Method for selecting heterogeneous network based on deep learning for traffic transmission according to ToS in smart gateway, apparatus and computer program for performing the method
A deep learning-based method for network selection in smart gateways optimizes transmission efficiency by selecting appropriate networks for military communications, addressing inefficiencies in conventional methods by using a pre-trained model to ensure service type compliance.
Patent Information
- Application Number
- KR1020240192423
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2024-12-20
- Publication Date
- 2026-11-11
- Estimated Expiration
- 2044-12-20
AI Technical Summary
Conventional network transmission methods in military communications fail to optimize end-to-end service efficiency due to the lack of consideration for the varying states of heterogeneous networks connected to gateways, leading to inefficient data traffic handling.
A deep learning-based method for heterogeneous network selection in smart gateways that monitors network status and selects appropriate networks for data traffic based on service type using a pre-trained model, combining auto-encoders and deep neural networks to optimize load balancing.
Enhances transmission efficiency by ensuring data is routed through networks that meet service type requirements, thereby improving overall service performance.
Smart Images

Figure 112024141821647-PAT00002_ABST
Abstract
Description
Technology Field
[0001] This document relates to a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway, a device for performing the same, and a computer program. Background Technology
[0002] In military communications, Software Defined Radio (SDR) gateways are connected to the combat radio network as a sub-network and to multiple networks (i.e., infrastructure, public, and satellite networks) as a backbone network to transmit data traffic. While the network environments required for military communications services may vary depending on the Type of Service (ToS), conventional technology transmits data traffic without considering the state of the heterogeneous networks connected to the gateway. This can lead to a problem where end-to-end service transmission efficiency is reduced when the system fails to meet the transmission requirements for each ToS due to changes in the state of the heterogeneous networks connected to the gateway. Although transport layer protocols are generally used to control traffic load on the network, this does not significantly improve service transmission efficiency. Therefore, to improve end-to-end service transmission efficiency at gateways connected to multiple backbone networks (heterogeneous networks), it is necessary to recognize the state of the heterogeneous networks connected to the gateway in advance and select and utilize the appropriate network for transmitting traffic for each ToS. The problem to be solved
[0003] According to one embodiment of the present document, a deep learning-based heterogeneous network selection method for transmitting traffic by Type of Service (ToS) in a smart gateway, which recognizes the status of the heterogeneous network connected to the gateway in advance and selects the network for transmitting traffic by Type of Service (ToS) in order to improve end-to-end service transmission efficiency in a gateway connected to multiple backbone networks (heterogeneous networks), and a device and computer program for performing the same may be provided.
[0004] The problems to be solved in this disclosure are not limited to those mentioned above and can be extended in various ways without departing from the spirit and scope of this disclosure. means of solving the problem
[0005] A deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway according to one embodiment of the present document comprises: a step of acquiring a data flow input from a mobile network; a step of monitoring status information of a plurality of heterogeneous networks; and a step of selecting the heterogeneous network to which the data flow will be serviced using a deep learning model that has been pre-trained and built.
[0006] The heterogeneous network selection step described above may be performed by using the deep learning model to determine which of the plurality of heterogeneous networks the data flow is suitable to be serviced in, and by distributing the data flow to the selected heterogeneous network based on the determination result to perform load balancing.
[0007] The above data flow acquisition step may be performed by acquiring an extraction value including an IP, a Type of Service (ToS), and a keyword based on the data flow input from the above mobile network.
[0008] The above monitoring step may be performed by monitoring the status information of a plurality of heterogeneous networks using report messages periodically transmitted from the heterogeneous network.
[0009] The heterogeneous network selection step may be performed by inputting a service type (ToS) based on the extraction value obtained from the data flow and state information of a plurality of heterogeneous networks into the deep learning model, and distributing the data flow to the selected heterogeneous network based on the probability value for each of the plurality of heterogeneous networks which is the output of the deep learning model and the extraction value obtained from the data flow.
[0010] The above deep learning model can be built by being trained in advance based on a training dataset consisting of data fields of service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), various heterogeneous network delays between gateways (HetNet Delay), and service availability (Able / Unable), taking service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), and various heterogeneous network delays between gateways (HetNet Delay) as inputs, and service availability (Able / Unable) as outputs.
