利用大型语言模型(LLMS)进行城市交叉口的交通管理：混合交通场景的案例（英）S-Masri-2024

This study aims to bridge this gap by investigating the capability of GPT-40-mini,an LLM tool,inmanaging urban intersections.Simulated scenarios are used to evaluate its performance in optimizingtraffic flow,reduc ing intersection wait times,and providing actionable guidance to drivers.By simulatingvarious traffic conditions and analyzing the model's responses,we seek to determine the feasibility ofdeploying LLMs for real-world traffic management.The potential impact of this research is significant.Effective traffic management systems can reducecongestion,lower emissions,and improve safety for all road users.Furthermore,integrating LLMs intotraffic control systems can enhance their adaptability and responsiveness to dynamic traffic conditions,providing a robust solution to the challenges faced by modern urban areas.Despite the promisingpotential,several challenges need to be addressed,including the model's response time and the need forreal-life data to fine-tune the system.This paper presents the methodology,results,and implications of GPT-40-mini for urban intersectionmanagement.We also discuss the challenges encountered during the study and propose future work toenhance the system's performance and applicability.Our research contributes to the growing knowledgeof AI-driven traffic management solutions,offering insights into the practical deployment of LLMs inurban environments.BackgroundTraffic management systems are part of city services,and artificial intelligence (Al)is a powerful toolthat has played an essential role in dealing with urban traffic.Traditional solutions are typically based onstatic algorithms and defined rules,which might not change effectively when traffic circumstanceschange.LLMs has the potential to enable more flexible,adaptable,explainable method as well as able toprovide actionable feedback to the drivers,traffic engineers,and policymakers [5],[6],[7],[8],[9].Earlier reviews have surfaced based on road traffic management solutions using Al and IoT technologies.The field reports on different methods such as routing mechanisms,intelligent transportation lightsolutions,or network traffic management strategies,thereby classifying them.AI is integrated to improvethe efficiency of current infrastructure,and it opens new paths for future work in managing urban roadtraffic [10].A survey on Traffic management with machine and deep learning has been conducted toexemplify the advantages and disadvantages of such techniques.They describe a general trafficmanagement architecture and discuss state-of-the-art research prototypes.These surveys highlight futureresearch directions and give insight into how machine learning and deep learning can help solve trafficmanagement problems [11].The GPT-4o-mini,a large language model,demonstrates a new traffic management approach.With theirunderstanding of human language,these pre-trained models can perform various tasks,from textgeneration to question answering [12],[13],[14],[15],[16].While th