KR20230094922A - Intelligent dynamic real-time spectrum resource management system and intelligent dynamic real-time spectrum resource management method using data mining and case-based reasoning - Google Patents

Abstract

The present invention provides an intelligent dynamic real-time spectrum resource management method in an intelligent dynamic real-time spectrum resource management system, comprising the steps of, by a CR master node, collecting and pre-processing history data as raw data; CR master node, generating a behavior pattern prediction model; The CR master node analyzes input cases and generates result data; CR user node determining whether to perform optimization; generating, by a CR user node, a transmission parameter that satisfies an interference probability criterion with an existing user within a channel, if it is decided to perform the optimization in the determining step; and controlling, by the CR user node, transmission power based on the transmission parameter. This prevents system performance from degrading due to incorrect prediction when there are misclassified cases due to irregular patterns through an accurate learning algorithm for each case data, reduces the complexity of intelligent dynamic real-time spectrum resource management, and enables rapid resource allocation. can make it possible.

Description

Intelligent dynamic real-time spectrum resource management system and intelligent dynamic real-time spectrum resource management method using data mining and case-based reasoning REASONING}

The present invention relates to an intelligent dynamic real-time spectrum resource management system and an intelligent dynamic real-time spectrum resource management method using data mining and case-based reasoning, and more particularly, to an intelligent dynamic real-time spectrum resource management method using data mining and case-based reasoning that enables rapid resource allocation. A real-time spectrum resource management system and an intelligent dynamic real-time spectrum resource management method.

Recently, with the development of the Internet of Things (IoT), a paradigm shift in computing technology is rapidly occurring, and accordingly, technologies that can efficiently process large amounts of data generated from various industrial IoT devices are emerging. According to International Data Corporation, by 2025 IoT devices connected to the Internet will generate data approaching 79.4 zettabytes, predicting that there will be as many as 44.1 billion of these devices. It is predicted that significant changes will be made to major end terminals.

Meanwhile, in the past, such vast amounts of data are processed in a centralized data center or an external cloud system, but these existing centralized computing systems provide various IoT services such as smart factories, smart farms, and self-driving cars that require real-time processing. There is a limit to accepting . Therefore, due to the exponentially increasing data volume and the risk of cloud server overload based on network traffic, a computing shift from the existing centralized computing system to the network edge within the wireless access network is expected to become a new paradigm. Edge computing is a distributed computing system that brings operations and data storage closer to the data source. As computing services are processed close to the terminal used by users, faster and more stable services can be obtained and flexible hybrid cloud computing can be performed. has the advantage of

However, the limited frequency resources of edge computing cause frequent frequency handoff in a dynamic spectrum environment that requires MTD operation in real time, and serious performance degradation may occur due to interference with other peer networks. Therefore, there is a need to dynamically optimize the distribution of frequency resources according to time-varying parameters such as channel state information.

To this end, in conventional intelligent dynamic spectrum resource management, for timely handoff to an idle channel, a learning engine generates a backup channel list based on historical data representing statistics of an existing local network, and an inference engine uses the backup channel list. Therefore, spectrum handoff was performed to another idle channel to enable continuous communication without disturbing existing users. At this time, when interference affects existing users existing in adjacent channels and locations, transmission parameters that can coexist between adjacent networks are dynamically reconstructed through an optimization engine using interference analysis and genetic algorithms.

However, decision-making based on such an optimization algorithm increases computation time and complexity, making it difficult to quickly allocate resources, and wastes valuable solution information optimized according to the evolution of spectral states between space-time and frequency domains.

Therefore, there is a need to develop an intelligent dynamic real-time spectrum resource management method capable of reducing the complexity of intelligent dynamic spectrum resource management and achieving rapid resource allocation.

Korean Patent Publication No. 10-2002-0030821

The present invention has been made to solve the above problems, and an object of the present invention is to reduce system performance due to incorrect prediction when there are misclassified cases due to irregular patterns using an accurate learning algorithm for each case data. To provide an intelligent dynamic real-time spectrum resource management system and an intelligent dynamic real-time spectrum resource management method using data mining and case-based reasoning that prevent the complexity of intelligent dynamic real-time spectrum resource management and enable rapid resource allocation.

