A wireless network resource allocation method based on domain adversarial network

Abstract

The application provides a wireless network resource allocation strategy based on a domain adversarial network, which is used for a multi-base station scene in a wireless network, uses an artificial intelligence algorithm, and solves a core problem: how to reasonably allocate corresponding wireless resources according to real-time network traffic demands of each base station to ensure the quality of the whole network. The algorithm combines an inter-cell interference coefficient obtained through theoretical analysis, uses network traffic and performance big data as a basis, and proposes a domain adversarial network using a performance evaluator for guidance to realize dynamic resource allocation based on real-time traffic, and then minimizes the number of base stations with a performance evaluation of "poor" in the network and reduces the overall wireless resource consumption of the network.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, specifically relating to a wireless network resource allocation method based on domain adversarial networks. Background Technology

[0002] Resource allocation is a crucial aspect of wireless communication systems, directly impacting mobile network performance and user experience. With the development of communication technologies and increasing network service demands, the number of base stations in wireless networks is constantly growing, leading to higher user density and making wireless resource allocation more complex and challenging. Firstly, facing a massive number of mobile users, base stations need more effective methods to manage limited wireless resources to meet users' demands for high-capacity, low-latency communication. Traditional wireless resource allocation algorithms generally rely on mathematical models or optimization theorems, such as convex optimization and heuristic algorithms, which often struggle to adapt to constantly changing network environments. New communication technologies offer higher bandwidth and lower latency, but also greater complexity, requiring new allocation methods to more intelligently allocate base station wireless resources to adapt to different service scenarios and network needs. Furthermore, interference between base stations and the limited spectrum resources also constrain wireless resource allocation. In densely deployed cellular networks, interference can degrade network performance, while limited spectrum resources demand more efficient bandwidth utilization. Therefore, allocation algorithms also need to allocate limited spectrum resources rationally to avoid or reduce interference.

[0003] With the development of big data technology and artificial intelligence, deep neural networks have begun to emerge in the field of wireless communication, such as wireless traffic prediction, network anomaly detection, and wireless network performance optimization. They can autonomously learn complex nonlinear relationships from large amounts of real data and effectively handle multidimensional and dynamic environmental features, further reducing system design complexity while also possessing a certain degree of generalization ability. These advantages enable deep neural networks to provide more intelligent and efficient resource allocation strategies for wireless communication systems. To further accelerate model convergence efficiency and improve adaptability to different scenarios, transfer learning applicable to different data distributions is widely used. Summary of the Invention

[0004] In view of this, the present invention considers domain adversarial networks as a method in transfer learning, which can reduce distributional differences between different domains and perform reasonable network mapping. Applied to wireless resource allocation, this feature is expected to ensure appropriate resource allocation to each base station while improving the overall performance of the cellular network.

[0005] This invention proposes a wireless network resource allocation method based on domain adversarial networks (DANs), which uses DANs from deep learning to solve the problem of real-time dynamic resource allocation among different base stations in multi-base station scenarios. This strategy allocates reasonable real-time resources to each base station while maximizing network demand, thereby effectively reducing the number of base stations with "poor" performance evaluations and lowering the overall wireless resource consumption of the network.

[0006] This strategy targets multi-base station scenarios in wireless networks, utilizing artificial intelligence algorithms to address a core problem: how to rationally allocate corresponding wireless resources based on the real-time network traffic demands of each base station to ensure overall network quality. The algorithm combines theoretically derived inter-cell interference coefficients with large-scale network traffic and performance data, proposing a domain adversarial network guided by a performance evaluator to achieve dynamic resource allocation based on real-time traffic. This minimizes the number of base stations with "poor" performance evaluations and reduces overall network wireless resource consumption.

[0007] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0008] A method for allocating wireless network resources based on domain adversarial networks, comprising:

[0009] In a multi-base station wireless network scenario, the base station parameters available for configuration include at least: engineering configuration parameters, real-time traffic and performance parameters;

[0010] A network performance evaluator is used to assess whether real-time resources can meet the demands of real-time traffic by utilizing the real-time traffic of each base station, the radio resources used, the inter-base station interference coefficient, and the real-time performance of the base stations.

[0011] Then, using a real-time wireless resource allocation algorithm based on domain adversarial networks, the information of base stations with "excellent" performance is used as a reference to optimize the corresponding real-time wireless resources for base stations with "poor" performance; and the interference caused by resource changes is recalculated, and the allocated wireless resources are continuously updated through iteration until the number of "poor" base stations is minimized.

