Self-adaptive agglomerate fog early warning and response control method and system for expressway

Abstract

According to the expressway self-adaptive agglomerate fog early warning and response control method and system, meteorological data and traffic flow data of an expressway are collected in real time, and a reinforcement learning algorithm is utilized to predict a space-time mode of occurrence of agglomerate fog. By calculating the agglomerate fog risk index, if the risk exceeds the standard, the system triggers agglomerate fog early warning and dynamically adjusts the traffic management strategy, so that the driving safety is improved to the maximum extent and the traffic jam is reduced to the maximum extent. The system continuously optimizes decisions through traffic flow and visibility information fed back in real time, realizes adaptive adjustment through reinforcement learning, and adapts to changes under different weather and traffic conditions. The method can accurately predict the occurrence of agglomerate fog, optimizes the traffic flow, reduces the occurrence rate of traffic accidents, improves the passing efficiency and safety of the expressway, and has a wide application prospect.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation, and particularly to a method and system for highway adaptive fog warning and response control. Background Art

[0002] With the continuous increase in traffic volume and the impact of climate change, the phenomenon of fog often occurs on highways and other traffic-intensive areas. Fog is a kind of haze formed by the condensation of water vapor, which often appears suddenly when environmental factors such as temperature and humidity are suitable, and its duration and influence range have strong locality and suddenness. Under fog conditions, visibility drops significantly, seriously affecting traffic safety and easily leading to traffic accidents, especially on sections of highways with higher vehicle speeds. In order to cope with the threat of fog to traffic, relevant research and technologies mainly focus on fog warning, detection, and response control.

[0003] Existing fog warning systems mainly rely on meteorological sensors and fixed traffic management measures. Common technical solutions include:

[0004] (1) Meteorological warning system: Using meteorological data (such as temperature, humidity, wind speed, etc.) to predict the occurrence of fog. Such systems usually give early warnings to traffic management departments based on historical meteorological data and statistical models. However, the accuracy of this method is relatively low, often only giving warnings after the fog has occurred, and the reaction speed is slow.

[0005] (2) Traditional traffic control system: Including fixed traffic control means such as speed limit signs and traffic signal regulation. When fog occurs, the traffic signal system makes simple speed limit adjustments according to the set rules. These methods lack flexibility and cannot be dynamically optimized according to real-time traffic flow and visibility changes, which may lead to unsmooth traffic flow, and even traffic congestion and accidents.

[0006] (3) Video monitoring and image recognition system: Some studies use cameras and image recognition technology to detect the occurrence of fog and control by recognizing traffic conditions. Such systems can monitor fog in real time, but their limitation is that image recognition depends on a clear camera view and requires a large amount of computing resources, making it difficult to respond quickly under complex traffic and weather conditions.

[0007] Although some progress has been made in fog warning and control in the existing technology, there are still the following disadvantages and problems:

[0008] (1) Poor prediction accuracy: Existing fog prediction systems usually rely on meteorological sensors and traditional statistical methods, with relatively low accuracy and real-time performance, and fail to effectively consider dynamically changing traffic and meteorological factors, resulting in the inability to give early warnings in time when fog occurs.

[0009] (2) Response lag: Traditional traffic control measures usually adjust based on predetermined rules when fog banks occur, lacking flexibility and intelligence, failing to respond in real time to changes in meteorology and traffic flow, easily leading to traffic control measures not adapting to the actual situation, causing traffic accidents or congestion.

[0010] (3) Lack of intelligent optimization: Existing technologies do not use intelligent algorithms to dynamically adjust traffic control strategies based on real-time data, resulting in traffic management means being too rigid to achieve optimal traffic flow control and accident prevention.

[0011] In summary, there are still many deficiencies in the existing technologies in the early warning, detection, and response to fog banks. There is an urgent need for a more intelligent, real-time, and flexible solution to better cope with the threats posed by fog banks to traffic safety on highways and other traffic-intensive areas. Summary of the Invention

[0012] In view of the problems existing in the above-mentioned existing technologies, the present invention provides a method and system for adaptive fog bank early warning and response control on highways, aiming to dynamically monitor and analyze information such as real-time meteorological data and traffic flow data of highways, and continuously optimize decisions using reinforcement learning algorithms, making the early warning and emergency response to fog banks more accurate and efficient, and capable of improving traffic safety, reducing traffic congestion, and optimizing traffic flow under fog bank conditions, providing a more intelligent traffic management solution for highways and other traffic-intensive areas.

[0013] To achieve the above object, the specific solutions of the present invention are as follows:

[0014] The method for adaptive fog bank early warning and response control on highways includes the following steps:

[0015] Step 1, collect real-time meteorological data and traffic flow data on highways. The meteorological data includes temperature data, air humidity data, wind speed data, and visibility data; the traffic flow data includes vehicle speed data, vehicle distance data, lane occupancy rate data, and traffic flow density data;

[0016] Step 2, perform denoising, standardization, and normalization processing on the meteorological data and traffic flow data collected in Step 1, input the processed meteorological data and traffic flow data into a reinforcement learning model, predict the spatio-temporal pattern of fog bank occurrence through a reinforcement learning model trained with historical data, construct a fog bank risk assessment model by combining a statistical model and a reinforcement learning algorithm, and use the fog bank risk assessment model to calculate a fog bank risk index based on the real-time collected meteorological data and traffic flow data; when the fog bank risk index exceeds a predetermined threshold, trigger a fog bank early warning, and the system automatically issues an early warning signal;

[0017] Step 3: According to the freeway fog risk index and real-time traffic flow data described in Step 2, use the reinforcement learning model to dynamically adjust traffic control strategies, which include speed limits, signal light regulation, and lane guidance;

[0018] Step 4: Optimize traffic management strategies through real-time feedback of traffic flow data, accident occurrence data, and visibility information, and take emergency response measures in advance when freeway fog forms;

[0019] Step 5: According to the real-time feedback of traffic flow data, accident occurrence data, and visibility information, use the reinforcement learning algorithm to update model parameters to adjust the specific parameters of the freeway fog warning threshold and traffic control strategies.

[0020] Furthermore, the temperature data in Step 1 is collected in real time through temperature sensors, the air humidity data is collected in real time through humidity sensors, the wind speed data is collected in real time through wind speed sensors, the visibility data is collected in real time through visibility sensors, the vehicle speed data is obtained by monitoring the driving speed of vehicles in real time through road surface radars or vehicle networking devices, the vehicle distance data is obtained by video monitoring or radar ranging devices to acquire the distance between vehicles, the lane occupancy rate data is monitored by lane perception devices to monitor the vehicle occupancy of each lane, and the traffic flow density data is obtained by calculating the number of vehicles passing through a specific section per unit time in real time by traffic flow detection devices.

