Traffic state estimation and queuing state discrimination method based on fuzzy traffic wave model

Abstract

The invention belongs to the technical field of intelligent traffic, and discloses a traffic state estimation and queuing state discrimination method based on a fuzzy traffic wave model, comprising the following steps: S1, collecting and preprocessing ETC gantry data, thunder-vision fusion trajectory data and artificial report event data, and extracting traffic flow parameters and event features after preprocessing; s2, fuzzy parameters are introduced to improve a traditional traffic wave model, the traffic wave velocity is calculated through the fuzzy parameters, wave front position transmission based on a state transition model is researched, and a fuzzy traffic wave model is constructed; s3, a VMD-GA-ConvLSTM network is constructed to carry out traffic state estimation; s4, determining a traffic state membership degree by adopting a GWO-FCM clustering algorithm, and judging a vehicle queuing state through an SVM classifier; and S5, improving the fuzzy traffic wave model in the network connection environment, and dynamically correcting the model from two dimensions of road section macroscopic characteristics and vehicle microscopic behaviors. The method provided by the invention can provide theoretical and technical support for accurate perception and dynamic regulation and control of the highway traffic situation.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent transportation, and in particular relates to a method for estimating traffic status and distinguishing queue status based on a fuzzy traffic wave model. Background Art

[0002] With the rapid development of social economy and the acceleration of urbanization, highways, as a core component of modern transportation networks, carry the growing travel demand. However, high-frequency use has led to increasingly acute contradictions between traffic supply and demand, and congestion problems have occurred frequently. In particular, when abnormal events such as traffic accidents, sudden large traffic flows, and road construction occur, the dynamic imbalance of traffic flow is particularly significant. Traditional traffic flow theory is based on deterministic models (such as the three-parameter model), which assume that traffic flow parameters are fixed values ​​and lack a detailed description of the dynamic evolution of the wavefront position and the traffic conditions on both sides. Such models usually simplify vehicle state changes to instantaneous behavior, ignoring the temporal and spatial gradual characteristics of vehicle behavior in actual traffic flow, resulting in limited applicability in complex traffic scenarios.

[0003] In recent years, the rapid development of connected vehicle (CV) technology has provided a new data dimension for traffic flow analysis. However, traditional models face two major challenges in mixed human-machine driving environments: first, the correlation mechanism between microscopic single-vehicle behavior and macroscopic traffic flow state has not been fully explored; second, existing models lack the ability to integrate multi-source heterogeneous data (such as ETC gantry data and radar-visual fusion trajectory data), making it difficult to effectively address data noise and spatiotemporal asynchrony. For example, patent document CN117373250-A proposes a method for analyzing traffic queue evolution based on a traffic flow wave model based on time occupancy. However, its definition of the wavefront position still relies on static assumptions and fails to fully utilize the complementary information between the macroscopic traffic flow of a road section and the microscopic trajectory of a single vehicle, resulting in insufficient prediction accuracy for dynamic traffic events. Existing research has mostly focused on traffic state analysis based on a single data source or a single scale (macro or micro), lacking collaborative optimization from the perspective of cyber-physical fusion. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a traffic state estimation and queue state discrimination method based on a fuzzy traffic wave model, aiming to solve the problem that existing methods are unable to fully utilize the complementary information of the macroscopic traffic flow of the road section and the microscopic trajectory of a single vehicle, resulting in low model prediction accuracy.

[0005] The present invention provides a method for estimating traffic status and distinguishing queue status based on a fuzzy traffic wave model, comprising the following steps:

[0006] S1. Collect and preprocess ETC gantry data, radar-visual fusion trajectory data, and manually reported event data, and extract traffic flow parameters and event characteristics after preprocessing;

[0007] Among them, trajectory data is collected by radars and cameras installed on highways. The trajectory data includes the vehicle's position, speed, and acceleration;

[0008] S2. Introduce fuzzy parameters to improve the traditional traffic wave model, calculate the traffic wave velocity using fuzzy parameters, study the wavefront position transfer based on the state transition model, and construct a fuzzy traffic wave model;

[0009] S3. Construct a VMD-GA-ConvLSTM network for traffic state estimation;

[0010] S4. Use the GWO-FCM clustering algorithm to determine the traffic state membership and use the SVM classifier to distinguish the vehicle queue status;

[0011] S5. Improve the fuzzy traffic wave model in the connected environment and dynamically modify the model from two dimensions: the macroscopic characteristics of the road section and the microscopic behavior of the vehicle.

