3Projection Neural Networks for Robot Arm Control

3.1 Introduction

With advances in mechanics, electronics, and computer engineering, automatic manipulators are becoming increasingly popular in industrial applications to reduce the burden on labor forces. Among the various types of available manipulators, redundant manipulators, which usually possess more degrees of freedom than general manipulators and thus offer increased control flexibility for complicated tasks, have attracted intensive research in recent decades.

Despite the great advantages offered by redundant manipulators in dexterous manipulation for complicated tasks, the efficient control of such manipulators remains a challenging problem. A redundant manipulator provides a nonlinear mapping from its joint space to a Cartesian workspace. The goal of kinematic control is to find a control action in the joint space that produces a desired motion in the workspace. However, the nonlinearity of the mapping makes it difficult to directly solve this problem at the angle level. Instead, in most approaches, the problem is first converted into a problem at the velocity or acceleration level, and solutions are then sought in the converted space. In early work [25], control solutions were directly found using the pseudoinverse of the Jacobian matrix of a manipulator. However, this control strategy suffers from an intensive computational burden because of the need to perform the pseudoinversion of matrices continuously over ...

Get Kinematic Control of Redundant Robot Arms Using Neural Networks now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.