[0011] The above deep learning model may be composed of a structure combining an auto-encoder and a deep neural network (DNN).
[0012] In the above deep learning model, 4-dimensional network data consisting of a Service Type (ToS), HetNet Kind, HetNet Background Traffic excluding data flow from the HetNet network, and HetNet Delay between gateways is input to the auto-encoder, and the 4-dimensional network data passes through the encoder layer to generate a 2-dimensional data distribution in the latent space, and the 2-dimensional data can be used as input data for the deep neural network (DNN).
[0014] A computer program according to one embodiment of the present document is stored on a computer-readable storage medium and executes on a computer any one of the deep learning-based heterogeneous network selection methods for traffic transmission by service type in the smart gateway described above.
[0016] An apparatus according to one embodiment of the present document comprises: a memory storing one or more programs for selecting a network suitable for transmitting traffic by Type of Service (ToS) using deep learning based on the state of a heterogeneous network; and one or more processors that perform operations for selecting a network suitable for transmitting traffic by Type of Service (ToS) using deep learning based on the state of the heterogeneous network according to the one or more programs stored in the memory; wherein the processors acquire a data flow input from a startup network, monitor state information of a plurality of the heterogeneous networks, and select the heterogeneous network to which the data flow will be serviced using a deep learning model that has been pre-trained and built.
[0017] The processor can use the deep learning model to determine which of the heterogeneous networks the data flow is suitable to be serviced, and perform load balancing by distributing the data flow to the selected heterogeneous network based on the determination result.
[0018] The above processor can obtain an extraction value including an IP, a service type (ToS), and a keyword based on the data flow input from the above startup network.
[0019] The processor can monitor status information of a plurality of heterogeneous networks by using report messages periodically transmitted from the heterogeneous networks.
[0020] The processor inputs a service type (ToS) based on the extraction value obtained from the data flow and state information of a plurality of heterogeneous networks into the deep learning model, and distributes the data flow to the selected heterogeneous network based on the probability value for each of the plurality of heterogeneous networks, which is the output of the deep learning model, and the extraction value obtained from the data flow.
[0021] The above deep learning model can be built by being trained in advance based on a training dataset consisting of data fields of service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), various heterogeneous network delays between gateways (HetNet Delay), and service availability (Able / Unable), taking service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), and various heterogeneous network delays between gateways (HetNet Delay) as inputs, and service availability (Able / Unable) as outputs.
[0022] The above deep learning model may be composed of a structure combining an auto-encoder and a deep neural network (DNN).
[0023] In the above deep learning model, 4-dimensional network data consisting of a Service Type (ToS), HetNet Kind, HetNet Background Traffic excluding data flow from the HetNet network, and HetNet Delay between gateways is input to the auto-encoder, and the 4-dimensional network data passes through the encoder layer to generate a 2-dimensional data distribution in the latent space, and the 2-dimensional data can be used as input data for the deep neural network (DNN). Effects of the invention
[0024] According to one embodiment of the present document, the transmission efficiency of a service can be increased by enabling transmission while avoiding communication networks that do not satisfy the Type of Service (ToS) transmission requirements.
[0025] The effects according to the various embodiments of this document are not limited to those described above, and it is obvious to those skilled in the art that various effects are inherent in this disclosure. Brief explanation of the drawing
[0026] FIG. 1 is a block diagram for explaining an apparatus according to one embodiment of the present document. FIG. 2 is a flowchart illustrating a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway according to one embodiment of the present document. FIG. 3 is a diagram illustrating the functional structure of a gateway equipped with a device for performing a deep learning-based heterogeneous network selection method according to one embodiment of the present document. FIG. 4 is a diagram illustrating the structure of a deep learning model according to one embodiment of the present document. Figure 5 is a diagram illustrating the structure of a learning algorithm for a deep learning model for a base network and a public network according to one embodiment of the present document. Figure 6 is a diagram illustrating the structure of a learning algorithm for a deep learning model for a satellite network according to one embodiment of the present document. FIG. 7 is a diagram illustrating the learning results for state recognition of a heterogeneous network according to one embodiment of the present document, and shows a base network. FIG. 8 is a diagram illustrating the learning results for recognizing the state of a heterogeneous network according to one embodiment of the present document, and shows a public network. FIG. 9 is a diagram illustrating the learning results for recognizing the state of a heterogeneous network according to one embodiment of the present document, and shows a satellite network. Specific details for implementing the invention
[0027] Hereinafter, embodiments of this document will be described in detail with reference to the attached drawings. The advantages and features of the embodiments of this document, and the methods for achieving them, will become clear by referring to the details described below in conjunction with the attached drawings. However, the embodiments of this document are not limited to those disclosed below but can be implemented in various different forms, and the embodiments of this document are defined only by the scope of the claims.
[0028] Throughout the specification, the same reference numerals refer to the same components. Unless otherwise defined, all terms used in this specification (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the embodiments of this document pertain. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.
[0029] In this specification, terms such as "first," "second," etc. are used to distinguish one component from another, and the scope of rights shall not be limited by these terms. For example, the first component may be named the second component, and similarly, the second component may be named the first component.
[0030] In this specification, identification symbols (e.g., a, b, c, etc.) for each step are used for convenience of explanation and do not indicate the order of the steps; the steps may occur differently from the specified order unless the context clearly indicates a specific order. That is, the steps may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.
[0031] In this specification, expressions such as “have,” “may have,” “include,” or “may include” refer to the existence of the relevant feature (e.g., a numerical value, function, operation, or component, etc.) and do not exclude the existence of additional features.
[0033] Hereinafter, with reference to the attached drawings, a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway according to one embodiment of the present document, a device for performing the same, and a computer program are described in detail.
[0035] First, an apparatus according to one embodiment of the present document will be described with reference to FIG. 1.
[0036] FIG. 1 is a block diagram for explaining an apparatus according to one embodiment of the present document.
[0037] Referring to FIG. 1, a device (100) according to one embodiment of the present document can recognize the status of a heterogeneous network connected to a gateway in advance and select a network for traffic transmission by Type of Service (ToS) in order to improve end-to-end service transmission efficiency at a gateway connected to multiple backbone networks (heterogeneous networks). Here, the device (100) is installed in a gateway and can perform a deep learning-based heterogeneous network selection method for traffic transmission by Type of Service.
[0038] Accordingly, the device (100) can increase the transmission efficiency of the service by avoiding communication networks that do not satisfy the service type (ToS) transmission requirements.
[0039] To this end, the device (100) may include one or more processors (110), a computer-readable storage medium (130), and a communication bus (150).
[0040] The processor (110) can control the device (100) to operate. For example, the processor (110) can execute one or more programs (131) stored in a computer-readable storage medium (130). One or more programs (131) may include one or more computer-executable instructions, and the computer-executable instructions may be configured to cause the device (100) to perform an operation to select a network suitable for traffic transmission by service type (ToS) using deep learning based on the state of a heterogeneous network when executed by the processor (110).
[0041] A computer-readable storage medium (130) is configured to store computer-executable instructions or program code, program data and / or other suitable forms of information for selecting a network suitable for traffic transmission by service type (ToS) using deep learning based on the state of a heterogeneous network. A program (131) stored in the computer-readable storage medium (130) includes a set of instructions executable by a processor (110). In one embodiment, the computer-readable storage medium (130) may be a memory (volatile memory such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, other forms of storage media that are accessed by the device (100) and capable of storing desired information, or a suitable combination thereof.
[0042] The communication bus (150) interconnects various other components of the device (100), including the processor (110) and the computer-readable storage medium (130).
[0043] The device (100) may also include one or more input / output interfaces (170) and one or more communication interfaces (190) that provide interfaces for one or more input / output devices. The input / output interfaces (170) and communication interfaces (190) are connected to a communication bus (150). An input / output device (not shown) may be connected to other components of the device (100) through the input / output interface (170).
[0045] Then, with reference to FIG. 2, a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway according to one embodiment of the present document will be described.
[0046] FIG. 2 is a flowchart illustrating a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway according to one embodiment of the present document.
[0047] Referring to FIG. 2, the processor (110) of the device (100) can obtain a data flow input from the startup network (S110).
[0048] That is, the processor (110) can obtain an extraction value including an IP, a Type of Service (ToS), and a keyword based on a data flow input from the startup network.
