Abstract: Visual servo control concerns the problem of controlling robots using real-time computer vision feedback. While numerous approaches have evolved over the years and many systems have been demonstrated in laboratories around the world, most can be classified as either position-based or image-based, depending on whether camera pose or image features are used in the control law. Each approach has well-documented performance problems.
In this talk, after a brief description of these methods and their respective shortcomings, I will describe two approaches to overcoming their performance problems. With the first approach, the control system is partitioned along its spatial degrees of freedom. Rotation about and translation along the optical axis are controlled by a specifically designed controller, and the remaining degrees of freedom are controlled with a traditional image-based method. With the second approach, a hybrid, switched control system selects either an image-based or position-based controller at each time instant, using a state-based switching scheme. The performance of both of these approaches have been verified by a variety of simulated and experimental results. If time permits, I will also describe recent results applying a switched method to the control of a unicycle robot with field-of-view constraints.