A framework for visual servoing tasks

A general framework for visual servoing tasks is proposed. The objective of the paper is twofold: a) how a complicated servoing task might be composed from a multitude of simple ones, and b) how the integration of basic and simple visual algorithms can be used in order to provide a robust input estimate to a control loop for a mobile platform or a robot manipulator For that purpose, voting schema and consensus theory approaches are investigated together with some initial vision based algorithms. Voting is known as a model-free approach to integration and therefore interesting for applications in real-world environments which are difficult to model. It is experimentally shown how servoing tasks like pick-and-place, opening doors and fetching mail can be robustly performed using the proposed approach.