Visual Navigation for Mobile Robots

摘要

Three methods for support of visual navigation have been presented: Artificial visual landmarks, corridor following using vanishing point, and road following using terrain classification based on data fusion of laser scanner and vision. All the methods have been verified experimentally so that their usefulness in real systems is demonstrated. It has been shown that a combination of artificial landmarks and odometry will be able to limit the navigation error to a given level if the landmarks are placed with a sufficiently small distance. Using both landmarks and corridor following the same navigation accuracy may be obtained with much fewer landmarks thus enhancing the usability of the system. The results for terrain classification show that fusing data from laser and vision gives a good foundation for path and road-following for outdoor robots. This may be used for service robots that are operating in e.g. public parks and gardens.