A systematic review of generative adversarial networks for traffic state prediction: Overview, taxonomy, and future prospects

In recent years, advances in deep learning have had a significant impact in the transportation domain, notably through the use of generative adversarial networks (GAN). As an unsupervised deep learning algorithm, GANs have been widely used in the field of traffic state prediction. This paper presents a exhaustive review of the key factors influencing the process of traffic state prediction, examining how GANs are applied in this area. Unlike previous surveys that mainly concentrated on models and algorithms, this paper investigates the essential elements influencing traffic prediction. To elucidate the application of GANs in traffic state prediction, we thoroughly explore how GANs are combined with other neural networks, and how other networks utilize GAN's adversarial mechanisms to solve complex traffic prediction problems. In addition, in order to promote the further development of GAN in Intelligent Transportation Systems (ITS), this paper highlights the challenges and emerging research directions in the development of GAN, thus facilitating a thorough understanding for researchers interested in leveraging GANs for traffic prediction.