[0012] Furthermore, in step S1, the specific content of the preprocessing is:

[0013] a.ETC data preprocessing:

[0014] Filter vehicle data collected from highway gantries and extract data fields that can reflect traffic flow status information;

[0015] De-noise ETC data to remove invalid and redundant data;

[0016] Perform spatiotemporal matching on the cleaned ETC data, match the vehicle's travel records on different road sections based on vehicle codes and passing times, and calculate the flow rate, density, and travel speed of each road section;

[0017] b. Trajectory data preprocessing:

[0018] The trajectory data is cleaned and filtered to remove abnormal data points; the trajectory data is integrated according to the time series to form the trajectory records of each vehicle in different time periods, and the corresponding road section density is extracted.

[0019] Furthermore, step S2 includes the following sub-steps:

[0020] S2.1 Analyze the temporal and spatial distribution characteristics of vehicle headway and speed based on radar-visual fusion trajectory data;

[0021] I. Taylor expansion of the Underwood model is as follows:

[0022]

[0023] Where v and k are the average vehicle speed and density of the road section respectively; v f represents the free flow speed; k j represents the blocking density;

[0024] II. Design the membership function, which is expressed as follows:

[0025]

[0026] Where d represents the vertical distance from the vehicle's current position to the wavefront, d = 0 corresponds to the wavefront position, and d > 0 represents the position behind the congestion; d0 represents the critical distance for the slowdown area change, which is taken as 250m in this study; μ N (d) is the corresponding proximal membership function, μ F (d) is the distal membership function

[0027] III. Calculate the headway h and speed v using the following expressions:

[0028]

[0029] Where, v b is the average speed of the traffic flow in the stop-and-go state; f v (d) represents the slowdown adjustment function; Q F is the upstream flow input;

[0030] S2.2 calibrates the speed and congestion tail distance of the improved Underwood model using simulated annealing method;

[0031] Here are the results:

[0032]

[0033] S2.3 Calculate the traffic wave speed using the following expression:

[0034]

[0035] S2.4 Define the traffic wave propagation process in a discretized framework, and define its spatiotemporal state variables as:

[0036]

[0037] Where, is the wavefront position at the nth time step; is the density upstream of the wave front; is the density downstream of the wave front;

[0038] The updating formula of the wavefront position is:

[0039]

[0040] The update formula for the downstream density of the wavefront is:

[0041]

[0042] Where, Indicates the flow rate entering the section from upstream; Expressed as the interface flow at the wavefront position; It represents the flow rate leaving the road section downstream; Δt represents the discrete time step; Δx represents the discrete space step.

[0043] Furthermore, step S3 includes the following sub-steps:

[0044] S3.1 adopts the LWR model in velocity expression and combines it with the improved Underwood model for theoretical derivation;

[0045] The physical loss function of the improved Underwood model when no traffic incident occurs is expressed as:

[0046]

[0047] The physical loss function of the improved Underwood model in the event of a traffic incident is:

[0048]

[0049] Where, L phy is the corresponding physical model loss; For k The density estimate at N c The total number of road sections divided; N o The number of road sections divided into congested sections in case of traffic incidents;

[0050] S3.2 establishes a traffic state estimation model based on CV real-time data and ETC detection data, and constructs a VMD-GA-ConvLSTM combined prediction model;

[0051] First, variational mode decomposition (VMD) is performed on the traffic state sequence. Then, a ConvLSTM network architecture is constructed. Finally, the optimized ConvLSTM network is used to perform parallel prediction of each IMF component. The traffic state estimation value is obtained through modal reconstruction and denormalization.

