Occlusion avoidance in corners-based SLAM with different data association algorithms

This paper proposes the occlusion avoidance method in comers-based simultaneous localization and mapping (SLAM) with different data association algorithms. The redundant or wrong features are extracted if part of the object is occluded. The comers are chosen by intersecting two adjacent line segments and selecting the end-points of some special line segment. When two segments are far enough, the nearest two end-points of these two lines are considered as candidate comers. Then one of two candidates is stored as final comer with shorter distance of laser beam. However, if the line segment with this corner is very short, this comer is ignored because it may be just part of the object with complex surface, such as column. After extracting theses comers, they have been used in estimating the state of mobile robot and previous landmarks. To have a better matching result, two data association algorithms are applied in constructing the correspondence between new features and stored map features. The experiment result in indoor environment shows the validity of proposed method.