ST-GAIN: Generative Dynamic Style Transfer Structure for Missing Traffic Speed Data Imputation

The sensing coverage of roadside sensing system usually does not cover the entire road network, resulting in block missing values in traffic data. Traditional methods either adopted simple hints or graph neural networks to capture the speed variation. However, these methods fail to consider that block missing values have less surrounding values and are less susceptible to the influence of adjacent data. The paper proposes Dynamic Style Transfer-based Generative Adversarial Imputation Network (ST-GAIN) for block missing traffic speed imputation. The core idea is to adopt temporal clustering to abstract a large volume of traffic speed into a series of style data. Subsequently, a dynamic encoding network of latent style codes is performed based on the similarity between the missing speed data and the style data. These style codes are then fed into a style transfer network to guide the imputation of the missing values. Additionally, a style discriminator is used to guide style transfer during training. The experimental results demonstrate that the proposed model outperforms state-of-the-art methods by an average of more than 15% in accuracy.

I. INTRODUCTION

Since the emergence of smart city, a multitude of advanced technologies have been extensively implemented in diverse urban traffic scenarios. Despite the rapid development of roadside sensing system, missing values in traffic data are still inevitable [1]. There are usually three types of missing values: random, fiber, and block [2]. Random missing values are typically attributed to accidental interference during data transmission. The fiber missing values are usually due to accidental failures of device connections. As show in Fig. 1, the cause of block missing values are that the coverage of the accessed sensor usually does not cover the entire road network. Furthermore, the missing traffic data has a significant impact on traffic planning and traffic research [3]. For instance, the accurate estimation of carbon emissions and the provision of emission reduction strategies by the carbon regulatory platform necessitate access to second-by-second vehicle speed data. The failure to effectively address these missing data can lead to resource allocation errors or biases. However, using GPS information to solve the block missing problem of vehicle speeds for all vehicles would be costly. The objective of this paper is to present a methodology for the imputation of block missing data in the only use sensors of roadside scenarios. In block missing scenarios, the optimal imputation method should aim to maintain high accuracy while preserving the distribution and temporal dynamics of the original data. This ensures the reliability of subsequent research.