Waliner: Lightweight and Resilient Plugin Mapping Method with Wall Features for Visually Challenging Indoor Environments
DongKi Noh, Byunguk Lee, Hanngyoo Kim, Seung-Hwan Lee, HyunSung Kim, Juwon Kim, Jeong-Sik Choi, Seung-Min Baek
AI summary
Problem
Vision-based SLAM systems struggle to extract reliable features in visually challenging indoor environments like featureless white walls, while existing robust solutions often demand expensive sensors or heavy computation unsuitable for affordable service robots.
Approach
Waliner uses an RGB-D sensor and embedded NPU to run semantic segmentation for wall detection, extracts structural wall lines, and refines robot pose via a Kalman filter under the Manhattan world assumption, functioning as a lightweight plugin for existing SLAM frameworks.
Key results
- Novel semantic line extraction algorithm optimized for resource-constrained embedded platforms
- Lightweight odometry refinement using Manhattan frames to enhance pose estimation and loop closure
- Over 5% improvement in mapping consistency (MSI) across real-world in-home scenes
- Seamless plugin integration into existing SLAM backends like RTAB-MAP with minimal computational overhead
Why it matters
Enables cost-effective, mass-produced in-home service robots to navigate reliably in feature-poor environments without relying on expensive LiDAR or high-end processors.
Abstract
Vision-based indoor navigation systems have been proposed previously for service robots. However, in real-world scenarios, many of these approaches remain vulnerable to visually challenging environments such as white walls. In- home service robots, which are mass-produced, require afford- able sensors and processors. Therefore, this paper presents a lightweight and resilient plugin mapping method called Waliner, using an RGB-D sensor and an embedded processor equipped with a neural processing unit (NPU). Waliner can be easily implemented in existing algorithms and enhances the accuracy and robustness of 2D/3D mapping in visually chal- lenging environments with minimal computational overhead by leveraging a) structural building components, such as walls; b) the Manhattan world assumption; and c) an extended Kalman filter-based pose estimation and map management technique to maintain reliable mapping performance under varying lighting and featureless conditions. As verified in various real-world in-home scenes, the proposed method yields over a 5 % improvement in mapping consistency as measured by the map similarity index (MSI) while using minimal resources.