Vehicle network channel congestion control method and system based on regional risk level

Abstract

This invention belongs to the technical field of intelligent transportation systems and specifically discloses a method and system for controlling in-vehicle network channel congestion based on regional risk. This system uses a combined fuzzy inference method to assess the driving risk of individual vehicles based on driver, vehicle, and environmental factors that influence driving safety. The system also assesses the driving risk of a road area and uses a convolutional long-short-term memory neural network to make short-term predictions of driving risk for that road area. A non-cooperative game model is constructed, incorporating driving risk and safety information for the road area within the vehicle's location into the model's utility function. The utility function is solved to derive an equilibrium strategy for controlling in-vehicle network channel congestion. This technical solution comprehensively and effectively assesses the impact of the interactions among driver, vehicle, and environmental factors on driving risk, resulting in a more comprehensive and accurate driving risk assessment.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent transportation systems and relates to a vehicle network channel congestion control method and system based on regional risk. Background Art

[0002] With the continuous development of the economy and the advancement of the automotive industry, the concept of Intelligent Traffic System (ITS) has emerged. As one of the key technologies of future ITS, C-V2X (Cellular-Vehicle-to-everything, C-V2X) communication can be used for message exchange between vehicles and their surroundings, covering areas such as safety, infotainment, and autonomous driving. It is undeniable that compared with Dedicated Short Range Communications (DSRC), the resource allocation strategy used by C-V2X has a relatively low probability of congestion. However, due to the increasing contention for available resources, C-V2X will also face the problem of performance degradation in dense network scenarios.

[0003] Compared to traditional wireless networks, the IoV must meet the communication needs of vehicles in high-speed mobile environments. However, IoV communication resources are significantly limited, making their rational utilization crucial. To achieve efficient resource allocation, the frequency of safety message transmissions can be adjusted based on the vehicle's driving risk. For high-risk vehicles, increasing the transmission frequency helps ensure the real-time delivery of critical safety information, thereby improving road safety. Conversely, for lower-risk, standard vehicles, reducing the transmission frequency can conserve channel resources and alleviate network congestion.

[0004] Current driving risk assessment methods fall into two main categories: single-factor-based and multi-factor-based. Single-factor-based driving risk assessments primarily assess a single factor among the driver, vehicle, and environment, which, to a certain extent, helps identify and prevent potential risks. However, with the development of connected vehicle technology and the continued advancement of traffic data collection in the "driver-vehicle-environment" domain, researchers are gradually turning to multi-factor-based driving risk assessment methods. These methods comprehensively consider various factors and can comprehensively and effectively assess the impact of the interactions between driver, vehicle, and environmental factors on driving risk. However, existing technologies do not dynamically formulate congestion control strategies based on vehicle driving status, potentially posing traffic safety hazards in congested environments. They also fail to consider the nonlinear, time-varying, and uncertain nature of the various factors influencing driving risk. The interdependence and difficulty in quantifying these factors make direct analysis difficult. Summary of the Invention

[0005] The purpose of the present invention is to provide a vehicle network channel congestion control method and system based on regional risk, which can comprehensively and effectively evaluate the impact of the interaction between driver factors, vehicle factors and environmental factors on driving risks, and obtain more comprehensive and accurate driving risk assessment results.

[0006] To achieve the above objectives, the basic solution of the present invention is: a method for controlling vehicle network channel congestion based on regional risk, comprising the following steps:

[0007] Based on the driver factors, vehicle factors and environmental factors that affect driving safety, the driving risk of a single vehicle is assessed using a combined fuzzy reasoning method.

[0008] Assess the driving risk of road areas and use convolutional long short-term memory neural networks to make short-term predictions of driving risk in road areas;

[0009] Construct a non-cooperative game model and incorporate the driving risk and safety information of the road area where the vehicle is located into the utility function of the non-cooperative game model;

[0010] Solving the utility function, we obtain the equilibrium strategy for transmission frequency control in vehicle network channel congestion control.

[0011] The working principle and beneficial effects of this basic solution are as follows: This technical solution first evaluates the driving risk of the vehicle, and based on the driving risk degree obtained from the evaluation, conducts a driving risk assessment of the road area, and performs a short-term road area driving risk prediction. Then, a non-cooperative game model is established and solved based on the driving risk degree to obtain the vehicle's equilibrium strategy. The vehicle adaptively adjusts the vehicle's safety message transmission frequency based on the obtained equilibrium strategy. The driving risk assessment method based on combined fuzzy reasoning can consider as many factors that affect driving safety as possible, thereby obtaining a more comprehensive and accurate driving risk assessment result. This patent takes driving risk and the transmission frequency of safety information into account in the utility function, and sets the optimal transmission frequency for vehicles with different communication needs to alleviate channel congestion and save communication resources.