[0052] The loss function of the VMD-GA-ConvLSTM neural network module based on data-driven flow is as follows:

[0053]

[0054] Under the PIML framework, the total loss function calculation expression for highway traffic state estimation is:

[0055] L=μL ML +(1-μ)L phy

[0056] Where, For the road section The actual average speed at For the neural network model The estimated value of velocity at ; μ represents the weighting coefficient of the machine learning loss function.

[0057] Furthermore, step S4 includes the following sub-steps:

[0058] S4.1 introduces the Grey Wolf Optimization Algorithm (GWO) to optimize and improve FCM;

[0059] Through the membership constraint The objective function of the FCM algorithm is transformed using the Lagrange multiplier method to obtain the unconstrained objective function:

[0060]

[0061] Where m is the fuzzy weighted index, m∈(1,+∞); λ j represents the Lagrange multiplier corresponding to the j-th sample;

[0062] S4.2 maps the original feature space to a high-dimensional space through the kernel function. The corresponding classification function is:

[0063]

[0064] Where a i is the Lagrange multiplier coefficient; K is the conversion kernel function; x∈R n ,y∈{1,-1},y∈{1,-1} represents the training sample set, x i is the i-th eigenvector, y i is the class label, when y i =1 is a positive example, y i =-1 is a negative example, l is the number of samples, and n is the dimension of x.

[0065] Furthermore, step S5 includes the following sub-steps:

[0066] S5.1 is based on the fuzzy traffic wave model and uses the traffic flow data provided by the upstream and downstream ETC gantries to calculate the initial wavefront position estimate;

[0067] S5.2 processes CV data in parallel, extracts single-vehicle feature values, and applies a physical fusion method for single-vehicle and road-section state discrimination information to determine a state judgment value based on the single vehicle and road section;

[0068] S5.3 Based on the physical constraints in the cyber-physical fusion queue state determination method, continuity correction is performed on the information processing results to obtain wavefront position estimates based on individual vehicles and road sections;

[0069] S5.4 combines the initial wavefront position obtained by the information processing layer with the CV single-vehicle and road section discrimination results, and integrates the dual constraints of the information layer and the physical layer through a physical constraint correction method based on continuity rules to finally obtain the corrected wavefront position estimate.

[0070] Beneficial effects:

[0071] This paper proposes a hybrid traffic flow state estimation and queue state discrimination method based on a fuzzy traffic wave model. This method can deeply integrate macroscopic and microscopic traffic data, dynamically modify wavefront propagation patterns, and adapt to changes in the penetration rate of connected vehicles. This paper aims to improve the traffic wave model through fuzzy parameters and, combined with a physical information machine learning framework, overcome the limitations of traditional models and provide theoretical and technical support for the precise perception and dynamic control of highway traffic conditions.

[0072] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 This is a flow chart of a method for estimating traffic status and distinguishing queue status based on a fuzzy traffic wave model according to the present invention;

[0074] Figure 2 Flowchart of the cyber-physical fusion method for wavefront correction in a networked environment. DETAILED DESCRIPTION

[0075] To make the technical solutions, advantages, and purposes of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0076] like Figure 1 and Figure 2 As shown, the present invention provides a method for estimating traffic status and distinguishing queue status based on a fuzzy traffic wave model, comprising the following steps:

[0077] S1. Data preprocessing: Collect and preprocess ETC gantry data, radar-visual fusion trajectory data, and manually reported event data, extract traffic flow parameters and event features, and use them as model input and labels.

[0078] ETC data preprocessing:

[0079] By screening the vehicle data collected from the highway gantries, key data fields that can effectively reflect the traffic flow status information are extracted.

[0080] De-noising is performed on ETC data to remove invalid and redundant data, such as invalid vehicle model data and redundant data that passes through the same gantry multiple times in a very short time interval.

[0081] The cleaned ETC data is matched in time and space, and the vehicle's travel records on different sections are matched according to the vehicle code and passing time, and the flow and density of each section and the travel speed of each vehicle are calculated.

