WO2022104504A1 - 一种公路交通状态预测方法、系统、终端以及存储介质 - Google Patents

一种公路交通状态预测方法、系统、终端以及存储介质 Download PDF

Info

Publication number
WO2022104504A1
WO2022104504A1 PCT/CN2020/129300 CN2020129300W WO2022104504A1 WO 2022104504 A1 WO2022104504 A1 WO 2022104504A1 CN 2020129300 W CN2020129300 W CN 2020129300W WO 2022104504 A1 WO2022104504 A1 WO 2022104504A1
Authority
WO
WIPO (PCT)
Prior art keywords
traffic
traffic data
time
time step
spatial attention
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/129300
Other languages
English (en)
French (fr)
Inventor
叶洁瑕
赵娟娟
叶可江
须成忠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Priority to PCT/CN2020/129300 priority Critical patent/WO2022104504A1/zh
Publication of WO2022104504A1 publication Critical patent/WO2022104504A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/01Detecting movement of traffic to be counted or controlled

Definitions

  • the present application belongs to the technical field of intelligent transportation, and in particular relates to a method, system, terminal and storage medium for predicting a road traffic state.
  • deep learning can provide end-to-end training, thus avoiding the feature engineering process that machine learning relies on expert knowledge.
  • deep learning can deal with a larger amount of data, richer and heterogeneous data features, and can theoretically approximate any function, which helps it to extract more complex traffic patterns in the transportation field. Therefore, more and more scholars are turning to various algorithms in deep learning to predict traffic data.
  • the recurrent neural network and its variants are often used to extract long-term time dependence in traffic data (time dependence refers to the future traffic data of each area in the highway network is affected by its own historical traffic data), sequence-to-sequence models
  • time dependence refers to the future traffic data of each area in the highway network is affected by its own historical traffic data
  • sequence-to-sequence models For multi-step prediction of traffic data, convolutional neural network is used to divide the traffic network into grids and extract the spatial dependence of the traffic network (spatial dependence refers to the future traffic data of each region in the highway network is affected by other regions. impact of historical traffic data).
  • most of the transportation networks are not regular grid structures, which are closer to the non-Euclidean distance graph structure in nature.
  • the spatial dependence extraction algorithm based on grid division can only approximate the spatial properties roughly. It is impossible to accurately characterize the traffic network based on the graph structure.
  • Graph neural networks can construct traffic networks as graphs, and transfer deep learning algorithms to graphs to extract spatial dependencies of graph-based highway networks, use recurrent neural networks to extract temporal dependencies in traffic data, and use sequence-to-sequence model to make multi-step predictions on traffic data.
  • the spatiotemporal dependencies in the transportation network are not independent of each other, but intertwined.
  • Existing algorithms extract spatiotemporal dependencies relatively independently. They only consider that the target area is affected by the traffic data of other areas in the past period of time, and do not consider the impact of the traffic data of other areas in the past time periods on the target area. The influence of traffic data in a certain period of time is different, which leads to rough extraction of spatio-temporal dependencies in the traffic network, and accurate extraction of spatio-temporal dependencies has an important impact on the accuracy of traffic data prediction.
  • the present application provides a road traffic state prediction method, system, terminal and storage medium, aiming to solve one of the above technical problems in the prior art at least to a certain extent.
  • a road traffic state prediction method comprising:
  • Obtain the traffic road network construct a directed weighted graph according to the traffic road network, take the area where each sensor in the traffic road network is located as a node on the directed weighted graph, and collect the data collected by each sensor.
  • the historical traffic data is used as the original time series data of the corresponding node;
  • the shortest path algorithm is used to calculate the shortest spending time of any two areas, and based on the shortest spending time, the spatial attention factor of the target area and other areas in a given time period is calculated;
  • the spatial attention factor represents the degree of influence of the traffic data v ip of other regions i at the input time step p on the traffic data v jq of the target region j at the output time step q;
  • the new traffic data is input into the sequence model, and the traffic conditions of the target area are predicted in multiple steps.
  • the technical solution adopted in the embodiment of the present application further includes: the constructing a directed and authorized graph according to the traffic road network includes:
  • the traffic data generated by area i in time period p is:
  • the traffic data generated by the entire traffic network in the time period p is
  • the total traffic data of the entire traffic network in the past P time periods is
  • the traffic data of the target area j in the next Q time periods are in Represents the traffic data of area j at the future time q
  • F O is the number of traffic data predicted.
  • the technical solution adopted in the embodiment of the present application further includes: the calculation of the spatial attention factor of the target area and other areas within a given time period based on the shortest time spent includes:
  • t s >qp ⁇ M ij , and the difference between t s and qp is less than the set threshold, it means that the vehicle can reach the target area j before time step q, then the influence of v ip on v jq is strong; where t s Indicates that the vehicle does not take the shortest path or is blocked, and the time spent is greater than M ij but still within the preset time range;
  • the technical solution adopted in the embodiment of the present application further includes: after calculating the spatial attention factor of the target area and other areas within a given time period based on the shortest time spent, the following further includes:
  • the spatial attention factor is modified by a graph embedding algorithm; the specific modification method of the spatial attention factor is:
  • the distance between the co-occurrence probability and the spatial attention factor is defined as The optimization objective is to minimize the distance
  • the technical solution adopted in the embodiment of the present application further includes: the said, based on the spatial attention factor, aggregating the traffic data of other areas at each historical time step, and splicing the aggregated traffic data with the traffic data of the target area, Generating new traffic data with dynamic spatiotemporal information includes:
  • the vector represents the importance of all other region input time step p to target region j at time step q, is the aggregated traffic data for all other regions at input time step p, and the aggregated weight vector is
  • the technical solution adopted in the embodiment of the present application further includes: after generating the new traffic data including dynamic spatiotemporal information, it further includes:
  • the technical solutions adopted in the embodiments of the present application further include: the inputting the new traffic data into the sequence model, and performing multi-step prediction on the traffic conditions of the target area, specifically:
  • a road traffic state prediction system comprising:
  • Graph building module used to obtain a traffic road network, construct a directed weighted graph according to the traffic road network, and take the area where each sensor in the traffic road network is located as a node on the directed weighted graph, Take the historical traffic data collected by each sensor as the original time series data of the corresponding node;
  • Spatial attention factor calculation module used to calculate the shortest spending time of any two areas based on the directed weighted graph, using the shortest path algorithm, and calculating the target area and other areas within a given time period based on the shortest spending time
  • the spatial attention factor of wherein, the spatial attention factor represents the degree of influence of the traffic data v ip of other regions i at the input time step p on the traffic data v jq of the target region j at the output time step q;
  • Data reconstruction module Based on the spatial attention factor, it is used to aggregate the traffic data of other regions at each historical time step, and splices the aggregated traffic data with the traffic data of the target region to generate a dynamic spatial and temporal information. new traffic data;
  • Traffic prediction module used to input the new traffic data into the sequence model, and perform multi-step prediction on the traffic conditions of the target area.
  • a terminal includes a processor and a memory coupled to the processor, wherein,
  • the memory stores program instructions for implementing the highway traffic state prediction method
  • the processor is configured to execute the program instructions stored in the memory to control highway traffic state prediction.
  • a storage medium storing program instructions executable by a processor, where the program instructions are used to execute the highway traffic state prediction method.
  • the beneficial effects of the embodiments of the present application are: the road traffic state prediction method, system, terminal and storage medium of the embodiments of the present application model the traffic network as a graph and combine the accessibility between regions.
  • Prior knowledge, shortest path algorithm and graph embedding technology calculate the spatial attention factor of the traffic data of each area in the traffic network in different time periods to the traffic data of the target area in different prediction time periods, and based on the spatial attention factor Constructing new traffic data containing spatiotemporal linkage information, inputting the reconstructed new traffic data into the sequence model for multi-step prediction of traffic data in the target area, which can more accurately and comprehensively consider the spatiotemporal characteristics of the entire traffic network for the target area.
  • the influence of traffic data in different forecasting time periods so as to achieve higher precision forecasting effect.
  • FIG. 1 is a flowchart of a method for predicting a road traffic state according to an embodiment of the present application
  • FIG. 2 is a schematic structural diagram of a highway traffic state prediction system according to an embodiment of the application.
  • FIG. 3 is a schematic structural diagram of a terminal according to an embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of a storage medium according to an embodiment of the present application.
  • FIG. 1 is a flowchart of a method for predicting a road traffic state according to an embodiment of the present application.
