Scholarlink

<p>Reliable localization in low-light and dynamic indoor environments remains a major challenge for Autonomous Mobile Robots (AMR) operating in GPS-denied areas. Existing visual SLAM systems typically address either illumination degradation or dynamic interference, but rarely handle both simultaneously, leading to unstable performance in real-world scenarios. This paper presents RID-SLAM, an integrated enhancement framework that jointly addresses illumination imbalance and dynamic disturbances within a unified pipeline. Instead of relying on a single improvement module, the proposed approach combines multiple complementary mechanisms to stabilize feature extraction, suppress dynamic interference, and maintain tracking continuity under adverse conditions. The system is evaluated on public datasets (TUM RGB-D and OpenLORIS) and real-world indoor environments. Experimental results demonstrate that RID-SLAM significantly improves localization robustness and reduces feature-matching errors, achieving up to 90% improvement over baseline methods. These results validate the effectiveness of the proposed integration strategy for reliable localization in challenging indoor environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rid-slam: a robust illumination-and-dynamics-aware RGB-D SLAM framework for indoor environments

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月11日

查看全文
  • Pei-Zhi Xie,
  • Kuei-Jung Hung,
  • Jau-Woei Perng

摘要

Reliable localization in low-light and dynamic indoor environments remains a major challenge for Autonomous Mobile Robots (AMR) operating in GPS-denied areas. Existing visual SLAM systems typically address either illumination degradation or dynamic interference, but rarely handle both simultaneously, leading to unstable performance in real-world scenarios. This paper presents RID-SLAM, an integrated enhancement framework that jointly addresses illumination imbalance and dynamic disturbances within a unified pipeline. Instead of relying on a single improvement module, the proposed approach combines multiple complementary mechanisms to stabilize feature extraction, suppress dynamic interference, and maintain tracking continuity under adverse conditions. The system is evaluated on public datasets (TUM RGB-D and OpenLORIS) and real-world indoor environments. Experimental results demonstrate that RID-SLAM significantly improves localization robustness and reduces feature-matching errors, achieving up to 90% improvement over baseline methods. These results validate the effectiveness of the proposed integration strategy for reliable localization in challenging indoor environments.