C-V2X Internet of Vehicles wireless communication resource allocation method

Abstract

The invention belongs to the field of wireless communication resource allocation, and provides a C-V2X Internet of Vehicles wireless communication resource allocation method, which comprises the steps of input feature integration, CAM generation interval prediction model prediction and RRI field updating. The pre-training process of the CAM generation interval prediction model comprises the steps of hyper-parameter optimization module construction, prediction module construction and core hyper-parameter optimization. By updating the CAM generation interval category prediction result into the RRI field in the sidelink control information, the problem that the existing SB-SPS algorithm cannot effectively deal with the non-periodic CAM is solved, and the rationality and flexibility of wireless communication resource allocation are improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication resource allocation, and in particular to a method for allocating wireless communication resources for C-V2X vehicle networking. Background Art

[0002] ITS is the abbreviation for Intelligent Transportation System; V2X is the abbreviation for Internet of Vehicles; C-V2X is the abbreviation for cellular Internet of Vehicles based on cellular mobile communication system; DENM is the abbreviation for Decentralized Environmental Notification Message; CAM is the abbreviation for Collaborative Awareness Message; 3GPP is the abbreviation for 3rd Generation Partnership Project; SB-SPS is the abbreviation for Perception-Based Semi-persistent Resource Allocation; RRI is the abbreviation for Resource Reservation Interval. Global technological innovation with intelligence and networking as important features is bringing major changes to the ITS field. As an important part of the intelligent transportation system, V2X has successfully transformed traffic participants such as cars, roads, and pedestrians from information islands to information nodes in the intelligent transportation system, thereby enabling more efficient, safer, and more environmentally friendly transportation. In recent years, in the field of Internet of Vehicles technology, C-V2X technology, a cellular Internet of Vehicles based on a cellular mobile communication system, has gradually developed into a mainstream technology. In particular, in the C-V2X communication mode, direct communication realizes direct data exchange between terminal devices through the PC5 interface, which is crucial for real-time application scenarios related to road safety. However, the direct communication mode faces challenges such as limited spectrum resources, a surge in the number of vehicles accessing the C-V2X system, and frequent changes in network topology caused by high-speed vehicle movement. These challenges have put a severe test on the existing distributed resource allocation mechanism, and there is an urgent need for in-depth theoretical research and practical innovation in the field of C-V2X distributed resource allocation. The European Telecommunications Standards Institute has introduced two key safety messages for reliable communication between vehicles in the C-V2X system: DENM and CAM. DENM is mainly used for road hazard warning. When the on-board system detects safety hazards, such as emergency braking of the vehicle in front or road construction warnings, the relevant application will quickly issue DENM. CAM is a broadcast message used for communication between vehicles. It carries real-time information about the vehicle, including location, speed, driving direction, and braking status, and is designed to enhance the vehicle's perception of the surrounding traffic environment. CAM is different from the on-demand transmission mechanism of DENM and adopts a continuous transmission method. This method requires that the transmission of CAM needs to occupy wireless resources at a high frequency and continuously, so that its resource demand presents a continuous and dynamic characteristic. In order to meet this demand, the system needs to continuously allocate wireless resources dynamically. 3GPP has developed the SB-SPS algorithm, which has become the standard distributed resource allocation algorithm in the direct communication mode. The main idea of ​​the SB-SPS algorithm is that each vehicle autonomously selects communication resources and decides which resources to use for communication by monitoring the communication resources in the resource pool to avoid interference with other vehicles. The SB-SPS algorithm is mainly divided into resource perception, resource selection and resource reservation. Although vehicles using the SB-SPS algorithm can decide on their own which communication resources to use based on local observations and reduce the probability of resource conflicts to a certain extent, they still face the problem of not being able to effectively deal with non-periodic CAM.The researchers proposed an adaptive resource reservation interval method based on SB-SPS, using fuzzy functions to fuzzify the road vehicle density and communication resource busy rate as input parameters of the fuzzy inference engine; the fuzzy inference engine calculates the RRI membership and infers the matching RRI value state based on the fuzzy rule base; defuzzification, selects the best RRI value and applies it to the SB-SPS scheduling mechanism to allocate communication resources.

[0003] However, the above methods have the following shortcomings: 1) Prediction accuracy: Although fuzzy reasoning has certain adaptability, its prediction accuracy may not be as good as that of deep learning models. The prediction results of fuzzy reasoning may be affected by the limitations and subjectivity of the rule base. 2) Real-time and computational efficiency: The computational efficiency of fuzzy reasoning is not as good as that of deep learning models, especially in the Internet of Vehicles environment where real-time decision-making is required, computational efficiency is an important consideration. 3) Establishment and maintenance of fuzzy rule base: Fuzzy reasoning technology is used to dynamically adjust resource reservation intervals, which requires the establishment and maintenance of a fuzzy rule base. The establishment and adjustment of the fuzzy rule base can be a complex and time-consuming process, and its performance is highly dependent on the quality and quantity of the rules. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a C-V2X vehicle network wireless communication resource allocation method to solve the problem that the existing SB-SPS algorithm cannot effectively deal with non-periodic CAM.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A C-V2X vehicle network wireless communication resource allocation method, comprising:

[0007] Integrate the lateral coordinate data, longitudinal coordinate data, speed data and heading angle data of the target vehicle as input features;

[0008] Inputting the input features into the CAM generation interval prediction model for prediction, and obtaining a CAM generation interval category prediction result;

[0009] Updating the CAM generation interval category prediction result to the RRI field in the side link control information;

[0010] The pre-training process of the CAM generation interval prediction model includes:

[0011] Build a hyperparameter optimization module that includes the Bayesian optimization traversal probability algorithm and sampling algorithm;

[0012] Construct a prediction module including input layer, LSTM layer, fully connected layer, Softmax layer, and classification layer;

[0013] The core hyperparameters of the LSTM layer are optimized using the hyperparameter optimization module, and the optimized prediction module is determined as the CAM generation interval prediction model.

[0014] Preferably, the input features are:

[0015]

[0016] Where x = [x 1 ,x 2 ,...,x n ] T ; y=[y 1 ,y 2 ,...,y n ] T ; v = [v 1 ,v 2 ,...,v n ] T ; θ=[θ 1 ,θ 2 ,...,θ n ] T ; X is the input feature; x is the horizontal coordinate data; x n is the nth data in the horizontal coordinate data; y is the vertical coordinate data; y n is the nth data in the longitudinal coordinate data; v is the speed data; v n is the nth data in the speed data; θ is the heading angle data; θ n is the nth data in the heading angle data; n represents the sequence length.

[0017] Preferably, the Bayesian optimization traversal probability algorithm includes:

[0018] w=gp(n(u),k(u,u′));

[0019] Wherein, n(x) represents the mean function; k(u,u') represents the covariance function; gp() is a Gaussian process; u is a training sample of the Gaussian process; and w is the function value of the Gaussian process.

[0020] Preferably, the sampling algorithm comprises:

[0021]

[0022] Among them, f(z) is the objective function value, is the distribution function; μ(z) and σ(z) are the mean and variance of the Gaussian process respectively; z is the input sample point for evaluation; α is the hyperparameter for exploring the maximum space; P() is the probability function; f(z max) is the optimal solution found.

[0023] Preferably, the CAM generates an interval category prediction result in a range of 100 to 1000 milliseconds.

[0024] Preferably, it also includes:

[0025] In the input layer, the input features are preprocessed using the Min-Max normalization formula.

[0026] Preferably, it also includes:

[0027] In the fully connected layer, the hidden states output by the LSTM layer are integrated to obtain fusion features;

[0028] In the Softmax layer, the fused features are converted into probability distribution;

[0029] At the classification layer, the candidate CAM generation interval category with the greatest probability is selected according to the probability distribution order as the CAM generation interval category prediction result.

[0030] Preferably, the core hyperparameters include: initial learning rate, number of hidden layer neurons, number of iterations and dropout coefficient.

[0031] The present invention discloses the following technical effects:

[0032] The present invention provides a C-V2X vehicle network wireless communication resource allocation method. By updating the CAM generation interval category prediction result to the RRI field in the side link control information, the problem that the existing SB-SPS algorithm cannot effectively deal with non-periodic CAM is solved, and the RRI field is adaptively configured according to input features. By integrating horizontal coordinate data, longitudinal coordinate data, speed data and heading angle data as input features and constructing a CAM generation interval prediction model, the problem of poor prediction accuracy and speed of the existing model is solved, and the prediction of the CAM generation interval using a deep learning model is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0034] Figure 1 A schematic diagram of a C-V2X vehicle network wireless communication resource allocation process provided by an embodiment of the present invention;

[0035] Figure 2A framework diagram of a resource allocation method provided by an embodiment of the present invention;

[0036] Figure 3 A framework diagram of a CAM generation interval prediction model provided by an embodiment of the present invention;

[0037] Figure 4 A schematic diagram of prediction-based resource selection and reservation provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the 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 the present invention.

[0039] The purpose of the present invention is to provide a C-V2X vehicle network wireless communication resource allocation method to solve the problem that the existing SB-SPS algorithm cannot effectively deal with non-periodic CAM.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] Figure 1 A schematic diagram of a C-V2X vehicle network wireless communication resource allocation process provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides a C-V2X vehicle network wireless communication resource allocation method, comprising:

[0042] Step 100: Integrate the lateral coordinate data, longitudinal coordinate data, speed data and heading angle data of the target vehicle as input features;

[0043] Step 200: inputting the input features into a CAM generation interval prediction model for prediction, and obtaining a CAM generation interval category prediction result;

[0044] Step 300: Update the CAM generation interval category prediction result to the RRI field in the side link control information;

[0045] The pre-training process of the CAM generation interval prediction model includes:

[0046] Step 201: construct a hyperparameter optimization module including a Bayesian optimization traversal probability algorithm and a sampling algorithm;

[0047] Step 202: construct a prediction module including an input layer, an LSTM layer, a fully connected layer, a Softmax layer, and a classification layer;

[0048] Step 203: Utilize the hyperparameter optimization module to optimize the core hyperparameters of the LSTM layer, and determine the optimized prediction module as the CAM generation interval prediction model.

[0049] Furthermore, the input features are:

[0050]

[0051] Where x = [x 1 ,x 2 ,...,x n ] T ; y=[y 1 ,y 2 ,...,y n ] T ; v = [v 1 ,v 2 ,...,v n ] T ; θ=[θ 1 ,θ 2 ,...,θ n ] T ; X is the input feature; x is the horizontal coordinate data; x n is the nth data in the horizontal coordinate data; y is the vertical coordinate data; y n is the nth data in the longitudinal coordinate data; v is the speed data; v n is the nth data in the speed data; θ is the heading angle data; θ n is the nth data in the heading angle data; n represents the sequence length.

[0052] Specifically, the Bayesian optimization traversal probability algorithm includes:

[0053] w=gp(n(u),k(u,u′));

[0054] Wherein, n(x) represents the mean function; k(u,u') represents the covariance function; gp() is a Gaussian process; u is a training sample of the Gaussian process; and w is the function value of the Gaussian process.

[0055] Furthermore, the sampling algorithm includes:

[0056]

[0057] Among them, f(z) is the objective function value, is the distribution function; μ(z) and σ(z) are the mean and variance of the Gaussian process respectively; z is the input sample point for evaluation; α is the hyperparameter for exploring the maximum space; P() is the probability function; f(z max ) is the optimal solution found.

[0058] Optionally, the CAM generates an interval category prediction result in a range of 100 to 1000 milliseconds.

[0059] Preferably, it also includes:

[0060] In the input layer, the input features are preprocessed using the Min-Max normalization formula.

[0061] Furthermore, it also includes:

[0062] In the fully connected layer, the hidden states output by the LSTM layer are integrated to obtain fusion features;

[0063] In the Softmax layer, the fused features are converted into probability distribution;

[0064] At the classification layer, the candidate CAM generation interval category with the greatest probability is selected according to the probability distribution order as the CAM generation interval category prediction result.

[0065] Preferably, the core hyperparameters include: initial learning rate, number of hidden layer neurons, number of iterations and dropout coefficient.

[0066] Specifically, the challenge encountered by the SB-SPS algorithm in non-periodic CAM transmission is mainly due to resource conflicts caused by the mismatch between the resource reservation interval and the CAM generation interval. However, ETSI (European Telecommunications Standards Institute) stipulates that the generation of CAM is closely related to changes in the vehicle's position, speed, heading angle, etc. Therefore, this embodiment considers using a deep learning method to predict the CAM generation interval by observing the vehicle's state changes. Based on the prediction results, the resource reservation mechanism in the SB-SPS algorithm is further adjusted and optimized to reduce resource conflicts and improve the overall performance of the algorithm. This embodiment designs a resource allocation method based on CAM generation interval prediction on the basis of the SB-SPS algorithm, and its structure is as follows: Figure 2 shown.

[0067] Furthermore, from the perspective of deep learning, the task of predicting the vehicle CAM generation interval belongs to the category of supervised learning, and it is appropriate to use the discriminant model method. By establishing a parameter model, the mapping relationship h(s) from the vehicle historical dynamic sequence s (position, speed, heading angle) to the next CAM generation interval data Δt is learned:

[0068] h:Δt=h(s)

[0069] The Long Short-Term Memory (LSTM) network has good time series data learning capabilities and can effectively model patterns and dependencies in time series data. It exhibits powerful information mining capabilities and deep representation capabilities when dealing with time series problems, and has achieved a series of major breakthroughs in speech recognition, machine translation, and other fields. Therefore, considering the close correlation between the kinetic sequence data s and the CAM generation interval, and the continuity and dependency of the kinetic sequence data in time, this embodiment selects the LSTM neural network as the basic model for CAM generation interval prediction.

[0070] Specifically, the selection of input features. In the process of building a prediction model, the selection of a set of representative input features is crucial to the performance and generalization ability of the model. These input features need to effectively extract the core information in the data and maintain a high correlation with the target variable, namely the generation time interval of the CAM. According to the CAM generation rules specified by ETSI, it can be seen that the generation of CAM is closely related to vehicle dynamic parameters such as vehicle position changes, speed fluctuations, and heading angle adjustments. In order to ensure that the model can deeply explore and effectively utilize key information directly related to CAM generation, thereby improving the accuracy of the prediction results, the time series data of the vehicle's lateral coordinates, longitudinal coordinates, speed, and heading angle are used as the input features of the model. Specifically, the lateral coordinate data x = [x 1 ,x 2 ,...,x n ] T , vertical coordinate data y=[y 1 ,y 2 ,...,y n ] T , speed data v = [v 1 ,v 2 ,...,v n ] T , heading angle data θ=[θ 1 ,θ 2 ,...,θ n ] T , where n represents the sequence length. A single input sample is expressed as follows:

[0071]

[0072] Furthermore, the model is constructed. The prediction model predicts the time interval data of the next CAM generation in the future through the historical moment dynamic sequence data and the current moment dynamic data. The CAM generation interval prediction model in this embodiment is composed of two modules: hyperparameter optimization and prediction. The overall architecture of the model is as follows Figure 3 As shown:

[0073] 1) Hyperparameter optimization module

[0074] The core hyperparameters of the CAM generation interval prediction model include: initial learning rate, number of hidden layer neurons, number of iterations, and discard coefficient. Taking into account the speed and optimal solution of hyperparameter optimization, this embodiment proposes a hyperparameter optimization algorithm for the LSTM model based on Bayesian Optimization to optimize the hyperparameters of the LSTM model. The Bayesian optimization algorithm is based on Bayes' theorem and continuously updates the posterior distribution of the objective function to guide the optimization process. The probabilistic proxy model and the sampling function are the two core parts of Bayesian optimization. The probabilistic proxy model is used as a probabilistic model substitute for the unknown objective function, and the prior probability is continuously corrected through the information collected during the iteration process, thereby improving the accuracy of the probabilistic proxy model and reducing the number of attempts at hyperparameters. The Gaussian process is the most common Bayesian optimization traversal probability model:

[0075] w=gp(n(u),k(u,u′))

[0076] By utilizing a sampling function based on the posterior distribution, the algorithm assigns priority to unexplored samples in the region that is considered most likely to contain the global optimal solution. The sample points with the greatest potential are selected from the candidate set to minimize the value of the loss function. The boost probability function is the most commonly used sampling function:

[0077]

[0078] Among them, α=0 makes the value converge to f(z max ).

[0079] 2) Prediction Module

[0080] The prediction module takes the vehicle kinematic time series information as input, and after LSTM learning, uses the Softmax function to calculate the probability of the next CAM generation interval being 100 to 1000 milliseconds (with 100 milliseconds as an interval), and finally selects the category with the highest probability as the prediction output. The prediction module consists of three parts: input layer, LSTM layer, and output layer:

[0081] Input layer: Determine the model input variables and preprocess the raw data. Before the original time series data of vehicle kinematics (including the vehicle's lateral coordinates, longitudinal coordinates, speed, and heading angle) is input into the network model, in order to narrow the differences between the data, avoid the disappearance or excessive expansion of the gradient during the training process, and speed up the convergence of the algorithm, a data standardization preprocessing technology, namely Min-Max normalization, is used. Through this operation, the data set is adjusted to the standardized interval of [-1,1] to ensure the uniformity of the data. The basic idea of ​​Min-Max normalization is to scale the raw data to a specific range so that the value of each feature is within a relatively fixed interval, thereby eliminating the influence of dimension and numerical size. This method is particularly suitable for features that have significant differences in order of magnitude, which can prevent certain features from dominating the model training process, thereby improving the performance and prediction accuracy of the model. The normalization formula used is as follows:

[0082]

[0083] Among them, m' is the normalized value, m is the original value, and m max and m min are the maximum and minimum values ​​of the sample data respectively.

[0084] LSTM layer: Let X be the input feature vector after data preprocessing. At each time t, the LSTM unit reads the input at the current time and the hidden state h at the previous time. t-1 , the hidden state h at the current moment t Through the hidden state h of the previous moment t-1 , cell status c t-1 And the current input X t The following formula describes the update process of the hidden layer state of the LSTM unit from time t-1 to time t:

[0085] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0086] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0087] a t =tanh(W c ·[h t-1 ,x t ]+bc )

[0088] c t =f t ·c t-1 +i t ·a t

[0089] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0090] h t =o t ·tanh(c t )

[0091] Among them, [W f ,b f ]、[W i ,b i ]、[W c ,b c ]、[W o ,b o ] respectively correspond to the forget gate (f t ), input gate (i t ), unit status (c t ) and output gate (o t )The weight matrix and bias term at time t; [h t-1 ,x t ] are the output of the hidden layer at time t-1 and the input at time t; σ and tanh are the sigmoid and tanh activation functions respectively. The update process of the above formula can be simplified as:

[0092] [h t ,c t ],=F(h t-1 ,X t )

[0093] The LSTM layer learns the long-term dependency between the vehicle's historical kinematic sequence data and the CAM generation interval in this way. The LSTM layer finally outputs the hidden state h t .

[0094] Output layer: The output of the LSTM layer is then passed to the fully connected layer, which performs deep integration processing on the input data to further enhance the representation ability of the features. The processed features pass through the Softmax layer, which converts the output into a probability distribution to reflect the probability of each time interval being predicted. Finally, the probability distribution is input to the classification layer, which is responsible for extracting the most likely time interval from the probability distribution to derive the next CAM generation interval. Specifically, let C = (c 1 ,c 2 ,...,c 10 ) is the CAM generated interval category vector output by the prediction module, c 1 ,c 2 ,...,c 10 Respectively represent 100 milliseconds, 200 milliseconds, ..., 1000 milliseconds 10 kinds of CAM generation interval categories; Ω is the vector composed of the probability of each CAM generation interval category, ω i (i=1,2,...,10) represent the probabilities of the categories being 100 milliseconds, 200 milliseconds, ..., 1000 milliseconds respectively, then the output of the Softmax layer is:

[0095] ω i =P(c i |X),Ω=(ω 1 ,ω 2 ,...,ω 10 )

[0096] Next, the output of the Softmax layer is input to the classification layer to convert the output of the network into a category label, as shown in the following formula:

[0097] c*=argmaxP(c i |X)

[0098] Where c* is the final output CAM generation interval category.

