WO2024183151A1 - 基于多源群智感知数据的城市交通速度估计方法 - Google Patents

基于多源群智感知数据的城市交通速度估计方法 Download PDF

Info

Publication number
WO2024183151A1
WO2024183151A1 PCT/CN2023/093404 CN2023093404W WO2024183151A1 WO 2024183151 A1 WO2024183151 A1 WO 2024183151A1 CN 2023093404 W CN2023093404 W CN 2023093404W WO 2024183151 A1 WO2024183151 A1 WO 2024183151A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
speed
road
matrix
urban traffic
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/CN2023/093404
Other languages
English (en)
French (fr)
Inventor
贺诗波
张莹倩
李超
陈积明
方毅
杨秦敏
程鹏
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.)
Zhejiang University
Original Assignee
Zhejiang University
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 Zhejiang University filed Critical Zhejiang University
Priority to US18/727,108 priority Critical patent/US12674680B2/en
Publication of WO2024183151A1 publication Critical patent/WO2024183151A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/052—Detecting movement of traffic to be counted or controlled with provision for determining speed or overspeed
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34—Route searching; Route guidance
    • G01C21/36—Input/output arrangements for on-board computers
    • G01C21/3691—Retrieval, searching and output of information related to real-time traffic, weather, or environmental conditions
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
    • G08G1/0112—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from the vehicle, e.g. floating car data [FCD]
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
    • G08G1/0116—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from roadside infrastructure, e.g. beacons
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • G08G1/0133—Traffic data processing for classifying traffic situation
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
    • G08G1/0141—Measuring and analyzing of parameters relative to traffic conditions for specific applications for traffic information dissemination

