Model-based Visual Servoing for Automatic Docking System of Circular Symmetrical Target with Large Displacement

In this paper, a model-based visual servoing strategy with dual cameras is proposed for a circular symmetrical target with large displacement. The automatic docking system is composed of the fast-moving stage and the position-based visual servoing (PBVS) control stage, which enables a fast, accurate and stable docking task. A simplified stereo matching method is used in the fast-moving stage, and the CAD model-based tracking method with virtual visual servoing (VVS) method is applied in the PBVS control stage to estimate and update the current pose of the circular symmetrical docking target. Experiments of a robotic docking system validate the effectiveness, precision and flexibility of this strategy. The Root Mean Square errors of the docking results are less than 0.5 millimeters in translation and 0.1 degrees in rotation, and the average time for the whole docking task is no more than 1 minutes.