This research introduces a novel system that uses Large Language Models to construct hierarchical 3D Scene Graphs, enabling robots to understand indoor environments more comprehensively.

• Creates multi-layered spatial representations with metric-semantic information
• Enhances robot navigation capabilities through improved environmental understanding
• Utilizes precise point-cloud representation to model objects in space
• Bridges the gap between AI language understanding and physical space perception

For engineering applications, this approach represents a significant advancement in spatial cognition for autonomous systems, potentially improving robot performance in complex indoor environments where contextual understanding is critical.

Intelligent Spatial Perception by Building Hierarchical 3D Scene Graphs for Indoor Scenarios with the Help of LLMs