Intelligent dynamic real-time spectrum in an intelligent dynamic real-time spectrum resource management system including a CR master node, at least one CR user node, and a case database constituting a CR network according to an embodiment of the present invention for achieving the above object The resource management method may include, by the CR master node, collecting and pre-processing history data including sensing data obtained by sensing a channel state of the CR network at the CR user node as raw data; learning, by the CR master node, a behavior pattern of the user of the CR network using the preprocessed data, and generating a behavior pattern prediction model; generating result data by analyzing input cases based on the generated behavior pattern prediction model by the CR master node; determining, by the CR user node, whether to perform optimization based on the result data; generating, by the CR user node, a transmission parameter that satisfies an interference probability criterion with an existing user within a channel, if the determining step determines to perform the optimization; and controlling, by the user node, transmission power based on the transmission parameter.

In addition, the generating of the resulting data may include verifying a similarity by comparing a predicted value predicted through the behavior pattern prediction model with an actual value of a case in which the CR master node is input; generating, by the CR master node, result data based on the similarity verification result; and transmitting, by the CR master node, generated result data to the CR user node.

In addition, the result data includes at least one of a previous solution, which is a solution for a previous case stored in the case database, and a backup channel list, which is a list of idle channels not used by a user at a specific time and location in the CR network, , In the step of generating the result data based on the similarity verification result, if the predicted value and the actual value are the same, the CR master node generates the result data so that both the previous solution and the backup channel list are included, If the predicted value and the actual value are not the same, the resulting data may be generated to include only the backup channel list.

In addition, the step of determining whether to perform the optimization may include: performing solution verification, which is a verifier for the solution received by the CR user node, when receiving result data including the solution; if the CR user node satisfies the interference probability criterion as a result of the solution verification, determining whether to perform the optimization and maintaining the received solution; and determining whether to perform the optimization if the CR user node does not satisfy the interference probability criterion as a result of the solution verification.

Also, in the step of performing the solution verification, the solution verification may be performed using interference analysis based on a Monte Carlo algorithm.

In addition, the determining whether to perform the optimization may include determining whether to perform the optimization when result data including only the backup channel list is received.

The method may further include allowing the CR user node to update the case database by matching the generated transmission parameter with a solution for the input case.

In the preprocessing step, the raw data may be preprocessed using at least one of data cleaning, data reduction, outlier detection, and data scaling.

Meanwhile, in an intelligent dynamic real-time spectrum resource management system including a CR master node constituting a CR network, at least one CR user node, and a case database according to another embodiment of the present invention for achieving the above object, the CR The master node includes a pre-processing unit that collects and pre-processes history data including sensing data obtained by sensing channel conditions of the CR network at the CR user node as raw data; a learning unit for learning a behavioral pattern of the CR network user using the preprocessed data and generating a behavioral pattern prediction model; and a verification unit generating result data by analyzing an input case based on the generated behavior pattern prediction model, wherein the CR user node includes: an interference analysis unit determining whether optimization is performed based on the result data; and an optimization unit generating transmission parameters that satisfy an interference probability criterion with an existing user within a channel when it is determined to perform the optimization, and controlling transmission power based on the transmission parameters.

In addition, the verification unit compares the predicted value predicted through the behavior pattern prediction model with the actual value of the input case to verify similarity, generates the result data based on the similarity verification result, and generates the result data It can be passed to the CR user node.

In addition, the result data includes at least one of a previous solution, which is a solution for a previous case stored in the case database, and a backup channel list, which is a list of idle channels not used by a user at a specific time and location in the CR network, , The verification unit, if the predicted value and the actual value are identical, the CR master node generates the result data so that both the previous solution and the backup channel list are included, and if the predicted value and the actual value are not identical, the backup The result data may be generated to include only the channel list.

In addition, when the interference analysis unit receives the result data including the solution, it performs solution verification, which is a verifier for the solution received by the CR user node, and if the solution verification result satisfies the interference probability criterion, whether to perform the optimization is determined not to be performed so that the received solution is maintained, and if the solution verification result does not satisfy the interference probability criterion, it may be determined whether or not to perform the optimization.

In addition, the interference analysis unit may perform the solution verification using an interference analysis based on a Monte Carlo algorithm.

In addition, the interference analysis unit may determine whether or not to perform the optimization upon receiving result data including only the backup channel list.

In addition, the CR user node may match the generated transmission parameter with a solution for an input case to be updated in the case database.

According to one aspect of the present invention described above, by providing an intelligent dynamic real-time spectrum resource management system and an intelligent dynamic real-time spectrum resource management method using data mining and case-based reasoning, irregular patterns are provided through an accurate learning algorithm for data for each case. When there is a case of misclassification due to misclassification, it is possible to prevent system performance degradation due to incorrect prediction, reduce the complexity of intelligent dynamic real-time spectrum resource management, and enable rapid resource allocation.

