A MIMO Approach to Supervisory Control in Robot Visual Feedback

Abstract

A two camera system providing sensory information between a robot and the environment enabling visual guidance and feedback can be modelled as a Multiple Input Multiple Output (MIMO) system with interaction. Condition Number of the matrix describing the system or Relative Gain Array can be used to measure the loop interaction in a MIMO system This paper presents a supervisory control in visual servoing, integrating the cooperation of two similar sensors motivated by the fact that no single strategy delivers solutions to the problems associated with visual servoing especially for large and complex workspaces. The movement of a manipulator is a challenge in physically limited or hazardous workspaces. Hybrid algorithms are more suitable in such cases rather than relying on any single scheme. The interaction of a stationary master camera with an eye in hand camera expands the scope of visual feedback. Simulation and experimentation suggest the validity of the proposed control strategy especially for manipulator poses from where the target is not visible and neural networks are used to improve the performance.