Integrated command and dispatch system and method based on cloud platform

Abstract

The present application relates to the field of cloud platform technology, and more specifically discloses a cloud platform-based integrated command and dispatch system and method, which utilizes real-time traffic data and road information, combines the adjacency relationship between nodes and the planned path, extracts traffic-road topology features and planned path features, and obtains dispatch path classification features by integrating these features. Finally, the rationality of the current planned path is judged by a classifier to optimize the operational efficiency of urban rail transit.

Description

Technical Field

[0001] The present application relates to the field of cloud platform technology, and more specifically, to a cloud platform-based integrated command and dispatch system and method. Background Art

[0002] In recent years, with the rapid development of urban rail transit, the number of urban rail transit lines has also increased dramatically. Many cities have changed from single lines to mesh lines. Accordingly, it is necessary to integrate the original independent command and control systems to form a unified dispatching system. Previously, the control of each line relied on a central controller, as well as the corresponding equipment and controllers, resulting in low utilization of a large number of resources. The operating cost of rail transit is positively correlated with the line. With the increase in lines, the operating cost naturally increases, which brings a considerable burden to the construction of urban rail transit. At the same time, due to the increase in lines, the application data generated during the operation of rail transit has increased exponentially. These newly added valid data need to be further processed in order to better serve urban rail transit. The processing of this part of the data is also a new problem encountered by the current rail transit operation. While challenging, it also means that the construction of urban rail transit is getting closer and closer to the era of big data.

[0003] Therefore, a cloud platform-based integrated command and dispatch system and method is desired. Summary of the invention

[0004] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a cloud platform-based integrated command and dispatch system and method, which uses real-time traffic and road information to determine the rationality of urban rail transit planning paths through feature extraction and classifiers.

[0005] Accordingly, according to one aspect of the present application, a cloud platform-based integrated command and dispatch system is provided, which includes:

[0006] The scheduling data acquisition module is used to obtain the traffic data of each node in each time period, the adjacency matrix between multiple nodes of the road information, and the current planned path;

[0007] A scheduling data processing module, used to extract a traffic-road topology feature matrix and a planned path feature vector from the flow data of each node in each time period, the adjacency matrix between multiple nodes, and the current planned path;

[0008] A scheduling data fusion module, used to construct a scheduling path classification feature vector between the traffic-road topology feature matrix and the planning path feature vector, and perform expected activation of the parameter space on the scheduling path classification feature vector to obtain an optimized scheduling path classification feature vector;

[0009] The scheduling data analysis module is used to pass the optimized scheduling path classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the current planned path is reasonable.

[0010] According to another aspect of the present application, a cloud platform-based integrated command and dispatch method is provided, which includes:

[0011] Obtain the traffic data of each node in each time period, obtain the adjacency matrix between multiple nodes of road information, and obtain the current planned path;

[0012] Extracting a traffic-road topology feature matrix and a planned path feature vector from the flow data of each node in each time period, an adjacency matrix between multiple nodes, and a current planned path;

[0013] Constructing a scheduling path classification feature vector between the traffic-road topology feature matrix and the planning path feature vector, and performing expected activation of the parameter space on the scheduling path classification feature vector to obtain an optimized scheduling path classification feature vector;

[0014] The optimized scheduling path classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the current planned path is reasonable.

[0015] Compared with the existing technology, the present application provides a cloud platform-based integrated command and dispatch system and method, which utilizes real-time traffic data and road information, combines the adjacency relationship between nodes and the planned path, extracts traffic-road topology characteristics and planned path characteristics, and obtains the dispatch path classification characteristics by fusing these characteristics. Finally, the rationality of the current planned path is judged by the classifier to optimize the operational efficiency of urban rail transit. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 It is a block diagram of a cloud platform-based integrated command and dispatch system according to an embodiment of the present application.

[0018] Figure 2 It is a block diagram of a dispatching data processing module in a cloud platform-based integrated command and dispatching system according to an embodiment of the present application.

[0019] Figure 3 It is a block diagram of a traffic data extraction unit in a cloud platform-based integrated command and dispatch system according to an embodiment of the present application.

[0020] Figure 4 It is a block diagram of a planning path extraction unit in a cloud platform-based integrated command and dispatch system according to an embodiment of the present application.

[0021] Figure 5 It is a flow chart of a cloud platform-based integrated command and dispatch method according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0023] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0024] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0025] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0026] Figure 1 The figure shows a block diagram of a cloud platform-based integrated command and dispatch system according to an embodiment of the present application. Figure 1As shown, according to the embodiment of the present application, the cloud platform-based fusion command and dispatch system 100 includes: a dispatch data acquisition module 110, which is used to obtain the flow data of each node in each time period, obtain the adjacency matrix between multiple nodes of road information, and obtain the current planned path; a dispatch data processing module 120, which is used to extract the traffic-road topology feature matrix and the planned path feature vector from the flow data of each node in each time period, the adjacency matrix between multiple nodes, and the current planned path; a dispatch data fusion module 130, which is used to construct a dispatch path classification feature vector between the traffic-road topology feature matrix and the planned path feature vector, and activate the expected activation of the parameter space of the dispatch path classification feature vector to obtain an optimized dispatch path classification feature vector; a dispatch data analysis module 140, which is used to pass the optimized dispatch path classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the current planned path is reasonable.

