Network simulation method, device and equipment based on signal category, medium and product

Abstract

The invention discloses a network simulation method and device based on signal categories, equipment, a medium and a product. The method comprises: for each communication signal, inputting a fusion feature to a signal identification model to determine the category of the communication signal, the fusion feature being determined according to a time sequence feature and a spatial feature of the communication signal; constructing a model of the communication network based on a graph representation method, wherein each node in a graph represents network equipment in the communication network; for each category of the communication signals, optimizing the network topology of the model of the communication network with the goals of meeting the network performance requirements and minimizing the cost; the propagation performance of each category of communication signals in the communication network is predicted and optimized. According to the method, the category of the communication signals is determined by using the fusion features, in the constructed network model, network topology and propagation performance are optimized according to the category, the accuracy and robustness of network simulation are improved, corresponding optimization schemes are provided for various communication signals, and the communication performance of various communication signals is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of communication network technology, and in particular, to a signal category-based network simulation method, device, equipment, medium and product. Background Art

[0002] With the rapid development and widespread application of communication technology, higher requirements are placed on the deployment, optimization and maintenance of communication networks. In this context, communication network simulation technology has become an important research field. Simulation technology can help researchers and engineers predict network performance and optimize network configuration before actual deployment, while providing a practical platform for education and training. For the simulation of communication networks, modules such as network planning, equipment configuration, signal transmission and performance evaluation are usually included. These modules can simulate the basic operations of 5G networks, including the transmission of data packets, network coverage and access of user equipment. Its process conditions depend on high-performance computing platforms and accurate network models. These models need to be configured according to actual network parameters, including the layout of base stations, the use of frequencies and the types of communication signals.

[0003] At present, the recognition of communication signals is usually based on the statistical characteristics of the signal, such as power spectral density or autocorrelation function. These methods require manual setting of recognition parameters, preprocessing and feature extraction of signals. When processing signals in complex or noisy environments, complex algorithms and a large amount of computing resources are required, and the accuracy is also limited. In general, the existing communication signal recognition methods have the following shortcomings: the level of automation is limited, and manual intervention is usually required to set parameters and make adjustments, which limits the flexibility and response speed of the system, and the accuracy and robustness of recognition need to be improved: in the face of complex communication environments, such as different types of communication signals, and the presence of multipath interference and signal attenuation, the accuracy and robustness of existing recognition methods are poor, making it difficult to meet the needs of high-precision recognition of communication signals, and unable to provide applicable optimization solutions for different types of communication signals. Summary of the invention

[0004] The present application provides a network simulation method, apparatus, device, medium and product based on signal category to achieve accurate identification of communication signals and improve the accuracy and robustness of network simulation optimization for various types of communication signals.

[0005] In a first aspect, an embodiment of the present application provides a network simulation method based on signal categories, comprising:

[0006] For each communication signal, inputting the fusion feature of the communication signal into a signal recognition model to determine the category of the communication signal through the signal recognition model, wherein the fusion feature is determined according to the temporal feature and the spatial feature of the communication signal;

[0007] A communication network model is constructed based on a graph representation, where each node in the graph represents a network device in the communication network;

[0008] For each category of communication signals, optimizing the network topology of the model of the communication network with the goal of meeting network performance requirements and minimizing costs;

[0009] The propagation performance of each category of communication signals in the communication network is predicted and optimized.

[0010] In a second aspect, an embodiment of the present application further provides a network simulation device based on signal categories, including:

[0011] an identification module, configured to input, for each communication signal, a fusion feature of the communication signal into a signal identification model, so as to determine the category of the communication signal through the signal identification model, wherein the fusion feature is determined according to a temporal feature and a spatial feature of the communication signal;

[0012] A building module, for building a model of a communication network based on a graph representation, where each node in the graph represents a network device in the communication network;

[0013] A topology optimization module, for optimizing the network topology of the model of the communication network for each category of the communication signal with the goal of meeting the network performance requirements and minimizing the cost;

[0014] The performance optimization module is used to predict and optimize the propagation performance of each category of communication signals in the communication network.

[0015] In a third aspect, an embodiment of the present application further provides an electronic device, including:

[0016] one or more processors;

[0017] A storage device for storing one or more programs;

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement the signal category-based network simulation method as described in the first aspect.

[0019] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the signal category-based network simulation method as described in the first aspect.

[0020] In a fifth aspect, an embodiment of the present application further provides a computer program product, comprising a computer program and / or instructions, which, when executed by a processor, implements a signal category-based network simulation method as described in any of the above embodiments.

[0021] The embodiment of the present application provides a network simulation method, device, equipment, medium and product based on signal category. The network simulation method based on signal category includes: for each communication signal, inputting the fusion feature of the communication signal into the signal recognition model to determine the category of the communication signal through the signal recognition model, wherein the fusion feature is determined according to the timing feature and spatial feature of the communication signal; constructing a model of the communication network based on a graph representation, wherein each node in the graph represents a network device in the communication network; for each category of the communication signal, optimizing the network topology of the model of the communication network with the goal of meeting the network performance requirements and minimizing the cost; predicting and optimizing the propagation performance of each category of communication signal in the communication network. The category of the communication signal is determined by the fusion feature, and in the network model constructed based on the graph representation, the network topology and propagation performance are optimized for the category, the accuracy and robustness of the network simulation are improved, and targeted optimization solutions are provided for various types of communication signals, thereby improving the communication performance of various types of communication signals in the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0023] Figure 1 A flowchart of a network simulation method based on signal categories provided in an embodiment of the present application;

[0024] Figure 2 A schematic diagram of building a communication network model based on a graph representation provided in an embodiment of the present application;

[0025] Figure 3 A flowchart of a network simulation method based on signal categories provided in an embodiment of the present application;

[0026] Figure 4 A schematic diagram of the structure of a network simulation device based on signal categories provided in an embodiment of the present application;

[0027] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only the parts related to the present application, rather than all structures, are shown in the accompanying drawings.

[0029] It should be mentioned before discussing the exemplary embodiments in more detail that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the steps as sequential processes, many of the steps therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of the steps can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the accompanying drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0030] It should be noted that the concepts of "first", "second", etc. mentioned in the embodiments of the present application are only used to distinguish different devices, modules, units or other objects, and are not used to limit the order or interdependence of the functions performed by these devices, modules, units or other objects.

[0031] In addition, the embodiments in the present application and the features in the embodiments may be combined with each other if there is no conflict.

[0032] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0033] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the relevant content of the solution.

[0034] Figure 1 A flowchart of a network simulation method based on signal categories is provided for an embodiment of the present application. This embodiment can be applied to the case of simulating communication signals and communication networks. Specifically, the network simulation method based on signal categories can be performed by a network simulation device based on signal categories, and the network simulation device based on signal categories can be implemented by software and / or hardware and integrated in an electronic device. Electronic devices include but are not limited to devices with computing functions such as computers, smart phones or servers.

