Musculoskeletal Bias on Infant Sensorimotor Development Driven by Predictive Learning

In the early developmental stages, infants learn to control complex and redundant body movements using sensory inputs. Reaching with the arm and hand position control are fundamental features of motor development. However, it remains unclear how infants aquire such kind ofmotor control. In the current study, we propose a network model that learns the relationship between motor commands and visual and proprioceptive sensory input using predictive learning to perform reaching based on infantile musculoskeletal body. Based on assumption that human motion is generated from combinations of muscle activation patterns deïňĄned as a motor primitive, we examine the contribution of motor primitive to sensorimotor development. The results of this predictive learning model revealed that acquisition of motor primitives promoted sensorimotor learning of reaching.