Method for automatically identifying and positioning congestion source based on graph neural network

Abstract

The invention relates to the field of artificial intelligence enabling traffic control, and particularly discloses a congestion source automatic identification and positioning method based on a graph neural network, and the method comprises the steps: obtaining the GPS driving tracks of N congestion vehicles in J congestion scenes through a map big data platform, and generating corresponding J training samples; constructing an inter-vehicle relation matrix of each frame of N congested vehicles of each training sample; taking congestion source vehicle positioning as an output label, and training a space-time convolutional network to obtain a prediction model in a congestion scene; and the model is used for identifying and positioning a congestion source in a real scene. According to the method, the geographic position of the congestion source can be more accurately given through the steps.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence-enabled traffic management, and in particular to a method for automatically identifying and locating the source of congestion based on a graph neural network. Background Art Congestion source location and identification can more accurately identify and analyze specific traffic congestion points, thereby facilitating the implementation of more effective measures to alleviate and resolve traffic congestion. This has significant implications for improving the refinement of traffic management, optimizing traffic planning and design, enhancing the public travel experience, and reducing the risk of traffic accidents. Traditional methods rely on analysis of fixed detector networks, manual inspections and reporting, analysis based on theoretical traffic flow models, and limited floating vehicle data. However, each of these approaches has its own drawbacks. Fixed detectors are expensive to install and maintain, and they can create blind spots. Manual inspections and monitoring centers require significant manpower. Traffic flow models struggle to identify complex or dynamic sources of congestion. Actual traffic systems are complex and dynamic, and simplified model assumptions and parameter errors can distort positioning results. Furthermore, methods based on floating vehicle data still exhibit poor accuracy. Summary of the Invention

[0002] The present invention provides a method for automatically identifying and locating the source of congestion based on a graph neural network. The technical problem it solves is how to more accurately identify and locate the source of congestion using the existing GPS data of congested road section maps.

[0003] To solve the above technical problems, the present invention provides a method for automatically identifying and locating congestion sources based on a graph neural network, comprising the following steps:

[0004] S1. Obtain the GPS driving trajectories of N congested vehicles in J congested scenarios through the map big data platform and generate corresponding J training samples;

[0005] S2, constructing the inter-vehicle relationship matrix of N congested vehicles in the training sample at each sampling time interval by constructing the mutual influence function of vehicle operation in the metric space;

[0006] S3. Using the training samples constructed from the vehicle's GPS driving trajectory as input and the location information of the vehicle at the source of the congestion as output, the spatiotemporal convolutional network is trained to obtain a prediction model.

[0007] S4. Input the GPS driving trajectory information of congested vehicles in real congestion scenarios into the trained model to obtain the GPS location information of the vehicles at the source of the congestion.

[0008] Furthermore, in step S1, the driving trajectories of N motor vehicles within a congestion time Δt are analyzed at equal time intervals T. fra The F frame GPS coordinates formed by sampling, that is, the training sample 1 frame of data The spatial positions of N motor vehicles at a certain moment are recorded, that is, The GPS position of vehicle n in frame f is identified as 1≤f≤F is the frame number, 1≤n≤N is the vehicle number, (x,y) is the GPS latitude and longitude coordinates of the vehicle; the vehicle number at the source of the congestion is N0, and its corresponding GPS location is

[0009] Furthermore, in step S2, the mutual influence function is used Metrics, expressed as:

[0010]

[0011] in, represents the relative displacement between vehicle i and vehicle j The model, Represents the relative displacement between vehicle i and the center coordinate of the congestion source area The modulus of the effect, D is the maximum range distance that can have an impact.

[0012] Furthermore, in step S2, the vehicle relationship matrix corresponding to the coordinates of the N vehicles in the t-th frame is The element in row i and column j is

[0013] Furthermore, the step S3 includes the steps of:

[0014] S31, constructing a continuous F frame set The matrix form of the position coordinates of N vehicles is

[0015]

[0016] S32, will Input spatial graph convolutional network layer, based on A t V t Perform convolution, the output of the t-th frame spatial graph convolution network layer 1≤t≤T0, where D is A t The degree matrix, W is a randomly initialized 2×2 dimensional matrix, I is the identity matrix, σ(·) is the activation function, and the output set of the graph convolution layer is obtained.

[0017] S33, Input Convolution is performed on the L-layer convolutional neural network, and the geographic location coordinate matrix is output after passing through the TCN network.

[0018] S34. The design uses the average displacement error between the model output and the center coordinates of the congestion source area during training as the loss function for calculating the training error, and uses the stochastic gradient descent algorithm to learn the weight value. The training termination condition is to meet the maximum number of training cycles and save the trained prediction model.

