Enhancing Urban Intelligent Traffic Simulations of Human-Driven Car-Following Behavior Using Deep Learning Techniques

This study presents an innovative human-driven car-following model, specifically tailored for urban traffic contexts. In comparison to traditional car-following models such as the Optimal Velocity Model (OVM) and the Intelligent Driver Model (IDM), our Deep Learning based Human-Driven Car-following Model (DHCM) offers a more realistic simulation of real-world scenarios. Utilizing deep learning methodologies, we constructed this model and subsequently assessed its performance across diverse traffic conditions. Furthermore, this novel model has been integrated into a control framework, designed to extract and simulate characteristics of human-driven vehicles (HDVs), catered particularly for the emerging domain of connected and autonomous vehicles, with an emphasis on connected cruise control. Evaluations based on metrics like average absolute velocity error, total fuel consumption, and total travel distance affirm that the new model not only delivers superior performance but also enhances traffic efficiency, especially in mixed-traffic scenarios.