Tightly Coupled SLAM with Imprecise Architectural Plans
Muhammad Shaheer, Jose Andres Millan Romera, Hriday Bavle, Marco Giberna, Jose Luis Sanchez-Lopez, Javier Civera, Holger Voos
AI summary
Problem
Real-world buildings frequently deviate from their architectural plans, introducing systematic errors in robot localization and mapping. Existing SLAM methods typically assume perfect alignment between "as-planned" and "as-built" environments, ignoring construction tolerances and modifications.
Approach
The authors introduce diS-Graphs, a method that tightly couples a hierarchical LiDAR-based SLAM graph with a semantic architectural plan graph. This unified framework jointly optimizes the robot's global pose and the geometric deviations of matched structural elements like walls and rooms in real-time.
Key results
- Robust performance under structural deviations up to 35 cm and 15°
- 43% lower localization error than baselines in simulation
- 7% reduction in alignment error for real-world "as-built" 3D maps
- Real-time joint estimation of global pose and structural deviations
Why it matters
Enables reliable indoor robot navigation and real-time construction accuracy assessment where architectural plans are inherently imprecise.
Abstract
Robots navigating indoor environments often have access to architectural plans, which can serve as prior knowledge to enhance their localization and mapping capabilities. While some SLAM algorithms leverage these plans for global localiza- tion in real-world environments, they typically overlook a critical challenge: the “as-planned” architectural designs frequently devi- ate from the “as-built” real-world environments. To address this gap, we present a novel algorithm that tightly couples LIDAR- based simultaneous localization and mapping with architectural plans in the presence of deviations. Our method utilizes a multi- layered semantic representation to not only localize the robot, but also to estimate global alignment and structural deviations between “as-planned” and “as-built” environments in real-time. To validate our approach, we performed experiments in simulated and real datasets demonstrating robustness to structural devia- tions up to 35 cm and 15◦. On average, our method achieves 43% less localization error than baselines in simulated environments, while in real environments, the “as-built” 3D maps show 7% lower average alignment error. Paper Video: https://www.youtube.com/watch?v=9O0qwNhTuqk