[0082] Trajectory data preprocessing:

[0083] Vehicle trajectory data, captured by radar and cameras installed on highways, includes information such as vehicle location, speed, and acceleration. This trajectory data is cleaned and filtered to remove abnormal data points, such as those caused by equipment failure or environmental interference. The trajectory data is then aggregated into time series to generate a track record for each vehicle over different time periods, and the density of each vehicle's road section is extracted.

[0084] S2. Analyze the temporal and spatial distribution characteristics of vehicle headway and speed based on radar-visual fusion trajectory data. Fuzzy parameters are introduced to improve the traditional traffic wave model. Traffic wave speed is calculated based on the fuzzy parameters. The wavefront position transfer based on the state transition model is studied to construct a fuzzy traffic wave model. This includes the following sub-steps:

[0085] Step 2.1: The basic graphical expression of the Underwood model is as follows:

[0086]

[0087] Where v and k are the average vehicle speed and density of the road section respectively; v f represents the free flow speed; k j represents the blocking density.

[0088] However, when the velocity approaches 0, the Underwood model does not produce a solution for the blocking density. To address this problem, Ardekani et al. proposed a Taylor series expansion, and obtained the Taylor expansion containing the first three terms as follows:

[0089]

[0090] The theoretical framework of random slowing down in the NaSch model is introduced and fuzzified into the gradual deceleration behavior of the vehicle when approaching the wavefront area. Considering that the fuzzy slowing down parameter α should be characterized as a spatial correlation function with the distance d from the wavefront position, the study defines the fuzzy slowing down parameter α as a function α(d) of the wavefront distance d, satisfying the physical law that "the closer to the wavefront, the higher the degree of slowing down." The specific implementation process is as follows:

[0091] Design the membership function and divide d into two fuzzy sets proximal A near and remote A far , corresponding to high slowdown and low slowdown areas respectively. Aiming at the change phenomenon in traffic phenomena, the SIGMOD function is selected as the membership function of this paper. Its membership function is as follows:

[0092]

[0093] Where d represents the vertical distance from the vehicle's current position to the wavefront, d = 0 corresponds to the wavefront position, and d > 0 represents the position behind the congestion; μ N (d) is the corresponding proximal membership function, μ F (d) is the distal membership function.

[0094] Through parameter coupling, the weighted velocity expression is obtained as follows:

[0095] v(d)=v b +(v f -v b )f v (d)

[0096] Considering that the vehicle flow Q in the road section is constant, combined with the flow conservation model, Then the corresponding relationship between h and v can be obtained as follows:

[0097]

[0098] Among them, v b is the average speed of the traffic flow in the stop-and-go state; Q F Enter the upstream flow rate.

[0099] Finally, after establishing the fuzzy slowing rule for traffic flow speed, we can construct a relationship between the speed density of vehicles belonging to the approaching road section. The corresponding expressions for the headway and speed are as follows:

[0100]

[0101] Step 2.2: Use simulated annealing to calibrate the speed and congestion tail distance of the improved Underwood model. The results are as follows:

[0102]

[0103] Step 2.3: Based on the improved Underwood model, first decouple the velocity and obtain the explicit expression of the velocity:

[0104]

[0105] From the relationship between the three parameters, we can get:

[0106]

[0107] At the same time, according to the definition of traffic wave speed, we can get the wave speed v w The calculation formula is:

[0108]

[0109] Step 2.4: Define the traffic wave propagation process within the discretization framework, and define its spatiotemporal state variables as:

[0110]

[0111] in, is the wavefront position at the nth time step; is the density upstream of the wave front; is the density downstream of the wave front.

[0112] After obtaining the spatiotemporal variables of the traffic wave propagation process, the position update formula can be obtained as follows:

[0113]

[0114] At the same time, the downstream and upstream densities of the wavefront are updated, and the update calculation formula is:

[0115] Congested areas (affected by vehicle accumulation):

[0116]

[0117] Approaching area (affected by free-flow vehicle compression):

[0118]

[0119] in, is the interface flow at the wavefront position.