  • the road traffic state prediction method of the embodiment of the present application includes the following steps:
  • S10 Obtain the traffic road network, construct a directed and authorized graph according to the traffic road network, take the area where each sensor used to collect traffic data in the road network is located as a node on the graph, and use the historical traffic data collected by each sensor As the original time series feature data of the corresponding node;
  • Each sensor in the road network is constantly generating new traffic data. Assuming that each sensor in the road network generates FI traffic data in each time period, the traffic data generated by area i in time period p is: The traffic data generated by the entire traffic network in the time period p is The total traffic data of the entire traffic network in the past P time periods are recorded as The traffic data of the target area j in the future Q (Q>1) time periods are: in represents the traffic data of area j at the future time q, and F O is the number of traffic data to be predicted.
  • the prediction target of the embodiment of the present invention is to use the total traffic data of the entire network in the past P time periods to predict the traffic data of the target area j in the future Q time periods, and the formula is as follows:
  • G j is the structural information related to the region j in the graph G
  • f is the graph-based deep learning framework.
  • the loss function used in the present invention is the mean square error, specifically:
  • the embodiment of the present invention uses the inter-regional accessibility prior knowledge to calculate the spatial attention factor, and because the spatial attention factor calculated based on the accessibility prior knowledge is inflexible, the spatial attention factor is further combined with the graph embedding algorithm.
  • the force factor is corrected to accurately and comprehensively describe the connection strength of different regions in different time periods, accurately capture the spatiotemporal linkage dependence in the transportation network, and facilitate the subsequent multi-step prediction of traffic data more effectively.
  • the spatial attention factor is calculated as follows: denote the traffic data of the target area j at the output time step q as v jq , denote the traffic data of other areas i at the input time step p as v ip , and denote the input time step p
  • the time difference from the output time step q is recorded as qp, (p ⁇ q)
  • the shortest path length between the target area j and other areas i is recorded as M ij (measured by time)
  • the embodiment of the present invention calculates the spatial attention factor to measure the reachability-based traffic data influence strength between other areas and the target area.
  • the calculation formula of the spatial attention factor is as follows:
  • the embodiment of the present invention adopts a graph embedding technique to correct the spatial attention factor.
  • the correction method of the spatial attention factor is as follows: Assuming that all regions learn a hidden vector, the hidden vector of the target region j at time step q is written as The hidden vectors of other regions i at time step p are The co-occurrence probability of target region j and other regions i in the hidden space is In order to make the co-occurrence probability as close as possible to the spatial attention factor e jqip , the distance between them is defined as The optimization objective is to minimize this distance In the embodiment of the present invention, the KL divergence is used as the distance function d for optimization, and the optimization formula is as follows:
  • the optimization objective can be trained separately, and the revised spatial attention factor is random.
  • the optimization objective can also be trained together with the prediction objective, in which case the revised spatial attention factor can be adapted to the traffic data.
  • the overall objective function formula is as follows:
  • the new traffic data includes the influence of the entire network in each historical time step on the traffic data of the target area at different time steps in the future;
  • the traffic data reconstruction method is as follows:
  • the vector represents the importance of all other regions input time step p to target region j at time step q. is the aggregated traffic data for all other regions at input time step p, and the aggregated weight vector is
  • the present invention aggregates the traffic data of all other areas and splices the historical traffic data of the target area at the input time step p, denoted as
  • the impact of all input time steps of the entire network on the target region is recorded as Concatenate all the output time steps of the entire network to the traffic data aggregation of all output time steps in the target area, denoted as
  • the formula is as follows:
  • S40 Input the new traffic data into the sequence model, and perform multi-step prediction on the traffic conditions of the target area in multiple time steps;
  • sequence-to-sequence model multi-step prediction formula is as follows:
  • the sequence-to-sequence model (Seq2Seq) includes an encoder and a decoder, and the encoder and the decoder are both gated recurrent unit network GRUs, specifically:
  • H p GRU_Encoder(X p ,H p-1 ) (10)
  • p is the input time step
  • q is the output time step
  • Y q-1 is the traffic data at time step q-1
  • C is the context vector output by the encoder
  • the embodiment of the present invention adopts the gated recurrent unit network GRU as the encoder and decoder of the sequence-to-sequence model, which can capture the long-term time dependence in the time series data, and has a moderate amount of parameters, short training time, and better performance. .
  • the highway traffic state prediction method of the embodiment of the present application calculates the time period of each region in the traffic network by modeling the traffic network as a graph, combining the prior knowledge of accessibility between regions, the shortest path algorithm and the graph embedding technology
  • the spatial attention factor of the traffic data of the target area in different prediction time periods, and reconstruct the new traffic data containing the spatiotemporal linkage information based on the spatial attention factor, and input the reconstructed new traffic data into the sequence model to carry out
  • the multi-step prediction of the traffic data of the target area can more accurately and comprehensively take into account the impact of the spatiotemporal characteristics of the entire traffic network on the traffic data of the target area in different prediction time periods, so as to achieve a more accurate prediction effect.
  • FIG. 2 is a schematic structural diagram of a highway traffic state prediction system according to an embodiment of the present application.
  • the highway traffic state prediction system 40 of the embodiment of the present application includes:
  • Graph building module 41 used to obtain the traffic road network, construct a directed weighted graph according to the traffic road network, take the area where each sensor used for collecting traffic data in the road network is located as a node on the graph, The collected historical traffic data is used as the original time series feature data of the corresponding node;
  • Spatial attention factor calculation module 42 used to calculate the shortest spending time of any two regions (nodes) based on the directed weighted graph, using the Dijkstra shortest path algorithm combined with the weighted adjacency matrix of the traffic road network, and calculating based on the shortest spending time
  • the spatial attention factor of the target area and other areas due to their accessibility within a given time period, and the spatial attention factor is corrected by the graph embedding algorithm to adapt to the current traffic data and dynamically extract traffic spatiotemporal dependence in the data; wherein, the spatial attention factor represents the degree of influence of the traffic data v ip of other regions i at the input time step p on the traffic data v jq of the target region j at the output time step q;
  • Data reconstruction module 43 used to aggregate the original traffic data of other regions at each historical time step based on the corrected spatial attention factor, and stitch the aggregated original traffic data with the original traffic data of the target region and perform linear enhancement , generating new traffic data containing dynamic spatiotemporal information;
  • Traffic prediction module 44 used to input the new traffic data into the sequence model, and perform multi-step prediction on the traffic conditions of the target area in multiple time steps.
  • FIG. 3 is a schematic structural diagram of a terminal according to an embodiment of the present application.
  • the terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51 .
  • the memory 52 stores program instructions for implementing the above-described road traffic state prediction method.
  • the processor 51 is configured to execute program instructions stored in the memory 52 to control the highway traffic state prediction.
  • the processor 51 may also be referred to as a CPU (Central Processing Unit, central processing unit).
  • the processor 51 may be an integrated circuit chip with signal processing capability.
  • the processor 51 may also be a general purpose processor, digital signal processor (DSP), application specific integrated circuit (ASIC), off-the-shelf programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component .
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FPGA off-the-shelf programmable gate array
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • FIG. 4 is a schematic structural diagram of a storage medium according to an embodiment of the present application.
  • the storage medium of this embodiment of the present application stores a program file 61 capable of implementing all the above methods, wherein the program file 61 may be stored in the above-mentioned storage medium in the form of a software product, and includes several instructions to make a computer device (which may It is a personal computer, a server, or a network device, etc.) or a processor that executes all or part of the steps of the methods of the various embodiments of the present invention.
  • a computer device which may It is a personal computer, a server, or a network device, etc.
  • a processor that executes all or part of the steps of the methods of the various embodiments of the present invention.
  • the aforementioned storage medium includes: U disk, removable hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes , or terminal devices such as computers, servers, mobile phones, and tablets.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种公路交通状态预测方法、系统、终端以及存储介质。所述方法包括:获取交通路网,根据交通路网构建有向有权图;基于有向有权图,采用最短路径算法计算任意两个区域的最短花费时间,基于最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子;基于空间注意力因子,在每个历史时间步聚合其他区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据;将新的交通数据输入序列到序列模型中,对目标区域的交通状况进行多步预测。上述方法可以更准确、更全面地考虑到整个交通网络的时空特征对目标区域在不同预测时间段内的交通数据影响,从而取得精度更高的预测效果。

