大语言模型在道路交通领域应用:创新与挑战

[Objective] To inspire and promote the application of large language models (LLMs) in the field of road transportation. [Background] The emergence of LLMs, represented by ChatGPT, has profoundly transformed human society, driving the development of artificial intelligence technologies and significantly influencing social interactions, including those in the field of road transportation. [Methods] A brief introduction to LLMs and their main characteristics is provided. Typical applications of LLMs in the road transportation domain are listed, and their common features are summarized. [Results] LLMs can empower road traffic research and applications in several ways: reducing technical barriers between users and outcomes, helping models adapt to practical needs, enabling automatic understanding of traffic videos, alleviating the burden of text processing, and supporting autonomous driving. From another perspective, LLMs can assist in problem-solving when encountering these technical barriers. However, there are certain limitations of current LLMs, including reproducibility, lack of domain knowledge, processing speed, differences in modality the hallucination problem, limited understanding of the physical world, and privacy and security concerns.