[0035] like Figure 1 As shown, the method specifically comprises the following steps:

[0036] S110. For each communication signal, input the fusion feature of the communication signal into a signal recognition model to determine the category of the communication signal through the signal recognition model, wherein the fusion feature is determined according to the temporal feature and the spatial feature of the communication signal.

[0037] In this embodiment, the communication signal may refer to a modulated signal in a communication network, such as a 5G communication signal or a 6G communication signal. The category of the communication signal may refer to the modulation mode of the communication signal, such as amplitude shift keying (ASK), phase shift keying (PSK), binary phase shift keying (BPSK), frequency shift keying (FSK) and quadrature amplitude modulation (QAM), etc. It may also refer to the length, importance or format of the communication signal, etc. It may also be divided into different categories according to frequency, format, service, etc., such as cellular mobile network signals, WIFI signals, low frequency signals, high frequency signals, amplitude modulation signals, frequency modulation signals, voice signals or image signals, etc. The basis for the classification of communication signals can be set according to actual needs. In the process of communication network simulation, the communication signal can be used as input data, and different categories of communication signals may affect the simulation results (such as network topology, network parameters, cost, etc.).

[0038] For each communication signal, the fusion features can be determined according to its spatial features and temporal features, and its category can be determined according to the fusion features. Spatial features can be understood as the properties of the communication signal in the spatial domain (or time domain), which may include the amplitude, frequency and / or phase of the signal. In the case of using CNN to extract spatial features, spatial features may refer to the local patterns of the communication signal in space, such as edges, textures and / or shapes. Temporal features may include the characteristics and laws of the communication signal changing over time. On this basis, its spatial features and temporal features can be fused, with the aim of merging the outputs of different neural networks into a unified representation for input into the signal recognition model. The feature fusion process can be implemented based on feature mapping and normalization, as well as cascading (or stacking).

[0039] The signal recognition model can be a trained artificial intelligence model such as a deep learning model, which has the ability to perform feature analysis and reasoning on the input signal and output the reasoning result (i.e., the category corresponding to the signal). Its input is the fusion feature of the communication signal, and its output is the corresponding recognition result, i.e., the category of the communication signal.

[0040] On the basis of the above, the signal recognition model can be used to automatically, accurately and efficiently identify the category of communication signals according to the fusion features, providing a reliable input basis for wireless network simulation modeling.

[0041] S120. Construct a model of the communication network based on a graph representation method, where each node in the graph represents a network device in the communication network.

[0042] Figure 2 A schematic diagram of a communication network model based on a graph representation provided in an embodiment of the present application. Figure 2 As shown in the figure, the network model of the communication network is constructed using the graph representation method in graph theory. Each node represents a network device in the communication network, and each edge represents a connection or link between network devices. The formula is as follows: G = (V, E). Where G is a graph, V is a node set, and E is an edge set.

[0043] S130. For each category of communication signals, optimizing the network topology of the model of the communication network with the goal of satisfying network performance requirements and minimizing costs.

[0044] Specifically, the entire communication network is modeled and simulated, the feasibility of the solution is measured through the simulation results, the most reasonable system configuration and parameter settings are selected, and then applied to the actual communication network. During the simulation of the communication network, the data of the input signal can be changed multiple times to observe and analyze the reaction of the constructed communication network model to the input signal and the performance of the communication network during the simulation.

[0045] The input of the communication network model includes the communication signals of the identified categories. During the network simulation, the network topology optimization can be performed for each category, that is, for each category, the network performance maximization and cost minimization of the communication signal network of the category are taken as the optimization goals to determine the optimal network topology.

[0046] Exemplarily, for a category of communication signals (there may be one or more communication signals of this category), its transmission area A in the communication network may involve x1 nodes, y1 edges and z1 links. In order to maximize the network performance and minimize the cost of the communication signal of this category, unnecessary nodes, edges or links in the transmission area A can be removed, or some redundant links in the transmission area A can be optimized to shorten its transmission path while ensuring that the communication signal can reach the destination device; the transmission of some heavily loaded nodes or edges in the transmission area A can also be dispersed to nodes or edges with lighter loads, so that the transmission area A The load within the network is more balanced, avoiding the situation where some nodes are overloaded or underloaded and causing unreasonable resource utilization; similarly, for another category of communication signals (there may be one or more communication signals of this category), its transmission area B in the communication network may involve x2 nodes, y2 edges and z2 links. Unnecessary nodes, edges or links in the transmission area B can be removed, or some links in the transmission area B can be optimized, or the transmission of some heavily loaded nodes or edges in the transmission area B can be dispersed to lightly loaded nodes or edges, etc., so as to maximize the network performance and minimize the cost of this category of communication signals.

[0047] It is understandable that different categories of communication signals may have overlapping transmission areas, nodes, edges or links, and there may be different or even conflicting optimization schemes for different categories of communication signals. In this case, the network topology optimization scheme within the common transmission area (intersection of transmission areas) can be determined with the optimization goals of maximizing the network performance and minimizing the cost of communication signals. The network topology optimization scheme within the common transmission area can also be determined with the optimization goals of maximizing the network performance and minimizing the cost of communication signals in the total transmission area (union of transmission areas). The network topology optimization scheme within the total transmission area can also be determined with the optimization goals of maximizing the network performance and minimizing the cost of communication signals in the entire network graph.

[0048] In one embodiment, for a class of communication signals, the specific method of optimizing the network topology may be: applying a dynamic programming algorithm to balance the goals of minimizing cost and maximizing network performance. Specifically, the cost may be minimized while meeting the network performance requirements. The objective function may be expressed as: Formula: C * =min e∈E {C(e)+Σ v∈V P(v)}, where C * is the optimal cost (minimum cost after optimization), C(e) is the cost of edge e, and P(v) is the performance indicator of node v. Costs can include economic and resource consumption related to network devices and connections. The optimized network topology is both economical and efficient and meets performance requirements.

[0049] S140: Predict and optimize the propagation performance of each category of communication signals in the communication network.

[0050] Specifically, the propagation performance of communication signals in the communication network can be set with corresponding key performance indicators, such as throughput (the amount of data successfully transmitted divided by the transmission time), delay (transmission time of a transmission packet divided by the number of data packets), bit error rate (the ratio of the number of erroneous code elements to the total number of transmitted code elements), signal coverage, network capacity and / or data transmission rate, etc. By predicting the throughput, delay and / or bit error rate of each type of communication signal transmitted in the communication network, the pros and cons of the propagation performance can be predicted, and the network topology and configuration parameters in the network can be adjusted accordingly to optimize the propagation performance. In this process, deep learning algorithms can be introduced to automatically analyze network conditions and make optimization suggestions to achieve dynamic learning and adaptive adjustment.