Description

一种公路交通状态预测方法、系统、终端以及存储介质 技术领域
本申请属于智慧交通技术领域,特别涉及一种公路交通状态预测方法、系统、终端以及存储介质。
背景技术
随着GDP的增长,中国的汽车保有量不断上升。据公安部统计,截至2019年底,全国汽车保有量达2.6亿辆,与2018年底相比,增长8.83%。不断增长的汽车保有量给公路网带来了运营压力,造成交通拥堵、交通事故、环境污染等一系列问题。为了增强公路网络的运营能力,构建智能公路网络,重要措施之一是提前对公路交通数据进行准确预测。准确的交通数据预测能有效地疏导车流,缓解交通堵塞,提高人们的出行效率。近年来,越来越多的传感器、摄像头等智能设备已经被大量部署在公路网络中。这些设备能不断收集车流量、车速度、车流密度、路面状况等各种交通数据,这些维度丰富的交通大数据为交通数据的准确预测提供了新的解决思路。
目前,交通数据预测领域已经有了大量的研究成果。在早期的研究中,统计学方法较为流行,包括自回归模型、差分整合移动平均自回归模型、线性回归模型等。这类模型有着严谨的数学基础和较强的可解释性,因此被广泛应用。但这类模型只能提取到交通数据中的线性特征,而交通数据具有高度的非线性特征,因此在较为复杂的交通场景中表现较差。随着机器学习的兴起,使用支持向量机、K近邻法、随机森林、梯度下降树等机器学习模型进行交通数据的预测。上述机器学习模型能够提取交通数据中高度的非线性特征,尤其是在大量交通数据的支持下,其模型性能优于传统的统计方法,应用场景也更为广泛。
相比于传统的机器学习方法,深度学习可以提供端到端的训练,从而避免了机器学习依赖于专家知识的特征工程环节。且深度学习能处理更大量的数据、更丰富和异构的数据特征,在理论上能逼近任意函数,这种特质有助于它提取交通领域中更为复杂的交通模式。因而,越来越多的学者转向用深度学习中的各种算法进行交通数据的预测。其中,循环神经网络及其变种常用于提取交通数据中的长期时间依赖性(时间依赖性指的是公路网络中的每个区域的未来交通数据受到自身历史交通数据的影响),序列到序列模型用于交通数据多步预测,卷积神经网络用于将交通网络划分为网格,提取交通网络的空间依赖性(空间依赖性指的是公路网络中每个区域的未来交通数据受到其他区域的历史交通数据的影响)。但是,大部分的交通网络都并非规则的网格结构,它本质上更接近非欧几里得距离的图结构,基于网格划分的空间依赖性提取算法只能粗略、近似地模拟空间性质,无法准确地刻画基于图结构的交通网络。
图神经网络可以将交通网络构建为图,并将深度学习算法迁移到图上来,提取基于图的公路网络的空间依赖性,用循环神经网络去提取交通数据中的时间依赖性,用序列到序列模型去进行交通数据多步预测。然而,交通网络中的时空依赖性并不是相互独立的,而是相互交织的。现有算法都是相对独立地提取时空依赖性,它们只考虑目标区域受到其他区域在过去一段时间内的交通数据的影响,并没有考虑到其他区域过去各个时间段的交通数据对目标区域在给定时间段内的交通数据影响是不同的,导致对交通网络中时空依赖性的提取较为粗糙,而时空依赖性的准确提取对交通数据预测的精度有着重要的影响。
发明内容
本申请提供了一种公路交通状态预测方法、系统、终端以及存储介质,旨在至少在一定程度上解决现有技术中的上述技术问题之一。
为了解决上述问题,本申请提供了如下技术方案:
一种公路交通状态预测方法,包括:
获取交通路网,根据所述交通路网构建有向有权图,将所述交通路网中的每个传感器所在区域作为所述有向有权图上的一个节点,将每个传感器所收集的历史交通数据作为对应节点的原始时间序列数据;
基于所述有向有权图,采用最短路径算法计算任意两个区域的最短花费时间,基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子;其中,所述空间注意力因子表示其他区域i在输入时间步p的交通数据v ip对目标区域j在输出时间步q的交通数据v jq的影响程度;
基于所述空间注意力因子,在每个历史时间步聚合其他区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据;
将所述新的交通数据输入序列到序列模型中,对所述目标区域的交通状况进行多步预测。
本申请实施例采取的技术方案还包括:所述根据所述交通路网构建有向有权图包括:
定义有向有权图G=(V,E,A),图中每个顶点代表所述交通路网中每个传感器所在的区域,V={v 1,…,v N}是顶点集,代表图G中共有N个顶点,v i是第个i顶点,即为区域i;
Figure PCTCN2020129300-appb-000001
是边集,其中e i,j代表区域i和j之间的连接关系;A=(w ij) N×N是图G的邻接矩阵,元素w ij代表从区域i到区域j所花费的时间;
设所述路网中每个传感器在每个时间段分别会产生F I个交通数据,则区域i在时间段p产生的交通数据为
Figure PCTCN2020129300-appb-000002
整个交通网络在时间段p产生的交通数据为
Figure PCTCN2020129300-appb-000003
整个交通网络在过去P个时间段的总的交 通数据为
Figure PCTCN2020129300-appb-000004
目标区域j在未来Q个时间段的交通数据为
Figure PCTCN2020129300-appb-000005
其中
Figure PCTCN2020129300-appb-000006
表示区域j在未来时刻q的交通数据,F O为预测的交通数据数目。
本申请实施例采取的技术方案还包括:所述基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子包括:
将所述输入时间步p与输出时间步q的时间差记为q-p(p<q),将目标区域j与其他区域i之间的最短路径长度记为M ij,假设车辆在时间步p从区域i出发到目标区域j,则有以下判定条件:
如果q-p<M ij,表示车辆无法在时间步q前到达目标区域j,则判定v ip对v jq的影响强度小;
如果t s>q-p≥M ij,且t s与q-p的差值小于设定阈值,表示车辆能够在时间步q前到达目标区域j,则v ip对v jq的影响强度大;其中,t s表示车辆不走最短路径或发生堵塞,花费的时间大于M ij但仍然在预设时间范围内;
如果p-q>>M ij,表示车辆在时间步q前已经到达目标区域j且已经离开,则v ip对v jq的影响强度小;
基于所述判定规则,所述空间注意力因子计算公式为:
Figure PCTCN2020129300-appb-000007
上述公式中,空间注意力因子
Figure PCTCN2020129300-appb-000008
表示v ip对v jq的重要程度,e jqip∈[0,1];当q-p<M ij时,e iqip=0。
本申请实施例采取的技术方案还包括:所述基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子后还包括:
通过图嵌入算法对所述空间注意力因子进行修正;所述空间注意力因子修正方式具体为:
假设所有区域均学习一个隐藏向量,记目标区域j在时间步q的隐藏向量为
Figure PCTCN2020129300-appb-000009
其他区域i在时间步p的隐藏向量为
Figure PCTCN2020129300-appb-000010
目标区域j和其他区域i在隐藏空间的共现概率为
Figure PCTCN2020129300-appb-000011
定义所述共现概率与空间注意力因子之间的距离为
Figure PCTCN2020129300-appb-000012
优化目标为最小化所述距离
Figure PCTCN2020129300-appb-000013
Figure PCTCN2020129300-appb-000014
对所述优化目标进行训练,得到修正后的空间注意力因子
Figure PCTCN2020129300-appb-000015
Figure PCTCN2020129300-appb-000016
本申请实施例采取的技术方案还包括:所述基于所述空间注意力因子,在每个历史时间步聚合其他区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据包括:
重构整个网络在输入时间步p对目标区域j在时间步q的影响:
Figure PCTCN2020129300-appb-000017
上式中,向量