[0027] In the embodiment of the present application, the scheduling data acquisition module 110 is used to obtain the flow data of each node in each time period, the adjacency matrix between multiple nodes of the road information, and the current planned path. It should be understood that by obtaining the flow data of each node in each time period and the adjacency matrix of the road information, the system can monitor the traffic conditions in real time, including vehicle density, vehicle speed, road congestion, etc., to help the system understand the current traffic conditions. Based on the data monitored in real time, the system can provide accurate real-time traffic information, including traffic flow, congestion, road condition changes, etc., so that drivers and traffic management departments can understand the road conditions in time and make corresponding adjustments and decisions. By obtaining the current planned path and the traffic-road topology feature matrix, the system can analyze the characteristics of the existing path, identify bottlenecks and congestion points, thereby optimizing the planned path, providing more efficient route selection, and reducing driving time and fuel consumption. After integrating the traffic-road topology feature matrix and the planned path feature vector, the system can generate a scheduling path classification feature vector, evaluate the planned path through the classifier of the data analysis module, help the system to make scheduling and command decisions, improve traffic efficiency and reduce congestion. Obtaining this data is crucial for the integrated command and dispatch system based on the cloud platform, which can improve the real-time, accuracy and intelligence level of the system, promote the intelligent development of urban rail transit systems, and meet the requirements of urban public transportation for safety, efficiency and reliability.

[0028] In particular, the level of cloud computing is constantly improving and has been well applied in many fields. It has the following advantages: (1) The data itself has high security performance; (2) The overall quality of the service is high and will not be greatly affected by network problems such as network instability; (3) Improve resource utilization and effectively utilize existing resources; (4) Do not affect the overall process of information technology management and ensure that the process can operate normally. Urban rail transit, as a public transportation in the city, has high requirements for security. As a new technology, cloud technology has been applied in many fields. The integrated command and dispatch system is based on the cloud platform and adopts a hierarchical data processing architecture. While meeting real-time requirements, it can effectively analyze the potential value of data and effectively improve resource utilization. With the continuous development of informatization, the integration of cloud technology and rail transit will become a mainstream. Cloud technology will play an increasingly important role in the integrated command and dispatch system and promote the development of urban construction.

[0029] In the embodiment of the present application, the scheduling data processing module 120 is used to extract the traffic-road topology feature matrix and the planning path feature vector from the flow data of each node in each time period, the adjacency matrix between multiple nodes and the current planning path. It should be understood that the extraction of the traffic-road topology feature matrix and the planning path feature vector is to extract more valuable features from the original data so that the system can better understand the structure of the traffic network and the characteristics of the planning path. By analyzing the traffic-road topology feature matrix and the planning path feature vector, the system can find the potential patterns and laws in the traffic network, thereby realizing the prediction and planning of future traffic conditions. The extracted features can help the system better make decision support, such as optimizing route planning, dispatching vehicles, adjusting signal light control, etc., to improve traffic efficiency and reduce congestion. By analyzing the traffic-road topology feature matrix and the planning path feature vector, the system can find potential bottlenecks and problems, thereby guiding the system to optimize, improve traffic management strategies, and improve overall system performance.

[0030] Specifically, in one embodiment of the present application, Figure 2 The figure shows a block diagram of a dispatching data processing module in a fusion command and dispatch system based on a cloud platform according to an embodiment of the present application. Figure 2As shown, in the above-mentioned cloud platform-based fusion command and dispatch system 100, the dispatch data processing module 120 includes: a flow data extraction unit 121, which is used to encode the flow data of each node in each time period to obtain a flow data feature matrix; an adjacency matrix extraction unit 122, which is used to extract an adjacency feature matrix from the adjacency matrix between multiple nodes of the road information; a traffic road fusion unit 123, which is used to fuse the flow data feature matrix and the adjacency feature matrix to obtain the traffic-road topology feature matrix; a planned path extraction unit 124, which is used to convert the current planned path into a matrix and then perform convolution encoding to obtain a planned path feature vector.

[0031] Specifically, the traffic data extraction unit 121 is used to encode the traffic data of each node in each time period to obtain a traffic data feature matrix. It should be understood that the original traffic data may be high-dimensional and contain a large amount of redundant information. Through a suitable encoding method, these high-dimensional data can be reduced in dimension and key features can be extracted to obtain a more compact and representative traffic data feature matrix. The traffic data of different nodes in different time periods may have different data formats, units or magnitudes. Encoding can standardize these data so that they have a unified representation form, which is convenient for subsequent data processing and analysis. By converting traffic data into a feature matrix, various machine learning algorithms and models can be more conveniently applied for data analysis, modeling and prediction. The feature matrix can be used as input data to help the model better understand the correlation between data. Processing the encoded traffic data feature matrix is ​​usually more efficient than processing the original data. In large-scale data sets and complex systems, the use of feature matrices can improve computing efficiency and speed up data processing and analysis. The encoded feature matrix may not contain sensitive information in the original data, which helps to protect data privacy and security. In the process of data sharing and processing, the use of feature matrices can reduce potential privacy leakage risks.