[0012] Furthermore, in the aforementioned multi-base station wireless network scenario:

[0013] Suppose there exists in the network The base station, of which the first The engineering parameters for each base station are: rated configuration resources. Including the rated bandwidth and maximum transmit power Center frequency and position coordinates ;No. Each base station during the time period The real-time parameters are: traffic load Base station performance indicators :when At that time, the base station performance was rated as "excellent" and marked as the source domain. ;on the contrary At that time, the base station performance was "poor" and it was marked as the target domain. Real-time resources This includes: real-time bandwidth utilization and real-time transmit power .

[0014] Furthermore, in the multi-base station wireless network scenario: the wireless resource allocation strategy aims to minimize the number of base stations with a "poor" performance index, and dynamically allocates the wireless resources of each base station based on the time-varying traffic load. The objective equation is expressed as:

[0015]

[0016] in As a constraint, the allocated real-time resources are guaranteed not to exceed the base station's rated configuration resources.

[0017] Furthermore, the network performance evaluator is a model based on an adaptive enhancement algorithm. According to the definition of channel capacity, the real-time performance index of a base station is determined by real-time traffic, the wireless resources used, and the interference coefficient between base stations, expressed as:

[0018]

[0019] Among them, for the first The set of base stations where interference exists is: The corresponding interference coefficient is:

[0020]

[0021] For the first The first base station and the first The distance between base stations is expressed as , This is the path loss index.

[0022] Furthermore, in the aforementioned real-time wireless resource allocation algorithm based on domain adversarial networks:

[0023] Domain adversarial neural networks consist of feature extraction layers Resource allocation layer and domain classification layer Composition, in which Input is all real-time data Mapped to high-dimensional features ; Input is Source domain data High-dimensional features Mapped to resource allocation ; The input is high-dimensional features of all real-time data. Mapped to performance tags ;

[0024] The objectives of domain adversarial networks include learning the resource allocation of "superior" base stations and the distribution alignment between different domains; for "superior" base stations, whose current resources can meet implementation requirements, the network objective is to learn these accurate resource allocations, i.e. ,in The network objective is represented as minimizing the following function:

[0025] ;

[0026] For base stations with "poor" performance, the network ensures accurate resource allocation by minimizing the distance between the high-dimensional feature distributions of the target domain and the source domain. When the sum of the probabilities of the discriminator correctly predicting the distribution decreases, the goal of effectively reducing the distance between distributions is expressed as maximizing the following function:

[0027] ;

[0028] Considering that the network needs to achieve two objectives simultaneously during training, therefore... and Introducing a gradient inversion layer During forward propagation, it is represented as , As input, it is represented as during backpropagation. , Let be the identity matrix; in summary, the loss function of the domain adversarial network is expressed as:

[0029]

[0030] in As a penalty for allocating resources beyond the allotted amount, i.e. when hour, The penalty is the smallest.

[0031] After one round of domain adversarial network training, real-time data is transmitted via... and reallocate resources This is used for a new round of iterations; the new resources will be used to recalculate the interference coefficients and evaluate network performance. After counting the total number of base stations with "poor" performance, the changed real-time data will be input into the domain adversarial network again, and resources will be reallocated again after one round of training; this iteration continues until... When the strategy converges to a fixed value, the iteration ends, and the current wireless resource is the optimal resource.

[0032] Compared to existing technologies, this invention and its preferred solution target multi-base station scenarios in wireless networks. Utilizing artificial intelligence algorithms, it addresses a core problem: how to rationally allocate corresponding wireless resources based on the real-time network traffic demands of each base station to ensure overall network quality. This algorithm combines theoretically analyzed inter-cell interference coefficients with big data on network traffic and performance. It proposes a domain adversarial network guided by a performance evaluator to achieve dynamic resource allocation based on real-time traffic, thereby minimizing the number of base stations with "poor" performance evaluations and reducing overall network wireless resource consumption. Attached Figure Description

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0034] Figure 1 This is a schematic diagram of a network scenario employing a wireless resource allocation strategy in a preferred embodiment of the present invention.

[0035] Figure 2 This is a structural diagram of the wireless network resource allocation algorithm based on domain adversarial networks in an embodiment of the present invention.

[0036] Figure 3 This is a performance diagram of network performance evaluators built using different machine learning models in embodiments of the present invention.

[0037] Figure 4 This is a schematic diagram comparing the resource allocation effect of the proposed algorithm in this embodiment of the invention with different benchmark algorithms.