[0049] Then, the processor (110) can monitor the status information of multiple heterogeneous networks (S120).
[0050] Here, multiple heterogeneous networks may include a base network, a public network, and a satellite network.
[0051] At this time, the processor (110) can monitor the status information of multiple heterogeneous networks by using report messages periodically transmitted from each heterogeneous network.
[0052] Afterwards, the processor (110) can select a heterogeneous network to which the data flow will be serviced using a deep learning model that has been trained and built in advance (S130).
[0053] That is, the processor (110) can use a deep learning model to determine which heterogeneous network among a plurality of heterogeneous networks is suitable for service of the data flow, and based on the determination result, distribute the data flow to the selected heterogeneous network to perform load balancing.
[0054] To explain in more detail, the processor (110) inputs the service type (ToS) based on the extraction value obtained from the data flow and the state information of a plurality of heterogeneous networks into a deep learning model, and can distribute the data flow to a selected heterogeneous network based on the probability value for each of the plurality of heterogeneous networks, which is the output of the deep learning model, and the extraction value obtained from the data flow.
[0055] Here, the deep learning model can be built by being trained in advance based on a training dataset consisting of data fields of service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), various heterogeneous network delays between gateways (HetNet Delay), and service availability (Able / unable), taking service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), and various heterogeneous network delays between gateways (HetNet Delay) as inputs, and service availability (Able / unable) as outputs.
[0056] In addition, the deep learning model can be structured by combining an auto-encoder and a deep neural network (DNN). That is, the deep learning model is input into an auto-encoder 4-dimensional network data consisting of a Service Type (ToS), HetNet Kind, HetNet Background Traffic (excluding data flow from the HetNet), and HetNet Delay between gateways. The 4-dimensional network data passes through the encoder layer to generate a 2-dimensional data distribution in the latent space, and the 2-dimensional data can be used as input data for the deep neural network (DNN).
[0057] At this time, deep learning models are constructed for each of the base network, public network, and satellite network, and the Service Type (ToS) based on the extracted value obtained from the data flow and the status information of multiple heterogeneous networks are input to each of the deep learning models for the base network, the deep learning model for the public network, and the deep learning model for the satellite network, and a probability value (service availability) for the base network of a specific data flow is output from the deep learning model for the base network, a probability value (service availability) for the public network of a specific data flow is output from the deep learning model for the public network, and a probability value (service availability) for the satellite network of a specific data flow may be output from the deep learning model for the satellite network.
[0058] Meanwhile, although FIG. 2 illustrates that the monitoring step (S120) is performed after the execution of the data flow acquisition step (S110), this is merely an example, and depending on the embodiment, the monitoring step (S120) may be performed before the execution of the data flow acquisition step (S110), or the monitoring step (S120) may be performed simultaneously (in parallel) with the data flow acquisition step (S110).
[0060] Then, with reference to FIGS. 3 to 9, an example of a deep learning-based heterogeneous network selection system for traffic transmission by service type according to one embodiment of the present document will be described.
[0061] FIG. 3 is a diagram illustrating the functional structure of a gateway equipped with a device for performing a deep learning-based heterogeneous network selection method according to one embodiment of the present document.
[0062] Referring to FIG. 3, a smart gateway equipped with a device (100) that performs a deep learning-based heterogeneous network selection method according to one embodiment of the present document may be functionally composed of a total of four modules (Input Manager module, Prob module, AI Model module and Data Flow Manager module).
[0063] The Input Manager module may be a module that recognizes and manages data flow input from the startup network.
[0064] - Extract three values from the data flow: IP, ToS (Type of Service), and Keyword
[0065] - For keywords, Start and Stop are used to denote the beginning and end of a specific data flow.
[0066] -> When a new data flow is input, generate a Start keyword
[0067] -> If packets are not delivered for a specific Data Flow for a certain period, generate the Stop keyword
[0068] - Output: Send the IP, ToS, and Keyword of the extracted Data Flow to the Data Flow Manager module using UDP socket communication.
[0069] Also, the Prob module can be a module that monitors the state information of various heterogeneous networks that must be input into the AI Model.