[0021] Furthermore, the denoising process in Step 2 includes the following steps:

[0022] Step 21: Automatically select a denoising method according to the characteristics of meteorological data and traffic flow data:

[0023] For traffic flow data, the moving average method of the smoothing filter method is used to calculate the average value in the data window and smooth out the fluctuations. The calculation formula is as follows:

[0024]

[0025] Where: Y t is the smoothed data, X i is the original data, and N is the window size, usually an odd number;

[0026] For cases with obvious outliers, the median filter method is used to calculate the median value of the data to replace the original data point. The calculation formula is as follows:

[0027]

[0028] Where: Y t is the denoised data, represents the data at time point at the left end of the window; represents the data at time point data; represent data, at the right end of the window.

[0029] For data with complex frequencies and large fluctuations, the wavelet denoising method is adopted. The signal is decomposed by wavelet transform to remove high-frequency noise;

[0030] Step 22, optimize the denoising strategy in combination with the expected effect: According to the influence of different denoising methods in Step 21 on meteorological data and traffic flow data, select a suitable denoising method to remove noise while not losing key information;

[0031] Step 23, combine the denoising method in Step 21 with a reinforcement learning model to form an adaptive feedback mechanism and dynamically select the most suitable denoising strategy.

[0032] Furthermore, the calculation formula for the normalization process in Step 2 is as follows:

[0033]

[0034] In the formula: X' is the normalized data; X is the original data; μ is the mean of the data, and σ is the standard deviation of the data.

[0035] Furthermore, the normalization process in Step 2 automatically selects the Min-Max normalization method or the Z-score normalization method according to the characteristics of each meteorological data and traffic flow data;

[0036] The Min-Max normalization method is used for the case where the data distribution is relatively uniform and there are no significant outliers. The calculation formula for the Min-Max normalization method is as follows:

[0037]

[0038] In the formula: X' is the normalized data, X is the original data, X min and X max are the minimum and maximum values of the data respectively;

[0039] The Z-score normalization method is used for the case where the data has large fluctuations and contains outliers or does not satisfy a uniform distribution. The calculation formula for the Z-score normalization method is as follows:

[0040]

[0041] In the formula: μ is the mean of the data, range(X) is the range of the data, that is, the maximum value minus the minimum value.

[0042] Further, the step 2 further includes outlier detection, and the outlier detection includes the box plot method and the Z-Score method. The box plot identifies outliers based on four numbers;

[0043] The Z-Score method determines whether meteorological data and traffic flow data are outliers. If the absolute value is greater than a certain threshold, it is regarded as an outlier. The calculation formula is as follows:

[0044]

[0045] If |Z| > 3, then X is considered an outlier, where μ is the mean of the data and σ is the standard deviation of the data.

[0046] Further, the reinforcement learning model in step 2 adopts the Q-learning or deep Q-network algorithm. The reinforcement learning model includes a state space, an action space, and a reward function;

[0047] The state space includes current traffic flow data, meteorological data, road visibility, and the risk index of patchy fog;

[0048] The calculation formula of the risk index of patchy fog is as follows:

[0049] F = α1T + α2RH + α3V + α4P + α5VIS

[0050] In the formula, α1, α2, α3, α4, and α5 are weight coefficients, which are automatically adjusted by the reinforcement learning model through historical data; T represents temperature; RH represents relative humidity; P represents air pressure; VIS represents visibility;

[0051] The action space includes adjustments to traffic control strategies, and the traffic control strategies also include speed limit adjustment, signal light adjustment, and lane guidance;

[0052] The speed limit adjustment dynamically adjusts the electronic speed limit sign according to the risk index of patchy fog and real-time visibility data. Specifically:

[0053] When the visibility is 200m ≤ VIS < 500m, the speed limit is adjusted to 80 km / h;

[0054] When the visibility is 100m ≤ VIS < 200m, the speed limit is adjusted to 60 km / h;

[0055] When the visibility VIS < 100m, the speed limit is 40 km / h, and some lanes are closed;

[0056] The signal lamp adjustment is to optimize the green and red light durations of traffic signal lamps, adjust the signal lamp cycle according to the fog risk, traffic flow density, vehicle speed, accident frequency, lane occupancy rate, and section passing capacity, and increase the flashing frequency of traffic signal lamps in high fog risk areas to remind drivers to pay attention to decelerating; dynamically adjust the phase and cycle of signal lamps according to traffic flow density and vehicle speed to reduce vehicle waiting time and congestion;

[0057] The lane guidance is that when the fog causes the visibility in a local area to be too low, the vehicle is guided into a safe lane through a lane guidance device or an electronic sign;

[0058] The reward function is comprehensively evaluated based on traffic flow data, accident rate, and passing efficiency, and the optimal strategy is selected according to the principle of maximizing the reward. The calculation formula of the reward function is as follows:

[0059] R = ω1·λ - ω2·α + ω3·η - ω4·T

[0060] In the formula: R represents the reward function; λ represents the traffic flow in the current area, and the traffic flow includes the number of vehicles and the lane occupancy rate. The higher it is, the more congested it indicates; α represents the number of accidents or the accident probability occurring within a specific time window; η is the passing capacity of the traffic system, that is, the number of vehicles passing through per unit time or the average vehicle speed; T is the fog risk index; ω1, ω2, ω3, and ω4 are the weight coefficients of each factor, indicating the importance of different factors in decision-making;

[0061] The warning signals include, but are not limited to, real-time release of warning information through variable speed limit signs, variable message signs, and in-vehicle information systems to remind drivers to decelerate or take a detour.

[0062] Furthermore, the emergency response measures in step 4 include measures such as early activation of speed limits, adjustment of traffic signals, and guiding vehicles into safe lanes.

[0063] The highway adaptive fog warning and response control system for implementing the method includes

[0064] A data acquisition system for real-time acquisition of meteorological data and traffic flow data of the highway;

[0065] A data processing module for denoising, standardizing, and normalizing the acquired meteorological data and traffic flow data;

[0066] A reinforcement learning model for predicting the spatio-temporal pattern of fog occurrence and calculating the fog risk index based on the processed meteorological data and traffic flow data;

[0067] A warning and response control system for triggering a fog warning and dynamically adjusting traffic control strategies when the fog risk index exceeds a predetermined threshold;

[0068] An adaptive learning module for continuously updating a reinforcement learning model based on real-time feedback of traffic flow data, accident occurrence data, and visibility information, and dynamically adjusting the specific parameters of the fog warning threshold and traffic control strategy.

