Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
166 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

BundledSLAM: An Accurate Visual SLAM System Using Multiple Cameras (2403.19886v2)

Published 28 Mar 2024 in cs.RO

Abstract: Multi-camera SLAM systems offer a plethora of advantages, primarily stemming from their capacity to amalgamate information from a broader field of view, thereby resulting in heightened robustness and improved localization accuracy. In this research, we present a significant extension and refinement of the state-of-the-art stereo SLAM system, known as ORB-SLAM2, with the objective of attaining even higher precision. To accomplish this objective, we commence by mapping measurements from all cameras onto a virtual camera termed BundledFrame. This virtual camera is meticulously engineered to seamlessly adapt to multi-camera configurations, facilitating the effective fusion of data captured from multiple cameras. Additionally, we harness extrinsic parameters in the bundle adjustment (BA) process to achieve precise trajectory estimation.Furthermore, we conduct an extensive analysis of the role of bundle adjustment (BA) in the context of multi-camera scenarios, delving into its impact on tracking, local mapping, and global optimization. Our experimental evaluation entails comprehensive comparisons between ground truth data and the state-of-the-art SLAM system. To rigorously assess the system's performance, we utilize the EuRoC datasets. The consistent results of our evaluations demonstrate the superior accuracy of our system in comparison to existing approaches.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (17)
  1. R. Pless, “Using many cameras as one,” in 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings., vol. 2, 2003, pp. II–587.
  2. J.-M. Frahm, K. Köser, and R. Koch, “Pose estimation for multi-camera systems,” in Joint Pattern Recognition Symposium.   Springer, 2004, pp. 286–293.
  3. J. Sola, A. Monin, M. Devy, and T. Vidal-Calleja, “Fusing monocular information in multicamera slam,” IEEE Transactions on Robotics, vol. 24, no. 5, pp. 958–968, 2008.
  4. A. Harmat, I. Sharf, and M. Trentini, “Parallel tracking and mapping with multiple cameras on an unmanned aerial vehicle,” in Intelligent Robotics and Applications, C.-Y. Su, S. Rakheja, and H. Liu, Eds.   Berlin, Heidelberg: Springer Berlin Heidelberg, 2012, pp. 421–432.
  5. A. Harmat, M. Trentini, and I. Sharf, “Multi-camera tracking and mapping for unmanned aerial vehicles in unstructured environments,” Journal of Intelligent & Robotic Systems, vol. 78, no. 2, pp. 291–317, May 2015.
  6. M. J. Tribou, A. Harmat, D. W. Wang, I. Sharf, and S. L. Waslander, “Multi-camera parallel tracking and mapping with non-overlapping fields of view,” The International Journal of Robotics Research, vol. 34, no. 12, pp. 1480–1500, 2015.
  7. S. Yang, S. A. Scherer, X. Yi, and A. Zell, “Multi-camera visual slam for autonomous navigation of micro aerial vehicles,” Robotics and Autonomous Systems, vol. 93, pp. 116 – 134, 2017.
  8. P. Liu, M. Geppert, L. Heng, T. Sattler, A. Geiger, and M. Pollefeys, “Towards Robust Visual Odometry with a Multi-Camera System,” IEEE/RSJ International Conference on Intelligent Robots and System, 2018.
  9. C. Forster, Z. Zhang, M. Gassner, M. Werlberger, and D. Scaramuzza, “SVO: Semi-Direct Visual Odometry for Monocular and Multi-Camera Systems,” IEEE Transactions on Robotics, vol. 33, no. 2, pp. 249–265, 2017.
  10. R. Mur-Artal and J. D. Tardos, “ORBSLAM2: an Open-Source SLAM System for Monocular, Stereo and RGB-D Cameras,” IEEE Transactions on Robotics, vol. 33, no. 5, p. 1255–1262, 2017.
  11. W. Wang, J. Li, Y. Ming, and P. Mordohai, “EDI: Eskf-based disjoint initialization for visual-inertial slam systems,” in Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2023.
  12. H. Strasdat, A. J. Davison, J. M. M. Montiel, and K. Konolige, “Double window optimisation for constant time visual SLAM,” IEEE International Conference on Computer Vision, pp. 2352–2359, 2011.
  13. B. Triggs, P. McLauchlan, R. Hartley, and Fitzgibbon, “Bundled adjustment - a modern synthesis,” in Vision Algorithms: Therory and Pracitce. Springer Verlag, pp. 298–375, 2000.
  14. R.Kuemmerle, G. Grisetti, H. Strasdat, K. Konolige, and W. Burgard, “g2o: A general framework for graph optimization,” International Conference on Robotics and Automation, pp. 3607–3613, 2011.
  15. D. Galvez-Lpez and J. D. Tardos, “Bags of binary words for fast place recognition in image sequences,” IEEE Transactions on Robotics, vol. 28, no. 5, pp. 1188–1197, 2012.
  16. R. Mur-Artal and J. D. Tardos, “Fast relocalisation and loop closing in keyframe-based slam,” IEEE Internatinal Conference on Robotics and Automation, pp. 846–853, 2014.
  17. M. Burri, J. Nikolic, P. Gohl, T. Schneider, J. Rehder, M. W. A. S. Omari, and R. Siegwart, “The EuRoC micro aerial vehicle datasets,” The International Journal of Robotics Research, vol. 35, no. 10, p. 1157–1163, 2016.
Citations (9)
View on Semantic Scholar

