Advancements in Gaussian Splatting for SLAM and 3D Reconstruction

The field of Simultaneous Localization and Mapping (SLAM) and 3D reconstruction is rapidly advancing with the integration of Gaussian Splatting (GS) techniques. Recent research has focused on improving the efficiency and accuracy of GS-based SLAM systems, particularly in dynamic environments. Innovations in algorithm-hardware co-design, adaptive feature extraction, and unified optimization frameworks have led to significant breakthroughs in reconstruction quality and pose estimation accuracy. Furthermore, the incorporation of multi-spectral imaging and opacity radiance fields has enhanced geometric mapping performance and enabled more robust tracking. Noteworthy papers include: AGS, which achieves up to 17.12x speedups against state-of-the-art 3DGS accelerators. DyPho-SLAM, which presents a real-time, resource-efficient visual SLAM system designed to address the challenges of localization and photorealistic mapping in environments with dynamic objects.

Sources

AGS: Accelerating 3D Gaussian Splatting SLAM via CODEC-Assisted Frame Covisibility Detection

DyPho-SLAM : Real-time Photorealistic SLAM in Dynamic Environments

UPGS: Unified Pose-aware Gaussian Splatting for Dynamic Scene Deblurring

Towards Integrating Multi-Spectral Imaging with Gaussian Splatting

SR-SLAM: Scene-reliability Based RGB-D SLAM in Diverse Environments

FGO-SLAM: Enhancing Gaussian SLAM with Globally Consistent Opacity Radiance Field

IL-SLAM: Intelligent Line-assisted SLAM Based on Feature Awareness for Dynamic Environments

Real-time Photorealistic Mapping for Situational Awareness in Robot Teleoperation

Online Dynamic SLAM with Incremental Smoothing and Mapping

Built with on top of