A Neural Network Model for Planning Force and Posture in Three-digit Object Holding

It is hypothesized that suitable forces and postures of digits for holding an object are planned simultaneously based on the same motor criterion, using an internal representation of the object. To demonstrate the computational potential of this hypothesis, a neural network model for learning and planning forces and postures of digits in object holding tasks is proposed. Integrating visual 2D images of 9 objects that are different in size and orientation with forces and postures of digits, the neural network internally represents the size and the orientation of the objects and memorizes 5 motor patterns per object. Moreover, the neural network restores forces and postures in natural holding tasks according to the same motor strategy in which a norm of forces of digits is minimized. These results support the computational plausibility of our hypothesis.