Traffic State Estimation
Traffic state estimation (TSE) aims to accurately determine the current conditions of traffic flow, such as speed and density, often using limited or incomplete data from various sources like sensors and connected vehicles. Recent research heavily utilizes deep learning models, particularly physics-informed deep learning (PIDL) and other neural network architectures like convolutional and recurrent networks, often incorporating spatial and temporal attention mechanisms to improve accuracy and efficiency, especially in sparse data scenarios. These advancements are crucial for improving traffic management, enabling real-time predictions of congestion and incidents, and optimizing traffic control strategies, ultimately leading to safer and more efficient transportation systems.
Papers
Spatial-Temporal Attention Model for Traffic State Estimation with Sparse Internet of Vehicles
Jianzhe Xue, Dongcheng Yuan, Yu Sun, Tianqi Zhang, Wenchao Xu, Haibo Zhou, Xuemin, Shen
Spatial-Temporal Generative AI for Traffic Flow Estimation with Sparse Data of Connected Vehicles
Jianzhe Xue, Yunting Xu, Dongcheng Yuan, Caoyi Zha, Hongyang Du, Haibo Zhou, Dusit Niyato