[0032] further, Figure 3 The figure shows a block diagram of a flow data extraction unit in a fusion command and dispatch system based on a cloud platform according to an embodiment of the present application. Figure 3 As shown, in the dispatching data processing module 120 of the above-mentioned cloud platform-based integrated command and dispatching system 100, the traffic data extraction unit 121 includes: a context encoding subunit 1211, which is used to pass the traffic data of each node in each time period through a context-based encoder including an embedding layer to obtain a plurality of traffic data feature vectors, and arrange the plurality of traffic data feature vectors to obtain the traffic data feature vectors of each node; a two-dimensional arrangement subunit 1212, which is used to arrange the traffic data feature vectors of each node in two dimensions to obtain a traffic data feature matrix.

[0033] Specifically, the context encoding subunit 1211 is used to pass the flow data of each node in each time period through a context-based encoder including an embedding layer to obtain multiple flow data feature vectors, and arrange the multiple flow data feature vectors to obtain the flow data feature vectors of each node. It should be understood that by considering the context information between nodes and time periods in the encoding process, the correlation between nodes and the sequence of time can be better captured. This helps to retain the mutual influence between nodes and the continuity of time in the feature vector and improve the characterization ability of data. The context-based encoder can convert the original flow data into a feature vector with more characterization ability. These feature vectors can better reflect the flow situation of the node in different time periods, including information such as change trends and periodicity, which is conducive to subsequent data analysis and modeling. After encoding the flow data in each time period into a feature vector, the dimension of the data can be effectively reduced, redundant information can be reduced, and the calculation efficiency can be improved. In addition, arranging multiple feature vectors into a feature vector of a node helps to integrate information in different time periods and better represent the overall situation of the node. Converting flow data into the form of feature vectors is more suitable for the application of many machine learning algorithms. The arrangement of feature vectors can better meet the requirements of machine learning models for input data and improve the performance and generalization ability of the model.

[0034] Correspondingly, the context encoding subunit includes: an embedded encoding secondary subunit, which is used to use the embedding layer of the encoder to convert the traffic data of each node in each time period into a traffic data input vector to obtain a sequence of traffic data input vectors; a converter secondary subunit, which is used to pass the sequence of traffic data input vectors through the converter of the encoder to obtain the multiple traffic data feature vectors; and a one-dimensional arrangement secondary subunit, which is used to arrange the multiple traffic data feature vectors in one dimension to obtain the traffic data feature vectors of each node.

[0035] Further, the embedded coding secondary subunit is used to use the embedding layer of the encoder to convert the traffic data of each node in each time period into a traffic data input vector to obtain a sequence of traffic data input vectors. It should be understood that converting the traffic data of different nodes in different time periods into a vector sequence can make the data have a uniform dimension before inputting the encoder. This helps to ensure that the model can effectively process and learn traffic data of different nodes and time periods. Through the embedding layer, the original traffic data can be converted into a more semantically informative representation. This helps the model better understand the features and patterns in the data. Converting the data into the form of a vector sequence can retain the time series information between the data. This is very important for capturing time correlations and sequence patterns, especially when processing time series data. The embedding layer helps to capture contextual information between nodes and time periods, thereby improving the model's understanding and expression of the relationship between data.

[0036] Furthermore, the converter secondary subunit is used to pass the sequence of the flow data input vector through the converter of the encoder to obtain the multiple flow data feature vectors. It should be understood that the role of the converter is to process the input vector sequence and extract key features from it. These features can help identify patterns, trends and important information in the data, and provide more meaningful data representation for subsequent analysis and decision-making. Through the converter, the original high-dimensional data can be converted into a low-dimensional feature vector with more abstraction and expressiveness. Doing so helps to reduce the complexity of the data while retaining important information, providing a more efficient data representation for subsequent processing and analysis. The feature vectors processed by the converter can be used as input to a neural network or other machine learning model. These feature vectors are more representative and distinguishable, which helps to improve the accuracy and generalization ability of the model. The converter can compress and optimize the original data to reduce the demand for calculation and storage. This helps to improve the efficiency and speed of data processing, especially in the case of large-scale data and complex models.

