Controlling An 8 Dof Manipulator By Means Of Neural Fields

Abstract

In this article we present a new approach for the control of a redundant robot arm. Our approach contains two parts: first, we have implemented a concept which deals with the underdetermined problem of the inverse kinematics of robot arms with more than six degrees of freedom. This concept guarantees a one-to-one mapping between the task coordinates (position and orientation) and the joint coordinates of the robot arm and allows to use the additional degrees of freedom for additional task requirements such as obstacle avoidance and smoothness of the trajectory. Second, we apply a mathematical concept known as neural fields to the control of the end-e#ector's position. The application of neural fields to the problem of trajectory generation solves two problems: a smooth end-e#ector trajectory is generated and obstacles are avoided. After presenting our hardware platform, an anthropomorphic assistance robot, we will describe the basic concepts of our approach.