Fast processing of shape information for online control of grasping

Abstract

When picking up objects with a precision grip, we tend to grasp at contact points that minimize slippage and torsion. For smooth objects with complex shape, the set of ideal grasp points is limited and depends on the particular shape. In normal conditions, grasp points could potentially be determined during a planning stage prior to movement. This study tested whether the visual motor system can identify appropriate grasp points during an ongoing movement. We used a perturbation method to isolate online control of grasping. Observers reached to grasp virtual planar objects with varied shapes. On a subset of trials, the object was changed after the onset of movement by either rotating the object by 45° (Experiment 1) or by replacing with an entirely different object (Experiment 2). Optimality of the final grasp points was evaluated by two measurements: torque control and force closure control (Blake, 1992). With or without perturbations, grasp axes passed close to the center of mass (1mm deviation on average), and had small angular deviations from the surface normals (average force closure angle of 25°). Corrective adjustments in response to perturbations were detectable within 100ms of perturbation onset. There was no slowing of the hand or increase in movement duration on perturbed trials, and final grasp points showed no reduction in optimality relative to unperturbed trials with the same object. These results demonstrate that the visual-motor system is capable of rapid, online processing of shape information for guiding the hand to optimal contact points during grasping.