[0012] Furthermore, the method for evaluating the driving risk of a single vehicle is as follows:

[0013] The traditional fuzzy reasoning model is selected for fuzzy reasoning. The driving risk assessment model influenced by driver factors belongs to a dual-input multi-rule reasoning model. The reasoning model processing process is as follows:

[0014]

[0015] Among them, X n is the input 1 under the nth rule, Y nis the input 2 under the nth rule, Z n is the inference result obtained by the nth rule, X * is the “∪” aggregation of input parameter 1 in all rules, Y * is the "∪" aggregation of input parameter 2 in all rules;

[0016] First process each rule:

[0017]

[0018] Then the results of each inference are Perform "∪" aggregation to obtain the final fuzzy result Z * ,Right now:

[0019]

[0020] Use the centroid method to defuzzify and get Z * Assuming that the exact value of the driving risk affected by the driver factor is A, then A is:

[0021]

[0022] Among them, N represents the number of points in the domain, A i is the i-th point, μ Z* (A i ) represents the fuzzy solution Z * The membership function of

[0023] Obtain the precise value D1 of the driving risk affected by the driver factor. Similarly, calculate the driving risk D2 and D3 of the vehicle factor and environmental factor respectively.

[0024] Perform fuzzy reasoning based on D1, D2, and D3 to output the corresponding total driving risk D0 using the triangle membership function:

[0025]

[0026] By adjusting the parameters a, b, and c, the sensitivity of the triangle membership function to input changes is controlled; x is the independent variable; a, b, and c represent the vertex positions of the triangle respectively;

[0027] The traditional fuzzy reasoning method is used for reasoning, and after defuzzification, the driving risk of the vehicle itself can be obtained.

[0028] Analyze the factors that affect driving safety, and based on these factors, use the combined fuzzy reasoning method to evaluate driving risks, which is easy to operate.

[0029] Furthermore, the method for evaluating the driving risk of the road area is as follows:

[0030] Divide the road into grids and use the grid as a unit to perform driving risk assessment on the road area. For the current road grid, it is necessary to determine the surrounding grids.

[0031] The communication range of C-V2X is 150m. However, in actual road scenarios, when the grid approaches the boundary, the grid range needs to be adjusted accordingly according to the actual situation. Therefore, for each road grid (i, j), its surrounding grids are defined as:

[0032] In the length direction, it extends from min(i-30,0) to max(i+30,199); in the width direction, it extends from min(j-50,0) to max(j+50,3);

[0033] For each road grid, its risk value is calculated by weighted average, using the Euclidean distance weighting method to represent the impact of vehicles at different distances on the current road grid;

[0034] Assume that the grid where the current vehicle is located is i, and the surrounding adjacent grids are j1, j2, ..., j k , the corresponding driving risk values ​​are R1, R2, …, Rk; then the driving risk assessment result of the current grid i is:

[0035] R i =(ω1R1+ω2R2+…+ω n R n ) / (ω1+ω2+…+ω n )

[0036] Among them, ω i Represents the current grid i to the adjacent grid j1, j2, ..., j k The reciprocal of the Euclidean distance, that is:

[0037]

[0038] ...

[0040]

[0041] Where d(i,j) represents the Euclidean distance between grid i and grid j, that is:

[0042]

[0043] Among them, (x i ,y i ) represents the coordinates of grid i, (x j ,y j) represents the coordinates of grid j;

[0044] Get the driving risk of the road grid where the vehicle is located, and use the matrix X t To express the driving risk distribution of the road area at time t:

[0045]

[0046] in, Represents the driving risk of the road grid in row i and column j at time t.

[0047] Carry out driving risk assessment to ensure accurate reflection of the driving risk conditions in the road area, thereby providing protection for driving safety.

[0048] Furthermore, the specific method for short-term prediction of driving risk in road areas is as follows:

[0049] The driving risk of the past m moments is used to predict the driving risk of the next n moments, which can be expressed as:

[0050] X t,t+n =Γ(X t-m+1,t )

[0051] Among them, X t,t+n represents the driving risk from time t to time t+n, Γ represents the prediction method, t represents the time, X t-m+1,t It represents the driving risk from time t-m+1 to time t.

[0052] Road traffic environments are characterized by diversity and uncertainty. The driving risk at one moment cannot fully reflect the driving risk at the next moment, resulting in inaccurate results when adaptively adjusting parameters for congestion control. To overcome these problems, short-term predictions of driving risk in road areas are needed to compensate for the shortcomings of real-time assessments.

[0053] Furthermore, the method of constructing a non-cooperative game model is:

[0054] The congestion control model based on driving risk game is:

[0055] G={V;r1,r2,…,r n ;u1,u2,…,u n}

[0056] Among them, V = {1,2,…,n} represents the set of players participating in the game, i.e., vehicles, and R = {r1,r2,…r n} represents the strategy set of the vehicle transmission frequency participating in the game, r i Represents the strategy of vehicle i. The transmission frequency range of vehicle i participating in the game is 0 <rmin ≤r i ≤r max The transmission frequency range in C-V2X is [1,50], i.e. 1≤ri≤50.

