Traffic Flow Prediction Using Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional traffic flow data collection methods in urban areas are plagued by high noise and missing data segments due to detector failures and transmission malfunctions, leading to incomplete and inaccurate data, which is insufficient for supporting effective intelligent transportation systems (ITS).
Innovation Solution
A traffic flow prediction system that collects data from multiple sources, models it in graph form, and processes it using a generative neural network to remove noise, represent spatial and temporal relationships, and utilize delayed sampling for accurate predictions, fusing feature vectors from different time windows to generate reliable traffic predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional fixed detectors are deployed to collect traffic flow data, then traffic monitoring coverage is improved, but data accuracy deteriorates due to high noise and missing segments from detector failures and transmission malfunctions
Solution Approach 1:
The patent introduces an intermediary system consisting of multiple data sources (GPS devices, mobile phones, portable computers) that act as mediators between the road network and the traffic monitoring system. These intermediaries collect and transmit traffic data without relying on fixed detectors, thereby maintaining monitoring coverage while eliminating the noise and failure issues associated with traditional detectors
Solution Approach 2:
The patent replaces the mechanical fixed detector system with an electronic/software-based solution using graph convolutional networks and recurrent units for data processing. This substitution eliminates the physical limitations and failures of mechanical detectors while maintaining or improving monitoring capabilities
2Measurement precision
If traffic data is collected from multiple sources and processed through complex graph modeling, then prediction accuracy is improved, but system complexity increases
Solution Approach 1:
The patent segments the traffic prediction problem into distinct functional modules: graph construction from multi-source data, noise filtering through encoding/decoding, spatial relationship extraction via graph convolutional networks, temporal relationship extraction via recurrent units, and prediction generation. This segmentation manages complexity by organizing the processing pipeline into manageable, specialized components
Solution Approach 2:
The patent transforms the traffic data into a graph structure that adds a spatial dimension to the traditional time-series analysis. By representing roads as nodes and connections as edges, the system captures spatial relationships alongside temporal patterns, improving prediction accuracy without proportionally increasing complexity
Data Source
AI summary
A method of generating a traffic prediction includes receiving traffic data from a plurality of reporting sources, forming a plurality of initial graphs, generating a plurality of completed graphs based on the plurality of initial graphs by at least removing noise from the plurality of initial graphs, generating a plurality of feature vectors that represent spatial relationships and temporal relationships in the plurality of completed graphs, outputting a first feature vector corresponding to a first time window as a first output, caching a copy of the first feature vector with a set of feature vectors of the plurality of feature vectors corresponding to a second time window, connecting the first feature vector with the set of feature vectors, outputting the result as a second output, fusing the first output with the second output and generating a traffic prediction based on the fused outputs.