[0038] Figure 5 This is a schematic diagram of the probability density function of resource allocation of the proposed algorithm in this embodiment of the invention compared with different benchmark algorithms. Detailed Implementation

[0039] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below for detailed explanation:

[0040] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0041] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0042] The embodiments of the present invention provide the following specific design process for a wireless network resource allocation strategy based on domain adversarial networks:

[0043] Step S1: Consider a multi-base station wireless network scenario; utilize big data information from network operation to set relevant parameters for each base station, including its engineering configuration parameters, real-time traffic and performance parameters.

[0044] Step S1 specifically includes the following steps:

[0045] Step S11: Set the base station parameters in the cellular network; assuming there are... The base station, of which the first The engineering parameters for each base station are: rated configuration resources. (Including the rated bandwidth) and maximum transmit power ), center frequency and position coordinates ;No. Each base station during the time period The real-time parameters are: traffic load Base station performance indicators (when At that time, the base station performance was rated as "excellent" and marked as the source domain. ;on the contrary At that time, the base station performance was "poor" and it was marked as the target domain. Real-time resources (Including real-time bandwidth) and real-time transmit power );

[0046] Step S12: Define the objective equation for wireless resource allocation; in this network, the wireless resource allocation strategy aims to minimize the number of base stations with a "poor" performance metric, and dynamically allocates wireless resources to each base station based on time-varying traffic load. Therefore, the objective equation is expressed as follows:

[0047]

[0048] in As a constraint, the allocated real-time resources are guaranteed not to exceed the base station's rated configuration resources.

[0049] Step S2: Construct a network performance evaluator; this evaluator uses the real-time traffic of each base station, the radio resources used, the inter-base station interference coefficient, and the real-time performance of the base stations to evaluate whether the real-time resources can meet the real-time traffic requirements.

[0050] Step S2 specifically includes the following steps:

[0051] Step S21: Calculate the interference coefficient between base stations in the wireless network; inter-cell interference usually only exists between base stations on the same frequency within a certain distance. For any given base station, the set of base stations that interfere with it is: Then its real-time interference coefficient can be calculated as

[0052]

[0053] in For the first The first base station and the first The distance between base stations is expressed as , The path loss index;

[0054] Step S22: Construct a network performance evaluator; based on the definition of channel capacity, assume that the real-time performance indicators of a base station are determined by real-time traffic, the radio resources used, and the interference coefficient between base stations, expressed as follows:

[0055]

[0056] By testing the performance evaluation results of different machine learning models, the adaptive augmentation algorithm achieved the best performance on the test set. Therefore, the model based on the adaptive augmentation algorithm will be used as a network performance evaluator.

[0057] Step S3: Construct a real-time wireless resource allocation algorithm based on a domain adversarial network; the domain adversarial network will use the base station information with "excellent" performance as a reference to optimize the real-time wireless resources for base stations with "poor" performance; at the same time, it will recalculate the interference caused by resource changes and continuously update the allocated wireless resources through iteration until the number of "poor" base stations is minimized.

[0058] Step S3 specifically includes the following steps:

[0059] Step S31: Construct a domain adversarial neural network; this network consists of a feature extraction layer. Resource allocation layer and domain classification layer Composition, in which Input is all real-time data Mapped to high-dimensional features ; Input is Source domain data High-dimensional features Mapped to allocate resources ; The input is high-dimensional features of all real-time data. Mapped to performance tags ;

[0060] Step S32: Design the loss function for the domain adversarial network; to satisfy the objective equation of resource allocation, namely, to reasonably allocate real-time resources to each base station to reduce the number of "poor" base stations, the objectives of the domain adversarial network include learning the resource allocation of "good" base stations and the distribution alignment between different domains. For "good" performance base stations, their current resources can meet the implementation requirements, and the network objective is to learn these accurate resource allocations, i.e. ,in Therefore, the network objective can be expressed as minimizing the following function:

[0061] ;

[0062] For base stations with "poor" performance, the network will ensure accurate resource allocation by minimizing the distance between the high-dimensional feature distributions of the target domain and the source domain. The difference distance between two distributions is closely related to the prediction results of the domain discriminator. That is, when the sum of the probabilities of correct predictions by the discriminator decreases, the distance between the distributions can be effectively reduced. Therefore, this objective can be expressed as maximizing the following function...