[0057] u i Represents the utility function of vehicle i participating in the game:

[0058] u i =(r i ,r -i ),

[0059] Among them, r -i =(r1,…,r i-1 ,r i+1 ,…,r n ) represents the transmission frequency strategy of other participating vehicles except vehicle i, where i is a natural number;

[0060] Considering that each participant needs to choose the best strategy to maximize his own interests according to his own interests and goals, the optimal transmission frequency of vehicle i is determined by maximizing u i =(r i ,r -i ) is obtained, the problem is stated as:

[0061] r i * =arg max u i (r i ,r -i )

[0062] r i ∈[r min ,r max ]

[0063] The final utility function is:

[0064]

[0065] Among them, r i represents the transmission frequency of vehicle i, d i represents the driving risk of vehicle i, α i represents the utility factor, β i is a constant coefficient, and α i >>β i ,λ i is the cost factor, It is related to the strategies of other vehicles, and N is the total number of vehicles.

[0066] Adaptively adjust parameters according to the driving risk of each vehicle in the road grid to ensure driving safety, and establish a non-cooperative game model suitable for the Internet of Vehicles to describe the competitive relationship and interaction between vehicles for easy use.

[0067] Furthermore, the proof of the existence of Nash equilibrium in the non-cooperative game model is as follows:

[0068] For any vehicle i, its strategy is discrete and is in the interval 0 <r min ≤r i ≤r max Inside, so R is nonempty, closed and bounded;

[0069] For each car's utility function u i (r i ,r -i ) is continuous on itself R, for u i To find the first-order partial derivative:

[0070]

[0071] Continue to find the second-order partial derivative and get:

[0072]

[0073] For any r in the interval i All have 0 <r min ≤r i ≤r max ,and Therefore, the utility function of the vehicle is a continuous concave function in the strategy space, so the non-cooperative game model has a Nash equilibrium solution;

[0074] Use proof by contradiction to prove the uniqueness of Nash equilibrium: Suppose there are two different Nash equilibria (r i1 ,r -i1 ) and (r i2 ,r -i2 ), and r i1 Not equal to r i2 , since each participant will choose its own optimal transmission frequency, we get:

[0075] u i (r i1 ,r -i1 )>u i (r i2 ,r -i1 )

[0076] u i (r i2 ,r -i2 )>u i(r i1 ,r -i2 )

[0077] Add the two equations together: i (r i1 ,r -i1 )+u i (r i2 ,r -i2 )>u i (r i2 ,r -i1 )+u i (r i1 ,r -i2 ), according to the given utility function u i (r i1 ,r -i1 ), we can get:

[0078]

[0079] Since all participants in Nash equilibrium will choose the transmission frequency that maximizes their own utility, r i1 and r i2 The conditions that the derivative of the corresponding utility function is equal to 0 must be satisfied, that is:

[0080]

[0081]

[0082] Dividing the two equations yields: 1+β i ·r i1 =1+β i ·r i2 , and get r i1 =r i2 , and r i1 ≠r i2 Contradiction, so the Nash equilibrium solution of the game model is unique.

[0083] Simple operation and easy to use.

[0084] Furthermore, the method of solving the utility function using the hybrid particle swarm optimization algorithm is:

[0085] Each vehicle is considered as a particle, the transmission frequency of the safety message is regarded as the speed of the particle, a solution to the problem is represented by the position of the particle, and the fitness function is the utility function u of vehicle i i definition:

[0086] f(r i )=u i (r i ,r -i )

[0087] Among them, the fitness function f(r i ) is the strategy r of vehicle i i is a function of the independent variable; r i represents the strategy of vehicle i, r -i =(r1,…,r i-1 ,r i+1 ,…,r n ) represents the transmission frequency strategy of other participating vehicles except vehicle i;

[0088] Calculate and compare the fitness value f of the particle position, perform iterative updates, and find the optimal solution to the problem;

[0089] By continuously adjusting the position and speed of the particles, and recording the currently searched individual optimal value and group optimal value, the particle swarm is continuously optimized within the search range, and finally the particle swarm converges to the optimal position gbest with the maximum fitness value f.

[0090] Based on the driving risk and safety message transmission frequency of the road grid area where the vehicle is located, a hybrid particle swarm optimization algorithm is used to solve the equilibrium strategy, and the vehicle adaptively adjusts the vehicle's safety message transmission frequency based on the obtained equilibrium strategy.

[0091] The present invention also provides an in-vehicle network channel congestion control system based on regional risk, including a data acquisition unit and a processing unit. The data acquisition unit is used to collect data on driver factors, vehicle factors and environmental factors that affect driving safety. The output end of the data acquisition unit is connected to the input end of the processing unit. The processing unit executes the method described in the present invention to perform in-vehicle network channel congestion control.

[0092] The system is used to obtain the vehicle's equilibrium strategy, which is convenient for use. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is a flow chart of the vehicle network channel congestion control method based on regional risk level of the present invention;

[0094] Figure 2 This is a schematic diagram of combined fuzzy reasoning of the vehicle network channel congestion control method based on regional risk level of the present invention;

[0095] Figure 3 This is a network structure diagram of driving risk prediction of the vehicle network channel congestion control method based on regional risk in the present invention. DETAILED DESCRIPTION

[0096] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0097] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0098] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0099] The present invention discloses a vehicle network channel congestion control method based on regional risk, which mainly studies the C-V2X congestion control strategy. First, the driving risk of the vehicle is evaluated. According to the driving risk degree obtained by the evaluation, the driving risk of the road area is evaluated, and the short-term driving risk of the road area is predicted. Then, a non-cooperative game model is established and solved according to the driving risk degree to obtain the vehicle's equilibrium strategy. Based on the obtained equilibrium strategy, the vehicle adaptively adjusts the vehicle's safety message transmission frequency.