[0051] By utilizing the self-learning ability of deep learning algorithms, the simulation environment can be adjusted in real time during the simulation process to adapt to changing network conditions. It can process communication signals with high fidelity, automatically extract key features of communication signals through neural networks, and achieve high-precision classification of communication signals through complex pattern recognition; on this basis, it integrates advanced network behavior simulation algorithms and real-time data streams to build a communication network model that is close to the actual operating environment. This highly simulated environment provides a reliable platform for testing, verification, and optimization of communication networks, effectively reducing the deviation between simulation and real-world operations.

[0052] The embodiment of the present application provides a network simulation method based on signal categories, which uses an advanced deep learning model to automatically extract the characteristics of communication signals and perform rapid and accurate identification. This method can significantly reduce manual intervention and improve the efficiency and accuracy of identification. On this basis, for the simulation of the communication network, high-fidelity processing can be performed on the communication signals to achieve high-precision classification of modulated signals, especially in complex communication environments such as multipath effects and signal attenuation, which significantly improves the accuracy of signal identification. In addition, the category of the communication signal is determined by the fusion feature, and in the constructed network model, the network topology and propagation performance are optimized according to the category, the accuracy and robustness of the network simulation are improved, and corresponding optimization solutions are provided for various types of communication signals to improve the communication performance of various types of communication signals.

[0053] In one embodiment, S110 includes:

[0054] S1110. Input the communication signal into a first neural network and a second neural network respectively, so as to extract spatial features of the communication signal through the first neural network and extract temporal features of the communication signal through the second neural network.

[0055] In this embodiment, the signal recognition model includes two neural networks, and the communication signal can be input into the first neural network and the second neural network respectively, so as to extract the spatial features of the communication signal through the first neural network and extract the temporal features of the communication signal through the second neural network. The first neural network and the second neural network are used to extract the spatial features and the temporal features of the communication signal respectively. The first neural network is, for example, various forms of convolutional neural networks (CNN), and the second neural network is, for example, a long short-term memory network (LSTM), a recurrent neural network (RNN) or a Transformer model, etc.

[0056] Optionally, the first neural network is a CNN, and the first neural network can also determine the pattern or structure of the communication signal based on the identifiable and recurring local features in the communication signal, so as to understand the nature of the communication signal and provide a basis for accurately identifying the communication signal. For example, local features may include specific frequency components or waveform features. The pattern or structure may include the following: 1) Modulation features: Different modulation techniques, such as Quadrature Amplitude Modulation (QAM) or Orthogonal Frequency Division Multiplexing (OFDM), will present specific patterns in the communication signal; 2) Signal waveform: The basic shape of the communication signal, such as a sine wave or a square wave, is a manifestation in the time domain; 3) Spectrum distribution: The distribution of the communication signal in the frequency domain, including the location and intensity of the main frequency components.

[0057] Optionally, the second neural network is LSTM, which is a special type of RNN that is sensitive to time and can learn features and patterns in time series data and capture temporal dependencies. It has advantages in tasks such as dynamic change analysis of communication signals or time series prediction. The update formula of the LSTM unit is as follows:

[0058] f t =σ(W f ·[h t-1 , x t ]+b f )

[0059] i t =σ(W i ·[h t-1 , x t ]+b i )

[0060]

[0061] o t =σ(W o ·[h t-1 , x t ]+b o )

[0062] h t =o t ⊙tanh(C t )

[0063] Among them, f t ,i t , o t are the activation values ​​of the forget gate, input gate, and output gate, respectively. are candidate memory cells, C t is the memory cell state, h t is the hidden state output, W represents the weight matrix, W f , W i , W o , W C Respectively represent the weights of the forget gate, input gate, output gate, and memory cell state, b represents the bias vector, and b f 、b i 、b o 、b C Respectively represent the bias vectors of the forget gate, input gate, output gate, and memory cell state, [h t-1 , x t ] represents the concatenation of the hidden state of the previous moment and the input of the current moment, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ represents element-by-element multiplication (Hadamard multiplication).

[0064] In this embodiment, the first neural network and the second neural network work together to extract the spatial features and temporal features of the communication signal respectively, and then combine the two features through feature fusion. The first neural network and the second neural network can share certain layers or interact in different parts of the network to achieve feature fusion and deeper analysis to provide a more comprehensive signal feature representation, which can enable the signal recognition model to more effectively cope with the complexity of the communication signal and improve the accuracy of network simulation based on signal category.

[0065] Optionally, by analyzing the multi-scale time series, the characteristics of the communication signal at different time scales can be extracted. Multi-scale analysis can be performed using wavelet transform: Among them, C A (a, b) are the coefficients of wavelet transform, ψ A is a wavelet function, a and b are scale and translation parameters respectively. It should be noted that the first neural network can be used to extract spatial features of communication signals, such as edges or textures in images, and can also identify local patterns and structures in communication signals. The use of multi-scale features can enhance the feature extraction capability of the first neural network, and provide information at different scales to help the first neural network better understand the complex structures and patterns in the signal. The second neural network is mainly used to process and predict long-term dependencies in time series data. The use of multi-scale features can provide the second neural network with the ability to analyze signals at different time scales, which helps the second neural network accurately capture the short-term and long-term dynamic changes of communication signals.

[0066] Optionally, both the first neural network and the second neural network use multi-scale convolutional layers to respectively extract multi-scale spatial features and temporal features of the communication signal.

[0067] In this embodiment, the communication signal may refer to a converted digital signal. By using a multi-scale convolutional layer, multi-scale features of the communication signal can be extracted. The multi-scale features extracted by the first neural network and the second neural network complement each other and can be used as inputs to the model, providing a more comprehensive signal representation for subsequent signal recognition models and improving the model's ability to recognize and classify signals. The multi-scale features enhance the first neural network's understanding of the local pattern of the signal, and provide support for the second neural network to capture the temporal dynamics of the signal, which helps to identify complex patterns in the signal.

[0068] Optionally, the training process of the signal recognition model includes: for each sample signal, inputting the sample signal into the first neural network and the second neural network respectively, so as to extract the sample space features of the sample signal through the first neural network, and extract the sample time series features of the sample signal through the second neural network; fusing the sample space features and the sample time series features to obtain the sample fusion features of the sample signal; selecting the target sample features from the sample fusion features; and training the signal recognition model based on the target sample features of each sample signal.

[0069] In this embodiment, for each sample signal, the multi-scale features extracted by the first neural network and the second neural network are fused, and then the target sample features are selected (DBN and feature selection method based on mutual information can be used to identify the feature subset most relevant to the target variable. The optimized target sample features will be used to train the signal recognition model to achieve high-precision recognition and classification of communication signals. The trained model will be deployed in an actual communication network environment to provide real-time signal analysis and network optimization decisions.

[0070] S1120: Fuse the spatial feature and the temporal feature to obtain an original fused feature of the communication signal.

[0071] Among them, the feature fusion process can be implemented based on feature mapping and normalization, as well as cascading (or stacking).

[0072] S1130, converting the original fusion features into abstract features through a deep belief network;

[0073] In this embodiment, the Deep Belief Network (DBN) is a generative model composed of multiple stacked restricted Boltzmann machines (RBM), which is suitable for feature fusion and abstract representation of high-level features. DBN processes signals in a layer-by-layer manner, and each layer extracts features from the input data and fuses these features into higher-level abstract features. Using DBN layer-by-layer abstraction, low-level original fused features (such as basic statistical characteristics and simple patterns of communication signals) can be converted into higher-level abstract features, thereby deeply understanding the essence of the signal and better representing the intrinsic properties and patterns of the communication signal. It provides a powerful feature representation for the recognition and classification of communication signals. Specifically, each RBM layer can be activated in the following ways: i =σ(∑ j W ij h j +b i )h j =σ(∑ i W ij v i +b j ), where v and h represent the activation values ​​of the visible layer and the hidden layer respectively, W is the weight matrix, and b is the bias term.

[0074] S1140: Based on the mutual information between features and categories, select some features from the abstract features as fusion features of the communication signal.

[0075] In this embodiment, for abstract features, feature selection can be used to optimize the feature set and remove redundant or irrelevant features to improve the recognition performance of the signal recognition model. The selected features can be understood as representative features that are conducive to distinguishing the types of communication signals, or high-level abstract features that can represent the intrinsic properties and complex patterns of communication signals, which can achieve efficient and accurate recognition of communication signals using fewer, high-quality features.

[0076] Exemplarily, feature selection can be performed based on mutual information (MI), whereby the abstract feature with the highest mutual information can be selected as the target feature, which helps to improve the prediction accuracy of the model. Based on mutual information, the interdependence or correlation between the abstract features and the target variable can be evaluated to determine which abstract features have the greatest information sharing with the target variable (the predicted or classified variable, i.e., the category of the communication signal), thereby selecting the most useful target feature, which contains the most informative information about the target variable.

[0077] Specifically, calculate the mutual information between each abstract feature and the target variable:

[0078] Among them, I(x;y) represents the mutual information between the abstract feature x and the target variable y, p(x,y) is the joint probability distribution of the abstract feature and the target variable, and p(x) and p(y) are the marginal probability distributions of the abstract feature and the target variable, respectively.

[0079] Optionally, in the process of extracting spatial features by the first neural network, frequency domain features may also be extracted as part of the fusion features, so as to more comprehensively analyze the features of the communication signal and further improve the accuracy of communication signal recognition. Specifically, the method of extracting frequency domain features may include at least one of the following:

[0080] Method 1: Convert the time domain signal into a frequency domain signal. For example, convert the time domain signal into a frequency domain representation through Fast Fourier Transform (FFT), and then use the first neural network to analyze the frequency domain representation. The convolution layer in the first neural network can identify patterns in the frequency domain, such as the distribution of specific frequencies or specific structures in the spectrum.

[0081] Method 2: By applying convolutional layers at different scales, the first neural network can capture both local and global features of the communication signal and analyze the behavior of the communication signal at different time scales, which may manifest as the importance of different frequency components in the frequency domain.

[0082] Method 3: Through forward propagation, forward propagation can refer to the process in which the communication signal passes through the first neural network, and the output of each convolutional layer becomes the input of the next convolutional layer. In this process, the first neural network can learn the hierarchical features of the signal, which may include patterns in the frequency domain.

[0083] Exemplarily, the forward propagation formula is: Z (l) =W (l) *A (l-1) +b (l) A (l) =σ(Z (l) , where W (l) represents the convolution kernel weight of the ,th layer, b (l) Represents the bias term of the,th layer, A (l-1) is the activation output of the previous layer, Z (l) is the linear output of the current layer, and σ is the activation function, such as ReLU (σ(x) = max(0, x)).

[0084] Optionally, before the communication signal to be identified is input into the first neural network and the second neural network respectively, the communication signal can be preprocessed, for example, using a high-precision analog-to-digital converter (ADC) to convert the analog signal into a digital signal to ensure the digitization accuracy of the signal; for example, the communication signal can be processed by an adaptive filter group and multi-scale time series analysis to eliminate the influence of noise and signal attenuation and improve the quality of the communication signal.

[0085] Exemplarily, the input of the first neural network can be a communication signal collected and digitized by an ADC. Before inputting into the first neural network, the digital signal can also be preprocessed, such as filtering and normalization, to ensure the quality of the signal and highlight important features, thereby automatically extracting the frequency domain features and time domain features of the communication signal using the convolutional layer of the CNN.

[0086] Exemplarily, the input of the second neural network can be a preprocessed digital signal sequence, so as to highlight the time series characteristics of the communication signal. These sequences can be digital signals obtained by ADC conversion of the communication signal, and the communication signal can also be preprocessed, such as segmentation, normalization, differentiation and / or feature extraction, so that the second neural network can more effectively learn and predict the time series characteristics of the signal. Among them, segmentation can refer to dividing the continuous signal into a series of shorter time series, each sequence containing a certain number of data points; normalization can refer to adjusting the amplitude range of the signal to make it suitable for neural network processing; differentiation can refer to calculating the difference between consecutive data points to highlight the change of the signal; feature extraction can refer to extracting the statistical characteristics or other time series characteristics of the signal. After preprocessing, the noise can be reduced as much as possible and the signal-to-noise ratio of the communication signal can be improved.

[0087] Communication signals may have different signal-to-noise ratios after different degrees of preprocessing. Communication signals with different modulation modes (such as 4ASK, BPSK, QPSK, OQPSK, 8PSK, 16QAM and 32QAM) have different processing effects under noises with different signal-to-noise ratios.

[0088] Optionally, before inputting the communication signal to be identified into the first neural network and the second neural network respectively, the method further includes:

[0089] S1010, converting the communication signal into a frequency domain signal through Fourier transform;

[0090] S1020, filtering the frequency domain signal through a bandpass filter;

[0091] S1030 , converting the filtered frequency domain signal into a time domain signal through inverse Fourier transform.

[0092] Specifically, an adaptive filter group can be established to filter the communication signal to perform real-time noise reduction, enhancement and frequency band selection to improve the quality of the communication signal. The adaptive filter group can dynamically adjust its parameters to adapt to the characteristics of the communication signal, ensuring that the useful information in the communication signal is retained and enhanced while reducing the impact of noise and other interference components. The pre-processed communication signal is more suitable for feature extraction and analysis of the deep learning model.

