Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Online Global Loop Closure Detection for Large-Scale Multi-Session Graph-Based SLAM (2407.15305v1)

Published 22 Jul 2024 in cs.RO

Abstract: For large-scale and long-term simultaneous localization and mapping (SLAM), a robot has to deal with unknown initial positioning caused by either the kidnapped robot problem or multi-session mapping. This paper addresses these problems by tying the SLAM system with a global loop closure detection approach, which intrinsically handles these situations. However, online processing for global loop closure detection approaches is generally influenced by the size of the environment. The proposed graph-based SLAM system uses a memory management approach that only consider portions of the map to satisfy online processing requirements. The approach is tested and demonstrated using five indoor mapping sessions of a building using a robot equipped with a laser rangefinder and a Kinect.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (22)
  1. F. Lu and E. Milios, “Globally consistent range scan alignment for environment mapping,” Autonomous robots, vol. 4, no. 4, pp. 333–349, 1997.
  2. M. Bosse, P. Newman, J. Leonard, and S. Teller, “Simultaneous localization and map building in large-scale cyclic environments using the Atlas framework,” Int. J. of Robotics Research, vol. 23, no. 12, pp. 1113–39, 2004.
  3. S. Thrun and M. Montemerlo, “The graph SLAM algorithm with applications to large-scale mapping of urban structures,” Int. J. of Robotics Research, vol. 25, no. 5-6, pp. 403–429, 2006.
  4. G. Grisetti, R. Kümmerle, C. Stachniss, and W. Burgard, “A tutorial on graph-based SLAM,” Intelligent Transportation Systems Magazine, IEEE, vol. 2, no. 4, pp. 31–43, 2010.
  5. J. McDonald, M. Kaess, C. Cadena, J. Neira, and J. Leonard, “Real-time 6-DOF multi-session visual SLAM over large scale environments,” Robotics and Autonomous Systems, vol. 61, no. 10, pp. 1144–58, 2012.
  6. B. Kim, M. Kaess, L. Fletcher, J. Leonard, A. Bachrach, N. Roy, and S. Teller, “Multiple relative pose graphs for robust cooperative mapping,” in Proc. IEEE Int. Conf. on Robotics and Automation.   IEEE, 2010, pp. 3185–3192.
  7. K. L. Ho and P. Newman, “Loop closure detection in SLAM by combining visual and spatial appearance,” Robotics and Autonomous Systems, vol. 54, no. 9, pp. 740–749, 2006.
  8. M. Cummins and P. Newman, “Appearance-only SLAM at large scale with FAB-MAP 2.0,” The Int. J. of Robotics Research, vol. 30, no. 9, pp. 1100–1123, 2011.
  9. A. Angeli, D. Filliat, S. Doncieux, and J.-A. Meyer, “Fast and incremental method for loop-closure detection using bags of visual words,” IEEE Trans. on Robotics, vol. 24, no. 5, pp. 1027–1037, October 2008.
  10. T. Botterill, S. Mills, and R. Green, “Bag-of-words-driven, single-camera simultaneous localization and mapping,” J. of Field Robotics, vol. 28, no. 2, pp. 204–226, 2011.
  11. K. Konolige, J. Bowman, J. Chen, P. Mihelich, M. Calonder, V. Lepetit, and P. Fua, “View-based maps,” The Int. J. of Robotics Research, vol. 29, no. 8, pp. 941–957, July 2010.
  12. O. Booij, Z. Zivkovic, and B. Kröse, “Efficient data association for view based SLAM using connected dominating sets,” Robotics and Autonomous Systems, vol. 57, no. 12, pp. 1225–1234, 2009.
  13. J. Folkesson and H. I. Christensen, “Closing the loop with graphical SLAM,” IEEE Trans. on Robotics, vol. 23, no. 4, pp. 731–41, 2007.
  14. G. Grisetti, S. Grzonka, C. Stachniss, P. Pfaff, and W. Burgard, “Efficient estimation of accurate maximum likelihood maps in 3D,” in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, 2007, pp. 3472–3478.
  15. H. Johannsson, M. Kaess, M. Fallon, and J. J. Leonard, “Temporally scalable visual SLAM using a reduced pose graph,” in RSS Workshop on Long-term Operation of Autonomous Robotic Systems in Changing Environments, Karlsruhe, Germany, May 2012.
  16. M. Labbe and F. Michaud, “Appearance-based loop closure detection for online large-scale and long-term operation,” IEEE Transactions on Robotics, vol. 29, no. 3, pp. 734–745, 2013.
  17. J. Sivic and A. Zisserman, “Video Google: A text retrieval approach to object matching in videos,” in Proc. 9th Int. Conf. on Computer Vision, Nice, France, 2003, pp. 1470–1478.
  18. P. J. Besl and N. D. McKay, “Method for registration of 3-D shapes,” in Robotics-DL tentative.   International Society for Optics and Photonics, 1992, pp. 586–606.
  19. R. Atkinson and R. Shiffrin, “Human memory: A proposed system and its control processes,” in Psychology of Learning and Motivation: Advances in Research and Theory.   Elsevier, 1968, vol. 2, pp. 89–195.
  20. F. Ferland, L. Clavien, J. Frémy, D. Letourneau, F. Michaud, and M. Lauria, “Teleoperation of azimut-3, an omnidirectional non-holonomic platform with steerable wheels,” in Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, Oct 2010, pp. 2515–2516.
  21. Y. Latif, C. D. C. Lerma, and J. Neira, “Robust loop closing over time.” in Robotics: Science and Systems, Sydney, Australia, July 2012.
  22. N. Sunderhauf and P. Protzel, “Towards a robust back-end for pose graph SLAM,” in Proc. IEEE Int. Conf. on Robotics and Automation, 2012, pp. 1254–1261.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (2)
  1. François Michaud (23 papers)
  2. Mathieu Labbe (1 paper)
Citations (386)
View on Semantic Scholar