[0100] like Figure 1 As shown, the vehicle network channel congestion control method includes the following steps:

[0101] like Figure 2 As shown in the figure, based on the driver factors, vehicle factors and environmental factors that affect driving safety, the driving risk of a single vehicle is assessed using a combined fuzzy reasoning method (first performing combined reasoning, then dividing the road area into grids of equal size, and using a weighted method to assess the driving risk of the road area); due to the mutual influence between vehicles and the complexity and variability of traffic flow, it is necessary to analyze the various factors that affect driving safety. Based on the above factors, a fuzzy reasoning model is established, and combined fuzzy reasoning is performed based on these models.

[0102] Assess the driving risk of road areas and use Convolutional Long Short-Term Memory (ConvLSTM) neural networks to make short-term predictions of driving risk in road areas.

[0103] Construct a non-cooperative game model and incorporate the driving risk and safety information of the road area where the vehicle is located into the utility function of the non-cooperative game model;

[0104] Solving the utility function, we obtain the equilibrium strategy for transmission frequency control in vehicle network channel congestion control, and the output gbest is the global optimal position.

[0105] In a preferred embodiment of the present invention, taking into account the complexity of influencing factors, this embodiment adopts a two-stage fuzzy inference model to evaluate the driving risk of a single vehicle as follows:

[0106] The first stage includes three fuzzy reasoning modules, which are used to evaluate the driving risk corresponding to driver factors, vehicle factors and environmental factors respectively.

[0107] Taking the driver factor as an example, the driver's continuous driving time and driving behavior are selected as input variables. Fuzzy reasoning is performed using IF-THEN fuzzy rules based on the knowledge base, and the output linguistic variable of the driver factor driving risk is obtained by defuzzification.

[0108] The fuzzy set for the input variable "driver's driving time DT" is divided into three fuzzy values: normal (S), long (M), and very long (B). The fuzzy set for the input variable "driver's driving time DT" is {S, M, B}. The driver's driving time can be obtained by onboard equipment. The domain is represented as [0, 4]. The longer the driving time, the larger the value mapped to the domain.

[0109] The "Driver Driving Behavior Database" serves as another input linguistic variable. The fuzzy set is divided into three fuzzy linguistic values: normal (S), poor (M), and dangerous (B). The fuzzy set is {S, M, B}, representing normal driving behavior, poor driving behavior (such as following too close, frequently changing lanes, frequent sudden braking and acceleration), and dangerous driving behavior (such as not following traffic rules and distracted driving). Driver behavior can be obtained through in-vehicle devices, smartphone apps, and other means. The domain is defined as [0, 3]. The more dangerous the driving behavior, the larger the value mapped to the domain.

[0110] The traditional fuzzy reasoning model is selected for fuzzy reasoning. The driving risk assessment model influenced by driver factors belongs to a dual-input multi-rule reasoning model. The reasoning model processing process is as follows:

[0111]

[0112] Among them, Xn is the input 1 under the nth rule, Y n is the input 2 under the nth rule, Z n is the inference result obtained by the nth rule, X * is the “∪” aggregation of input parameter 1 in all rules, Y * is the "∪" aggregation of input parameter 2 in all rules;

[0113] First process each rule:

[0114]

[0115] Then the results of each inference are Perform "∪" aggregation to obtain the final fuzzy result Z * ,Right now:

[0116]

[0117] Use the centroid method to defuzzify and get Z * Assuming that the exact value of the driving risk affected by the driver factor is A, then A is:

[0118]

[0119] Among them, N represents the number of points in the domain, A i is the i-th point, μ Z* (A i ) represents the fuzzy solution Z * The membership function of

[0120] Obtain the precise value D1 of driving risk affected by driver factors (A is a general term, and different values ​​can be used to obtain the precise values ​​D1, D2, and D3 of driving risk affected by different factors). Similarly, calculate the driving risk D2 and D3 of vehicle factors and environmental factors respectively.

[0121] The second stage is to perform fuzzy reasoning based on D1, D2, and D3, and output the corresponding total driving risk D0 using the triangle membership function:

[0122]

[0123] By adjusting the parameters a, b, and c, the sensitivity of the triangle membership function to input changes can be controlled; x is the independent variable; a, b, and c represent the positions of the vertices of the triangle respectively; these parameters determine the shape and width of the triangle; if the distance between a, b, and c is small, that is, the triangle is narrower, then the function is more sensitive to input changes; conversely, if the triangle is wider, it is less sensitive to input changes, and a larger input value change is required to cause a significant change in the membership function.

