Intention-driven digital twin modeling object selection method and device

Abstract

The invention provides an intention-driven digital twinning modeling object selection method and device, and relates to the technical field of digital twinning. The method comprises the following steps: acquiring a multi-business demand; defining modeling object parameters according to multi-business requirements; the modeling object parameters comprise attributes of a multi-dimensional QoS target and a modeling object; according to the attributes of the multi-dimensional QoS target and the modeling object, a multi-target Gaussian process regression method is adopted for prediction, and predicted QoS distribution is obtained; the predicted QoS distribution comprises predicted mean values and covariances of the plurality of QoS targets; according to the predicted mean value and covariance of the multiple QoS targets, KL divergence is adopted for evaluation, and the comprehensive importance of each modeling object is obtained; according to the comprehensive importance of each modeling object, a recursive feature elimination method is adopted to adaptively select the modeling objects, and the modeling objects are ranked to select the most relevant modeling object meeting the current task demand. By adopting the method, the construction complexity of the twinborn model can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of digital twins, and particularly to a method and device for selecting an object for intention-driven digital twin modeling. Background Art

[0002] With the rapid development of industrial intelligence and Internet of Things technologies, the modeling method based on multi-granularity twin models has gradually become an important means to optimize production efficiency, predict equipment failures, and improve resource allocation. The multi-granularity twin model simulates physical entities and processes at multiple levels and dimensions to achieve comprehensive perception and real-time monitoring of the production system, providing accurate support for decision-makers. However, when constructing a complex multi-granularity twin model, how to select appropriate modeling objects, especially in the face of massive data and a dynamically changing environment, has become a key issue in improving the efficiency and accuracy of the model. At present, many existing modeling methods often rely on static rules or means based on simple data analysis, and it is difficult to flexibly adapt to different business requirements and environmental changes. Traditional methods for selecting modeling objects usually lack a deep understanding of business goals and actual needs, resulting in the inability to achieve efficient and accurate object selection in a dynamic environment. This lack of targeted modeling object selection method not only increases the time and computational cost of model construction, but may also affect the performance and stability of the system.

[0003] In recent years, data-driven modeling methods have emerged, which use deep learning and machine learning technologies to dynamically select key objects, dynamically adjust data mapping by updating the model in real time and intelligently selecting data most relevant to user needs, thereby reducing data volume and resource consumption. However, the above methods still face the problems of over-reliance on data and inability to adapt to environmental changes in real time. The problems existing in the current prior art include: (1) The existing technologies cannot effectively combine business goals and system requirements to flexibly and dynamically select modeling objects, and it is difficult to ensure the effectiveness and reliability of the model in a changing environment; (2) The existing unified granularity model construction method may cause resource waste and even hinder the timely response to changes in real-time task requirements and resource status; (3) The construction complexity of the twin model increases exponentially. Affected by factors such as the required data volume, inherent data complexity, and its pattern stability, there are significant differences in the modeling costs of different modeling objects, resulting in a large amount of computational and storage resources being required during the model construction process; (4) There are differences in the relevance of different modeling objects to the current business requirements, resulting in different contributions of each object. The prior art often ignores this difference and fails to quantitatively evaluate each object, resulting in some objects being over-modeled and wasting computational resources. Summary of the Invention

[0004] To solve the technical problems existing in the prior art, such as the waste of resources caused by the existing unified granularity model construction method, the obstruction of the timely response to the changes in the real-time task requirements and resource status, the large differences in the modeling costs of different modeling objects, resulting in the need to invest a large amount of computing and storage resources during the model construction process, and the differences in the relevance between different modeling objects and the current business requirements, resulting in different contributions of each object and some objects being over-modeled, wasting computing resources, the embodiments of the present invention provide an intention-driven digital twin modeling object selection method and device. The technical solutions are as follows:

[0005] On the one hand, an intention-driven digital twin modeling object selection method is provided. This method is implemented by an intention-driven digital twin modeling object selection device, and the method includes:

[0006] S1. Obtain diversified business requirements; define modeling object parameters according to the diversified business requirements; the modeling object parameters include: multi-dimensional QoS objectives and the attributes of the modeling object;

[0007] S2. According to the multi-dimensional QoS objectives and the attributes of the modeling object, use the multi-objective Gaussian process regression method for prediction to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted means and covariances of multiple QoS objectives;

[0008] S3. According to the predicted means and covariances of multiple QoS objectives, use the KL divergence for evaluation to obtain the comprehensive importance of each modeling object;

[0009] S4. According to the comprehensive importance of each modeling object, use the recursive feature elimination method to adaptively select modeling objects, and by ranking the modeling objects, select the modeling objects most relevant to meeting the current task requirements.

