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

📅 2025-08-30
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
To address the throughput bottleneck in 3D Gaussian Splatting SLAM systems—caused by frame-wise multiple iterative optimization and excessive Gaussian primitives—this paper proposes AGS, an algorithm-hardware co-optimization framework. Methodologically: (i) it introduces a coarse-to-fine pose tracking scheme; (ii) it pioneers cross-frame co-visibility detection leveraging intermediate video encoder (CODEC) data to enable Gaussian contribution reuse; and (iii) it establishes an end-to-end hardware-software co-design architecture integrating custom hardware accelerators and a dynamic scheduler. Experiments demonstrate that AGS achieves up to 17.12× and 5.41× throughput improvement over state-of-the-art methods—including GSCore—on mobile and high-end GPU platforms, respectively, significantly enhancing real-time performance. This work establishes a novel paradigm for efficient deployment of Gaussian Splatting SLAM.

Technology Category

Application Category

📝 Abstract
Simultaneous Localization and Mapping (SLAM) is a critical task that enables autonomous vehicles to construct maps and localize themselves in unknown environments. Recent breakthroughs combine SLAM with 3D Gaussian Splatting (3DGS) to achieve exceptional reconstruction fidelity. However, existing 3DGS-SLAM systems provide insufficient throughput due to the need for multiple training iterations per frame and the vast number of Gaussians. In this paper, we propose AGS, an algorithm-hardware co-design framework to boost the efficiency of 3DGS-SLAM based on the intuition that SLAM systems process frames in a streaming manner, where adjacent frames exhibit high similarity that can be utilized for acceleration. On the software level: 1) We propose a coarse-then-fine-grained pose tracking method with respect to the robot's movement. 2) We avoid redundant computations of Gaussians by sharing their contribution information across frames. On the hardware level, we propose a frame covisibility detection engine to extract intermediate data from the video CODEC. We also implement a pose tracking engine and a mapping engine with workload schedulers to efficiently deploy the AGS algorithm. Our evaluation shows that AGS achieves up to $17.12 imes$, $6.71 imes$, and $5.41 imes$ speedups against the mobile and high-end GPUs, and a state-of-the-art 3DGS accelerator, GSCore.
Problem

Research questions and friction points this paper is trying to address.

Accelerating 3D Gaussian Splatting SLAM systems
Reducing redundant Gaussian computations across frames
Improving real-time performance through algorithm-hardware co-design
Innovation

Methods, ideas, or system contributions that make the work stand out.

Algorithm-hardware co-design framework for 3DGS-SLAM
Coarse-then-fine pose tracking with movement adaptation
CODEC-assisted frame covisibility detection engine
🔎 Similar Papers
No similar papers found.
H
Houshu He
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China
N
Naifeng Jing
Department of Micro-Nano Electronics, Shanghai Jiao Tong University, Shanghai, China
L
Li Jiang
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China
Xiaoyao Liang
Xiaoyao Liang
Shanghai Jiao Tong University
Computer Architecture
Z
Zhuoran Song
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China