The invention provides a deep neural network for traffic flow simulation, the traffic flow is represented by the number of vehicles at different moments, and the deep neural network comprises an input preprocessing module used for coding input element data in a traffic scene, and outputting the coded element data to an output preprocessing module; the elements in the traffic scene comprise a road network structure based on formalized expression, periodic changes of signal lamps and traffic demands; the spatial information coding module is used for extracting spatial information from the input various element data in formalized expression; and the time sequence information coding module is used for calculating the vehicle number information of each node at the current moment according to the coded various element data and the extracted space information. Compared with an existing macroscopic simulation method, the method based on the deep neural network has the advantages that the accuracy is greatly improved, meanwhile, the phenomena of accumulation and dissipation of a vehicle queue, periodic change of a vehicle proportion and the like at an intersection with signal lamps are accurately reproduced, and the method is also suitable for other places with relatively simple traffic conditions.
本发明提供一种用于交通流仿真的深度神经网络,其中交通流以不同时刻车辆的数量表示,深度神经网络包括:输入预处理模块,用于对输入的交通场景内的要素数据进行编码,交通场景内的要素包括:基于形式化表达的路网结构、信号灯的周期变化以及交通需求;空间信息编码模块,用于从输入的形式化表达的各类要素数据中提取空间信息;时序信息编码模块,用于根据编码后的各类要素数据和提取的空间信息计算当前时刻每个节点的车辆数目信息。基于本发明深度神经网络的方法相比于现有的宏观仿真方法准确性有较大的提升,同时准确的再现了有信号灯的路口处车辆队列的累积与消散、车辆比例的周期变化等现象,并且对于其它交通状况相对简单的地方同样适用。
Urban road network traffic flow simulation method based on graph neural network
一种基于图神经网络的城市路网交通流仿真方法
2024-05-31
Patent
Elektronische Ressource
Chinesisch
Simulation of the Dynamic Traffic Flow on Urban Road Network
British Library Conference Proceedings | 1993
|Urban traffic flow space-time prediction scheme based on graph convolutional neural network
Europäisches Patentamt | 2021
|Europäisches Patentamt | 2021
|Research on Urban Road Network Traffic Simulation Model
ASCE | 2014
|Traffic flow prediction method based on dynamic graph neural network
Europäisches Patentamt | 2022
|