30 NTT Technical Review 1. Introduction Humans are not consciously aware of the cognitive computation that is performed to produce motor com- mands for their physical movements, such as grasp- ing objects like cups and doorknobs. When we learn new sports, we never imagine the processes occurring in our brain as our performance improves. Infants cannot smoothly reach out their hands to grasp an attractive toy placed in front of them. Older children, however, can easily extend their hands and take hold of any object without any practice; they can also speak fluently through the unconscious coordination of multiple speech organs, including the jaw, upper and lower lips, tongue, and vocal cords. These fine motor skills are honed through unconscious improve- ments in their information processing as they learn how to coordinate many muscles to generate smooth and complicated movements. Understanding these computational mechanisms for human movements will enable us to design new types of man-machine interface for communication devices and of artificial brain processing mechanisms for humanoid robots. 2. Hierarchical motor control It has been reported that the hand trajectory is almost straight for any directional arm reaching Selected Papers Hiroaki Gomi † Abstract This paper introduces a hierarchical structure of information processing for generating movements and an experimental test of a major hypothesis of the human movement production mechanism called ‘equi- librium-point control’ from the viewpoint of stiffness control. It also presents experimental observations of stiffness control of the human arm in interactions with the external world and of coordinative speech articulation. This research has future applications in the design of new types of man-machine interfaces for communication devices and of artificial brain processing mechanisms for humanoid robots. Human Motor Control Mechanism † NTT Communication Science Laboratories Atsugi-shi, 243-0198 Japan E-mail: gomi@idea.brl.ntt.co.jp Positional reference Trajectory formation Inverse kinematics Inverse dynamics (a) Hierarchical information processing for robot arm control (b) Equilibrium-point control hypothesis for human arm movement Muscle and reflex servo mechanism Movement target Hand trajectory Joint trajectory Motor command Limb movement Movement Fig. 1. Computational framework for arm control.