[0120] S3. Considering the potential noise interference and heterogeneous characteristics of multi-source data, the PIPM framework is used to embed the Underwood dynamic equations constructed in step 2) as physical constraints into the neural network architecture, and a VMD-GA-ConvLSTM network is constructed for traffic state estimation.

[0121] Step 3.1: Construct the PIML strategy through the physical information loss function, the loss function L of the data-driven term ML , which is defined as the mean square error (MSE) between the predicted result and the actual result, and is defined as follows:

[0122]

[0123] Among them, v x i is the speed value of the road section x at the i-th moment, is the estimated speed value of the road segment x at the i-th moment of the corresponding model.

[0124] The LWR model in velocity expression is used in combination with the improved Underwood model for theoretical derivation. The specific process is as follows:

[0125]

[0126] Considering the LWR model represented by density and combined with the improved Underwood model, when no traffic incident occurs, the corresponding constraints should be:

[0127]

[0128] In the event of a traffic incident, the corresponding derivation process is as follows:

[0129]

[0130] Based on the above derivation, in the framework of physical information learning, the physical loss function of the improved Underwood model when no traffic incident occurs can be expressed as:

[0131]

[0132] Accordingly, when a traffic incident occurs, the corresponding physical loss function is:

[0133]

[0134] Step 3.2: Build a traffic state estimation model based on real-time CV data and ETC detection data, and construct a VMD-GA-ConvLSTM combined prediction model. First, perform variational mode decomposition (VMD) on the traffic state sequence, balancing decomposition accuracy and smoothness using a penalty factor. Next, build a ConvLSTM network architecture. Finally, use the optimized ConvLSTM network to perform parallel predictions on each IMF component, and obtain traffic state estimates through modal reconstruction and denormalization.

[0135] The loss function of the VMD-GA-ConvLSTM neural network module based on data-driven flow is as follows:

[0136]

[0137] Under the PIML framework, the total loss function calculation expression for highway traffic state estimation is:

[0138] L=μL ML +(1-μ)L phy

[0139] Among them, L phy is the corresponding physical model loss; For k The density estimate at N c The total number of road sections divided; N o The number of road sections divided into congested sections in case of traffic incidents; For the road section The actual average speed; For the neural network model The estimated value of speed.

[0140] S4. The vehicle state recognition method based on cyber-physical fusion first collects key state indicators such as headway, speed, and vehicle type, and inputs them into the GWO-FCM clustering algorithm to obtain vehicle traffic state membership values. These membership values ​​are then passed to the SVM model for identification, and the results are corrected based on the physical constraints of traffic flow continuity.

[0141] Step 4.1: Introduce the Grey Wolf Optimization (GWO) algorithm to optimize and improve FCM. The objective function of the FCM algorithm is as follows:

[0142]

[0143] Among them, U is the membership matrix; V represents the class center of c categories; V={v1,v2,...,v n}; m is the fuzzy weighted index, m∈(1,+∞).

[0144] Through the membership constraint The objective function is transformed using the Lagrange multiplier method to obtain the unconstrained objective function:

[0145]

[0146] Derivative and solution of the above formula can be obtained to get the membership center u ij and the corresponding cluster center v i The update formula is:

[0147]

[0148] Step 4.2: The SVM algorithm is based on the principle of maximum class separation and searches for the optimal hyperplane to divide samples in the feature space. Its basic mathematical expression is:

[0149] {x i ,y i |i=1,2,...,l},x∈R n ,y∈{1,-1}

[0150] Among them, y∈{1,-1} represents the training sample set, x i is the i-th eigenvector, y i is the class label, when y i =1 is a positive example, y i =-1 is a negative example, l is the number of samples, and n is the dimension of x.

[0151] Taking into account the nonlinear characteristics of the traffic queue state recognition problem, the original feature space is mapped to a high-dimensional space through the kernel function. The corresponding classification function is:

[0152]

[0153] Among them, a i is the Lagrange multiplier coefficient; K is the conversion kernel function.