Definitions

  • the present invention relates to a traffic speed estimation method, and in particular to an urban traffic speed estimation method based on multi-source crowd intelligence perception data.
  • Fine-grained large-scale urban traffic speed estimation is of great significance for urban traffic management and improvement.
  • Traditional coarse-grained traffic speed estimation only calculates the speed of road segments within a small range based on a limited number of traffic sensors.
  • mobile phones are increasingly used for navigation purposes. When mobile phone users use maps or ride-hailing apps, service providers record GPS coordinates.
  • These on-road mobile navigation data have become an important data source for traffic monitoring and sensing, and are widely used for traffic state estimation.
  • the spatial coverage of mobile navigation data is uneven. Usually more data is collected in hot spots and little or no data is collected in suburban areas, making it impossible to achieve fine-grained traffic speed estimation.
  • the purpose of this invention is to improve and standardize the deficiencies of existing research and technology, and propose a method for estimating urban traffic speed based on multi-source crowd intelligence perception data.
  • This method proposes an overall data processing flow for the traffic speed estimation method, which can promote urban traffic planning and management and has practical value.
  • a method for estimating urban traffic speed based on multi-source crowd intelligence perception data comprises the following steps:
  • Step 1 data set preprocessing: clean the original data set collected by the smartphone to obtain roadside pedestrian data and on-road navigation data;
  • Step 2 average speed calculation: use the data set in step 1 to calculate the speed X of each current road section in different time periods;
  • Step 3 fill in missing data: Use the speed X of each current section in different time periods obtained in step 2, combined with the speed H of each historical section in different time periods, to fill in the missing speed data in X and obtain the filled speed
  • Step 4 self-view velocity aggregation: use the filled velocity calculated in step 3 According to the historical speed H, the spatial dependency between different road sections is quantified, and useful neighbor information is collected to obtain the aggregated roadside pedestrian speed data Vd and on-road mobile navigation speed data Vw ;
  • Step 5 multi-view speed fusion: Based on the aggregated roadside pedestrian speed data Vd and the on-road mobile navigation speed data Vw , a multi-layer perceptron (MLP) is used to fuse the multi-source speed data according to the timestamp and whether the current speed data is filler data to obtain the fused speed. Finally, the fusion speed is corrected according to the road type characteristics to obtain a fine-grained and large-scale urban traffic speed estimation.
  • MLP multi-layer perceptron
  • step 1 specifically includes: obtaining on-road navigation data by filtering the APP usage list in the original data, that is, personal location data reported by users when using programs such as AutoNavi Map Navigation and Didi Taxi, and obtaining roadside pedestrian data based on whether there is a vehicle-mounted WIFI signal in the list by filtering the scanned WIFI signal list in the original data, that is, when the user inadvertently scans the WIFI signal of a passing vehicle while using a mobile phone and reports his or her personal location, the position of the roadside pedestrian is approximately regarded as the position of the moving vehicle.
  • step 2 specifically includes: using the data obtained by cleaning and filtering in step 1, and using the hidden Markov road network matching algorithm to project the trajectory data into the road network, so as to obtain the speed X of each current road section in different time periods, wherein the hidden Markov road network matching algorithm is also called the hidden Markov model map matching algorithm, which is a well-known technology.
  • the hidden Markov road network matching algorithm is also called the hidden Markov model map matching algorithm, which is a well-known technology.
  • the step 3 specifically includes: firstly introducing a mask matrix M to represent the missing units of the speed X:
  • the historical data is used to establish the historical speed matrix H to provide additional information to help fill in the missing data
  • another mask matrix N is introduced to represent the missing units of H:
  • ⁇ and ⁇ are learning rate parameters
  • ⁇ represents the penalty parameter
  • U and V represent the two sub-matrices decomposed from the original matrix
  • the step 4 specifically includes: using self-view speed aggregation to capture the spatial correlation between adjacent roads, aggregating neighboring road segment information that is highly correlated with the central road segment, firstly, calculating the spatial correlation e i,j between road segment i and road segment j according to the historical speed matrix, and retaining the highly correlated part and ignoring irrelevant information:
  • the fusion coefficient ai ,j between road sections is calculated according to the spatial correlation ei ,j , and then the roadside pedestrian speed data Vd and the road mobile navigation speed data Vw after self-view aggregation are obtained:
  • ⁇ represents a very small constant to prevent overflow
  • k represents a constant scaling value
  • Ni represents the adjacent set of road segment i
  • Vi uniformly represents the feature representation of the i-th row, i.e., road segment i, in the roadside pedestrian speed data Vd and the road mobile navigation speed data Vw , and will not be described separately.