Figure PCTCN2020129300-appb-000018
表示所有其他区域输入时间步p对目标区域j在时间步q的重要性,
Figure PCTCN2020129300-appb-000019
是所有其他区域在输入时间步p的交通数据聚合,聚合的权重向量是
Figure PCTCN2020129300-appb-000020
本申请实施例采取的技术方案还包括:所述生成包含动态时空信息的新的交通数据后还包括:
对所述拼接后的每个输入时间步p的交通数据进行线性增强:
Figure PCTCN2020129300-appb-000021
上式中,
Figure PCTCN2020129300-appb-000022
重构后的新的交通数据为
Figure PCTCN2020129300-appb-000023
Figure PCTCN2020129300-appb-000024
本申请实施例采取的技术方案还包括:所述将所述新的交通数据输入序列到序列模型中,对所述目标区域的交通状况进行多步预测具体为:
Figure PCTCN2020129300-appb-000025
本申请实施例采取的另一技术方案为:一种公路交通状态预测系统,包括:
图构建模块:用于获取交通路网,根据所述交通路网构建有向有权图,将所述交通路网中的每个传感器所在区域作为所述有向有权图上的一个节点,将每个传感器所收集的历史交通数据作为对应节点的原始时间序列数据;
空间注意力因子计算模块:用于基于所述有向有权图,采用最短路径算法计算任意两个区域的最短花费时间,基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子;其中,所述空间注意力因子表示其他区域i在输入时间步p的交通数据v ip对目标区域j在输出时间步q的交通数据v jq的影响程度;
数据重构模块:用于基于所述空间注意力因子,在每个历史时间步聚合其他区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据;
交通预测模块:用于将所述新的交通数据输入序列到序列模型中,对所述目标区域的交通状况进行多步预测。
本申请实施例采取的又一技术方案为:一种终端,所述终端包括处理器、与所述处理器耦接的存储器,其中,
所述存储器存储有用于实现所述公路交通状态预测方法的程序指令;
所述处理器用于执行所述存储器存储的所述程序指令以控制公路交通状态预测。
本申请实施例采取的又一技术方案为:一种存储介质,存储有处理器可运行的程序指令,所述程序指令用于执行所述公路交通状态预测方法。
相对于现有技术,本申请实施例产生的有益效果在于:本申请实施例的公路交通状态预测方法、系统、终端及存储介质通过将交通网络建模为图,结合区域之间的可达性先验知识、最短路径算法和图嵌入技术计算交通网络中每个区域在不同时间段内的交通数据对目标区域在不同预测时间段的交通数据的空间注意力因子,并基于空间注意力因子重构包含时空联动信息的新交通数据,将重构的新交通数据输入序列到序列模型进行目标区域的交通数据多步预测,可以更准确、更全面地考虑到整个交通网络的时空特征对目标区域在不同预测时间段内的交通数据影响,从而取得精度更高的预测效果。
附图说明
图1是本申请实施例的公路交通状态预测方法的流程图;
图2为本申请实施例的公路交通状态预测系统结构示意图;
图3为本申请实施例的终端结构示意图;
图4为本申请实施例的存储介质的结构示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅 用以解释本申请,并不用于限定本申请。
请参阅图1,是本申请实施例的公路交通状态预测方法的流程图。本申请实施例的公路交通状态预测方法包括以下步骤:
S10:获取交通路网,根据交通路网构建有向有权图,将路网中用于收集交通数据的每个传感器所在区域作为图上的一个节点,将每个传感器所收集的历史交通数据作为对应节点的原始时间序列特征数据;
本步骤中,首先,定义有向有权图G=(V,E,A),图中每个顶点代表路网中每个传感器所在的区域,V={v 1,…,v N}是顶点集,代表图G中共有N个顶点,v i是第个i顶点,本发明实施例称其为区域i。
Figure PCTCN2020129300-appb-000026
是边集,其中e i,j代表区域i和j之间的连接关系。A=(w ij) N×N是图G的邻接矩阵,元素w ij代表从区域i到区域j所花费的时间。
路网中每个传感器都在不断地产生新的交通数据,设路网中每个传感器在每个时间段分别会产生F I个交通数据,则区域i在时间段p产生的交通数据为
Figure PCTCN2020129300-appb-000027
整个交通网络在时间段p产生的交通数据为
Figure PCTCN2020129300-appb-000028
Figure PCTCN2020129300-appb-000029
整个交通网络在过去P个时间段的总的交通数据记为
Figure PCTCN2020129300-appb-000030
目标区域j在未来Q(Q>1)个时间段的交通数据为
Figure PCTCN2020129300-appb-000031
其中
Figure PCTCN2020129300-appb-000032
表示区域j在未来时刻q的交通数据,F O为要预测的交通数据数目。本发明实施例的预测目标即为利用整个网络在过去P个时间段的总的交通数据预测目标区域j在未来Q个时间段的交通数据,公式如下:
Figure PCTCN2020129300-appb-000033
公式(1)中,G j是图G中与区域j有关的结构信息,f为基于图的深度学习框架本发明中使用的损失函数是均方误差,具体为:
Figure PCTCN2020129300-appb-000034
S20:基于有向有权图,采用Dijkstra最短路径算法结合交通路网的带权邻接矩阵计算任意两个区域(节点)的最短花费时间,基于最短花费时间计算目标区域与其他区域在各自给定的时间段内由于可达性而产生的空间注意力因子,并通过图嵌入算法对该空间注意力因子进行修正,使其自适应当前交通数据并动态提取交通数据中的时空依赖性;
本步骤中,由于车辆在交通网络中会不断移动,从而导致各区域之间的交通数据相互作用,因此,目标区域的交通数据会受到其他相关区域的历史交通数据的影响。基于该规律,本发明实施例利用区域间可达性先验知识计算空间注意力因子,又由于基于可达性先验知识计算的空间注意力因子缺乏灵活性,进一步结合图嵌入算法对空间注意力因子进行修正,从而准确全面地刻画不同区域在不同时间段的连接强度,精确捕获交通网络中的时空联动依赖性,便于后续更加有效地进行交通数据的多步预测。
具体的,空间注意力因子计算方式为:将目标区域j在输出时间步q的交通数据记为v jq,将其他区域i在输入时间步p的交通数据记为v ip,将输入时间步p与输出时间步q的时间差记为q-p,(p<q),将目标区域j与其他区域i之间的最短路径长度记为M ij(采用时间衡量),如果车辆在时间步p从区域i出发到目标区域j,则有以下判定条件:
(1)如果q-p<M ij,表示车辆无法在时间步q前到达目标区域j,此时,则判定v ip对v jq的影响强度很小。
(2)如果t s>q-p≥M ij,且t s与q-p的差值小于设定阈值(相互接近),表示车辆能够在时间步q前到达目标区域j,此时,判定v ip对v jq的影响强度很大。其中,t s表示车辆可能不走最短路径,或者路上发生了堵塞,花费的时间大于M ij但仍然在预设时间范围内。
(3)如果p-q>>M ij,表示车辆可能在时间步q前已经到达目标区域j且已经离开了一段时间,此时,判定v ip对v jq的影响很小。且时间差p-q越大,v ip对v jq的影响强度越小。
基于上述判定规则,本发明实施例通过计算空间注意力因子以衡量其他区域与目标区域之间基于可达性的交通数据影响强度,空间注意力因子计算公式如下:
Figure PCTCN2020129300-appb-000035
公式(3)中,空间注意力因子
Figure PCTCN2020129300-appb-000036
表示v ip对v jq的重要程度,其中e jqip∈[0,1]。由于本发明只考虑其他区域对目标区域的影响,因此当q-p<M ij时,表示目标区域是不可达的,即区域之间的注意力为零(即e iqip=0)。当时间差q-p越大,则注意力因子e jqip越小。
进一步地,由于基于先验可达性定义的空间注意力因子缺乏灵活性,无法自适应当前数据,因此本发明实施例采用图嵌入技术对空间注意力因子进行修正。空间注意力因子修正方式具体为:假设所有区域均学习一个隐藏向量,记目标区域j在时间步q的隐藏向量为
