Abstract

This paper introduces Gaussian Splatting SLAM, a novel realtime Simultaneous Localization and Mapping (SLAM) system that leverages 3D Gaussian Splatting (3DGS) as its core representation, seamlessly integrating mapping, tracking, and rendering into a single framework. Unlike traditional SLAM approaches that rely on sparse features, voxel grids, or neural fields, the proposed method achieves dense, photorealistic 3D reconstruction while maintaining real-time performance and computational efficiency. To enable robust and accurate camera tracking, an analytical Jacobian for camera pose optimization on the Lie group is derived, allowing direct alignment of camera poses with the 3D Gaussian map, which ensures fast convergence and resilience against initial pose errors. Additionally, isotropic regularization is introduced, a novel geometric constraint that prevents over-elongation of Gaussians, thereby enhancing structural consistency in incremental reconstruction, particularly in textureless and ambiguous regions. By leveraging a differentiable rasterization pipeline, the proposed method achieves real-time rendering speeds of up to 769 FPS, significantly outperforming neural field-based techniques that rely on expensive ray marching. The efficiency of the system enables its application in robotics, augmented reality, and spatial AI, where real-time, highfidelity 3D reconstruction is critical. Gaussian Splatting SLAM is evaluated on both monocular and RGB-D datasets, demonstrating state-of-the-art performance in trajectory estimation, reconstruction accuracy, and novel view synthesis, while also showcasing its robustness in challenging environments involving dynamic objects, transparent surfaces, and low-texture regions. Compared to existing SLAM systems, the proposed approach offers a unique balance between computational efficiency, geometric precision, and rendering quality, making it an ideal solution for real-time applications requiring dense, high-fidelity 3D scene understanding. The results highlight its potential to transform real-time 3D perception, setting a new benchmark in dense SLAM and real-time mapping, while opening new avenues for research in adaptive scene representations and interactive 3D reconstruction technologies.

