Lane-level Traffic Flow Prediction by Integrating Heterogeneous Data and Dynamic Graphs: An Adaptive Multi-scale Spatio-Temporal Convolutional Network and Reinforcement Learning Collaborative Optimization Method

The dataset covers data from multiple cities and regions, including various scenarios such as urban roads and highways. The data sources are diverse, encompassing historical traffic flow, real-time vehicle speeds, traffic events, weather conditions, and road conditions. The data is collected from various sources, including traffic monitoring systems, vehicle sensors, traffic management departments, and meteorological departments. These data enable the paper to capture the spatio-temporal characteristics of traffic flow, micro-driving behavior, and external environmental factors in multiple dimensions. All the data undergoes precise spatio-temporal alignment. Time alignment is performed using linear interpolation to unify real-time vehicle speed and traffic flow data to the same timestamp. Spatial alignment uses GIS (Geographic Information System) technology to ensure the data is accurately matched to each lane, ensuring spatial precision.