Integration of vision in autonomous mobile robotics

An architecture for integrating various vision devices in an autonomous agent is presented and demonstrated in the case of vision-based navigation in autonomous mobile robotics. The architecture is based on the behavioural approach. Applying this approach to vision results in a collection of vision-based behaviours that perform just simple tasks when considered isolated but are much more capable when working in cooperation. In the presented work, an experimental behavioural architecture was applied to vision by developing a large number and variety of vision devices and behaviours. Shown to be successful in robot navigation tasks, we believe the behavioural architecture to be relevant also for other complex vision-based applications involving real-world and real-time interactions