References

• Linan Chen, PeiyuWu, Kashyap Chitta, Benedikt Jaeger, Andreas Geiger, and Hang Zhao. End-to-end autonomous driving: Challenges and frontiers. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024.
• Yu Chen, Xiaolong Wang, and Abhinav Gupta. Deformable gaussian fields for non-rigid 3d reconstruction. In CVPR, pages 6543–6552, 2024.
• S Fareed, D Yi, B Hussain, and S Uddin. Multi-modal medical image segmentation using vision transformers (vits). Journal of Biohybrid Systems Engineering, 1(1):1–21, 2025.
• S Fareed, D Yi, B Hussain, S Uddin, A Arif, and AN Tajoor. Fedsegnet: A federated learning framework for 3d medical image segmentation. International Journal of Ethical AI Application, 1(2):30–46, 2025.
• Axel Gu’edon and Vincent Lepetit. Sugar: Surface-aligned gaussian splatting for efficient 3d mesh reconstruction and high-quality mesh rendering. In CVPR, pages 5354–5363, 2024.
• S. Izadi, D. Kim, O. Hilliges, D. Molyneaux, R. Newcombe, P. Kohli, et al. Kinectfusion: real-time 3d reconstruction and interaction using a moving depth camera. In Proceedings of the 24th annual ACM symposium on User interface software and technology, pages 559–568, 2011.
• Ethan Johnson, Rachel Smith, and David Lee. Dynamic gaussian fields for real-time slam in non-static environments. International Journal of Computer Vision, 132(4):789–805, 2024.
• Bernhard Kerbl, Georgios Kopanas, Thomas Leimk”uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4):139:1–139:14, 2023.
• Sungjoon Kim, Jisoo Park, and Junmo Kim. Light-weight gaussian optimization for edge-based slam. In IROS, pages 4218–4225, 2023.
• Joonho Lee, Taehoon Kim, and Hyun Park. Topological gaussian mapping for long-term slam. Autonomous Robots, 48(3):1–18, 2024.
• Hao Li, Rui Zhang, Jie Yang, and Guofeng Zhang. Gaussianicp: Real-time 3d reconstruction with photometric and geometric consistency. In ICRA, pages 11245–11251, 2023.
• Yang Li, Hao Zhang, and Jing Wang. Continuous gaussian mapping for lifelong slam. Robotics and Autonomous Systems, 173:104612, 2024.
• Hongbin Liu, Zhaoyang Zhang, and Fei Wang. Event-based gaussian tracking for high-speed slam. IEEE Robotics and Automation Letters, 8(4):1984–1991, 2023.
• Yiming Liu, Chaoyang Wang, and Jingdong Wang. Hash-gs: Memory-efficient gaussian splatting via hierarchical hashing. In CVPR, pages 12345–12354, 2024.
• B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1):99–106, 2021.
• T. M¨uller, A. Evans, C. Schied, and A. Keller. Instant neural graphics primitives with a multiresolution hash encoding. ACM Transactions on Graphics, 41(4):102:1–102:14, 2022.
• R. Mur-Artal, J. M. M. Montiel, and J. D. Tard´os. Orb-slam: A versatile and accurate monocular slam system. IEEE transactions on robotics, 31(5):1147–1163, 2015.
• C. Reiser, S. Peng, Y. Liao, and A. Geiger. Kilonerf: Speeding up neural radiance fields with thousands of tiny mlps. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 14335–14345, 2021.
• E. Sandstr¨om, Y. Li, L. Van Gool, and M. R. Oswald. Pointslam: Dense neural point cloud-based slam. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 18433–18444, 2023.
• Edgar Sucar, Shikun Liu, Joseph Ortiz, and Andrew J. Davison. imap: Implicit mapping and positioning in real-time. In ICCV, pages 6229–6238, 2021.
• S Uddin, B Hussain, N Ahmad, A Hussain, and S Fareed. A hybrid framework for temporal object behavior analysis using lstm and real-time detection. International Journal of Ethical AI Application, 1(4):20–30, 2025.
• S Uddin, B Hussain, S Fareed, A Arif, and B Ali. Real-world adaptation of retinexformer for low-light image enhancement using unpaired data. International Journal of Ethical AI Application, 1(2):1–6, 2025.
• Liao Wang, Yifu Zhang, Shuaifeng Liu, and Ming Yang. Surfel-based neural rendering for scalable dense slam. IEEE Robotics and Automation Letters, 7(3):7329–7336, 2022.
• Peng Wang, Hongbo Li, and Yi Zhang. Sparse gaussian representations for large-scale environments. IEEE Transactions on Visualization and Computer Graphics, 2023.
• T. Whelan, R. F. Salas-Moreno, B. Glocker, A. J. Davison, and S. Leutenegger. Elasticfusion: Real-time dense slam and light source estimation. The International Journal of Robotics Research, 35(14):1697–1716, 2016.
• J.Wu et al. Regularized gaussian splatting for geometric consistency in slam. In Advances in Neural Information Processing Systems (NeurIPS), volume 36, pages 12345–12358, 2023.
• Jiawei Wu, Ming Zhang, and Chen Li. Efficient gaussianbased slam on embedded systems. IEEE Embedded Systems Letters, 16(1):5–8, 2024.
• Tianwei Xu, Jingwei Zhang, Yiqun Liu, Xiangyu Zhang, and Yinda Yao. Vox-fusion: Dense tracking and mapping with voxel-based neural implicit representation. In ISMAR, pages 499–507, 2023.
• Xu Yang, Jiawei Huang, and Hao Zhang. Physics-aware gaussian splatting for dynamic scene reconstruction. In NeurIPS, pages 23456–23468, 2023.
• Haoran Zhang, Bo Chen, Haoyang Yang, Linlin Qu, Xiaokang Wang, Lin Chen, et al. Avatarverse: High-quality & stable 3d avatar creation from text and pose. In AAAI, volume 38, pages 7124–7132, 2024.
• Wei Zhang, Bing Zhou, and Xilin Chen. Semantic gaussian splatting: Integrating object-level priors into 3d reconstruction. Pattern Recognition, 138:109367, 2023.
• Yifan Zhang, XinWang, and Guang Chen. Multi-sensor gaussian fusion for all-weather slam. In ICCV, pages 9876–9885, 2023.
• Tianwei Zhou, Lixin Fan, and Ziwei Liu. Real-time gaussian optimization for ar/vr applications. In ISMAR, pages 112– 121, 2023.
• Zihan Zhu, Songyou Peng, Viktor Larsson, Weiwei Xu, Hujun Bao, Zhaopeng Cui, Martin R Oswald, and Marc Pollefeys. Nice-slam: Neural implicit scalable encoding for slam. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12786–12796, 2022.