3D motion representations in visual servo control

A new approach to visual servoing and vision-guided robotics is introduced. This approach uses visual information for autonomous behavior. It amounts to using robust, global spatiotemporal representations easily extracted from the dynamic imagery. Specifically, the geometrical patterns of normal flows are used as the input to the servo mechanism. It is shown that the positions of these patterns are related to the three dimensional motion parameters. By locating the positions of these patterns, we can solve a variety of navigational problems with little computational effort.