Sensor fusion localization system for outdoor mobile robot

Abstract

The purpose of this study is to develop sensor fusion loclization of autonomous mobile robot for outdoor environment. We merge 2 localization models; a Camera sensor with Particle filter and two GPS sensors with Kalman filter. The marks on the load such as lane, arrow, cross road and speed hump, etc., are most important information in urban environment. The camera device is useful to detect road landmarks, because the information printed on the road are color-based model. We use Delaunay triangulation method to extract landmark points in the interest region. The robot finds out the position by comparing these landmark points with a map with Particle filter algorithm. We also use GPS that provides global position directly. The GPS, however, has large drift position data caused by intentional pseudo noise and slow acquisition time for mobile robot. In this paper, a sensor fusion localization system is proposed to get the robot's reasonable position by fusing these 2 localization models that compensate the weak points each other.