[0037] Specifically, the one-dimensional arrangement secondary subunit is used to arrange the multiple traffic data feature vectors in one dimension to obtain the traffic data feature vectors of each node. It should be understood that by arranging multiple feature vectors in one dimension, it can be ensured that the node traffic data feature vectors finally obtained have the same dimension. This helps to maintain the consistency and comparability of the data. Arranging multiple feature vectors into a one-dimensional vector can integrate the feature information in each time period. This helps to comprehensively consider the data features in different time periods, so as to more comprehensively describe the traffic situation of the node. One-dimensional arrangement can simplify data representation, making data easier to process and analyze. This form of representation is usually more suitable for the input requirements of many machine learning models. Through one-dimensional arrangement, features in different time periods can be arranged together in an orderly manner, thereby retaining the time series information while also retaining the correlation between features.

[0038] Specifically, the two-dimensional arrangement subunit 1212 is used to arrange the flow data feature vectors of each node in two dimensions to obtain a flow data feature matrix. It should be understood that in many cases, there are spatial associations and structures between each node, such as in graphic data or spatial data. By arranging the node feature vectors in two dimensions, the spatial structure information between the nodes can be better retained, which helps the model to better understand and utilize this association. Arranging the node feature vectors in a matrix form can conveniently perform matrix operations, such as matrix multiplication, matrix transposition, etc. This form helps to more efficiently process node feature data in neural networks or other models and accelerate the calculation process. Many deep learning models and machine learning algorithms are more suitable for processing matrix data rather than vector data. Therefore, organizing the node feature vectors into a matrix form can better adapt to the needs of different models and improve the performance and applicability of the model. By arranging the node feature vectors into a matrix, the feature information of each node can be better considered, thereby obtaining a more comprehensive and comprehensive feature representation. This helps to improve the model's understanding and modeling capabilities of the overall data.

[0039] Accordingly, in a specific example of the present application, the adjacency matrix extraction unit 122 is used to extract an adjacency feature matrix from the adjacency matrix between multiple nodes of the road information. It should be understood that the adjacency matrix is ​​an important tool for representing the connection relationship between nodes. Extracting the adjacency feature matrix from the adjacency matrix can help encode the relationship between nodes into a feature form, which is necessary for many graph neural network models. In a graph neural network, the characteristics of a node are usually affected by the characteristics of its neighboring nodes. Extracting the adjacency feature matrix can help consider the relationship between nodes during information transmission, thereby better capturing the dependencies between nodes. The adjacency matrix contains the topological structure information of the graph, and extracting the adjacency feature matrix can help utilize this structural information to better understand the connection mode and correlation between nodes. In a graph neural network, the adjacency matrix is ​​one of the foundations for performing graph convolution operations. By extracting the adjacency feature matrix, the input for the graph convolution operation can be better provided, thereby achieving effective processing of graph structure data.

[0040] Specifically, the adjacency matrix extraction unit is used to: obtain an adjacency feature matrix by passing the adjacency matrix between multiple nodes of the road information through a first convolutional neural network model based on a filter, wherein the values ​​of each non-diagonal position in the adjacency matrix are the distances between adjacent road intersections, and the values ​​of each diagonal position in the adjacency matrix are zero.

[0041] Furthermore, the adjacency matrix extraction unit includes: an output feature map subunit, which is used for each layer of the first convolutional neural network except the last layer to perform convolution processing, pooling processing and activation processing on the input data during the forward transmission process of the layer so as to output the adjacency feature map from the second last layer of the first convolutional neural network; and a generation matrix subunit, which is used for the last layer of the first convolutional neural network to perform convolution processing, global mean pooling processing along the channel dimension and activation processing on the adjacency feature map to generate the adjacency feature matrix.

[0042] Accordingly, in a specific example of the present application, the traffic road fusion unit 123 is used to fuse the flow data feature matrix and the adjacency feature matrix to obtain the traffic-road topology feature matrix. It should be understood that the flow data feature matrix provides information about traffic flow and can reflect the vehicle flow situation at different time points and regions. The adjacency feature matrix provides the topological structure information of the road network, describing the connection relationship and spatial position between roads. The fusion of these two types of information can comprehensively consider the association between traffic flow and road topology, so that the model can better understand the operating status of the traffic system. By fusing flow data and road topology features, richer input features can be provided for the model, which helps to improve the accuracy and robustness of traffic state prediction or other related tasks. Combining two different types of information can make up for the shortcomings of each information, so that the model can more comprehensively capture the dynamic changes of the traffic system. Fusion of flow data and adjacency features can provide richer feature representations, which helps to learn more complex and abstract traffic-road topology features. This comprehensive feature can better reflect the complexity and diversity of the traffic system, thereby improving the performance of the model in traffic analysis and prediction tasks.

[0043] Specifically, the flow data feature matrix and the adjacency feature matrix are fused through the graph neural network to generate a traffic-road topology feature matrix containing flow data and irregular topological structure information. This process uses the learnable neural network parameters of the graph neural network to better capture the complex relationships and topological features in traffic data, thereby improving the effect of traffic data processing and analysis. Through this fusion method, the data in the traffic system can be more fully understood and more accurate feature representation can be provided for further traffic analysis and prediction.

