Abstract
The advancement of digital twin technology has significantly impacted the utilization of virtual cities in the realm of smart cities and mobility. Digital twins provide a platform for the development and testing of various mobility systems, algorithms, and policies. In this research, we introduce DTUMOS, a digital twin framework for urban mobility operating systems. DTUMOS is a versatile, open-source framework that can be flexibly and adaptably integrated into various urban mobility systems. Its novel architecture, combining an AI-based estimated time of arrival model and vehicle routing algorithm, allows DTUMOS to achieve high-speed performance while maintaining accuracy in the implementation of large-scale mobility systems. DTUMOS exhibits distinct advantages in terms of scalability, simulation speed, and visualization compared to current state-of-the-art mobility digital twins and simulations. The performance and scalability of DTUMOS are validated through the use of real data in large metropolitan cities including Seoul, New York City, and Chicago. DTUMOS' lightweight and open-source environment present opportunities for the development of various simulation-based algorithms and the quantitative evaluation of policies for future mobility systems.