Large Language Model-Powered Digital Traffic Engineers: The Framework and Case Studies

Abstract

This paper presents a novel Digital Traffic Engineers (DTEs) framework, leveraging Large Language Models (LLMs) to intelligently interpret human language and automate the creation of traffic control strategies. This advancement eliminates the need for manual scheme creation, reducing the workload of human traffic engineers (HTEs) and significantly improving the efficiency from requirement to control scheme generation. Experimental results in scenario understanding and traffic control underscore the potential of DTEs to effectively perform tasks traditionally managed by HTEs. This synergy between HTEs and DTEs not only streamlines traffic management processes but also paves the way for more adaptive, responsive, and environmentally friendly urban transportation solutions.