[0070] - Collect network status information by utilizing periodically transmitted report messages
[0071] - Input: "Heterogeneous network status information via Prob operation" and "Heterogeneous network interface monitoring values"
[0072] - Output: Passes various network information to the AI Model for use in model computation.
[0073] In addition, the AI Model module may be a module that determines which network is suitable for a specific Data Flow to be serviced.
[0074] - A structure that loads and applies weight values (in the form of a Saved Model) learned in advance through a training server.
[0075] - Input: "Data Flow Service Type (ToS)" and "Network Status Monitoring Value (Prob Data)"
[0076] - Output: The most suitable heterogeneous network probability value for data flow transmission
[0077] In addition, the Data Flow Manager module may be a module that achieves load balancing by distributing Data Flow based on the results of AI Model operations.
[0078] - Input: "Data Flow (IP, ToS)" and "AI Model Decision Result"
[0079] - Output: Distribute multiple data flows to a specific heterogeneous network based on AI model results
[0080] In addition, the data transmission format for each module, that is, the data format in which the In / Out of each module is defined, is as shown in [Table 1] below.
[0081] Interaction Module data form Input Manager→Data Flow Manager DataFlow "IP:ToS:Keyword" - Keyword signifies the start and end of a specific data flow - The colon (:), IP, ToS, and Keyword are all of type String consisting of characters - Ex) "10.10.100.2,0x04,start" Data Flow Manager → AI Model Request Socket connection complete message AI Model → Data Flow Manager Result "HetNet[n]" : { "ToS": Probability value of Able} - In a specific heterogeneous network, it returns the probability value that a specific service can be properly provided - The data is in JSON format - Ex)"HetNet[1]" : { "0x04" : 0.96 "0x08" : 0.21 … "0x1C" : 0.87},"HetNet[2]" : { "0x04" : 0.77 "0x08" : 0.98 … "0x1C" : 0.33},"HetNet[3]" : { "0x04": 0.99 "0x08": 0.97 … "0x1C": 0.55}
[0082] In addition, the final data fields used to build the training dataset of the AI Model to achieve load balancing are as shown in [Table 2] below.
[0083] AI Data ToS HetNet Kind HetNet Background Traffic HetNet Delay Able or Unable Learn ○ ○ ○ ○ ○ Input ○ ○ ○ ○ × Output × × × × ○
[0084] - Construct a training dataset for AI model learning using data fields such as ToS, HetNet Kind (type of heterogeneous network), HetNet Background Traffic (traffic excluding data flow from heterogeneous networks), HetNet Delay (delay between various heterogeneous networks and gateways), and Able or Unable (service availability).
[0085] - Use ToS, HetNet Kind, HetNet Background Traffic, and HetNet Delay values as input to the AI model, and output probability values for Able / Unable.
[0086] - Transmits the probability value of the Enable status for each ToS to the Data Flow Manager module, and the Data Flow Manager module performs traffic control for heterogeneous networks based on this value.
[0087] - AI model training can be performed in the AI Model module, or if the gateway system lacks the computing power for model training, the model can be trained on a remote learning server, and the resulting weight file (Saved Model) can be transferred to the AI Model module for use in traffic load balancing.
[0089] FIG. 4 is a diagram illustrating the structure of a deep learning model according to one embodiment of the present document, FIG. 5 is a diagram illustrating the structure of a learning algorithm for a deep learning model for a base network and a public network according to one embodiment of the present document, and FIG. 6 is a diagram illustrating the structure of a learning algorithm for a deep learning model for a satellite network according to one embodiment of the present document.
[0090] Referring to FIGS. 4 to 6, a deep learning model (AI Model) according to one embodiment of the present document may be composed of a structure combining an Auto-Encoder and a DNN.
[0091] 4-dimensional network data input to the Auto-Encoder passes through the Encoder layer to generate a 2-dimensional data distribution in the Latent Space, and this 2-dimensional data is used as input data for the DNN model.