[0154] S5. Based on the state estimation model constructed in steps S3 and S4, the fuzzy traffic wave model in the connected environment is improved, and the model is dynamically modified from two dimensions: the macroscopic characteristics of the road section and the microscopic behavior of the vehicle. This specifically includes the following sub-steps:

[0155] Step 5.1: Based on the fuzzy traffic wave model constructed in step S2, the traffic flow data provided by the upstream and downstream ETC gantries are used to calculate the initial wavefront position estimate to obtain the initial wavefront position estimate;

[0156] Step 5.2: Process the CV data in parallel, extract the single-vehicle feature value, and apply the physical fusion method of single-vehicle and road segment state judgment information to determine the state judgment value based on the single vehicle and road segment;

[0157] Step 5.3: Based on the physical constraints in the cyber-physical fusion queue state determination method, the information processing results are corrected for continuity to obtain the wavefront position estimate based on the individual vehicle and road segment.

[0158] Step 5.4: Combine the initial wavefront position obtained from the information processing layer with the CV single-vehicle and road section discrimination results. Through the physical constraint correction method based on continuity rules, the dual constraints of the information layer and the physical layer are integrated to finally obtain the corrected high-precision wavefront position estimate.

[0159] It is hereby stated that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for estimating traffic status and distinguishing queue status based on fuzzy traffic wave model, characterized in that: The following steps are involved: S1. Collect and preprocess ETC gantry data, radar-visual fusion trajectory data, and manually reported event data, and extract traffic flow parameters and event characteristics after preprocessing; Among them, trajectory data is collected by radars and cameras installed on highways. The trajectory data includes the vehicle's position, speed, and acceleration; S2. Introduce fuzzy parameters to improve the traditional traffic wave model, calculate the traffic wave velocity using fuzzy parameters, study the wavefront position transfer based on the state transition model, and construct a fuzzy traffic wave model; S3. Construct a VMD-GA-ConvLSTM network for traffic state estimation; S4. Use the GWO-FCM clustering algorithm to determine the traffic state membership and use the SVM classifier to distinguish the vehicle queue status; S5. Improve the fuzzy traffic wave model in the connected environment and dynamically modify the model from two dimensions: the macroscopic characteristics of the road section and the microscopic behavior of the vehicle.

2. The method for estimating traffic status and distinguishing queue status based on fuzzy traffic wave model according to claim 1, characterized in that: In step S1, the specific content of the preprocessing is: a.ETC data preprocessing: Filter vehicle data collected from highway gantries and extract data fields that can reflect traffic flow status information; De-noise ETC data to remove invalid and redundant data; Perform spatiotemporal matching on the cleaned ETC data, match the vehicle's travel records on different road sections based on vehicle codes and passing times, and calculate the flow rate, density, and travel speed of each road section; b. Trajectory data preprocessing: The trajectory data is cleaned and filtered to remove abnormal data points; the trajectory data is integrated according to the time series to form the trajectory records of each vehicle in different time periods, and the corresponding road section density is extracted.

3. The method for estimating traffic status and distinguishing queue status based on fuzzy traffic wave model according to claim 2, characterized in that: The step S2 includes the following sub-steps: S2.1 Analyze the temporal and spatial distribution characteristics of vehicle headway and speed based on radar-visual fusion trajectory data; I. Taylor expansion of the Underwood model is as follows: Where v and k are the average vehicle speed and density of the road section respectively; v f represents the free flow speed; k j represents the blocking density; II. Design the membership function, which is expressed as follows: Where d represents the vertical distance from the vehicle's current position to the wavefront, d = 0 corresponds to the wavefront position, and d > 0 represents the position behind the congestion; d0 represents the critical distance for the slowdown area to change; μ N (d) is the corresponding proximal membership function, μ F (d) is the distal membership function III. Calculate the headway h and speed v using the following expressions: Where, v b is the average speed of the traffic flow in the stop-and-go state; f v (d) represents the slowing down adjustment function, whose value is the membership degree of the corresponding congestion state; Q F is the upstream flow input; S2.2 calibrates the speed and congestion tail distance of the improved Underwood model using simulated annealing method; Here are the results: S2.3 Calculate the traffic wave speed using the following expression: S2.4 Define the traffic wave propagation process in a discretized framework, and define its spatiotemporal state variables as: Where, is the wavefront position at the nth time step; is the density upstream of the wave front; is the density downstream of the wave front; The updating formula of the wavefront position is: The update formula for the downstream density of the wavefront is: Where, Indicates the flow rate entering the section from upstream; Expressed as the interface flow at the wavefront position; Indicates the flow rate leaving the link downstream; Δt represents the discrete step length in time; Δx represents the discrete step length in space.