[0044] Furthermore, the traffic road fusion unit is used to: pass the flow data feature matrix and the adjacency feature matrix through a graph neural network, and the graph neural network is used to generate a traffic-road topology feature matrix containing flow data and irregular topological structure information through learnable neural network parameters.

[0045] Accordingly, in a specific example of the present application, the planned path extraction unit 124 is used to convert the current planned path into a matrix and then perform convolution coding to obtain a planned path feature vector. It should be understood that converting the planned path into a matrix form can better retain the structural information of the path, such as the order and relationship in the path. Through convolution coding, features can be effectively extracted from this structured data to capture local patterns and global associations in the path. Convolution coding has been widely used in image and sequence data, and local patterns in data can be effectively identified through convolution operations. For the planned path, this method can help extract key features in the path, thereby better representing the feature vector of the path. After the planned path is converted into a feature vector, the dimension of the original data can be reduced to a more compact representation. This helps to reduce the complexity of the data and extract the most important information in the planned path, so that subsequent analysis and processing can be more convenient.

[0046] further, Figure 4 The figure shows a block diagram of a planning path extraction unit in a fusion command and dispatch system based on a cloud platform according to an embodiment of the present application. Figure 4 As shown, in the dispatching data processing module 120 of the above-mentioned cloud platform-based fusion command and dispatching system 100, the planned path extraction unit 124 includes: a conversion matrix subunit 1241, which is used to convert the current planned path into a two-dimensional planned path numerical matrix, wherein the current planned path includes multiple target nodes and the distances between each target node; a convolutional coding subunit 1242, which is used to pass the planned path numerical matrix through a second convolutional neural network model to obtain a planned path feature map corresponding to the planned path numerical matrix; a dimensionality reduction subunit 1243, which is used to perform global mean pooling on each feature matrix along the channel dimension of the planned path feature map to obtain the planned path feature vector.

[0047] Specifically, the conversion matrix subunit 1241 is used to convert the current planned path into a two-dimensional planned path numerical matrix, wherein the current planned path includes multiple target nodes and the distances between each target node. It should be understood that representing the planned path as a two-dimensional matrix can better retain the structural information in the path. Through the matrix form, the distance relationship between each target node can be clearly displayed, so as to better understand the spatial structure and order of the path. Matrix is ​​a common data structure, which is convenient for processing and operation in a computer. Representing the planned path in matrix form can facilitate subsequent mathematical operations and algorithm processing, such as convolution operations, dimensionality reduction processing, etc. By converting the planned path into a matrix, the key features in the path can be better extracted. This helps in subsequent feature extraction and analysis work, so as to better understand the characteristics and patterns of the planned path. Deep learning methods usually need to represent data as tensors or matrices for processing. Representing the planned path as a two-dimensional matrix can be better combined with a deep learning model, and deep learning technology is used to learn and extract path features.

[0048] Specifically, the convolutional encoding subunit 1242 is used to pass the planned path numerical matrix through the second convolutional neural network model to obtain a planned path feature map corresponding to the planned path numerical matrix. It should be understood that the convolutional neural network can effectively extract features from the input data. Convolution operation on the planned path numerical matrix can identify features at different positions, which helps to capture key information in the planned path. The convolutional neural network has translation invariance, which means that no matter how the features are located in the image, the convolutional neural network can identify these features. In the planned path, the position of the target node may change, and the convolutional neural network can better handle such position changes. Convolutional neural network is a hierarchical feature learning method that can learn abstract feature representations at different levels. Through multi-layer convolution operations, the convolutional neural network can gradually extract more abstract and advanced planned path features, which is conducive to the system to better understand the path information. The parameter sharing mechanism in the convolutional neural network can reduce the number of parameters of the model and improve the generalization ability of the model. This is particularly important when the amount of planned path data is large, and can effectively reduce the risk of overfitting of the model.

[0049] Specifically, the dimension reduction subunit 1243 is used to perform global mean pooling on each feature matrix along the channel dimension of the planning path feature map to obtain the planning path feature vector. It should be understood that global mean pooling can reduce the spatial dimension of the feature map, thereby reducing the number of parameters, reducing the computational complexity, and reducing the risk of overfitting. Through global mean pooling, the average value of each feature map can be retained, which helps to retain important feature information and reduce information loss. The global mean pooling operation is performed on the entire feature map, so it will not be affected by the specific position in the feature map, has a certain position invariance, and helps to improve the generalization ability of the model. Global mean pooling converts each feature map into a feature vector of fixed length, which is convenient for subsequent processing and analysis, such as classification, clustering and other tasks.

[0050] In an embodiment of the present application, the scheduling data fusion module 130 is used to construct a scheduling path classification feature vector between the traffic-road topology feature matrix and the planning path feature vector, and to perform expected activation of the parameter space of the scheduling path classification feature vector to obtain an optimized scheduling path classification feature vector. It should be understood that the traffic road topology feature matrix provides information about the road network structure, while the planning path feature vector contains the characteristics of the specific path. Fusion of these two types of information can comprehensively consider the road network structure and the specific path characteristics, making the feature vector more representative. Fusion of different types of information can generally improve the classification performance of the model. By combining traffic road topology features and planning path features, the feature vector can be made more discriminative, which helps the classification model to better distinguish different scheduling path categories. Fusion of different types of features can enrich feature expression capabilities. Traffic road topology features and planning path features provide useful information in different aspects. Combining them can make the feature vector more expressive and help improve the accuracy of classification tasks.

[0051] In particular, in the technical solution of the present application, it is considered that when obtaining the flow data of each node, the data may be missing or incomplete due to sensor failure, data transmission problems or other external factors. This incompleteness will introduce outliers in the flow data feature vector, thereby affecting the entire scheduling path classification feature vector. When using context-based encoders and convolutional neural network models to process data, if overfitting occurs during model training, the model may be overly sensitive to the features of the training data, while ignoring some important but uncommon features in the training data. This will produce local abnormal distributions in the scheduling path classification feature vector. Local abnormal distributions may make it difficult for the classifier to correctly identify and classify scheduling paths, thereby reducing the accuracy of the classification results. This may cause unreasonable paths to be mistakenly judged as reasonable, or reasonable paths to be mistakenly judged as unreasonable. In actual vehicle scheduling and path planning, if the classification results are inaccurate, it may lead to decision-making errors. For example, it may cause vehicles to be directed to congested or closed roads, thereby affecting transportation efficiency and safety. If unreasonable roads are mistakenly classified as reasonable, it may lead to a waste of resources (such as vehicles, fuel, time, etc.) because these resources may be used for inefficient or invalid paths. Therefore, in order to solve this problem, in the technical solution of the present application, the scheduling path classification feature vector is activated in the parameter space.

[0052] Accordingly, in one embodiment of the present application, the scheduling data fusion module 130 includes: a fusion unit, used to fuse the traffic-road topology feature matrix and the planned path feature vector to obtain the scheduling path classification feature vector; an optimization unit, used to perform expected activation of the parameter space of the optimized scheduling path classification feature vector to obtain the optimized scheduling path classification feature vector.

[0053] Further, the optimization unit includes: an adjustment exponent eigenvalue subunit, which is used to calculate the natural exponential function value of the eigenvalue at the predetermined position of the scheduling path classification feature vector as a power and then subtract one to obtain the adjustment exponent eigenvalue; an adjustment exponent eigenvalue ratio subunit, which is used to calculate the adjustment exponent eigenvalue divided by the eigenvalue at the predetermined position of the scheduling path classification feature vector, and take the absolute value of the calculation result of the adjustment exponent eigenvalue to obtain the adjustment exponent eigenvalue ratio; an activation adjustment subunit, which is used to activate the adjustment exponent eigenvalue ratio through a sigmoid function to obtain an activation adjustment exponent eigenvalue ratio; an exponential attenuation coefficient subunit, which is used to calculate the natural exponential of the opposite number of the eigenvalue at the predetermined position of the scheduling path classification feature vector as a power. function value to obtain the exponential decay coefficient of the eigenvalue; an adjustment decay difference subunit, used to calculate the eigenvalue of the predetermined position of the scheduling path classification feature vector divided by the exponential decay coefficient of the eigenvalue and then subtract one, and take the absolute value of the calculation result of the exponential decay coefficient to obtain the adjusted decay difference of the eigenvalue; an activation feature subunit, used to activate the adjusted decay difference of the eigenvalue through a sigmoid function to obtain the activated eigenvalue adjusted decay difference; an optimization feature subunit, used to calculate the weighted sum of the activation adjustment exponential eigenvalue ratio and the activated eigenvalue adjusted decay difference, and multiply the weighted calculation result by the eigenvalue of the predetermined position of the scheduling path classification feature vector to obtain the eigenvalue of the predetermined position of the optimized scheduling path classification feature vector.

[0054] Here, in view of the local abnormal distribution in the scheduling path classification feature vector, in the technical solution of the present application, the scheduling path classification feature vector is activated in the parameter space of the expected activation, which activates the feature vector elements that have an important impact on the classification decision in an expected manner by dynamically adjusting the parameter space. When implementing the expected activation, not only the importance of a single feature vector element is considered, but also the interaction and synergy between the feature vector elements are considered. Specifically, by performing the expected activation of the parameter space of the scheduling path classification feature vector, it is possible to automatically identify and enhance those feature vector elements that contribute more to the model performance, and dynamically adjust the parameter space according to the current model's learning state and the activation of the feature vector elements to maintain the activity and diversity of the feature representation. In this way, the problem of reduced classification performance due to local abnormal distribution can be effectively alleviated, and the sensitivity and adaptability of the model to key features in the data can also be improved, so that a high classification accuracy and robustness can be maintained in various complex data environments.

