Infant Neural Controller for Adaptive Sensory-Motor Coordination*

摘要

The human brain develops accurate sensory-motor coordination in the face of many unforeseen changes in the dimensions of the body, strength of the muscles, and placements of the sensory organs. This chapter presents a theory and implementation that suggest how three types of adaptive sensory-motor coordinations might be learned and maintained by animals as well as robot controllers. They are: locating stationary targets with movable sensors, reaching arbitrarily positioned and oriented targets in three-dimensional space with multijoint arms, and positioning an unforeseen payload with accurate and stable movements despite unknown sensor feedback delay. The chapter also presents the motivations, constraints, experiments, results, and interpretations of the INFANT controller. It provides a representation that accommodates both coordinates and satisfies all of the constraints for an invariant target position map. The chapter addresses the problem controlling a multijoint arm to reach objects in three-dimensional space.