[0092] - Auto-Encoder structure
[0093] -> Input Data = hetnet_kind, hetnet_Background Traffic, hetnet_delay, ToS
[0094] -> Latent dim (latent space dimension) = 2
[0095] -> Intermediate dim(number of neurons in the intermediate layer) = 128, Activation Function = ReLU
[0096] -> Optimizer = Adam, Loss = MSE
[0097] -> Batch size = 64, Epochs = 100
[0098] - DNN structure
[0099] The number of nodes in each layer of the model is designed as 32, 32, 16, and 1 for the public and infrastructure networks, and 64, 64, 32, and 1 for the satellite network.
[0100] -> Use ReLU for the Hidden Layer's Activation Function and Sigmoid for the Output Layer. Apply Dropout (0.5) between layers to prevent overfitting of the training model.
[0101] -> Train 3 identical networks with 200 epochs and a batch size of 64
[0102] -> Use Binary Cross-entropy as the loss function, Adam as the optimizer for optimization, and a learning rate of 0.001.
[0104] FIG. 7 is a diagram illustrating the learning results for state recognition of a heterogeneous network according to one embodiment of the present document, showing a base network; FIG. 8 is a diagram illustrating the learning results for state recognition of a heterogeneous network according to one embodiment of the present document, showing a public network; and FIG. 9 is a diagram illustrating the learning results for state recognition of a heterogeneous network according to one embodiment of the present document, showing a satellite network.
[0105] A smart gateway equipped with a device (100) that performs a deep learning-based heterogeneous network selection method according to one embodiment of the present document derives the probability of transmission possibility for each service type (ToS) through deep learning based on the state of the heterogeneous network connected to the gateway, and by selecting and transmitting a heterogeneous network according to the transmission possibility probability for each service type (ToS), it can recognize the heterogeneous network environment capable of transmission for each service type (ToS) and attempt to transmit traffic. Through this, it is possible to transmit by avoiding communication networks that do not satisfy the transmission requirements of the service type (ToS), thereby increasing the transmission efficiency of the service.
[0106] Referring to FIGS. 7 to 9, a high state recognition rate can be confirmed through the learning results of heterogeneous network state recognition, and through this, a smart gateway equipped with a device (100) according to one embodiment of the present document can increase service transmission efficiency by service type (ToS).
[0108] The operation according to the embodiments of this document described above may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable storage medium. A computer-readable storage medium refers to any medium that participates in providing instructions to a processor for execution. A computer-readable storage medium may include program instructions, data files, data structures, or a combination thereof. Examples include magnetic media, optical storage media, memory, etc. Computer programs may be distributed over networked computer systems, and computer-readable code may be stored and executed in a distributed manner. Functional programs, code, and code segments for implementing the embodiments of this document will be readily deducible by programmers in the art to which the embodiments of this document belong.
[0109] The embodiments of this document are intended to illustrate technical concepts, and the scope of the technical concepts of the embodiments of this document is not limited by these embodiments. The scope of protection of the embodiments of this document shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the embodiments of this document. Explanation of the symbols
[0110] 100 : Device, 110 : Processor, 130 : Computer-readable storage media, 131 : Program, 150 : Communication bus, 170 : Input / Output Interface, 190: Communication interface
Claims
Claim 1 A step of acquiring a data flow input from a mobile network; a step of monitoring status information of multiple heterogeneous networks; and a step of selecting the heterogeneous network to be serviced by the data flow using a pre-trained and constructed deep learning model; wherein the heterogeneous network selection step comprises determining, using the deep learning model, which of the plurality of heterogeneous networks is suitable for the data flow to be serviced, and performing load balancing by distributing the data flow to the selected heterogeneous network based on the determination result; and wherein the deep learning model takes the service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow flowing in the heterogeneous network (HetNet Background Traffic), various heterogeneous network delays between gateways (HetNet Delay), and service availability (Able / Unable) as inputs and outputs service availability (Able / Unable) based on a training dataset consisting of data fields of service type (ToS), heterogeneous network type (HetNet Kind), traffic excluding data flow flowing in the heterogeneous network (HetNet Background Traffic), and various heterogeneous network delays between gateways (HetNet Delay). Thus, a deep learning-based heterogeneous network selection method for service-type traffic transmission in a smart gateway, which is built by prior learning and consists of a structure combining an Auto-Encoder and a Deep Neural Network (DNN). Claim 2 