Traffic flow analysis method and server based on spatio-temporal graph neural network

By constructing a traffic flow analysis method based on a spatiotemporal graph neural network, the system dynamically tracks the correlation graph of traffic node state changes, eliminates nonlinear causal inhibition conditions, and generates a causal propagation and diffusion path graph. This solves the problems of dynamism and reliability in existing traffic flow propagation path analysis and achieves highly reliable traffic flow propagation path analysis.

CN122392296APending Publication Date: 2026-07-14GUIZHOU INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-03
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve dynamic and highly reliable analysis of traffic flow propagation paths. Graph structures based on static geographic topology cannot adapt to dynamic changes in traffic flow propagation patterns. Methods based on statistical correlation lack the ability to determine the direction of causality in time series, resulting in a large number of noisy edges in the analysis results, making it difficult to identify the real sources of congestion and key propagation paths.

Method used

A traffic flow analysis method based on spatiotemporal graph neural network is constructed. By obtaining the traffic state parameter sequence of traffic nodes, a correlation graph of traffic node state changes is constructed, the evolution of graph structure is dynamically tracked, a causal propagation and diffusion path graph is generated, nonlinear causal inhibition conditions are eliminated, and a set of temporal causal coupling pairs reflecting the real causal transmission relationship is generated.

Benefits of technology

It achieves a breakthrough from static spatial correlation to dynamic causal logic, significantly improving the statistical stability and nonlinear consistency of causal relationships, autonomously discovering the true causal topology of traffic flow propagation, and providing highly reliable traffic congestion tracing and prediction of abnormal event propagation paths.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122392296A_ABST
    Figure CN122392296A_ABST
Patent Text Reader

Abstract

The application provides a traffic flow analysis method and a server based on a space-time graph neural network. The method comprises the following steps: first, acquiring traffic state parameter sequences of each traffic node in continuous collection periods to form a sequence set; constructing a traffic node state change correlation graph of each collection period according to the traffic state parameter change amount between adjacent periods to represent the synchronization relationship of the change amount between nodes; tracking the dynamic evolution of the correlation graph in time sequence to generate a time sequence causal coupling pair set representing the state-induced relationship between nodes; performing a nonlinear causal edge pruning operation on the time sequence causal coupling pair set to eliminate the coupling pairs satisfying the nonlinear causal inhibition condition to obtain a pruned coupling pair set; and finally generating a causal propagation diffusion path graph structure according to the state-induced direction in the pruned coupling pair. The causal influence relationship between nodes is autonomously discovered from time sequence data and a propagation path graph is constructed, so that dynamic and high-fidelity analysis of the traffic flow propagation path is realized.
Need to check novelty before this filing date? Find Prior Art