[0093] Taking the Least Mean Square (LMS) error algorithm as an example, the weight update formula of the adaptive filter is: Among them, W n is the filter weight of the nth iteration, μ is the learning rate, e n is the error signal, is the conjugate of the input signal.

[0094] In this embodiment, the adaptive filter bank can use a bandpass filter to better remove high-frequency and low-frequency noise in the communication signal. The filtering process can be expressed as: in, and denote the Fourier transform and inverse transform respectively, and H(f) is the frequency response of the bandpass filter.

[0095] Specifically, the communication signal (analog signal, x(t)) can be converted to the frequency domain through Fourier transform, and then the frequency response H(f) of the bandpass filter can be used to selectively retain the frequency components related to the signal in the frequency domain. Finally, the processed signal is converted back to the time domain through inverse Fourier transform to obtain the filtered signal y(t), thereby improving the quality of the communication signal.

[0096] Optionally, before inputting the communication signal to be identified into the first neural network and the second neural network respectively, the method further includes:

[0097] S1040, decomposing the communication signal by wavelet transform to obtain wavelet coefficients of different frequencies;

[0098] S1050, processing wavelet coefficients of different frequencies according to a threshold value;

[0099] S1060: Perform inverse wavelet transform on the processed wavelet coefficients to obtain a preprocessed communication signal.

[0100] In this embodiment, in order to ensure efficient application of denoising, the following operations may also be performed on the communication signal: filtered (t) = WPT -1 {WPT{x(t)}·Ψ}, where WPT and WPT -1 and WPT-1 They represent wavelet transform and inverse transform respectively, Ψ represents the denoising threshold, x filtered (t) represents the preprocessed communication signal.

[0101] Specifically, through wavelet decomposition, the communication signal can be decomposed into wavelet coefficients of different frequencies. The wavelet coefficients represent the energy distribution of the communication signal at different frequencies. The following wavelet functions can be used in this process: Haar wavelet, Daubechies wavelet, Morlet wavelet, etc.; threshold processing is performed on the wavelet coefficients to set the wavelet coefficients with lower energy to zero, thereby suppressing noise. Hard threshold processing or soft threshold processing can be used. Hard threshold processing can be understood as setting coefficients less than the threshold to zero, and soft threshold processing can be understood as attenuating coefficients less than the threshold; the wavelet coefficients after threshold processing are inversely transformed to obtain the denoised communication signal. The inverse transformation process can restore the original signal by linearly combining the wavelet coefficients with the wavelet basis functions.

[0102] Through the above wavelet transform denoising process, the communication signal can be analyzed simultaneously in the time domain and the frequency domain, the noise can be suppressed more accurately, and the time-frequency characteristics of the communication signal can be retained. By decomposing the communication signal into wavelet coefficients of different frequencies, the communication signal can be analyzed at multiple resolutions, and the noise of different frequencies can be suppressed to different degrees. In some embodiments, the threshold can be adaptively selected according to the energy characteristics of the communication signal, so that the wavelet transform can better adapt to the noise characteristics of different communication signals, and further improve the denoising effect.

[0103] Optionally, for communication signals, frequency domain analysis and filtering are first performed using Fourier transform and its inverse transform, and then wavelet transform is used to perform denoising processing in the time-frequency domain to improve the quality of communication signals.

[0104] In one embodiment, S1120 includes:

[0105] S11210, normalizing the spatial feature and the temporal feature respectively, so that the spatial feature and the temporal feature have the same scale and distribution;

[0106] S11220. Cascade the normalized spatial features and the temporal features to obtain fused features of the communication signal.

[0107] S1210, normalizing the spatial features and the temporal features respectively, so that the spatial features and the temporal features have the same scale and distribution;

[0108] S1220, cascade the normalized spatial features and temporal features to obtain a fusion feature of the communication signal. In this embodiment, the normalization of the spatial features and the temporal features can be achieved by the following formula: Where F is the feature vector of spatial features or the feature vector of temporal features. Correspondingly, F norm is the normalized feature vector of the spatial feature or the feature vector of the temporal feature, μ is the mean of the feature vector of the spatial feature or the mean of the feature vector of the temporal feature, and σ is the standard deviation of the feature vector of the spatial feature or the standard deviation of the feature vector of the temporal feature. The normalized spatial feature and temporal feature are concatenated to form: F concat =[F CNN ; F LSTM ], where F CNN The feature vector representing the normalized spatial features corresponding to the first neural network, F LSTM The feature vector representing the normalized time series features corresponding to the second neural network, F concat The feature vector represents the fused features after cascading. The semicolon represents the vertical stacking of feature vectors.

[0109] Figure 3 A flowchart of a network simulation method based on signal categories provided in an embodiment of the present application.

[0110] like Figure 3 As shown, the method includes:

[0111] S1. Convert the communication signal to be identified into a frequency domain signal by Fourier transform;

[0112] S2, filtering the frequency domain signal through a bandpass filter;

[0113] S3, converting the filtered frequency domain signal into a time domain signal by inverse Fourier transform;

[0114] S4, decomposing the communication signal by wavelet transform to obtain wavelet coefficients of different frequencies;

[0115] S5, processing the wavelet coefficients of different frequencies according to the threshold value;

[0116] S6. Perform inverse wavelet transform on the processed wavelet coefficients to obtain the preprocessed communication signal

[0117] S7, inputting the preprocessed communication signal into the first neural network and the second neural network respectively to obtain the spatial characteristics and temporal characteristics of the communication signal;

[0118] S8, normalize the spatial features and temporal features respectively;

[0119] S9, cascading the normalized spatial features and temporal features to obtain the original fusion features of the communication signal;

[0120] S10, converting the original fusion features into abstract features through a deep belief network;

[0121] S11, select target features from abstract features based on the mutual information between features and categories;

[0122] S12: Input the target feature into a signal recognition model to determine the category of the communication signal through the signal recognition model.

[0123] The network simulation method based on signal category in this embodiment adopts an advanced feature extraction mechanism, such as using multi-scale convolution kernels and pooling layers in convolutional neural networks to achieve multi-level feature abstraction of communication signals and extract key features in frequency domain and time domain; it also adopts temporal dependency modeling, that is, through the gating mechanism of long short-term memory network, including input gate, forget gate and output gate, the temporal features of communication signals are effectively captured to deal with the long-term dependency problem of time series data. On this basis, through the feature fusion and feature selection process, the features most relevant to the target variable are selected and identified through the signal recognition model, so that the category of the communication signal can be accurately determined.

