CHOW-SLAM: Compact Hybrid Representation with Complementary Overlap Window Optimization for RGB-D SLAM

2026-08-03

AI Summary

This work addresses the challenge that existing NeRF-based RGB-D SLAM systems struggle to balance compactness and discriminability under limited online computational resources, while temporal constraints often degrade in effectiveness. To overcome these limitations, we propose CHOW-SLAM, a novel framework that integrates a parameterized–hash hybrid representation with a multi-scale plane–grid structure to jointly model TSDF and density fields. Our approach employs a complementary overlapping window mechanism to synergistically optimize short-term geometric consistency and long-term temporal coherence. Coupled with loss-aware keyframe scheduling and geometry-informed initialization, CHOW-SLAM achieves significantly improved reconstruction quality and tracking accuracy within a fixed computational budget, outperforming state-of-the-art methods across multiple benchmark datasets.

Abstract

Simultaneous localization and mapping (SLAM) based on Neural Radiance Fields (NeRF) enables dense, continuous scene reconstruction. However, existing systems operating with limited online resources struggle to simultaneously construct two types of constraints, namely, compact yet discriminative spatial constraints derived from scene representations and persistent temporal constraints derived from historical observations. To address this challenge, we propose CHOW-SLAM, a dense RGB-D SLAM framework that explicitly constructs these complementary spatial and temporal constraints. Spatially, we propose a compact parametric-hash (P-H) hybrid representation that organizes components based on planes and grids across scales in P and H branches. A unified multi-output decoder further aligns the ray termination distributions induced by TSDF and density, preserving geometry and appearance under a compact parameter budget. Temporally, we propose a complementary overlap-window strategy to prevent optimization from being dominated by short-term overlap or weakly related historical observations. Within a fixed budget, the strategy retains recent frames, selects high-overlap local frames, and introduces temporally distributed historical keyframes. Loss-aware keyframe insertion and bundle adjustment scheduling further adapt optimization to tracking quality. In addition, ORB-based tracking and geometric pose estimation are used for pose initialization, followed by neural rendering optimization to improve tracking stability. Extensive evaluations on multiple datasets demonstrate that CHOW-SLAM outperforms state-of-the-art methods in both scene reconstruction quality and camera tracking accuracy. The source code is available at https://github.com/jinjidexiaohuoban/CHOW-SLAM.

Problem

• Neural Radiance Fields
• RGB-D SLAM
• spatial constraints
• temporal constraints
• online resource limitation

Innovation

• Neural Radiance Fields
• Hybrid Representation
• Overlap Window Optimization
• RGB-D SLAM
• Compact Scene Representation