Zhang, Hua (Tongji University), Zhang, Zijing (Tongji University), Lu, Ruicheng (tongji university)

Community-Aware Road Networks Embedding for Shortest-Path Distances Queries

Scheduled for presentation during the Regular Session "S34a-Data-Driven Optimization and Governance in Intelligent Urban Mobility" (FR-LM-T34), Friday, November 21, 2025, 11:10−11:30, Surfers Paradise 1

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

Fast and accurate acquisition of the shortest-path distance (SPD) is a key technology in dynamic route planning. Traditional graph search algorithms (e.g., Dijkstra’s algorithm) suffer from inefficiency, while index-based methods struggle to balance storage costs and accuracy. To address these limitations, road network embedding (RNE) methods have emerged. However, existing RNE models fail to capture directional asymmetry in directed road networks and incur high training costs. This paper proposes a community-aware directed road network embedding model. Our approach introduces a directional distance metric based on projection length to represent asymmetric shortest-path distances in the embedding space and employs an accelerated training method based on the Girvan-Newman community detection algorithm, which significantly reduces training time without compromising prediction accuracy. Experiments on real-world road networks in Shanghai and Seoul demonstrate that our model achieves 22.7%–32.6% lower prediction errors (RMSE and MAPE) and up to 98% faster training compared to state-of-the-art RNE methods for SPD, which is proven this technology holds great potential for lowering training overhead and accelerating learning in real-time route planning tasks