[0124] On this basis, the network simulation method of this embodiment takes into account that the current simulation of communication networks mainly relies on static, preset parameter configurations, lacks the ability to dynamically learn and adapt to different network conditions, and usually requires users to manually configure network parameters and topology structures, and cannot be adaptively adjusted; and considering that the simulation accuracy of communication networks needs to be improved, existing simulation tools often cannot fully capture network dynamics and complex scenarios when simulating actual network operations, resulting in deviations between simulation results and the real world, affecting the practical value of simulation. In this regard, by integrating advanced network behavior simulation algorithms and real-time data streams, a simulation model that is closer to the actual operating environment is constructed. This highly simulated environment provides a reliable platform for testing, verification and optimization of communication networks, effectively reducing the deviation between simulation and real operations.

[0125] In one embodiment, S130 includes:

[0126] S1310, dividing the network topology into at least two sub-networks;

[0127] S1320. For each of the sub-networks and each category of communication signals, optimize the network topology of the sub-network with the goal of meeting network performance requirements and minimizing costs.

[0128] In this embodiment, by dividing the network topology into at least two sub-networks, targeted simulation and optimization can be performed on different sub-networks. A sub-network can be understood as a sub-graph of a graph representation of a model of the entire communication network.

[0129] Exemplarily, subnetworks can be divided according to different categories of communication signals. For example, the transmission area of ​​each category of communication signal is respectively regarded as a corresponding subnetwork. For N categories of communication signals, there are N subnetworks. The advantage of doing this is that it is more conducive to targeted simulation, optimization and prediction for each category of communication signal.

[0130] Optionally, the subnetworks can also be divided according to the degree of co-reference, where the nodes in the network topology represent network devices, and the edges represent the degree of co-reference between network devices. The principle for dividing the subnetworks is that the ratio of the weight of a hyperedge (an edge that allows the connection of at least two nodes) that starts from a node in the subgraph and crosses its boundary to the sum of the weights of all hyperedges starting from the node of the subgraph is less than a set threshold, and the scale of each subgraph is relatively close, and the difference is within a set range. Among them, for a category of communication signals, the weight of the hyperedge can be comprehensively determined based on the transmission performance (transmission time, transmission quality and / or transmission distance, etc.) of the communication signal of this type on the link corresponding to the hyperedge. On this basis, each device and link in a subnetwork has similar performance, different subnetworks have significantly different performance, and the complexity of different subnetworks is close, thereby improving the reliability and rationality of subnetwork simulation and optimization.

[0131] Exemplarily, optimizing the network topology of a subnetwork can be understood as removing unnecessary nodes, edges or links in the subnetwork, or optimizing some redundant links in the subnetwork to shorten the transmission path while ensuring that the communication signal can reach the destination device; it is also possible to disperse the transmission on some heavily loaded nodes or edges in the subnetwork to lightly loaded nodes or edges, so that the load within the subnetwork is more balanced, avoiding unreasonable resource utilization caused by overload or underload of some nodes, etc., and the purpose is to maximize the network performance and minimize the cost of communication signals of various categories within the subnetwork.

[0132] In one embodiment, S130 includes:

[0133] S1330. Based on a graph neural network algorithm, extract node features and edge features from the network topology of the model of the communication network.

[0134] Among them, the characteristics of nodes can be understood as the characteristics of network devices, such as the type or model of network devices, traffic size, computing power, failure rate, the number of communication signals that can be carried per unit time or the communication frequency, etc.; the characteristics of edges can be understood as the characteristics of links between network devices, such as bandwidth, latency, and channel quality, etc. The Graph Neural Network (GNN) algorithm is an algorithm that uses neural networks to learn graph structure data, extract and discover features and patterns in graph structure data. Each graph convolution layer in the neural network can learn the interactions and dependencies between nodes, aggregate the information of adjacent nodes, increase the perception range, and enable each node to capture a wide range of topological structure information; the last layer is a global pooling layer, which can aggregate the features of all nodes and edges for scoring network performance and cost.

[0135] S1340. For each category of the communication signal, determine a network performance score and a cost score corresponding to the category according to the characteristics of the node and the characteristics of the edge.

[0136] Among them, a scoring model can be trained, which has performance evaluation capabilities, can learn the relationship and rules between the characteristics of nodes and edges and the performance scores and cost scores, and input the characteristics of nodes and edges into the scoring model to infer and output network performance scores and cost scores. Exemplarily, the more advanced the type or model of network equipment, the larger the traffic, the stronger the computing power, the lower the failure rate, the larger the number of communication signals or communication frequency that can be carried per unit time, the larger the bandwidth of the link, the lower the delay or the better the channel quality, the higher the network performance score and the lower the cost score.

[0137] S1350. Optimize nodes and edges of the network topology according to the network performance score and the cost score.

[0138] Among them, according to the network performance score and cost score, redundant nodes and edges in the network topology can be determined and removed; nodes and edges with heavier loads can be determined and their loads can be dispersed to other nodes and edges with lighter loads to achieve load balancing and reasonable bandwidth allocation; the transmission path of communication signals can be optimized, such as reducing the number of hops, increasing link bandwidth, etc., thereby shortening the transmission path and reducing latency; the number of nodes and links can be increased, redundant paths can be established, etc., to improve the capacity and scalability of the network, meet the growing communication needs, and on this basis improve network performance and reduce costs.

[0139] It should be noted that the network topology can be divided into at least two sub-networks, and for each sub-network of the network topology, a graph neural network algorithm is used to extract node features and edge features and perform network performance scoring and cost scoring and optimization.

[0140] In one embodiment, S140 includes: predicting and optimizing the propagation performance of the communication signal in the communication network by random forest, gradient boosting machines (GBM) or deep deterministic policy gradient (DDPG).

[0141] Taking GBM as an example, the principles of predicting and optimizing network performance are as follows: in, is the predicted value, η is the intercept, T k (x) is a decision tree. GBM builds a strong prediction model by combining multiple weak prediction models (usually decision trees). In the context of network performance optimization, GBM can be used to predict key performance indicators such as throughput, latency, and bit error rate, and then adjust the network configuration based on these predictions to optimize performance. However, before training the GBM model, a set of training data needs to be collected and labeled. These data include the input features of the network and the corresponding performance indicators (such as throughput, latency, etc.), and these training data are used to train the GBM model. The model will learn the mapping relationship between input features and performance indicators. The finally trained GBM model can be used to predict performance indicators under new network configurations or conditions, or to evaluate the impact of different configurations on performance.

[0142] The input of GBM includes features extracted from the signal processing stage and parameters obtained from the network model construction stage. These features comprehensively describe the state and configuration of the network. The output of GBM is the predicted performance indicator value. By training the GBM model to learn the relationship between input features and performance indicators, the model's predictions can be used to guide the adjustment of network configuration to optimize network performance. For example, if GBM predictions show that adding certain network resources can improve throughput, resource allocation strategies can be adjusted based on these predictions.