Figure PCTCN2020129300-appb-000037
其他区域i在时间步p的隐藏向量为
Figure PCTCN2020129300-appb-000038
目标区域j和其他区域i在隐藏空间的共现概率为
Figure PCTCN2020129300-appb-000039
Figure PCTCN2020129300-appb-000040
为了让该共现概率能尽可能接近空间注意力因子e jqip,定义它们之间的距离为
Figure PCTCN2020129300-appb-000041
优化目标即最小化该距离
Figure PCTCN2020129300-appb-000042
本发明实施例采取KL散度作为距离函数d进行优化,优化公式如下:
Figure PCTCN2020129300-appb-000043
由于∑e jqiplog(e jqip)是常数,不影响优化,因此优化目标变为:
Figure PCTCN2020129300-appb-000044
该优化目标可以单独训练,此时修正后的空间注意力因子便具备随机性。该优化目标也可以跟预测目标一起训练,此时修正后的空间注意力因子便能自适应交通数据。总目标函数公式如下:
Figure PCTCN2020129300-appb-000045
经过数据训练后,得到修正后的空间注意力因子
Figure PCTCN2020129300-appb-000046
S30:基于修正后的空间注意力因子,在每个历史时间步聚合其他区域的原始交通数据,并将聚合后的原始交通数据与目标区域的原始交通数据拼接以及线性增强,生成包含动态时空信息的新的交通数据;
本步骤中,新的交通数据中包含了每个历史时间步中整个网络对目标区域在未来不同时间步的交通数据的影响;交通数据重构方式具体为:
首先,重构整个网络在输入时间步p对目标区域j在时间步q的影响,公式如下:
Figure PCTCN2020129300-appb-000047
公式(7)中,向量
Figure PCTCN2020129300-appb-000048
表示所有其他区域输入时间步p对目标区域j在时间步q的重要性。
Figure PCTCN2020129300-appb-000049
是所有其他区域在输入时间步p的交通数据聚合,聚合的权重向量是
Figure PCTCN2020129300-appb-000050
由于目标区域自身的历史交通数据也会对自身的未来交通数据产生影响,因此本发明通过将所有其他区域的交通数据聚合与目标区域在输入时间步p的历史交通数据进行拼接,记为
Figure PCTCN2020129300-appb-000051
将整个网络所有的输入时间步对目标区域的影响记为
Figure PCTCN2020129300-appb-000052
将整个网络所有输出时间步对目标区域所有输出时间步的交通数据聚合进行拼接,记为
Figure PCTCN2020129300-appb-000053
为了减少输入特征的数目,以减少参数量,降低过拟合,进一步对拼接后的每个输入时间步p的交通数据进行线性增强,公式如下:
Figure PCTCN2020129300-appb-000054
公式(8)中,
Figure PCTCN2020129300-appb-000055
整个重构后的新的交通数据为
Figure PCTCN2020129300-appb-000056
S40:将新的交通数据输入序列到序列模型中,对目标区域在多个时间步的交通状况进行多步预测;
本步骤中,序列到序列模型多步预测公式如下:
Figure PCTCN2020129300-appb-000057
本发明实施例中,序列到序列模型(Seq2Seq)包含一个编码器和一个解码器,编码器和解码器均为门控循环单元网络GRU,具体为:
H p=GRU_Encoder(X p,H p-1)   (10)
C=H P   (11)
S q=GRU_Decoder([C||Y q-1],S q-1)   (12)
Figure PCTCN2020129300-appb-000058
上式中,p是输入时间步,q是输出时间步,
Figure PCTCN2020129300-appb-000059
是编码器的隐藏状态,
Figure PCTCN2020129300-appb-000060
是解码器的隐藏状态。Y q-1是时间步q-1的交通数据,C是编码器输出的上下文向量,
Figure PCTCN2020129300-appb-000061
是时间步q的预测输出,
Figure PCTCN2020129300-appb-000062
是可训练的参数。
基于上述,本发明实施例采用门控循环单元网络GRU作为序列到序列模型的编码器和解码器,能够捕获时间序列数据中的长期时间依赖性,且参数量适中,训练时间短,性能更好。
本申请实施例的公路交通状态预测方法通过将交通网络建模为图,结合区域之间的可达性先验知识、最短路径算法和图嵌入技术计算交通网络中每个区域在不同时间段内的交通数据对目标区域在不同预测时间段的交通数据的空间注意力因子,并基于空间注意力因子重构包含时空联动信息的新交通数据,将重构的新交通数据输入序列到序列模型进行目标区域的交通数据多步预测,可以更准确、更全面地考虑到整个交通网络的时空特征对目标区域在不同预测时间段内的交通数据影响,从而取得精度更高的预测效果。
请参阅图2,是本申请实施例的公路交通状态预测系统的结构示意图。本申请实施例的公路交通状态预测系统40包括:
图构建模块41:用于获取交通路网,根据交通路网构建有向有权图,将路网中用于收集交通数据的每个传感器所在区域作为图上的一个节点,将每个传感器所收集的历史交通数据作为对应节点的原始时间序列特征数据;
空间注意力因子计算模块42:用于基于有向有权图,采用Di jkstra最短路径算法结合交通路网的带权邻接矩阵计算任意两个区域(节点)的最短花费时间,基于最短花费时间计算目标区域与其他区域在各自给定的时间段内由于可达性而产生的空间注意力因子,并通过图嵌入算法对该空间注意力因子进行 修正,使其自适应当前交通数据并动态提取交通数据中的时空依赖性;其中,所述空间注意力因子表示其他区域i在输入时间步p的交通数据v ip对目标区域j在输出时间步q的交通数据v jq的影响程度;
数据重构模块43:用于基于修正后的空间注意力因子,在每个历史时间步聚合其他区域的原始交通数据,并将聚合后的原始交通数据与目标区域的原始交通数据拼接以及线性增强,生成包含动态时空信息的新的交通数据;
交通预测模块44:用于将新的交通数据输入序列到序列模型中,对目标区域在多个时间步的交通状况进行多步预测。
请参阅图3,为本申请实施例的终端结构示意图。该终端50包括处理器51、与处理器51耦接的存储器52。
存储器52存储有用于实现上述公路交通状态预测方法的程序指令。
处理器51用于执行存储器52存储的程序指令以控制公路交通状态预测。
其中,处理器51还可以称为CPU(Central Processing Unit,中央处理单元)。处理器51可能是一种集成电路芯片,具有信号的处理能力。处理器51还可以是通用处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、现成可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
请参阅图4,为本申请实施例的存储介质的结构示意图。本申请实施例的存储介质存储有能够实现上述所有方法的程序文件61,其中,该程序文件61可以以软件产品的形式存储在上述存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)或处理器(processor)执行本发明各个实施方式方法的全部或部分步骤。而前述的存储介质包括:U 盘、移动硬盘、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质,或者是计算机、服务器、手机、平板等终端设备。
对所公开的实施例的上述说明,使本领域专业技术人员能够实现或使用本申请。对这些实施例的多种修改对本领域的专业技术人员来说将是显而易见的,本申请中所定义的一般原理可以在不脱离本申请的精神或范围的情况下,在其它实施例中实现。因此,本申请将不会被限制于本申请所示的这些实施例,而是要符合与本申请所公开的原理和新颖特点相一致的最宽的范围。