Summary

Online Global Loop Closure Detection for Large-Scale Multi-Session Graph-Based SLAM

The paper presented by Labbé and Michaud introduces a sophisticated approach to solving challenges associated with large-scale and long-term simultaneous localization and mapping (SLAM) in autonomous robotics, particularly under multi-session conditions. It addresses the complexities of unknown initial positioning—a problem that arises when a robot is either relocated without awareness or when maps are accumulated across multiple sessions. The solution proposed integrates a global loop closure detection mechanism within a graph-based SLAM framework, optimized for online performance, an aspect particularly challenging on a scale where the SLAM domain expands over extensive environments.

Technical Details

The proposed system constructs a map using a graph of nodes and links. Nodes encapsulate odometric data and various forms of sensory information including laser scans and RGB-D images, which are used for visualization and loop closure detection. There are two primary types of links within this graph: neighbor links that represent odometric transformations, and loop closure links that are formed when a robot revisits a previously mapped area.

Key to the paper's contribution is the combination of loop closure detection with a memory management strategy to achieve online processing efficiency. The loop closure detection is implemented using a bag-of-words model, leveraging visual words extracted from RGB images. This approach utilizes a Bayesian filter to compute loop closure hypotheses, which are confirmed if they surpass a defined threshold. Transformations detected through loop closures are computed via RANSAC, ensuring robustness against outliers.

The graph optimization is managed through the TORO algorithm, which uses tree-based network optimization to refine the map by correcting errors propagated through odometry using links as constraints. Nonetheless, as the size of the environment grows, maintaining all map nodes in working memory for real-time processing becomes infeasible. To tackle this, the paper introduces a memory management technique inspired by human memory models, maintaining a balance between short-term and long-term storage, effectively ensuring that the map's size does not compromise the online processing requirements.

Results and Implications

The approach is assessed via indoor mapping experiments using multiple sessions, demonstrating that the system processes data efficiently within a set time limit, despite the size of the environment. The results validate that the method can create maps that remain cognizant of previously explored areas, enabling the merging of multiple maps into a single cohesive representation across sessions. This is a critical feature for applications requiring persistent environment mapping, such as in reconnaissance and search-and-rescue missions.

While the experimental outcomes demonstrate the feasibility and efficacy of the proposed solution, the discussion recognizes that the complexity of real-world environments may demand enhancements. For instance, the paper notes the potential need for more sophisticated strategies to judiciously manage memory as the number of mapping sessions escalates.

Conclusion and Future Directions

Overall, the research contributes significantly to the field of SLAM by presenting an approach capable of supporting large-scale and long-term mapping over multiple sessions, with practical implications for autonomous navigation in dynamic environments. The integration of loop closure detection with a robust memory management framework paves the way for maintaining efficient online operations irrespective of environmental complexity or size.

The authors also hint at further research directions, such as exploring autonomous exploration strategies that could optimize exploration based on nodes retained in working memory. Such future developments could further refine the efficiency and reliability of SLAM systems, underscoring an ongoing evolution in autonomous robotics.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown