Communication performance analysis method based on time-space frequency resources

Abstract

The invention discloses a communication performance analysis method based on space-time-frequency resources, which belongs to the technical field of communication performance analysis, and comprises the following steps of: acquiring a space-time-frequency resource type and a communication performance analysis index, and then generating a communication performance analysis test task according to the space-time-frequency resource type and the communication performance analysis index; after the communication performance analysis test task is executed, the relationship between the actual value of the space-time-frequency resource type and the actual value of the communication performance analysis index can be learned through artificial intelligence, so that predictive analysis can be performed on the communication performance analysis index through the actual value of the space-time-frequency resource type in the subsequent process; therefore, the staff can quickly adjust the distribution of the space-time-frequency resources according to the requirements.

Description

Technical Field

[0001] The present invention belongs to the technical field of communication performance analysis, and particularly relates to a communication performance analysis method based on spatio-temporal-frequency resources. Background Art

[0002] With the continuous emergence of emerging scenarios such as vehicle-to-everything (V2X) networks and unmanned aerial vehicle (UAV) networks, communication-sensing integrated systems have received extensive attention in recent years. Through integrated design, a communication-sensing integrated system realizes the dual functions of data transmission and target detection based on the same hardware and spectrum resources, reducing the hardware cost and improving the system integration and spectrum resource utilization performance. Spatio-temporal-frequency resources refer to the concept of effectively managing and utilizing time, space, and frequency resources in communication and network technologies. These resources are particularly important in wireless communication and networks, especially in 5G and future communication technologies. In the prior art, in the process of analyzing communication performance, a single index is often used for analysis, and there is no predictability, making it difficult to adjust the system communication performance in advance. Summary of the Invention

[0003] The present invention provides a communication performance analysis method based on spatio-temporal-frequency resources to solve the problems that the prior art uses a single index for analysis, has no predictability, and is difficult to adjust the system communication performance in advance.

[0004] A communication performance analysis method based on spatio-temporal-frequency resources includes:

[0005] Obtain the spatio-temporal-frequency resource type and communication performance analysis index input by a staff member, and generate a communication performance analysis test task according to the spatio-temporal-frequency resource type and communication performance analysis index;

[0006] Execute the communication performance analysis test task to execute the sample value corresponding to the spatio-temporal-frequency resource type, so as to obtain the actual value corresponding to the communication performance analysis index;

[0007] Construct a communication performance analysis model using an artificial intelligence model, and use the sample value of the spatio-temporal-frequency resource type as the input and the actual value of the corresponding communication performance analysis index as the expected output to optimize the communication performance analysis model, and obtain the optimized communication performance analysis model;

[0008] Collect the allocation value corresponding to the spatio-temporal-frequency resource type, and use the optimized communication performance analysis model to identify the allocation value corresponding to the spatio-temporal-frequency resource type to obtain the communication performance analysis result.

[0009] Further, obtaining the spatio-temporal-frequency resource type and communication performance analysis index input by a staff member, and generating a communication performance analysis test task according to the spatio-temporal-frequency resource type and communication performance analysis index includes:

[0010] Obtain the spatio-temporal-frequency resource types input by the staff; among them, each spatio-temporal-frequency resource type has its corresponding upper limit and lower limit;

[0011] Based on the fact that each spatio-temporal-frequency resource type has its corresponding upper limit and lower limit, use the equal division method to divide the numerical interval corresponding to the spatio-temporal-frequency resource type into a fixed number of parts, and obtain multiple numerical division points corresponding to each spatio-temporal-frequency resource type;

[0012] Arrange and combine the multiple numerical division points corresponding to all spatio-temporal-frequency resource types to obtain sample values corresponding to multiple spatio-temporal-frequency resource types;

[0013] Generate a task to be executed for each sample value corresponding to a spatio-temporal-frequency resource type.

[0014] Further, execute the communication performance analysis test task to execute the sample values corresponding to the spatio-temporal-frequency resource types, so as to obtain the actual values of the communication performance analysis indicators, including:

[0015] Execute the communication performance analysis test task to use the sample values corresponding to the spatio-temporal-frequency resource types as the control parameters of the communication system to be analyzed, and obtain the actual values of the communication performance analysis indicators of the communication system to be analyzed.

[0016] Further, use an artificial intelligence model to construct a communication performance analysis model, including: using a CNN model to construct a communication performance analysis model.

[0017] Further, use the sample values of the spatio-temporal-frequency resource types as the input and the actual values of the corresponding communication performance analysis indicators as the expected output to optimize the communication performance analysis model, and obtain the optimized communication performance analysis model, including:

[0018] For the hyperparameters of the communication performance analysis model, initialize the hyperparameters and encode the initialized hyperparameters into vectors to obtain hyperparameter vectors, and at the same time obtain multiple different hyperparameter vectors;

[0019] For any hyperparameter vector, use the sample values of the spatio-temporal-frequency resource types as the input and the actual values of the corresponding communication performance analysis indicators as the expected output to obtain the loss function value corresponding to the hyperparameter vector;

[0020] Determine the optimal hyperparameter vector according to the loss function values corresponding to all hyperparameter vectors;

[0021] For any hyperparameter vector, according to the optimal hyperparameter vector, and use the local personalized search strategy to perform the first local search on the hyperparameter vector to obtain the hyperparameter vector after the first local search;

[0022] For any hyperparameter vector after the first local search, according to the optimal hyperparameter vector, a second local search is performed on the hyperparameter vector using a non-linear local search strategy to obtain the hyperparameter vector after the second local search;

[0023] For the hyperparameter vector after the second local search, a first global search is performed on the hyperparameter vector using a reverse global greedy search strategy to obtain the hyperparameter vector after the first global search;

[0024] For the hyperparameter vector after the first global search, a second global search is performed on the hyperparameter vector using a random global greedy search strategy to obtain the hyperparameter vector after the second global search;

[0025] Determine whether the optimization end condition is satisfied. If so, according to the hyperparameter vector after the second global search, the optimal hyperparameter vector is re-obtained, and the parameters in the optimal hyperparameter vector are applied to the communication performance analysis model to obtain the optimized communication performance analysis model. Otherwise, return to the step of obtaining the loss function value.

[0026] Further, for any hyperparameter vector, according to the optimal hyperparameter vector, a first local search is performed on the hyperparameter vector using a local personalized search strategy to obtain the hyperparameter vector after the first local search, including:

[0027] Based on the current number of optimization times, an adaptively decreasing weight control factor is generated;

[0028] According to the weight control factor and the optimal hyperparameter vector, an optimal direction search is performed on the hyperparameter vector to obtain an optimal information search term;

[0029] The center vector of all hyperparameter vectors is obtained, and according to the weight control factor and the center vector, a comprehensive direction search is performed on the hyperparameter vector to obtain a comprehensive information search term;

[0030] According to the optimal information search term and the comprehensive information search term, a first local search is performed on the hyperparameter vector to obtain the hyperparameter vector after the first local search.

[0031] Further, for any hyperparameter vector after the first local search, according to the optimal hyperparameter vector, a second local search is performed on the hyperparameter vector using a non-linear local search strategy to obtain the hyperparameter vector after the second local search, including:

[0032] For any hyperparameter vector after the first local search, a cosine information search term between the hyperparameter vector and the optimal hyperparameter vector is obtained;

[0033] Perform a second local search on the hyperparameter vector according to the cosine information search term and the weight control factor to obtain the hyperparameter vector after the second local search.

[0034] Further, for the hyperparameter vector after the second local search, perform a first global search on the hyperparameter vector using a reverse global greedy search strategy to obtain the hyperparameter vector after the first global search, including:

[0035] Based on the current optimization times, obtain a non-linear perturbation factor, and for the hyperparameter vector after the second local search, use the non-linear perturbation factor to perturb the hyperparameter vector to generate a perturbation information vector;

[0036] Generate a reference information vector according to the non-linear perturbation factor, and perform a global search on the hyperparameter vector according to the reference information vector and the perturbation information vector to obtain a first global search vector;

[0037] Judge whether the loss function value of the first global search vector decreases. If so, use the first global search vector as the hyperparameter vector after the first global search; otherwise, directly use the original hyperparameter vector after the second local search as the hyperparameter vector after the first global search.

[0038] Further, for the hyperparameter vector after the first global search, perform a second global search on the hyperparameter vector using a random global greedy search strategy to obtain the hyperparameter vector after the second global search, including:

[0039] For the hyperparameter vector after the first global search, generate a random flight individual for the hyperparameter vector according to the optimal hyperparameter vector;

[0040] According to the random flight individual corresponding to the hyperparameter vector, perform a second global search on the hyperparameter vector using a cosine function to obtain a second global search vector;

[0041] Judge whether the loss function value of the second global search vector decreases. If so, use the second global search vector as the hyperparameter vector after the second global search; otherwise, directly use the original hyperparameter vector after the first global search as the hyperparameter vector after the second global search.

[0042] Further, judge whether the optimization end condition is satisfied, including:

[0043] Judge whether the current optimization times are greater than or equal to the maximum optimization times. If so, determine that the optimization end condition is satisfied; otherwise, determine that the optimization end condition is not satisfied.

[0044] A communication performance analysis method based on spatio-temporal-frequency resources provided by the present invention collects the types of spatio-temporal-frequency resources and communication performance analysis metrics, then generates a communication performance analysis test task according to the types of spatio-temporal-frequency resources and communication performance analysis metrics. After executing the communication performance analysis test task, the relationship between the actual values of the spatio-temporal-frequency resource types and the actual values of the communication performance analysis metrics can be learned through artificial intelligence, so that in the subsequent process, the communication performance analysis metrics can be predictively analyzed based on the actual values of the spatio-temporal-frequency resource types, enabling staff to quickly adjust the allocation of spatio-temporal-frequency resources according to requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings are incorporated herein and form a part of this specification, showing embodiments consistent with the present invention and, together with the specification, are used to explain the principles of the present invention.

[0046] Figure 1 It is a flowchart of a communication performance analysis method based on spatio-temporal-frequency resources provided by an embodiment of the present invention.

[0047] Through the above accompanying drawings, specific embodiments of the present invention have been shown, and there will be more detailed descriptions hereinafter. These drawings and written descriptions are not intended to limit the scope of the inventive concept in any way, but to illustrate the concept of the present invention to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0049] Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0050] As Figure 1 shown, an embodiment of the present invention provides a communication performance analysis method based on spatio-temporal-frequency resources, including:

[0051] S1. Obtain the types of spatio-temporal-frequency resources and communication performance analysis metrics input by staff, and generate a communication performance analysis test task according to the types of spatio-temporal-frequency resources and communication performance analysis metrics;

[0052] Spatio-temporal-frequency resources refer to the concept of effectively managing and utilizing time, space, and frequency resources in communication and network technologies. These resources are the core components of wireless communication and networks, especially in 5G, 6G, and future communication technologies. Spatio-temporal-frequency resources often have an impact on communication performance. Therefore, after determining the type of spatio-temporal-frequency resources input by the staff, communication performance analysis and test tasks can be generated based on the value range of the type of spatio-temporal-frequency resources, so as to obtain the actual values of communication performance analysis indicators.

[0053] S2. Execute the communication performance analysis and test tasks to execute the sample values corresponding to the spatio-temporal-frequency resource type, so as to obtain the actual values corresponding to the communication performance analysis indicators;

[0054] The communication performance analysis and test tasks can be combined test tasks of multiple spatio-temporal-frequency resources. These spatio-temporal-frequency resources can be applied to the communication system, and then the actual values corresponding to the communication performance analysis indicators can be obtained. By analyzing and learning these data, the impact of spatio-temporal-frequency resources on communication performance can be determined.

[0055] S3. Use an artificial intelligence model to construct a communication performance analysis model, and use the sample values of the spatio-temporal-frequency resource type as the input and the actual values of the corresponding communication performance analysis indicators as the expected output to optimize the communication performance analysis model, and obtain the optimized communication performance analysis model;

[0056] The artificial intelligence model is the core component in the field of artificial intelligence (AI). It refers to a collection of a series of algorithms and mathematical models. Therefore, in the embodiments of the present invention, an artificial intelligence model is used to construct a communication performance analysis model, and the relationship between the sample values of the spatio-temporal-frequency resource type and the actual values of the communication performance analysis indicators can be learned through artificial intelligence.

[0057] S4. Collect the allocation values corresponding to the spatio-temporal-frequency resource type, and use the optimized communication performance analysis model to identify the allocation values corresponding to the spatio-temporal-frequency resource type to obtain the communication performance analysis result.

[0058] After the communication performance analysis model is optimized, it has the ability to identify the allocation values corresponding to the spatio-temporal-frequency resource type. Therefore, after the staff formulates the allocation plan of the spatio-temporal-frequency resources, they can pre-analyze the communication performance corresponding to the allocation plan of the spatio-temporal-frequency resources, which is convenient for the staff to optimize the allocation plan of the spatio-temporal-frequency resources.

[0059] A communication performance analysis method based on spatio-temporal-frequency resources provided by the present invention collects spatio-temporal-frequency resource types and communication performance analysis metrics, and then generates a communication performance analysis test task according to the spatio-temporal-frequency resource types and communication performance analysis metrics. After executing the communication performance analysis test task, the relationship between the actual values of the spatio-temporal-frequency resource types and the actual values of the communication performance analysis metrics can be learned through artificial intelligence, so that in the subsequent process, the communication performance analysis metrics can be predictively analyzed through the actual values of the spatio-temporal-frequency resource types, enabling staff to quickly adjust the allocation of spatio-temporal-frequency resources according to requirements.

[0060] In an embodiment of the present invention, obtaining the spatio-temporal-frequency resource types and communication performance analysis metrics input by a staff member, and generating a communication performance analysis test task according to the spatio-temporal-frequency resource types and communication performance analysis metrics includes:

[0061] Obtaining the spatio-temporal-frequency resource types input by a staff member; wherein each spatio-temporal-frequency resource type has its corresponding upper limit and lower limit;

[0062] For example: The spatio-temporal-frequency resources can be transmission rate, beam scheduling, time allocation, etc. For any communication system, the spatio-temporal-frequency resources have corresponding upper and lower limits. If the spatio-temporal-frequency resources are certain fixed options, then the numerical division points of this spatio-temporal-frequency resource are those fixed options. For example, if the spatio-temporal-frequency resources are of types A1, A2, and A3, then the numerical division points are A1, A2, and A3.

[0063] Based on the fact that each spatio-temporal-frequency resource type has its corresponding upper limit and lower limit, the numerical interval corresponding to the spatio-temporal-frequency resource type is divided into a fixed number of parts by an equal division method to obtain multiple numerical division points corresponding to each spatio-temporal-frequency resource type (for example, if the transmission rate interval is (B, C), then (B, C) can be equally divided into N sub-intervals, and each boundary is a numerical division point);

[0064] Arranging and combining the multiple numerical division points corresponding to all spatio-temporal-frequency resource types to obtain sample values corresponding to multiple spatio-temporal-frequency resource types;

[0065] Generating a task to be executed for each sample value corresponding to a spatio-temporal-frequency resource type.

[0066] For example, if the numerical division points of the spatio-temporal-frequency resource A are a1 and a2, and the numerical division points of the spatio-temporal-frequency resource B are b1, b2, and b3, then the sample values corresponding to the spatio-temporal-frequency resource types can be (a1, b1), (a1, b2), (a1, b3), (a2, b1), (a2, b2), and (a2, b3). Any one of the sample values can be input into the communication system to obtain the actual value of the corresponding communication performance analysis index, so as to obtain the sample data for optimizing the artificial intelligence model.

[0067] In the embodiment of the present invention, performing the communication performance analysis test task to execute the sample value corresponding to the spatio-temporal-frequency resource type, so as to obtain the actual value corresponding to the communication performance analysis index, includes:

[0068] Performing the communication performance analysis test task to use the sample value corresponding to the spatio-temporal-frequency resource type as the control parameter of the communication system to be analyzed, and obtaining the actual value corresponding to the communication performance analysis index of the communication system to be analyzed.

[0069] In the embodiment of the present invention, an artificial intelligence model is used to construct a communication performance analysis model, including: using a CNN (Convolutional Neural Network) model to construct a communication performance analysis model. It should be noted that other artificial intelligence models can also be used to construct a communication performance analysis model.

[0070] In the embodiment of the present invention, using the sample value of the spatio-temporal-frequency resource type as the input and the actual value of the corresponding communication performance analysis index as the expected output, the communication performance analysis model is optimized to obtain the optimized communication performance analysis model, including:

[0071] For the hyperparameters of the communication performance analysis model, the hyperparameters are initialized, and the initialized hyperparameters are encoded as vectors to obtain hyperparameter vectors. At the same time, multiple different hyperparameter vectors are obtained;

[0072] For example, the hyperparameters of the communication performance analysis model generally have upper and lower limits, and can be randomly initialized between the upper and lower limits, so as to obtain a hyperparameter vector.

[0073] For any one of the hyperparameter vectors, using the sample value of the spatio-temporal-frequency resource type as the input and the actual value of the corresponding communication performance analysis index as the expected output, the loss function value corresponding to the hyperparameter vector is obtained; for example: the root mean square error function value or the cross-entropy error function value can be obtained;

[0074] According to the loss function values corresponding to all the hyperparameter vectors, the optimal hyperparameter vector is determined;

[0075] For any hyperparameter vector, based on the optimal hyperparameter vector, perform a first local search on the hyperparameter vector using a local personalized search strategy to obtain the hyperparameter vector after the first local search;

[0076] For any hyperparameter vector after the first local search, based on the optimal hyperparameter vector, perform a second local search on the hyperparameter vector using a non - linear local search strategy to obtain the hyperparameter vector after the second local search;

[0077] For the hyperparameter vector after the second local search, perform a first global search on the hyperparameter vector using a reverse global greedy search strategy to obtain the hyperparameter vector after the first global search;

[0078] For the hyperparameter vector after the first global search, perform a second global search on the hyperparameter vector using a random global greedy search strategy to obtain the hyperparameter vector after the second global search;

[0079] Determine whether the optimization end condition is satisfied. If so, based on the hyperparameter vector after the second global search, re - obtain the optimal hyperparameter vector, and apply the parameters in the optimal hyperparameter vector to the communication performance analysis model to obtain the optimized communication performance analysis model. Otherwise, return to the step of obtaining the loss function value.

[0080] In the prior art, the gradient descent algorithm is often used to optimize the hyperparameters of the artificial intelligence model, and there are often technical problems such as being easily trapped in local optima. Therefore, the embodiments of the present invention provide an intelligent optimization algorithm to solve the technical problems existing in the prior art and improve the prediction ability of communication performance.

[0081] In the embodiments of the present invention, for any hyperparameter vector, based on the optimal hyperparameter vector, perform a first local search on the hyperparameter vector using a local personalized search strategy to obtain the hyperparameter vector after the first local search, including:

[0082] Based on the current number of optimizations, generate an adaptively decreasing weight control factor as: where ω represents the weight control factor, π represents pi, sin represents the sine function, T represents the maximum number of optimizations, and t represents the current number of optimizations;

[0083] According to the weight control factor and the optimal hyperparameter vector, perform an optimal direction search on the hyperparameter vector to obtain the optimal information search term as: where, represents the i - th hyperparameter vector in the t - th optimization process, i = 1, 2, …, I, I represents the total number of hyperparameter vectors, Δ best represents the optimal information search term, r 1Represents a random number between (0, 1), Represents the optimal hyperparameter vector;

[0084] Obtain the central vector of all hyperparameter vectors (i.e., the vector composed of the means of each dimension), and perform a comprehensive direction search on the hyperparameter vectors according to the weight control factor and the central vector, and the comprehensive information search term is: Among them, Δ zx Represents the comprehensive information search term, r 2 Represents a random number between (0, 1), Represents the central vector of all hyperparameter vectors;

[0085] According to the optimal information search term and the comprehensive information search term, perform a first local search on the hyperparameter vectors, and the hyperparameter vectors after the first local search are: Among them, Represents the hyperparameter vector after the first local search

[0086] The local personalized search strategy provided by the embodiments of the present invention can maintain the diversity of hyperparameter vectors during the algorithm search process, thereby effectively avoiding the algorithm from falling into local optimality. At the same time, learning towards the optimal hyperparameter vector can effectively improve the convergence speed of the algorithm. In the later stage of the algorithm, the weight control factor decreases non-linearly, which can gradually improve the convergence accuracy of the algorithm.

[0087] In the embodiments of the present invention, for any hyperparameter vector after the first local search, according to the optimal hyperparameter vector, a second local search is performed on the hyperparameter vector using a non-linear local search strategy to obtain the hyperparameter vector after the second local search, including:

[0088] For any hyperparameter vector after the first local search, obtain the cosine information search term between the hyperparameter vector and the optimal hyperparameter vector as: Among them, r 4 Represents a random number between (0, 1), r 5 Represents a random number between (0, 1), Represents the jth hyperparameter vector after the first local search, j = 1, 2,..., I;

[0089] According to the cosine information search term and the weight control factor, perform a second local search on the hyperparameter vector, and the hyperparameter vector after the second local search is: Among them, Represents the hyperparameter vector after the second local search

[0090] The non - linear local search strategy provided by the embodiments of the present invention can enable all hyperparameter vectors to perform non - linear search around the optimal position, effectively improving the convergence speed and convergence accuracy of the algorithm.

[0091] In the embodiments of the present invention, for the hyperparameter vector after the second local search, a reverse global greedy search strategy is adopted to perform the first global search on the hyperparameter vector, and the hyperparameter vector after the first global search is obtained, including:

[0092] Based on the current number of optimizations, the non - linear perturbation factor is obtained as: where δ represents the non - linear perturbation factor, and T represents the preset maximum number of optimizations;

[0093] And for the hyperparameter vector after the second local search, the non - linear perturbation factor is used to perturb the hyperparameter vector to generate a perturbation information vector as: where represents the hyperparameter vector after the n - th second local search in the t - th optimization process, n = 1, 2, …, I, represents the perturbation information vector;

[0094] A reference information vector is generated according to the non - linear perturbation factor, and the hyperparameter vector is globally searched according to the reference information vector and the perturbation information vector to obtain the first global search vector as: where r 6 represents a random number between (0, 1), lb represents the upper - bound vector corresponding to the hyperparameter vector, ub represents the lower - bound vector corresponding to the hyperparameter vector, represents the first global search vector;

[0095] It is judged whether the loss function value of the first global search vector decreases. If so, the first global search vector is used as the hyperparameter vector after the first global search; otherwise, the original hyperparameter vector after the second local search is directly used as the hyperparameter vector after the first global search.

[0096] The reverse global greedy search strategy provided by the embodiments of the present invention can perform reverse search on each hyperparameter vector, thereby improving the search efficiency of the unknown region, effectively improving the traversability of the solution space, and thus avoiding falling into local optimality.

[0097] In the embodiments of the present invention, for the hyperparameter vector after the first global search, a random global greedy search strategy is adopted to perform the second global search on the hyperparameter vector, and the hyperparameter vector after the second global search is obtained, including:

[0098] For the hyperparameter vector after the first global search, according to the optimal hyperparameter vector, a random flight individual is generated for the hyperparameter vector as: Among them, represents a randomly flying individual, represents except random individuals outside, represents the hyperparameter vector after the m-th first global search in the t-th optimization process, m = 1, 2, …, I, α represents the random flight coefficient, u represents the first normal distribution random number, v represents the second normal distribution random number, and β represents the constant term;

[0099] According to the randomly flying individual corresponding to the hyperparameter vector, the cosine function is used to perform the second global search on the hyperparameter vector, and the second global search vector is obtained as: Among them, represents the second global search vector, r 7 represents a random number between (0, 1), r 8 represents a random number between (0, 1), and π represents the circumference ratio;

[0100] Judge whether the loss function value of the second global search vector decreases. If so, use the second global search vector as the hyperparameter vector after the second global search. Otherwise, directly use the hyperparameter vector after the original first global search as the hyperparameter vector after the second global search.

