Experimental evaluations of prediction servoing to moving object by hand-eye robotic system

In the field of robot vision, a control method called visual servoing attracts attention. The visual servoing is a method to control robots by visual information in a feedback loop, which is obtained by video cameras. So, this method is expected to be able to make robots adapt to tasks in changing or unknown environment. However, when the target object moves quickly, it happens to be unable for the robots to track it due to dynamical effect, i.e., motion delay. To decrease the delay, we have proposed prediction servoing control method, which utilizes prediction of the target position based on the past observed position data of the object and learning by neural networks, and utilize predicted position as a desired position for the visual servoing. In this research, we have confirmed how the learning function in neural networks work for precise prediction of target´s future position through visual servoing experiments.