1 is a diagram for explaining the performance of spectrum resource management using a conventional optimization engine;
2 is a diagram for explaining spectrum resource management performance through an intelligent dynamic real-time spectrum resource management system according to an embodiment of the present invention;
3 is a configuration diagram of an intelligent dynamic real-time spectrum resource management system according to an embodiment of the present invention;
4 is a block diagram showing a specific configuration of the CR master node shown in FIG. 1;
5 is a block diagram showing a specific configuration of a CR user node shown in FIG. 1;
6 is a diagram showing an example of preprocessing in the preprocessing unit shown in FIG. 4;
7 is a diagram showing a process of learning and reasoning in the learning unit shown in FIG. 4;
8 is a diagram showing an intelligent dynamic real-time spectrum resource management method according to another embodiment of the present invention, and
9 and 10 are diagrams specifically illustrating the intelligent dynamic real-time spectrum resource management process shown in FIG. 8 .

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The detailed description of the present invention which follows refers to the accompanying drawings which illustrate, by way of illustration, specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable one skilled in the art to practice the present invention. It should be understood that the various embodiments of the present invention are different from each other but are not necessarily mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in another embodiment without departing from the spirit and scope of the invention in connection with one embodiment. Additionally, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the detailed description set forth below is not to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all equivalents as claimed by those claims. Like reference numbers in the drawings indicate the same or similar function throughout the various aspects.

Components according to the present invention are components defined not by physical division but by functional division, and may be defined by the functions each performs. Each of the components may be implemented as hardware or program codes and processing units that perform respective functions, and the functions of two or more components may be implemented by being included in one component. Therefore, the names given to the components in the following embodiments are not to physically distinguish each component, but to imply the representative function performed by each component, and the names of the components indicate the present invention. It should be noted that the technical idea of is not limited.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

First, FIG. 1 is a diagram for explaining performance of spectrum resource management using a conventional optimization engine, and FIG. 2 is a diagram for explaining performance of spectrum resource management through an intelligent dynamic real-time spectrum resource management system according to an embodiment of the present invention. , and, Figure 3 is a configuration diagram of an intelligent dynamic real-time spectrum resource management system according to an embodiment of the present invention.

As shown in FIG. 1, the spectrum resource management based on the conventional optimization algorithm increases the calculation time and complexity, making it difficult to allocate resources quickly within a limited time, and wasting valuable solution information optimized according to the evolution of each spectrum state in the space-time and frequency domains. There is a problem that you have no choice but to do.

Accordingly, the intelligent dynamic real-time spectrum resource management system (hereinafter, the system) of the present invention, as shown in FIG. 2, is provided to reduce the complexity of spectrum resource management using an existing optimization engine and achieve rapid resource allocation. To this end, this system uses case-based reasoning. If there are cases that are misclassified due to irregular patterns, incorrect predictions can cause system performance degradation, so the need for an accurate learning algorithm for each case data is emphasized. do.

Therefore, in this system, data mining can be used along with case-based reasoning.

In addition, the system according to the present embodiment includes at least one CR user node (CR User, 200) and a CR master node (CR Master, 100) to form a CR (Cognitive Radio) network.

Here, the CR network is IEEE 802.22. A base station using a TV white space (TVWS) band according to a wireless area network (WRAN) standard may be an infrastructure-based CR network that controls IoT devices in suburban areas where it is difficult to support broadband access, but is not limited thereto.

At this time, the CR master node 100, which is each base station constituting a single wireless network, independently manages frequency resources within the CR network through database contact for self-coexistence between cells of the same type existing in adjacent channels and locations, The CR user node 200 is a terminal under the management of the CR master node 100, and may opportunistically use idle frequency resources in the CR network.

To this end, the CR user node 200 senses the channel state in each CR network, and the sensing data is a CR master who regularly uses data mining to analyze the activity patterns of existing users in the history data stored in the database (DB). It is transmitted (ηdata) to the database through the node 100. In addition, the CR master node 100 determines (② retrieve) whether to provide a solution for a case similar to the present case through similarity verification when reasoning using previous cases classified through data mining. In addition, the CR master node 100 preprocesses historical data from the database (DB) side (data cleaning, data reduction, data scaling), and performs machine learning based on the preprocessed data to determine the pattern of the user pattern. After extracting (pattern extract), it can be transferred to and stored in a database (DB).

And in the process of inferring an idle channel using the backup channel list (③), the CR user node 200 operates an optimization engine according to the similarity verification result in the case of an interference effect on an existing user existing in an adjacent channel or location (⑤ optimization engine) Alternatively, the solution provided by the CR master node 100 is reused (④ reuse (solved case)). In addition, the CR user node 200 may determine the solution verification for determining the operation of the optimization engine through interference analysis based on the Monte Carlo algorithm. In addition, the optimized solution is updated (⑥ retain (optimized case)) as a case database in the database (DB) for the next reasoning.