[0010] Optionally, the step S2 of using the multi-objective Gaussian process regression method for prediction according to the multi-dimensional QoS objectives and the attributes of the modeling object to obtain the predicted QoS distribution includes:

[0011] S21. Obtain training data for the multi-objective Gaussian process regression; wherein, the training data includes training input data and training output data;

[0012] S22. Set the joint distribution of all target QoS objectives to satisfy the multivariate Gaussian distribution for the given input;

[0013] S23. According to the set multivariate Gaussian distribution, obtain the joint distribution of the training output data and any new predicted input data;

[0014] S24. Use the radial basis function kernel to define the similarity between the training input data to obtain the kernel function;

[0015] S25. Calculate the kernel matrix of the training input data according to the kernel function; optimize the hyperparameters of the Gaussian process by using the maximum log marginal likelihood optimization method according to the kernel matrix of the training input data, and obtain the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution. Process to optimize the hyperparameters of the Gaussian process by using the maximum log marginal likelihood optimization method is optionally represented by the following formula (1):

[0016] Optionally, the process of optimizing the hyperparameters of the Gaussian process by using the maximum log marginal likelihood optimization method is represented by the following formula (1):

[0017] (1)

[0018] Where, represents the kernel matrix; represents the noise variance; M represents the dimension of the Qos target; I represents the identity matrix; Y represents the observed target; represents the number of samples; represents the conditional probability distribution.

[0019] Optionally, the process of obtaining the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution is represented by the following formula (2):

[0020] (2)

[0021] Where,

[0022]

[0023] Where, represents the predicted mean of QoS targets, represents the predicted covariance matrix, capturing the uncertainty and correlation between targets; represents the kernel matrix of the predicted input data; represents the kernel matrix of the training input data; represents the cross-kernel matrix between the training input and the predicted input; represents the predicted average of M targets;

[0024] Optionally, the process of obtaining the comprehensive importance of each modeling object is represented by the following formula (3):

[0025] (3)

[0026] Where, represents the mean vector of the predicted QoS distribution; represents the covariance matrix of the predicted QoS distribution; represents the target QoS distribution, represents the dimension of the QoS target, represents the determinant of the matrix, represents the trace of the matrix; represents the modeling object; represents the predicted mean; represents the comprehensive importance of each modeling object.

[0027] Optionally, the S4 adaptively selects modeling objects by using the recursive feature elimination method. By ranking the modeling objects, the most relevant modeling objects that meet the current task requirements are selected, including:

[0028] Rank the comprehensive importance of each modeling object in descending order. Use the recursive feature elimination method. By iterative means, eliminate the modeling objects with the lowest comprehensive importance to obtain the remaining modeling objects; recalculate the comprehensive importance of the remaining modeling objects until the number of remaining modeling objects is equal to the predefined target number, and output the most relevant modeling objects.

[0029] On the other hand, an intent-driven digital twin modeling object selection device is provided. This device is applied to the intent-driven digital twin modeling object selection method. The device includes:

[0030] An acquisition unit for acquiring multi-service requirements; defining modeling object parameters according to the multi-service requirements; the modeling object parameters include: multi-dimensional QoS targets and the attributes of the modeling objects;

[0031] A prediction unit for predicting using the multi-objective Gaussian process regression method according to the multi-dimensional QoS targets and the attributes of the modeling objects to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted means and covariances of multiple QoS targets;

[0032] An evaluation unit for evaluating using the KL divergence according to the predicted means and covariances of multiple QoS targets to obtain the comprehensive importance of each modeling object;

[0033] A selection unit for adaptively selecting modeling objects by using the recursive feature elimination method according to the comprehensive importance of each modeling object. By ranking the modeling objects, the most relevant modeling objects that meet the current task requirements are selected.

[0034] Optionally, the prediction unit is used for:

[0035] Obtain training data for multi-objective Gaussian process regression; wherein, the training data includes training input data and training output data;

[0036] Set the joint distribution of all target QoS targets to satisfy a multivariate Gaussian distribution for a given input;

[0037] Obtain the joint distribution of the trained output data and any new predicted input data according to the set multivariate Gaussian distribution;

[0038] Define the similarity between the trained input data using a radial basis function kernel to obtain a kernel function;

[0039] According to the kernel function, calculate the kernel matrix of the training input data; according to the kernel matrix of the training input data, use the maximum log marginal likelihood optimization method to optimize the hyperparameters of the Gaussian process, and obtain the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution.

[0040] Optionally, the process of optimizing the hyperparameters of the Gaussian process using the maximum log marginal likelihood optimization method is represented by the following formula (1):

[0041] (1)

[0042] where, represents the kernel matrix; represents the noise variance; M represents the dimension of the Qos target; I represents the identity matrix; Y represents the observed target; represents the number of samples; represents the conditional probability distribution.

