Vision-based autonomous navigation based on motion estimation

This paper addresses a navigation method which copes with dynamic environments, e.g. objects moving in the environment and environment changes. Because many vision-based navigation methods mainly focus on finding consistent corresponding parts with one of the database images, dynamic environments can cause the failure of autonomous navigation due to visual occlusion. To solve these problems, we propose a motion-based navigation method in contrast with appearance-based approaches. We also solve the kidnapping problem, which is defined as the problem of re-localizing a mobile robot after it is undergone an unknown motion or visual occlusion. This algorithm is based on motion estimation by a camera to plan the next movement and an efficient pose recovery scheme after kidnapping problems. Various experimental results demonstrate the capability of vision-based autonomous navigation against dynamic environments.