Cooperative dynamic speed limit control method based on construction area risk conduction theory

Abstract

The invention discloses a collaborative dynamic speed limit control method based on a construction area risk conduction theory, and belongs to the field of highway reconstruction and extension project construction safety risk prevention and control speed limit control. In order to solve the problems of static response lagging and space cooperation deficiency in traditional construction area speed limit control, the traffic flow safety is obtained according to the average speed, the traffic density, the blocking density and the load-to-vehicle ratio; the construction activity index is used as construction safety; obtaining the environmental safety according to the visibility and the road surface slippery degree; obtaining a comprehensive risk value based on the three security to determine a risk level; on the basis of the comprehensive risk value of the construction area, conduction comprehensive risk values at different distances from the construction area are obtained, corresponding risk levels are determined, the speed limit value with the highest speed limit value probability in the risk levels corresponding to the nearest distance serves as the speed limit value of the construction area, and then the speed limit values at different distances are determined in sequence; and determining the risk conduction value of each position and the speed limit value of the corresponding variable speed limit board in combination with the constraint, and further giving a dynamic speed limit control scheme.

Description

A Collaborative Dynamic Speed ​​Limit Control Method Based on the Theory of Risk Transmission in Construction Zones Technical Field

[0001] This invention belongs to the field of speed limit control for construction safety risk prevention and control in highway reconstruction and expansion projects, and specifically relates to a collaborative dynamic speed limit control method. Background Technology

[0002] In recent years, the scale of highway reconstruction and expansion has continued to expand, leading to a significant increase in accident risks in construction areas due to lane reduction and more complex traffic organization. Under the combined effects of various weather conditions such as rain, snow, and fog, and dynamic lane-changing demands, vehicle safety and traffic efficiency in construction areas face serious challenges. To ensure traffic safety and improve vehicle traffic efficiency in construction areas, dynamic speed limit control has become an important direction for research on safety risk prevention and control in highway reconstruction and expansion projects.

[0003] Current speed limit systems face two major drawbacks: static response lag and risk gradient discontinuity. Static thresholds are decoupled from dynamic risks. Fixed speed limits cannot respond in real-time to sudden weather changes and traffic flow variations. When visibility drops sharply or traffic volume surges, the preset value is maintained, significantly weakening safety protection capabilities. Spatial coordination mechanisms are lacking. The absence of a mechanism for sensing the upstream and downstream conditions of the entire construction area and linking speed limit sections results in discrete control, failing to achieve smooth speed transitions through step-by-step speed changes, and inducing traffic speed oscillations and conflicts due to discrepancies between speed limit commands and actual operating conditions. Therefore, there is an urgent need to develop a dynamic speed limit control architecture that integrates real-time risk assessment and multi-machine collaboration in highway reconstruction and expansion construction areas. Summary of the Invention

[0004] This invention aims to solve the problems of static response lag and lack of spatial coordination in traditional construction zone speed limit control.

[0005] A collaborative dynamic speed limit control method based on the risk transmission theory of construction zones includes:

[0006] First, obtain the traffic flow, average speed, and road occupancy data for the entire affected section of the highway construction; then, obtain the road surface slippage of the affected section; and finally, obtain the construction activity index of the affected section.

[0007] Then, based on the acquired data, speed limit decisions are made, including:

[0008] Traffic density is determined based on traffic flow and construction zone length, and congestion density is determined based on historical traffic density; traffic flow safety is obtained based on average speed, traffic density, congestion density, and vehicle load ratio. The construction activity index is used as a measure of construction safety. Environmental safety is determined based on visibility and road surface slipperiness. ;based on , , Obtain the overall risk value Risk level is determined based on Risk;

[0009] Determine the probabilistic relationship between risk level and speed limit:

[0010] Based on the speed limit standards of the construction section, the speed variation space S of the variable speed limit sign is determined; based on the historical average speed under different risk levels, and combined with the speed area corresponding to the speed variation space S, the probability of the speed limit value corresponding to each speed area under different risk levels is obtained, and a risk level-speed limit value probability correlation table is obtained.

[0011] The risk transmission direction is determined by the distance from the construction area to the nearest to the farthest point, and the comprehensive risk value of the construction area is then calculated. As ,according to Calculate the comprehensive risk value of transmission at a distance d from the construction area, where, The attenuation coefficient;

[0012] Each variable speed limit information board (VMS) is designated as VMS 1, VMS 2, ..., VMS q, ... according to its distance from the construction area, from farthest to closest.

[0013] Based on the risk level corresponding to the comprehensive risk value of transmission at different distances from the construction area, for VMS 1, the speed limit value with the highest probability of speed limit value in the corresponding risk level is taken as the speed limit value of VMS 1. Then, based on the speed limit value order in the risk level-speed limit value probability association table, the speed limit value corresponding to VMS q is determined sequentially. Then, the gradient adjustment constraint VMS 1≥VMS 2≥……≥VMS q≥…… is used as one of the constraints. Combined with the constraints, the risk transmission value at each location and the corresponding variable speed limit value are determined, and then a dynamic speed limit control scheme is given.

[0014] Furthermore, the process of obtaining traffic flow, average speed, and road occupancy data for the entire affected section of the highway construction includes:

[0015] Traffic data, including traffic flow, average speed, road occupancy, and heavy vehicle ratio, is collected at each of the at least two traffic detection devices set up in the affected section of the highway construction. Then, based on the traffic data from two of the at least two traffic detection devices, the overall traffic data of the affected section of the highway construction is obtained through a neural network model.

[0016] Furthermore, the processing procedure of the neural network model is as follows:

[0017] The output of the input layer is fed into the bidirectional LSTM layer for processing. The bidirectional LSTM layer adopts a three-layer stacked structure. After the output of each layer is regularized by Dropout, the forward and backward hidden states are concatenated in the feature dimension to form a higher-order feature representation.