[0069] Furthermore, the warning and response control system includes:

[0070] A warning signal issuing device for real-time issuing of fog warning information through variable speed limit signs, variable message signs, and vehicle information systems;

[0071] A dynamic traffic control device for dynamically adjusting electronic speed limit signs, optimizing traffic signal cycles, and guiding vehicles into safe lanes according to real-time visibility data;

[0072] The adaptive learning module includes:

[0073] A model updating unit for updating the reinforcement learning model according to real-time feedback of traffic flow data and visibility information;

[0074] A threshold adjustment unit for dynamically adjusting the fog warning threshold;

[0075] A strategy optimization unit for optimizing the specific parameters of the traffic control strategy.

[0076] Advantages of the present invention

[0077] The technical advantages of the present invention mainly include the following aspects:

[0078] 1. Improved accuracy of fog prediction: Using the reinforcement learning algorithm combined with real-time meteorological data, historical data, and traffic flow information, deep learning and optimization are performed on the spatio-temporal patterns of fog occurrence, significantly improving the accuracy and timeliness of fog prediction, and avoiding the common false alarm and missed alarm problems in traditional methods.

[0079] 2. Intelligent dynamic response control: Through the reinforcement learning model, according to real-time visibility, traffic flow density, and the influence range of fog, the speed limit strategy, signal regulation, and lane guidance are adaptively adjusted to ensure the best traffic flow management strategy, effectively improving traffic safety and energy efficiency to reduce traffic congestion, overcoming the deficiencies of existing fixed speed limit and signal control methods.

[0080] 3. Strong adaptive optimization ability: Using reinforcement learning technology, the system can continuously learn and optimize strategies during operation. Through the real-time feedback mechanism, the fog warning threshold and response measures are dynamically adjusted to adapt to different weather, seasons, and traffic conditions, enhancing the adaptive ability of the system.

[0081] 4. Reduce traffic accidents and improve safety: By providing early warnings, intelligent speed limits, and optimizing traffic flow control, the accident rate on highways can be effectively reduced, and overall driving safety can be improved, thereby reducing chain-rear-end accidents caused by delayed driver reaction times due to low visibility in foggy weather and minimizing casualties and economic losses.

[0082] 5. Enhance road traffic efficiency: By optimizing speed limits and signal control through reinforcement learning, traffic management becomes more precise. While ensuring safety, unnecessary traffic delays are minimized to the greatest extent, improving the traffic capacity of highways in adverse weather conditions and overcoming the problem of reduced traffic efficiency caused by traditional conservative speed limit measures.

[0083] 6. High degree of intelligence and reduced manual intervention: Through automated data collection, real-time learning, and decision optimization, dependence on manual monitoring and decision-making is reduced, and the level of intelligent traffic management is improved, making traffic control more scientific and reasonable.

[0084] 7. Wide range of applications and universality: It is applicable not only to highways but also to urban expressways, bridges, tunnels, and mountain roads where fog is frequent. Through reinforcement learning training in different environments, the system can adapt to various complex traffic scenarios and achieve cross-regional deployment and application.

[0085] In summary, the present invention constructs an adaptive fog warning and response control method and system for highways through reinforcement learning technology. Compared with traditional methods, it has higher prediction accuracy, more flexible dynamic response, and stronger adaptive optimization ability. It can effectively improve traffic safety and efficiency, reduce manual intervention, achieve intelligent traffic management, and has broad application value and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a flowchart of the adaptive fog warning and response control method and system for highways of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0087] The present invention will be further explained and illustrated below in conjunction with the drawings and specific embodiments. It should be noted that the specific embodiments do not limit the scope of the rights of the present invention.

[0088] As Figure 1 shown, the specific embodiment provides an adaptive fog warning and response control method and system for highways, including the following steps:

[0089] Step 1, Collect real-time meteorological data and traffic flow data on highways. The meteorological data includes temperature data, air humidity data, wind speed data, and visibility data; the traffic flow data includes vehicle speed data, vehicle distance data, lane occupancy data, and traffic flow density data.

[0090] (1) Meteorological data:

[0091] Temperature (T): Use a temperature sensor to monitor the temperature in real time.

[0092] Humidity (RH): Use a humidity sensor to monitor the air humidity in real time.

[0093] Wind speed (V): Use a wind speed sensor to measure the wind speed.

[0094] Visibility (VIS): Use a visibility sensor to monitor the visibility data in real time.

[0095] (2) Traffic flow data:

[0096] Vehicle speed (S): Use a road surface radar or vehicle networking (such as V2X) device to monitor the driving speed of vehicles in real time.

[0097] Vehicle distance (D): Obtain the distance between vehicles through video surveillance or radar ranging devices.

[0098] Lane occupancy rate (LOR): Monitor the vehicle occupancy of each lane through lane perception devices.

[0099] Traffic flow density (TD): Calculate the number of vehicles passing through a specific section per unit time in real time by traffic flow detection devices.

[0100] Step 2: Denoise, standardize, and normalize the meteorological data and traffic flow data collected in Step 1 so that the reinforcement learning model can be better trained and optimized.

[0101] (1) Denoising

[0102] The purpose of denoising is to remove unnecessary noise in the data, which may come from sensor errors, outliers, etc. In this embodiment, the most suitable denoising method is selected according to the characteristics and expected effects of the meteorological data and traffic flow data. This method is different from the traditional fixed use of a single denoising technique, but flexibly selects the most suitable denoising strategy according to different situations of the input meteorological data and traffic flow data (such as traffic flow data, accident data, visibility information, etc.).

[0103] The steps of the denoising process are as follows:

[0104] Step 21: Automatically select the denoising method according to the characteristics of the data:

[0105] For traffic flow data, which may be greatly affected by factors such as weather, time, and geography, a smoothing filter method is adopted to calculate the average value in the data window and smooth out the fluctuations. The calculation formula is as follows:

[0106]

[0107] Where: Y t is the smoothed data, X i is the original data, and N is the window size, usually an odd number is selected.

[0108] For cases with obvious outliers (such as when an accident occurs and the traffic flow suddenly changes), the median filtering method is adopted. By calculating the median value of the data to replace the original data points, sudden noise can be effectively removed. The specific calculation formula is as follows:

[0109]

[0110] Where: Y t is the denoised data, represents the data at time point at the left end of the window; represents the data at time point ; represents the data at the right end of the window.

[0111] For meteorological data and traffic flow data with relatively complex frequencies and large fluctuations, the wavelet denoising method is adopted. The signal is decomposed through wavelet transform to remove high-frequency noise.