  • the step 5 specifically includes: using multi-view velocity fusion to effectively fuse multi-source data, according to the timestamp feature and whether the current velocity data is filler data, the features are F d and F w , and then the features are passed through the embedding layer, The features are concatenated, and the fused speed is obtained through a multi-layer perceptron (MLP) based on the aggregated roadside pedestrian speed data Vd and the road navigation speed data Vw obtained in step 4.
  • MLP multi-layer perceptron
  • the estimated speed is corrected according to external factors (i.e., road type). Based on historical data, the speed distribution of each type of road can be easily obtained.
  • the speed distribution is regarded as a normal distribution, and the speed at the tail end of the distribution is corrected to obtain the final estimated speed.
  • the present invention has the following innovative advantages and significant effects:
  • the present invention integrates data from multiple sources, combines roadside pedestrian data and on-road navigation data, achieves 100% coverage of speed estimation for the entire road network, fills in more than 70% of missing data, and greatly reduces the cost of speed estimation, without the need to install additional loop detectors or cameras and other equipment with a single price of tens of thousands of yuan;
  • the present invention proposes a standardized processing flow for data, and the specific implementation method of each step can be varied, with strong flexibility and scalability.
  • FIG1 is a flow chart of a fine-grained large-scale urban traffic speed estimation method according to the present invention.
  • FIG2 is a schematic diagram of the correlation between neighboring road segments in step 3 of the embodiment of the present invention, in which (1) represents the mobile navigation data on the road, and (2) represents the roadside pedestrian data;
  • FIG3 is a schematic diagram of the improvement of the self-view velocity aggregation accuracy in step 3 of the embodiment of the present invention, in which (1) is the on-road mobile navigation data, and (2) is the roadside pedestrian data.
  • This example processes user data collected from a certain place from March 21, 2020 to March 28, 2020, and the data collection process is anonymous.
  • the specific variables included in the data set are shown in Table 1:
  • the implementation data set for realizing fine-grained large-scale urban traffic speed estimation is the above-mentioned user data of a certain place, and the detailed implementation steps are as follows:
  • Step 1 Clean the original data set in Table 1, delete duplicate records, etc.
  • the mobile navigation data on the road is obtained by filtering the APP usage list in the original data, such as Didi, Amap and other navigation applications.
  • the pedestrian data on the road is obtained by filtering the scanned WIFI list and filtering the data containing vehicle WIFI signals.
  • Step 2 Use the data obtained by cleaning and filtering in step 1, and use the hidden Markov network matching algorithm to project the trajectory data into the road network, so as to obtain the speed X of each road section in different time periods. Then, the example calculates the road coverage of different data sets from 8 am to 8:30 am, as shown in Table 2. Each item in the table represents the coverage rate of data under different data sets for different types of road sections in the example time period. The larger the coverage rate, the smaller the data missing:
  • Step 3 fill in the data with the speed X of each current road section in different time periods obtained in step 2.
  • the example estimates the missing speed data of the road sections with records on March 28, 2020.
  • the on-road mobile navigation data and road test pedestrian data from March 21, 2020 to March 27, 2020 are used as historical data.
  • the data of the on-road mobile navigation data and road test pedestrian data on the same day are 74.5% and 76.8%, respectively.
  • this embodiment adopts a learning rate ⁇ of 1e-4, ⁇ of 1e-4, ⁇ 1 of 0.9, and ⁇ 2 of 0.999 in the Adam optimizer.
  • the example first pre-trains the regular matrix decomposition for 10K steps to obtain a good initialization, and then runs the meta-learning process of the weighted matrix for 30K steps. Finally, the matrix decomposition process based on the weighted matrix is trained for 10K steps.
  • the proposed method is similar to the ordinary method. Matrix decomposition, tensor decomposition, linear difference, GAIN and KNN filling methods are compared, and the experimental results are shown in Table 3:
  • MAE represents mean absolute error
  • RMSE represents root mean square error
  • MAPE mean absolute percentage error. The lower the error value, the better the method. It can be seen from the table that the method proposed in this embodiment has the lowest average error value under multiple evaluation error standards compared with other models, is significantly better than other methods, and has a good missing data filling effect.
  • Step 4 using self-view speed aggregation to capture the spatial correlation between adjacent roads, aggregate the neighboring road segment information that is highly correlated with the central road segment.
  • this embodiment calculates the spatial correlation between road segments based on the historical speed matrix, and retains the highly correlated parts and ignores irrelevant information:
  • this embodiment calculates the fusion coefficient a i,j between the road sections according to the spatial correlation, and obtains the speed V after self-view aggregation:
  • the result is shown in FIG3 .
  • the aggregation method proposed in this embodiment can well combine the neighboring road segment information to improve the speed estimation accuracy of the intermediate road segment.
  • Step 5 Based on the data aggregated from the self-view in step 4, multi-view speed fusion is used to effectively fuse multi-source data, and the estimated speed is corrected according to external factors, namely the road type. The results are shown in Table 4:

