Machine learning-based BIM component and robot association positioning method
By constructing component relation units, their reversible observation signature library, and a topological modulation LightGlue network, the problem of unstable matching in robot localization in complex environments was solved, achieving high robustness and stable localization recovery.
Patent Information
- Application Number
- CN202610462738.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-09
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for robot localization suffer from unstable matching in environments with weak textures, areas with repetitive structures, or dynamic occlusion. They also lack global structural consistency constraints, leading to localization drift or failure and making it difficult to achieve robust localization recovery in complex indoor environments.
By constructing component relationship units and their reversible observation signature library, introducing a topological modulation LightGlue network, and combining multi-source relationship chain closed-loop consistency determination, the system utilizes prior information from the Building Information Model and robot ontology observation data to achieve cross-perspective component relationship observation and relationship consistency inversion.
It achieves highly robust positioning recovery under conditions of positioning degradation or failure, has strong anti-occlusion ability, adapts to weak texture and repetitive structure environments, has good positioning stability, and requires no additional hardware.
Smart Images
Figure CN122415968A_ABST