Self-organization of a task-oriented visuo-motor map for a redundant arm

In this paper, we propose a new method that solves the inverse kinematics problem by self-organizing a task-oriented visuo-motor map for a redundant manipulator. The method makes good use of the redundancy of the manipulator with excess degrees of freedom, although conventional methods using a self-organization algorithm of a visuo-motor map do not utilize the redundancy, and eliminates the redundancy by maximizing a dexterity measure for task execution. In this paper, two types of the map can be teamed up by applying an extended manipulability measure and a force-controllability measure as dexterity measure. The efficiency of the proposed method is ascertained by computer simulations for learning the task-oriented visuo-motor maps of a three-link manipulator and a four-link manipulator in horizontal plane. The values of dexterity measures became higher than those of conventional methods that did not consider the characteristics of tasks and did not make use of the redundant degrees of freedom of the manipulator