Forward Model Creation for Visual Servoing in a Six Link Manipulator

Abstract

This paper shows the utility of a forward model for improving the visual servoing performance of a robotic manipulator. Following a developmental robotics approach, the manipulator undergoes a motor babbling phase which is used to create a forward model using an ANFIS neural network. The forward model maps the relationship between the joint positions and the image of the end effector in two camera views. Using the obtained forward model, an initial image Jacobian is estimated and is used with a visual servoing controller. Simulation results demonstrate that errors are significantly lower when the estimated Jacobian is used.