[0099] Furthermore, in the C-V2X system, the chain reservation of equally spaced resources is achieved through the resource reservation interval RRI field in the sidelink control information (Sidelink Control Information, SCI). When a time-frequency resource RB needs to be used continuously for k times, the RRI field of the currently sent CAM will carry the information of the next reserved resource (including the resource location and time interval). With the transmission of each CAM, the reservation chain structure will continue to expand until the randomly selected reselection counter (RC) is reached, at which time the RRI value is set to 0, indicating that the RB is no longer reserved. For the case where the RRI value is 100 milliseconds, the RC value is randomly selected between 5 and 15, and then a new time-frequency resource will be selected. Finally, the RRI value of 0 indicates to other vehicles that the time-frequency resource is no longer reserved, and other vehicles can reselect the resource for transmission. In the method proposed in this embodiment, the predictive reservation of resources is achieved by replacing the original fixed RRI value with a predicted value in the RRI field of the SCI. For example, if the predicted value is 400ms, the value is encoded in the RRI field. Resource selection and reservation based on predictions Figure 4 shown.

[0100] The beneficial effects of the present invention are as follows:

[0101] The present invention solves the problem that the existing SB-SPS algorithm cannot effectively deal with non-periodic CAM by updating the CAM generation interval category prediction result to the RRI field in the side link control information, and improves the rationality and flexibility of wireless communication resource allocation.

[0102] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0103] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A C-V2X vehicle network wireless communication resource allocation method, characterized in that: include: Integrate the lateral coordinate data, longitudinal coordinate data, speed data and heading angle data of the target vehicle as input features; Inputting the input features into the CAM generation interval prediction model for prediction, and obtaining a CAM generation interval category prediction result; Updating the CAM generation interval category prediction result to the RRI field in the side link control information; The pre-training process of the CAM generation interval prediction model includes: Build a hyperparameter optimization module that includes the Bayesian optimization traversal probability algorithm and sampling algorithm; Construct a prediction module including input layer, LSTM layer, fully connected layer, Softmax layer, and classification layer; The core hyperparameters of the LSTM layer are optimized using the hyperparameter optimization module, and the optimized prediction module is determined as the CAM generation interval prediction model.

2. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: The input features are: Where x=[x1,x2,...,x n ] T ; y=[y1,y2,...,y n ] T ; v = [v1, v2, ..., v n ] T ; θ=[θ1,θ2,...,θ n ] T ; X is the input feature; x is the horizontal coordinate data; x n is the nth data in the horizontal coordinate data; y is the vertical coordinate data; y n is the nth data in the longitudinal coordinate data; v is the speed data; v n is the nth data in the speed data; θ is the heading angle data; θ n is the nth data in the heading angle data; n represents the sequence length.

3. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: The Bayesian optimization traversal probability algorithm includes: w=gp(n(u),k(u,u′)); Wherein, n(x) represents the mean function; k(u,u') represents the covariance function; gp() is a Gaussian process; u is a training sample of the Gaussian process; and w is the function value of the Gaussian process.

4. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: The sampling algorithm includes: Among them, f(z) is the objective function value, is the distribution function; μ(z) and σ(z) are the mean and variance of the Gaussian process respectively; z is the input sample point for evaluation; α is the hyperparameter for exploring the maximum space; P() is the probability function; f(z max ) is the optimal solution found.

5. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: The CAM generates interval category prediction results ranging from 100 to 1000 milliseconds.

6. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: Also includes: In the input layer, the input features are preprocessed using the Min-Max normalization formula.

7. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: Also includes: In the fully connected layer, the hidden states output by the LSTM layer are integrated to obtain fusion features; In the Softmax layer, the fused features are converted into probability distribution; At the classification layer, the candidate CAM generation interval category with the greatest probability is selected according to the probability distribution order as the CAM generation interval category prediction result.

8. A C-V2X vehicle network wireless communication resource allocation method according to claim 1, characterized in that: The core hyperparameters include: initial learning rate, number of hidden layer neurons, number of iterations, and dropout coefficient.

Images