[0112] Selecting the most suitable denoising method according to the characteristics of meteorological data and traffic flow data (such as noise type, change trend, etc.) can effectively improve the accuracy and stability of the model.

[0113] Step 22, optimize the denoising strategy in combination with the expected effect: Different denoising methods have different effects on meteorological data and traffic flow data. For example, some methods may be more suitable for retaining the boundary features of the data, while others are better at removing high-frequency noise. Especially when preprocessing traffic data in a complex environment (such as in a situation of advection fog), selecting the appropriate denoising method can remove noise without losing key information.

[0114] This embodiment can maintain the quality of meteorological data and traffic flow data, and at the same time achieve a better prediction effect. Especially in a complex environment such as advection fog, it can effectively adjust the warning threshold and traffic control strategy.

[0115] Step 23, the combination of reinforcement learning and denoising:

[0116] In a traffic control system based on reinforcement learning, the model needs to adapt to various complex dynamic changes (such as weather changes, traffic flow fluctuations, etc.). If the most suitable denoising strategy can be dynamically selected during the training process, this may help the reinforcement learning model more accurately identify and respond to traffic states, thereby improving the performance of the system.

[0117] Combine denoising techniques with reinforcement learning algorithms to form an adaptive feedback mechanism, continuously adjusting and optimizing traffic control strategies. Flexibly select denoising methods instead of rigidly using a certain fixed denoising technique. This approach of dynamically adjusting denoising strategies according to data characteristics and expected effects can better adapt to different situations, thereby improving the accuracy and practicality of the entire fog warning and response control system.

[0118] (2) Standardization processing

[0119] Standardization is to perform a linear transformation on the data so that its mean is 0 and the standard deviation is 1, which helps to reduce the dimensionality difference between features and promote the convergence of model training.

[0120] The steps of standardization processing are as follows:

[0121] Calculate the mean μ and standard deviation σ of meteorological data and traffic flow data;

[0122] Perform standardization processing on each data point and convert it into a form with a mean of 0 and a standard deviation of 1. The standardization calculation formula:

[0123]

[0124] In the formula: X' is the standardized data, X is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0125] (3) Normalization

[0126] Normalization is to scale the data to a fixed range, usually [0, 1] or [-1, 1], which helps to avoid the influence of some feature values being too large or too small on model training.

[0127] The normalization processing in step 2 automatically selects the standardization processing method according to the characteristics of each meteorological data and traffic flow data: The standardization processing methods include the Min - Max normalization method and the Z - score normalization method.

[0128] For the case where the data distribution is relatively uniform and there are no significant outliers, use the Min - Max normalization method. Calculate the maximum and minimum values of each meteorological data and traffic flow data feature, and scale the data to the [0, 1] interval. The Min - Max normalization calculation formula is as follows:

[0129]

[0130] In the formula: X' is the normalized data, X is the original data, X min and X max are the minimum and maximum values of the data respectively.

[0131] For cases where the data fluctuates greatly and contains outliers or does not satisfy a uniform distribution, the Z-score normalization method is adopted, and the calculation formula is as follows:

[0132]

[0133] In the formula: μ is the mean of the data, and range(X) is the range of the data (the maximum value minus the minimum value).

[0134] (4) Hybrid processing combining standardization and normalization

[0135] For some machine learning tasks, combining standardization and normalization can further optimize model training. For example, performing standardization first and then normalization can map the data to a unified range while retaining the original distribution characteristics of the data.

[0136] Hybrid method example: First, perform standardization processing on the data so that its mean is 0 and the standard deviation is 1. Then, use the Min-Max normalization method to map the standardized data to the interval [0,1].

[0137] (5) Outlier detection

[0138] After denoising and standardization, it is also necessary to further check for outliers in the data. Outlier detection includes the box plot method and the Z-Score method:

[0139] The box plot (Box Plot) identifies outliers based on quartiles.

[0140] The Z-Score method determines whether meteorological data and traffic flow data are outliers. If the absolute value is greater than a certain threshold (such as 3), it is considered an outlier.

[0141] Z-Score outlier detection formula:

[0142]

[0143] If |Z| > 3, then X is considered an outlier.

[0144] Input the processed meteorological data and traffic flow data into the reinforcement learning model. The reinforcement learning model trained through historical data predicts the spatio-temporal patterns of the occurrence of advection fog, and constructs an advection fog risk assessment model by combining a statistical model and a reinforcement learning algorithm. Use the advection fog risk assessment model to calculate the advection fog risk index based on the real-time collected meteorological data and traffic flow data;

[0145] The calculation formula of the advection fog risk index is as follows:

[0146] F = α1T + α2RH + α3V + α4P + α5VIS

[0147] Wherein, α1, α2, α3, α4, α5 are weight coefficients, which are automatically adjusted by the reinforcement learning model through historical data; T represents temperature; RH represents relative humidity; P represents air pressure; VIS represents visibility.

[0148] When the fog patch risk index F exceeds the predetermined threshold F th a fog patch warning is triggered, and the system automatically issues a warning signal.

[0149] The warning signal includes, but is not limited to, real-time release of fog patch warning information through variable speed limit signs, variable message signs (such as VMS), vehicle information systems (such as C-V2X), etc., to remind drivers to slow down or take a detour.

[0150] The reinforcement learning model uses Q-learning or deep Q-network algorithm to optimize traffic management strategies. By learning from real-time data, the optimal traffic response strategy is selected.

[0151] The reinforcement learning model includes a state space, an action space, and a reward function;

[0152] (1) State space S

[0153] The state space S includes current traffic flow data, meteorological data, road visibility, and fog patch risk index;

[0154] Traffic flow data: such as vehicle speed, vehicle distance, lane occupancy rate, traffic flow density, etc.;

[0155] Meteorological data: such as temperature, humidity, wind speed, visibility, etc.;

[0156] Fog patch risk index T: the result calculated by the risk assessment model;

[0157] Road visibility: including traffic flow, accident rate, vehicle spacing, traffic congestion index, etc.

[0158] (3) Action space A

[0159] The action space A includes adjustments to traffic control strategies, and the traffic control strategies also include speed limit adjustment, signal light adjustment, and lane guidance;

[0160] The speed limit adjustment is to dynamically adjust the electronic speed limit sign according to the fog patch risk index and real-time visibility data, specifically:

[0161] When the visibility 200m ≤ VIS < 500m, the speed limit is adjusted to 80 km / h;

[0162] When the visibility is such that 100m ≤ VIS < 200m, the speed limit is adjusted to 60 km / h;

[0163] When the visibility VIS < 100m, the speed limit is 40 km / h and some lanes are closed;

[0164] The signal light adjustment is based on the risk of radiation fog, traffic flow density, vehicle speed, accident frequency, lane occupancy rate, and section capacity. It optimizes the green and red light durations of traffic signal lights, adjusts the signal cycle, and reduces traffic congestion and lane conflicts. Specifically as follows:

[0165] In areas with a high risk of radiation fog, increase the flashing frequency of traffic signal lights to remind drivers to pay attention to decelerating.

