Camera-based localization for indoor service robots using pictographs

In our current research, we are developing a practical localization system for an indoor service vehicle. In order to realize the practical use, required self-position estimating system should have high reliability and be low cost. Usually, a localization scheme with special landmarks, which are equipped in navigation area beforehand, is high reliability. However, the special marker causes an increasing of its cost and in general environment, it is difficult to set many special markers all over the service area of the robot, in practical use. Therefore, in this paper, we propose a novel positioning system that does not require special markers. Our key idea is that our proposed localization system estimates its position using pictographs. Pictographs are widely used in general environment with human such as offices, stations and welfare care facilities. Using them as a landmark, our system does not require special markers and realizes high reliability. Our system equips only a low-cost camera and uses SURF (Speeded Up Robust Feature) algorithm for detecting pictographs. Furthermore, our system estimates the relative position using pictographs as a landmark, and it derives the robot's path to the goal position. We verify the effectiveness of our scheme by experiments with our prototype.