A Network-Level Traffic Operation State Inference Method for Unobservable Road Sections
By using the public security checkpoint detection data and Bayesian network model, combined with the dual attention mechanism-directed graph-long short-term memory network, the problem of traffic flow characteristic parameters completion in unobserved sections is solved, and accurate inference of network-level traffic operation status and effective prediction of morning and evening rush hour are achieved.
Patent Information
- Application Number
- CN202411653767.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-19
- Publication Date
- 2025-08-01
- Estimated Expiration
- 2044-11-19
AI Technical Summary
Under actual conditions, traffic detection equipment cannot be installed on some sections of roads, resulting in traffic managers being unable to effectively grasp the traffic operation status. The existing deep learning algorithms lack special considerations for morning and evening rush hour, and the complexity of the traffic network and data loss problems are difficult to quickly and accurately integrate traffic data from different time and space dimensions, resulting in increased difficulty in inferring the network-level traffic operation status.
The public security checkpoint detection data is used for preprocessing, a model based on Bayesian network and basic traffic flow chart is built, a dual attention mechanism-directed graph long short-term memory network model is built, combined with spatial syntax theory, the influencing factors of traffic flow in upstream and downstream sections are captured, and traffic flow characteristic parameter completion and state inference are realized through directed graph convolution operators.
It improves the balance and accuracy of the training data, and can accurately infer the traffic operation status of the road network in a short period of time, especially to provide effective traffic management and control support during morning and evening rush hours.
Smart Images

Figure CN119516782B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent traffic control, and in particular to a method for inferring network-level traffic operation status for unobservable road sections. Background Art
[0002] With the progress of technology, the development of intelligent transportation systems provides an important means for traffic management departments to master the urban road traffic operation status and its changing trends. Compared with detection devices such as infrared rays, induction coils, and GPS, the bayonet detection device based on visual perception technology has the advantages of wide layout range, high data validity, and good stability, and has become the most commonly used data source for traffic management departments. Although the bayonet detection device has the widest application range, there are still a considerable number of roads that have not installed traffic detection devices. That is, for traffic managers, there are a large number of road sections in the urban road where the traffic operation status cannot be effectively grasped. In response to this problem, researchers have sought solutions from two perspectives: traffic flow theory modeling and artificial intelligence modeling. However, the method based on traffic flow theory modeling has poor universality, harsh application conditions, and low feasibility; the method based on artificial intelligence modeling requires a large amount of reliable data. Under real conditions, some road sections are completely unobservable, that is, the data is completely missing, and the traffic state estimation task cannot be effectively completed.
[0003] The rapid development of deep learning provides a new opportunity for traffic operation status inference. It has a high degree of self-organization and self-adaptability, information storage ability, and strong computing ability. It can adapt to the diversity and complexity of input data and express any linear and non-linear relationships. It can obtain a pattern from the training data and infer the results of the test data based on this pattern. However, the current invention considers less about the ability of the road network to attract traffic flow itself, and traffic managers are more concerned about the changing trends of traffic operation status during the morning and evening rush hours. The existing deep learning algorithms lack special consideration for the morning and evening rush hours. Especially in the data environment where some road sections are unobservable in reality, there is a lack of effective means to infer the network-level traffic operation status.
[0004] More importantly, the traffic network is a complex dynamic system. Its flow state is affected by both time series and spatial position changes at the same time, and there are direct or indirect correlations and influences between traffic flows in different regions. Therefore, it is difficult to quickly and accurately fuse traffic data in different spatio-temporal dimensions, which is not conducive to the extraction of key information therein, and further exacerbates the difficulty of inferring the network-level traffic operation status. Therefore, it is urgent for technical personnel to solve the above problems. Summary of the Invention
[0005] Therefore, in view of the above existing problems and defects, the designers of the present invention collected relevant materials, conducted multi-party evaluations and considerations, and through continuous experiments and modifications by technicians with many years of R & D experience in this industry, finally led to the emergence of the network-level traffic operation state inference method for unobservable road sections.
