Graph SLAM for AGV using geometrical arrangement based on lamp and SURF features in a factory environment

摘要

In this study, we propose a method to extract features for graph SLAM using an upward-looking camera mounted in an automated guided vehicle (AGV) in a factory environment. To this end, a novel feature detector and descriptor GALS (geometrical arrangement based on lamps and SURF) is proposed in this study. This algorithm consists of two parts: 1) geometrical arrangement of SURF features around a lamp that is often observed in a factory environment; and 2) a SURF descriptor. The use of the geometrical arrangement of several SURF features provides a more distinguishable feature than the use of a single SURF feature, and thus it can be robustly used for feature matching. As a result of GALS-based graph SLAM, a graph is created and optimized through a Georgia Tech Smoothing and Mapping Library (GTSAM) tool so that the accumulated error due to the slip is minimized. A series of experiments in indoor factory environments showed that the proposed scheme resulted in reliable navigation.