Existing grasp synthesis methods are either analytical or data-driven. The former one is oftentimes limited to specific application scope. The latter one depends heavily on demonstrations, thus suffers from generalization issues; e.g., models trained with human grasp data would be difficult to transfer to 3-finger grippers. To tackle these deficiencies, we formulate a fast and differentiable force closure estimation method, capable of producing diverse and physically stable grasps with arbitrary hand structures, without any training data. Although force closure has commonly served as a measure of grasp quality, it has not been widely adopted as an optimization objective for grasp synthesis primarily due to its high computational complexity; in comparison, the proposed differentiable method can test a force closure within 4ms. In experiments, we validate the proposed method's efficacy in 8 different settings.

3D View of Synthesized Example

Please follow this link to view the final state of the synthesis trajectory in the above video.

3D View Before/After Refinement

The following table contains the 3D views of corresponding examples depicted in Fig 6.
Click on each image to view them in 3D.

3D View Grasp of Different Hand

Click on each image to view them in 3D.

Synthesizing Specific Grasp Type

As mentioned in Section IV-B, the choice of contact points on the hand surface primarily determines the grasp type. Hence, specific grasp types can be synthesized by mandating the choice of contact point.

(a)(d)(g)(j) show the query contact points in red, each followed by two synthesized examples using the queried contact points.