Claims (10)

  1. 一种公路交通状态预测方法,其特征在于,包括:
    获取交通路网,根据所述交通路网构建有向有权图,将所述交通路网中的每个传感器所在区域作为所述有向有权图上的一个节点,将每个传感器所收集的历史交通数据作为对应节点的原始时间序列数据;
    基于所述有向有权图,采用最短路径算法计算任意两个区域的最短花费时间,基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子;其中,所述空间注意力因子表示其他区域i在输入时间步p的交通数据v ip对目标区域j在输出时间步q的交通数据v jq的影响程度;
    基于所述空间注意力因子,在每个历史时间步聚合其他区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据;
    将所述新的交通数据输入序列到序列模型中,对所述目标区域的交通状况进行多步预测。
  2. 根据权利要求1所述的公路交通状态预测方法,其特征在于,所述根据所述交通路网构建有向有权图包括:
    定义有向有权图G=(V,E,A),图中每个顶点代表所述交通路网中每个传感器所在的区域,V={v 1,…,v N}是顶点集,代表图G中共有N个顶点,v i是第个i顶点,即为区域i;
    Figure PCTCN2020129300-appb-100001
    是边集,其中e i,j代表区域i和j之间的连接关系;A=(w ij) N×N是图G的邻接矩阵,元素w ij代表从区域i到区域j所花费的时间;
    设所述路网中每个传感器在每个时间段分别会产生F I个交通数据,则区域i在时间段p产生的交通数据为
    Figure PCTCN2020129300-appb-100002
    整个交通网络在时间段p产生的交通数据 为
    Figure PCTCN2020129300-appb-100003
    整个交通网络在过去P个时间段的总的交通数据为
    Figure PCTCN2020129300-appb-100004
    目标区域j在未来Q个时间段的交通数据为
    Figure PCTCN2020129300-appb-100005
    其中
    Figure PCTCN2020129300-appb-100006
    表示区域j在未来时刻q的交通数据,F O为预测的交通数据数目。
  3. 根据权利要求2所述的公路交通状态预测方法,其特征在于,所述基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子包括:
    将所述输入时间步p与输出时间步q的时间差记为q-p(p<q),将目标区域j与其他区域i之间的最短路径长度记为M ij,假设车辆在时间步p从区域i出发到目标区域j,则有以下判定条件:
    如果q-p<M ij,表示车辆无法在时间步q前到达目标区域j,则判定v ip对v jq的影响强度小;
    如果t s>q-p≥M ij,且t s与q-p的差值小于设定阈值,表示车辆能够在时间步q前到达目标区域j,则v ip对v jq的影响强度大;其中,t s表示车辆不走最短路径或发生堵塞,花费的时间大于M ij但仍然在预设时间范围内;
    如果p-q>>M ij,表示车辆在时间步q前已经到达目标区域j且已经离开,则v ip对v jq的影响强度小;
    基于所述判定规则,所述空间注意力因子计算公式为:
    Figure PCTCN2020129300-appb-100007
    上述公式中,空间注意力因子
    Figure PCTCN2020129300-appb-100008
    表示v ip对v jq的重要程度,e jqip∈[0,1];当q-p<M ij时,e iqip=0。
  4. 根据权利要求3所述的公路交通状态预测方法,其特征在于,所述基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子后还包括:
    通过图嵌入算法对所述空间注意力因子进行修正;所述空间注意力因子修正方式具体为:
    假设所有区域均学习一个隐藏向量,记目标区域j在时间步q的隐藏向量为
    Figure PCTCN2020129300-appb-100009
    其他区域i在时间步p的隐藏向量为
    Figure PCTCN2020129300-appb-100010
    目标区域j和其他区域i在隐藏空间的共现概率为
    Figure PCTCN2020129300-appb-100011
    定义所述共现概率与空间注意力因子之间的距离为
    Figure PCTCN2020129300-appb-100012
    优化目标为最小化所述距离
    Figure PCTCN2020129300-appb-100013
    Figure PCTCN2020129300-appb-100014
    对所述优化目标进行训练,得到修正后的空间注意力因子
    Figure PCTCN2020129300-appb-100015
    Figure PCTCN2020129300-appb-100016
  5. 根据权利要求4所述的公路交通状态预测方法,其特征在于,所述基于所述空间注意力因子,在每个历史时间步聚合其他区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据包括:
    重构整个网络在输入时间步p对目标区域j在时间步q的影响:
    Figure PCTCN2020129300-appb-100017
    上式中,向量
    Figure PCTCN2020129300-appb-100018
    表示所有其他区域输入时间步p对目标区域j在时间步q的重要性,
    Figure PCTCN2020129300-appb-100019
    是所有其他区域在输入时间步p的交通数据聚合,聚合的权重向量是
    Figure PCTCN2020129300-appb-100020
  6. 根据权利要求5所述的公路交通状态预测方法,其特征在于,所述生成包含动态时空信息的新的交通数据后还包括:
    对所述拼接后的每个输入时间步p的交通数据进行线性增强:
    Figure PCTCN2020129300-appb-100021
    上式中,
    Figure PCTCN2020129300-appb-100022
    重构后的新的交通数据为
    Figure PCTCN2020129300-appb-100023
    Figure PCTCN2020129300-appb-100024
  7. 根据权利要求6所述的公路交通状态预测方法,其特征在于,所述将所述新的交通数据输入序列到序列模型中,对所述目标区域的交通状况进行多步预测具体为:
    Figure PCTCN2020129300-appb-100025
  8. 一种公路交通状态预测系统,其特征在于,包括:
    图构建模块:用于获取交通路网,根据所述交通路网构建有向有权图,将所述交通路网中的每个传感器所在区域作为所述有向有权图上的一个节点,将每个传感器所收集的历史交通数据作为对应节点的原始时间序列数据;
    空间注意力因子计算模块:用于基于所述有向有权图,采用最短路径算法计算任意两个区域的最短花费时间,基于所述最短花费时间计算目标区域与其他区域在给定时间段内的空间注意力因子;其中,所述空间注意力因子表示其他区域i在输入时间步p的交通数据v ip对目标区域j在输出时间步q的交通数据v jq的影响程度;
    数据重构模块:用于基于所述空间注意力因子,在每个历史时间步聚合其他 区域的交通数据,并将聚合后的交通数据与目标区域的交通数据进行拼接,生成包含动态时空信息的新的交通数据;
    交通预测模块:用于将所述新的交通数据输入序列到序列模型中,对所述目标区域的交通状况进行多步预测。
  9. 一种终端,其特征在于,所述终端包括处理器、与所述处理器耦接的存储器,其中,
    所述存储器存储有用于实现权利要求1-7任一项所述的公路交通状态预测方法的程序指令;
    所述处理器用于执行所述存储器存储的所述程序指令以控制公路交通状态预测。
  10. 一种存储介质,其特征在于,存储有处理器可运行的程序指令,所述程序指令用于执行权利要求1至7任一项所述公路交通状态预测方法。
PCT/CN2020/129300 2020-11-17 2020-11-17 一种公路交通状态预测方法、系统、终端以及存储介质 Ceased WO2022104504A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/129300 WO2022104504A1 (zh) 2020-11-17 2020-11-17 一种公路交通状态预测方法、系统、终端以及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/129300 WO2022104504A1 (zh) 2020-11-17 2020-11-17 一种公路交通状态预测方法、系统、终端以及存储介质