[0124] The number of inference rules in a fuzzy inference system is calculated by taking the Cartesian product of the membership set and the fuzzy linguistic variable. In this fuzzy inference system, each input variable has three membership sets. Therefore, the number of fuzzy rules in this fuzzy inference system is the product of the number of membership sets of the two input linguistic variables, which is 9. The formulated rules are shown in Table 1. Using the traditional fuzzy inference method (Mamdani), the fuzzy output is defuzzified using the maximum method, centroid method, center method, or weighted average method to obtain the vehicle's driving risk D0.

[0125] Table 1 Reasoning rules for driver factor driving risk assessment

[0126]

[0127] In a preferred embodiment of the present invention, the method for evaluating the driving risk of a road area is as follows:

[0128] The driving status and behavior of the current vehicle will affect the driving status and behavior of other vehicles, and the driving status and behavior of other vehicles will also affect the current vehicle. Therefore, after obtaining the driving risk assessment results of a single vehicle, it is necessary to conduct a comprehensive assessment of the driving risk of the road area. Divide the road into grids and perform the road area driving risk assessment as a unit. For the current road grid, it is necessary to determine the surrounding grids.

[0129] The communication range of C-V2X is 150m. However, in actual road scenarios, when the grid approaches the boundary, the grid range needs to be adjusted accordingly according to the actual situation. Therefore, for each road grid (i, j), its surrounding grids are defined as:

[0130] In the length direction, it extends from min(i-30,0) to max(i+30,199); in the width direction, it extends from min(j-50,0) to max(j+50,3);

[0131] For each road grid, its risk value is calculated by weighted average, using the Euclidean distance weighting method to represent the impact of vehicles at different distances on the current road grid;

[0132] Assume that the grid where the current vehicle is located is i, and the surrounding adjacent grids are j1, j2, ..., j k , the corresponding driving risk values ​​are R1, R2, …, Rk; then the driving risk assessment result of the current grid i is:

[0133] R i =(ω1R1+ω2R2+…+ω n R n ) / (ω1+ω2+…+ωn )

[0134] Among them, ω i Represents the current grid i to the adjacent grid j1, j2, ..., j k The reciprocal of the Euclidean distance, that is:

[0135]

[0136] ...

[0138]

[0139] Where d(i,j) represents the Euclidean distance between grid i and grid j, that is:

[0140]

[0141] Among them, (x i ,y i ) represents the coordinates of grid i, (x j ,y j ) represents the coordinates of grid j;

[0142] Get the driving risk of the road grid where the vehicle is located, and use the matrix X t To express the driving risk distribution of the road area at time t:

[0143]

[0144] in, Represents the driving risk of the road grid in row i and column j at time t.

[0145] In a preferred embodiment of the present invention, a specific method for performing short-term prediction of driving risk in a road area is as follows:

[0146] Road traffic environments are characterized by diversity and uncertainty. The driving risk at one moment cannot fully reflect the driving risk at the next moment, resulting in inaccurate results when adaptively adjusting parameters for congestion control. To overcome these problems, short-term predictions of driving risk in road areas are needed to compensate for the shortcomings of real-time assessments.

[0147] Traffic flow changes are not only related to time, but also to space. Therefore, the distribution changes of driving risk also have temporal and spatial correlations.

[0148] The driving risk of the past m moments is used to predict the driving risk of the next n moments, which can be expressed as:

[0149] X t,t+n =Γ(Xt-m+1,t )

[0150] Among them, X t,t+n represents the driving risk from time t to time t+n, Γ represents the prediction method, t represents the time, X t-m+1,t represents the driving risk from time t-m+1 to time t;

[0151] To more accurately capture the distribution of driving risk across time and space, this solution uses a convolutional long short-term memory (ConvLSTM) neural network for short-term prediction. ConvLSTM combines the advantages of convolutional neural networks and long short-term memory networks, processing both time series and spatial data simultaneously, which is consistent with the characteristics of road traffic.

[0152] Unlike traditional LSTMs, ConvLSTMs use convolutions for computation rather than the matrix multiplication used in LSTMs. Furthermore, the input to the ConvLSTM model is a 3D tensor, rather than the 2D input of LSTMs. Existing convolutional long-short-term memory (CSTM) neural networks can be used for short-term predictions. For example, the convLstm formula for a CSTM is:

[0153]

[0154] Among them, i t 、f t 、o t 、X t 、H t 、C t Both represent 3D tensors, “*” represents convolution operation; σ is a sigma function, and ⊙ is the same or equal symbol.

[0155] The road area driving risk prediction model of the present invention superimposes multiple ConvLSTM layers. The model consists of two parts: the encoding network and the prediction network. Figure 3 As shown in Figure 2, the encoding network and prediction network are both composed of multiple ConvLSTM layers and are used to process spatiotemporal sequence data. The encoding network compresses the input sequence to generate a hidden state tensor, which carries information about driving risk. The prediction network expands the hidden state tensor into an output sequence and predicts the value of each position in the sequence step by step. To ensure continuity and consistency of the neural network, the initial state and unit output of the prediction network are copied from the final state of the encoding network.

[0156] Ultimately, all states in the prediction network are connected and passed through convolutional layers to generate the final prediction result. This approach enables the model to effectively handle spatiotemporal correlations and achieve accurate road risk prediction.