In addition, the case-based inference performed in the system of the present invention is performed by classifying the roles of the CR master node 100 and the CR user node 200 as shown, and using the case database generated through data mining for each case Similarity verification, solution reuse and verification, solution maintenance, and update operations can be performed.

Hereinafter, detailed configurations of each CR master node 100 and CR user node 200 will be described with reference to FIGS. 4 to 7 .

FIG. 4 is a block diagram showing the specific configuration of the CR master node 100 shown in FIG. 1, FIG. 5 is a block diagram showing the specific configuration of the CR user node 200 shown in FIG. 1, and FIG. A diagram showing an example of pre-processing in the pre-processing unit 121 shown in FIG. 4, and FIG. 7 is a diagram showing a process of learning and reasoning in the learning unit 123 shown in FIG.

In addition, the CR master node 100 and the CR user node 200 constituting the system may install and execute software (application) for performing an intelligent dynamic real-time spectrum resource management method.

The CR master node 100 includes a communication unit 110, a control unit 120 and a storage unit 130.

The communication unit 110 communicates with at least one or more CR user nodes 200 in the CR network to receive channel state sensing data from the CR user nodes 200, and transmits resultant data including a backup channel list to the CR user nodes 200. ) to the side.

In addition, the communication unit 110 communicates with the database (DB) associated with the CR network to transmit the channel state sensing data, solutions, and optimization cases from the CR user node 200 to the database (DB), and the previous cases and for a certain period of time. Collected sensing data may be received from the database (DB) as history data.

In particular, the control unit 120 of the CR master node 100 according to this embodiment creates a backup channel list through characterization, frequency determination, and classification based on history data representing statistical information of the CR network, which is a local network, to generate a list of CR users. It may facilitate a timely transition of node 200 to an idle channel.

In addition, the control unit 120 performs data pre-processing and data using an artificial neural network based on information sensed by the CR user node 200 for each pre-set inference cycle unit in order to improve prediction performance through accurate pattern learning for the sensed data. mining can be performed.

To this end, the control unit 120 may include a pre-processing unit 121, a learning unit 123, and a verification unit 125.

The pre-processing unit 121 may collect and pre-process history data including sensing data obtained by sensing the channel state of the CR network in the CR user node 200 as raw data. The preprocessed data is used for machine learning in the learning unit 123 .

Data pre-processing performed by the pre-processing unit 121 is a technique for improving the quality of raw data, and as shown in FIG. 6, data cleaning using history data, which is raw data, data reduction, and outlier identification Raw data may be preprocessed using at least one of outlier detection and data scaling.

At this time, the pre-processing unit 121 of the present invention performs a data pre-processing process consisting of data cleaning, data reduction, outlier detection, and data scaling steps to make the pre-processing of the data more desirable. can be done

Data cleaning aims to construct new features based on a linear combination of existing variables. Representative linear feature extraction is a statistical feature extraction method that calculates statistics such as the average or median value of raw data within a specific time range. However, if the time-series fluctuations are severe, data loss may occur, so the pre-processing unit 121 of the present invention may perform data reduction from statistical data to categorical data.

Categorical data refers to the frequency of occurrence of values belonging to the range when the range in which statistical data values exist is divided into several intervals of the same size according to user definition, and this method can minimize information loss and data size. When analyzing large-scale data, the amount of calculation can be effectively reduced.

을 하기의 수학식 1에 기초하여 산출할 수 있다. More specifically, the pre-processing unit 121 determines the occupancy probability representing statistical characteristics of data sensed within a total of M time slots per channel.

Can be calculated based on Equation 1 below.

[Equation 1]

는 j번째 채널을 의미하고,

는 i시점의 타임슬롯을 의미할 수 있다. here

denotes the jth channel,

may mean a timeslot of point i.

를 나눈 값은 하기의 수학식 2와 같이 정의될 수 있다. At this time, the width of the interval based on the number of intervals L specified within the range of the categorical data set (min, max)

The value divided by may be defined as in Equation 2 below.

[Equation 2]

이 통계자료 중 k번째 중앙값

를 기준으로

의 범위에 있으면 1로 정의하고, 그렇지 않으면 0으로 정의한다. 이때

의 출현빈도는 구간당

까지 카운트한 결과값을 의미할 수 있다. 이를 하기의 수학식 3과 같이 정의할 수 있다. Therefore

The kth median of these statistics

based on

If it is in the range of , it is defined as 1, otherwise it is defined as 0. At this time

The frequency of occurrence of

It may mean a result value counted up to . This can be defined as in Equation 3 below.