Landscapes

  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Environmental & Geological Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Ecology (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Atmospheric Sciences (AREA)
  • Environmental Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Quality & Reliability (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Traffic Control Systems (AREA)
  • Navigation (AREA)

Abstract

一种基于多源群智感知数据的城市交通速度估计方法，该方法基于智能手机收集的路侧行人数据和路上导航数据，经过缺失数据填充、自视图速度聚合和多视图速度融合等步骤得到最终的估计速度。该细粒度大规模城市交通速度估计方法能够实现所有类型的道路速度估计，包含了郊区路段与小路，而不是只关注市中心主路。该方法基于数据驱动，无需在道路上安装额外装置，成本低，普适性较强，对改善交通管理和规划具有重要意义。

Description

基于多源群智感知数据的城市交通速度估计方法 技术领域
本发明涉及一种交通速度估计方法，特别涉及一种基于多源群智感知数据的城市交通速度估计方法。
背景技术
细粒度大规模城市交通速度估计对城市交通管理和改善具有着重大意义。传统的粗粒度交通速度估计仅基于有限数目的交通传感器对小范围内的路段速度进行计算。如今手机已被越来越多地用于导航目的的，当手机用户使用地图或叫车APP时，服务提供商会记录GPS坐标。这些路上移动导航数据已成为交通监测和传感的重要数据源，并被广泛应用于交通状态估计。然而，移动导航数据空间覆盖不均匀，通常在热点地区会收集到更多的数据而在郊区则很少或没有数据，无法实现细粒度的交通速度估计。除了通过导航目的获得的移动导航数据外，当用户使用微博、美团等移动应用程序时，涉及到基于位置的服务，而许多路边行人在走路时会同时使用手机。与此同时，当一些行人使用具有基于位置服务的手机应用程序时，他们会随机扫描附近的车辆WIFI信号，并报告其当前的位置。我们可以通过行人报告的WIFI列表来过滤车辆的WIFI信号，并根据路边行人的位置来近似出车辆的位置，获得的数据无需部署任何额外设备同时能够覆盖更多的人行道方面。因此可以融合路测行人数据与路上移动导航数据，以一种低成本且精确的方式获得细粒度大规模城市交通速度估计。
发明内容
本发明目的在于对现有研究和技术存在的不足之处加以完善与规范化，提出一种基于多源群智感知数据的城市交通速度估计方法。该方法对交通速度估计方法提出整体的数据处理流程，可以促进城市交通规划和管理，具有实用价值。
本发明的目的通过以下的技术方案实现：
一种基于多源群智感知数据的城市交通速度估计方法，包括以下步骤：
步骤1，数据集预处理：对智能手机收集的原始数据集进行清洗，分别得到路侧行人数据和路上导航数据；
步骤2，平均速度计算：利用步骤1中的数据集计算当前各个路段在不同时间段内的速度X；
步骤3，缺失数据填充：利用步骤2中获取的当前各个路段在不同时间段内的速度X，结合历史各个路段在不同时间段内的速度H，对X中的缺失速度数据进行填充，得到填充后的速度
步骤4，自视图速度聚合：利用步骤3中计算得到的填充后的速度根据历史速度H量化不同路段间的空间依赖性，集合有用的邻居信息，得到聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw；
步骤5，多视图速度融合：根据聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw，利用多层感知器(MLP)，根据时间戳以及当前速度数据是否为填充数据，对多源速度数据进行融合，得到融合速度最终根据道路类型特征对融合速度进行修正，得到细粒度大规模的城市交通速度估计
进一步地，所述步骤1具体包括：通过过滤原始数据中的APP使用列表，得到路上导航数据，即用户使用高德地图导航、滴滴打车等程序时所上报的个人位置数据，通过过滤原始数据中的扫描WIFI信号列表，根据列表内是否存在车载WIFI信号，得到路侧行人数据，即用户在使用手机时无意中扫描到了行驶过车辆的WIFI信号并上报个人位置时，将路侧行人位置近似看作行驶车辆位置。
进一步地，所述步骤2具体包括：利用步骤1清洗过滤得到的数据，利用隐马尔科夫路网匹配算法将轨迹数据投射到路网中，从而获得当前各个路段在不同时间段内的速度X，其中，隐马尔科夫路网匹配算法也称作隐马尔可夫模型地图匹配算法，为公知技术。
进一步地，所述步骤3具体包括：首先引入掩码矩阵M代表速度X的缺失单元：
其次，利用历史数据建立历史速度矩阵H提供额外的信息帮助填充缺失数据，引入另一个掩码矩阵N代表H的缺失单元：
定义一个加权矩阵W来衡量历史速度矩阵H中每个项的重要性，然后，利用H、W、M、N和X进行矩阵分解：


之后基于更新后的U+和V+通过矩阵分解构造损失：
之后对权重W进行更新：
其中，α和β为学习率参数；
经过迭代更新，得到一个学习完成的加权矩阵W，我们可以同时使用X和H来估算缺失数据

其中，λ表示惩罚项参数，U和V代表原矩阵分解而成的两个子矩阵。
进一步地，所述步骤4具体包括：采用自视图速度聚合来捕捉相邻道路之间的空间相关性，聚合与中心路段高度相关的邻居路段信息，首先，根据历史速度矩阵来计算路段i和路段j间的空间相关性ei,j，并且且保留高度相关的部分，忽略不相关信息：

之后，根据空间相关性ei,j计算得到了路段间的融合系数ai,j，进而得到自视图聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw：

