A self-organizing neural model of motor equivalent reaching and tool use by a multijoint arm

摘要: This paper describes a self-organizing neural model for eye-hand coordination. Called the DIRECT model, it embodies solution of classical motor equivalence problem. Motor computations allow humans and other animals to flexibly employ an arm with more degrees freedom than space in which moves carry out spatially defined tasks under conditions that may require novel joint configurations. During babbling phase, endogenously generates movement commands activate correlated visual, spatial, information are used learn its internal coordinate transformations. After learning occurs, is capable controlling reaching movements prescribed spatial targets using many different combinations joints. When allowed visual feedback, can automatically perform, without additional learning, reaches tools variable lengths, clamped joints, distortions input by prism, unexpected perturbations. These compensatory occur within single accurate movement. No corrective needed. Blind feedback have also been simulated. The achieves competence transforming about target position end effector 3-D into body-centered representation direction must move contact target. vector adaptively transformed vector, represents rotations desired from present configuration. Properties compared psychophysical data on human movements, neurophysiological tuning curves neurons monkey cortex, alternative models control.