Open AccessAn ANN and Virtual Vector Function Approach for the Computation of the Inverse Kinematics of Redundant Manipulators
Author(s) -
Shahnaz Habibkhah,
René V. Mayorga
Publication year - 2021
Publication title -
wseas transactions on information science and applications
Language(s) - English
Resource type - Journals
eISSN - 2224-3402
pISSN - 1790-0832
DOI - 10.37394/23209.2021.18.15
Subject(s) - workspace , inverse kinematics , kinematics , computer science , computation , forward kinematics , artificial neural network , function (biology) , task (project management) , set (abstract data type) , control theory (sociology) , mathematics , mathematical optimization , algorithm , artificial intelligence , robot , engineering , physics , control (management) , classical mechanics , evolutionary biology , biology , systems engineering , programming language
This Paper presents an Artificial Neural Network (ANN) method for the solution of the Inverse Kinematics (IK) of redundant manipulators. This problem normally has an infinite number of solutions. Several conventional approaches based on numerical methods have been proposed over the years. However, it is well known that ANN implementations for the solution of the problem for redundant manipulators are inefficient since they lead to inaccurate solutions. The main issue on the implementation of ANN solutions is that an attempt is made on mapping a relation from a given number of variables in the manipulator task space to a larger number of variables in the joint space. This Paper proposes the inclusion of a virtual vector function in the task space to “complete” it; so that the number of variables in the task space is at least equal to the number of variables in the joint space. Here, the proposed approach is successfully tested on a 3 DOF planar redundant manipulator performing four diverse target trajectories inside the manipulator’s workspace. Additionally, in order to define the target trajectories, some manipulator’s links limitations are considered and some conditions are set for the target trajectories.