其中，ε代表极小的常量，防止溢出，k代表常量缩放值，Ni代表路段i的邻接集合，Vi,统一代表路侧行人速度数据Vd和路上移动导航速度数据Vw中的第i行即路段i的特征表示，不再分开赘述。
进一步地，所述步骤5具体包括：采用多视图速度融合有效融合多源数据，根据表示时间戳特征以及当前速度数据是否为填充数据特征为Fd和Fw，之后将特征通过embedding层， 并将特征拼接，根据步骤4得到的聚合后的路侧行人速度数据Vd和路上导航速度数据Vw，通过多层感知器(MLP)得到融合速度
Z＝Embedding(Concat(Fd,Fw))
最后，根据外部因素(即道路类型)来修正估计的速度，根据历史数据，可以很容易地得到每种类型道路的速度分布，将速度分布看作正态分布，将落在分布尾端的速度进行修正，得到最终估计的速度
与现有技术相比，本发明具有以下创新优势及显著效果：
1)本发明集成了来自多个来源的数据，融合了路侧行人数据和路上导航数据，实现了路网全路段100％覆盖的速度估计，填充了超过70％的缺失数据，并极大降低了速度估计成本，无需额外安装环路探测器或摄像头等单件金额上万设备；
2)本发明针对数据提出规范化处理流程，每个步骤的具体实现方式可以有所变化，具有较强的灵活性与扩展性。
附图说明
图1是本发明的细粒度大范围城市交通速度估计方法流程图；
图2是本发明应用于实施例的步骤3中邻居路段相关性示意图，图中(1)表示路上移动导航数据，(2)表示路侧行人数据；
图3是本发明应用于实施例的步骤3中自视图速度聚合精度提升示意图，图中(1)为路上移动导航数据，(2)为路侧行人数据。
具体实施方式
以下结合附图对本发明的具体实施方法与工作原理作如下详述：
实施例
本实施例对从某地获取的自2020年3月21日至2020年3月28日采集到的用户数据进行处理，数据收集过程进行了匿名保护。数据集包括的具体变量如表1所示：
表1群智感知数据集

本实施例中实现细粒度大规模城市交通速度估计的实施数据集即为上述某地用户数据，其详细实施步骤具体如下：
步骤1，对表1中的原始数据集进行清洗，删除重复记录项等，路上移动导航数据通过过滤原始数据中的APP使用列表，筛选例如滴滴，高德地图等导航应用程序中获得。路测行人数据通过通过过滤扫描的WIFI列表，筛选含车辆WIFI信号的数据得到。
步骤2，利用步骤1清洗过滤得到的数据，利用隐马尔科夫路网匹配算法将轨迹数据投射到路网中，从而获得当前各个路段在不同时间段内的速度X。然后，实例计算了上午8点至8：30不同数据集的道路覆盖情况，如表2所示，表中每项代表了在所示例时间段内，不同数据集下的数据对不同类型路段的覆盖率，覆盖率越大则说明数据缺失越小：
表2同数据集上午8点至上午8：30的道路覆盖情况比较
对比发现，不同数据覆盖情况表现不同，在不同的路段中占主导地位而不是在所有情况下总是表现得更好。路上移动导航数据主要集中在主干道，而移动路侧数据分布更均匀。
步骤3，将步骤2中得到当前各个路段在不同时间段内的速度X进行数据填充，实例对2020年3月28日有记录的路段缺失的速度数据进行推算。为了提供额外的速度模式信息，以2020年3月21日至2020年3月27日的路上移动导航数据和路测行人数据作为历史数据。经计算，路上移动导航数据和路测行人数据当天的数据分别为74.5％和76.8％。具体来说，本实施例采用学习率α为1e-4，β为1e-4，Adam优化器中β1为0.9，β2为0.999。实例首先对正规矩阵分解进行10K步的预训练，得到好的初始化，然后对加权矩阵的元学习过程运行30K步。最后，对基于加权矩阵的矩阵分解过程进行10K步的训练，所提出方法与普通 矩阵分解、张量分解、线性差值、GAIN和KNN填充方法进行了对比，实验结果如表3中所示：
表3不同方法数据填充效果比较
表中MAE表示平均绝对误差，RMSE表示均方根误差，MAPE表示平均绝对百分比误差，误差值越低则方法更优越，从表中看出本实施例所提出的方法相比其他模型在多种评价误差标准下均误差值最低，明显优于其他方法，有良好的缺失数据填充效果。
步骤4，采用自视图速度聚合来捕捉相邻道路之间的空间相关性，聚合与中心路段高度相关的邻居路段信息。首先，本实施例根据历史速度矩阵来计算路段间的空间相关性，并且且保留高度相关的部分，忽略不相关信息：

结果如图2所示，各路段的地理距离越近，速度相似性越大；同时，距离越远，差异就越大。
之后，本实施例根据空间相关性计算得到了路段间的融合系数ai,j，得到了自视图聚合后的速度V：

