Epipolar-kinematics relations estimation neural approximation for robotics closed loop visual servo system

Abstract

This article studies a possibility of using a learning system for learning the complicated kinematics relating object features to robotics arm joint space. To achieve visual tracking, visual servoing and control for object manipulation without losing it from a robotics system, it is essential to relate a number of object's geometrical features to a robotics system joint space. Object visual data play important role in such sense. Most robotics visual servo systems rely on object features Jacobian, in addition to the inverse. Object visual features inverse Jacobian is not easily put together and computed, hence to use such relation in a visual loops. A neural system have been used to approximate such relations, hence avoiding computing object's feature inverse Jacobian, even at singular Jacobian postures. To validate the concept, the visual servo loop developed by Rives [1] has been rather updated and used as a test bench problem.