[0157] In a preferred embodiment of the present invention, the method for constructing a non-cooperative game model is:

[0158] According to the components of game theory, the congestion control model based on driving risk game is:

[0159] G={V;r1,r2,…,r n ;u1,u2,…,u n}

[0160] Among them, V = {1,2,…,n} represents the set of players participating in the game, i.e., vehicles, and R = {r1,r2,…r n} represents the strategy set of the vehicle transmission frequency participating in the game, r i Represents the strategy of vehicle i. The transmission frequency range of vehicle i participating in the game is 0 <r min ≤r i ≤r max The transmission frequency range in C-V2X is [1,50], i.e. 1≤ri≤50.

[0161] u i Represents the utility function of vehicle i participating in the game:

[0162] u i =(r i ,r -i )

[0163] Among them, r -i =(r1,…,r i-1 ,r i+1 ,…,r n ) represents the transmission frequency strategy of other participating vehicles except vehicle i, where i is a natural number;

[0164] Considering that each participant needs to choose the best strategy to maximize his own interests according to his own interests and goals, the optimal transmission frequency of vehicle i is determined by maximizing u i =(r i ,r -i ) is obtained, the problem is stated as:

[0165] r i * =arg max u i (r i ,r -i )

[0166] r i ∈[r min ,r max ]

[0167] The final utility function is:

[0168]

[0169] Among them, r i represents the transmission frequency of vehicle i (i.e., the strategy of vehicle i), d i represents the driving risk of vehicle i, α i represents the utility factor, β i is a constant coefficient, and α i >>β i ,λ i is the cost factor, It is related to the strategies of other vehicles, and N is the total number of vehicles.

[0170] The first term of the utility function is the benefit function, which represents the benefit obtained by vehicle i in the game and is related to the transmission frequency r i Related, with r i As the value of utility increases, the utility value will also increase. i represents the sensitivity of the i-th vehicle to the benefit of the transmission frequency, and the coefficient β i The second term of the utility function represents the cost function of vehicle i, which is related to the driving risk d i The cost function is proportional to the transmission frequency of the vehicle and the driving risk d i This indicates that the greater the driving risk of a vehicle, the easier it is to obtain a higher transmission frequency.

[0171] This utility function balances the risks and benefits of driving. For high-risk vehicles, lower penalties lead to higher transmission frequencies, ensuring timely and reliable delivery of safety information. For ordinary vehicles, higher penalties lead to lower transmission frequencies, helping to conserve communication resources and alleviate channel congestion.

[0172] A preferred embodiment of the present invention further includes a method for proving the existence of a Nash equilibrium of a non-cooperative game model as follows:

[0173] For any vehicle i, its strategy is discrete and is in the interval 0 <r min ≤r i ≤r max Inside, so R is nonempty, closed and bounded;

[0174] For each car's utility function u i (r i ,r -i ) is continuous on itself R, for u i To find the first-order partial derivative:

[0175]

[0176] Continue to find the second-order partial derivative and get:

[0177]

[0178] For any r in the interval i All have 0 <r min ≤r i ≤r max ,and Therefore, the utility function of the vehicle is a continuous concave function in the strategy space, so the non-cooperative game model has a Nash equilibrium solution;

[0179] Use proof by contradiction to prove the uniqueness of Nash equilibrium: Suppose there are two different Nash equilibria (r i1 ,r -i1 ) and (r i2 ,r -i2 ), and r i1 Not equal to r i2 , since each participant will choose its own optimal transmission frequency, we get:

[0180] u i (r i1 ,r -i1 )>u i (r i2 ,r -i1 )

[0181] u i (r i2 ,r -i2 )>u i (r i1 ,r -i2 )

[0182] Add the two equations together: i (r i1 ,r -i1 )+u i (r i2 ,r -i2 )>u i (r i2 ,r -i1 )+u i (r i1 ,r -i2 ), according to the given utility function u i (r i1 ,r -i1 ), we can get:

[0183]

[0184] Since all participants in Nash equilibrium will choose the transmission frequency that maximizes their own utility, r i1 and r i2 The conditions that the derivative of the corresponding utility function is equal to 0 must be satisfied, that is:

[0185]

[0186]

[0187] Dividing the two equations yields: 1+β i ·r i1 =1+β i ·r i2 , and get r i1 =r i2 , and r i1 ≠r i2 Contradiction, so the Nash equilibrium solution of the game model is unique.

[0188] In a preferred embodiment of the present invention, the method for solving the utility function using the hybrid particle swarm optimization algorithm (HPSO) is:

[0189] Each vehicle is considered as a particle, the transmission frequency of the safety message is regarded as the speed of the particle, a solution to the problem is represented by the position of the particle, and the fitness function is the utility function u of vehicle i i definition:

[0190] f(r i )=u i (r i ,r -i )

[0191] Among them, the fitness function f(r i ) is the strategy r of vehicle i i is a function of the independent variable; r i represents the strategy of vehicle i, r -i =(r1,…,r i-1 ,r i+1 ,…,r n ) represents the transmission frequency strategy of other participating vehicles except vehicle i;

[0192] Calculate and compare the fitness value f of the particle position, perform iterative updates, and find the optimal solution to the problem;

[0193] By continuously adjusting the position and speed of the particles, and recording the currently searched individual optimal value and group optimal value, the particle swarm is continuously optimized within the search range, and finally the particle swarm converges to the optimal position gbest with the maximum fitness value f.