[0055] In an embodiment of the present application, the scheduling data analysis module 140 is used to pass the optimized scheduling path classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the current planned path is reasonable. It should be understood that the classifier can learn the law and make classification decisions based on the input feature vector. By training a classifier, it can learn the characteristic pattern of a reasonable planned path, so that the rationality of the new planned path can be judged. The classifier can be used to realize the automatic judgment of the planned path, avoiding the tedious process of manual verification one by one. This can improve efficiency and make decisions quickly in real-time scenarios. The classification result provides a quantitative evaluation method, mapping the planned path to different categories, thereby intuitively indicating the rationality of the planned path. This quantitative evaluation helps to better understand the characteristics of the planned path. If the classification result shows that the planned path is unreasonable, the decision basis of the classifier can be further analyzed to find out the reason why the planned path is unreasonable, and make targeted adjustments and optimizations to improve the quality of the planned path. Therefore, the classification feature vector of the optimized scheduling path is evaluated through a classifier and the classification result is used to represent the rationality of the current planned path, thereby improving the intelligence level, decision-making efficiency and system operation quality of the system, which is helpful to optimize the scheduling strategy of the urban rail transit system and improve the efficiency and safety of traffic operation.

[0056] Accordingly, in one embodiment of the present application, the scheduling data analysis module 140 is used to: use the classifier to process the optimized scheduling path classification feature vector using the following classification formula to obtain the classification result;

[0057] Wherein, the classification formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is a bias vector, X is a classification feature vector of the optimized scheduling path, softmax represents a softmax function, and O represents the classification result.

[0058] In summary, the cloud-based integrated command and dispatch system and method based on the embodiment of the present application utilizes real-time traffic data and road information, combines the adjacency relationship between nodes and the planned path, extracts traffic-road topology characteristics and planned path characteristics, and obtains the dispatch path classification characteristics by integrating these characteristics. Finally, the rationality of the current planned path is judged by the classifier to optimize the operational efficiency of urban rail transit.

[0059] As described above, the cloud-based integrated command and dispatch system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server of the cloud-based integrated command and dispatch system. In one example, the cloud-based integrated command and dispatch system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the cloud-based integrated command and dispatch system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the cloud-based integrated command and dispatch system 100 can also be one of the many hardware modules of the terminal device.

[0060] Alternatively, in another example, the cloud platform-based integrated command and dispatch system 100 and the terminal device may also be separate devices, and the cloud platform-based integrated command and dispatch system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0061] Figure 5 FIG. 1 is a flow chart of a cloud platform-based integrated command and dispatch method according to an embodiment of the present application. Figure 5 As shown, according to the cloud platform-based fusion command and dispatch method of the embodiment of the present application, the steps include: S110, obtaining the flow data of each node in each time period, obtaining the adjacency matrix between multiple nodes of road information, and obtaining the current planned path; S120, extracting the traffic-road topology feature matrix and the planned path feature vector from the flow data of each node in each time period, the adjacency matrix between multiple nodes, and the current planned path; S130, constructing a scheduling path classification feature vector between the traffic-road topology feature matrix and the planned path feature vector, and performing expected activation of the parameter space of the scheduling path classification feature vector to obtain an optimized scheduling path classification feature vector; S140, passing the optimized scheduling path classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the current planned path is reasonable.

[0062] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned cloud platform-based integrated command and dispatch method have been referred to above. Figures 1 to 4 The description of the cloud platform-based integrated command and dispatch system has been introduced in detail, and therefore, its repeated description will be omitted.