[0043] Optionally, the process of obtaining the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution is represented by the following formula (2):

[0044] (2)

[0045] where,

[0046]

[0047] where, represents the predicted mean of QoS targets, represents the predicted covariance matrix, capturing the uncertainty and correlation between targets; represents the kernel matrix of the predicted input data; represents the kernel matrix of the training input data; represents the cross-kernel matrix between the training input and the predicted input; represents the predicted average value of M targets;

[0048] Optionally, the process of obtaining the comprehensive importance of each modeling object is represented by the following formula (3):

[0049] (3)

[0050] where represents the mean vector of the predicted QoS distribution; represents the covariance matrix of the predicted QoS distribution; represents the target QoS distribution, represents the dimension of the QoS target, represents the determinant of the matrix, represents the trace of the matrix; represents the modeling object; represents the predicted mean; represents the comprehensive importance of each modeling object.

[0051] Optionally, the selection unit is configured to:

[0052] Rank the comprehensive importance of each modeling object in descending order, adopt the recursive feature elimination method, and iteratively exclude the modeling object with the smallest comprehensive importance to obtain the remaining modeling objects; recalculate the comprehensive importance of the remaining modeling objects until the number of the remaining modeling objects is equal to the predefined target number, and output the most relevant modeling objects.

[0053] On the other hand, an intention-driven digital twin modeling object selection device is provided. The intention-driven digital twin modeling object selection device includes: a processor; a memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, any one of the methods in the above-mentioned intention-driven digital twin modeling object selection method is implemented.

[0054] On the other hand, a computer-readable storage medium is provided. At least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement any one of the methods in the above-mentioned intention-driven digital twin modeling object selection method.

[0055] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include:

[0056] In the embodiments of the present invention, firstly, multi - service requirements are obtained; according to the multi - service requirements, modeling object parameters are defined; the modeling object parameters include: multi - dimensional QoS objectives and the attributes of the modeling object; secondly, according to the multi - dimensional QoS objectives and the attributes of the modeling object, a multi - objective Gaussian process regression method is used for prediction to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted means and covariances of multiple QoS objectives; according to the predicted means and covariances of multiple QoS objectives, KL divergence is used for evaluation to obtain the comprehensive importance of each modeling object; finally, according to the comprehensive importance of each modeling object, a recursive feature elimination method is used to adaptively select modeling objects, and by ranking the modeling objects, the most relevant modeling objects that meet the current task requirements are selected.

[0057] This application significantly improves the response speed and resource utilization rate of the digital twin system by optimizing the modeling object selection method. By accurately selecting key modeling objects, it avoids the resource waste caused by modeling all objects in traditional methods, enabling the system to quickly respond to changes in real - time task requirements and improving the overall efficiency. Compared with traditional methods, this application simplifies the modeling methods of various computing elements in heterogeneous networks, reduces the complexity of constructing the twin model, and reduces the computational burden caused by modeling too many objects. Even in a large - scale dynamic heterogeneous network environment, it still maintains a low computational overhead and good modeling effects. In addition, by dynamically selecting modeling objects highly relevant to business requirements, it effectively reduces the modeling cost of low - value elements, reduces unnecessary investments in large - scale modeling, and thus improves the modeling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0059] Figure 1 It is a schematic structural diagram of an intention - driven digital twin modeling object selection method provided by the embodiments of the present invention;

[0060] Figure 2 It is a flowchart of an intention - driven digital twin modeling object selection method provided by the embodiments of the present invention;

[0061] Figure 3 It is a block diagram of an intention - driven digital twin modeling object selection device provided by the embodiments of the present invention;

[0062] Figure 4 It is a schematic structural diagram of an intention - driven digital twin modeling object selection device provided by the embodiments of the present invention. Detailed implementation manners

[0063] The technical solutions in the present invention will be described below with reference to the accompanying drawings.

[0064] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or more advantageous than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0065] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0066] In the embodiments of the present invention, sometimes subscripts such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0067] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0068] The embodiments of the present invention provide an intention-driven digital twin modeling object selection method. This method can be implemented by an intention-driven digital twin modeling object selection device, and the intention-driven digital twin modeling object selection device can be a terminal or a server. As Figure 1 shown, it is a structural schematic diagram of an intention-driven digital twin modeling object selection method provided by the embodiments of the present invention and is implemented by the steps as Figure 2 shown. Among them, as Figure 1 shown is the flowchart of the intention-driven digital twin modeling object selection method. The processing flow of this method can include the following steps:

[0069] S1. Obtain diversified service requirements; define modeling object parameters according to the diversified service requirements; the modeling object parameters include: multi-dimensional QoS objectives and attributes of the modeling object.

