Traffic Forecasting with Adversarial Domain Adaptation in Edge-Computing Systems

Traffic forecasting is one of the important functions of intelligent transportation systems (ITS), which is of great significance to user experience and urban traffic control. Edge computing, as an emerging technology, has been a promising candidate for real-time and accurate traffic flow prediction. In this paper, we propose an adversarial domain adaptation model for traffic forecasting in an edge computing system. We design the feature mapping and the discriminator optimization objective functions. The traffic features of target domain can be aligned with the traffic features of source domain by adversarial domain adaptation training. We utilize the proposed model for different traffic prediction tasks such as traffic flow and occupancy. We also present extensive simulations by using real-world traffic dataset. We show that the proposed model can achieve better prediction accuracy than the other algorithms when the number of training dataset is insufficient.