[0018] The output of the bidirectional LSTM layer is fed into a multi-head attention mechanism to obtain the feature matrix of the multi-head attention output; then a fully connected layer is used for prediction, and the output dimension corresponds to the three prediction targets [flow rate, average speed, road occupancy].

[0019] Furthermore, in the process of obtaining overall traffic data for the road construction-affected section through a neural network model, a Bayesian correction module is used to correct the output of the neural network model. The specific process includes:

[0020] During the inference phase, all Dropout layers in the neural network model are kept active, and D forward propagations are performed for the same input to obtain the prediction set. Each element corresponds to a three-dimensional prediction target; the prediction mean is calculated based on this set. and prediction variance ;

[0021] Then, adaptive corrections are performed using historical prediction errors, including:

[0022] Maintain a sliding window of length D to store the most recent prediction error sequence and obtain the exponential moving average of the historical prediction error sequences. ;

[0023] For the current predicted value Make corrections, the correction formula is as follows ,in, To adjust the weights, This is an adjustable parameter that controls the effect of variance on the correction strength.

[0024] Furthermore, the process of obtaining the road surface slippage of the affected section of the highway construction includes:

[0025] First, the basic friction coefficient is calculated based on temperature and rainfall. ,in: The maximum coefficient of friction under ideal conditions. The attenuation coefficient is... R represents rainfall; T represents temperature.

[0026] Then humidity correction is introduced. Where: H is relative humidity, Humidity influence coefficient;

[0027] The final standardized road surface slip resistance index is obtained. .

[0028] Furthermore, the process of obtaining the construction activity index of the affected sections of the highway construction includes:

[0029] (1) Obtaining population density indicators Mechanical strength index Job type index W;

[0030] (2) Standardized three-dimensional data Principal component analysis was performed to obtain the principal component contribution rates. Then calculate the construction activity index. , The score is the score of the i-th principal component.

[0031] Furthermore, the acquisition of personnel density indicators Mechanical strength index The process of the job type indicator W includes:

[0032] Personnel density index ,in, A represents the number of construction workers, and A represents the area of ​​the construction area.

[0033] Mechanical strength index ,in, L represents the number of mechanical equipment, and L represents the activity level.

[0034] The job type index W = {1.0, maintenance job; 2.0, expansion job; 3.0, new job}.

[0035] Furthermore, traffic flow safety is obtained based on average speed, traffic density, congestion density, and vehicle load ratio. The process is as follows:

[0036] Traffic flow safety = (Speed, density, congestion density, road occupancy, heavy vehicle ratio), traffic flow safety indicators as follows:

[0037] (18)

[0038] Where v is the current average speed. k is the free-flow velocity; k is the traffic density. Blocking density; denoted as road occupancy rate; r represents the proportion of heavy vehicles.

[0039] Furthermore, the process of determining environmental safety based on visibility and road surface slipperiness is as follows:

[0040] Environmental safety = (Visibility, Slipperiness), Environmental Safety Indicators as follows:

[0041] (19)

[0042] in, Visibility; The degree of slipperiness of the road surface; , These are the fitting parameters.

[0043] Furthermore, in proposing a dynamic speed limit control scheme, it is necessary to consider the risk of accidents to arrive at the final dynamic speed limit control scheme; the aforementioned risk of accidents... Where TTC is the distance-to-collision time.

[0044] The beneficial effects of this invention are as follows:

[0045] This invention addresses the core issues of static response lag and lack of spatial coordination in traditional construction zone speed limit control by establishing risk transmission theory and a hidden Markov chain decision-making mechanism. First, to address the inability of fixed speed limits to respond to dynamic risks in real time, it integrates LSTM traffic flow prediction, meteorological data mapping, and construction activity index calculation to automatically trigger adaptive speed limit adjustments under adverse weather conditions or high construction intensity scenarios, accurately quantifying the level of sudden risks. Second, to address the operational disconnect caused by the independent operation of discrete control units, it achieves coordinated control of multi-level VMS (Vehicle Management Systems) upstream and downstream of the construction zone based on risk transmission theory. Combining distance decay functions and traffic flow propagation delay effects, it significantly distinguishes the differences in risk evolution at different spatial locations, effectively eliminating potential traffic conflicts caused by sudden speed changes. By deeply integrating the above dynamic risk perception and spatial coordinated transmission mechanism, and within the framework of hidden Markov chain decision-making, combined with the multi-dimensional coupled risk characteristics of traffic, weather, and construction, an active protection system with spatiotemporal adaptive characteristics is constructed, significantly improving the systematicness and comprehensiveness of risk prevention and control. This invention is particularly suitable for long-distance reconstruction and expansion construction areas of highways. Under adverse conditions such as low visibility, poor road surface and complex construction activities, it can achieve rapid risk response and smooth traffic flow, providing all-weather protection for traffic safety in the construction area. Attached Figure Description

[0046] Figure 1 shows the overall system architecture.

[0047] Figure 2 is a schematic diagram of the VMS layout and construction area location.

[0048] Figure 3 is a flowchart of the LSTM traffic flow prediction module.

[0049] Figure 4 is a flowchart of the meteorological data mapping module.

[0050] Figure 5 is a flowchart of the construction activity index calculation module.

[0051] Figure 6 is a flowchart of the risk transmission model.

[0052] Figure 7 shows the evaluation and analysis method.

[0053] Figure 8 shows the comparison results of the test data schemes.

[0054] Figure 9 is a schematic diagram of the application scenario. Detailed Implementation

[0055] The purpose of this invention is to address the two core defects in existing speed limit control in construction areas: static response lag and lack of spatial coordination mechanism. Fixed speed limit values ​​cannot respond in real time to dynamic weather changes and traffic flow fluctuations, and discrete control units cannot achieve coordinated regulation of upstream and downstream speed gradients. Therefore, this invention proposes a dynamic speed limit control method for highway construction areas that integrates dynamic risk assessment and multi-machine coordination.

[0056] Specific Implementation Method 1: This implementation method is explained with reference to Figure 1.