[0063] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0064] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0065] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0066] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0067] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0068] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0069] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A cloud platform-based integrated command and dispatch system, characterized in that: include: The scheduling data acquisition module is used to obtain the traffic data of each node in each time period, the adjacency matrix between multiple nodes of the road information, and the current planned path; A scheduling data processing module, used to extract a traffic-road topology feature matrix and a planned path feature vector from the flow data of each node in each time period, the adjacency matrix between multiple nodes, and the current planned path; A scheduling data fusion module, used to construct a scheduling path classification feature vector between the traffic-road topology feature matrix and the planning path feature vector, and perform expected activation of the parameter space on the scheduling path classification feature vector to obtain an optimized scheduling path classification feature vector; A scheduling data analysis module, used for passing the optimized scheduling path classification feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the current planned path is reasonable; Wherein, the scheduling data processing module includes: A flow data extraction unit, used for encoding the flow data of each node in each time period to obtain a flow data feature matrix; An adjacency matrix extraction unit, used to extract an adjacency feature matrix from the adjacency matrix between the plurality of nodes of the road information; A traffic-road fusion unit, used for fusing the flow data feature matrix and the adjacency feature matrix to obtain the traffic-road topology feature matrix; A planned path extraction unit, used for converting the current planned path into a matrix and then performing convolution coding to obtain a planned path feature vector; Wherein, the scheduling data fusion module includes: A fusion unit, used for fusing the traffic-road topology feature matrix and the planned path feature vector to obtain the scheduling path classification feature vector; An optimization unit, configured to perform expected activation of the parameter space of the optimized scheduling path classification feature vector to obtain the optimized scheduling path classification feature vector; Wherein, the optimization unit comprises: An adjustment index eigenvalue subunit, used to calculate a natural exponential function value raised to a power of the eigenvalue of a predetermined position of the scheduling path classification eigenvector and then subtract one to obtain an adjustment index eigenvalue; an adjustment index eigenvalue ratio subunit, configured to calculate the adjustment index eigenvalue divided by the eigenvalue of a predetermined position of the scheduling path classification eigenvector, and to take an absolute value of the calculation result of the adjustment index eigenvalue to obtain an adjustment index eigenvalue ratio; an activation adjustment subunit, configured to activate the adjustment index eigenvalue ratio through a sigmoid function to obtain an activation adjustment index eigenvalue ratio; An exponential decay coefficient subunit, used for calculating a natural exponential function value of a power of the eigenvalue at a predetermined position of the scheduling path classification eigenvector to obtain an exponential decay coefficient of the eigenvalue; an adjusted attenuation difference subunit, configured to calculate the characteristic value of a predetermined position of the scheduling path classification characteristic vector divided by an exponential attenuation coefficient of the characteristic value minus one, and to take an absolute value of a calculation result of the exponential attenuation coefficient to obtain an adjusted attenuation difference of the characteristic value; An activation feature subunit, used for activating the adjusted attenuation difference of the feature value through a sigmoid function to obtain an activated feature value adjusted attenuation difference; The optimization feature subunit is used to calculate the weighted sum of the activation adjustment index eigenvalue ratio and the activation eigenvalue adjustment attenuation difference, and multiply the weighted calculation result by the eigenvalue of the predetermined position of the scheduling path classification feature vector to obtain the eigenvalue of the predetermined position of the optimized scheduling path classification feature vector.

2. The cloud platform-based integrated command and dispatch system according to claim 1 is characterized in that: The flow data extraction unit comprises: A context encoding subunit, configured to pass the flow data of each node in each time period through a context-based encoder including an embedding layer to obtain a plurality of flow data feature vectors, and arrange the plurality of flow data feature vectors to obtain a flow data feature vector of each node; The two-dimensional arrangement subunit is used to perform two-dimensional arrangement on the flow data feature vectors of each node to obtain a flow data feature matrix.

3. The cloud platform-based integrated command and dispatch system according to claim 2 is characterized in that: The adjacency matrix extraction unit is used to obtain an adjacency feature matrix by applying the adjacency matrix between multiple nodes of the road information to a first convolutional neural network model based on a filter, wherein the value of each non-diagonal position in the adjacency matrix is ​​the distance between adjacent road intersections, and the value of each diagonal position in the adjacency matrix is ​​zero.

4. The cloud platform-based integrated command and dispatch system according to claim 3 is characterized in that: The adjacency matrix extraction unit comprises: An output feature map subunit, used for each layer except the last layer of the first convolutional neural network to perform convolution processing, pooling processing and activation processing on the input data during the forward transmission process of the layer to output an adjacent feature map from the second last layer of the first convolutional neural network; A matrix generation subunit is used for the last layer of the first convolutional neural network to perform convolution processing, global mean pooling processing along the channel dimension and activation processing on the adjacent feature map to generate the adjacent feature matrix.

5. The cloud platform-based integrated command and dispatch system according to claim 4 is characterized in that: The traffic road fusion unit is used to: pass the flow data feature matrix and the adjacency feature matrix through a graph neural network, and the graph neural network is used to generate a traffic-road topology feature matrix containing flow data and irregular topological structure information through learnable neural network parameters.

6. The cloud platform-based integrated command and dispatch system according to claim 5 is characterized in that: The planned path extraction unit comprises: A conversion matrix subunit, used to convert the current planned path into a two-dimensional planned path numerical matrix, wherein the current planned path includes a plurality of target nodes and distances between the target nodes; A convolutional encoding subunit, used for passing the planned path numerical matrix through a second convolutional neural network model to obtain a planned path feature graph corresponding to the planned path numerical matrix; The dimension reduction subunit is used to perform global mean pooling on each feature matrix along the channel dimension of the planned path feature map to obtain the planned path feature vector.

7. A cloud platform-based integrated command and dispatch method, using the cloud platform-based integrated command and dispatch system according to claim 1, characterized in that: include: Obtain the traffic data of each node in each time period, obtain the adjacency matrix between multiple nodes of road information, and obtain the current planned path; Extracting a traffic-road topology feature matrix and a planned path feature vector from the flow data of each node in each time period, an adjacency matrix between multiple nodes, and a current planned path; Constructing a scheduling path classification feature vector between the traffic-road topology feature matrix and the planning path feature vector, and performing expected activation of the parameter space on the scheduling path classification feature vector to obtain an optimized scheduling path classification feature vector; The optimized scheduling path classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the current planned path is reasonable.

Images