Construction of SLAM Algorithm for Window Cleaning Robot Moving Along Window Frame

Abstract

The purpose of this study is to build SLAM algorithm for window cleaning robot which move along window frame. Automation of window cleaning in skyscraper by a robot is desired. It’s necessary for cleaning to the corner of window that the robot which move on the windowpane wall recognizes shape of the window frame and to estimations self-localization on the windowpane wall. However, since it is difficult to detect the glass surface directly with a non-contact sensor, it is not possible to acquire the features necessary for SLAM. Therefore, the window frame is detected as a feature point by the RGB-D sensor, and the SLAM algorithm on the windowpane wall is constructed using the detected feature value. In this paper, various window frame shapes are prepared, distance information from the RGB-D sensor to the window frame is acquired, and the results obtained as 3D information and the feature value acquisition results by image processing are described.