Research Analyzer
← Back ICRA 2026

ED-SLAM: Event-Depth Gaussian Splatting SLAM

Jian Huang, Haotian Shen, Xinhao Lou, Peidong Liu

PDF

AI summary

Key figure (auto-extracted from paper)
ED-SLAM achieves robust, high-fidelity 3D reconstruction and accurate pose estimation in challenging conditions by tightly coupling a novel event-depth tracker with 3D Gaussian Splatting.
Event cameras 3D Gaussian Splatting SLAM depth estimation visual tracking robotics

Problem

Existing event-based and 3DGS SLAM methods struggle with robust pose estimation under fast motion, low illumination, or textureless scenes, often degrading over long trajectories without ground-truth poses.

Approach

The pipeline uses a bidirectional patch-based tracker on time-surface maps and depth data for stable pose estimation, which directly initializes and constrains a continuous-time 3D Gaussian Splatting mapping process.

Key results

  • Novel bidirectional patch-based event-depth tracker for robust pose estimation
  • First tight integration of event-depth tracking with 3DGS-based SLAM
  • Significantly improved tracking stability and novel view synthesis quality on synthetic and real-world datasets
  • Demonstrates superior robustness in low-light and high-speed motion scenarios

Why it matters

Enables reliable autonomous navigation and high-fidelity scene understanding for robotics and AR/VR systems in challenging real-world environments.

Abstract

Event-based Gaussian splatting (GS) reconstruc- tion have recently attracted considerable attention. Existing methods usually assume the camera poses are known as a prior, or struggle to process long event streams due to the robustness of the method while poses are not known. In this work, we present ED-SLAM, an Event-Depth Gaussian Splatting-based simultaneous localization and mapping (SLAM) pipeline, which is robust to long event streams and does not require ground- truth camera poses. The pipeline achieves high-accuracy pose estimation and high-fidelity 3D reconstruction thanks to the impressive 3D representation capability of Gaussian splatting. In particular, we propose a novel bidirectional patch-based event-depth tracking algorithm and seamlessly integrate it into the Gaussian splatting mapping pipeline. Extensive experiments on both synthetic and real-world datasets demonstrate that our method significantly improves tracking accuracy and robust- ness, and also delivers improved reconstruction performance.

Index terms

Computer Vision for Automation Sensor Fusion

Related papers