[0057] The collaborative dynamic speed limit control method based on the risk transmission theory of construction areas described in this embodiment includes the following steps:

[0058] Step 1: Detection equipment deployed within the construction impact zone (0-1600m) of the secondary highway includes upstream traffic detection equipment 1 (0m location), traffic detection equipment 2 (300m location), a small weather station (800m location), and traffic detection equipment 3 (1600m location). The traffic detection equipment consists of video surveillance and microwave radar to acquire real-time traffic flow data, meteorological data, and construction activity data. Considering the requirements for traffic safety and efficiency in the construction area, as well as avoiding data gaps and outliers, the detection equipment was installed at key locations upstream and downstream of the construction section of the secondary highway to obtain optimal monitoring coverage. The data obtained so far is shown in Table 1.

[0059] Table 1 Data collected

[0060]

[0061] Meanwhile, four variable message signs, spaced 200m apart, were deployed at 800m-1400m in the construction-affected area to provide the hardware foundation for subsequent coordinated dynamic speed limiting. The locations of the detection equipment, variable message signs, construction-affected area, and actual construction area are shown in Figure 2.

[0062] The specific data collection methods involved are as follows:

[0063] (1) Traffic flow data collection:

[0064] Traffic flow data is collected through traffic detection equipment 1 and traffic detection equipment 2, including four key indicators: flow rate, speed, occupancy rate, and heavy vehicle ratio. Flow rate data is collected every minute to calculate the total traffic volume in the previous minute; speed data is used to calculate the average speed in the previous minute; occupancy rate data is used to calculate the road occupancy rate in the previous minute; and heavy vehicle ratio data is used to calculate the proportion of heavy vehicles in the previous minute.

[0065] (2) Meteorological data collection:

[0066] Meteorological data is collected through weather stations and includes five key indicators: temperature, humidity, rainfall, wind speed, and air pressure. All meteorological data is collected every minute to provide the basis for subsequent calculations of road surface slippage.

[0067] (3) Construction data collection:

[0068] Construction data is collected through monitoring equipment in the construction area, including four key indicators: number of personnel, number of mechanical equipment, activity level, and construction type. The number of personnel is the total number of workers in the construction area; the number of equipment is the total number of mechanical equipment in the construction area; the activity level is divided into 1-5 levels according to the equipment's operating status; and the construction type is divided into three categories: maintenance, expansion, and new construction.

[0069] Based on the data collected in step 1, the data processing and prediction process from step 2 to step 4 needs to be executed. Each step can be executed in parallel, and the execution order does not affect the results of subsequent steps.

[0070] Step 2: Traffic Flow Data Processing and Prediction. Using a combination of LSTM-based short-term prediction and Bayesian correction, traffic flow data collected from two upstream detection devices is used to predict traffic flow data in the construction area, providing a theoretical basis for subsequent risk transmission and speed limit decisions.

[0071] As shown in Figure 3, the specific data processing flow involved includes:

[0072] (1) Data Preprocessing. The system receives raw traffic flow data streams in real time from upstream detection device 1 (located 0m upstream of the construction area) and device 2 (located 300m upstream of the construction area). The data streams contain four key indicators: flow rate (unit: veh / h), average speed (unit: km / h), road occupancy (percentage), and heavy vehicle proportion (percentage). To address the time discontinuity problem in the raw data, a hybrid interpolation method combining forward filling and backward filling is used to handle missing values, ensuring the integrity of the time series. The outlier detection module is based on the improved interquartile range algorithm, dynamically calculating Q1 (first quartile) and Q3 (third quartile) for each feature dimension, setting the effective data range as [Q1-1.5IQR, Q3+1.5IQR], and performing smooth replacement processing on data points outside this range.

[0073] After data cleaning, the Min-Max normalization algorithm is used to linearly map each feature value to the [0,1] interval. The algorithm calculation formula is shown in equation (1).

[0074] (1)

[0075] in, , These represent the historical minimum and maximum values ​​for each feature dimension.

[0076] (2) Temporal Data Construction. The heterogeneous data streams from the two detectors are spatiotemporally aligned and feature fused. Specifically, using 5 minutes as the basic time unit, historical data from 8 consecutive time slices (i.e., 40 minutes) are extracted to form a three-dimensional input tensor [N, T, F]. Here, N represents the training batch size, which is dynamically adjusted according to the hardware memory; T is fixed at 8 time steps, corresponding to a 40-minute observation window; and F integrates the 4 feature indicators from device 1 and device 2 to form an 8-dimensional feature vector. This tensor construction process is implemented using a sliding window technique, with the window step size set to 1 time unit (5 minutes) to ensure the continuity of temporal prediction.

[0077] (3) Model structure design. The core components of the network architecture adopt a modular design concept, and the input and output dimensions and data flow paths of each level are strictly defined.

[0078] After receiving the [N,8,8]-dimensional preprocessed tensor, the input layer first performs normalization on the eight-dimensional features using the Layer Normalization module. This technique differs from traditional batch normalization; its normalization calculation is performed along the feature dimensions, and the specific calculation formula is shown in equation (2):

[0079] (2)

[0080] in, For the i-th original value of a certain feature, ; ; and The parameters are learnable; F=8 represents the feature dimension. This design effectively addresses the statistical bias problem during mini-batch training and improves model training stability.

[0081] The bidirectional LSTM layer employs a three-layer stacked structure, with each layer containing 80 hidden units. The forward and backward LSTM units process the temporal data independently. The forward LSTM processes time steps sequentially from t=1 to 8, capturing historical information dependencies; the backward LSTM processes data in reverse order from t=8 to 1, extracting potential future patterns. Each LSTM unit internally includes an input gate. Forgotten Gate Output gate and candidate memory cells After each layer's output is regularized with a probability of 0.2 using Dropout, the forward and backward hidden states are concatenated along the feature dimension to form a [N,8,160]-dimensional high-order feature representation.