[0194] This technical solution first establishes a non-cooperative game model suitable for the Internet of Vehicles (IoV) to describe the competitive relationships and interactions between vehicles. Based on the driving risk and safety message transmission frequency of the road grid area where the vehicle is located, a vehicle utility function is designed, and the uniqueness and existence of the Nash equilibrium of this model are proven. Finally, a hybrid particle swarm optimization (HPSO) algorithm is used to solve the problem, resulting in an equilibrium strategy that can be used to adaptively adjust the vehicle transmission frequency.

[0195] The present invention also provides a regional risk-based in-vehicle network channel congestion control system, comprising a data acquisition unit and a processing unit. The data acquisition unit (e.g., a camera, temperature sensor, speed sensor, etc.) is used to collect data on driver, vehicle, and environmental factors that affect driving safety. The output of the data acquisition unit is electrically connected to the input of the processing unit, which executes the method described in the present invention to control in-vehicle network channel congestion. This system can be used to obtain a vehicle balancing strategy, facilitating its use.

[0196] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0197] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A vehicle network channel congestion control method based on regional risk, characterized in that: The steps include: Based on the driver factors, vehicle factors and environmental factors that affect driving safety, the driving risk of a single vehicle is assessed using a combined fuzzy reasoning method. Assess the driving risk of road areas and use convolutional long short-term memory neural networks to make short-term predictions of driving risk in road areas; Construct a non-cooperative game model and incorporate the driving risk and safety information of the road area where the vehicle is located into the utility function of the non-cooperative game model; Solve the utility function and obtain the equilibrium strategy of transmission frequency control for vehicle network channel congestion control; The method of constructing a non-cooperative game model is: The congestion control model based on driving risk game is: G={V;r1,r2,…,r n ;u1,u2,…,u n } Among them, V = {1,2,…,n} represents the set of players participating in the game, i.e., vehicles, and R = {r1,r2,…r n } represents the strategy set of the vehicle transmission frequency participating in the game, r i Represents the strategy of vehicle i. The transmission frequency range of vehicle i participating in the game is 0 <r min ≤r i ≤r max ; The transmission frequency range in C-V2X is [1,50], that is, 1≤r i ≤50; u i Represents the utility function of vehicle i participating in the game: u i =(r i ,r -i ) Among them, r -i =(r1,…,r i-1 ,r i+1 ,…,r n ) represents the transmission frequency strategy of other participating vehicles except vehicle i, where i is a natural number; Considering that each participant needs to choose the best strategy to maximize his own interests according to his own interests and goals, the optimal transmission frequency of vehicle i is determined by maximizing u i =(r i ,r -i ) is obtained, the problem is stated as: r i * =arg max u i (r i ,r -i ) r i ∈[r min ,r max ] The final utility function is: Among them, r i represents the transmission frequency of vehicle i, d i represents the driving risk of vehicle i, α i represents the utility factor, β i is a constant coefficient, and α i >>β i ,λ i is the cost factor, It is related to the strategies of other vehicles, and N is the total number of vehicles.

2. The method for controlling vehicle network channel congestion based on regional risk as claimed in claim 1, characterized in that: The method for assessing the driving risk of a single vehicle is as follows: The traditional fuzzy reasoning model is selected for fuzzy reasoning. The driving risk assessment model influenced by driver factors belongs to a dual-input multi-rule reasoning model. The reasoning model processing process is as follows: Among them, X n is the input 1 under the nth rule, Y n is the input 2 under the nth rule, Z n is the inference result obtained by the nth rule, X * is the “∪” aggregation of input parameter 1 in all rules, Y * is the "∪" aggregation of input parameter 2 in all rules; First process each rule: Then the results of each inference are Perform "∪" aggregation to obtain the final fuzzy result Z * ,Right now: Use the centroid method to defuzzify and get Z * Assuming that the exact value of the driving risk affected by the driver factor is A, then A is: Among them, N represents the number of points in the domain, A i is the i-th point, Represents the fuzzy solution Z * The membership function of Obtain the precise value D1 of the driving risk affected by the driver factor. Similarly, calculate the driving risk D2 and D3 of the vehicle factor and environmental factor respectively. Perform fuzzy reasoning based on D1, D2, and D3 to output the corresponding total driving risk D0 using the triangle membership function: By adjusting the parameters a, b, and c, the sensitivity of the triangle membership function to input changes is controlled; x is the independent variable; a, b, and c represent the vertex positions of the triangle respectively; The traditional fuzzy reasoning method is used for reasoning, and after defuzzification, the driving risk of the vehicle itself can be obtained.

