On July 1, 2025, the TRANS Research Group hosted an invited academic lecture by Dr. Wei Ma, Associate Professor in the Department of Civil and Environmental Engineering at The Hong Kong Polytechnic University (PolyU) and Director of the PolyU Mobility AI Lab, titled “Harnessing the Power of Large Language Models in Traffic Management: Insights from Two Case Studies.”
Large Language Models (LLMs) are emerging as transformative tools in traffic management, leveraging their advanced reasoning and contextual learning capabilities. This presentation delves into the practical applications of LLMs through two compelling case studies.
The first case highlights the use of an LLM-enhanced reinforcement learning framework for dynamic bus holding control strategies. By automating the generation and refinement of reward functions, this approach significantly improves the operational efficiency and stability of both single-line and multi-line bus systems. The results demonstrate enhanced adaptability, reduced passenger waiting times, and improved headway regularity compared to traditional methods.
The second case examines LLMGeovec, a novel geolocation representation paradigm that utilizes LLMs to encode rich geographic semantics from textual descriptions. Applied to spatio-temporal tasks, such as traffic flow forecasting and urban mobility optimization, LLMGeovec boosts model performance by seamlessly integrating geographic and temporal data, showcasing its effectiveness as a plug-and-play enhancement for various predictive frameworks.
These studies not only underscore the versatility of LLMs in addressing diverse challenges in traffic management but also provide actionable insights into their deployment for real-world applications.