Manhattan World Constraint for Indoor Line-based Mapping Using Ultrasonic Scans

Abstract

Ultrasonic rangefinders are low-cost sensors that have been utilized in the past to perform mapping, localization and Simultaneous Localization and Mapping (SLAM). Ultrasonic has a low angular resolution, and it is challenging to identify and correspond features in ultrasonic scans. In the past, occupancy mapping has been proposed to address this issue. However, for the localization of the sensor with respect to such occupancy grids, post-processing for landmark identification is required (which increases the computational cost). In contrast to the occupancy grid approach, direct feature-based mapping methods are also proposed in the literature, where landmarks are extracted using many consecutive ultrasonic sensor scans instead of using a single sparser scan. In this article, a novel direct feature-based mapping method using ultrasonic sensors in the indoor manmade environment is proposed. The proposed method utilizes the prior knowledge of the structure of indoor environments (such as most houses or warehouses), where the walls are either parallel or perpendicular (Manhattan World Constraint (MWC)). This prior knowledge is used with multiple consecutive ultrasonic scans to provide accurate line-based maps. Compared to the occupancy grid mapping, the proposed method avoids the discretization of the point cloud coordinates (which can be a source of inaccuracy). Further, it does not require a second step for landmark detection. Compared to the direct feature-based methods, which are general-purpose, our approach is the first method- to the best of the authors' knowledge- that imposes MWC to map using panoramic ultrasonic scans. The accuracy of the proposed mapping method is demonstrated in a cluttered indoor environment. The results indicate that the proposed method accurately detects and distinguishes most edges and corners.