finger, aimed at fine force control tasks like handling small electronic components without crushing them.

[matthew43]

To answer this directly: Compliant wrists that absorb impact during a bad approach or misjudged contact reduce mechanical stress on the whole arm, which matters a lot for long-term reliability even though it's a less visible feature than the hand itself.