The simultaneous localization and mapping (SLAM): An overview

Positioning is needed for many applications related to mapping and navigation, either in civilian or military domains. The si gnificant developments in satellite-based techniques, sensors, telecommunications, computer hardware and software, image processing, etc. positively influenced solving the positioning problem efficiently and instantaneously. Accordingly, the mentioned development empowered the applications and advancement of autonomous navigation. One of the most interestingly developed positioning techniques is what is called in robotics Simultaneous Localization and Mapping (SLAM). The SLAM problem solution has witnessed a quick improvement in the last decades, either using active sensors like the RAdio Detection and Ranging (Radar) and Light Detection and Ranging (LiDAR) or passive sensors like cameras. Definitely, positioning and mapping is one of the main tasks for geomatics engineers, and therefore it's of high importance for them to understand the SLAM topic, which is not easy because of the huge documentation and algorithms available and the various SLAM solutions in terms of the mathematical models, complexity, the sensors used, and the type of applications. In this paper, a clear and simplified explanation of SLAM from a geometrical viewpoint is introduced, avoiding going into the complicated algorithmic details behind the presented techniques. In this way, a general overview of SLAM is presented, showing the relationship between its different components and stages, like the core part of the front-end and back-end, and their relation to the SLAM paradigm. Furthermore, we explain the major mathematical techniques of filtering and pose graph optimization, either using visual or LiDAR SLAM, and introduce a summary of the efficient contribution of deep learning to the SLAM problem. Finally, we address examples of some existing practical applications of SLAM in our reality.