GARAD-SLAM: 3D GAussian splatting for Real-time Anti Dynamic SLAM
🤖 AI Summary
Dynamic objects in real-world scenes degrade the mapping accuracy and tracking stability of 3D Gaussian Splatting (3DGS)-based SLAM. To address this, we propose a real-time SLAM framework tailored for dynamic environments. Our method integrates 3DGS representation with Gaussian pyramid feature extraction, front-end dynamic label mapping, and rendering-aware optimization. Key contributions include: (1) a novel hierarchical Gaussian-based dynamic segmentation mechanism that enables fine-grained detection and separation of moving points; and (2) a differentiable rendering penalty term that regularizes dynamic Gaussian parameter updates, preserving geometric consistency of the static map and ensuring invertibility of incremental updates. Evaluated on real-world dynamic datasets, our approach significantly reduces tracking error, suppresses rendering artifacts, and achieves notable improvements in PSNR and SSIM—outperforming existing baselines in both reconstruction quality and robustness.
Application Category
📝 Abstract
The 3D Gaussian Splatting (3DGS)-based SLAM system has garnered widespread attention due to its excellent performance in real-time high-fidelity rendering. However, in real-world environments with dynamic objects, existing 3DGS-based SLAM systems often face mapping errors and tracking drift issues. To address these problems, we propose GARAD-SLAM, a real-time 3DGS-based SLAM system tailored for dynamic scenes. In terms of tracking, unlike traditional methods, we directly perform dynamic segmentation on Gaussians and map them back to the front-end to obtain dynamic point labels through a Gaussian pyramid network, achieving precise dynamic removal and robust tracking. For mapping, we impose rendering penalties on dynamically labeled Gaussians, which are updated through the network, to avoid irreversible erroneous removal caused by simple pruning. Our results on real-world datasets demonstrate that our method is competitive in tracking compared to baseline methods, generating fewer artifacts and higher-quality reconstructions in rendering.
Problem
Research questions and friction points this paper is trying to address.
Addresses mapping errors in dynamic environments
Improves tracking accuracy with dynamic segmentation
Enhances rendering quality by reducing artifacts
Innovation
Methods, ideas, or system contributions that make the work stand out.
Dynamic segmentation on Gaussians
Gaussian pyramid network for labeling
Rendering penalties on dynamic Gaussians
🔎 Similar Papers
No similar papers found.