[0143] For DDPG, Q-Learning or Policy Gradient methods can be used. In addition, cloud computing platforms and distributed computing resources can also be used for analytical calculations. The DDPG algorithm learns and optimizes the strategy of network operations through the interaction between the intelligent agent and the simulation environment, and dynamically adjusts the simulation parameters to maximize network performance indicators such as throughput and reduce latency. DDPG is able to handle high-dimensional network configuration space and respond to changes in network conditions in real time to achieve better network performance. During the simulation process, the DDPG algorithm uses neural networks to approximate the optimal strategy and continuously updates the strategy to adapt to the dynamic changes of the network environment. Through reinforcement learning, the algorithm can self-optimize without manual parameter adjustment, greatly reducing the complexity and cost of simulation. In addition, the DDPG algorithm has shown significant technical advantages in dealing with nonlinear and highly interactive network performance optimization problems, including faster convergence speed and higher policy execution accuracy.

[0144] Optionally, during the simulation process, a frequency domain smoothing method may be used to accurately simulate the propagation and changes of signals in an actual network, thereby narrowing the gap between simulation and actual deployment.

[0145] Optionally, the operation and maintenance platform provides a user-friendly interactive interface, allowing users to easily build, configure and debug the network.

[0146] The network simulation process of this embodiment can automatically adjust simulation parameters according to network conditions to improve the intelligence and accuracy of the simulation; the simulation results are close to the actual deployment environment and performance and can be directly applied to actual network optimization, providing more reliable data support for network optimization.

[0147] In one embodiment, the method further includes: S1410, simulating the propagation of communication signals in the communication network by using a simulation engine based on a network simulation technology software package or a network simulation tool.

[0148] Among them, the network simulation technology software package is, for example, OPNET, and the network simulation tool is, for example, NS-3. The node configuration and parameters in the NS-3 network simulation together define the simulated network environment, which is used to simulate and test the behavior and performance of the network under specific conditions. Use a simulation engine based on NS-3 to simulate operations in the communication network to evaluate and optimize network performance. Simulation includes building a network model, configuring network parameters and traffic patterns, and running simulations to collect performance data. By analyzing the simulation results, it is possible to identify bottlenecks in network performance, evaluate the impact of different network configurations, and optimize network design to increase throughput, reduce latency, and improve overall network efficiency. The ultimate goal of simulation is to ensure that the communication network can meet performance requirements and provide users with high-quality services.

[0149] The core of the signal category-based network simulation method of this embodiment is the use of a highly integrated deep learning architecture that integrates convolutional neural networks and long short-term memory networks to achieve efficient extraction of signal features and accurate modeling of timing dependencies. Through end-to-end simulation optimization strategies, it can provide a continuous solution from network topology planning to performance evaluation, ensuring high fidelity and real-time performance of the simulation process; the adaptive algorithms embedded in the system, such as the parameter adjustment mechanism based on deep reinforcement learning, can achieve instant response to dynamic network conditions, thereby optimizing network configuration and improving overall performance.

[0150] It should be noted that the network simulation process based on signal categories and the simulation process of the communication network in the above embodiment are two complementary processes, and both jointly support the performance analysis and optimization of the communication system. The features extracted in the feature processing stage of the communication signal (such as timing features, spatial features, fusion features, abstract features and target features, etc.) can provide input data for the simulation of the communication network, which is used to define the properties and behaviors of the communication signal in the communication network simulation, such as signal strength, modulation mode, etc.; the communication network simulation can evaluate and optimize the propagation performance of the communication signal in the network based on the characteristics of the communication signal, such as signal coverage, network capacity and / or data transmission rate, etc. These evaluations can be based on the features obtained from the signal feature processing stage. The optimization of signal features can improve the accuracy of network simulation, and the optimization of network models can improve the efficiency and effect of network operations. The feature processing of communication signals and the communication network simulation work together to provide reliable support for the design and optimization of communication systems.

[0151] Optionally, in order to take user interaction into consideration during the simulation process, an intelligent user interaction interface can be built by integrating natural language processing (NLP) and sentiment analysis technology, so that users can interact with the system in a more natural and intuitive way; the performance evaluation system adopts a multi-objective optimization algorithm, combined with artificial intelligence-assisted decision analysis tools, to provide comprehensive and in-depth network performance prediction and optimization, and can build a network performance evaluation system based on multi-objective optimization, using the Pareto optimal theory to balance the trade-offs between different performance indicators and achieve comprehensive optimization of network performance.

[0152] Optionally, in terms of security, quantum encryption and blockchain technology can be used to provide multi-level security protection for data transmission and processing, ensuring the absolute security of user data and privacy.

[0153] Optionally, microservices and containerization technologies can be adopted in the system architecture. Containerization technology and continuous integration (CI) and continuous deployment (CD) processes can be used to achieve high modularity, rapid iteration, and scalability of the microservice architecture, and support rapid iteration and elastic scaling of cloud-native applications.

[0154] Optionally, in terms of multimodal data processing, deep coupling networks and multi-task learning frameworks can be used to achieve effective fusion and analysis of cross-modal information. Multimodal data fusion technology based on DBN or multimodal deep learning frameworks combines data from signal processing and machine vision to perform cross-modal feature learning.

[0155] Optionally, the network can be given adaptability and autonomy by integrating self-organizing network (Self-Organizing Network) technology, including self-configuration, self-healing and self-optimization functions, so that it can automatically respond to changes in the external environment and achieve intelligent management and optimization of network resources.

[0156] Optionally, software defined network (SDN) and network function virtualization (NFV) technologies are used to achieve flexible configuration and optimization of network resources.

[0157] Optionally, a heterogeneous computing platform is used, combining the computing power of the central processing unit (CPU) and the graphics processing unit (GPU), and using the Compute Unified Device Architecture (CUDA) or Open Computing Language (OpenCL) for parallel computing optimization to accelerate the training and reasoning process of deep learning models.

[0158] Optional, dynamic network behavior simulation: Simulate dynamic behaviors and adaptive mechanisms in the network using agent-based modeling (ABM) and meta-heuristic algorithms such as genetic algorithms (GA) or particle swarm optimization (PSO).

[0159] Figure 4 A schematic diagram of a network simulation device based on signal categories provided in an embodiment of the present application. The network simulation device based on signal categories provided in this embodiment includes:

[0160] An identification module 210, for each communication signal, inputting the fusion feature of the communication signal into a signal identification model, so as to determine the category of the communication signal through the signal identification model, wherein the fusion feature is determined according to the temporal feature and the spatial feature of the communication signal;

[0161] A construction module 220, for constructing a model of the communication network based on a graph representation, wherein each node in the graph represents a network device in the communication network;

[0162] A topology optimization module 230, for optimizing the network topology of the model of the communication network for each category of the communication signal with the goal of meeting the network performance requirements and minimizing the cost;

[0163] The performance optimization module 240 is used to predict and optimize the propagation performance of each category of communication signals in the communication network.

