Learning and generation of goal-directed arm reaching from scratch

In this paper, we propose a computational model for arm reaching control and learning. Our model describes not only the mechanism of motor control but also that of learning. Although several motor control models have been proposed to explain the control mechanism underlying well-trained arm reaching movements, it has not been fully considered how the central nervous system (CNS) learns to control our body. One of the great abilities of the CNS is that it can learn by itself how to control our body to execute required tasks. Our model is designed to improve the performance of control in a trial-and-error manner which is commonly seen in human’s motor skill learning. In this paper, we focus on a reaching task in the sagittal plane and show that our model can learn and generate accurate reaching toward various target points without prior knowledge of arm dynamics. Furthermore, by comparing the movement trajectories with those made by human subjects, we show that our model can reproduce human-like reaching motions without specifying desired trajectories.