Relationship Modelling For Road Geometric Digital Twins Using Spatial Analysis And LLMs

Abstract

Geometric Digital Twins offer a robust framework for managing road environments by integrating high-fidelity object geometries and their interrelationships. However, existing relationship modelling approaches rely on manual processes or are not tailored for road objects, creating bottlenecks in large-scale digitisation. This paper proposes a hybrid method combining spatial analysis with large language models to automate topological and functional relationship modelling among road assets. Our approach captures and enriches the spatial context with semantic insights derived from relevant textual data. The resulting holistic GDT leads to more efficient information and asset management, facilitating advanced use cases, including safety assessments and simulations.