[0082] The multi-head attention mechanism uses eight independent computation heads, each projecting 160-dimensional input features onto a 20-dimensional query (Q), key (K), and value (V) subspace. Attention weights are calculated using a scaled dot product, and the formula is given in equation (3):

[0083] (3)

[0084] in, =20 represents the subspace dimension. The outputs of each head are concatenated along the feature dimension and then processed by a trainable weight matrix. A linear transformation is performed to maintain the output dimension consistent with the input, ultimately yielding an [N, 8, 160]-dimensional feature matrix for the multi-head attention output. This design enables the model to capture temporal dependency patterns in different subspaces in parallel.

[0085] A fully connected layer is used for prediction. First, the [N, 8, 160]-dimensional feature matrix output by the multi-head attention is flattened into a [N, 1280]-dimensional vector. The parameter matrix of the first fully connected layer has a dimension of [1280, 128]. A BatchNorm layer and a ReLU activation function are added, as shown in equation (4):

[0086] (4)

[0087] in, This is the weight matrix; This is the bias vector.

[0088] The second fully connected layer is directly mapped to the output space, and the parameter matrix has dimensions [128,3], as shown in equation (5):

[0089] (5)

[0090] in, This is the weight matrix; The bias vector corresponds to the three prediction targets: [flow rate, average speed, and road occupancy].

[0091] The inputs include the traffic flow, average speed, road occupancy, and heavy vehicle percentage corresponding to each of devices 1 and 2. The outputs include traffic flow, average speed, and road occupancy. This is because, for a road, traffic detection devices 1 and 2 represent the situation at two nodes. Since information such as vehicle traffic flow and average speed is not constant, the situation at the two nodes cannot characterize the situation in the highway construction area (construction location). The network model is used to input the traffic flow, average speed, and road occupancy corresponding to the highway construction area.

[0092] The training adopts a multi-task learning framework, and the loss function is the weighted mean square error of the three predicted targets, as shown in equation (6):

[0093] (6)

[0094] in, These are the predicted and actual values ​​for traffic, speed, and occupancy, respectively. These are the weighting coefficients of the loss function, reflecting a slightly higher weighting for the accuracy of speed prediction.

[0095] The Bayesian correction module is designed to address the uncertainty of prediction results. This module first constructs the prediction distribution: during the inference phase (the inference phase of the training process), all Dropout layers in the network are kept active (dropout rate of 0.2), and 100 forward propagations are performed on the same input to obtain the prediction set. Each element corresponds to a three-dimensional prediction target. The prediction mean is calculated based on this set. and prediction variance The correction process utilizes historical prediction errors for adaptive adjustment (adaptive adjustment means that this value gradually changes as historical data accumulates; the system continuously generates new data as it runs, and this new data is also applied to the calculation of this value, thus achieving adaptive adjustment). Specifically, a sliding window of length 100 is maintained to store the most recent prediction error sequence. For the current predicted value... The corrected formula is:

[0096]

[0097] Where w is the adjusted weight. , It is the exponential moving average of the historical prediction error series. This is an adjustable parameter that controls the effect of variance on the correction strength.

[0098] When prediction variance When the variance is small, w approaches 1, resulting in stronger error correction; when the variance is small... When the value is large, w approaches 0, maintaining the original predicted value. This design ensures that the system actively learns historical error patterns at high confidence levels and maintains a conservative prediction strategy at low confidence levels.

[0099] The final output is the corrected predicted value and its 95% confidence interval. When the predicted variance is detected to exceed the threshold of 0.25, the system automatically reduces the weight of the predicted result in subsequent decisions and relies more on historical statistical values, thereby improving the system's robustness in high uncertainty scenarios.

[0100] For example, the prediction results for a certain period are shown in Table 2.

[0101] Table 2 Examples of Model Prediction Results

[0102]

[0103] Model performance metrics: MAE=2.3, RMSE=3.1, MAPE=3.2%, CORR=0.95.

[0104] Step 3: Meteorological Data Processing and Road Surface Slipperiness Calculation. Based on temperature, humidity, and rainfall data collected from meteorological stations, road surface slipperiness is calculated using a physical model to provide basic data for subsequent risk assessment. The specific processing flow is shown in Figure 4.

[0105] Firstly, based on temperature ( The basic friction coefficient is calculated based on the rainfall (mm / h) and the amount of precipitation, then a humidity correction is introduced to obtain the standardized pavement slippage index. The specific calculation process is as follows:

[0106] The basic friction coefficient is calculated using an improved exponential decay model, as shown in equation (7):

[0107] (7)

[0108] in: The coefficient of friction is the maximum under ideal conditions. According to the "Technical Standard for Highway Engineering" (JTG B01-2014) and the "Technical Specification for Construction of Asphalt Pavement" (JTG F40-2004), the standard coefficient of friction for asphalt pavement under dry conditions is 0.45-0.55. Since this refers to asphalt pavement, therefore... The value is 0.55; This is the attenuation coefficient, which is temperature-dependent. R represents rainfall; T represents temperature.

[0109] This model reflects the nonlinear effect of rainfall on road surface friction performance and considers the moderating effect of temperature on the attenuation rate. When the temperature is far from 20... At that time, the road surface is more sensitive to rainfall, and the coefficient of friction decreases faster.

[0110] Humidity correction uses a linear adjustment factor, as shown in equation (8):

[0111] (8)

[0112] Where: H represents relative humidity (%); The humidity influence coefficient is set to 0.2.

[0113] Finally, the coefficient of friction is converted into a standardized index of slipperiness, as shown in equation (9):

[0114] (9)

[0115] in, The value represents the degree of road slipperiness, ranging from [0,1], where 0 indicates completely dry and 1 indicates completely slippery. The value increases as the coefficient of friction decreases.

[0116] Taking meteorological data for a certain period as an example, the degree of road slipperiness was calculated, and the results are shown in Table 3.