[0006] To solve the above technical problems, the present invention relates to a network-level traffic operation state inference method for unobservable road sections, which includes the following steps:
[0007] S1. A data preprocessing method based on public security checkpoint detection data is used to correct or delete abnormal data, and on this basis, traffic flow characteristic parameters such as flow, speed, and density are extracted;
[0008] S2. A traffic flow characteristic parameter completion model based on Bayesian network and fundamental diagram of traffic flow is constructed to complete the traffic flow characteristic parameters of unobservable road sections in the road network;
[0009] S3. A dual-attention mechanism - directed graph long short-term memory network model is built, and the attention mechanism of the GAT model is introduced to capture traffic flow influencing factors of upstream and downstream road sections;
[0010] S4. Based on space syntax theory, the topological values of each road section in the road network and the vehicle speed characteristics of each road section in the road network are calculated. Among them, the topological values of each road section in the road network are used as spatial dimension features; the topological values of each road section in the road network are combined with the vehicle speed characteristics to construct a directed graph convolution operator;
[0011] S5. The spatial dimension features and the directed graph convolution operator obtained in step S4 are reversely input into the dual-attention mechanism - directed graph long short-term memory network model to evaluate the effectiveness of the model.
[0012] As a further improvement of the technical solution disclosed by the present invention, step S1 includes the following sub-steps:
[0013] S11. The traffic detection data of two adjacent public security checkpoints are matched and compared, and invalid license plate data that cannot be correctly read and unreasonable license plate data that appear within a fixed observation time window are excluded;
[0014] S12. The signal cycle to which the current vehicle belongs is divided according to the time stamp of the public security checkpoint detection data, and 4s start-up loss is deducted from the time of the first vehicle passing the stop line in each cycle to obtain the start time of the green light in each cycle;
[0015] S13. Compare the green light on-time with the actual green light on-time to obtain an offset, perform normal distribution fitting on the offset distribution in the entire study area, and select the offset with the highest normal distribution probability as the calibration value of the drift correction algorithm to correct the offset error of the green light on-time;
[0016] S14. Select a fixed observation time window, count the traffic volume, average travel speed and average vehicle density of the observable road sections in the road network, and eliminate abnormal data by calculating the statistical distribution characteristics of the average travel speed of vehicles.
[0017] As a further improvement of the technical solution disclosed in the present invention, step S2 includes the following sub-steps:
[0018] S21. Based on the Greenshields model, the traffic flow basic graph function relationship is fitted by the least squares method;
[0019] S22. Construct a Bayesian network graph structure for unobservable road sections in a road network based on basic traffic flow theory.
[0020] S23. Through the functional relationship of the basic traffic flow graph constructed in S21, the relationship between the nodes in the Bayesian network model in S22 is determined to complete the traffic flow characteristic parameters of the unobservable sections in the road network.
[0021] As a further improvement of the technical solution disclosed in the present invention, step S3 includes the following sub-steps:
[0022] S31. Introducing the graph attention module to explore the influence of traffic flow parameters between adjacent road segments;
[0023] S32. Long short-term memory network is integrated with attention mechanism to optimize the weight of the function.
[0024] As a further improvement of the technical solution disclosed in the present invention, step S4 includes the following sub-steps:
[0025] S41. Calculate the topological value of each road segment in the road network based on space syntax theory and use it as a spatial dimension feature;
[0026] S42. By obtaining the average travel speed of the vehicle in each section, a speed variance matrix of each section is constructed as the vehicle running speed feature, and together with the spatial dimension feature constructed in S41, it constitutes the feature vector of the model input, and constructs a directed graph convolution operator.
[0027] As a further improvement of the technical solution disclosed in the present invention, step S5 includes the following sub-steps:
[0028] S51. Select three evaluation metrics, namely mean absolute error, root mean square error, and mean absolute percentage error, in the neural network algorithm as the evaluation metrics for model effectiveness.
[0029] S52. Input actual data and evaluate the effectiveness of introducing the dual attention mechanism in the model through model ablation analysis.
[0030] In practical applications, the network-level traffic operation state inference method for unobservable road segments disclosed in the present invention can achieve at least the following beneficial technical effects: specifically:
[0031] 1) Oriented towards solving practical problems, using the most widely used and easily obtainable public security checkpoint detection data as the data source, and complementing the traffic flow characteristic parameters of unobservable road segments through the Bayesian model and the traffic flow fundamental diagram model. Construct a dual attention mechanism - directed graph long short-term memory network model to capture the traffic flow influencing factors of upstream and downstream road segments while improving the balance of training data, and achieve accurate inference of the actual traffic operation state of the road network.
[0032] 2) Verified by actual data, the dual attention mechanism - directed graph long short-term memory network model meets expectations in short-term traffic flow prediction and achieves excellent results. It provides a theoretical support for predicting the traffic flow operation status of road segments and road networks in the short term in the future, and provides a basis for traffic management and control. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0034] Figure 1 It is the overall flowchart of the network-level traffic operation state inference method for unobservable road segments disclosed in the present invention.
[0035] Figure 2 It is the layout schematic diagram of a form road network disclosed in the present invention.
[0036] Figure 3 It is the architecture schematic diagram of the Bayesian network constructed in the network-level traffic operation state inference method for unobservable road segments disclosed in the present invention.
[0037] Figure 4 It is the architecture schematic diagram of the dual attention mechanism - directed graph long short-term memory network model constructed in the network-level traffic operation state inference method for unobservable road segments disclosed in the present invention.
[0038] Figure 5 Schematic diagram of the architecture of the attention mechanism introduced in the long short-term memory network for the network-level traffic operation state inference method for unobservable road sections disclosed by the present invention.
[0039] Figure 6 Schematic layout diagram of the example road network for verifying the effectiveness of the network-level traffic operation state inference method for unobservable road sections disclosed by the present invention.
[0040] Figure 7 Schematic diagram of the result after completing the traffic flow characteristic parameters for unobservable road sections in the network-level traffic operation state inference method for unobservable road sections disclosed by the present invention.
[0041] Figure 8 Comparison chart of MAE and RMSE results of the dual-attention mechanism - directed graph long short-term memory network model compared with the baseline model in the network-level traffic operation state inference method for unobservable road sections disclosed by the present invention.
[0042] Figure 9 Ablation analysis results of the dual-attention mechanism - directed graph long short-term memory network model in the network-level traffic operation state inference method for unobservable road sections disclosed by the present invention. Detailed implementation manners
[0043] To facilitate those skilled in the art to fully understand the technical solutions disclosed by the present invention, Figure 2 The schematic layout diagram of a form of road network disclosed by the present invention is shown. It can be clearly known that some road sections lack bayonet detection equipment or are affected by other difficult-to-overcome factors, which are extremely likely to lead to the problem that the traffic flow operation parameters cannot be obtained completely, and thus the traffic operation state of the entire road network after a period of time cannot be inferred. In view of this, the present invention proposes a network-level traffic operation state inference method for unobservable road sections. The present invention applies the public security bayonet detection data, which is the most extensive data source for traffic management departments, as the data source, and completes the data preprocessing process based on the statistical distribution characteristics of the bayonet detection data, deleting duplicate, missing, and abnormal data, and effectively extracting traffic flow characteristic parameters. By fitting the traffic flow fundamental diagram function with the Greenshields model, a Bayesian network model for under-observed road sections is constructed. The relationship between the nodes of the Bayesian network model is determined through the constructed traffic flow fundamental diagram function, and the solution is completed to obtain the traffic flow characteristic parameters of the under-observed road sections. The road sections are abstracted as nodes, a spatial similarity matrix and a traffic flow similarity matrix are defined, and a directed graph convolution operator is built to complete the construction of the dual-attention mechanism - directed graph long short-term memory network model, comprehensively considering the traffic flow characteristics and infrastructure characteristics, and realizing the inference of the traffic operation state of the road network considering under-observed road sections.
[0044] This invention is oriented towards solving practical problems. It uses the most widely used and easily accessible police checkpoint detection data as the data source. It uses the Bayesian model and the basic traffic flow graph model to supplement the traffic flow characteristic parameters of the under-observed road sections, and constructs a dual attention mechanism-directed graph long short-term memory network model to capture the traffic flow influencing factors of upstream and downstream sections while improving the balance of training data, thus realizing the inference of road network traffic operation status. Figure 1 As shown in , the specific steps of the network-level traffic operation status inference method for unobservable road sections are as follows:
[0045] S1. Through the data preprocessing method based on the police checkpoint detection data, the abnormal data is corrected or deleted, and the traffic flow characteristic parameters such as flow rate, speed and density are extracted on this basis;
[0046] S11, matching the vehicle license plate recognition data of the upstream and downstream adjacent intersections, and eliminating invalid license plate data that cannot be correctly read and unreasonable license plate data that appears within a fixed observation time window;
[0047] S12. Divide the signal cycle to which the current vehicle belongs according to the timestamp of the license plate recognition data, deduct 4 seconds of starting loss from the time of the first vehicle passing the stop line in each cycle, and obtain the green light start time of each cycle;
[0048] S13. Compare the green light on-time with the actual green light on-time to obtain an offset, perform normal distribution fitting on the distribution of the offsets in the entire study area, and select the offset with the highest normal distribution probability as the calibration value of the drift correction algorithm to correct the offset error of the green light on-time;
[0049] S14, select a fixed observation time window, through the formula T i =(t k+1 ,-t k ) Statistically calculate the average travel speed of observable sections within the road network, obtain the average vehicle density within the time window through the entry and exit method, and calculate the statistical distribution characteristics of the average travel speed of vehicles using the following formula: i When -10%<α<200%, the data is considered valid, otherwise the data stamp is deleted, thereby reducing abnormal data caused by on-street parking, recognition errors, etc.
[0050]
[0051] Among them, T i represents the travel time of the i-th vehicle through the road section; T k represents the time when the i-th vehicle passes through the k-th intersection; Indicates the average travel speed of vehicles in the time window;
[0052] S2. Construct a traffic state estimation model based on the Bayesian network and the fundamental diagram of traffic flow to estimate the traffic state of unobserved sections in the road network;
[0053] S21. Based on the Greenshields model Using the traffic flow characteristic parameters of flow, speed, and density of the observable sections obtained in step S14 as the data basis, the functional relationship of the fundamental diagram of traffic flow is fitted by the least squares method to obtain V of the observable sections f , K j ;
[0054] Where: V f represents the free flow speed of the section; K j represents the jam density;
[0055] S22. According to the principle of vehicle inflow and outflow traffic volume conservation of the section, construct a Bayesian network graph structure as shown in Figure 3 ;
[0056] S23. Through the functional relationship of the fundamental diagram of traffic flow constructed in S21, determine the relationship between the nodes in the Bayesian network model in S22. Based on traffic flow theory, the traffic flow characteristic parameters of the unobserved sections satisfy the following formula:
[0057] Q 欠-1 =Q 入-1 +Q 入-2 +Q 入-3
[0058] Q 欠-2 =Q 入-4 +Q 入-5 +Q 欠-1
[0059] ······
[0060] Q 出 =Q 欠-n +Q n-1 +Q n
[0061]
[0062] Where: Q 欠-1 represents the traffic volume within the time window of the first unobserved section; Q 入-1 , Q 入-2 , Q 入-3 represent the traffic volumes flowing into the time window of the first unobserved section respectively, and the other parameters can be obtained in the same way; k 欠-1 , k 欠-2 represent the traffic densities of the unobserved sections respectively, and the other parameters can be obtained in the same way. Estimate the traffic state of the unobserved sections;
[0063] S3. Build a model based on the dual attention mechanism - directed graph long short - term memory network model (ADALSTM);
[0064] S31. The model constructed in this step is as shown in Figure 4 . First, introduce a graph attention module into the long short - term memory network LSTM, abstract each road section in the road network as a directed graph, and apply the attention sharing mechanism in the weight matrix . Calculate the attention coefficient through the formula e ij = a(WX i , WX j ), where e ij represents the importance of section j to section i. Use the softmax function for normalization so that the coefficients between different nodes can be quantitatively represented . Perform an operation on the input feature V and the weight matrix introduced with the attention mechanism to obtain a new feature V′, and explore the influence of traffic flow parameters between adjacent road sections;
[0065] S32. Introduce the attention mechanism into the long - short - term memory network model again, as shown in Figure 5 . The formula is as follows. Among them, and are the section feature representations obtained through LSTM, tanh is the hyperbolic tangent activation function, is the parameter of the attention layer, nw is the number of time windows in a day, is to calculate the relative distance between time windows i and j and take the remainder. α i,j converts the importance score of section wj to section wi into a probability value between 0 and 1. The data sampling time period of a day is specified as 7:00 - 19:00, so that the morning and evening rush hours can receive more attention and the effectiveness of the model can be improved;
[0066]
[0067] S4. Based on the space syntax theory, calculate the topological value of each road section in the road network as the spatial dimension feature; calculate the vehicle speed feature of each road section in the road network, and combine the two to construct a directed graph convolution operator;
[0068] S41. Based on the space syntax theory, calculate the integration degree of each road section in the road network, and measure the influence of the spatial dimension features of the road section through the integration degree matrix. The integration degree can be calculated by the formula. Define the spatial similarity matrix I ∈ R N*N , and I is calculated using the following formula;
[0069]
[0070]
[0071] Wherein:
[0072] MD(xi): The average depth of element xi;
[0073] RA(xi): The relative asymmetry of element xi;
[0074] n: The total number of elements;
[0075] RRA(xi): The comparative asymmetry of element xi;
[0076] RA of Diamond: The relative asymmetry value of the diamond-shaped topological structure;
[0077] Integration(xi): The integration degree of element xi;
[0078] jm = inti - intj is the difference between road segments vi and vj in the m-th dimension; T is the length of the time series;
[0079] S42. Construct a speed variance matrix to measure the impact of traffic flow operation. The element vi,j in the matrix is calculated by the following formula:
[0080]
[0081] Wherein: gm = zi - zj is the difference between road segments vi and vj in the m-th dimension; T is the length of the time series.
[0082] S5. Input data to effectively infer the traffic operation state of the entire road network in the context of unobservable road segments in the road network, set reasonable evaluation indicators, and evaluate the effectiveness of the model;
[0083] S51. Select three evaluation indicators commonly used in neural network algorithms, namely mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), as the evaluation indicators for model effectiveness;
[0084] S52. Input actual data and evaluate the effectiveness of the model introducing the dual attention mechanism through model ablation analysis;
[0085] S53. Compare with four neural network algorithms, namely ANN, CNN, LSTM, and DGCLSTM, to verify the effectiveness of the model in inferring the network-level traffic operation state for unobservable road segments.
[0086] The technical content of the present invention will be explained below with a specific example.
[0087] Take the example where traffic flow detection devices are completely missing for two road segments in the road network, such as Figure 6As shown, the traffic operation status changes of section A and section B cannot be inferred. Taking 15 minutes as the time window, the traffic flow characteristic parameters of the under-observed sections are first completed by the method proposed in the present invention, and the results are as Figure 7 shown. The completed data reflects the time-varying characteristics of the traffic flow and has high data stability, and the results are credible.
[0088] The completed traffic flow characteristic parameters of the whole road section are input into the dual-attention mechanism-directed graph long short-term memory network model constructed in this study. Three evaluation indexes, namely Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE), are selected. The inference results of the model are compared with those of the Artificial Neural Network (ANN), Recurrent Neural Network (RNN), Long Short-Term Memory Network (LSTM), and Directed Graph Convolution-Long Short-Term Memory Neural Network (DGC-LSTM). Table 1 records the average prediction performance of ADALSTM and the baseline model at 15 min, 30 min, and 60 min. Figure 8 Describes the comparison of the MAE and RMSE results with the baseline model. As shown in the table, the short-term prediction performance of the ADALSTM model is better than that of other comparison models. Specifically, the CNN cannot reflect complex spatio-temporal relationships. The RNN extracts the time-dependent characteristics of the data on the basis of the CNN, so its effect is better than that of the CNN, but the model has deficiencies in extracting the spatial relationships in the traffic network. The DGC-LSTM model considers the spatio-temporal characteristics of the road network and refines the traffic characteristics of the road sections. Its prediction performance is significantly better than that of the LSTM at the three time steps, but the model does not consider the time characteristics of the road network traffic flow sufficiently. The ADALSTM adds a time attention module to improve the balance of the data, and its prediction performance is significantly better than that of the DGC-LSTM in the short term, proving the effectiveness of the model.
[0089] Table 1 Comparison of model performance
[0090]
[0091] Three days of data are selected from the original dataset for ablation analysis of the ADALSTM model to verify the effectiveness of the model's attention mechanism. Figure 9An ablation experiment of the ADALSTM model was conducted on two graph structures. Among them, the nonA-DALSTM model represents the original model without the road segment spatial attention mechanism, and the AD-nonALSTM model represents the original model without the temporal attention mechanism. It can be seen from the results that the ADALSTM model has better performance. The traffic flow parameters of the two graph structures have different performances during the morning and evening rush hours, but overall, ADALSTM can achieve good prediction results and has good fitting ability. Generally speaking, the performance of the ADALSTM model is better than that of the other baseline models. It can be seen that the dual attention mechanism and the graph convolution operator can effectively improve the ability of the model to extract spatio-temporal information and effectively improve the effect of the model.
[0092] In summary, using real traffic flow data as verification, compared with other algorithms, the inference model proposed by the present invention is superior to other models in short-term traffic flow prediction, and can effectively judge the traffic flow operation status of road segments and road networks in the short term in the future, providing a basis for traffic management and control.
[0093] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A network-level traffic operation state inference method for unobservable road sections, characterized in that, The following steps are involved: S1. Data preprocessing method based on police checkpoint detection data to achieve correction or deletion of abnormal data, and on this basis extract traffic flow characteristic parameters such as flow rate, speed, and density; S11. Match and compare traffic detection data from two adjacent public security checkpoints, and eliminate invalid license plate data that cannot be correctly read and unreasonable license plate data that appears repeatedly within a fixed observation time window; S12. Divide the signal cycle to which the current vehicle belongs according to the timestamp of the public security checkpoint detection data, deduct 4 seconds of starting loss from the time of the first vehicle passing the stop line in each cycle, and obtain the green light start time of each cycle; S13. Compare the green light on-time with the actual green light on-time to obtain an offset, perform normal distribution fitting on the offset distribution in the entire study area, and select the offset with the highest normal distribution probability as the calibration value of the drift correction algorithm to correct the offset error of the green light on-time; S14, selecting a fixed observation time window, and counting the traffic volume, average travel speed, and average vehicle density of observable road sections within the road network, and eliminating abnormal data by calculating the statistical distribution characteristics of the average travel speed of vehicles; S2. Construct a traffic flow characteristic parameter completion model based on Bayesian network and traffic flow basic graph to complete the traffic flow characteristic parameters of unobservable sections in the road network; S3. Build a directed graph long short-term memory network model based on a dual attention mechanism, introduce the attention mechanism of the GAT model, and capture the factors affecting traffic flow in upstream and downstream sections; S4. Based on space syntax theory, calculate the topological value of each road segment in the road network and the vehicle speed characteristics of each road segment in the road network, wherein the topological value of each road segment in the road network is used as a spatial dimension feature; combine the topological value of each road segment in the road network with the vehicle speed characteristics to construct a directed graph convolution operator; S5. The spatial dimension features and directed graph convolution operator obtained in step S4 are reversely input into the dual-attention mechanism-directed graph long short-term memory network model to evaluate the effectiveness of the model.
2. The method for inferring network-level traffic operation status for unobservable road sections according to claim 1, wherein Step S2 includes the following sub-steps: S21. Based on the Greenshields model, the traffic flow basic graph function relationship is fitted by the least squares method; S22. Construct a Bayesian network graph structure for unobservable road sections in a road network based on basic traffic flow theory. S23. Through the functional relationship of the basic traffic flow graph constructed in S21, the relationship between the nodes in the Bayesian network model in S22 is determined to complete the traffic flow characteristic parameters of the unobservable sections in the road network.
3. The method for inferring network-level traffic operation state for unobservable road sections according to claim 1, wherein Step S3 includes the following sub-steps: S31. Introducing the graph attention module to explore the influence of traffic flow parameters between adjacent road segments; S32. Long short-term memory network is integrated with attention mechanism to optimize the weight of the function.
4. The network-level traffic operation state inference method for unobservable road sections according to claim 1, wherein Step S4 includes the following sub-steps: S41. Calculate the topological value of each road segment in the road network based on space syntax theory and use it as a spatial dimension feature; S42. By obtaining the average travel speed of the vehicle on each road section, construct the speed variance matrix of each road section as the vehicle operation speed feature, which together with the spatial dimension feature constructed in S41 forms the feature vector of the model input, and construct the directed graph convolution operator.
5. The method for inferring network-level traffic operation status for unobservable road sections according to claim 1, wherein Step S5 includes the following sub-steps: S51. Select three evaluation indicators, namely mean absolute error, root mean square error, and mean absolute percentage error in the neural network algorithm as the model effectiveness evaluation indicators; S52. Input the actual data and evaluate the effectiveness of introducing the dual attention mechanism in the model through model ablation analysis.
Citation Information
Patent Citations
Traffic state prediction method for urban road network based on key road sections
CN109285346A
Traffic flow prediction method and system, and terminal device
CN110969854A