结果如图3所示，本实施例所提出聚合方法可以很好地结合邻居路段信息来提高中间路段的速度估计精度。
步骤5，基于步骤4中经过自视图聚合的数据，采用多视图速度融合有效融合多源数据，并根据外部因素，即道路类型来修正估计的速度，结果如表4所示：
表4不同方法数据融合效果比较
从表中看出本发明所提出的方法在三种评价下的误差值均小于其他模型，效果明显优于其他方法，有良好的数据融合效果。
以上所述仅是本发明的实施方式，虽然本发明已以较佳实施例披露如上，然而并非用以限定本发明。任何熟悉本领域的技术人员，在不脱离本发明技术方案范围情况下，都可利用上述揭示的方法和技术内容对本发明技术方案做出许多可能的变动和修饰，或修改为等同变化的等效实施例。因此，凡是未脱离本发明技术方案的内容，依据本发明的技术实质对以上实施例所做的任何的简单修改、等同变化及修饰，均仍属于本发明技术方案保护的范围内。

Claims (6)

  1. 一种基于多源群智感知数据的城市交通速度估计方法，其特征在于，包括以下步骤：
    步骤1，数据集预处理：对智能手机收集的原始数据集进行清洗，分别得到路侧行人数据和路上导航数据；
    步骤2，平均速度计算：利用步骤1中的数据集计算当前各个路段在不同时间段内的速度X；
    步骤3，缺失数据填充：利用步骤2中获取的当前各个路段在不同时间段内的速度X，结合历史各个路段在不同时间段内的速度H，对X中的缺失速度数据进行填充，得到填充后的速度
    步骤4，自视图速度聚合：利用步骤3中计算得到的填充后的速度根据历史速度H量化不同路段间的空间依赖性，集合有用的邻居信息，得到聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw；
    步骤5，多视图速度融合：根据聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw，利用多层感知器，根据时间戳以及当前速度数据是否为填充数据，对多源速度数据进行融合，得到融合速度最终根据道路类型特征对融合速度进行修正，得到细粒度大规模的城市交通估计速度
  2. 根据权利要求1所述的一种基于多源群智感知数据的城市交通速度估计方法，其特征在于，所述步骤1具体包括：通过过滤原始数据中的APP使用列表，得到路上导航数据；通过过滤原始数据中的扫描WIFI信号列表，根据列表内是否存在车载WIFI信号，得到路侧行人数据。
  3. 根据权利要求1所述的一种基于多源群智感知数据的城市交通速度估计方法，其特征在于，所述步骤2具体包括：利用步骤1清洗过滤得到的数据，利用隐马尔科夫路网匹配算法将轨迹数据投射到路网中，从而获得当前各个路段在不同时间段内的速度X。
  4. 根据权利要求1所述的一种基于多源群智感知数据的城市交通速度估计方法，其特征在于，所述步骤3具体包括：首先引入掩码矩阵M代表速度X的缺失单元：
    其次，利用历史数据建立历史速度矩阵H提供额外的信息帮助填充缺失数据，引入另一个掩码矩阵N代表历史速度矩阵H的缺失单元：
    定义一个加权矩阵W来衡量历史速度矩阵H中每个项的重要性，然后，利用H、W、M、N和X进行矩阵分解：


    之后基于更新后的U+和V+通过矩阵分解构造损失：
    之后对权重W进行更新：
    其中，α和β为学习率参数；
    经过迭代更新，得到一个学习完成的加权矩阵W，可以同时使用X和H来估算缺失数据

    其中，λ表示惩罚项参数，U和V代表原矩阵分解而成的两个子矩阵。
  5. 根据权利要求1所述的一种基于多源群智感知数据的城市交通速度估计方法，其特征在于，所述步骤4具体包括：采用自视图速度聚合来捕捉相邻道路之间的空间相关性，聚合与中心路段高度相关的邻居路段信息，首先，根据历史速度矩阵来计算路段i和路段j间的空间相关性ei,j，并且且保留高度相关的部分，忽略不相关信息：

    之后，根据空间相关性ei,j计算得到路段间的融合系数ai,j，进而得到自视图聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw：

