Robustness traffic flow prediction method based on multitask graph convolutional network

Abstract

The invention relates to a robustness traffic flow prediction method based on a multitask graph convolutional network. The spatial-temporal relationship of traffic data is extracted by adopting GraphConvolutional Networks and combining time domain convolution, and the generalization ability of the model is enhanced by adopting a multi-task learning MTL (Multi-Task Learning) structure so as to resist traffic data missing and abnormality. According to the invention, a multitask graph convolution model combining graph convolution and multitask learning is designed, so that the model can predicttraffic flow more stably under the condition of abnormal data. The model provided by the invention has a multi-output structure, and can output three tasks at the same time. One task is to output traffic flow data in a target time period, and the remaining two tasks are auxiliary tasks for predicting the same road network in different time periods, namely a time period 15 minutes before the targettask and a time period 15 minutes after the target task. The robustness of model parameters can be realized by simultaneously training a target time period and early and later traffic flow predictiontasks thereof.

Description

technical field [0001] The invention relates to the fields of intelligent transportation and deep learning, in particular to a robust traffic flow prediction method based on a multi-task graph convolutional network. Background technique [0002] Traffic flow forecasting is a critical part of realizing Intelligent Transportation Systems (ITS) in smart cities. The purpose of traffic forecasting is to predict the traffic conditions of the future road network based on historical traffic data. It plays an important role in many practical applications. Accurate traffic condition prediction is the basis of effective traffic management, and it is a key method to guide vehicles more reasonably and improve the operating efficiency of road network. In addition, traffic flow is an important indicator to detect traffic conditions in the traffic system. It will provide important traffic information for other important tasks in ITS, such as estimated time of arrival and route planning. ...

AI Technical Summary

Problems solved by technology

However, in reality, traffic transmission network delays and sensor failures, as well as traffic anomalies caused by weather, traffic accidents, and large-scale events (such as: international football matches, marathons, etc.), make traffic data produce outliers and missing values, which will lead to We Wrongly Forecast Traffic Data and Plan Long-Term Traffic Routes

Specifically, intelligent transportation systems can collect traffic data from various fixed and mobile sensors, but fixed sensors such as loop detectors and roadside cameras have limited spatial coverage, while mobile sensors such as GPS sensors have limited spatial coverage. High-precision data collection is also unstable

And in terms of communication transmission, the information transmission of equipment and data centers will be restricted by the environment and network delay, which will lead to the lack and abnormality of traffic information

Therefore, how to solve the abnormal data, the model can still predict the traffic conditions stably, is a very challenging topic

Images