Publications (1)

Publication Number Publication Date
WO2022104504A1 true WO2022104504A1 (zh) 2022-05-27

Family

ID=81707947

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/129300 Ceased WO2022104504A1 (zh) 2020-11-17 2020-11-17 一种公路交通状态预测方法、系统、终端以及存储介质

Country Status (1)

Country Link
WO (1) WO2022104504A1 (zh)

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115116227A (zh) * 2022-06-30 2022-09-27 西安交通大学 一种交通流量预测方法、系统、装置及可读存储介质
CN115293428A (zh) * 2022-08-08 2022-11-04 中国科学技术大学 一种基于多模态数据融合的交通预测方法及应用
CN115469646A (zh) * 2022-09-28 2022-12-13 吉林大学 一种基于分布式模型预测控制的半挂汽车列车异质队列的协同控制方法
CN115631622A (zh) * 2022-09-21 2023-01-20 清华大学 交通状态预测方法、装置、电子设备及存储介质
CN116403397A (zh) * 2022-12-21 2023-07-07 中国人民解放军国防科技大学 一种基于深度学习的交通预测方法
CN116634560A (zh) * 2023-05-29 2023-08-22 湖南工商大学 一种基于联邦迁移学习的车辆协作定位方法
CN116884215A (zh) * 2023-07-14 2023-10-13 北京百度网讯科技有限公司 交通状态的预测、模型的生成方法、装置、设备及介质
CN116913097A (zh) * 2023-09-14 2023-10-20 江西方兴科技股份有限公司 一种交通状态预测方法及系统
CN117173894A (zh) * 2023-11-02 2023-12-05 深圳市城市交通规划设计研究中心股份有限公司 基于天气因素影响的高速公路节假日出行交通量预测方法
CN117912230A (zh) * 2023-12-12 2024-04-19 淮阴工学院 一种基于isaa-vmd-gru-rnn的交通流预测方法
CN118571021A (zh) * 2024-07-31 2024-08-30 杭州电子科技大学 基于多层注意力的图融合交通流量预测方法、介质及设备
CN118803210A (zh) * 2024-09-14 2024-10-18 北京合盛环境工程技术有限公司 一种基于云边协同的视频传输管理系统及方法
CN119400009A (zh) * 2024-10-22 2025-02-07 北京航空航天大学 一种应用于低空空域的无人机无冲突航迹规划方法及系统
CN120123439A (zh) * 2025-02-11 2025-06-10 中山大学 一种个体出行地点预测方法及系统
CN120356325A (zh) * 2025-04-16 2025-07-22 安徽宇凯公路建设有限公司 一种智能道路报警预警巡查救援联动系统及其方法
CN120998030A (zh) * 2025-09-01 2025-11-21 山东建筑大学 一种基于多尺度分解与动态图建模的交通流量预测方法
CN121278319A (zh) * 2025-12-10 2026-01-06 上海路明星光智能科技有限公司 电力动态数据处理方法、装置、设备以及存储介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180137412A1 (en) * 2016-11-16 2018-05-17 Cisco Technology, Inc. Network traffic prediction using long short term memory neural networks
CN110619430A (zh) * 2019-09-03 2019-12-27 大连理工大学 一种用于交通预测的时空注意力机制方法
CN111145541A (zh) * 2019-12-18 2020-05-12 深圳先进技术研究院 交通流量数据预测方法、存储介质和计算机设备

