METHOD FOR ESTIMATING AND PREDICTING SHORT-TERM TRAFFIC CIRCULATION STATE OF URBAN ROAD NETWORK

摘要

A method for estimating and predicting a short-term traffic circulation state of an urban road network, comprising: (1) acquiring heterogeneous data and pre-processing same, using a road section between two urban signal intersections as a study unit, and reconstructing a velocity field of the study unit using a GASM algorithm; (2) constructing a space weight matrix of the urban road network, calculating a space-time correlation between road sections, and using TOPSIS to identify and quantify vulnerable road sections; (3) according to the reconstructed velocity field of the study unit, using an average value of velocities and selecting reasonable vulnerable road sections to construct a space-time characteristic matrix of the urban road network; and (4) estimating and predicting a traffic state of the whole road network according to Bi-ConvLSTM. The method reconstructs a velocity field of a study unit by means of heterogeneous data fusion, overcoming the limitations of prediction caused by a single data source, and uses Bi-ConvLSTM to consider the influence of traffic velocities upstream and downstream of the study unit, fully exploiting the space-time characteristics of traffic flow, further improving the accuracy of prediction, etc.