[0166] Traffic flow density refers to the number of vehicles passing through a specific section per unit time. When the density is relatively high, it may lead to traffic congestion and reduce the driving speed; conversely, when the density is relatively low, the traffic efficiency is relatively high. According to the real-time traffic flow density data, adjust the green light duration and red light duration of traffic signal lights to optimize the traffic flow.

[0167] Vehicle speed refers to the driving speed of a vehicle within a specific time. A relatively low vehicle speed may indicate traffic congestion or poor road conditions. Especially under the conditions of radiation fog weather, the vehicle speed will be affected by the visibility. According to the traffic flow density and vehicle speed, dynamically adjust the phase and cycle of the signal lights to reduce the vehicle waiting time and congestion;

[0168] Accident frequency refers to the number of traffic accidents occurring within a specific section and time period. A high accident frequency indicates an increased traffic safety risk and requires a more rigorous signal light control strategy to avoid further accidents.

[0169] Lane occupancy rate refers to the proportion of occupied vehicles on each lane. A relatively high lane occupancy rate may lead to slow traffic flow, especially during radiation fog weather or peak traffic hours.

[0170] Section capacity refers to the maximum number of vehicles that a road can carry. If the capacity reaches or exceeds the road's carrying capacity, the traffic flow will be restricted, which will in turn affect the traffic efficiency and safety.

[0171] Lane guidance is when the radiation fog causes the visibility in a local area to be too low, guiding vehicles to drive into a safe lane through lane guidance devices or electronic signs.

[0172] The reward function R is comprehensively evaluated based on traffic flow data, accident rate, and traffic efficiency, and the optimal strategy is selected according to the principle of maximizing the reward. Specifically,

[0173] (1) Design of the reward function

[0174] In this scenario, the reward function R considers the following factors:

[0175] Traffic flow λ: Represents the traffic flow in the current area. The higher it is, the more congested the traffic is.

[0176] Accident rate α: Represents the number of accidents or the probability of accidents occurring within a specific time window. The lower it is, the better the safety.

[0177] Traffic efficiency η: Represents the traffic capacity of the traffic system. The higher it is, the more effectively the traffic management system works.

[0178] Fog patch risk index T: The risk level of fog patches. The higher it is, the greater the risk of fog patches.

[0179] The goal of the reward function is to maximize traffic efficiency and reduce accidents, especially in areas with high fog patch risk. A reward function is constructed, and the calculation formula of the reward function is as follows:

[0180] R = ω1·λ - ω2·α + ω3·η - ω4·T

[0181] In the formula: λ represents the traffic flow (number of vehicles, lane occupancy rate, etc.); α is the accident rate (probability of accidents occurring); η is the traffic efficiency (e.g., number of vehicles passing through per unit time or average vehicle speed); T is the fog patch risk index; ω1, ω2, ω3, and ω4 are the weight coefficients of each factor, indicating the importance of different factors in decision-making.

[0182] (2) Goals of reward function optimization

[0183] Maximize traffic flow (λ): Improve traffic capacity and reduce traffic congestion.

[0184] Minimize accident rate (α): Improve safety and reduce traffic accidents.

[0185] Maximize traffic efficiency (η): Ensure a smooth traffic flow.

[0186] Minimize fog patch risk index (T): Ensure effective control strategies are adopted in areas with high fog patch risk to reduce the probability of accidents occurring.

[0187] To evaluate the effect of each action, a reward function R is constructed, and based on the principle of maximizing rewards through reinforcement learning, the optimal traffic control strategy is automatically selected.

[0188] Step 3: According to the fog risk index and real-time traffic flow data described in Step 2, use the reinforcement learning model to dynamically adjust traffic control strategies, which include speed limits, signal light regulation, and lane guidance. The purpose of dynamically adjusting traffic control strategies is to dynamically adjust traffic control strategies based on real-time traffic flow data and meteorological data through the reinforcement learning model. The reinforcement learning model predicts the fog risk based on real-time data and continuously optimizes decisions through the Q-learning algorithm. Each state-action pair has a Q value, and the system selects the best traffic control strategy according to the Q value, such as speed limit adjustment, signal light regulation, and lane guidance. Eventually, the system can efficiently respond to adverse weather conditions such as fog and ensure traffic safety and smoothness.

[0189] (1) Q-learning algorithm:

[0190] In the reinforcement learning model, Q-learning or Deep Q-Network (DQN) is usually used to optimize decisions. Its purpose is to continuously update and optimize strategies by interacting with the environment. Each state-action pair (s, a) has a Q value Q(s, a), which represents the expected reward that can be obtained after performing action a in state s.

[0191] The Q-value update formula is:

[0192] Q(s,a)←Q(s,a)+α·(r+γ·max a Q(s′,a)-Q(s,a))

[0193] In the formula:

[0194] Q(s,a) is the Q value of taking a certain action in the current state;

[0195] α is the learning rate, which determines the speed of Q-value update;

[0196] r is the reward of the current action (calculated by the reward function R);

[0197] γ is the discount factor, indicating the impact of future rewards;

[0198] max a Q(s',a) is the maximum Q value in state s', representing the expected reward of the future optimal decision.

[0199] (2) Model training and decision-making

[0200] At each time step, the system selects an action according to the current state (i.e., the collected real-time meteorological data and traffic flow data) and updates the Q value through the Q-value update formula.

[0201] The reinforcement learning model continuously helps train and optimize traffic control decisions based on training data, enabling the reinforcement learning model to adjust decisions according to real-time data in practical applications and optimize traffic management. The training data mainly includes historical traffic flow data, meteorological data, visibility data, fog cluster risk index, accident data, historical traffic management strategies and result data, etc.

[0202] (3) Final decision

[0203] Through continuous training and feedback, the reinforcement learning model can ultimately select the optimal traffic control strategies (such as speed limits, signal light regulation, lane guidance, etc.) according to each state. When the fog cluster risk index T exceeds a predetermined threshold, the model will trigger an appropriate warning and ensure smooth traffic, reduce accidents, and improve safety by dynamically adjusting traffic control strategies.

[0204] Step 4, optimize traffic management strategies through real-time feedback of traffic flow data and visibility information, and take emergency response measures in advance when fog clusters form; the emergency response measures include measures such as starting speed limits in advance, adjusting traffic signals, and guiding vehicles into safe lanes.

