SELF-ORGANIZATION AND FUNCTION FORMATION ON MOTION AND/ACTION CONTROL

点击翻译按钮获取中文摘要

Let's imagine the phenomenal finger movements of a pianist, the dynamic carriage of a soccer playr, and the hand movements of a potter. The complexity and dexterity of their motions are worthy of great admiration. However, we feel more admiration for the diversity and flexibility of their motions produced from the same body and the adaptability to its surroundings. By what mechanisms are these motions/actions controlled? Our body is a very redundant system having more than one hundred joint degrees of freedom and a complex non-linear dynamic system. As a consequence, to achieve smooth and dexterous movements, these redundant degrees of freedom must be constrained in one form or another.For example, when reaching our hand to pick up a cup, we put our fingers in order into the shape suitable for it, depending on the size and shape of cup and presence of its grip. In addition, we preset the stiffness of the wrist and other joints depending on the weight of the cup. Moreover, in the moment … More that a soccer playr shoot a ball, he draws his body to the full like a bow and presets well-directed impedance (visco-elasticity) about the joint of a pivoting foot. In skilled actions, thus, the degrees of freedom and impedances of the body have been preset according to the goal of the motion and the characteristics of the object. The phrases "taking a stance" and "carrying oneself well" which are used on a daily basis represent the above directly.In the present research, we took up the problem of the redundant degrees of freedom and gave self-organization and function formation methods on skilled motion control. We obtained the following results. 1) Self-organization of task-oriented sensory-motor maps, 2) Hierarchical internal representation method for hand pre-shaping based on visual information, 3) Arm impedance identification during postural control and bimanual control, 4) Mathematical model of adaptation in rhythmic motion to environmental changes, 5) Learning of inverse kinematics and dynamics maps based on artificial neural networks, 6) Motion control scheme for dynamic manipulation 7) Distributed cooperative control of multiple manipulator systems based on passive velocity field control (PVFC). 8) Learning of robot impedance based on repetitive learning. Less