3. The method for controlling vehicle network channel congestion based on regional risk as claimed in claim 1, characterized in that: The method for assessing driving risk in road areas is: Divide the road into grids and use the grid as a unit to perform driving risk assessment on the road area. For the current road grid, it is necessary to determine the surrounding grids. The communication range of C-V2X is 150m. However, in actual road scenarios, when the grid approaches the boundary, the grid range needs to be adjusted accordingly according to the actual situation. Therefore, for each road grid (i, j), its surrounding grids are defined as: In the length direction, it extends from min(i-30,0) to max(i+30,199); in the width direction, it extends from min(j-50,0) to max(j+50,3); For each road grid, its risk value is calculated by weighted average, using the Euclidean distance weighting method to represent the impact of vehicles at different distances on the current road grid; Assume that the grid where the current vehicle is located is i, and the surrounding adjacent grids are j1, j2, ..., j k , the corresponding driving risk values ​​are R1, R2,…, R k ; The driving risk assessment result of the current grid i is: R i =(ω1R1+ω2R2+…+ω n R n ) / (ω1+ω2+…+ω n ) Among them, ω i Represents the current grid i to the adjacent grid j1, j2, ..., j k The reciprocal of the Euclidean distance, that is: Where d(i,j) represents the Euclidean distance between grid i and grid j, that is: Among them, (x i ,y i ) represents the coordinates of grid i, (x j ,y j ) represents the coordinates of grid j; Get the driving risk of the road grid where the vehicle is located, and use the matrix X t To express the driving risk distribution of the road area at time t: in, Represents the driving risk of the road grid in row i and column j at time t.

4. The method for controlling vehicle network channel congestion based on regional risk as claimed in claim 1, characterized in that: The specific method for short-term prediction of driving risks in road areas is as follows: The driving risk of the past m moments is used to predict the driving risk of the next n moments, which can be expressed as: X t,t+n =Γ(X t-m+1,t ) Among them, X t,t+n represents the driving risk from time t to time t+n, Γ represents the prediction method, t represents the time, X t-m+1,t It represents the driving risk from time t-m+1 to time t.

5. The method for controlling vehicle network channel congestion based on regional risk as claimed in claim 1, characterized in that: The proof of the existence of Nash equilibrium of the non-cooperative game model is also included as follows: For any vehicle i, its strategy is discrete and is in the interval 0 <r min ≤r i ≤r max Inside, so R is nonempty, closed and bounded; For each car's utility function u i (r i ,r -i ) is continuous on itself R, for u i To find the first-order partial derivative: Continue to find the second-order partial derivative and get: For any r in the interval i All have 0 <r min ≤r i ≤r max ,and Therefore, the utility function of the vehicle is a continuous concave function in the strategy space, so the non-cooperative game model has a Nash equilibrium solution; Use proof by contradiction to prove the uniqueness of Nash equilibrium: Suppose there are two different Nash equilibria (r i1 ,r -i1 ) and (r i2 ,r -i2 ), and r i1 Not equal to r i2 , since each participant will choose its own optimal transmission frequency, we get: u i (r i1 ,r -i1 )>u i (r i2 ,r -i1 ) u i (r i2 ,r -i2 )>u i (r i1 ,r -i2 ) Add the two equations together: i (r i1 ,r -i1 )+u i (r i2 ,r -i2 )>u i (r i2 ,r -i1 )+u i (r i1 ,r -i2 ), according to the given utility function u i (r i1 ,r -i1 ), we can get: Since all participants in Nash equilibrium will choose the transmission frequency that maximizes their own utility, r i1 and r i2 The conditions that the derivative of the corresponding utility function is equal to 0 must be satisfied, that is: Dividing the two equations yields: 1+β i ·r i1 =1+β i ·r i2 , and get r i1 =r i2 , and r i1 ≠r i2 Contradiction, so the Nash equilibrium solution of the game model is unique.

6. The method for controlling vehicle network channel congestion based on regional risk as claimed in claim 1, characterized in that: The method of solving the utility function using hybrid particle swarm optimization algorithm is: Each vehicle is considered as a particle, the transmission frequency of the safety message is regarded as the speed of the particle, a solution to the problem is represented by the position of the particle, and the fitness function is the utility function u of vehicle i i definition: f(r i )=u i (r i ,r -i ) Among them, the fitness function f(r i ) is the strategy r of vehicle i i is a function of the independent variable; r -i =(r1,…,r i-1 ,r i+1 ,…,r n ) represents the transmission frequency strategy of other participating vehicles except vehicle i; Calculate and compare the fitness value f of the particle position, perform iterative updates, and find the optimal solution to the problem; By continuously adjusting the position and speed of the particles, and recording the currently searched individual optimal value and group optimal value, the particle swarm is continuously optimized within the search range, and finally the particle swarm converges to the optimal position gbest with the maximum fitness value f.

7. A vehicle network channel congestion control system based on regional risk, characterized in that: It includes a data acquisition unit and a processing unit. The data acquisition unit is used to collect data on driver factors, vehicle factors and environmental factors that affect driving safety. The output end of the data acquisition unit is connected to the input end of the processing unit. The processing unit executes the method described in one of claims 1 to 6 to perform in-vehicle network channel congestion control.

Images