[0070] Among them, the diversified service requirements include: QoS requirements, energy efficiency, computing requirements, large storage task requirements, network topology and resource scheduling.

[0071] In a feasible implementation manner, according to the multi-service requirements, it is assumed that the multi-dimensional QoS target set is defined as , where each target represents different metrics, including latency, energy consumption, or throughput; among them, the set of modeling objects is represented as , and each modeling object has an attribute vector representing the modeling characteristics; among them, the attribute vector is represented as , represents the th attribute of the modeling object . Among them, the attribute space of the attribute vector is represented by the following formula (1):

[0072] (1)

[0073] Among them, the attribute vector of each modeling object can be extended and represented by the following formula (2):

[0074] (2)

[0075] Among them, represents the total number of unique attributes in the attribute space .

[0076] S2. According to the multi-dimensional QoS target and the attributes of the modeling object, use the multi-objective Gaussian process regression method for prediction to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted mean and covariance of multiple QoS targets.

[0077] Among them, use the multi-objective Gaussian process regression method to model the QoS distribution, and estimate the service quality and its uncertainty by predicting the mean and covariance of each QoS metric.

[0078] Optionally, the specific implementation process of S2 includes S21 - S25:

[0079] S21. Obtain the training data for multi-objective Gaussian process regression; among them, the training data includes the training input data and the training output data;

[0080] Among them, the training data consists of multiple samples; among them, the input data and the corresponding output can be represented by the following formula (3):

[0081] (3)

[0082] Among them, observations of QoS targets for

[0083] S22. Set the joint distribution of all target QoS targets to satisfy a multivariate Gaussian distribution for a given input;

[0084] In a feasible implementation, the multivariate Gaussian distribution is represented by the following formula (4):

[0085] (4)

[0086] where, represents the predicted mean of all targets; represents the covariance matrix between targets; represents the value of the multivariate Gaussian distribution function.

[0087] S23. Obtain the joint distribution of the trained output data and any new predicted input data according to the set multivariate Gaussian distribution;

[0088] where, the joint distribution of the trained output data and any new predicted input data is represented by the following formula

[0089] (5)

[0090] (5)

[0091] where, represents the kernel matrix of the training input, represents the cross-kernel matrix between the training input and the predicted input, represents the kernel matrix of the predicted input, represents the noise variance; represents the cross-kernel matrix between the predicted input and the training input; represents the predicted output.

[0092] S24. Define the similarity between the training input data using a radial basis function kernel to obtain a kernel function;

[0093] where, the kernel function is represented by the following formula (6):

[0094] (6)

[0095] where, represents the input vector and the input vector the squared Euclidean distance between them, is the length scale parameter that controls the smoothness of the function.

[0096] where, the kernel matrix of the training input is obtained by calculating the kernel function for all inputs in the dataset The specific calculation process is represented by the following formula (7):

[0097] (7)

[0098] Among them, represents the similarity measure between the $i$-th and $j$-th input points in the kernel matrix.

[0099] S25. Calculate the kernel matrix of the training input data according to the kernel function; according to the kernel matrix of the training input data, use the maximum log marginal likelihood optimization method to optimize the hyperparameters of the Gaussian process, and obtain the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution.

[0100] Optionally, the process of optimizing the hyperparameters of the Gaussian process by using the maximum log marginal likelihood optimization method is represented by the following formula (8):

[0101] (8)

[0102] Among them, represents the kernel matrix; represents the noise variance; $M$ represents the dimension of the Qos target; $I$ represents the identity matrix; $Y$ represents the observed target; represents the number of samples; represents the conditional probability distribution.

[0103] Optionally, the process of obtaining the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution is represented by the following formula (9):

[0104] (9)

[0105] Among them,

[0106]

[0107] Among them, represents the predicted mean of $M$ QoS targets, represents the predicted covariance matrix, capturing the uncertainty and correlation between the targets; represents the kernel matrix of the predicted input data; represents the kernel matrix of the training input data; represents the cross-kernel matrix between the training input and the predicted input; represents the predicted average of $M$ targets; represents the cross-kernel matrix between the predicted input and the training input.

[0108] S3. Evaluate according to the predicted means and covariances of multiple QoS targets using the KL divergence, and obtain the comprehensive importance of each modeling object.

[0109] Optionally, the process of obtaining the comprehensive importance of each modeling object is represented by the following formula (10):

[0110] (10)

[0111] where represents the mean vector of the predicted QoS distribution; represents the covariance matrix of the predicted QoS distribution; represents the target QoS distribution, represents the dimension of the QoS target, represents the determinant of the matrix, represents the trace of the matrix; represents the modeling object; represents the predicted average value; represents the comprehensive importance of each modeling object.

[0112] S4. According to the comprehensive importance of each modeling object, adopt the recursive feature elimination method to adaptively select modeling objects. By ranking the modeling objects, select the most relevant modeling objects that meet the current task requirements.

[0113] Optionally, the step S4 of adaptively selecting modeling objects by using the recursive feature elimination method includes:

[0114] Rank the comprehensive importance of each modeling object in descending order. Adopt the recursive feature elimination method. By means of iteration, exclude the modeling objects with the smallest comprehensive importance to obtain the remaining modeling objects; recalculate the comprehensive importance of the remaining modeling objects until the number of the remaining modeling objects is equal to the predefined target number, and select the most relevant modeling objects that meet the current task requirements.

[0115] where the remaining modeling objects are represented as , which consists of the most relevant modeling objects that have a significant impact on the QoS target optimization. In a feasible implementation manner, after obtaining the most relevant modeling objects, evaluate the connection relationships among the most relevant modeling objects, and identify the preferences that need to be optimized by analyzing the network diagram, including: minimizing throughput, latency, or energy consumption. Assume that the topological structure of the original network is

[0116] where represents the set of nodes, represents the set of edges, represents that there is a connection between node and node and node ; It means there is no connection. For each link , its contribution depends on the transmission capacity and bandwidth. In a feasible implementation, the addition process of the most relevant modeling objects is a recursive execution process, and the optimal modeling object and link set are directly selected. The specific implementation process includes:

[0117] (1) Input the set of selected key nodes , the original network , select two nodes and the performance threshold of the link between them and the maximum number of modeling links for each node; initialize; initialize the nodes to be detected in each recursion as ; initialize the currently selected set of key links as an empty set, the candidate link set is initialized as , and the access node set is initialized as ;

[0118] (2) In each recursion, calculate for each node in the current node set, including: traverse all its neighbor nodes, if the contribution is greater than or equal to the threshold, and add the link contribution between the node and its neighbor nodes to the list; among them, for the links in the list, if the number of links exceeds the maximum allowed number of connections, sort them in descending order of contribution and only keep the first links, and discard the others; if no link meets the required threshold, that is, the list is empty, add the link with the largest contribution among all the edges of the node to the list; add the link and its corresponding neighbor node from the list to the selected link set and the selected node set; if the neighbor node has not been visited before, add it to the current detection node set and the access node set at the same time. After processing each link, remove the current node from the node set. This process will be recursively executed until all nodes are processed;

[0119] (3) Output the modeled key area, denoted as , and this area is the original network . Among them, as shown in Table 1 is the pseudocode of the algorithm for the intent-driven key modeling object selection method.

[0120] Table 1

[0121]

[0122] In a feasible implementation, by combining multi-objective Gaussian process regression and recursive feature elimination techniques, the process of dynamically optimizing the selection of modeling objects can meet different business requirements while reducing waste of system resources. A QoS evaluation mechanism based on KL divergence is introduced to dynamically measure the contribution of modeling objects to multi-dimensional service quality. This mechanism can intelligently identify the modeling objects most relevant to the target, avoid interference from low-value attributes, and improve the accuracy and real-time response ability of the twin model. This application breaks through the limitations of traditional unified granularity model methods, can adapt to complex and changing industrial production environments, and flexibly adjusts the fine-grained and granularity selection of the twin model according to changes in system performance requirements and business needs, thereby improving the overall system efficiency and modeling efficiency.

[0123] In the embodiment of the present invention, first, diverse business requirements are obtained; according to the diverse business requirements, modeling object parameters are defined; the modeling object parameters include: multi-dimensional QoS objectives and attributes of the modeling object; secondly, according to the multi-dimensional

[0124] QoS objectives and attributes of the modeling object, a multi-objective Gaussian process regression method is used for prediction to obtain a predicted QoS distribution; the predicted QoS distribution includes predicted means and covariances of multiple QoS objectives; according to the predicted means and covariances of multiple QoS objectives, KL divergence is used for evaluation to obtain the comprehensive importance of each modeling object; finally, according to the comprehensive importance of each modeling object, a recursive feature elimination method is used to adaptively select modeling objects, and by ranking the modeling objects, the modeling objects most relevant to meeting the current task requirements are selected.

[0125] This application significantly improves the response speed and resource utilization rate of the digital twin system by optimizing the modeling object selection method. By accurately selecting key modeling objects, it avoids the resource waste caused by modeling all objects in traditional methods, enabling the system to quickly respond to changes in real-time task requirements and improving overall efficiency. Compared with traditional methods, this application simplifies the modeling method of various computing elements in heterogeneous networks, reduces the complexity of constructing the twin model, and reduces the computational burden caused by modeling too many objects. Even in a large-scale dynamic heterogeneous network environment, it still maintains low computational overhead and good modeling effects. In addition, by dynamically selecting modeling objects highly relevant to business requirements, it effectively reduces the modeling cost of low-value elements, reduces unnecessary investment in large-scale modeling, and thus improves modeling efficiency.

[0126] Figure 3 is a block diagram of an intent-driven digital twin modeling object selection device shown according to an exemplary embodiment. This device is used for the intent-driven digital twin modeling object selection method. Referring to Figure 3 , this device includes an acquisition unit 310, a prediction unit 320, an evaluation unit 330, and a selection unit 340. Among them:

[0127] An acquisition unit 310, configured to acquire multi-service requirements; define modeling object parameters according to the multi-service requirements; the modeling object parameters include: multi-dimensional QoS objectives and attributes of the modeling object;

[0128] A prediction unit 320, configured to perform prediction by using a multi-objective Gaussian process regression method according to the multi-dimensional QoS objectives and the attributes of the modeling object, so as to obtain a predicted QoS distribution; the predicted QoS distribution includes predicted means and covariances of multiple QoS objectives;

[0129] An evaluation unit 330, configured to perform evaluation by using KL divergence according to the predicted means and covariances of multiple QoS objectives, so as to obtain the comprehensive importance of each modeling object;

[0130] A selection unit 340, configured to adaptively select modeling objects by using a recursive feature elimination method according to the comprehensive importance of each modeling object, and select the most relevant modeling objects that meet the current task requirements by ranking the modeling objects.

[0131] Optionally, the prediction unit 320 is configured to:

[0132] Acquire training data for multi-objective Gaussian process regression; wherein, the training data includes training input data and training output data;

[0133] Set that the joint distribution of all target QoS objectives satisfies a multi-variate Gaussian distribution for a given input;

[0134] According to the set multi-variate Gaussian distribution, obtain the joint distribution of the training output data and any new predicted input data;

[0135] Define the similarity between the training input data by using a radial basis function kernel to obtain a kernel function;

[0136] According to the kernel function, calculate the kernel matrix of the training input data; according to the kernel matrix of the training input data, use a maximum log-likelihood optimization method to optimize the hyperparameters of the Gaussian process, and obtain the predicted QoS distribution through conditional probability calculation of the multi-variate Gaussian distribution.

[0137] Optionally, the process of using the maximum log-likelihood optimization method to optimize the hyperparameters of the Gaussian process is represented by the following formula (1):

[0138] (1)

[0139] Wherein, represents the kernel matrix; denotes the noise variance; M denotes the dimension of the QoS objective; I denotes the identity matrix; Y denotes the observed objective; denotes the number of samples; denotes the conditional probability distribution.

[0140] Optionally, the process of obtaining the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution is represented by the following formula (2):

[0141] (2)

[0142] where,

[0143]

[0144] where, denotes the predicted mean of the QoS objectives, denotes the predicted covariance matrix, capturing the uncertainty and correlation between the objectives; denotes the kernel matrix of the predicted input data; denotes the kernel matrix of the training input data; denotes the cross-kernel matrix between the training input and the predicted input; denotes the predicted average value of the M objectives; denotes the cross-kernel matrix between the predicted input and the training input.

[0145] Optionally, the process of obtaining the comprehensive importance of each modeling object is represented by the following formula (3):

[0146] (3)

[0147] where, denotes the mean vector of the predicted QoS distribution; denotes the covariance matrix of the predicted QoS distribution; denotes the target QoS distribution, denotes the dimension of the QoS objective, denotes the determinant of the matrix, denotes the trace of the matrix; denotes the modeling object; denotes the predicted mean; denotes the comprehensive importance of each modeling object.

[0148] Optionally, the selection unit 340 is used for:

[0149] Rank the comprehensive importance of each modeling object in descending order. Use the recursive feature elimination method to iteratively exclude the modeling object with the lowest comprehensive importance and obtain the remaining modeling objects. Recalculate the comprehensive importance of the remaining modeling objects until the number of remaining modeling objects equals the predefined target number, and output the most relevant modeling objects.

[0150] In the embodiment of the present invention, first, obtain the multi-service requirements; according to the multi-service requirements, define the modeling object parameters; the modeling object parameters include: multi-dimensional QoS objectives and the attributes of the modeling object; secondly, according to the multi-dimensional QoS objectives and the attributes of the modeling object, use the multi-objective Gaussian process regression method for prediction to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted means and covariances of multiple QoS objectives; according to the predicted means and covariances of multiple QoS objectives, use the KL divergence for evaluation to obtain the comprehensive importance of each modeling object; finally, according to the comprehensive importance of each modeling object, use the recursive feature elimination method to adaptively select the modeling objects, and by ranking the modeling objects, select the most relevant modeling objects that meet the current task requirements.

[0151] This application significantly improves the response speed and resource utilization rate of the digital twin system by optimizing the modeling object selection method. By accurately selecting key modeling objects, it avoids the resource waste caused by modeling all objects in the traditional method, enabling the system to quickly respond to changes in real-time task requirements and improving the overall efficiency. Compared with the traditional method, this application simplifies the modeling method of various computing elements in the heterogeneous network, reduces the complexity of building the twin model, and reduces the computational burden caused by modeling too many objects. Even in a large-scale dynamic heterogeneous network environment, it still maintains a low computational overhead and good modeling effect. In addition, by dynamically selecting modeling objects highly relevant to the business requirements, it effectively reduces the modeling cost of low-value elements, reduces unnecessary investment in large-scale modeling, and thus improves the modeling efficiency.

[0152] Figure 4 is a schematic structural diagram of an intention-driven digital twin modeling object selection device provided by an embodiment of the present invention, as Figure 4 shown, the intention-driven digital twin modeling object selection device may include the above-mentioned Figure 3 shown intention-driven digital twin modeling object selection device. Optionally, the intention-driven digital twin modeling object selection device 410 may include a first processor 2001.

[0153] Optionally, the intention-driven digital twin modeling object selection device 410 may further include a memory 2002 and a transceiver 2003.

[0154] Among them, the first processor 2001, the memory 2002, and the transceiver 2003 can be connected through a communication bus, for example.

[0155] Next, in combination with Figure 4 each component of the intent-driven digital twin modeling object selection device 410 will be specifically introduced:

[0156] Among them, the first processor 2001 is the control center of the intent-driven digital twin modeling object selection device 410, which can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or can be an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. For example: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0157] Optionally, the first processor 2001 can execute various functions of the intent-driven digital twin modeling object selection device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0158] In a specific implementation, as an embodiment, the first processor 2001 can include one or more CPUs, such as Figure 4 CPU0 and CPU1 shown in

[0159] In a specific implementation, as an embodiment, the intent-driven digital twin modeling object selection device 410 can also include multiple processors, such as Figure 4 the first processor 2001 and the second processor 2004 shown in

[0160] Among them, the memory 2002 is used to store software programs for implementing the solution of the present invention and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiments and will not be elaborated here.

[0161] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and is coupled to the first processor 2001 through the interface circuit of the intent-driven digital twin modeling object selection device 410 ( Figure 4 not shown in the figure), and the embodiments of the present invention do not make specific limitations on this.

[0162] The transceiver 2003 is used to communicate with a network device or communicate with a terminal device.

[0163] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 4 not shown separately in the figure). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0164] Optionally, the transceiver 2003 may be integrated with the first processor 2001 or may exist independently and is coupled to the first processor 2001 through the interface circuit of the intent-driven digital twin modeling object selection device 410 ( Figure 4 not shown in the figure), and the embodiments of the present invention do not make specific limitations on this.

[0165] It should be noted that Figure 4 the structure of the intent-driven digital twin modeling object selection device 410 shown in the figure does not constitute a limitation on the router. The actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0166] In addition, the technical effects of the intent-driven digital twin modeling object selection device 410 may refer to the technical effects of the intent-driven digital twin modeling object selection method described in the above method embodiments, and will not be elaborated here.

[0167] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0168] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0169] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0170] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context.

[0171] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0172] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0173] Those of ordinary skill in the art will realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0174] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0175] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the devices or units can be in electrical, mechanical, or other forms.

[0176] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0177] In addition, the functional units in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0178] When the above-mentioned function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0179] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An intention-driven digital twin modeling object selection method, characterized in that The method includes: S1. Obtain multi-service requirements; define modeling object parameters according to the multi-service requirements; the modeling object parameters include: multi-dimensional QoS objectives and attributes of the modeling object; S2. Perform prediction using the multi-objective Gaussian process regression method according to the multi-dimensional QoS objectives and attributes of the modeling object to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted means and covariances of multiple QoS objectives; S3. Evaluate using the KL divergence according to the predicted means and covariances of multiple QoS objectives to obtain the comprehensive importance of each modeling object; S4. Adaptively select modeling objects using the recursive feature elimination method according to the comprehensive importance of each modeling object, and select the most relevant modeling objects that meet the current task requirements by ranking the modeling objects.