[0101] The second global search provided by the embodiments of the present invention can enable the algorithm to effectively jump out of the local optimal solution and provide strong global search capabilities, thereby improving the hyperparameter search ability.

[0102] Optionally, after each search for the hyperparameter vector, the hyperparameter vector can be processed to prevent it from exceeding the boundary, thereby ensuring the effectiveness of the algorithm.

[0103] In the embodiments of the present invention, determining whether the optimization end condition is satisfied includes:

[0104] Judge whether the current number of optimization times is greater than or equal to the maximum number of optimization times. If so, it is determined that the optimization end condition is satisfied. Otherwise, it is determined that the optimization end condition is not satisfied.

[0105] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily think of other embodiments of the present invention. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It should be understood that the present invention is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A communication performance analysis method based on time-space-frequency resources, characterized in that: include: Acquire the time-space-frequency resource type and communication performance analysis index input by the staff, and generate a communication performance analysis test task according to the time-space-frequency resource type and communication performance analysis index; Executing the communication performance analysis test task to execute the sample value corresponding to the time-space-frequency resource type, thereby obtaining the actual value corresponding to the communication performance analysis indicator; An artificial intelligence model is used to construct a communication performance analysis model, and the sample value of the time-space-frequency resource type is used as input, and the actual value of the corresponding communication performance analysis indicator is used as the expected output to optimize the communication performance analysis model to obtain the optimized communication performance analysis model; The allocation values ​​corresponding to the time-space-frequency resource types are collected, and the allocation values ​​corresponding to the time-space-frequency resource types are identified using the optimized communication performance analysis model to obtain communication performance analysis results.

2. The communication performance analysis method based on time-space-frequency resources according to claim 1 is characterized in that: Acquiring the time-space-frequency resource type and the communication performance analysis index input by the staff, and generating a communication performance analysis test task according to the time-space-frequency resource type and the communication performance analysis index, including: Obtaining the time-space-frequency resource type input by the staff; wherein each time-space-frequency resource type has its corresponding upper limit and lower limit; Based on the fact that each time-space-frequency resource type has its corresponding upper limit and lower limit, the numerical interval corresponding to the time-space-frequency resource type is divided into a fixed number of parts by an equal division method, so as to obtain multiple numerical division points corresponding to each time-space-frequency resource type; Arrange and combine multiple numerical division points corresponding to all time-space-frequency resource types to obtain sample values ​​corresponding to multiple time-space-frequency resource types; For each sample value corresponding to the time-space-frequency resource type, a task to be executed is generated.

3. The communication performance analysis method based on time-space-frequency resources according to claim 1 is characterized in that: Executing the communication performance analysis test task to execute the sample value corresponding to the time-space-frequency resource type, thereby obtaining the actual value corresponding to the communication performance analysis indicator, including: The communication performance analysis test task is executed to use the sample value corresponding to the time-space-frequency resource type as the control parameter of the communication system to be analyzed, and to obtain the actual value corresponding to the communication performance analysis index of the communication system to be analyzed.

4. The communication performance analysis method based on time-space-frequency resources according to claim 1 is characterized in that: An artificial intelligence model is used to build a communication performance analysis model, including: using a CNN model to build a communication performance analysis model.

5. The communication performance analysis method based on time-space-frequency resources according to claim 1 is characterized in that: The sample value of the time-space-frequency resource type is used as input, and the actual value of the corresponding communication performance analysis indicator is used as the expected output, and the communication performance analysis model is optimized to obtain the optimized communication performance analysis model, including: For the hyperparameters of the communication performance analysis model, the hyperparameters are initialized, and the initialized hyperparameters are encoded into vectors to obtain a hyperparameter vector, and multiple different hyperparameter vectors are obtained at the same time; For any hyperparameter vector, take the sample value of the time-space-frequency resource type as input, take the actual value of the corresponding communication performance analysis indicator as the expected output, and obtain the loss function value corresponding to the hyperparameter vector; Determine the optimal hyperparameter vector based on the loss function values ​​corresponding to all hyperparameter vectors; For any hyperparameter vector, according to the optimal hyperparameter vector, a first local search is performed on the hyperparameter vector using a local personalized search strategy to obtain a hyperparameter vector after the first local search; For any hyperparameter vector after the first local search, a second local search is performed on the hyperparameter vector using a nonlinear local search strategy according to the optimal hyperparameter vector to obtain a hyperparameter vector after the second local search; For the hyperparameter vector after the second local search, a reverse global greedy search strategy is used to perform a first global search on the hyperparameter vector to obtain the hyperparameter vector after the first global search; For the hyperparameter vector after the first global search, a random global greedy search strategy is used to perform a second global search on the hyperparameter vector to obtain the hyperparameter vector after the second global search; Determine whether the optimization end condition is met. If so, re-obtain the optimal hyperparameter vector based on the hyperparameter vector after the second global search, and apply the parameters in the optimal hyperparameter vector to the communication performance analysis model to obtain the optimized communication performance analysis model. Otherwise, return to the step of obtaining the loss function value.