[0117] Table 3. Examples of meteorological data mapping results

[0118]

[0119] Step 4: Calculation of Construction Activity Index. The construction activity index is calculated using principal component analysis (PCA). The specific processing flow is shown in Figure 5.

[0120] (1) Multidimensional data preprocessing:

[0121] This invention first standardizes the original construction data and establishes a three-dimensional index system:

[0122] Personnel density index The calculation is shown in equation (10):

[0123] (10)

[0124] in, Let A represent the number of construction workers and A represent the area of ​​the construction area (m²). This indicator converts the absolute number of people into personnel density per unit area, eliminating the influence of the size of the construction area.

[0125] The mechanical strength index M is calculated as shown in equation (11):

[0126] (11)

[0127] in, L represents the number of mechanical equipment, and L represents the activity level (levels 1-5). This indicator comprehensively considers the coupling effect between the number of equipment and the intensity of work.

[0128] The calculation of the job type index W is shown in equation (12):

[0129] W = {1.0, maintenance job; 2.0, expansion job; 3.0, new job} (12)

[0130] (2) Principal component analysis:

[0131] Standardized 3D data The specific steps for performing principal component analysis are as follows:

[0132] Construct the sample matrix X:

[0133] (13)

[0134] in These are standardized data for personnel density, mechanical intensity, and job type, respectively.

[0135] Calculate the covariance matrix S:

[0136] (14)

[0137] Where n is the number of samples.

[0138] Solve the characteristic equation:

[0139] (15)

[0140] It is an identity matrix. Eigenvalues ​​are obtained. and the corresponding feature vector .

[0141] Calculate the principal component contribution rate:

[0142] (16)

[0143] Principal components with a cumulative contribution rate of 85% were selected.

[0144] (3) Calculation of construction activity index:

[0145] Based on the principal component analysis results, the construction activity index is calculated:

[0146] (17)

[0147] in, For the score of the i-th principal component, Its corresponding contribution rate.

[0148] Taking construction data for a certain period as an example, the calculation results are shown in Table 4.

[0149] Table 4 Example of Construction Activity Index Calculation Results

[0150]

[0151] Step 5: Make a speed limit decision based on the risk transmission theory. The specific process is shown in Figure 6.

[0152] (1) Multi-dimensional security indicator calculation model:

[0153] Traditional safety assessment methods typically consider only a single factor, such as traffic flow or weather conditions, failing to comprehensively reflect the overall safety status of the construction area. This invention establishes a multi-dimensional safety indicator calculation model that comprehensively considers safety indicators from three dimensions: traffic flow, weather, and construction.

[0154] Traffic flow safety = (Speed, density, congestion density, road occupancy, heavy vehicle ratio), traffic flow safety indicators used in this implementation method. as follows:

[0155] (18)

[0156] Where v is the current average speed (km / h). is the free-flow velocity (taken as 80km / h); k is the traffic density (veh / h·km), which is equal to the flow rate obtained in step 2 divided by the length of the construction zone. In this embodiment, the length of the construction zone is taken as 1.6km. Congestion density (taken as the historical highest traffic density of 120veh / h·km); r represents the road occupancy rate; r represents the proportion of heavy vehicles. The weights for the speed-density term and the heavy vehicle term are respectively ( =0.7, =0.3). When speeds are too high and vehicle density is high, the safe distance between vehicles is insufficient, increasing safety risks; when speeds are too low and vehicle density is low, traffic efficiency decreases, but safety risks are relatively low. Meanwhile, an increased proportion of heavy vehicles significantly increases safety risks because heavy vehicles have longer braking distances and greater interference with traffic flow. When road occupancy is high, road operating space is compressed, increasing safety risks.

[0157] Construction safety =Construction Activity Index. This index takes into account the impact of construction activity on traffic risks. In construction zones, the Construction Activity Index largely determines the speed limit. When construction activity is high, the speed limit should be appropriately reduced to ensure the safety of vehicles and construction workers; when the Construction Activity Index is low, the speed limit can be appropriately increased to improve vehicle traffic efficiency.

[0158] Environmental safety = (Visibility, Slipperyness) Environmental safety indicators used in this embodiment as follows:

[0159] (19)

[0160] in, Visibility (km); The road surface slippage is calculated in step 3; These are the fitting parameters. This function considers the nonlinear coupling effect of visibility and slipperiness. When visibility decreases or slipperiness increases, A decrease in the value indicates a decline in environmental safety. When both occur simultaneously, the risk-additional effect is significant.

[0161] The formula for calculating comprehensive risk is shown in equation (20):

[0162] (20)

[0163] In the formula, Risk is the comprehensive risk value, with a value range of [0,1]; is a safety index based on average speed, traffic density, and heavy vehicle ratio; is a safety index based on traffic density and heavy vehicle ratio; and is a safety index based on visibility and slipperiness.

[0164] The weighting is based on the degree of influence of each factor on safety: traffic flow safety has a weight of 0.4 because vehicle operating status is the main factor affecting the severity of traffic accidents; construction safety and environmental safety both have a weight of 0.3, as their influence on safety risk is comparable.

[0165] (2) Dynamic risk level determination:

[0166] Traditional risk level determination methods typically employ fixed thresholds or simple classifications, which cannot adapt to the risk characteristics under varying construction conditions. This invention designs a dynamic risk level determination algorithm that determines the risk level based on a comprehensive risk value.

[0167] Low risk level: When Risk < 0.3, it is considered low risk. This threshold is based on historical data analysis. When the overall risk value is below 0.3, the safety condition of the construction area is good, and a relatively lenient speed limit strategy can be adopted.

[0168] Medium risk level: When 0.3 ≤ Risk < 0.6, it is judged as medium risk. This level indicates that there is a certain safety risk in the construction area, and appropriate speed limit measures need to be taken to maintain traffic efficiency as much as possible while ensuring safety.

[0169] High-risk level: When 0.6 ≤ Risk < 0.8, it is judged as high-risk. This level indicates that the safety risk in the construction area is relatively high, and stricter speed limit measures need to be taken to ensure safety first.

