Ground-plane based indoor mobile robot localization using RGB-D sensor1

This paper addresses the problem of absolute localization in an indoor environment using a RGB-Depth camera. The approach is based on the use of the ground region perceived by the RGB camera to detect and decode its position and edges. The localization system uses this data to match it with a known on-board map. The ground plane detection algorithm is designed to be robust to vibration or disturbances during the robot motion. The localization system is based on the particle filter, fusing odometry with ground region matching where each particle’s weight is proportional to the quality of correspondence between the ground edge estimation and the nearest walls. Promising results were obtained and are presented in this article.