Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generation of skill-specific maps from graph world models for robotic systems

Koen de Vos, Gijs van den Brandt, Jordy Senden, Pieter Pauwels, Rene van de Molengraft, Elena Torta

TL;DR

The paper tackles the challenge of sharing and reusing maps across heterogeneous robotic systems by introducing a robot-independent graph world model derived from BIM/IFC data (or scan-to-BIM when BIM is unavailable). It encodes environmental knowledge as an RDF graph of semantic elements with geometry and properties, enabling robot-specific map generation through targeted queries. For 2D localization and navigation with 2D LIDAR, it computes maps by intersecting 3D meshes with height-specific volumes, and demonstrates integration with ROS MoveBase. The approach is validated in a large university building model and a lab environment, showing feasibility for automatic initialization and seamless integration with existing planning and localization frameworks, while highlighting future work on dynamic objects and broader sensor support.

Abstract

With the increase in the availability of Building Information Models (BIM) and (semi-) automatic tools to generate BIM from point clouds, we propose a world model architecture and algorithms to allow the use of the semantic and geometric knowledge encoded within these models to generate maps for robot localization and navigation. When heterogeneous robots are deployed within an environment, maps obtained from classical SLAM approaches might not be shared between all agents within a team of robots, e.g. due to a mismatch in sensor type, or a difference in physical robot dimensions. Our approach extracts the 3D geometry and semantic description of building elements (e.g. material, element type, color) from BIM, and represents this knowledge in a graph. Based on queries on the graph and knowledge of the skills of the robot, we can generate skill-specific maps that can be used during the execution of localization or navigation tasks. The approach is validated with data from complex build environments and integrated into existing navigation frameworks.

Generation of skill-specific maps from graph world models for robotic systems

TL;DR

The paper tackles the challenge of sharing and reusing maps across heterogeneous robotic systems by introducing a robot-independent graph world model derived from BIM/IFC data (or scan-to-BIM when BIM is unavailable). It encodes environmental knowledge as an RDF graph of semantic elements with geometry and properties, enabling robot-specific map generation through targeted queries. For 2D localization and navigation with 2D LIDAR, it computes maps by intersecting 3D meshes with height-specific volumes, and demonstrates integration with ROS MoveBase. The approach is validated in a large university building model and a lab environment, showing feasibility for automatic initialization and seamless integration with existing planning and localization frameworks, while highlighting future work on dynamic objects and broader sensor support.

Abstract

With the increase in the availability of Building Information Models (BIM) and (semi-) automatic tools to generate BIM from point clouds, we propose a world model architecture and algorithms to allow the use of the semantic and geometric knowledge encoded within these models to generate maps for robot localization and navigation. When heterogeneous robots are deployed within an environment, maps obtained from classical SLAM approaches might not be shared between all agents within a team of robots, e.g. due to a mismatch in sensor type, or a difference in physical robot dimensions. Our approach extracts the 3D geometry and semantic description of building elements (e.g. material, element type, color) from BIM, and represents this knowledge in a graph. Based on queries on the graph and knowledge of the skills of the robot, we can generate skill-specific maps that can be used during the execution of localization or navigation tasks. The approach is validated with data from complex build environments and integrated into existing navigation frameworks.
Paper Structure (10 sections, 1 equation, 9 figures, 1 table, 3 algorithms)

This paper contains 10 sections, 1 equation, 9 figures, 1 table, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Graph World Model
  4. Graph World Model: implementation and initialization
  5. Graph initialization
  6. Generation of skill-specific maps
  7. 2D maps for localization
  8. 2D navigation maps
  9. Integration
  10. Discussion and conclusion

Figures (9)

  • Figure 1: Conceptual architecture. Blue: provided input. Green: novel algorithms. Orange: algorithm's output.
  • Figure 2: Representation of the mesh for Wall1 and positioning of the global and local reference frames.
  • Figure 3: Render of the BIM model and physical setup of the simplified lab environment
  • Figure 4: RDF graph representing some of the adjacent elements to the Space. Red: node representing a space. Blue: nodes representing adjacent elements to the space. Green nodes representing properties of the elements.
  • Figure 5: Generated occupancy grid maps of the Neuron building at Eindhoven University of Technology introduced in Figure \ref{['fig:neuron']}. Note the absence of dynamic objects and furniture in the maps compared to the real world. Left column: Localization maps without glass elements. Right column: Navigation maps. Bottom row: Closeup of maps on the top row, which includes the space depicted in Figure \ref{['fig:neuron']}.
  • ...and 4 more figures

Theorems & Definitions (1)

  • Definition 1: Mesh