Hand Synergies

The many joints of a human hand tend to move together in a few fixed combinations, describable by a handful of synergy parameters.

The idea of hand synergies comes from neuroscience: Santello, Flanders, and Soechting had subjects imagine grasping 57 common objects in 1998, measured the 15 joint angles of the fingers and thumb, and ran principal component analysis (PCA), finding that just the first two principal components explained over 80% of the variance. In other words, although the human hand has many degrees of freedom, most grasp postures fall within a low-dimensional subspace, with the higher-order components only fine-tuning the details. Robotics has borrowed this idea: the GraspIt! simulator's eigengrasp uses a few synergy amplitudes in place of the full set of joint angles when searching for grasps. On the hardware side, the Pisa/IIT SoftHand (Catalano et al., 2014) was designed around 'adaptive synergies,' and a related product, the qb SoftHand Research, drives 19 humanlike degrees of freedom with just one motor. When learning dexterous manipulation, synergies can also be used to compress the action space.

ExampleThe qb SoftHand has only one motor, so the only control input is how much to close the hand; the five fingers close together in proportions set by their tendon routing, letting it grasp objects of many different shapes.

Also called

Postural Synergies, Eigengrasps

Related

Dexterous Hand · Underactuation · Tendon-Driven Actuation · Motion Retargeting · Grasp Taxonomy (Power Grasp vs. Precision Grasp / Pinch) · Data Glove

Sources

Santello, Flanders & Soechting 1998, Postural Hand Synergies for Tool Use（J Neurosci，PMC 全文） (Chinese)
GraspIt! Documentation: Eigengrasps
qbrobotics: qb SoftHand Research