[0063] ;

[0064] Considering that the network needs to achieve two objectives simultaneously during training, therefore... and A gradient inversion layer was introduced between them. It is represented as during forward propagation. , As input, and represented as during backpropagation , Let be the identity matrix; in summary, the loss function of this network is expressed as:

[0065]

[0066] in As a penalty for allocating resources beyond the allotted amount, i.e. when hour, The penalty is the smallest.

[0067] Step S33: Construct an iterative resource allocation algorithm framework; after one round of domain adversarial network training, real-time data will be processed through... and reallocate resources This new data is then used for a new round of iterations; the new resources will be used to recalculate the interference coefficients and evaluate network performance. After counting the total number of base stations with "poor" performance, the changed real-time data will be input into the domain adversarial network again, and resources will be reallocated again after one round of training; this iteration continues until... When the strategy converges to a fixed value, the iteration ends, and the current wireless resource is the optimal resource.

[0068] The complete flow of the proposed algorithm can be obtained from the following pseudocode.

[0069]

[0070] To enable those skilled in the art to further understand the wireless network resource allocation strategy based on domain adversarial networks proposed in this invention, a detailed description is provided below with reference to specific embodiments. These embodiments are implemented based on the technical solution of this invention.

[0071] like Figure 1 The diagram shown is a schematic of a network scenario employing a wireless resource allocation strategy.

[0072] The model primarily consists of multiple independent base stations serving different users and devices, and a central controller that collects data and allocates resources via wireless links. Each base station has different engineering parameters, resulting in variations in its rated configuration resources, user equipment, and geographical location information. After collecting real-time data from each base station, the controller calculates the optimal resources through a deployed wireless resource allocation network, and then performs real-time resource scheduling for each base station.

[0073] like Figure 2 The diagram shown is a structural diagram of a wireless network resource allocation algorithm based on domain adversarial networks.

[0074] The algorithm is an iterative strategy, comprising a domain adversarial resource allocation network and a base station performance evaluator. After each round of resource allocation by the domain adversarial network, the interference coefficient is recalculated due to changes in real-time resources and input into the performance evaluator for classification. After each base station receives a new performance label, it is determined whether the number of base stations with "poor" performance meets the convergence condition. If not, resource allocation is performed again through the domain adversarial network. This process iterates until the condition is met, and then the optimal real-time resources are output.

[0075] like Figure 3 The figure shows a performance diagram of network performance evaluators built using different machine learning models.

[0076] To compare the performance of network performance evaluators built using different machine learning algorithms, ROC curves were plotted for each classifier on the same test dataset. The ROC curve is a graphical tool used to represent the performance of a classification model. It plots the true positive rate (the proportion of samples that were correctly identified as true on the horizontal axis) and the false positive rate (the proportion of samples that were correctly identified as false on the vertical axis). The area under the ROC curve, or AUC, measures the classifier's performance; the closer the AUC value is to 1, the better the classifier's performance. In the graph, the adaptive boosting classifier achieved the best performance (highest AUC value) compared to other classifiers such as random forests and decision trees. Therefore, this classifier was chosen as the network evaluator.

[0077] like Figure 4 The diagram shows the resource allocation performance of the proposed algorithm compared to different benchmark algorithms.

[0078] The experimental results illustrate the bandwidth allocation of a base station in a wireless network scenario over one week using different algorithms. The base station's traffic exhibits a clear periodicity, with 44 periods of poor network performance before allocation. The experiment demonstrates that the proposed algorithm, along with other benchmark algorithms, can allocate bandwidth according to the base station's real-time traffic load. Specifically, the domain adversarial network allocates less bandwidth during peak traffic periods and more during off-peak periods, resulting in a smoother bandwidth allocation. For example, during periods of sudden traffic surges, some benchmark algorithms allocate excessive bandwidth to meet network demands, while the domain adversarial network does not. After resource allocation, the proposed domain adversarial network reduced the number of periods of poor performance to 13, fewer than other algorithms and occurring only during a few peak traffic periods. In conclusion, the proposed allocation algorithm achieves optimal resource allocation performance while ensuring a smooth allocation, avoiding drastic changes in interference between different base stations, and thus enabling other adjacent base stations to achieve better performance.

[0079] like Figure 5 As shown, this is a schematic diagram of the probability density function of resource allocation by the proposed algorithm compared with different benchmark algorithms.

