Simultaneous Localization and Mapping in the Epoch of Semantics: A Survey

• Published in: International journal of control, automation, and systems Vol. 17; no. 3; pp. 729 - 742
• Main Authors: Sualeh, Muhammad, Kim, Gon-Woo
• Format: Journal Article
• Language: English
• Published: Bucheon / Seoul Institute of Control, Robotics and Systems and The Korean Institute of Electrical Engineers 01.03.2019; Springer Nature B.V; 제어·로봇·시스템학회
• Subjects: Algorithms; Control; Control theory; Engineering; Localization; Machine learning; Mechatronics; Robotics; Semantics; Simultaneous localization and mapping; Taxonomy; 제어계측공학; Bundle adjustment; deep learning; SLAM; semantics; pose graph optimization
• ISSN: 1598-6446; 2005-4092
• DOI: 10.1007/s12555-018-0130-x

Simultaneous Localization and Mapping (SLAM) with an astonishing research history of over three decades has brought the concept to the door step of truly autonomous robotic systems. The concept has advanced beyond the map building and self-localization of robot on the map. On the other hand, the long-standing challenges pertaining to the provision of out of the box solution for range of conditions still needs to be addressed. However, the technological advancements in the area is steadily making its ways into industry. This paper surveys state-of-the-art SLAM and discuss the insights of existing methods. Starting with a classical definition of SLAM, a brief conceptual overview, and formulation of a standard SLAM system is carried out. While discussing the auxiliaries for solving SLAM, the influx of machine learning into SLAM is also addressed. Moreover, recent SLAM algorithms have been reviewed with a focus on emerging concept of semantics to augment the system. In this paper a taxonomy of recently developed SLAM algorithms with a detailed comparison metrics, is presented. Furthermore, open challenges, future directions and emerging research issues have also been laid down.