    其中，ε代表极小的常量，防止溢出，k代表常量缩放值，Ni代表路段i的邻接集合，Vi,统一代表路侧行人速度数据Vd和路上移动导航速度数据Vw中的第i行即路段i的特征表示。
  6. 根据权利要求1所述的一种基于多源群智感知数据的城市交通速度估计方法，其特征在于，所述步骤5具体包括：采用多视图速度融合多源速度数据，根据表示时间戳特征是否为填充数据特征为Fd，当前速度数据是否为填充数据特征为Fw，之后将特征通过embedding层，并将特征拼接，根据步骤4得到的聚合后的路侧行人速度数据Vd和路上移动导航速度数据Vw，通过多层感知器得到融合速度
    Z＝Embedding(Concat(Fd,Fw))
    最后，根据即道路类型特征来修正估计的速度，根据历史数据，得到每种类型道路的速度分布，将速度分布看作正态分布，将落在分布尾端的速度进行修正，得到最终估计的速度
PCT/CN2023/093404 2023-03-09 2023-05-11 基于多源群智感知数据的城市交通速度估计方法 Ceased WO2024183151A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/727,108 US12674680B2 (en) 2023-03-09 2023-05-11 Urban traffic velocity estimation method based on multi-source crowd sensing data

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310221863.0 2023-03-09
CN202310221863.0A CN116340879B (zh) 2023-03-09 2023-03-09 基于多源群智感知数据的城市交通速度估计方法

Publications (1)

Publication Number Publication Date
WO2024183151A1 true WO2024183151A1 (zh) 2024-09-12

Family

ID=86890842

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/093404 Ceased WO2024183151A1 (zh) 2023-03-09 2023-05-11 基于多源群智感知数据的城市交通速度估计方法

Country Status (3)

Country Link
US (1) US12674680B2 (zh)
CN (1) CN116340879B (zh)
WO (1) WO2024183151A1 (zh)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070208501A1 (en) * 2006-03-03 2007-09-06 Inrix, Inc. Assessing road traffic speed using data obtained from mobile data sources
CN102737510A (zh) * 2012-07-03 2012-10-17 浙江大学 一种基于移动智能终端的实时交通路况采集方法
CN104217593A (zh) * 2014-08-27 2014-12-17 北京航空航天大学 一种面向手机移动速度的路况信息实时获取方法
CN105513359A (zh) * 2016-01-27 2016-04-20 北京工业大学 一种基于智能手机移动检测的城市快速路交通状态估计方法
CN108335482A (zh) * 2017-01-20 2018-07-27 亚信蓝涛（江苏）数据科技有限公司 一种城市交通态势感知方法及可视化方法
CN109410586A (zh) * 2018-12-13 2019-03-01 中南大学 一种基于多元数据融合的交通状态检测方法
CN114120635A (zh) * 2021-11-05 2022-03-01 兴民智通(武汉)汽车技术有限公司 基于张量分解的城市路网线状缺失流量估计方法及系统
CN114445252A (zh) * 2021-11-15 2022-05-06 南方科技大学 数据补全方法、装置、电子设备及存储介质
CN114925836A (zh) * 2022-07-20 2022-08-19 中国海洋大学 基于动态多视图图神经网络的城市交通车流量推理方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
ES2373336T3 (es) * 2006-03-03 2012-02-02 Inrix, Inc. Evaluación de condiciones de tráfico de carretera utilizando datos de fuentes móviles de datos.
WO2010124138A1 (en) * 2009-04-22 2010-10-28 Inrix, Inc. Predicting expected road traffic conditions based on historical and current data
CN107765212B (zh) * 2017-10-19 2020-11-13 太原理工大学 基于群智感知思想的城市街道行人定位方法
CN109191840A (zh) * 2018-09-13 2019-01-11 电子科技大学 一种基于智能终端的实时交通状况判定方法
US11435751B2 (en) * 2018-12-28 2022-09-06 Beijing Voyager Technology Co., Ltd. Vehicle-based road obstacle identification system
CN113538898A (zh) * 2021-06-04 2021-10-22 南京美慧软件有限公司 一种基于多源数据的高速公路拥堵管控系统