6. The communication performance analysis method based on time-space-frequency resources according to claim 5 is characterized in that: For any hyperparameter vector, a first local search is performed on the hyperparameter vector according to the optimal hyperparameter vector and a local personalized search strategy is adopted to obtain a hyperparameter vector after the first local search, including: Based on the current optimization times, an adaptively reduced weight control factor is generated; According to the weight control factor and the optimal hyperparameter vector, the optimal direction search is performed on the hyperparameter vector to obtain the optimal information search item; Obtaining the center vector of all hyperparameter vectors, and performing a comprehensive direction search on the hyperparameter vectors according to the weight control factor and the center vector to obtain a comprehensive information search item; According to the optimal information search item and the comprehensive information search item, a first local search is performed on the hyperparameter vector to obtain the hyperparameter vector after the first local search.

7. The communication performance analysis method based on time-space-frequency resources according to claim 6 is characterized in that: For any hyperparameter vector after the first local search, a second local search is performed on the hyperparameter vector using a nonlinear local search strategy according to the optimal hyperparameter vector to obtain the hyperparameter vector after the second local search, including: For any hyperparameter vector after the first local search, obtaining a cosine information search item between the hyperparameter vector and the optimal hyperparameter vector; A second local search is performed on the hyperparameter vector according to the cosine information search item and the weight control factor to obtain the hyperparameter vector after the second local search.

8. The communication performance analysis method based on time-space-frequency resources according to claim 7 is characterized in that: For the hyperparameter vector after the second local search, a reverse global greedy search strategy is used to perform a first global search on the hyperparameter vector to obtain the hyperparameter vector after the first global search, including: Based on the current optimization times, a nonlinear perturbation factor is obtained, and for the hyperparameter vector after the second local search, the nonlinear perturbation factor is used to perturb the hyperparameter vector to generate a perturbation information vector; Generate a reference information vector according to the nonlinear perturbation factor, and perform a global search on the hyperparameter vector according to the reference information vector and the perturbation information vector to obtain a first global search vector; Determine whether the loss function value of the first global search vector is reduced. If so, use the first global search vector as the hyperparameter vector after the first global search; otherwise, use the original hyperparameter vector after the second local search directly as the hyperparameter vector after the first global search.

9. The communication performance analysis method based on time-space-frequency resources according to claim 8 is characterized in that: For the hyperparameter vector after the first global search, a random global greedy search strategy is used to perform a second global search on the hyperparameter vector to obtain the hyperparameter vector after the second global search, including: For the hyperparameter vector after the first global search, a random flight individual is generated for the hyperparameter vector according to the optimal hyperparameter vector; According to the random flight individuals corresponding to the hyperparameter vector, a cosine function is used to perform a second global search on the hyperparameter vector to obtain a second global search vector; Determine whether the loss function value of the second global search vector is reduced. If so, use the second global search vector as the hyperparameter vector after the second global search. Otherwise, use the original hyperparameter vector after the first global search directly as the hyperparameter vector after the second global search.

10. The communication performance analysis method based on time-space-frequency resources according to claim 5, characterized in that: Determine whether the optimization end conditions are met, including: It is determined whether the current optimization times are greater than or equal to the maximum optimization times. If so, it is determined that the optimization end condition is met; otherwise, it is determined that the optimization end condition is not met.

Images