[0170] Extremely High Risk Level: When Risk ≥ 0.8, it is classified as extremely high risk. This level indicates that the safety risk in the construction area is extremely high, requiring the strictest speed limit measures, and temporary closure of the construction area may be considered if necessary.

[0171] (3) Probabilistic correlation between risk level and speed limit:

[0172] Based on the speed limit standards of the construction section, the speed variation range of the variable speed limit sign is determined as S={20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70}. This design is based on the actual displayed speed limit values ​​of the signs, with 5 km / h intervals, conforming to the speed limit standards of the traffic management department. In this embodiment, the speed limit for the construction section is 70 km / h. For other sections, the specific speed limit shall prevail, with the minimum unit for speed limit changes being 5 km / h.

[0173] Statistical analysis was performed on the historical average speed under different risk levels. At the same time, combined with the speed change space S, a probability correlation table of risk level-speed limit value was formulated. In this embodiment, based on the historical average speed under different risk levels and combined with the speed area corresponding to the speed change space S, the speed limit risk corresponding to each speed area under different risk levels was obtained, and then the probability correlation table of risk level-speed limit value was obtained, as shown in Table 5.

[0174] Table 5. Risk Level-Speed ​​Limit Probability Correlation Table

[0175]

[0176] The data in Table 5 represents speed limit risk, which is actually a probability value. For low-risk situations, the probability leans towards higher speed limits to prioritize traffic efficiency; for high-risk situations, the probability leans towards lower speed limits to prioritize safety; and for medium-risk situations, the probability is distributed across medium speed limits to achieve a balance between safety and efficiency.

[0177] It should be noted that speed limit risk and Risk are different. Risk is used to determine the risk level, and then the speed limit value corresponding to the highest probability (highest speed limit risk value) is found in the corresponding risk level in Table 5 based on the determined risk level.

[0178] (4) Risk transmission theory model:

[0179] Traditional speed limit decision-making methods typically consider each variable message sign (VMS) independently, neglecting the spatial propagation effect of risks in construction zones. This invention, for the first time, applies risk transmission theory to speed limit decision-making in construction zones, establishing a spatial propagation model of risk:

[0180] The risk transmission direction is VMS4 (200m from the construction area) → VMS3 (400m from the construction area) → VMS2 (600m from the construction area) → VMS1 (800m from the construction area). This direction design is based on the fact that the construction area is located near VMS4, and the risk spreads from the construction area to the direction away from the construction area, which is consistent with the actual physical phenomenon.

[0181] The formula for the distance attenuation of the comprehensive risk value Risk(d) at a distance d from the construction area is given in equation (21):

[0182] (twenty one)

[0183] in, The overall risk value of the construction area; is the attenuation coefficient, with a value of 0.50; d is the distance (km).

[0184] This formula reflects the spatial propagation and attenuation characteristics of risk. As distance increases, the risk decreases exponentially; the farther away from the construction area, the smaller the risk impact. This realizes risk transmission from the construction area to each VMS (Vehicle Management System), enabling speed limit decisions to consider spatial risk distribution and achieve a more scientific speed limit strategy.

[0185] Taking a Risk of 0.5 as an example, based on 0.5, the transmission risk values ​​at distances of 200m, 400m, 600m, and 800m from the construction site are calculated according to formula (21). (This implementation method calculates based on this distance, but the actual distance can be determined according to the actual design requirements. It should be noted that Table 5 is based on this distance division. If other distance division methods are used, then Table 5 needs to be recalculated.) The risk level of each location is determined, and the transmission risk values ​​at the four locations are 0.45 and 0.41, respectively. 0.37 and 0.34 all correspond to medium risk. Speed ​​limits are determined starting from the location closest to the construction area: 200m corresponds to a maximum probability speed limit of 50 (probability 0.40) in Table 5; 400m corresponds to 55 (based on 50, the probability is 0.30); 600m corresponds to 60 (probability 0.20); and 800m shows no speed limit higher than 60. Therefore, the speed limits at 200m, 400m, and 600m are all reduced by 5 km / h. Considering the constraints: ① Gradient adjustment ensures VMS1 ≥ VMS2 ≥ VMS3 ≥ VMS4, therefore the corresponding speed limits for the four locations in Table 5 are 60, 55, 50, and 45 respectively.

[0186] If the risk level of a certain location changes when calculating the transmission risk value of the four locations, the speed limit value is determined based on the highest probability among the risk levels corresponding to the current location. Then, the speed limit value of the next location is determined forward based on the speed limit value determined by the highest probability.

[0187] (5) Gradient constraint strategy:

[0188] Based on traffic safety management principles, relevant physical constraints are added to the risk transmission model to ensure that vehicles gradually decelerate as they approach the construction zone, avoiding safety hazards caused by sudden changes in speed limits.

[0189] ① Gradient adjustment ensures that VMS1 ≥ VMS2 ≥ VMS3 ≥ VMS4;

[0190] ② The difference in speed limit between two adjacent speed limit signs is less than or equal to 10 km / h;

[0191] ③ The speed limit for VMS1 is greater than or equal to 60 km / h and less than or equal to 70 km / h (the speed limit is 70 km / h on non-construction sections, therefore, as the first variable speed limit sign before entering the construction zone, the difference between the value of VMS1 and the original speed limit should be less than or equal to 10 km / h).

[0192] Taking construction zone risk = 0.50 and 0.90 as examples, construction zone risk refers to Risk0, which is assumed to be 0.5 and 0.9 here. The risk transmission value of each location and the corresponding speed limit value of the variable speed limit sign are shown in Table 6 for the calculation demonstration.

[0193] Table 6 Risk transmission values ​​at each location and corresponding speed limits on variable speed signs

[0194]

[0195]

[0196] Step 6: Compare and analyze the four speed limit schemes. The specific comparison and analysis schemes are shown in Figure 7.