[0205] Through real-time feedback of traffic data, accident data, and visibility information, the reinforcement learning model continuously optimizes traffic control strategies. The system uses the Q-learning algorithm to update the Q value at each time step, thereby dynamically adjusting strategies such as speed limits, signal lights, and lane guidance according to different traffic conditions and meteorological conditions, ultimately achieving efficient traffic management, ensuring traffic safety, reducing traffic congestion, especially under adverse weather conditions such as fog clusters.

[0206] (1) Real-time data feedback mechanism

[0207] In step 3, the system has adjusted traffic control strategies (such as speed limits, signal light regulation, lane guidance) according to the fog cluster risk index and real-time traffic flow data. In step 4, the system will further optimize and correct the strategies through the following real-time feedback data:

[0208] Traffic flow data: including information such as vehicle speed, vehicle distance, lane occupancy rate, traffic flow density, etc.

[0209] Accident occurrence data: including accident rate, frequency of traffic accidents, etc.

[0210] Visibility information: real-time updated road visibility data.

[0211] The key role of these feedback data is to evaluate the effectiveness of existing traffic management strategies and further adjust and optimize them through reinforcement learning.

[0212] (2) Optimization strategy: Adjust traffic control strategies

[0213] Based on the real-time feedback data, the system optimizes the following traffic control strategies:

[0214] A. Speed limit adjustment: Dynamically adjust the speed limit according to the real-time feedback of visibility and traffic flow. For example, if the visibility suddenly decreases in a fog bank area, the system will automatically lower the speed limit; if the traffic flow density is low, moderately increase the speed limit to keep the traffic flowing smoothly.

[0215] Formula example:

[0216] Speed limit = min(80, speed limit upper bound)

[0217] When the visibility VIS is lower than a certain threshold, the speed limit is dynamically adjusted, such as:

[0218] When VIS < 100m, the speed limit is adjusted to 40 km / h.

[0219] When 100 ≤ VIS < 200m, the speed limit is adjusted to 60 km / h.

[0220] When 200m ≤ VIS < 500m, the speed limit is adjusted to 80 km / h.

[0221] B. Signal light adjustment: Real-time monitor the traffic flow density and accident occurrence, and dynamically adjust the traffic signal light cycle and green / red light duration. For example:

[0222] In high-risk fog bank areas, the green light time of the signal lights can be appropriately extended to reduce waiting time and traffic congestion.

[0223] In accident-prone areas, the signal light cycle can be adjusted to give priority to the passing traffic flow and reduce vehicle stagnation.

[0224] Formula example:

[0225] Signal light adjustment = f(flow density, accident rate, visibility)

[0226] Adjust the green and red light duration according to the real-time flow density, accident rate and visibility. In high-flow areas, increase the green light duration and reduce the red light duration.

[0227] C. Lane guidance: When the visibility is too low and the traffic flow is large on certain sections, the system can guide vehicles into safe lanes through lane guidance devices or signs.

[0228] (3) Combine with the Q-learning algorithm for strategy optimization

[0229] In step 3, the reinforcement learning model (such as Q-learning) has selected a preliminary traffic control strategy through training with historical data and feedback. In step 4, based on real-time traffic flow data, accident data, and visibility feedback, the Q-learning algorithm updates the strategy according to the new data to further optimize the traffic management strategy.

[0230] The optimization process of Q-learning is based on the state-action value (Q-value). Under real-time feedback, the Q-value is updated, thereby gradually improving the accuracy and efficiency of decision-making. The Q-value update formula is:

[0231] Q(s,a)←Q(s,a)+α·(r+γ·max a Q(s′,a)-Q(s,a))

[0232] In the formula:

[0233] s is the current state, that is, real-time data such as traffic flow, accident rate, visibility, etc.;

[0234] a is the current action, that is, strategies such as speed limit, signal light adjustment, lane guidance, etc.;

[0235] r is the reward calculated based on real-time data (such as reducing traffic congestion, reducing accidents, etc.);

[0236] α is the learning rate, which determines the speed of Q-value update;

[0237] γ is the discount factor, indicating the impact of future rewards;

[0238] Q(s,a) is the Q-value of taking a certain action in the current state;

[0239] max a Q(s',a) is the maximum Q-value in state s', representing the expected return of the future optimal decision.

[0240] (4) Final decision

[0241] According to the real-time feedback data, the system continuously adjusts and optimizes the traffic management strategy through the reinforcement learning model. For example, when the real-time traffic flow increases and the visibility decreases, the system may decide to increase the speed limit in the fog area and optimize the signal light cycle to reduce traffic accidents and congestion. In case of poor visibility, the system may guide vehicles into safe lanes through lane guidance to ensure the stability of the traffic flow.

[0242] Through this real-time feedback and optimization process, the traffic management system can make timely adjustments under extreme weather conditions such as fog, ensuring traffic safety and smoothness to the greatest extent and reducing accidents and traffic congestion.

[0243] Step 5: Continuously feedback and update the reinforcement learning model based on the real-time feedback of traffic flow data, accident occurrence data, and visibility information. Through continuous learning and adjustment, the optimal response strategy can be achieved in different fog bank environments. Use the reinforcement learning algorithm to update the model parameters to adjust the specific parameters of the fog bank warning threshold and traffic control strategy. The purpose is to continuously adjust the traffic control strategy according to the real-time traffic flow, accidents, and visibility data using the reinforcement learning algorithm. By updating the parameters of the model through Q-learning or other reinforcement learning algorithms, the system can learn the optimal warning threshold and traffic control strategy, so as to make the optimal response in different fog bank environments.

[0244] (1) Basic framework of reinforcement learning

[0245] In the reinforcement learning model, decisions are learned by interacting with the environment. After each decision, the reinforcement learning model adjusts its strategy according to the feedback from the environment (i.e., the reward signal). The strategy is updated through real-time traffic data, enabling the traffic control system to make the optimal response in the fog bank environment.

[0246] (2) State Space S

[0247] The state space represents the combination of information such as traffic flow, accident occurrence, and visibility. For the response to the fog bank environment, the state is defined as:

[0248] S = {S1, S2, …, S n}

[0249] where each S i contains traffic flow, visibility, accident data, etc. at the current time point.

[0250] (3) Action Space A

[0251] The action space represents the control strategies that the system can take. Possible actions include: adjusting the traffic light cycle, changing the road speed limit, issuing a fog bank warning, and dynamically adjusting traffic flow control (such as restricting entry).

