Splatonic: Architecture Support for 3D Gaussian Splatting SLAM via Sparse Processing

2025-11-23

AI Summary

To address the high computational overhead of 3D Gaussian Splatting (3DGS) in SLAM—hindering real-time deployment on mobile platforms—this paper proposes Splatonic, an algorithm–hardware co-design framework. Methodologically, it introduces: (i) adaptive sparse pixel sampling to reduce rendering load; (ii) parallel Gaussian rendering with predictive alpha detection to accelerate projection and aggregation; and (iii) a pipelined architecture to maximize GPU hardware utilization. Evaluated on off-the-shelf GPUs, Splatonic achieves up to 14.6× end-to-end speedup and an average 274.9× acceleration across four representative 3DGS-SLAM algorithms, while improving energy efficiency by up to 4738.5×—all without compromising reconstruction accuracy. This work is the first to tightly couple sparsity-aware rendering with hardware-level pipelining, establishing a scalable, efficient implementation paradigm for lightweight 3DGS-SLAM.

Technology Category

• Machine Learning: Hardware-aware ML
• Computer Vision: 3D Computer Vision
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

3D Gaussian splatting (3DGS) has emerged as a promising direction for SLAM due to its high-fidelity reconstruction and rapid convergence. However, 3DGS-SLAM algorithms remain impractical for mobile platforms due to their high computational cost, especially for their tracking process. This work introduces Splatonic, a sparse and efficient real-time 3DGS-SLAM algorithm-hardware co-design for resource-constrained devices. Inspired by classical SLAMs, we propose an adaptive sparse pixel sampling algorithm that reduces the number of rendered pixels by up to 256$ imes$ while retaining accuracy. To unlock this performance potential on mobile GPUs, we design a novel pixel-based rendering pipeline that improves hardware utilization via Gaussian-parallel rendering and preemptive $α$-checking. Together, these optimizations yield up to 121.7$ imes$ speedup on the bottleneck stages and 14.6$ imes$ end-to-end speedup on off-the-shelf GPUs. To further address new bottlenecks introduced by our rendering pipeline, we propose a pipelined architecture that simplifies the overall design while addressing newly emerged bottlenecks in projection and aggregation. Evaluated across four 3DGS-SLAM algorithms, Splatonic achieves up to 274.9$ imes$ speedup and 4738.5$ imes$ energy savings over mobile GPUs and up to 25.2$ imes$ speedup and 241.1$ imes$ energy savings over state-of-the-art accelerators, all with comparable accuracy.

Problem

• Reducing computational cost of 3DGS-SLAM for mobile platforms
• Enabling real-time performance on resource-constrained devices
• Optimizing hardware utilization through algorithm-hardware co-design

Innovation

• Adaptive sparse pixel sampling reduces rendered pixels
• Pixel-based rendering pipeline improves hardware utilization
• Pipelined architecture simplifies design and addresses bottlenecks