delete Claim 3 A deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway, wherein the data flow acquisition step comprises acquiring an extraction value including an IP, a Type of Service (ToS), and a keyword based on the data flow input from the operating network. Claim 4 In claim 3, the monitoring step comprises monitoring status information of a plurality of heterogeneous networks using report messages periodically transmitted from the heterogeneous networks, a deep learning-based heterogeneous network selection method for service-type traffic transmission in a smart gateway. Claim 5 In claim 4, the heterogeneous network selection step comprises inputting a service type (ToS) based on the extraction value obtained from the data flow and state information of a plurality of heterogeneous networks into the deep learning model, and distributing the data flow to the selected heterogeneous network based on the probability value for each of the plurality of heterogeneous networks which is the output of the deep learning model and the extraction value obtained from the data flow, a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway. Claim 6 delete Claim 7 delete Claim 8 In claim 1, the deep learning model is a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway, wherein 4-dimensional network data consisting of a service type (ToS), a heterogeneous network type (HetNet Kind), remaining traffic excluding data flow in the heterogeneous network (HetNet Background Traffic), and various heterogeneous network delays between gateways (HetNet Delay) is input to the auto-encoder, the 4-dimensional network data passes through the encoder layer to generate a 2-dimensional data distribution in the latent space (Latent Space), and the 2-dimensional data is used as input data for the deep neural network (DNN). Claim 9 A computer program stored on a computer-readable storage medium for executing on a computer a deep learning-based heterogeneous network selection method for traffic transmission by service type in a smart gateway described in any one of paragraphs 1, 3 through 5, and 8. Claim 10 Memory for storing one or more programs for selecting a network suitable for traffic transmission by Type of Service (ToS) using deep learning based on the state of a heterogeneous network; and one or more processors that perform operations to select a network suitable for traffic transmission by Service Type (ToS) based on the state of the heterogeneous network using deep learning according to one or more programs stored in the memory; wherein the processor acquires a data flow input from a startup network, monitors state information of a plurality of the heterogeneous networks, and selects the heterogeneous network to be serviced by the data flow using a pre-trained and constructed deep learning model; wherein the processor determines which of the plurality of the heterogeneous networks is suitable for service of the data flow using the deep learning model, and performs load balancing by distributing the data flow to the selected heterogeneous network based on the determination result; and wherein the deep learning model comprises training data consisting of data fields of Service Type (ToS), HetNet Kind, remaining traffic excluding data flow flowing from the heterogeneous network (HetNet Background Traffic), various heterogeneous network delays between gateways (HetNet Delay), and service availability (Able / Unable). A device that is built by prior training based on a set, taking Service Type (ToS), HetNet Kind, HetNet Background Traffic (excluding data flow from the heterogeneous network), and HetNet Delay between gateways as inputs, and Service Availability (Able / Unable) as outputs, and is composed of a structure combining an Auto-Encoder and a Deep Neural Network (DNN). Claim 11 delete Claim 12 In claim 10, the processor is a device that obtains an extraction value including an IP, a service type (ToS), and a keyword based on the data flow input from the startup network. Claim 13 In paragraph 12, the processor is a device that monitors status information of a plurality of heterogeneous networks by using a report message periodically transmitted from the heterogeneous network. Claim 14 In claim 13, the device comprises a processor that inputs a service type (ToS) based on the extraction value obtained from the data flow and state information of a plurality of heterogeneous networks into a deep learning model, and distributes the data flow to a selected heterogeneous network based on a probability value for each of the plurality of heterogeneous networks which is the output of the deep learning model and the extraction value obtained from the data flow. Claim 15 delete Claim 16 delete Claim 17 In claim 10, the deep learning model is a device in which 4-dimensional network data, consisting of a Service Type (ToS), a HetNet Kind, remaining traffic excluding data flow from the HetNet, and various HetNet Delays between gateways, is input to the auto-encoder, the 4-dimensional network data passes through the encoder layer to generate a 2-dimensional data distribution in the latent space, and the 2-dimensional data is used as input data for the deep neural network (DNN).
Citation Information
Patent Citations
Heterogeneous network switching method based on SDN and SDR
CN106851757A
Method and apparatus for causing information related to mutual communication property between a terminal and access points of a wireless network to be used in accessing the wireless network
KR1020130118673A
