Abstract
Significant breakthroughs have been made in aerial and terrestrial autonomous navigation, taking great strides toward fully autonomous technologies. However, their underwater counterpart, marine robotics, remains an emerging field. This is partly because many essential sensors, such as optical cameras, LIDAR or GPS, face significant challenges when operating in submerged environments. Therefore, research on underwater-compatible sensors sensors will contribute significantly to the advancement of autonomous underwater vehicles. This study explores the application of Forward-looking SONAR (FLS) for underwater autonomous navigation, including conventional Simultaneous Localization and Mapping (SLAM), as well as novel parent-child localization techniques. Specifically, we examine the use of Constant False Alarm Rate (CFAR) filter as a feature extractor for FLS images. The proposed algorithm's performance is evaluated through controlled experiments, demonstrating that CFAR filter is effective and robust as a feature extractor for FLS images. Additionally, it achieves near real-time operation and maintains a consistent processing rate due to the proposed algorithm's window-size-invariant characteristic.