[0164] The device uses fusion features to determine the category of communication signals. In the constructed network model, it optimizes the network topology and propagation performance according to the category, improves the accuracy and robustness of network simulation, provides corresponding optimization solutions for various communication signals, and improves the communication performance of various communication signals.

[0165] Optionally, the device further includes: a fusion feature determination module, configured to:

[0166] Inputting the communication signal into a first neural network and a second neural network respectively, so as to extract spatial features of the communication signal through the first neural network and extract temporal features of the communication signal through the second neural network;

[0167] Fusing the spatial feature and the temporal feature to obtain an original fused feature of the communication signal;

[0168] Converting the original fusion features into abstract features through a deep belief network;

[0169] Based on the mutual information between features and categories, some features are selected from the abstract features as fused features of the communication signal.

[0170] Optionally, fusing the spatial feature and the temporal feature to obtain a fused feature of the communication signal includes:

[0171] Normalizing the spatial features and the temporal features respectively so that the spatial features and the temporal features have the same scale and distribution;

[0172] The normalized spatial features and the temporal features are cascaded to obtain fused features of the communication signal.

[0173] Optionally, for each category of communication signals, optimizing the network topology of the model of the communication network with the goal of satisfying network performance requirements and minimizing costs includes:

[0174] Based on the graph neural network algorithm, extracting node features and edge features from the network topology of the model of the communication network;

[0175] For each category of the communication signal, determine a network performance score and a cost score corresponding to the category according to the characteristics of the node and the characteristics of the edge;

[0176] Nodes and edges of the network topology are optimized according to the network performance score and the cost score.

[0177] Optionally, according to the category of each communication signal, the network topology of the model of the communication network is optimized with the goal of meeting network performance requirements and minimizing costs, including:

[0178] Dividing the network topology into at least two sub-networks;

[0179] For each of the sub-networks and each category of communication signals, the network topology of the sub-network is optimized with the goal of meeting network performance requirements and minimizing costs.

[0180] Optionally, predicting and optimizing the propagation performance of each category of communication signals in the communication network includes:

[0181] The propagation performance of each category of communication signals in the communication network is predicted and optimized by random forest, gradient boosting machine or deep deterministic policy gradient DDPG.

[0182] The signal category-based network simulation device provided in the embodiment of the present application can be used to execute the signal category-based network simulation method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0183] Figure 5 A schematic diagram of an electronic device 10 that can be used to implement an embodiment of the present application is shown. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 10 can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, user equipment, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0184] like Figure 5As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0185] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, wireless networks.

[0186] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs the various methods and processes described above.

[0187] In some embodiments, the method of the above embodiment may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform any of the above embodiment methods in any other appropriate manner (e.g., by means of firmware).

[0188] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0189] The computer programs for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer programs are executed by the processor, the functions / operations specified in the flow charts and / or block diagrams are implemented. The computer programs may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0190] In the context of the present application, a computer readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device or equipment. A computer readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer readable storage medium may be a machine readable signal medium. A more specific example of a machine readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0191] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device 10 having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device 10. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0192] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0193] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0194] An embodiment of the present application further provides a computer program product, including a computer program and / or instructions, which, when executed by a processor, implements the signal category-based network simulation method as described in any of the above embodiments.

[0195] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0196] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A network simulation method based on signal category, characterized in that: include: For each communication signal, inputting the fusion feature of the communication signal into a signal recognition model to determine the category of the communication signal through the signal recognition model, wherein the fusion feature is determined according to the temporal feature and the spatial feature of the communication signal; A communication network model is constructed based on a graph representation, where each node in the graph represents a network device in the communication network; For each category of communication signals, optimizing the network topology of the model of the communication network with the goal of meeting network performance requirements and minimizing costs; The propagation performance of each category of communication signals in the communication network is predicted and optimized.

2. The method according to claim 1, characterized in that Also includes: Inputting the communication signal into a first neural network and a second neural network respectively, so as to extract spatial features of the communication signal through the first neural network and extract temporal features of the communication signal through the second neural network; Fusing the spatial feature and the temporal feature to obtain an original fused feature of the communication signal; Converting the original fusion features into abstract features through a deep belief network; Based on the mutual information between features and categories, some features are selected from the abstract features as fused features of the communication signal.

3. The method according to claim 1, characterized in that The spatial feature and the temporal feature are fused to obtain a fused feature of the communication signal, including: Normalizing the spatial features and the temporal features respectively so that the spatial features and the temporal features have the same scale and distribution; The normalized spatial features and the temporal features are cascaded to obtain fused features of the communication signal.

4. The method according to claim 1, characterized in that: According to the category of each communication signal, the network topology of the model of the communication network is optimized with the goal of meeting network performance requirements and minimizing costs, including: Dividing the network topology into at least two sub-networks; For each of the sub-networks and each category of communication signals, the network topology of the sub-network is optimized with the goal of meeting network performance requirements and minimizing costs.

5. The method according to claim 1, characterized in that: For each category of communication signals, the network topology of the model of the communication network is optimized with the goal of meeting network performance requirements and minimizing costs, including: Based on the graph neural network algorithm, extracting node features and edge features from the network topology of the model of the communication network; For each category of the communication signal, determine a network performance score and a cost score corresponding to the category according to the characteristics of the node and the characteristics of the edge; Nodes and edges of the network topology are optimized according to the network performance score and the cost score.

6. The method according to claim 1, characterized in that Predicting and optimizing the propagation performance of each category of communication signals in the communication network, including: The propagation performance of each category of communication signals in the communication network is predicted and optimized by random forest, gradient boosting machine or deep deterministic policy gradient DDPG.

7. A network simulation device based on signal category, characterized in that: include: an identification module, configured to input, for each communication signal, a fusion feature of the communication signal into a signal identification model, so as to determine the category of the communication signal through the signal identification model, wherein the fusion feature is determined according to a temporal feature and a spatial feature of the communication signal; A building module, for building a model of a communication network based on a graph representation, where each node in the graph represents a network device in the communication network; A topology optimization module, for optimizing the network topology of the model of the communication network for each category of the communication signal with the goal of meeting the network performance requirements and minimizing the cost; The performance optimization module is used to predict and optimize the propagation performance of each category of communication signals in the communication network.

8. An electronic device, characterized in that: include: at least one processor; a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the signal category-based network simulation method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the signal category-based network simulation method as described in any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program and / or instructions, characterized in that: When the computer program and / or the instructions are executed by a processor, the signal category-based network simulation method as described in any one of claims 1 to 7 is implemented.

Images