How NeRFs and 3D Gaussian Splatting areReshaping SLAM: a Survey

Abstract---Over the past two decades, research in the field of Simultaneous Localization and Mapping (SLAM) has undergone a

significant evolution, highlighting its critical role in enabling autonomous exploration of unknown environments. This evolution ranges

from hand-crafted methods, through the era of deep learning, to more recent developments focused on Neural Radiance Fields

(NeRFs) and 3D Gaussian Splatting (3DGS) representations. Recognizing the growing body of research and the absence of a

comprehensive survey on the topic, this paper aims to provide the first comprehensive overview of SLAM progress through the lens of

the latest advancements in radiance fields. It sheds light on the background, evolutionary path, inherent strengths and limitations, and

serves as a fundamental reference to highlight the dynamic progress and specific challenges


TABLE 1: SLAM Systems Overview. We categorize the different methods into main RGB-D, RGB, and LiDAR-based
frameworks. In the leftmost column, we identify sub-categories of methods sharing specific properties, detailed in Sections
3.2.1 to 3.3.2 . Then, for each method, we report, from the second leftmost column to the second rightmost, the method name
and publication venue, followed by (a) the input modalities they can process: RGB, RGB-D, D ( e.g. LiDAR, ToF, Kinect,
etc.), stereo, IMU, or events; (b) mapping properties: scene encoding and geometry representations learned by the model;
(c) additional outputs learned by the method, such as object/semantic segmentation, or uncertainty modeling (Uncert.);
(d) tracking properties related to the adoption of a frame-to-frame or frame-to-model approach, the utilization of external
trackers, Global Bundle Adjustment (BA), or Loop Closure; (e) advanced design strategies, such as modeling sub-maps or
dealing with dynamic environments (Dyn. Env.); (f) the use of additional priors. Finally, we report the link to the project
page or source code in the rightmost column. † indicates code not released yet

相关推荐
Wellijohn7 小时前
我把公司内部的 12 个 CRUD 接口交给了 AI Agent,然后省出了两周
人工智能·python·mcp
伞伞悦读8 小时前
【第92期】逻辑回归的分数概率与阈值的实验与排查
人工智能·python·深度学习·机器学习
7177779 小时前
互联网企业研发平台推荐:从工具堆叠到一体化协同的选型逻辑
人工智能·gitee
立青.9 小时前
AI大模型术语
人工智能
电科_银尘10 小时前
【大模型算法与工程 27】训练配方:批量、学习率、精度与梯度裁剪
学习·算法·机器学习
土豆125012 小时前
你只说要什么,它负责拆活交付:任务编排插件上手指南
人工智能·llm
GoldenSpider.AI13 小时前
讲透 Claude Code 的 Code Mode:原理、流程、用法与 24 轮实测
人工智能·ai编程·claude code
oak隔壁找我13 小时前
一款运行在 DSH(DeepSeek Harness)中的工作区文件管理插件
人工智能
江屿风14 小时前
AI Agent架构全解析:从规划到记忆的智能跃迁
人工智能·笔记·架构·agent·c/s·mcp工作流