[0197] (1) Multi-scheme comparison design:

[0198] To comprehensively evaluate the effectiveness of different speed limiting strategies, this invention designs four representative speed limiting schemes for comparative analysis:

[0199] Option 1 is unrestricted (free passage), which represents the traditional management method. There is no speed limit control for vehicles, and vehicles can pass freely at the designed speed.

[0200] Option 2 is a fixed speed limit (uniform 40km / h). This option represents the traditional fixed speed limit management method, where all VMSs display the same speed limit value, without taking into account actual traffic and weather conditions.

[0201] Option 3 is a step-down speed limit (70-60-50-40km / h). This option represents an improved fixed speed limit method, which gradually reduces the speed limit from VMS1 to VMS4 to form a speed gradient.

[0202] Scheme 4 is dynamic optimization (the scheme of this invention). This scheme is based on the risk transmission theory and dynamically adjusts the speed limit values ​​of each VMS according to real-time traffic, weather and construction conditions.

[0203] This multi-scheme comparative design covers different speed limiting strategies from traditional to intelligent, and can comprehensively evaluate the effects of different schemes, verifying the advantages of the present invention.

[0204] (2) Multi-dimensional evaluation indicator system:

[0205] Traditional evaluation methods typically consider only a single indicator, such as safety or efficiency, failing to comprehensively reflect the overall effectiveness of a solution. This invention establishes a multi-dimensional evaluation indicator system, comprehensively evaluating the effectiveness of different solutions from two dimensions: safety (weight 0.6) and efficiency (weight 0.4).

[0206] The safety index is the risk of accidents. The risk of accidents is calculated based on the traffic conflict index, which reflects the probability of accidents occurring in the construction area under different schemes. The number of times the TTC (Time To Collision) is less than 3 seconds within 15 minutes is counted, and the accident risk index is obtained after normalization. Its calculation formula is shown in equation (22):

[0207] (twenty two)

[0208] Efficiency indicators include traffic delay, traffic flow, and service level. Traffic delay indicators... = (Actual travel time - Ideal travel time) / Ideal travel time. Traffic delay is based on the time it takes for vehicles to pass through the construction zone, reflecting the degree of traffic delay under different scenarios; Traffic flow index = Actual traffic flow / Design capacity. Traffic flow is based on road capacity analysis and reflects the road's efficiency under different scenarios; Service level index = Actual speed / Free-flow speed. The level of service is based on traffic flow state analysis and reflects the quality of traffic services under different scenarios.

[0209] This multi-dimensional evaluation system is designed based on actual engineering needs and can quantitatively evaluate the performance of different solutions in terms of safety and efficiency, providing a scientific basis for solution selection.

[0210] (3) Comparison of results:

[0211] Taking one week of test data as an example, the comparison results are shown in Table 7, and the performance comparison is shown in Figure 8.

[0212] Table 7 Comparison Results of Schemes

[0213]

[0214] Comparative analysis based on one week's test data shows that the dynamic optimization scheme has significant advantages over other schemes:

[0215] In terms of safety, the dynamic optimization scheme improves safety by 26% compared to the fixed speed limit scheme, 17% compared to the gradual speed reduction scheme, and 111% compared to the unrestricted scheme. This advantage is mainly due to the application of risk transmission theory, which enables dynamic adjustment of the speed limit strategy based on real-time risk conditions.

[0216] In terms of efficiency, the dynamic optimization scheme is 54% more efficient than the fixed speed limit scheme, 16% more efficient than the gradual speed reduction scheme, and 6% less efficient than the unrestricted scheme. This result indicates that the dynamic optimization scheme can maintain high traffic efficiency while ensuring safety.

[0217] Step 7: The system is applicable to various road construction scenarios, as shown in Figure 9.

[0218] Traditional speed limit systems are typically designed for specific scenarios, lacking versatility and struggling to adapt to different road types and construction conditions. This invention designs a multi-scenario applicable system capable of adapting to various road construction scenarios:

[0219] (1) Ordinary highway construction scenarios, applicable to long-distance, high-speed construction sections. The system can handle the characteristics of complex traffic flow, realize long-distance risk propagation through risk transmission theory, and ensure the safety of high-speed vehicles.

[0220] (2) Urban road maintenance scenario, applicable to the maintenance of urban main roads and secondary roads. The system can adapt to the complexity and diversity of urban traffic and handle the special needs of urban traffic flow through a multi-dimensional safety index calculation model.

[0221] (3) Bridge and tunnel construction scenarios, applicable to construction areas with special structures. The system can handle the special environmental conditions of bridges and tunnels, and adapt to the special meteorological environment inside the tunnel through the meteorological data mapping module.

[0222] (4) Mountain road construction scenario, applicable to construction under complex terrain conditions. The system can adapt to the complex terrain and severe weather conditions of mountain roads, and respond to the special needs of mountain transportation through dynamic weight adjustment algorithm.

[0223] This multi-scenario applicability design reflects the system's versatility and adaptability, providing a foundation for the system's large-scale promotion and application.