[0080] The experimental diagram illustrates the probability density function of bandwidth allocation for a base station in a cellular network scenario after applying different algorithms for wireless resource allocation. The base station's rated bandwidth is 20MHz. The experiment demonstrates that some benchmark algorithms allocate bandwidth similar to or larger than the initial resources to achieve performance improvements. The proposed allocation algorithm, however, avoids outputting excessively large or small bandwidths. This not only saves resource consumption but also implies that the strategy is more robust to extreme network conditions and less sensitive to network fluctuations. In conclusion, the proposed allocation algorithm can lead to a more stable network environment, resulting in better network performance.

[0081] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0083] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0084] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0086] This patent is not limited to the above-described preferred embodiments. Anyone can derive other forms of wireless network resource allocation methods based on domain adversarial networks under the guidance of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A method for allocating wireless network resources based on domain adversarial networks, characterized in that: In a multi-base station wireless network scenario, the base station parameters available for configuration include at least: engineering configuration parameters, real-time traffic and performance parameters; A network performance evaluator is used to assess whether real-time resources can meet the demands of real-time traffic by utilizing the real-time traffic of each base station, the radio resources used, the inter-base station interference coefficient, and the real-time performance of the base stations. Then, using a real-time wireless resource allocation algorithm based on domain adversarial networks, the information of base stations with "excellent" performance is used as a reference to optimize the corresponding real-time wireless resources for base stations with "poor" performance; and the interference caused by resource changes is recalculated, and the allocated wireless resources are continuously updated in an iterative manner until the number of "poor" base stations is minimized. In the multi-base station wireless network scenario: Suppose there exists in the network The base station, of which the first The engineering parameters for each base station are: rated configuration resources. Including the rated bandwidth and maximum transmit power Center frequency and position coordinates ;No. Each base station during the time period The real-time parameters are: traffic load Base station performance indicators :when At that time, the base station performance was "excellent" and it was marked as the source domain. ;on the contrary At that time, the base station performance was "poor" and it was marked as the target domain. Real-time resources This includes: real-time bandwidth utilization and real-time transmit power ; The network performance evaluator is a model based on an adaptive enhancement algorithm. According to the definition of channel capacity, the real-time performance index of a base station is determined by real-time traffic, the radio resources used, and the interference coefficient between base stations, expressed as: Among them, for the first The set of base stations where interference exists is: The corresponding interference coefficient is: For the first The first base station and the first The distance between base stations is expressed as , This is the path loss index.

2. The wireless network resource allocation method based on domain adversarial networks according to claim 1, characterized in that: In the multi-base station wireless network scenario: the wireless resource allocation strategy aims to minimize the number of base stations with a "poor" performance index, and dynamically allocates wireless resources to each base station based on time-varying traffic load. The objective equation is expressed as: in As a constraint, the allocated real-time resources are guaranteed not to exceed the base station's rated configuration resources.

3. The wireless network resource allocation method based on domain adversarial networks according to claim 2, characterized in that: In the aforementioned real-time wireless resource allocation algorithm based on domain adversarial networks: Domain adversarial neural networks consist of feature extraction layers Resource allocation layer and domain classification layer Composition, in which Input is all real-time data Mapped to high-dimensional features ; Input is Source domain data High-dimensional features Mapped to allocate resources ; The input is high-dimensional features of all real-time data. Mapped to performance tags ; The objectives of domain adversarial networks include learning the resource allocation of "superior" base stations and the distribution alignment between different domains; for "superior" base stations, their current resources can meet the implementation requirements, and the network objective is to learn these accurate resource allocations, i.e. ,in The network objective is represented as minimizing the following function: ； For base stations with "poor" performance, the network ensures accurate resource allocation by minimizing the distance between the high-dimensional feature distributions of the target domain and the source domain. When the sum of the probabilities of the discriminator correctly predicting the distribution decreases, the goal of effectively reducing the distance between distributions is expressed as maximizing the following function: ； Considering that the network needs to achieve two objectives simultaneously during training, therefore... and Introducing a gradient inversion layer During forward propagation, it is represented as , As input, it is represented as during backpropagation. , Let be the identity matrix; in summary, the loss function of the domain adversarial network is expressed as: in As a penalty for allocating resources beyond the allotted amount, i.e. when hour, The penalty is the smallest. After one round of domain adversarial network training, real-time data is transmitted via... and reallocate resources This is used for a new round of iterations; the new resources will be used to recalculate the interference coefficients and evaluate network performance. After counting the total number of base stations with "poor" performance, the changed real-time data will be input into the domain adversarial network again, and resources will be reallocated again after one round of training; this iteration continues until... When the strategy converges to a fixed value, the iteration ends, and the current wireless resource is the optimal resource.

Images