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070208501A1 (en) * 2006-03-03 2007-09-06 Inrix, Inc. Assessing road traffic speed using data obtained from mobile data sources
CN102737510A (zh) * 2012-07-03 2012-10-17 浙江大学 一种基于移动智能终端的实时交通路况采集方法
CN104217593A (zh) * 2014-08-27 2014-12-17 北京航空航天大学 一种面向手机移动速度的路况信息实时获取方法
CN105513359A (zh) * 2016-01-27 2016-04-20 北京工业大学 一种基于智能手机移动检测的城市快速路交通状态估计方法
CN108335482A (zh) * 2017-01-20 2018-07-27 亚信蓝涛（江苏）数据科技有限公司 一种城市交通态势感知方法及可视化方法
CN109410586A (zh) * 2018-12-13 2019-03-01 中南大学 一种基于多元数据融合的交通状态检测方法
CN114120635A (zh) * 2021-11-05 2022-03-01 兴民智通(武汉)汽车技术有限公司 基于张量分解的城市路网线状缺失流量估计方法及系统
CN114445252A (zh) * 2021-11-15 2022-05-06 南方科技大学 数据补全方法、装置、电子设备及存储介质
CN114925836A (zh) * 2022-07-20 2022-08-19 中国海洋大学 基于动态多视图图神经网络的城市交通车流量推理方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"Thesis", 1 January 2021, SOUTHEAST UNIVERSITY, China, article XING JIPING: "Study on the Urban Network Volume Estimation Based on Multisource Data Fusion", pages: 1 - 160, XP093208470 *
ZHANG RONG; SHU YUANCHAO; YANG ZEQU; CHENG PENG; CHEN JIMING: "Hybrid Traffic Speed Modeling and Prediction Using Real-World Data", 2015 IEEE INTERNATIONAL CONGRESS ON BIG DATA, IEEE, 27 June 2015 (2015-06-27), pages 230 - 237, XP033207640, DOI: 10.1109/BigDataCongress.2015.40 *

Also Published As

Publication number Publication date
US20250102316A1 (en) 2025-03-27
CN116340879B (zh) 2026-03-03
US12674680B2 (en) 2026-07-07
CN116340879A (zh) 2023-06-27

Similar Documents

Publication Publication Date Title
Fang et al. FTPG: A fine-grained traffic prediction method with graph attention network using big trace data
CN111027430B (zh) 一种面向无人车智能评估的交通场景复杂度计算方法
US8880323B2 (en) Combining road and vehicle traffic information
CN111522893B (zh) 一种高时空分辨率道路扬尘源排放清单制备方法
Li et al. Knowledge-based trajectory completion from sparse GPS samples
CN109902259A (zh) 一种轻量级的缺失时空数据的重构方法
Liu et al. Real-time multi-task environmental perception system for traffic safety empowered by edge artificial intelligence
CN120853333A (zh) 一种基于遥感技术的滑坡预测与预警系统及方法
CN118609036A (zh) 基于图像分析的城市运行中流动商贩管理方法
CN120339962A (zh) 一种路面塌陷风险监测方法及系统
CN120659013B (zh) 终端轨迹生成方法、装置、介质和产品
Hu et al. Turning traffic surveillance cameras into intelligent sensors for traffic density estimation
CN112818935A (zh) 基于深度学习的多车道拥堵检测及持续时间预测方法及系统
Zhang et al. High-resolution large-scale urban traffic speed estimation with multi-source crowd sensing data
CN114120635B (zh) 基于张量分解的城市路网线状缺失流量估计方法及系统
CN115658666A (zh) 多源数据融合的稀疏交通流量补全方法及系统
CN120372267A (zh) 一种基于图神经网络的稀疏轨迹恢复方法
Jiang et al. Graph-based machine learning for high-resolution assessment of pedestrian-weighted exposure to air pollution
Hermes et al. A graph-based U-net model for predicting traffic in unseen cities
CN121034083A (zh) 交通信息预测方法、装置、设备、存储介质及产品
CN112487995B (zh) 一种高速公路能见度时空连续性观测方法
CN111177297B (zh) 一种基于视频和gis的动态目标速度计算优化方法
Hu et al. Turning traffic monitoring cameras into intelligent sensors for traffic density estimation
US11686590B2 (en) Correcting speed estimations using aggregated telemetry data
US20250102316A1 (en) Urban traffic velocity estimation method based on multi-source crowd sensing data

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 18727108

Country of ref document: US

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

Ref document number: 23925866

Country of ref document: EP

Kind code of ref document: A1

WWP Wipo information: published in national office

Ref document number: 18727108

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 23925866

Country of ref document: EP

Kind code of ref document: A1

WWG Wipo information: grant in national office

Ref document number: 18727108

Country of ref document: US