[0224] The above embodiments are merely one example and do not limit the scope of the invention. The data in the examples are only for reference in the specific application of the method of the invention and do not guarantee that the exemplified data is completely true and reasonable. It should be understood that the data calculations in the examples may result in slight differences in some results due to differences in the number of significant figures retained; since they are only for reference in the calculation process, the retention of significant figures in the original data and calculation results (including intermediate and final results) in the examples does not consider practical rationality. In actual applications, a reasonable number of significant figures should be retained according to the actual meaning and calculation accuracy requirements. The invention may also have other various embodiments. Without departing from the spirit and essence of the invention, those skilled in the art can make various corresponding changes and modifications according to the invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A collaborative dynamic speed limit control method based on the risk transmission theory of construction zones, characterized in that, include: First, obtain the traffic flow, average speed, and road occupancy data for the entire affected section of the highway construction; then, obtain the road surface slippage of the affected section. The construction activity index of the affected sections of the highway is obtained; then, speed limit decisions are made based on the obtained data, including: determining traffic density based on traffic flow and construction zone length, and determining congestion density based on historical traffic density; and determining traffic flow safety based on average speed, traffic density, congestion density, and vehicle load ratio. The construction activity index is used as a measure of construction safety. Environmental safety is determined based on visibility and road surface slipperiness. ;based on 、 、 Obtain the overall risk value The risk level is determined based on the risk assessment; the probabilistic relationship between the risk level and the speed limit is established: based on the speed limit standards of the construction section, the speed variation space S of the variable speed limit sign is determined; based on the historical average speed under different risk levels, and combined with the speed area corresponding to the speed variation space S, the probability of the speed limit value corresponding to each speed area under different risk levels is obtained, resulting in a risk level-speed limit probability correlation table; the distance from the construction area, from near to far, is used as the risk transmission direction, and the comprehensive risk value of the construction area is calculated. As ,according to Calculate the comprehensive risk value of transmission at a distance d from the construction area, where, The attenuation coefficient is used. Each variable speed limit sign (VMS) is denoted as VMS1, VMS2, ..., VMSq, ... based on its distance from the construction area, from farthest to closest. According to the risk level corresponding to the comprehensive risk value transmitted from different distances to the construction area, the speed limit value for VMS1 is determined based on the speed limit value with the highest probability among the corresponding risk levels. Then, the speed limit values ​​corresponding to VMSq are determined sequentially based on the speed limit values ​​in the risk level-speed limit probability association table. Finally, the gradient adjustment constraint VMS1≥VMS2≥……≥VMSq≥…… is used as one of the constraints. Combining these constraints, the risk transmission value at each location and the corresponding variable speed limit sign speed limit value are determined, thus providing a dynamic speed limit control scheme.

2. The collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 1, characterized in that, The process of obtaining traffic flow, average speed and road occupancy data for the entire affected section of highway construction includes: collecting traffic data at each of the at least two traffic detection devices set up in the affected section of highway construction, including traffic flow, average speed, road occupancy and heavy vehicle ratio data; and then obtaining the overall traffic data of the affected section of highway construction through a neural network model based on the traffic data from two of the at least two traffic detection devices.

3. The collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 2, characterized in that, The processing procedure of the neural network model is as follows: the output of the input layer is fed into a bidirectional LSTM layer for processing. The bidirectional LSTM layer adopts a three-layer stacked structure. After Dropout regularization of the output of each layer, the forward and backward hidden states are concatenated in the feature dimension to form a high-order feature representation. The output of the bidirectional LSTM layer is fed into a multi-head attention mechanism to finally obtain the feature matrix output by the multi-head attention mechanism. Then, a fully connected layer is used for prediction, and the output dimension corresponds to the three prediction targets [flow rate, average speed, and road occupancy].

4. The collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 3, characterized in that, In the process of obtaining overall traffic data for the road construction-affected area using a neural network model, a Bayesian correction module is used to correct the output of the neural network model. Specifically, during the inference phase, all Dropout layers in the neural network model are kept active, and D forward propagations are performed on the same input to obtain the prediction set. Each element corresponds to a three-dimensional prediction target; the prediction mean is calculated based on this set. and prediction variance Then, adaptive correction is performed using historical prediction errors, including: maintaining a sliding window of length D, storing the most recent prediction error sequence, and obtaining the exponential moving average of the historical prediction error sequence. For the current predicted value Make corrections, the correction formula is as follows ,in, To adjust the weights, This is an adjustable parameter that controls the effect of variance on the correction strength.

5. The collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 1, characterized in that, The process of obtaining the pavement slipperiness of the affected section of highway construction includes: firstly, calculating the base friction coefficient based on temperature and rainfall. ,in: The maximum coefficient of friction under ideal conditions. The attenuation coefficient is... R represents rainfall; T represents temperature; then humidity correction is introduced. Where: H is relative humidity, The humidity influence coefficient is used to obtain the standardized road surface slipperiness index. 。 6. The collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 1, characterized in that, The process of obtaining the construction activity index of the affected section of highway construction includes: (1) obtaining the personnel density index Mechanical strength index 1. Job type index W; (2) Standardized three-dimensional data Principal component analysis was performed to obtain the principal component contribution rates. Then calculate the construction activity index. ， The score is the score of the i-th principal component.

7. The collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 6, characterized in that, The acquisition of personnel density index Mechanical strength index The process for the job type indicator W includes: personnel density indicator ,in, A represents the number of construction workers, and A represents the area of ​​the construction area; mechanical strength index ,in, L represents the number of mechanical equipment, and L represents the activity level; the operation type index W = {1.0, maintenance operation; 2.0, expansion operation; 3.0, new construction operation}.

8. A collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to any one of claims 1 to 7, characterized in that, Traffic flow safety is determined based on average speed, traffic density, congestion density, and vehicle load ratio. The process is as follows: Traffic flow safety = (Speed, density, congestion density, road occupancy, heavy vehicle ratio), traffic flow safety indicators as follows: (18) Where v is the current average speed, k is the free-flow velocity; k is the traffic density. Blocking density; denoted as road occupancy rate; r represents the proportion of heavy vehicles.

9. A collaborative dynamic speed limit control method based on the risk transmission theory of construction zones as described in claim 8, characterized in that, The process of determining environmental safety based on visibility and road surface slipperiness is as follows: Environmental Safety = (Visibility, Slipperiness), Environmental Safety Indicators as follows: (19) Among them, Visibility; The degree of slipperiness of the road surface; 、 These are the fitting parameters.

10. A collaborative dynamic speed limit control method based on the risk transmission theory of construction zones according to claim 9, characterized in that, In the process of proposing a dynamic speed limit control scheme, it is necessary to combine the risk of accident occurrence to give the final dynamic speed limit control scheme. Therefore, the risk occurs. Where TTC is the distance-to-collision time.

Images