Harnessing Large Language Models for Intelligent Transportation Systems: A Systematic Review

Traffic modelling is crucial in transportation research, serving as a cornerstone for effective urban planning. Intelligent transportation systems (ITS) increasingly depend on accurate traffic modelling to manage congestion, improve safety, and optimize operations. While traditional machine learning (ML) methods have advanced the field, they often face limitations in capturing complex spatio-temporal patterns, require large labelled datasets, and lack adaptability in real-world settings. Large language models (LLMs) have emerged as promising alternatives, capable of capturing contextual relationships and multimodal patterns. Their integration into ITS is an emerging research area with the potential to transform predictive analytics and decision-making. However, challenges remain, including irregular spatial topologies, inconsistent empirical gains over simpler statistical and ML models, computational overhead, limited interpretability, and absence of domain-specific pre-training pipelines. The rapid growth of publications has made it difficult to track progress and gain a clear overview of the field. This paper presents a systematic review of 129 peer-reviewed publications addressing the application of LLMs in ITS. Beyond descriptive synthesis, it provides comparative insights and theoretical analysis by linking model architectures, fine-tuning strategies, ablation evidence, and dataset trends with empirical outcomes. Unlike prior surveys, this review goes beyond surface-level summaries by drawing cross-paper interrelationships, thus offering a fine-grained and technically grounded synthesis of LLM applications in ITS. The survey concludes by identifying open challenges and outlining future research directions, including opportunities to enhance scalability, efficiency, multimodal integration, privacy, and decision support in ITS. Together, these contributions establish a roadmap for positioning LLMs within next-generation ITS.