City Research Online

Abstract

This paper introduces a self-supervised deep learning keypoint model SiLK specifically designed for loop closure detection in Cooperative Visual SLAM (Simultaneous Localization and Mapping). Firstly, the paper proposes a self-supervised learning framework that improves the robustness and accuracy of loop closure detection. Secondly, it presents the implementation of this deep learning keypoint model which replaces traditional manual feature descriptors, showcasing substantial enhancements in detecting loop closures across diverse and dynamic environments. This advancement ensures greater adaptability to variations in lighting and scene changes. Third, the paper illustrates the benefits of leveraging self-supervised learning within a cooperative visual SLAM context, where multiple agents share and fuse their local observations. This collaborative effort leads to refined pose estimations and more accurate mapping outcomes, enhancing the overall system's performance in complex settings. The effectiveness of the proposed model was rigorously tested in real-world scenarios, demonstrating its superiority in robustness and mapping precision compared to traditional methods in cooperative robotic systems, including those used for environmental monitoring and exploration tasks. The results highlight the potential of self-supervised deep learning models to revolutionize loop closure detection in visual SLAM, offering a promising avenue for future research in autonomous systems and cooperative robotics in challenging operational environments.

Publication Type:	Conference or Workshop Item (Paper)
Additional Information:	© 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Publisher Keywords:	Deep learning, Point cloud compression, Visualization, Simultaneous localization and mapping,Accuracy,Pose estimation,Collaboration,Self-supervised learning,Robustness, Robots
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) T Technology > TJ Mechanical engineering and machinery
Departments:	School of Science & Technology School of Science & Technology > Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads