Method and system for predicting congestion events on a reconstruction and expansion road section based on machine learning

Abstract

The application provides a reconstruction and expansion road section congestion event prediction method and system based on machine learning, belonging to the technical field of traffic prediction and management and control. First, the road alignment index data, traffic organization state data and real-time traffic flow data of the reconstruction and expansion road section are collected, and feature correlation processing is performed to generate a correlation feature set. Then, a pre-trained congestion event prediction model is called to perform congestion trend analysis on the correlation feature set to generate a prediction result containing congestion occurrence probability and congestion influence range. Then, according to the prediction result, combined with the current construction progress information, management and control strategy adaptation processing is performed to generate lane management suggestions and speed limit suggestions. Finally, the prediction result, lane management suggestions and speed limit suggestions are output to the traffic management and control terminal. In this way, various influencing factors are comprehensively considered, congestion events are dynamically predicted in real time, and the traffic operation efficiency and safety of the reconstruction and expansion road section are improved.

Description

Technical Field

[0001] This invention relates to the field of traffic prediction and control technology, and more specifically, to a method and system for predicting congestion events on reconstructed and expanded road sections based on machine learning. Background Technology

[0002] In the field of traffic engineering, road reconstruction and expansion sections often face congestion problems due to their unique construction environment and changes in traffic organization. Currently, congestion prediction and control for reconstruction and expansion sections mainly rely on traditional traffic flow theories and empirical judgments. On the one hand, traditional methods often only consider single traffic flow parameters, such as flow rate and speed, while ignoring the impact of road alignment indicators (such as curvature and gradient) and traffic organization status (such as lane closures and construction zone settings) on traffic flow. These factors are intertwined in actual traffic, jointly influencing the occurrence and development of congestion. The one-sidedness of traditional methods leads to low accuracy in prediction results.

[0003] On the other hand, most existing prediction methods lack dynamic analysis of real-time traffic flow data, failing to capture sudden changes and anomalies in traffic flow in a timely manner. Traffic conditions on road widening and reconstruction sections are complex and variable; factors such as construction progress and weather conditions can trigger sudden changes in traffic flow. Traditional static prediction methods are ill-suited to these dynamic changes and cannot provide timely and effective decision support for traffic management. Therefore, there is an urgent need for a method that can comprehensively consider multiple factors and dynamically predict congestion events on widening and reconstruction sections in real time. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a machine learning-based method for predicting congestion events on reconstructed or expanded road sections, the method comprising:

[0005] Collect road alignment data, traffic organization status data, and real-time traffic flow data for the road sections undergoing reconstruction and expansion;

[0006] The road alignment index data, traffic organization status data, and real-time traffic flow data are subjected to feature association processing to generate a set of associated features for the reconstructed and expanded road sections.

[0007] The pre-trained congestion event prediction model is invoked to perform congestion trend analysis on the associated feature set of the reconstructed and expanded road section, generating congestion event prediction results that include the probability of congestion occurrence and the scope of congestion impact.

[0008] Based on the congestion event prediction results and the current construction progress information of the road section to be reconstructed and expanded, the control strategy is adapted and processed to generate lane management suggestions and speed limit suggestions;

[0009] The congestion event prediction results, lane management suggestions, and speed limit suggestions are output to the traffic control terminal.

[0010] In another aspect, embodiments of the present invention also provide a machine learning-based system for predicting congestion events on reconstructed and expanded road sections, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0011] Based on the above, this embodiment of the invention collects road alignment index data, traffic organization status data, and real-time traffic flow data of the road section being reconstructed and expanded, and performs feature association processing to generate a set of associated features. A pre-trained congestion event prediction model is then invoked to perform congestion trend analysis on the associated feature set. This allows for real-time dynamic capture of traffic flow changes, generating prediction results that include the probability of congestion and the scope of congestion impact, significantly improving the accuracy and timeliness of prediction. Based on the prediction results and current construction progress information, control strategies are adapted to generate lane management suggestions and speed limit suggestions, helping to take preventative measures to avoid or mitigate congestion. Finally, the prediction results and control suggestions are output to the traffic control terminal, achieving timely information transmission and sharing, and effectively improving the traffic operation efficiency and safety of the road section being reconstructed and expanded. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the execution flow of the machine learning-based method for predicting congestion events on reconstructed and expanded road sections provided in an embodiment of the present invention.

[0013] Figure 2 This is a schematic diagram of exemplary hardware and software components of the machine learning-based road congestion event prediction system provided in an embodiment of the present invention. Detailed Implementation

[0014] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a machine learning-based method for predicting congestion events on reconstructed and expanded road sections, as provided in one embodiment of the present invention. The following is a detailed description of this machine learning-based method for predicting congestion events on reconstructed and expanded road sections.

[0015] This embodiment uses a section of highway undergoing reconstruction and expansion as an application scenario. This section is several kilometers long and includes various road alignment features, such as straight sections, curved sections, uphill sections, and downhill sections. There are also adjustments to the number of lanes and temporary traffic sign placement due to construction. Real-time traffic flow conditions will also change due to construction and natural conditions. The following embodiment will be described in detail based on this scenario.

[0016] Step S110: Collect road alignment index data, traffic organization status data, and real-time traffic flow data for the road section to be reconstructed and expanded.

[0017] In this embodiment, to comprehensively acquire the basic data for congestion event prediction, it is necessary to collect various relevant data for the road section to be reconstructed and expanded. Road alignment index data describes the geometric characteristics of the road, including curvature data, slope data, and road length data. This data can be obtained through road design drawings, on-site measurements of the road section using professional surveying equipment, etc. For example, using surveying equipment such as a total station to measure the road section point by point, recording the location information of each measurement point and the corresponding curvature and slope values, and then combining this with the start and end location information of the road section, the curvature distribution data and slope distribution data of the road section can be obtained.

[0018] Traffic organization status data mainly involves changes in traffic organization caused by reconstruction and expansion construction, including lane number adjustments and temporary traffic sign locations. Lane number adjustments can be obtained through construction plans and on-site surveys. For example, the construction plan will specify which road sections will have their lane numbers reduced or increased, while on-site surveys can verify the actual lane number changes. Temporary traffic sign locations can be obtained by locating each temporary traffic sign using GPS positioning devices. These signs include lane merging signs, lane separating signs, speed limit signs, etc., and their specific coordinates are recorded.

[0019] Real-time traffic flow data reflects the current traffic conditions of a road segment, including vehicle speed data, traffic volume data, and traffic density data. This data can be acquired through traffic detection equipment installed on the road segment, such as loop detectors, microwave detectors, and video surveillance equipment. Loop detectors can detect the number and speed of vehicles passing overhead; microwave detectors can monitor vehicle speed and quantity within a certain range; and video surveillance equipment can obtain information such as vehicle status and quantity through image recognition technology. By summarizing and processing the data acquired by these detection devices, real-time traffic flow data, including vehicle speed characteristics and traffic density characteristics, can be obtained.

[0020] In the process of collecting this data, privacy protection and leak prevention technologies are required for data that may contain sensitive information, such as vehicle information. For example, the obtained vehicle license plate information should be anonymized to remove identifiers that can identify specific vehicles and owners; encryption technologies, such as SSL encryption protocols, should be used during data transmission to ensure that data is not leaked during transmission; and access permissions should be set for the stored data so that only authorized personnel can access and process this data to prevent the data from being illegally obtained and used.

[0021] Step S120: Perform feature association processing on the road alignment index data, traffic organization status data, and real-time traffic flow data to generate a set of associated features for the reconstructed and expanded road sections.

[0022] After obtaining road alignment index data, traffic organization status data, and real-time traffic flow data, it is necessary to perform feature extraction and correlation processing on the above data to generate a set of correlated features that can comprehensively reflect the road segment conditions.

[0023] Step S121: Extract the curvature change characteristics of road segments from the road alignment index data, calculate the curvature change rate of continuous road segment units, and divide the curvature smooth segments and curvature abrupt segments based on the curvature change rate to obtain curvature segmentation characteristics.

[0024] Step S1211: Divide the road alignment index data into multiple continuous road segment units according to the preset road segment length, and each road segment unit corresponds to a curvature value.

[0025] In this embodiment, to facilitate the analysis of curvature changes in road segments, the curvature data in the road alignment index data needs to be divided according to a preset road segment length. The preset road segment length can be determined based on the actual situation of the road segment and the analysis requirements; for example, a suitable length can be selected as a road segment unit. The entire reconstruction and expansion road segment is sequentially divided into multiple continuous and non-overlapping road segment units according to the preset length, and each road segment unit has its corresponding start and end positions. For each road segment unit, the curvature value of the road segment unit is determined based on the collected curvature data. For example, the average curvature value of all measurement points within the road segment unit can be taken as the curvature value of the road segment unit, or the curvature value at the midpoint of the road segment unit can be taken as the curvature value of the road segment unit.

[0026] Step S1212: Calculate the curvature value difference between adjacent road segment units, and use the ratio of the curvature value difference to the length of the road segment unit as the curvature change rate.

[0027] After obtaining the curvature value of each road segment unit, the difference in curvature values ​​between two adjacent road segment units is calculated, which is the difference between the curvature values ​​of the later road segment unit and the curvature value of the earlier road segment unit. Then, this difference in curvature values ​​is divided by the length of the road segment unit. The result is the rate of change of curvature between these two adjacent road segment units. The rate of change of curvature reflects the change in road segment curvature per unit length, and can be used to determine whether the curvature change of the road segment is gradual or drastic.

[0028] Step S1213: When the curvature change rate of multiple consecutive road segment units is within the curvature change rate threshold range, the corresponding area is marked as a curvature smooth segment.

[0029] In this embodiment, a curvature change rate threshold range needs to be set. This threshold range is determined based on road design specifications and actual traffic operation experience, and is used to determine whether the curvature change of a road segment is gentle. When the curvature change rate between multiple consecutive road segment units is within this curvature change rate threshold range, it indicates that the curvature change within the area composed of these road segment units is relatively gentle, and therefore this area is marked as a gentle curvature segment. For example, on a long straight road segment, the curvature change rate of adjacent road segment units is usually small and will fall within the curvature change rate threshold range, thus being marked as a gentle curvature segment.

[0030] Step S1214: When the curvature change rate of a single road segment unit exceeds the threshold range, the road segment unit is marked as a curvature abrupt change segment.

[0031] If the rate of curvature change between a road segment unit and its adjacent road segment units exceeds a set threshold range, it indicates that the curvature of that road segment unit has undergone a relatively drastic change, and therefore that road segment unit is marked as a curvature abrupt change segment. For example, at a sharp bend in the road, the rate of curvature change of its adjacent road segment units will be large, possibly exceeding the threshold range, and thus it will be marked as a curvature abrupt change segment.

[0032] Step S1215: Integrate the marking information of the smooth curvature segments and the abrupt curvature segments according to the positional order of the road segment units to generate curvature segmentation features containing the start and end positions of each segment.

[0033] After marking the sections with gentle curvature and those with abrupt curvature changes, this marking information is integrated according to the positional order of the road segment units within the entire reconstruction and expansion section. For each section with gentle curvature and abrupt curvature change, its start and end positions are recorded, which can be represented by the distance from the start point of the road segment. Through this integration, curvature segmentation features are generated, which clearly reflect which areas in the entire reconstruction and expansion section are sections with gentle curvature, which areas are sections with abrupt curvature changes, and their respective location ranges.

[0034] Step S122: Extract the slope distribution characteristics from the road alignment index data, analyze the slope value change trend of adjacent road segment units, determine the uphill transition section and downhill transition section, and obtain the slope transition characteristics.

[0035] Step S1221: Divide the slope data in the road alignment index data into the same preset road segment length as the road segment unit, so that each road segment unit corresponds to a slope value.

[0036] Similar to processing curvature data, to analyze slope variations, the slope data is also divided into multiple segment units of the same preset segment length, with each segment unit corresponding to a slope value. The slope value can be determined using a method similar to that for curvature values, such as taking the average slope of multiple measurement points within the segment unit.

[0037] Step S1222: Calculate the slope difference between adjacent road segment units, and determine the slope change trend based on the sign of the difference.

[0038] Calculate the difference in slope values ​​between two adjacent road segment units, which is the slope value of the latter road segment unit minus the slope value of the former road segment unit. If the difference is positive, it indicates that the slope is increasing, meaning the road segment has an upward tendency; if the difference is negative, it indicates that the slope is decreasing, meaning the road segment has a downward tendency.

[0039] Step S1223: When the slope difference of multiple consecutive road segment units is positive and the trend of change is stable, the area composed of these road segment units is determined as the uphill transition section.

[0040] Establish a criterion for judging the stability of the slope change trend, such as the fluctuation range of the slope value difference of multiple consecutive road segment units within a certain range. When the slope value difference of multiple consecutive road segment units is positive and meets the criterion of stable change trend, it indicates that the area composed of these road segment units is gradually uphill, and therefore it is identified as an uphill transition section.

[0041] Step S1224: When the slope difference of multiple consecutive road segment units is negative and the trend of change is stable, the area composed of these road segment units is determined as the downhill transition section.

[0042] Similarly, when the difference in slope values ​​of multiple consecutive road segment units is negative and the trend of change is stable, it indicates that the area composed of these road segment units is gradually sloping downhill, and therefore it is identified as a downhill transition section.

[0043] Step S1225: Integrate the information of uphill transition sections and downhill transition sections according to the positional order of the road segment units to generate slope transition features that include the starting position, ending position and slope change trend of each section.

[0044] Based on the location sequence of road segment units, the information of uphill and downhill transition sections is integrated, and the start and end positions of each transition section, as well as the slope change trend (i.e., uphill or downhill), are recorded to generate slope transition features. These features can reflect the distribution of uphill and downhill transition sections and their slope change trends throughout the entire reconstruction and expansion section.

[0045] Step S123: Extract lane number adjustment information from the traffic organization status data, identify areas where the number of lanes decreases and areas where the number of lanes increases, and combine the temporary traffic sign placement information from the traffic organization status data to determine the lane change guidance area and obtain lane organization characteristics.

[0046] Step S1231: Analyze the lane number adjustment information in the traffic organization status data, extract the lane number values ​​at different road segment locations, and generate a lane number distribution sequence.

[0047] The lane number adjustment information in the collected traffic organization status data is analyzed to determine the number of lanes at different locations on the road segment being reconstructed or expanded. For example, a road segment may have a certain number of lanes before construction, and after construction, a certain number of lanes may be reduced in one area and a certain number of lanes may be added in another area. These lane number values ​​from different road segment locations are arranged in chronological order to generate a lane number distribution sequence, which reflects the changes in lane number along the road segment.

[0048] Step S1232: Compare the lane number values ​​of adjacent road segments. When the lane number value of the later road segment is less than the lane number value of the earlier road segment, mark the later road segment as a lane number reduction area. When the lane number value of the later road segment is greater than the lane number value of the earlier road segment, mark the later road segment as a lane number increase area.

[0049] The lane number values ​​of adjacent road segments in the lane number distribution sequence are compared. If the lane number value of the later road segment is less than that of the earlier road segment, it indicates that the number of lanes in the later road segment has decreased, and this area is marked as a lane number decrease area; if the lane number value of the later road segment is more than that of the earlier road segment, this area is marked as a lane number increase area.

[0050] Step S1233: Extract the location information of the temporary traffic signs and determine the coordinates of the lane change signs. The lane change signs include lane merging signs and lane separating signs.

[0051] The coordinates of lane change signs were extracted from the collected location information of temporary traffic signs. Lane change signs mainly include lane merging signs and lane separating signs. Lane merging signs indicate that the lanes ahead will merge, and lane separating signs indicate that the lanes ahead will separate.

[0052] Step S1234: Match the lane reduction area with the layout coordinates of the lane merging sign, and determine the successfully matched area as the lane merging guidance area.

[0053] The location of the marked lane reduction area is compared with the coordinates of the lane merging sign. If the coordinates of the lane merging sign are within the location range of the lane reduction area, or overlap with the location range of the lane reduction area, the two are considered to be matched successfully, and the lane reduction area is designated as the lane merging guidance area. This is because lane merging signs are usually placed in the area where lanes are about to merge to guide vehicles in lane changes.

[0054] Step S1235: Match the location of the lane number increase area with the layout coordinates of the lane separation sign, and determine the successfully matched area as the lane separation guidance area.

[0055] Similarly, the location range of the lane increase area is matched with the coordinates of the lane separation sign. When the coordinates of the lane separation sign are within or overlap with the lane increase area, the lane increase area is designated as the lane separation guidance area. Lane separation signs are generally placed in areas where lanes are about to separate to guide vehicles into the corresponding lanes.

[0056] Step S1236: Integrate the location information of the lane merging guidance area and the lane separation guidance area to generate lane organization features that include guidance area type and range coordinates.

[0057] The location information of the determined lane merging guidance areas and lane separating guidance areas is integrated to clarify the type of each guidance area (whether it is a lane merging guidance area or a lane separating guidance area) and their range coordinates (starting position coordinates and ending position coordinates), thereby generating lane organization features. These features can clearly reflect the regional distribution of lane change guidance in the road segment.

[0058] Step S124: Extract vehicle speed features from the real-time traffic flow data, calculate speed difference values ​​for different lanes, and combine them with traffic flow density features from the real-time traffic flow data to analyze the correlation between speed and density, thereby obtaining traffic flow state features.

[0059] Step S1241: Extract vehicle speed data for each lane from real-time traffic flow data. This data can be the average speed per unit time.

[0060] Real-time traffic flow data includes vehicle speed information for each lane, from which the average speed data for each lane over a unit of time is extracted. For example, the average speed of vehicles in each lane is calculated every five minutes to obtain vehicle speed data for each lane at different time periods.

[0061] Step S1242: Calculate the speed difference between adjacent lanes or lanes in different directions, which is the absolute value of the difference between the average speeds of the two lanes.

[0062] For adjacent lanes traveling in the same direction, or lanes traveling in different directions, calculate the speed difference between them. Specifically, subtract the average speed of one lane from the average speed of the other lane, and then take the absolute value to obtain the speed difference between the two lanes. These difference values ​​allow us to understand the changes in vehicle speeds across different lanes.

[0063] Step S1243: Extract traffic flow density data from real-time traffic flow data. Traffic flow density data represents the number of vehicles per unit length of road segment.

[0064] Traffic density data is an important indicator reflecting the degree of traffic congestion. It is the number of vehicles per unit length of road segment extracted from real-time traffic flow data. For example, the number of vehicles per kilometer of road segment.

[0065] Step S1244: Analyze the correlation between vehicle speed and traffic density. By observing the trend of vehicle speed change when traffic density changes, determine the correlation between the two.

[0066] The extracted vehicle speed and traffic density data were analyzed to observe whether vehicle speed increased or decreased as traffic density increased, and by what magnitude; and how vehicle speed changed when traffic density decreased. Through this analysis, the correlation between speed and density was determined; for example, generally, as traffic density increases, vehicle speed decreases, and the two are negatively correlated.

[0067] Step S1245: Integrate the speed difference values ​​of different lanes, traffic flow density data, and the correlation information between speed and density to generate traffic flow state features.

[0068] The calculated speed differences between different lanes, extracted traffic flow density data, and analyzed correlation information between speed and density are integrated to form traffic flow state characteristics. These characteristics can comprehensively reflect the current traffic flow operation status of the road segment.

[0069] Step S125: Input the curvature segmentation feature, the slope transition feature, the lane organization feature, and the traffic flow state feature into the feature association network. Calculate the mutual influence coefficients between each feature through the weighted connections of the network layers of the feature association network to generate a set of associated features for the reconstructed and expanded road section that includes feature association weights.

[0070] Step S1251: Construct a feature association network, which includes an input layer, a hidden layer and an output layer. The input layer is used to receive curvature segmentation features, slope transition features, lane organization features and traffic flow state features. The hidden layer is used to process the features and calculate the mutual influence coefficients. The output layer is used to output a set of associated features of the reconstructed and expanded road segment containing feature association weights.

[0071] The Feature Association Network (FAN) is a neural network model used to calculate the interaction coefficients between features. The input layer contains neurons corresponding to each feature dimension, receiving input features such as curvature segmentation, slope transition, lane organization, and traffic flow status. The hidden layer contains multiple neurons connected to the input layer neurons via weighted connections to process the input features. The output layer outputs the final set of associated features for the reconstructed / expanded road segment, which includes the association weights between each feature.

[0072] Step S1252: Convert the curvature segmentation features, slope transition features, lane organization features, and traffic flow state features into vector forms suitable for the input feature association network.

[0073] Since the feature association network receives data in vector form, it is necessary to transform the curvature segmentation features, slope transition features, lane organization features, and traffic flow state features separately. For example, for curvature segmentation features, information such as the start and end positions of each segment can be converted into numerical vectors; for traffic flow state features, data such as speed difference values ​​and traffic density can be composed into vectors.

[0074] Step S1253: Input the transformed feature vector into the input layer of the feature association network, and the neurons in the input layer pass the feature vector to the neurons in the hidden layer.

[0075] The transformed feature vectors are input into the input layer of the feature association network. Each neuron in the input layer corresponds to an element in the feature vector. Then, the neurons in the input layer pass the received feature data to the neurons in the hidden layer through weighted connections.

[0076] Step S1254: The neurons in the hidden layer process the received feature data and calculate the mutual influence coefficients between the features through weighted connections.

[0077] After receiving feature data from the input layer, the neurons in the hidden layer process this data according to the connection weights between neurons. The connection weights between neurons corresponding to different features represent the degree of influence between them. By calculating and adjusting these weights, the mutual influence coefficients between each feature are obtained. For example, the curvature abrupt change segment feature in the curvature segmentation feature may have a significant impact on the vehicle speed feature in the traffic flow state feature, and the corresponding mutual influence coefficient will be large.

[0078] Step S1255: The hidden layer passes the calculated mutual influence coefficients to the output layer, and the output layer integrates these coefficients to generate a set of associated features of the reconstructed and expanded road sections, which includes feature association weights.

[0079] After calculating the interaction coefficients between various features in the hidden layer, these coefficients are passed to the output layer. The neurons in the output layer receive these interaction coefficients and integrate them. During integration, the interaction coefficients can be adjusted and combined according to the importance of each feature in the overall road segment state description, ultimately forming a set of associated features for the reconstructed and expanded road segment that includes feature association weights. These feature association weights reflect the strength of the interaction between different features. For example, the association weight between the lane merging guidance area feature in lane organization features and the traffic flow density feature in traffic flow state features reflects the degree of influence of lane merging on changes in traffic flow density. Through this integration, the set of associated features for the reconstructed and expanded road segment can comprehensively reflect the intrinsic relationships between various factors such as road alignment, traffic organization, and traffic flow.

[0080] Step S130: Call the pre-trained congestion event prediction model to perform congestion trend analysis on the associated feature set of the reconstructed and expanded road section, and generate congestion event prediction results including the probability of congestion and the scope of congestion impact.

[0081] After obtaining the set of associated features of the road sections to be reconstructed and expanded, it is necessary to use a pre-trained congestion event prediction model to analyze them in order to predict relevant information about possible congestion events.

[0082] Step S131: Input the set of associated features of the reconstructed and expanded road section into the feature adaptation layer of the congestion event prediction model, and standardize the feature dimensions to match the preset input dimensions of the model to obtain standardized associated features.

[0083] Step S1311: The feature adaptation layer receives the set of features associated with the reconstructed and expanded road sections, and analyzes the number of feature dimensions and the feature distribution of each dimension in the set.

[0084] The feature adaptation layer of the congestion event prediction model first receives the input set of features associated with the reconstructed and expanded road sections. It then parses this set to determine the number of feature dimensions it contains, such as dimensions related to curvature segmentation features, slope transition features, lane organization features, traffic flow state features, and the dimensions related to the correlation weights between them. Simultaneously, it analyzes the data distribution of each feature dimension to understand the range and distribution pattern of feature values ​​for each dimension.

[0085] Step S1312: Based on the preset number of input dimensions and feature distribution standards of the model, perform dimension adjustment and distribution standardization processing on the associated feature set of the reconstructed and expanded road sections.

[0086] The preset number of input dimensions for the model is determined during model training. To ensure that the input feature set matches the model's processing requirements, the dimensions of the feature set associated with the road reconstruction and expansion sections need to be adjusted. If the number of dimensions in the feature set exceeds the preset input dimensions, feature selection is used to retain the feature dimensions more important for predicting congestion events. If the number of dimensions is less than the preset input dimensions, new relevant feature dimensions are generated to supplement them through feature expansion. Simultaneously, the distribution of each feature dimension is normalized to ensure that the distribution of feature values ​​conforms to the model's preset standards. For example, feature values ​​are transformed to the same value range to avoid affecting the model's processing performance due to excessively large differences in the magnitude of feature values.

[0087] Step S1313: After dimensional adjustment and distribution normalization, standardized association features that match the preset input dimensions of the model are generated.

[0088] Through the aforementioned dimensional adjustments and distribution normalization, the set of associated features for the reconstructed and expanded road sections is made to match the model's preset input requirements in both feature dimensions and distribution, thus obtaining standardized associated features. These standardized associated features can be better processed by subsequent network layers of the congestion event prediction model.

[0089] Step S132: The standardized associated features are extracted collaboratively by the road segment feature extraction layer of the congestion event prediction model, which extracts local and global features. The local features include the curvature and slope combination features of the road segment unit, and the global features include the lane organization and traffic flow distribution features of the entire road segment.

[0090] Step S1321: Divide the standardized associated features into multiple local feature blocks according to road segment units, with each local feature block corresponding to the feature data of a road segment unit.

[0091] The road segment feature extraction layer first divides the standardized associated features into multiple local feature blocks according to the previously defined road segment units. Each local feature block contains the curvature segmentation features, slope transition features, lane organization features related to the corresponding road segment unit, traffic flow status features, and the association weight information between these features.

[0092] Step S1322: Perform feature combination processing on the curvature segmentation features and slope transition features in each local feature block, calculate the product of curvature value and slope value as the combination coefficient, and generate the curvature and slope combination features of the road segment unit.

[0093] For each local feature block, the curvature value contained in the curvature segmentation feature and the slope value contained in the slope transition feature are extracted. These two values ​​are multiplied together, and the result is used as a combination coefficient. Through the above combination processing, the two parameters reflecting the local features of road alignment, curvature and slope, are combined to generate the curvature and slope combination feature of each road segment unit. This curvature and slope combination feature can more comprehensively reflect the impact of the road alignment condition of the road segment unit on traffic operation.

[0094] Step S1323: Summarize the curvature and slope combination features of all local feature blocks to obtain a local feature set containing local information of all road segment units.

[0095] The curvature and slope combination features of each road segment unit are summarized according to the positional order of the road segment units to form a local feature set containing local information of all road segment units. This set can comprehensively reflect the road alignment combination features of various local areas in the entire reconstruction and expansion section.

[0096] Step S1324: Perform statistical analysis on the lane organization features and traffic flow status features in the standardized correlation features across the entire road segment, calculate the average number of lanes and average traffic flow density of the entire road segment, and generate the lane organization distribution features and traffic flow distribution features of the entire road segment.

[0097] Within the entire road segment, statistical analysis is performed on the lane organization characteristics in the standardized correlation features to calculate the average number of lanes across the entire segment. This average number of lanes reflects the overall lane organization situation across the entire segment, thus generating lane organization distribution characteristics. Simultaneously, statistical analysis is performed on the traffic flow density data in the traffic flow state features across the entire road segment to calculate the average traffic flow density across the entire segment, generating traffic flow distribution characteristics. These traffic flow distribution characteristics reflect the overall traffic congestion level across the entire road segment.

[0098] Step S1325: Perform feature concatenation between the lane organization distribution features and the traffic flow distribution features to obtain a global feature containing overall information of the entire road segment.

[0099] The generated lane organization distribution features and traffic flow distribution features are concatenated in a predetermined order to form a global feature. For example, the vector of lane organization distribution features is concatenated with the vector of traffic flow distribution features to form a new vector. This vector contains information on the overall lane organization and traffic flow distribution of the entire road segment, and can reflect the traffic operation environment of the entire road segment from a macroscopic perspective.

[0100] Step S133: Input the local features and the global features into the spatiotemporal correlation layer of the congestion event prediction model, construct a feature correlation matrix based on the road segment location relationship and time series continuity, and calculate the feature transfer coefficient of different road segment units in the time dimension.

[0101] Step S1331: The spatiotemporal correlation layer receives local and global features and parses out the location information and feature data of each road segment unit as a function of time.

[0102] The spatiotemporal correlation layer first parses the input local and global features. It obtains the location information of each road segment unit (such as the distance from the starting point of the road segment) and the value of the feature data of each road segment unit at different time points from the local features; and obtains the change of the feature data of the entire road segment over time from the global features.

[0103] Step S1332: Determine the adjacency and distance relationships between each road segment unit based on the location information of the road segment units, and construct a spatial association matrix.

[0104] Based on the location information of each road segment unit, determine which road segment units are adjacent and their distances. Represent these spatial relationships in the form of a matrix, namely the spatial correlation matrix. The element values ​​in the matrix represent the degree of spatial correlation between two road segment units. Adjacent road segment units that are close to each other have a relatively high degree of spatial correlation, and vice versa.

[0105] Step S1333: Based on the changes in feature data over time, analyze the continuity and correlation between feature data at different time points, and construct a time correlation matrix.

[0106] Analyzing the values ​​of feature data at different time points allows us to observe trends and determine the degree of correlation between feature data at adjacent time points. For example, if the change in traffic flow density feature value between a previous time point and a subsequent time point is small, it indicates a strong temporal correlation between the two. These temporal correlations are then represented in matrix form, known as a temporal correlation matrix, where each element represents the degree of temporal correlation between feature data at different time points.

[0107] Step S1334: Merge the spatial correlation matrix and the temporal correlation matrix to construct a feature correlation matrix, which simultaneously reflects the spatial and temporal correlations between different road segment units.

[0108] The spatial correlation matrix and the temporal correlation matrix are merged according to a set rule, such as through matrix addition or matrix multiplication, to obtain the feature correlation matrix. The element values ​​in the feature correlation matrix reflect both the spatial correlation between different road segment units and their temporal correlation, thus comprehensively representing the spatiotemporal correlation of features.

[0109] Step S1335: Based on the feature correlation matrix, calculate the feature transfer coefficient of different road segment units in the time dimension. This coefficient represents the degree of influence of the feature change of one road segment unit on the feature change of another road segment unit in subsequent time.

[0110] Based on the element values ​​in the feature correlation matrix, the feature transfer coefficients between different road segment units in the time dimension are obtained. For example, for two road segment units, if their spatial correlation is high and their temporal correlation is also high, the feature transfer coefficient between them is relatively large, indicating that the feature changes of one road segment unit will have a significant impact on the feature changes of the other road segment unit in subsequent time.

[0111] Step S134: The feature correlation matrix is ​​processed by the congestion probability calculation layer of the congestion event prediction model, and the probability of congestion occurrence for each road segment unit is generated by combining the probability distribution pattern of historical congestion data.

[0112] Step S1341: The congestion probability calculation layer receives the feature correlation matrix, processes the element values ​​in the matrix, and converts them into intermediate parameters related to the probability of congestion.

[0113] The congestion probability calculation layer first processes the input feature correlation matrix, transforming the element values ​​into intermediate parameters that reflect the probability of congestion occurring through matrix operations. The range of values ​​for these intermediate parameters is related to the possible range of congestion probabilities.

[0114] Step S1342: Introduce the probability distribution pattern of historical congestion data. This pattern is obtained by statistical analysis of historical congestion data of the road section under reconstruction and expansion over a period of time, including the distribution of the frequency of congestion in different road sections and the duration of congestion.

[0115] The probability distribution patterns of historical congestion data are important information learned during model training, reflecting the statistical characteristics of congestion occurrences on that road segment in the past. For example, certain road segments may experience relatively high congestion frequencies due to road alignment or traffic organization, which will be reflected in the probability distribution patterns.

[0116] Step S1343: Match and adjust the intermediate parameters with the probability distribution of historical congestion data to obtain the probability of congestion occurrence for each road segment unit.

[0117] Based on the probability distribution patterns of historical congestion data, the intermediate parameters obtained from the transformation are adjusted and calibrated to make the results more consistent with the actual congestion situation of the road segment. Through the above processing, the congestion probability of each road segment unit is finally generated, which represents the likelihood of congestion occurring in each road segment unit under the current traffic conditions and road conditions.

[0118] Step S135: Based on the congestion occurrence probability of each road segment unit and the transmission coefficient in the feature correlation matrix, predict the path of congestion spreading from high-probability road segments to adjacent road segments, and determine the scope of congestion impact.

[0119] Step S1351: Select road segment units whose probability values ​​exceed a preset probability threshold from the congestion occurrence probabilities of each road segment unit, and mark them as initial congested road segment units.

[0120] The preset probability threshold is set based on the traffic management needs and historical congestion data of the road segment. It is used to determine whether a road segment unit belongs to a high-probability road segment that may experience congestion. The probability of congestion in each road segment unit is compared with this threshold. Road segment units that exceed the threshold are marked as initially congested road segment units. These road segment units are the areas where congestion is most likely to occur first.

[0121] Step S1352: Extract the row vector of transfer coefficients corresponding to the initial congested road segment unit in the feature correlation matrix. The row vector of transfer coefficients contains the feature transfer coefficients of the initial congested road segment unit and all other road segment units.

[0122] In the feature correlation matrix, each row corresponds to the feature transfer coefficient between a road segment unit and all other road segment units. The row vector corresponding to the initial congested road segment unit is extracted. Each element in this vector represents the feature transfer coefficient between the initial congested road segment unit and another road segment unit, reflecting the degree of influence of the feature changes of the initial congested road segment unit on the feature changes of other road segment units.

[0123] Step S1353: Sort other road segment units according to the numerical value of the transmission coefficient row vector, and select the top few road segment units with the largest transmission coefficient as potential diffusion road segment units.

[0124] The elements in the row vector of transmission coefficients are sorted in descending order, and the top-ranked road segment units are selected as potential diffusion road segment units. These potential diffusion road segment units have larger characteristic transmission coefficients than the initial congested road segment units, meaning that after congestion occurs in the initial congested road segment unit, the congestion state is more likely to spread to these potential diffusion road segment units.

[0125] Step S1354: Calculate the road segment distance between the initial congested road segment unit and each potential diffusion road segment unit, and generate a diffusion probability score by combining the transmission coefficient and the road segment distance.

[0126] Using the location information of road segment units, the actual road segment distance between the initial congested road segment unit and each potential spreading road segment unit is calculated. Then, the transmission coefficient and road segment distance are combined to generate a diffusion probability score. For example, the larger the transmission coefficient and the closer the road segment distance, the higher the diffusion probability score, indicating that the congestion is more likely to spread from the initial congested road segment unit to that potential spreading road segment unit.

[0127] Step S1355: Mark potential diffusion road segment units whose diffusion probability scores exceed the preset scoring threshold as secondary congestion road segment units.

[0128] A preset scoring threshold is used to determine whether a potential spreading road segment unit is likely to become a secondary congestion area. The diffusion probability score of each potential spreading road segment unit is compared with this threshold. Potential spreading road segment units that exceed the threshold are marked as secondary congestion road segment units, which are likely to become congested due to the influence of the initial congestion road segment unit.

[0129] Step S1356: Repeat the steps of extracting the row vector of the transmission coefficient corresponding to the secondary congested road segment unit, screening potential diffusion road segment units, and calculating the diffusion probability score until no new secondary congested road segment units are generated.

[0130] For the identified secondary congested road segment units, their corresponding transmission coefficient row vectors are extracted in the same manner as those for the initial congested road segment units. New potential diffusion road segment units are then selected, and their diffusion probability scores are calculated. These new secondary congested road segment units are then identified. This process is repeated until no new secondary congested road segment units are identified, at which point all road segment units potentially affected by congestion have been covered.

[0131] Step S1357: Integrate the location information of the initial congested road segment unit and all secondary congested road segment units, determine the boundary road segment units of the congestion impact range, and generate range description information containing all congestion-related road segment units.

[0132] The location information of the initial congested road segment units and all marked secondary congested road segment units is integrated to determine the distribution range of these road segment units throughout the entire road reconstruction and expansion section. The boundary road segment units of this range are identified, i.e., the road segment units located at the outermost edge of the range, and these boundary road segment units are used to define the scope of congestion impact. Finally, a range description information containing all congestion-related road segment units is generated, clearly indicating the areas where congestion may spread.

[0133] Step S136: Integrate the congestion occurrence probability of each road segment unit and the congestion impact range to generate a congestion event prediction result containing a probability distribution sequence and range boundary coordinates.

[0134] The congestion probability of each road segment is arranged in order of location to form a probability distribution sequence, which reflects the distribution of congestion probability across the entire road segment. Simultaneously, the coordinates of the boundary road segments within the congestion impact area are compiled to form boundary coordinates. The probability distribution sequence and boundary coordinates are then fused to obtain the congestion event prediction result, which comprehensively includes information on the probability of congestion and the potential affected areas.

[0135] Step S140: Based on the congestion event prediction results and the current construction progress information of the road section to be reconstructed and expanded, the control strategy is adapted to generate lane management suggestions and speed limit suggestions.

[0136] After obtaining the congestion event prediction results, it is necessary to combine the current construction progress information of the road sections to construct and expand them, and formulate corresponding traffic control strategies to alleviate or avoid congestion.

[0137] Step S141: Analyze the congestion occurrence probability distribution sequence in the congestion event prediction result, and determine the location and corresponding probability value of the candidate congested road segment unit, wherein the candidate congested road segment unit is a road segment unit whose congestion occurrence probability is greater than a set probability threshold.

[0138] Step S1411: Analyze the congestion occurrence probability distribution sequence in the congestion event prediction results and extract the congestion occurrence probability value for each road segment unit.

[0139] By analyzing the probability distribution sequence in the congestion event prediction results, the probability value of congestion occurrence corresponding to each road segment unit is extracted one by one to understand the likelihood of congestion occurring in each road segment unit.

[0140] Step S1412: Compare the congestion probability value of each road segment unit with the set probability threshold, and select road segment units with probability values ​​greater than the set probability threshold as candidate congested road segment units.

[0141] The probability threshold is set based on the strictness of traffic control and the traffic demand of road segments, and is used to screen out road segments that are likely to be congested and require special attention. After comparing the congestion probability value of each road segment with the threshold, road segments with a probability greater than the threshold are identified as candidate congested road segments, which are the areas that the control strategy needs to focus on.

[0142] Step S1413: Record the location information of the candidate congested road segment unit and the corresponding congestion probability value.

[0143] The location of the identified candidate congested road segment units (such as the distance range from the start of the road segment) and the corresponding probability value of congestion will be recorded. This information will serve as an important reference for formulating lane management recommendations and speed limit recommendations.

[0144] Step S142: Obtain the current construction progress information of the road section to be reconstructed and expanded, and extract the location range and construction stage identifier of the construction area. The construction stage identifier includes the roadbed construction stage, the pavement construction stage, and the ancillary facility construction stage.

[0145] Step S1421: Obtain the current construction progress information of the road section to be reconstructed and expanded through the construction management system or on-site reporting.

[0146] Current construction progress information can be obtained from the construction unit's construction management system or through regular reports from on-site management personnel. This information reflects the progress and current status of the construction.

[0147] Step S1422: Extract the location range of the construction area from the current construction progress information. This range is usually represented by the start and end station numbers of the road segment.

[0148] The location of the construction area clearly defines the road section currently under construction or affected by construction. By extracting information such as the starting and ending station numbers, the location of the construction area within the entire road reconstruction and expansion section can be accurately determined.

[0149] Step S1423: Extract the construction stage identifier to determine the current construction stage, i.e., the roadbed construction stage, the pavement construction stage, or the ancillary facilities construction stage.

[0150] Construction phase markers reflect the main tasks being carried out at any given time, and different construction phases have varying degrees and methods of impact on traffic. For example, the roadbed construction phase may cause greater damage to the pavement structure, resulting in a more severe impact on traffic; while the ancillary facilities construction phase may have a relatively smaller impact on traffic.

[0151] Step S143: Perform spatial overlay analysis between the location of the candidate congested road segment unit and the location range of the construction area to determine whether the candidate congested road segment unit is located within the construction area or within the influence range of the construction area.

[0152] Step S1431: Obtain the set of location coordinates of candidate congested road segment units. The set of location coordinates includes the start and end location coordinates of each candidate congested road segment unit. The location coordinates are described by the linear distance from the start point of the road segment and the lateral offset distance.

[0153] In this embodiment, the location information of candidate congested road segment units is precisely defined by their starting and ending coordinates. These coordinates are described using the starting point of the entire reconstruction and expansion section as a reference point. For each candidate congested road segment unit, not only is its linear distance from the reference point along the length of the road segment recorded, but also its lateral offset distance relative to the road segment centerline (perpendicular to the length of the road segment). This clarifies the specific location of the road segment unit on the road cross-section. For example, the starting position of a candidate congested road segment unit might be described as a certain linear distance from the reference point with a certain lateral offset, while the ending position would be a further linear distance and a corresponding lateral offset. Through this set of coordinates, the spatial range of the candidate unit within the road can be accurately delineated.

[0154] Step S1432: Extract the location range data of the construction area. This data includes the set of boundary coordinates of the construction area. The boundary coordinates are also based on the starting point of the road segment and are represented by linear distance and lateral offset distance. They also include the leftmost boundary, rightmost boundary, starting boundary and ending boundary of the construction area.

[0155] The location of the construction area is determined by its set of boundary coordinates. The reference points of these boundary coordinates are consistent with the location coordinates of the candidate congested road segment units, both based on the starting point of the road segment. The boundaries of the construction area include the leftmost boundary, the rightmost boundary (used to define the lateral range), and the starting boundary and ending boundary (used to define the range along the length of the road segment). The coordinates of each boundary point are described by both the linear distance from the reference point and the lateral offset distance. For example, the starting boundary of the construction area may be a certain linear distance from the reference point and a lateral range covering a range from one offset distance to another, while the ending boundary is a further linear distance and the corresponding lateral range. These boundary coordinates allow for a complete determination of the specific coverage area of ​​the construction area in the road space.

[0156] Step S1433: Construct a spatial overlay analysis model. This model can calculate the spatial relationship between the set of location coordinates of candidate congested road segment units and the set of boundary coordinates of the construction area. The calculation includes whether the candidate unit is completely located inside the construction area, whether it partially overlaps, whether it is located outside the construction area but within the influence range, and whether it is completely outside the influence range.

[0157] The core function of the spatial overlay analysis model is to compare and calculate the coordinate data of two spatial ranges to determine their spatial relationship. First, the model compares the start and end coordinates of candidate congested road segment units with the boundary coordinates of the construction area one by one. It determines whether the lateral extent of the candidate unit is completely within the lateral boundary of the construction area, and whether its length along the road segment is completely within the start and end boundaries of the construction area. If both conditions are met, it is determined to be completely within the construction area. If part of the lateral extent or part of the length of the candidate unit overlaps with the boundary of the construction area, it is determined to be partially overlapping. For candidate units located outside the construction area, the model further calculates the shortest distance to the boundary of the construction area and, combined with a preset construction influence range distance threshold, determines whether it is within the influence range of the construction area (i.e., the shortest distance is less than or equal to the threshold). If the shortest distance is greater than the threshold, it is determined to be completely outside the influence range.

[0158] Step S1434: Input the set of location coordinates of candidate congested road segment units into the spatial overlay analysis model. The model calls the boundary comparison algorithm to compare the boundary coordinates of the candidate units with those of the construction area point by point.

[0159] When performing spatial overlay analysis, the set of location coordinates of candidate congested road segment units is first input into the constructed spatial overlay analysis model. The spatial overlay analysis model then activates a boundary comparison algorithm, which compares each coordinate point of the candidate unit (including the start point, end point, and key nodes on the boundary line) with the boundary coordinate points of the construction area one by one. For example, it can first check whether the lateral offset range of the candidate unit intersects with the lateral boundary range of the construction area, and then check whether its range along the length of the road segment intersects with the start and end boundary ranges of the construction area. Through the above point-by-point comparison, a preliminary judgment is made on the spatial relationship between the two.

[0160] Step S1435: Based on the boundary comparison results and combined with the spatial relationship determination rules, determine the spatial location category of the candidate congested road segment unit, namely, within the construction area, within the influence range of the construction area, or outside both.

[0161] The spatial relationship determination rules are based on the results of the boundary comparison algorithm. If a candidate congested road segment is determined to be completely within the boundary of the construction area, it is classified as "within the construction area"; if a candidate segment is determined to be outside the construction area, but the shortest distance to the boundary of the construction area is less than or equal to a preset influence range threshold, it is classified as "within the influence range of the construction area"; if a candidate segment neither meets the conditions of being completely within the construction area nor within the influence range, it is classified as "outside of both". Through this classification, the spatial positional relationship between each candidate congested road segment and the construction area is clearly defined.

[0162] Step S144: When the candidate congested road segment unit is located in the construction area, query the corresponding lane traffic restriction rules according to the construction stage identifier, and combine the lane number adjustment information in the lane organization characteristics to generate lane closure suggestions or lane one-way traffic suggestions as lane management suggestions.

[0163] Step S144: When the candidate congested road segment unit is located in the construction area, query the corresponding lane traffic restriction rules according to the construction stage identifier, and combine the lane number adjustment information in the lane organization characteristics to generate lane closure suggestions or lane one-way traffic suggestions as lane management suggestions.

[0164] Step S1441: Query the corresponding lane traffic restriction rules from the preset construction stage rule base according to the construction stage identifier. The lane traffic restriction rules include the maximum number of lanes allowed to pass and the minimum lane width requirement.

[0165] In this embodiment, the pre-defined construction phase rule base is established based on a large number of construction cases and traffic management regulations. For different construction phase identifiers, corresponding lane restriction rules are stored. When a candidate congested road segment is determined to be located within a construction area, the construction phase identifier of that construction area is first obtained, such as the roadbed construction phase, pavement construction phase, or ancillary facility construction phase. Then, the construction phase rule base is searched according to this identifier to find the corresponding lane restriction rules. These lane restriction rules specify in detail the maximum number of lanes allowed to pass within the construction area under that construction phase, i.e., the maximum number of lanes that can be used by vehicles; they also specify the minimum lane width requirements, i.e., the width standard that each allowed lane must meet to ensure safe passage for vehicles.

[0166] Step S1442: Extract the lane number adjustment information from the lane organization features to determine the current number of lanes and the width parameters of each lane within the construction area.

[0167] Lane organization characteristics include detailed information on lane number adjustments within the construction area. By extracting and analyzing this information, the actual number of lanes within the construction area can be determined. For example, if a road segment originally had a certain number of lanes, some lanes may be occupied or closed due to construction needs, resulting in a change in the current number of lanes. Simultaneously, the width parameters of each lane also need to be obtained. These parameters can be acquired through on-site measurements or construction design drawings. The width of each lane may vary depending on the construction situation and needs to be recorded separately.

[0168] Step S1443: Compare the current number of lanes with the maximum number of lanes allowed in the lane traffic restriction rules. If the current number of lanes is greater than the maximum number of lanes allowed, calculate the number of lanes that need to be closed.

[0169] The current number of lanes determined in step S1442 is compared with the maximum number of lanes allowed to pass as retrieved in step S1441. If the current number of lanes exceeds the maximum number of lanes allowed to pass, it indicates that there are too many lanes in the construction area, which does not meet the traffic restrictions during the construction phase. Some lanes need to be closed to ensure construction safety and traffic order. The number of lanes to be closed is the current number of lanes minus the maximum number of lanes allowed to pass.

[0170] Step S1444: Analyze the traffic flow distribution of each lane based on the traffic flow density characteristics, select the lane with the lowest traffic flow as the lane to be closed, and generate a lane closure suggestion that includes the lane number to be closed and the closure time period.

[0171] Traffic flow density characteristics reflect the distribution of traffic flow across lanes within the construction area. Analyzing these characteristics reveals the vehicle density of each lane. When lanes need to be closed, to minimize the impact on traffic flow, the lanes with the lowest traffic flow should be prioritized for closure. Specifically, traffic flow data for each lane should be statistically analyzed and compared to identify the lanes with the lowest traffic flow, which should then be designated as the lanes to be closed. Simultaneously, the time period for lane closure needs to be determined. This time period should be comprehensively determined based on factors such as the construction schedule and peak traffic hours to avoid closures during busy periods and reduce traffic congestion. Finally, a lane closure recommendation should be generated, clearly specifying the lane number to be closed and the specific closure time period.

[0172] Step S1445: When the current number of lanes is equal to the maximum number of lanes allowed to pass but the lane width parameter does not meet the minimum lane width requirement, query the lane change guidance area in the lane organization characteristics to determine the lane direction that can be used for one-way traffic.

[0173] If the current number of lanes is equal to the maximum allowed number of lanes, but the width parameters of some or all lanes do not meet the minimum lane width requirements, two-way traffic on these lanes may pose a safety hazard. In this case, it is necessary to consider adopting a one-way traffic system to ensure traffic safety. By querying the lane change guidance area in the lane organization characteristics, we can understand the layout and changes of lanes within the construction area, such as which areas have lane merging or separation guidance signs, and which directions of lanes are suitable for one-way traffic. Based on this information, we can determine the lane directions that can be used for one-way traffic, such as only allowing traffic from east to west or only allowing traffic from west to east.

[0174] Step S1446: Combine the vehicle speed characteristics in the real-time traffic flow data to determine the feasibility of one-way traffic direction and generate lane one-way traffic suggestions that include the one-way traffic direction and the implementation time period.

[0175] Real-time traffic flow data reveals vehicle speed characteristics that reflect the current traffic conditions in and around the construction area. After identifying potential one-way traffic directions, these speed characteristics are used to assess their feasibility. For example, observing vehicle speeds in that direction reveals stability and frequent decelerations or stops. Stable speeds indicate one-way traffic is feasible; otherwise, the direction needs reassessment or adjustment. The implementation timeframe for one-way traffic must also be determined, aligning with construction progress and avoiding peak traffic periods. Finally, lane one-way traffic recommendations are generated, clearly specifying the direction and implementation timeframe.

[0176] Step S1447: Determine the lane closure suggestion or the lane one-way traffic suggestion as the lane management suggestion within the construction area.

[0177] Based on the analysis and judgment in the preceding steps, if it is necessary to close some lanes, the generated lane closure suggestions will be used as lane management suggestions within the construction area; if it is necessary to adopt a one-way traffic system, the lane one-way traffic suggestions will be used as lane management suggestions within the construction area. These suggestions will provide specific guidance to traffic management departments to ensure traffic order and safe vehicle passage within the construction area.

[0178] Step S145: When the candidate congested road segment unit is located within the influence range of the construction area, the flow carrying capacity at the current speed is calculated based on the vehicle speed characteristics and traffic flow density characteristics in the traffic flow state characteristics. Combined with the size of the congestion influence range, a speed limit value suggestion is generated as a speed limit suggestion.

[0179] Step S1451: Extract vehicle speed characteristics within the construction area from traffic flow state characteristics. These vehicle speed characteristics include the average speed of different lanes per unit time, speed fluctuation range (i.e., the difference between the highest and lowest speeds), and speed change trend over time.

[0180] The vehicle speed characteristics within the construction area's impact zone can reflect the traffic flow status of that area. The information extracted from this includes: the average speed of each lane per unit time (e.g., per hour), which reflects the overall traffic efficiency of that lane; the speed fluctuation range, i.e., the difference between the highest and lowest speeds of that lane within the same time period, with a larger fluctuation range indicating a more unstable traffic flow; and the speed change trend over time, such as whether the speed gradually increases, gradually decreases, or remains stable, which helps predict future speed changes.

[0181] Step S1452: Extract the traffic flow density features within the affected area, including the total number of vehicles per unit length of road segment, the density distribution of different types of vehicles (such as small cars and large cars), and the trend of traffic flow density over time.

[0182] Traffic flow density characteristics reflect the degree of traffic congestion within the construction area. The extracted content includes: the total number of vehicles per unit length of road segment, which directly reflects the traffic density; the density distribution of different types of vehicles, i.e., the proportion of small cars, large cars, etc., per unit length of road segment, because different types of vehicles have different impacts on traffic flow (e.g., large cars are slower and can affect the passage of subsequent vehicles); and the trend of traffic flow density over time, to determine whether the density is increasing, decreasing, or remaining stable.

[0183] Step S1453: Construct a traffic flow carrying capacity calculation model. This model takes vehicle speed and traffic flow density as input parameters, and calculates the maximum number of vehicles that the road segment can carry per unit time at the current speed by analyzing the correlation between the two (such as the law that speed decreases as density increases). That is, the traffic flow carrying capacity.

[0184] The traffic flow carrying capacity calculation model is built upon the relationship between speed and density in traffic flow theory. Its core logic is: within a certain range, an increase in traffic flow density leads to a decrease in vehicle speed, and when the density reaches a certain threshold, the speed drops sharply, at which point the traffic flow carrying capacity of the road segment reaches its upper limit. The model takes the extracted vehicle speed and traffic flow density as inputs, and through built-in correlation analysis algorithms (such as speed-density curves fitted based on historical data), calculates the maximum number of vehicles that can pass smoothly on the road segment per unit time at the current speed, i.e., the current traffic flow carrying capacity.

[0185] Step S1454: Input the extracted vehicle speed features and traffic flow density features into the traffic flow carrying capacity calculation model to obtain the traffic flow carrying capacity value at the current speed.

[0186] The vehicle speed (e.g., average speed) and traffic density (e.g., total number of vehicles per unit length) within the construction area's impact zone are input into the traffic flow capacity calculation model. The model then performs calculations based on its built-in algorithm. For example, the model first finds the corresponding theoretical speed value on the speed-density curve based on the current traffic density and compares it with the actual extracted vehicle speed. If the actual speed is lower than the theoretical value, it indicates that the current traffic flow is already congested to some extent, and the traffic flow capacity is low; conversely, if the actual speed is higher, the capacity is higher, and the model ultimately outputs the traffic flow capacity value at the current speed.

[0187] Step S1455: Analyze the size of the congestion impact range in the congestion event prediction results. This size is described by the number of road segment units involved in the congestion, the total length of the road segment, and the number of lanes affected.

[0188] The extent of congestion impact is a crucial factor in setting speed limit recommendations. Analysis of relevant information extracted from congestion event predictions includes: the number of road segments involved (a higher number indicates a wider impact); the total length of these road segments (a longer total length indicates a larger impact); and the number of lanes affected (multiple lanes affected suggest potentially more severe congestion). These indicators are used to comprehensively assess the size of the congestion impact.

[0189] Step S1456: Based on the flow capacity value and the size of the congestion impact range, and in conjunction with the preset speed limit gradient table, determine the appropriate speed limit value.

[0190] The preset speed limit gradient table is developed based on road design speed, the degree of construction impact, and traffic flow conditions. The table contains suggested speed limit values ​​corresponding to different flow capacity and congestion impact ranges. For example, when the flow capacity is low and the congestion impact range is large, the corresponding speed limit value is low; when the flow capacity is high and the congestion impact range is small, the corresponding speed limit value is high. The calculated flow capacity value and the analyzed congestion impact range are matched with the speed limit gradient table to determine the appropriate speed limit value for the current situation.

[0191] Step S1457: Combine the peak hours of real-time traffic flow and the operation time of the construction area to determine the effective time period of the speed limit value, and generate a speed limit recommendation that includes the speed limit value, the effective time period, the applicable road section range, and the reason for the restriction.

[0192] To make speed limit recommendations more targeted, their effective time periods need to be determined. Referring to peak traffic flow times in real-time, lower speed limits should be implemented during peak hours to control traffic volume and prevent further congestion; limits can be appropriately increased during off-peak hours. Simultaneously, the work schedule of the construction area should be considered to ensure that speed limits are effective while construction is underway. Finally, the determined speed limit values, effective time periods, applicable road segments (i.e., specific road segments within the construction area's impact range), and reasons for the restrictions (such as alleviating congestion and increasing traffic capacity) should be compiled into a speed limit recommendation.

[0193] Step S146: When the candidate congested road segment unit is neither located within the construction area nor within the influence range of the construction area, analyze the road segment alignment on vehicle speed based on the curvature segmentation characteristics and slope transition characteristics, and generate lane speed limit differentiation suggestions as speed limit suggestions in combination with traffic flow state characteristics.

[0194] In this embodiment, curvature information of the candidate congested road segment unit is extracted from the curvature segmentation features. This includes the distribution of curvature within the region, identifying which parts belong to gently curvature segments and which belong to segments with abrupt curvature changes, as well as the specific location range and continuous length of these segments. Simultaneously, combined with slope transition features, slope change trend information within the region is extracted to determine whether it is an uphill transition segment, a downhill transition segment, or a relatively stable slope segment. The location range and duration of slope changes for these segments are also recorded.

[0195] Then, based on the extracted curvature and slope information, the limiting factors of road alignment on vehicle speed are analyzed. For sections with abrupt changes in curvature, vehicles need to decelerate to ensure driving safety due to the sudden change in road direction, thus the speed limit is more significant; while for sections with gentle curvature, the speed limit is relatively small. In uphill transition sections, vehicles need to overcome gravity, which affects their speed, especially in areas with large slope changes, where the speed may decrease significantly; although vehicles may tend to accelerate due to gravity in downhill transition sections, there are usually speed limits for safety reasons, especially on long downhill sections where speed limits are more stringent.

[0196] Next, by combining vehicle speed and traffic density characteristics from the traffic flow status features, the current traffic flow operation status of the area is analyzed. By observing the differences in vehicle speeds across different lanes and the distribution of traffic density, it is determined whether the current traffic flow is in a congested state or whether there is a trend towards congestion. For example, when the traffic density is high and vehicle speeds are generally low, it indicates that the current traffic flow pressure is high, and stricter speed limits are needed to prevent further congestion.

[0197] Finally, considering the impact of road alignment on speed limits and current traffic flow conditions, differentiated speed limit recommendations are formulated for each lane. For sections with abrupt changes in curvature and significant gradient changes, relatively lower speed limits should be set; for sections with gentle curvature and stable gradients, relatively higher speed limits can be set, taking into account traffic flow conditions. Furthermore, based on the traffic flow characteristics of different lanes—for example, lanes with high traffic volume and slow speeds can have their speed limits appropriately reduced, while lanes with low traffic volume and smooth traffic flow can maintain or appropriately increase their speed limits—this aims to achieve a balanced distribution of traffic flow across lanes and improve the overall traffic efficiency of the road segment.

[0198] Step S147: Integrate the lane management suggestions and the speed limit suggestions to generate a control strategy adaptation result that includes the effective time range of the suggestions and the applicable road segment range.

[0199] In this embodiment, the applicable conditions and effective duration of lane management recommendations and speed limit recommendations are first determined. The effective time range of lane management recommendations needs to be determined in conjunction with factors such as construction schedule and peak traffic periods. For example, during periods of heavy construction, lane closures or one-way traffic recommendations may need to be maintained for a longer period; while at night when traffic flow is low, some management measures may be appropriately relaxed. The effective time of speed limit recommendations needs to take into account the changing patterns of traffic flow. For example, during morning and evening rush hours, speed limits may need to take effect earlier and have a longer duration.

[0200] Next, the specific road segment range to which each recommendation applies should be clearly defined to ensure that the recommendations match the actual conditions of the road segment. The applicable road segment range for lane management recommendations should be precise to specific road segments within lane merging guidance areas, lane separation guidance areas, or construction impact areas; the applicable road segment range for speed limit recommendations should take into account the characteristics of different road segments reflected by curvature segmentation features, slope transition features, and traffic flow status features. For example, in the overlapping areas of curvature abrupt change sections and uphill transition sections, the applicable range of speed limit recommendations should accurately cover the overlapping area.

[0201] Next, a coordination check is conducted on the lane management recommendations and speed limit recommendations to ensure there are no conflicts between them. For example, lane closure recommendations should not conflict with lanes requiring higher speeds in speed limit recommendations, to avoid traffic chaos caused by contradictions between recommendations. If conflicts are found, the recommendations need to be readjusted to ensure the consistency and effectiveness of the control strategy.

[0202] Finally, the coordinated and confirmed lane management recommendations, speed limit recommendations, and their respective effective time ranges and applicable road sections are integrated to form a complete control strategy adaptation result.

[0203] Step S150: Output the congestion event prediction result, the lane management suggestion, and the speed limit suggestion to the traffic control terminal.

[0204] In this embodiment, congestion event prediction results, lane management suggestions, and speed limit suggestions are converted into a data format that the traffic control terminal can recognize and display. Different traffic control terminals may support different data formats, so format conversion is required according to the specific requirements of the terminal to ensure that the data can be correctly received and parsed. For example, text-based suggestions are converted into structured data that the terminal can recognize, such as JSON or XML format, which includes information such as the type, content, effective time, and scope of application of the suggestion.

[0205] Next, establish a communication connection with the traffic control terminal to ensure the stability and security of data transmission. This can be achieved using wired or wireless communication methods, such as Ethernet or 4G / 5G networks. During transmission, encryption technology should be employed to encrypt the data, preventing tampering or leakage and ensuring data integrity and security.

[0206] Therefore, the converted congestion event prediction results, lane management suggestions, and speed limit suggestions are sent to the traffic control terminal. During the transmission process, the data transmission status is monitored in real time. If a transmission interruption or error is detected, retransmission is initiated promptly to ensure successful data delivery. After receiving the data, the traffic control terminal analyzes and processes it, and displays it in an intuitive way, such as marking the range and probability of congestion event predictions on a map, and displaying lane management suggestions and speed limit suggestions in text or icon form. This facilitates traffic management personnel in understanding relevant information and taking appropriate control measures in a timely manner.

[0207] Figure 2 The illustration shows exemplary hardware and software components of a machine learning-based road congestion event prediction system 100 for implementing the ideas of this application, provided in some embodiments of this application. For example, a processor 120 can be used in the machine learning-based road congestion event prediction system 100 for performing the functions described in this application.

[0208] The machine learning-based congestion event prediction system 100 for road reconstruction and expansion sections can be a general-purpose server or a special-purpose server; both can be used to implement the machine learning-based congestion event prediction method for road reconstruction and expansion sections of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0209] For example, the machine learning-based road congestion prediction system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the machine learning-based road congestion prediction system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The machine learning-based road congestion prediction system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0210] For ease of explanation, only one processor is described in the machine learning-based road congestion event prediction system 100. However, it should be noted that the machine learning-based road congestion event prediction system 100 of this application may also include multiple processors. Therefore, the steps performed by one processor described in this application may also be performed jointly by multiple processors or individually. For example, if the processor of the machine learning-based road congestion event prediction system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0211] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned machine learning-based method for predicting congestion events on reconstructed and expanded road sections is implemented.

[0212] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A machine learning-based method for predicting congestion events on reconstructed and expanded road sections, characterized in that, The method includes: Collect road alignment data, traffic organization status data, and real-time traffic flow data for the road sections undergoing reconstruction and expansion; The road alignment index data, traffic organization status data, and real-time traffic flow data are subjected to feature association processing to generate a set of associated features for the reconstructed and expanded road sections. The pre-trained congestion event prediction model is invoked to perform congestion trend analysis on the associated feature set of the reconstructed and expanded road section, generating congestion event prediction results that include the probability of congestion occurrence and the scope of congestion impact. Based on the congestion event prediction results and the current construction progress information of the road section to be reconstructed and expanded, the control strategy is adapted and processed to generate lane management suggestions and speed limit suggestions; The congestion event prediction results, lane management suggestions, and speed limit suggestions are output to the traffic control terminal. The process involves performing feature association processing on the road alignment index data, traffic organization status data, and real-time traffic flow data to generate a set of associated features for the reconstructed and expanded road sections, including: Extract the curvature change characteristics of road segments from the road alignment index data, calculate the curvature change rate of continuous road segment units, and divide the curvature smooth segments and curvature abrupt segments based on the curvature change rate to obtain curvature segmentation characteristics; Extract the slope distribution characteristics from the road alignment index data, analyze the slope value change trend of adjacent road segment units, determine the uphill transition section and downhill transition section, and obtain the slope transition characteristics; Extract lane number adjustment information from the traffic organization status data, identify areas where the number of lanes decreases and areas where the number of lanes increases, and combine this with the temporary traffic sign placement information from the traffic organization status data to determine lane change guidance areas and obtain lane organization characteristics. Extract vehicle speed features from the real-time traffic flow data, calculate speed differences between different lanes, and combine them with traffic density features from the real-time traffic flow data to analyze the correlation between speed and density, thereby obtaining traffic flow state features. The curvature segmentation feature, the slope transition feature, the lane organization feature, and the traffic flow state feature are input into the feature association network. The mutual influence coefficient between each feature is calculated through the weighted connection of the network layer of the feature association network, and a set of associated features of the reconstructed and expanded road section containing feature association weights is generated. The pre-trained congestion event prediction model is invoked to perform congestion trend analysis on the associated feature set of the reconstructed and expanded road section, generating congestion event prediction results that include the probability of congestion occurrence and the scope of congestion impact, including: The set of associated features of the reconstructed and expanded road section is input into the feature adaptation layer of the congestion event prediction model. The feature dimensions are standardized to match the preset input dimensions of the model, thus obtaining standardized associated features. The standardized associated features are extracted collaboratively by the road segment feature extraction layer of the congestion event prediction model, which extracts local and global features. The local features include the curvature and slope combination features of the road segment unit, and the global features include the lane organization and traffic flow distribution features of the entire road segment. The local features and the global features are input into the spatiotemporal correlation layer of the congestion event prediction model. A feature correlation matrix is ​​constructed based on the road segment location relationship and time series continuity. The feature transfer coefficients of different road segment units in the time dimension are calculated. The congestion probability calculation layer of the congestion event prediction model performs probability mapping processing on the feature correlation matrix, and generates the congestion occurrence probability of each road segment unit by combining the probability distribution pattern of historical congestion data. Based on the congestion occurrence probability of each road segment unit and the transmission coefficient in the feature correlation matrix, the path of congestion spreading from high-probability road segments to adjacent road segments is predicted, and the scope of congestion impact is determined. By integrating the congestion occurrence probability of each road segment unit and the congestion impact range, a congestion event prediction result containing a probability distribution sequence and range boundary coordinates is generated; The process involves adapting control strategies based on the congestion event prediction results and the current construction progress information of the road section being upgraded, generating lane management suggestions and speed limit suggestions, including: The congestion occurrence probability distribution sequence in the congestion event prediction results is analyzed to determine the location and corresponding probability value of candidate congested road segment units, wherein the candidate congested road segment unit is a road segment unit whose congestion occurrence probability is greater than a set probability threshold; Obtain the current construction progress information of the road section to be reconstructed and expanded, and extract the location range and construction stage identifier of the construction area. The construction stage identifier includes the roadbed construction stage, pavement construction stage and ancillary facility construction stage. The location of the candidate congested road segment unit is spatially superimposed with the location range of the construction area to determine whether the candidate congested road segment unit is located within the construction area or within the influence range of the construction area. When a candidate congested road segment unit is located within a construction area, the corresponding lane traffic restriction rules are queried according to the construction stage identifier. Combined with the lane number adjustment information in the lane organization characteristics, lane closure suggestions or lane one-way traffic suggestions are generated as lane management suggestions. When a candidate congested road segment unit is located within the influence range of the construction area, the flow carrying capacity at the current speed is calculated based on the vehicle speed characteristics and traffic flow density characteristics in the traffic flow state characteristics. Combined with the size of the congestion influence range, a speed limit value suggestion is generated as a speed limit suggestion. When a candidate congested road segment unit is neither located within the construction area nor within the influence range of the construction area, the road segment alignment is analyzed to limit vehicle speed based on the curvature segmentation characteristics and slope transition characteristics. Combined with traffic flow state characteristics, lane speed limit differentiation suggestions are generated as speed limit suggestions. By integrating the lane management recommendations and the speed limit recommendations, a control strategy adaptation result is generated, which includes the effective time range of the recommendations and the applicable road segment range.

2. The machine learning-based method for predicting congestion events on reconstructed and expanded road sections according to claim 1, characterized in that, The process involves extracting road segment curvature variation features from the road alignment index data, calculating the curvature variation rate of continuous road segment units, and dividing the road into smooth curvature segments and abrupt curvature segments based on the curvature variation rate to obtain curvature segmentation features, including: The road alignment index data is divided into multiple continuous road segment units according to a preset road segment length, and each road segment unit corresponds to a curvature value; Calculate the curvature value difference between adjacent road segment units, and use the ratio of the curvature value difference to the length of the road segment unit as the curvature change rate; When the rate of curvature change of multiple consecutive road segment units is within the range of the rate of curvature change threshold, the corresponding area is marked as a smooth curvature segment. When the rate of curvature change of a single road segment exceeds the threshold range, the road segment is marked as a curvature abrupt change segment. The marking information of the gently curvature segments and the abruptly curvature segments is integrated according to the positional order of the road segment units to generate curvature segmentation features that include the start and end positions of each segment.

3. The machine learning-based method for predicting congestion events on reconstructed and expanded road sections according to claim 1, characterized in that, The process involves extracting lane number adjustment information from the traffic organization status data, identifying areas where the number of lanes decreases and areas where the number of lanes increases, and combining this with the location information of temporary traffic signs in the traffic organization status data to determine lane change guidance areas, thereby obtaining lane organization characteristics, including: The lane number adjustment information in the traffic organization status data is analyzed, the lane number values ​​at different road segment locations are extracted, and a lane number distribution sequence is generated; By comparing the number of lanes in adjacent road segments, if the number of lanes in the later road segment is less than the number of lanes in the earlier road segment, the later road segment is marked as a lane reduction area; and if the number of lanes in the later road segment is greater than the number of lanes in the earlier road segment, the later road segment is marked as a lane increase area. Extract the location information of the temporary traffic signs and determine the layout coordinates of the lane change signs, which include lane merging signs and lane separating signs; The location of the lane reduction area is matched with the layout coordinates of the lane merging sign, and the area with a successful match is determined as the lane merging guidance area. The location of the area where the number of lanes increases is matched with the layout coordinates of the lane separation signs, and the area where the match is successful is determined as the lane separation guidance area. By integrating the location information of the lane merging guidance area and the lane separating guidance area, a lane organization feature containing the guidance area type and range coordinates is generated.

4. The machine learning-based method for predicting congestion events on reconstructed and expanded road sections according to claim 1, characterized in that, The step of collaboratively extracting local and global features from the standardized correlated features through the road segment feature extraction layer of the congestion event prediction model includes: The standardized associated features are divided into multiple local feature blocks according to road segment units, and each local feature block corresponds to the feature data of a road segment unit. For each local feature block, the curvature segmentation feature and slope transition feature are combined. The product of the curvature value and the slope value is calculated as the combination coefficient to generate the curvature and slope combination feature of the road segment unit. The curvature and slope combination features of all local feature blocks are summarized to obtain a local feature set containing local information of all road segment units; Statistical analysis is performed on the lane organization features and traffic flow status features in the standardized correlation features across the entire road segment. The average number of lanes and average traffic flow density of the entire road segment are calculated to generate the lane organization distribution features and traffic flow distribution features of the entire road segment. The lane organization distribution features and the traffic flow distribution features are concatenated to obtain a global feature that contains overall information about the entire road segment.

5. The machine learning-based method for predicting congestion events on reconstructed and expanded road sections according to claim 1, characterized in that, The step of predicting the path of congestion spreading from high-probability road segments to adjacent road segments based on the congestion occurrence probability and the transmission coefficient in the feature correlation matrix of each road segment unit, and determining the congestion impact range, includes: From the congestion occurrence probabilities of each road segment unit, road segment units with probability values ​​exceeding a preset probability threshold are selected and marked as initial congested road segment units; Extract the row vector of the transfer coefficients corresponding to the initial congested road segment unit from the feature correlation matrix. The row vector of the transfer coefficients contains the feature transfer coefficients between the initial congested road segment unit and all other road segment units. The other road segment units are sorted according to the numerical value of the transmission coefficient row vector, and the top N road segment units with the largest transmission coefficients are selected as potential diffusion road segment units, where N is a positive integer. Calculate the road segment distance between the initial congested road segment unit and each potential diffusion road segment unit, and generate a diffusion probability score by combining the transmission coefficient and the road segment distance; Potentially congested road segment units whose diffusion probability scores exceed a preset scoring threshold are marked as secondary congested road segment units; Repeat the steps of extracting the row vector of the transfer coefficient corresponding to the secondary congested road segment unit, screening potential diffusion road segment units, and calculating the diffusion probability score until no new secondary congested road segment units are generated. By integrating the location information of the initial congested road segment unit and all secondary congested road segment units, the boundary road segment units of the congestion impact range are determined, and range description information containing all congestion-related road segment units is generated.

6. The machine learning-based method for predicting congestion events on reconstructed and expanded road sections according to claim 1, characterized in that, When a candidate congested road segment unit is located within a construction area, the corresponding lane traffic restriction rules are queried based on the construction phase identifier. Combined with lane number adjustment information in the lane organization characteristics, lane closure suggestions or lane one-way traffic suggestions are generated as lane management suggestions, including: Based on the construction stage identifier, the corresponding lane traffic restriction rules are queried from the preset construction stage rule base. The lane traffic restriction rules include the maximum number of lanes allowed to pass and the minimum lane width requirements. Extract the lane number adjustment information from the lane organization features to determine the current number of lanes and the width parameters of each lane within the construction area; The current number of lanes is compared with the maximum number of lanes allowed in the lane traffic restriction rules. When the current number of lanes is greater than the maximum number of lanes allowed, the number of lanes that need to be closed is calculated. Based on the traffic flow density characteristics, the traffic flow distribution of each lane is analyzed, the lane with the lowest traffic flow is selected as the lane to be closed, and a lane closure suggestion containing the lane number to be closed and the closure time period is generated. When the current number of lanes equals the maximum number of lanes allowed to pass, but the lane width parameter does not meet the minimum lane width requirement, query the lane change guidance area in the lane organization characteristics to determine the lane direction that can be used for one-way traffic. By combining vehicle speed characteristics in real-time traffic flow data, the feasibility of one-way traffic direction is determined, and lane one-way traffic suggestions containing one-way traffic direction and implementation time period are generated; The lane closure suggestion or the lane one-way traffic suggestion is identified as the lane management suggestion within the construction area.

7. A machine learning-based system for predicting congestion events on reconstructed and expanded road sections, characterized in that, The system includes a processor and a memory, the memory being connected to the processor. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the machine learning-based method for predicting congestion events on reconstructed and expanded road sections as described in any one of claims 1-6.

Images