[0252] It is defined as:

[0253] A = {α1, α2, …, α n}, and each α j is a control strategy.

[0254] (4) Reward Function R

[0255] The reward function evaluates the performance of the system based on the actions selected by the model. For example, if the selected strategy makes the traffic flow smoothly, reduces accidents and delays, a positive reward is given; if there is a traffic jam or an accident, a negative reward is given.

[0256] For example:

[0257] R(S t ,a t ) represents the reward obtained at time t when taking action a t in state S t .

[0258] (5) Update steps of the reinforcement learning algorithm

[0259] Common reinforcement learning algorithms include Q-learning, SARSA, Deep Q-Network (DQN), etc. In this step, the reinforcement learning model updates its policy based on the feedback (reward signal) obtained. Assuming the Q-learning algorithm is used, the model parameters can be adjusted through the following update formula:

[0260] Q-value update formula:

[0261] Q(S t ,a t ) ← Q(S t ,a t ) + α · [R(S t ,a t ) + γ max a Q(S t+1 ,a) - Q(S t ,a t )]

[0262] In the formula:

[0263] Q(S t ,a t ) is the Q-value of taking action a t in state S t ;

[0264] α is the learning rate, which controls the step size of each update;

[0265] γ is the discount factor, which controls the weight of future rewards;

[0266] R(S t ,a t ) is the immediate reward obtained after taking action a t in state S t ;

[0267] max a Q(S t+1 ,a) is the state St+1 The maximum Q-value of all the following actions represents the future rewards expected by the model.

[0268] (6) Warning threshold and adjustment of traffic control strategy parameters

[0269] In step 5, through the process of reinforcement learning, the model will adjust the warning threshold of the freeway fog warning and the specific parameters of the traffic control strategy. For example, the reinforcement learning algorithm may optimize the following parameters: the warning threshold of the fog visibility, the upper limit of traffic flow control, the adjustment of the speed limit policy, and the control strategy of traffic lights. These parameters can be dynamically updated through the parameter adjustment process of the reinforcement learning model. By continuously interacting with the actual traffic environment, the model can optimize these strategies to maximize traffic safety and smoothness in the case of fog.

[0270] Through this adaptive learning and optimization mechanism, the system can better cope with the fog challenges in different environments and improve traffic safety and efficiency.

[0271] A system for implementing the above-mentioned freeway adaptive fog warning and response control method includes

[0272] A data acquisition system for collecting meteorological data and traffic flow data of the freeway in real time;

[0273] A data processing module for denoising, standardizing, and normalizing the collected meteorological data and traffic flow data;

[0274] A reinforcement learning model for predicting the spatio-temporal pattern of fog occurrence based on the processed meteorological data and traffic flow data and calculating the fog risk index;

[0275] A warning and response control system for triggering a fog warning and dynamically adjusting the traffic control strategy when the fog risk index exceeds a predetermined threshold;

[0276] The warning and response control system includes:

[0277] A warning signal release device for real-time releasing fog warning information through variable speed limit signs, variable message signs, and in-vehicle information systems;

[0278] A dynamic traffic control device for dynamically adjusting electronic speed limit signs, optimizing traffic signal cycles, and guiding vehicles into safe lanes according to real-time visibility data.

[0279] An adaptive learning module for continuously updating the reinforcement learning model and dynamically adjusting the specific parameters of the fog warning threshold and traffic control strategy according to the real-time feedback of traffic flow data and visibility information.

[0280] The adaptive learning module includes:

[0281] A model update unit for updating the reinforcement learning model according to the real-time feedback traffic flow data and visibility information;

[0282] A threshold adjustment unit for dynamically adjusting the warning threshold of advection fog;

[0283] A policy optimization unit for optimizing the specific parameters of the traffic control policy.

[0284] The system of this embodiment can adaptively adjust the warning threshold of advection fog and the traffic control policy to adapt to the characteristics of advection fog in different regions and seasons. For example, the occurrence pattern of advection fog in winter may be different from that in summer. Suppose on a certain highway, advection fog in winter mainly appears at night and early morning, and the visibility drops rapidly; while advection fog in summer mostly appears in the early morning and evening, and lasts for a longer time. Through real-time data feedback, the system learns these seasonal differences and dynamically adjusts the warning threshold of advection fog and the traffic control policy. For example, in winter, the system may issue warning signals in advance and adjust the speed limit signs more frequently; in summer, the system may extend the warning time and optimize the lane guiding policy.

[0285] The above-mentioned highway adaptive advection fog warning and response control method and system are applied to highways, urban expressways, bridges and tunnels.

Claims

1. A highway adaptive fog early warning and response control method, characterized in that: The steps include: Step 1: collect meteorological data and traffic flow data on the highway in real time. The meteorological data includes temperature data, air humidity data, wind speed data and visibility data; the traffic flow data includes vehicle speed data, vehicle distance data, lane occupancy rate data and traffic flow density data; Step 2, denoising, standardizing and normalizing the meteorological data and traffic flow data collected in step 1, inputting the processed meteorological data and traffic flow data into the reinforcement learning model, predicting the spatiotemporal pattern of fog cluster occurrence through the reinforcement learning model obtained by historical data training, and building a fog cluster risk assessment model by combining the statistical model with the reinforcement learning algorithm, and using the fog cluster risk assessment model to calculate the fog cluster risk index based on the meteorological data and traffic flow data collected in real time; When the fog risk index exceeds the preset threshold, the fog warning is triggered and the system automatically issues a warning signal; Step 3, according to the fog risk index and real-time traffic flow data described in step 2, dynamically adjust the traffic control strategy using a reinforcement learning model, the traffic control strategy including speed limit, traffic light adjustment and lane guidance; Step 4: Optimize traffic management strategies through real-time feedback of traffic flow data, accident data, and visibility information, and take emergency response measures in advance when fog forms; Step 5: Based on the real-time feedback of traffic flow data, accident data and visibility information, the reinforcement learning algorithm is used to update the model parameters to adjust the specific parameters of the fog warning threshold and the traffic control strategy.

2. The method according to claim 1, characterized in that: The temperature data described in step 1 is collected in real time through a temperature sensor, the air humidity data is collected in real time through a humidity sensor, the wind speed data is collected in real time through a wind speed sensor, the visibility data is collected in real time through a visibility sensor, the vehicle speed data is monitored in real time by a road radar or a vehicle networking device, the vehicle distance data is obtained by obtaining the distance between vehicles through video monitoring or radar ranging equipment, the lane occupancy rate data is monitored by a lane sensing device to monitor the vehicle occupancy of each lane, and the traffic flow density data is calculated by a traffic flow detection device to count the number of vehicles passing through a specific road section in real time per unit time.