2. The intention-driven digital twin modeling object selection method according to claim 1, characterized in that The step S2 of performing prediction using the multi-objective Gaussian process regression method according to the multi-dimensional QoS objectives and attributes of the modeling object to obtain the predicted QoS distribution includes: S21. Obtain training data for the multi-objective Gaussian process regression; wherein, the training data includes training input data and training output data; S22. Set the joint distribution of all target QoS objectives to satisfy the multi-variate Gaussian distribution for the given input; S23. Obtain the joint distribution of the training output data and any new predicted input data according to the set multi-variate Gaussian distribution; S24. Define the similarity between the training input data using the radial basis function kernel to obtain the kernel function; S25. Calculate the kernel matrix of the training input data according to the kernel function; optimize the hyperparameters of the Gaussian process by using the maximum log marginal likelihood optimization method according to the kernel matrix of the training input data, and obtain the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution. ​ 3. The intention-driven digital twin modeling object selection method according to claim 2, characterized in that The process of optimizing the hyperparameters of the Gaussian process using the maximum log marginal likelihood optimization method is represented by the following formula (1): （1） Among them, represents the kernel matrix; represents the noise variance; M represents the dimension of the Qos objective; I represents the identity matrix; Y represents the observed objective; represents the number of samples; represents the conditional probability distribution.

4. The intention-driven digital twin modeling object selection method according to claim 2, wherein The process of obtaining the predicted QoS distribution by calculating the conditional probability of the multi-variate Gaussian distribution is represented by the following formula (2): （2） Among them, ； wherein, represents the predicted mean of QoS targets, represents the predicted covariance matrix, capturing the uncertainty and correlation between targets; represents the kernel matrix of the predicted input data; represents the kernel matrix of the training input data; represents the cross-kernel matrix between the training input and the predicted input; represents the predicted average of M targets; represents the cross-kernel matrix between the predicted input and the training input.

5. The method for selecting an intention-driven digital twin modeling object according to claim 1, wherein The process of obtaining the comprehensive importance of each modeling object is represented by the following formula (3): （3） Among them, represents the mean vector of the predicted QoS distribution; represents the covariance matrix of the predicted QoS distribution; represents the target QoS distribution, represents the dimension of the QoS target, represents the determinant of the matrix, represents the trace of the matrix; represents the modeling object; represents the predicted mean; represents the comprehensive importance of each modeling object.

6. The intention-driven digital twin modeling object selection method according to claim 1, characterized in that, The step S4 of adaptively selecting modeling objects using the recursive feature elimination method and selecting the most relevant modeling objects that meet the current task requirements by ranking the modeling objects includes: Rank the comprehensive importance of each modeling object in descending order, use the recursive feature elimination method, and iteratively exclude the modeling objects with the minimum comprehensive importance to obtain the remaining modeling objects; recalculate the comprehensive importance of the remaining modeling objects until the number of remaining modeling objects is equal to the predefined target number, and output the most relevant modeling objects.

7. An intent-driven digital twin modeling object selection device, which is used to implement the intent-driven digital twin modeling object selection method according to any one of claims 1-6, characterized in that, The device includes: An acquisition unit, configured to obtain multi-service requirements; define modeling object parameters according to the multi-service requirements; the modeling object parameters include: multi-dimensional QoS objectives and attributes of the modeling object; A prediction unit, configured to perform prediction using the multi-objective Gaussian process regression method according to the multi-dimensional QoS objectives and attributes of the modeling object to obtain the predicted QoS distribution; the predicted QoS distribution includes the predicted means and covariances of multiple QoS objectives; An evaluation unit, configured to evaluate using the KL divergence according to the predicted means and covariances of multiple QoS objectives to obtain the comprehensive importance of each modeling object; A selection unit, which is configured to adaptively select modeling objects according to the comprehensive importance of each modeling object by using a recursive feature elimination method, and select the most relevant modeling objects that meet the requirements of the current task by ranking the modeling objects.

8. The intent-driven digital twin modeling object selection device according to claim 7, wherein The prediction unit is configured to: Obtain training data for multi-objective Gaussian process regression; wherein the training data includes training input data and training output data; Set that the joint distribution of all target QoS objectives satisfies a multivariate Gaussian distribution for a given input; Obtain the joint distribution of the training output data and any new prediction input data according to the set multivariate Gaussian distribution; Define the similarity between the training input data by using a radial basis function kernel to obtain a kernel function; Calculate the kernel matrix of the training input data according to the kernel function; optimize the hyperparameters of the Gaussian process by using the maximum log marginal likelihood optimization method, and obtain the predicted QoS distribution through the conditional probability calculation of the multivariate Gaussian distribution.

9. An intention-driven digital twin modeling object selection device, characterized in that, The intention-driven digital twin modeling object selection device includes: A processor; A memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that, Program code is stored in the computer-readable storage medium, and the program code can be called by the processor to execute the method described in any one of claims 1 to 6.

Images