[0019] The present invention provides a method for automatically identifying and locating the source of congestion based on a graph neural network. Compared with existing technical solutions, it has the following advantages:

[0020] 1) There is no need to consider the complex and specific causes of congestion (congestion sources include traffic accidents, abnormal vehicle driving, natural disasters, and other complex factors. If on-site image recognition is used, there are many factors to consider, which involves complex image recognition and is difficult to consider completely). The vehicle at the source of congestion can be accurately located only through specific map information data; 2) Comprehensively consider the relative displacement between vehicles on congested roads and the impact of congestion sources to construct a relationship matrix between vehicles and provide prior knowledge for the spatiotemporal convolutional network. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flowchart of a method for identifying and locating the source of congestion based on a graph neural network provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following describes the embodiments of the present invention in detail with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and are not to be construed as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of protection of the present invention. Many changes may be made to the present invention without departing from the spirit and scope of the present invention.

[0023] The map big data platform in step S1 mainly comes from vehicle GPS coordinate data provided by Baidu Maps or Amap;

[0024] In step S2, the proportional coefficients λ0 = 0.8, λ1 = 1, and the maximum range distance between vehicles that can have an impact is D = 3.5;

[0025] Activation function in step S3 Here, a = 0.25, the convolution kernel size is 2, the number of graph convolution layers is 1, and the total number of space-time convolution layers is L = 5. The loss function is the average displacement error between the predicted coordinates and the center of the actual congestion source area. Each training batch has 128 samples, and the model is trained using stochastic gradient descent (SGD) for 100 cycles with a learning rate of 0.01.

[0026] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A method for automatically identifying and locating congestion sources based on graph neural networks, characterized in that: Including steps: S1. Obtain the GPS driving trajectories of N congested vehicles in J congested scenarios through the map big data platform and generate corresponding J training samples; S2, constructing the inter-vehicle relationship matrix of N congested vehicles in the training sample at each sampling time interval by constructing the mutual influence function of vehicle operation in the metric space; S3. Using the training samples constructed from the vehicle's GPS driving trajectory as input and the location information of the vehicle at the source of the congestion as output, the spatiotemporal convolutional network is trained to obtain a prediction model. S4. Input the GPS driving trajectory information of congested vehicles in real congestion scenarios into the trained model to obtain the GPS location information of the vehicles at the source of the congestion.

2. The method for automatically identifying and locating congestion sources based on graph neural networks according to claim 1 is characterized in that: In step S1, the driving trajectories of N motor vehicles within a congestion time Δt are recorded at equal time intervals T. fra The F frame GPS coordinates formed by sampling, that is, the training sample 1 frame of data The spatial positions of N motor vehicles at a certain moment are recorded, that is, The GPS position of vehicle n in frame f is identified as 1≤f≤F is the frame number, 1≤n≤N is the vehicle number, (x,y) is the GPS latitude and longitude coordinates of the vehicle; the vehicle number at the source of the congestion is N0, and its corresponding GPS location is 3. The method for automatically identifying and locating congestion sources based on graph neural networks according to claim 2 is characterized in that: In step S2, the mutual influence function is used Metrics, expressed as: in, represents the relative displacement between vehicle i and vehicle j The model, Represents the relative displacement between vehicle i and the center coordinate of the congestion source area The modulus of the effect, D is the maximum range distance that can have an impact.

4. The method for automatically identifying and locating congestion sources based on graph neural networks according to claim 3 is characterized in that: In step S2, the vehicle relationship matrix corresponding to the coordinates of the N motor vehicles in the tth frame is The element in row i and column j is 5. The method for automatically identifying and locating congestion sources based on graph neural networks according to claim 4 is characterized in that: The step S3 comprises the steps of: S31, constructing a continuous F frame set The matrix form of the position coordinates of N vehicles is S32, will Input spatial graph convolutional network layer, based on A t V t Perform convolution, the output of the t-th frame spatial graph convolution network layer 1≤t≤T0, where D is A t The degree matrix, W is a randomly initialized 2×2 dimensional matrix, I is the identity matrix, σ(·) is the activation function, and the output set of the graph convolution layer is obtained. S33, Input Convolution is performed on the L-layer convolutional neural network, and the geographic location coordinate matrix is output after passing through the TCN network. S34. The design uses the average displacement error between the model output and the center coordinates of the congestion source area during training as the loss function for calculating the training error, and uses the stochastic gradient descent algorithm to learn the weight value. The training termination condition is to meet the maximum number of training cycles and save the trained prediction model.

Images