4. The method for estimating traffic status and distinguishing queue status based on fuzzy traffic wave model according to claim 3, characterized in that: The step S3 includes the following sub-steps: S3.1 adopts the LWR model in velocity expression and combines it with the improved Underwood model for theoretical derivation; The physical loss function of the improved Underwood model when no traffic incident occurs is expressed as: The physical loss function of the improved Underwood model in the event of a traffic incident is: Where, L phy is the corresponding physical model loss; For k The density estimate at N c The total number of road sections divided; N o The number of road sections divided into congested sections in case of traffic incidents; S3.2 establishes a traffic state estimation model based on CV real-time data and ETC detection data, and constructs a VMD-GA-ConvLSTM combined prediction model; First, variational mode decomposition (VMD) is performed on the traffic state sequence. Then, a ConvLSTM network architecture is constructed. Finally, the optimized ConvLSTM network is used to perform parallel prediction of each IMF component. The traffic state estimation value is obtained through modal reconstruction and denormalization. The loss function of the VMD-GA-ConvLSTM neural network module based on data-driven flow is as follows: Under the PIML framework, the total loss function calculation expression for highway traffic state estimation is: L=μL ML +(1-μ)L phy Where, For the road section The actual average speed at For the neural network model The estimated value of velocity at ; μ represents the weighting coefficient of the machine learning loss function.

5. The method for estimating traffic status and distinguishing queue status based on fuzzy traffic wave model according to claim 4, characterized in that: The step S4 includes the following sub-steps: S4.1 introduces the Grey Wolf Optimization Algorithm (GWO) to optimize and improve FCM; Through the membership constraint The objective function of the FCM algorithm is transformed using the Lagrange multiplier method to obtain the unconstrained objective function: Where m is the fuzzy weighted index, m∈(1,+∞); λ j represents the Lagrange multiplier corresponding to the j-th sample; S4.2 maps the original feature space to a high-dimensional space through the kernel function. The corresponding classification function is: Where a i is the Lagrange multiplier coefficient; K is the conversion kernel function; x∈R n ,y∈{1,-1},y∈{1,-1} represents the training sample set, x i is the i-th eigenvector, y i is the class label, when y i =1 is a positive example, y i =-1 is a negative example, l is the number of samples, and n is the dimension of x.

6. The method for estimating traffic status and distinguishing queue status based on fuzzy traffic wave model according to claim 5, characterized in that: The step S5 includes the following sub-steps: S5.1 is based on the fuzzy traffic wave model and uses the traffic flow data provided by the upstream and downstream ETC gantries to calculate the initial wavefront position estimate; S5.2 processes CV data in parallel, extracts single-vehicle feature values, and applies a physical fusion method for single-vehicle and road-section state discrimination information to determine a state judgment value based on the single vehicle and road section; S5.3 Based on the physical constraints in the cyber-physical fusion queue state determination method, continuity correction is performed on the information processing results to obtain wavefront position estimates based on individual vehicles and road sections; S5.4 combines the initial wavefront position obtained by the information processing layer with the CV single-vehicle and road section discrimination results, and integrates the dual constraints of the information layer and the physical layer through a physical constraint correction method based on continuity rules to finally obtain the corrected wavefront position estimate.

Images