Summary

  • The paper presents BundledSLAM's innovative approach that integrates multiple camera inputs into a unified BundledFrame for precise mapping and tracking.
  • It leverages advanced bundle adjustment with extrinsic parameters to refine pose estimation and significantly enhance trajectory accuracy.
  • Evaluations on the EuRoC dataset show that BundledSLAM outperforms ORB-SLAM2 and VINS-Stereo, demonstrating robustness in challenging conditions.

Enhancing Multi-Camera SLAM Systems for Precise Localization: A Dive into BundledSLAM

Introduction to Multi-Camera SLAM Systems

Recent advancements in Visual Simultaneous Localization and Mapping (SLAM) have ushered in a myriad of applications ranging from autonomous navigation to augmented reality. Despite the prolific research in SLAM systems, the implementation of multi-camera systems to overcome the limitations of traditional monocular or stereo systems has remained significantly less explored. Multi-camera systems, characterized by their wide Field of View (FoV), offer substantial benefits in terms of robustness and accuracy, which are critical for applications requiring comprehensive environmental perception. The paper introduces BundledSLAM, an extension of ORB-SLAM2, aiming to harness the potential of multi-camera configurations for enhanced SLAM performance.

Key Contributions

  • Comprehensiveness and Extensibility: The paper presents a full-fledged SLAM system tailored for multi-camera setups, extending capabilities to include loop closure and map reuse. By introducing an efficient structure termed "Bundled," which amalgamates data from multiple cameras into a unified entity, the system facilitates seamless operations such as tracking, optimization, and place recognition. The introduction of the virtual camera concept, or BundledFrame, and the utilization of Bundle Adjustment (BA) with extrinsic parameters significantly contribute to the system's novelty and effectiveness.

System Overview

BundledSLAM operates on the principle of converting measurements from multiple real cameras into a virtually constructed BundledFrame, optimizing for the SLAM pipeline's efficiency and accuracy. The system is segmented into three primary threads: tracking, local mapping, and loop closing, each meticulously designed to handle the complexity introduced by multi-camera inputs. Specifically, the introduction of a virtual camera model supports the system's ability to perform intricate tasks such as motion estimation and optimization across multiple cameras without compromising computational efficiency.

Motion Estimation and Bundle Adjustment

At the core of BundledSLAM's methodology is a sophisticated approach to motion estimation and Bundle Adjustment (BA), tailored for multi-camera systems. This segment of the paper delves deeply into the mathematical formulations and optimizations that enable accurate pose estimations. Leveraging the extrinsic parameters between cameras, the system optimizes for the pose of a primary camera, subsequently updating other camera positions to maintain relative consistency. The adaptation of BA, segmented into motion-only, local, and global adjustments, plays a pivotal role in refining pose estimations and map reconstructions, substantially enhancing the overall system accuracy.

Evaluation and Findings

BundledSLAM underwent rigorous evaluation using the EuRoC dataset, benchmarking against state-of-the-art systems like ORB-SLAM2 and VINS-Stereo. The results underscored BundledSLAM's superior performance in terms of trajectory accuracy and pose error metrics across varied sequences. Notably, the system demonstrated exceptional resilience and accuracy in scenarios challenging for traditional SLAM systems, such as those with rapid motion or limited textural features. These findings reinforce the potential of multi-camera SLAM systems in improving localization precision and robustness, setting a new standard for future research and applications in the field.

Implications and Future Directions

The introduction and evaluation of BundledSLAM highlight the substantial gains in accuracy and robustness achievable with multi-camera SLAM systems. Beyond immediate practical applications, this research opens avenues for further exploration into sensor fusion, particularly with IMUs, to address challenges posed by rapid movements or poor lighting conditions. However, the integration of additional sensors warrants a thoughtful consideration of computational complexity, which the future research endeavors of BundledSLAM aim to address. By focusing on optimizing performance without exacerbating computational demands, BundledSLAM sets a foundation for the next generation of SLAM systems, characterized by their precision, efficiency, and adaptability to complex environments.

In conclusion, BundledSLAM emerges as a highly promising approach to multi-camera SLAM, showcasing significant improvements over the benchmark systems. As future work propels this research forward, the fusion of multi-camera systems with other sensors holds the promise of unlocking unprecedented levels of SLAM performance, pushing the boundaries of what's possible in autonomous navigation and beyond.

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

Tweets