A Full Overview of Visual SLAM Algorithms

Simultaneous Localization and Mapping, or SLAM, is an essential approach for autonomous robotic systems. Simultaneous mapping and sensor pose estimation are made possible by SLAM in an unknown environment. Visual simultaneous localization and mapping, often known as V-SLAM, is an important field in robotics, particularly for cooperative and interactive mobile robot environments. Faster development of Visual SLAM can be attributed to low-cost sensors, easy integration of additional sensors, and improved ambient information. Numerous strategies and techniques for implementing visual-based SLAM systems are presented in the literature. It might be challenging for a novice in this field to sort through the range of publications, recognize and evaluate the key algorithms, and ultimately select the best one for their intended use. Therefore, we present the three main visual-based SLAM approaches (visual alone, visual inertial, and RGB-D SLAM), emphasizing their salient features and limitations. We also use flowcharts and diagrams to examine the main algorithms of each approach. It tracks the development of SLAM techniques historically and offers contrastive evaluations of concepts and salient ideas. The research examines important Visual SLAM benchmark datasets and offers process-level visualization for every method. This research aims to cover the essential elements and characteristics of SLAM methodologies, providing a foundational resource for understanding and selecting appropriate techniques.