An improved view-based navigation by adjustable position resolution

Abstract

View-based navigation has been proposed for navigating a robot with an image sensor in the field of robot vision. In existing view-based navigations, their position resolution is defined depending on the change of scenery around robots. However, when applying the view-based navigation in actual environments, requirements for the position resolution are different in accordance with the environments given and the tasks assigned to a robot. Therefore, the position resolution in the view-based navigation is required to be easily controlled to consider the environments and the tasks. We have previously proposed the view-based navigation robust to the change of illumination conditions such as the sunshine from windows and fluorescent lights by using edge images. However, using edge images tends to become sensitive in capture noises. To prevent this over-sensitiveness, we used a smoothing effect of approximate plane of brightness. We also found that the smoothing effect used here is not only robust to delicate changes in an image, but also available the changes of scenery around robots. Thus, in this paper we suggest an improved view-based navigation that is applicable to the changes of scenery by easily adjusting the position resolution. We also could confirm through the experiments that the improved navigation method is able to control the position resolution, and then flexibly respond to the change of surroundings and tasks of a robot.