3. The method according to claim 1, characterized in that The denoising process in step 2 includes the following steps: Step 21, automatically select a denoising method based on the characteristics of meteorological data and traffic flow data: For traffic flow data, the moving average method of the smoothing filter method is used to remove noise, calculate the average value in the data window and smooth out the fluctuations. The calculation formula is as follows: Where: Yt is the smoothed data, Xi is the original data, N is the window size, usually an odd number; For cases with obvious outliers, the median filter method is used to remove noise and calculate the middle value of the data to replace the original data point. The calculation formula is as follows: Where: Yt is the denoised data, Indicates time point The data is at the left end of the window; Indicates time point data; express The data is at the right end of the window; For data with complex frequencies and large fluctuations, the wavelet denoising method is used to decompose the signal through wavelet transform to remove high-frequency noise; Step 22, optimizing the denoising strategy based on the expected effect: according to the impact of different denoising methods on the meteorological data and traffic flow data in step 21, selecting a suitable denoising method to remove noise without losing key information; Step 23, combining the denoising method of step 21 with the reinforcement learning model to form an adaptive feedback mechanism to dynamically select the most appropriate denoising strategy.

4. The method according to claim 1, characterized in that: The calculation formula for the normalization process described in step 2 is as follows: Where: X' is the standardized data; X is the original data; μ is the mean of the data, and σ is the standard deviation of the data.

5. The method according to claim 1, characterized in that: The normalization process in step 2 automatically selects the Min-Max normalization method or the Z-score normalization method according to the characteristics of each meteorological data and traffic flow data; The Min-Max normalization method is used when the data distribution is relatively uniform and there are no significant outliers. The calculation formula of the Min-Max normalization method is as follows: In the formula: X' is the normalized data, X is the original data, Xmin and Xmax are the minimum and maximum values ​​of the data respectively; The Z-score normalization method is used for data with large fluctuations, outliers, or data that does not satisfy uniform distribution. The calculation formula of the Z-score normalization method is as follows: Where: μ is the mean of the data, range(X) is the range of the data, that is, the maximum value minus the minimum value.

6. The method according to claim 1, characterized in that The step 2 also includes outlier detection, which includes a box plot and a Z-Score method, wherein the box plot identifies outliers based on four digits; The Z-Score method determines whether the meteorological data and traffic flow data are outliers. If the absolute value is greater than a certain threshold, it is considered an outlier. The calculation formula is as follows: If |Z|>3, then X is considered an outlier, μ is the mean of the data, and σ is the standard deviation of the data.

7. The method according to claim 1, characterized in that The reinforcement learning model described in step 2 adopts Q-learning or deep Q network algorithm, and the reinforcement learning model includes state space, action space and reward function; The state space includes current traffic flow data, meteorological data, road visibility and fog risk index; The calculation formula of the fog risk index is as follows: F=α1T+α2RH+α3V+α4P+α5VIS Where α1, α2, α3, α4, and α5 are weight coefficients, which are automatically adjusted by the reinforcement learning model based on historical data; T represents temperature; RH represents relative humidity; and P represents air pressure. VIS stands for visibility; The action space includes adjustments to traffic control strategies, including speed limit adjustments, signal light adjustments, and lane guidance; The speed limit adjustment is to dynamically adjust the electronic speed limit sign according to the fog risk index and real-time visibility data. Specifically, when the visibility is 200m≤VIS<500m, the speed limit is adjusted to 80km / h. When visibility is 100m≤VIS<200m, the speed limit is adjusted to 60km / h; When visibility VIS < 100m, the speed limit is 40km / h and some lanes are closed; The traffic light adjustment is to optimize the green and red light durations of traffic lights and adjust the traffic light cycle according to the risk of fog, traffic flow density, vehicle speed, accident frequency, lane occupancy rate and road section capacity. In areas with high risk of fog, the flashing frequency of traffic lights is increased to remind drivers to slow down. According to the traffic flow density and vehicle speed, the phase and cycle of traffic lights are dynamically adjusted to reduce vehicle waiting time and congestion. The lane guidance is to guide the vehicle into a safe lane through a lane guidance device or electronic sign when the visibility in a local area is too low due to fog; The reward function is comprehensively evaluated based on traffic flow data, accident rate and traffic efficiency, and the optimal strategy is selected according to the principle of maximizing rewards. The calculation formula of the reward function is as follows: R=ω1·λ-ω2·α+ω3·η-ω4·T Where: R represents the reward function; λ represents the traffic flow in the current area, which includes the number of vehicles and lane occupancy rate. The higher the traffic flow, the more congested it is. α represents the number of accidents or the probability of accidents that occur within a specific time window. η is the capacity of the traffic system, that is, the number of vehicles passing per unit time or the average speed; T is the fog risk index; ω1, ω2, ω3 and ω4 are the weight coefficients of each factor, indicating the importance of different factors in decision-making; The warning signal includes but is not limited to issuing warning information in real time through variable speed limit signs, variable information display screens and on-board information systems to remind drivers to slow down or take detour measures.

8. The method according to claim 1, characterized in that: The emergency response measures described in step 4 include measures such as early activation of speed limits, adjustment of traffic signals, and guidance of vehicles into safe lanes.

9. A highway adaptive fog early warning and response control system implementing the method described in claims 1 to 8, characterized in that: include Data collection system, used to collect meteorological data and traffic flow data of expressways in real time; Data processing module, used for denoising, standardizing and normalizing the collected meteorological data and traffic flow data; A reinforcement learning model is used to predict the spatiotemporal pattern of fog occurrence based on processed meteorological data and traffic flow data, and calculate the fog risk index; The early warning and response control system is used to trigger a fog warning and dynamically adjust the traffic control strategy when the fog risk index exceeds a predetermined threshold; The adaptive learning module is used to continuously update the reinforcement learning model based on real-time feedback of traffic flow data, accident data and visibility information, and dynamically adjust the specific parameters of the fog warning threshold and traffic control strategy.

10. The system according to claim 8, characterized in that The early warning and response control system comprises: Warning signal issuing device, used to issue fog warning information in real time through variable speed limit signs, variable information display screens and vehicle information systems; Dynamic traffic control devices to dynamically adjust electronic speed limit signs, optimize traffic light cycles and guide vehicles into safe lanes based on real-time visibility data; The adaptive learning module includes: A model updating unit, used to update the reinforcement learning model according to the real-time feedback of traffic flow data and visibility information; A threshold adjustment unit, used to dynamically adjust the fog warning threshold; The strategy optimization unit is used to optimize the specific parameters of the traffic control strategy.

Images