AUTOSIM: Automated Urban Traffic Operation Simulation via Meta-Learning Yuanqi Qin, Wen Hua, Junchen Jin, Member, IEEE, Jun Ge, Xingyuan Dai, Lingxi Li, Senior Member, IEEE, Xiao Wang, Senior Member, IEEE, and Fei-Yue Wang, Fellow, IEEE Abstract—Online traffic simulation that feeds from online information to simulate vehicle movement in real-time has recently seen substantial advancement in the development of intelligent transportation systems and urban traffic management. It has been a challenging problem due to three aspects: 1) The diversity of traffic patterns due to heterogeneous layouts of urban intersections; 2) The nature of complex spatiotemporal correla- tions; 3) The requirement of dynamically adjusting the parame- ters of traffic models in a real-time system. To cater to these chal- lenges, this paper proposes an online traffic simulation frame- work called automated urban traffic operation simulation via meta-learning (AUTOSIM). In particular, simulation models with various intersection layouts are automatically generated using an open-source simulation tool based on static traffic geometry attributes. Through a meta-learning technique, AUTOSIM enables an automated learning process for dynamic model set- tings of traffic scenarios featured with different spatiotemporal correlations. Besides, AUTOSIM is capable of adapting traffic model parameters according to dynamic traffic information in real-time by using a meta-learner. Through computational exper- iments, we demonstrate the effectiveness of the meta-learning- based framework that is capable of providing reliable supports to real-time traffic simulation and dynamic traffic operations. Index Terms—Conditional generative adversarial network, signal- ized urban networks, short-term link speed prediction.    I. Introduction T RAFFIC simulation plays a crucial role in modeling, planning, management, and control of urban traffic sys- tems [1]–[3]. In the past decades, traffic simulation was pri- marily performed offline mainly because building a traffic simulation model with high fidelity requires extensive work of model calibrations (e.g., estimating traffic flows, turning ratios, and origin-destination matrices) [4]. A set of model cal- ibration parameters is merely suitable for a specific simula- tion scenario. However, an urban road network has complex and heterogeneous configurations for intersections. Mean- while, the traffic pattern changes dynamically because of the high stochasticity of traffic systems. This poses a big chal- lenge for the practice of using traffic simulations for traffic management and control. An online traffic simulation frame- work dynamically creates suitable traffic simulation models according to real-time traffic information. Such a framework becomes increasingly important for implementing traffic con- trol measures and analyzing their influence on various control measures. Nevertheless, most online models rely on deploy- ment of sophisticated traffic detectors and corresponding data collection in a large scale. The procedure is not a cost-effec- tive solution for practical implementation [5]. Besides, online traffic simulation meets challenges in adapt- ing its dynamic model settings for complex and heteroge- neous urban road networks traffic with real-time and limited data [5]. Real-time online traffic simulation encounters a con- tinual stream of calibration tasks for heterogeneous urban net- works with time-variant traffic patterns, which requires a sys- tem with fast-learning adaptability from limited data. Previ- ous traffic simulation studies usually rely on a large set of training samples and require a pre-defined simulation sce- nario [4]. According to the best of the authors’ knowledge, few existing studies can support real-time traffic model esti- mation for heterogeneous road networks in simulation sys- tems. To tackle the aforementioned challenges, we propose a real- time online simulation framework, AUTOSIM, for urban traf- fic operations based on an open-source microscopic traffic simulation software tool, called Simulation of Urban MObil- ity (SUMO) [6]. The AUTOSIM automatically generates static traffic simulation models with dynamic model settings according to real-time traffic observations. The core model- ing component of the framework is a meta-learner [7]–[9]. The meta-learner provides a paradigm to learn new tasks much faster than from scratch with an optimal learning model. The contributions of the framework are summarized below. 1) Create traffic simulation models considering heteroge- neous layouts of urban intersections; 2) Map traffic spatiotemporal characteristics with a wide Manuscript received June 7, 2022; revised September 7, 2022; accepted November 9, 2022. This work was supported by the National Natural Science Foundation of China (62173329). Recommended by Associate Editor Zhen Song. (Corresponding author: Xiao Wang.) Citation: Y. Q. Qin, W. Hua, J. C. Jin, J. Ge, X. Y. Dai, L. X. Li, X. Wang, and F.-Y. Wang, “AUTOSIM: Automated urban traffic operation simulation via meta-learning,” IEEE/CAA J. Autom. Sinica, vol. 10, no. 9, pp. 1871–1881, Sept. 2023. Y. Q. Qin and J. Ge are with Zhejiang Lab, Hangzhou 310030, China (e- mail: qinyq@zhejianglab.com; gejun@zhejianglab.com). W. Hua is with ECARX Technology Co., Ltd., Shanghai 201811, China (e- mail: wen.hua@ecarxgroup.com). J. C. Jin is with Zhejiang Supcon Information Company Ltd., Hangzhou 310052, China (e-mail: junchen@kth.se). X. Y. Dai and F.-Y. Wang are with the State Key Laboratory for Management and Control of Complex Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China (e-mail: xingyuan. dai@ia.ac.cn; feiyue.wang@ia.ac.cn). L. X. Li is with Indiana University-Purdue University Indianapolis (IUPUI), Indianapolis, Indiana 46202 USA (e-mail: ll7@iupui.edu). X. Wang is with School of Artificial Intelligence, Anhui University, Hefei 230039, China (e-mail: xiao.wang@ahu.edu.cn). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JAS.2023.123264 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 10, NO. 9, SEPTEMBER 2023 1871