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180137412A1 (en) * 2016-11-16 2018-05-17 Cisco Technology, Inc. Network traffic prediction using long short term memory neural networks
CN110619430A (zh) * 2019-09-03 2019-12-27 大连理工大学 一种用于交通预测的时空注意力机制方法
CN111145541A (zh) * 2019-12-18 2020-05-12 深圳先进技术研究院 交通流量数据预测方法、存储介质和计算机设备

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ZHENG CHUANPAN, XIAOLIANG FAN, CHENG WANG, JIANZHONG QI: "GMAN: A Graph Multi-Attention Network for Traffic Prediction", ARXIV -CORNELL UNIVERSITY, 26 November 2019 (2019-11-26), XP055932183, Retrieved from the Internet <URL:https://arxiv.org/abs/1911.08415v2> [retrieved on 20220616] *
ZHISHUAI LI, LYU YISHENG, XIONG GANG: "Short-term Traffic Flow Prediction Based on Graph Convolutional Neural Network and Attention Mechanism", JOURNAL OF TRANSPORTATION ENGINEERING, vol. 19, no. 4, 15 August 2019 (2019-08-15), pages 15 - 19, 28, XP055814309, ISSN: 2096-3432, DOI: 10.13986 /j.cnki.jote.2019.04.003 *

Cited By (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115116227A (zh) * 2022-06-30 2022-09-27 西安交通大学 一种交通流量预测方法、系统、装置及可读存储介质
CN115293428A (zh) * 2022-08-08 2022-11-04 中国科学技术大学 一种基于多模态数据融合的交通预测方法及应用
CN115631622A (zh) * 2022-09-21 2023-01-20 清华大学 交通状态预测方法、装置、电子设备及存储介质
CN115469646A (zh) * 2022-09-28 2022-12-13 吉林大学 一种基于分布式模型预测控制的半挂汽车列车异质队列的协同控制方法
CN116403397A (zh) * 2022-12-21 2023-07-07 中国人民解放军国防科技大学 一种基于深度学习的交通预测方法
CN116634560A (zh) * 2023-05-29 2023-08-22 湖南工商大学 一种基于联邦迁移学习的车辆协作定位方法
CN116884215A (zh) * 2023-07-14 2023-10-13 北京百度网讯科技有限公司 交通状态的预测、模型的生成方法、装置、设备及介质
CN116884215B (zh) * 2023-07-14 2024-05-31 北京百度网讯科技有限公司 交通状态的预测、模型的生成方法、装置、设备及介质
CN116913097A (zh) * 2023-09-14 2023-10-20 江西方兴科技股份有限公司 一种交通状态预测方法及系统
CN116913097B (zh) * 2023-09-14 2024-01-19 江西方兴科技股份有限公司 一种交通状态预测方法及系统
CN117173894B (zh) * 2023-11-02 2024-03-26 深圳市城市交通规划设计研究中心股份有限公司 基于天气因素影响的高速公路节假日出行交通量预测方法
CN117173894A (zh) * 2023-11-02 2023-12-05 深圳市城市交通规划设计研究中心股份有限公司 基于天气因素影响的高速公路节假日出行交通量预测方法
CN117912230A (zh) * 2023-12-12 2024-04-19 淮阴工学院 一种基于isaa-vmd-gru-rnn的交通流预测方法
CN118571021A (zh) * 2024-07-31 2024-08-30 杭州电子科技大学 基于多层注意力的图融合交通流量预测方法、介质及设备
CN118803210A (zh) * 2024-09-14 2024-10-18 北京合盛环境工程技术有限公司 一种基于云边协同的视频传输管理系统及方法
CN119400009A (zh) * 2024-10-22 2025-02-07 北京航空航天大学 一种应用于低空空域的无人机无冲突航迹规划方法及系统
CN120123439A (zh) * 2025-02-11 2025-06-10 中山大学 一种个体出行地点预测方法及系统
CN120356325A (zh) * 2025-04-16 2025-07-22 安徽宇凯公路建设有限公司 一种智能道路报警预警巡查救援联动系统及其方法
CN120998030A (zh) * 2025-09-01 2025-11-21 山东建筑大学 一种基于多尺度分解与动态图建模的交通流量预测方法
CN121278319A (zh) * 2025-12-10 2026-01-06 上海路明星光智能科技有限公司 电力动态数据处理方法、装置、设备以及存储介质

Similar Documents

Publication Publication Date Title
WO2022104504A1 (zh) 一种公路交通状态预测方法、系统、终端以及存储介质
CN112489420B (zh) 一种公路交通状态预测方法、系统、终端以及存储介质
CN110827544B (zh) 一种基于图卷积循环神经网络的短时交通流控制方法
CN110598558B (zh) 人群密度估计方法、装置、电子设备及介质
CN112884742B (zh) 一种基于多算法融合的多目标实时检测、识别及跟踪方法
CN111612243A (zh) 交通速度预测方法、系统及存储介质
CN115424223B (zh) 图神经网络训练方法、点云特征提取方法、设备及介质
CN110400332A (zh) 一种目标检测跟踪方法、装置及计算机设备
CN113159428A (zh) 交通流量预测方法、电子设备及存储介质
US11875523B2 (en) Adaptive stereo matching optimization method and apparatus, device and storage medium
CN113052401A (zh) 盲人行走轨迹预测方法、电子设备及存储介质
CN115113165A (zh) 雷达回波外推方法、装置及系统
CN118522155B (zh) 基于自适应的交通流量预测方法、系统、装置、存储介质
CN116610919B (zh) 一种基于循环图算子神经网络的空间时间序列预测方法
CN118568470B (zh) 一种基于自适应边生成的车辆轨迹预测方法、系统及设备
CN111047078A (zh) 交通特征预测方法、系统及存储介质
CN113743594A (zh) 网络流量预测模型建立方法、装置、电子设备及存储介质
CN117456738A (zh) 一种基于etc门架数据的高速公路交通量预测方法
CN117713053A (zh) 一种融合图结构表征的电力负荷预测方法及装置
CN115830865A (zh) 基于自适应超图卷积神经网络的车流量预测方法及装置
CN119378727A (zh) 一种基于位置感知动态图卷积循环网络的交通预测方法
CN113887704A (zh) 交通信息预测方法、装置、设备及存储介质
CN110275895A (zh) 一种缺失交通数据的填充设备、装置及方法
CN119169817B (zh) 一种基于图同构时空网络的交通流预测方法、设备及介质
CN120355036A (zh) 基于堆叠模型的水位预测系统、方法、介质及计算机程序产品

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20961796

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20961796

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20961796

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 24.05.2024 )

122 Ep: pct application non-entry in european phase

Ref document number: 20961796

Country of ref document: EP

Kind code of ref document: A1