[Equation 3]

일 수 있다. here

can be

On the other hand, repetitive and irregular data due to various transmission errors, sensor malfunctions, or external interference in the data collection process are highly likely to generate outliers during data preprocessing. This excessively increases the variance in estimating the mean of the data population, causing a decrease in the accuracy of data analysis.

In order to solve this problem, the pre-processing unit 121 of the present invention presents the characteristics of each channel by calculating statistical data on raw data, which is difficult to distinguish between an outlier and noise, through data reduction and outlier identification. At this time, when the frequency of appearance is calculated by dividing several sections within the range of statistical data values, the data of the section with the smallest observed value is identified as an outlier and converted to 0 as shown in FIG. 6 .

Conventionally, a method of encoding data scaling in binary is used, which facilitates efficient search by categorizing patterns for each data feature. However, in order to classify categorical data with binary numbers of 0 and 1, a high-resolution categorical data set that can be distinguished from neighboring data sets having other characteristics is required.

Accordingly, in order to solve this problem, the pre-processing unit 121 according to the present embodiment rescales the range of each component in the categorical data set to a range between 0 and 1 using Min-Max Normalization to achieve overall uniformity in the data. can be granted. The general formula for normalization is shown in Equation 4 below.

[Equation 4]

Here, max and min may mean the categorical data set size and x' may mean a normalized value. The pre-processing unit 121 of the present invention can minimize the amount of calculation by pre-processing the data in the above order.

Meanwhile, the learning unit 123 may learn a behavior pattern of a CR network user by using the preprocessed data, and generate a behavior pattern prediction model.

Specifically, the learning unit 123 classifies the preprocessed history data for each case, and determines whether or not to deliver optimized solution information to the CR user node 200 according to a result of inference obtained using the classified case database.

Artificial neural networks are used in a wide range of areas such as communication, signal processing, and intelligent control. In particular, in CRN (Cognitive Radio Network), artificial neural networks achieve high prediction accuracy for existing user patterns in the future by learning the behavior patterns of existing users from a data set. have Accordingly, the learning unit 123 of the present invention may use the artificial neural network configured as shown in FIG. 7 for learning and reasoning for each feature of history data preprocessed according to the categorical data size and data scaling technique.

Referring to FIG. 7 , the learning unit 123 of the present invention may use an ANN structure to learn a behavior pattern of an existing user from a data set and achieve high prediction accuracy for a pattern of an existing user in the future. The ANN structure consists of an input layer, a hidden layer, and an output layer. And the input layer plays a role of delivering content image information composed of N neurons to the hidden layer. The hidden layer checks whether the feature pattern is included in the content image received from the input layer and transfers the content cost to the output layer. Assuming that feature patterns are defined for each hidden layer as shown in FIG. 7 , content cost calculation for a content image included in the feature pattern may be determined by the thickness (weight) of arrows connecting the hidden layers. The output layer may output certainty as a value between 0 and 1 in consideration of the content cost information calculated in the hidden layer, and finally compare the certainty factors of each output layer from 1 to 9 to determine the final content image.

에 가중치

와 임계값

가 주어졌을 때 각 은닉층에 대한 출력

는

로 정의될 수 있으며, 이는 하기의 수학식 5에 기초하여 산출될 수 있다. The learning unit 123 may define variables required to explain the relationship between neurons constituting the ANN and arrange relational expressions for the content cost and certainty factor of the hidden layer and the output layer in the previously constructed ANN. First, the ith input layer neuron

weighted on

and threshold

Output for each hidden layer given

Is

It can be defined as , which can be calculated based on Equation 5 below.

[Equation 5]

일 수 있다. here

can be

는 0과 1 사이의 간격으로 연속적으로 표현되는 시그모이드 함수이며, j번째 은닉층의 입력 선형합 a는 하기의 수학식 6과 같이 정의될 수 있다. At this time

is a sigmoid function continuously expressed at intervals between 0 and 1, and the input linear sum a of the j-th hidden layer may be defined as in Equation 6 below.

는 도 7에 표시된 두꺼운 화살표로 연결된 입력층의 전송 신호 이외의 전송 소스에서 오는 노이즈 신호를 차단하기 위함이다. Threshold here

is to block noise signals coming from transmission sources other than the transmission signals of the input layer connected by the thick arrows in FIG. 7 .

Through this, the learning unit 123 determines the content image output by the sigmoid function as an output certainty factor having a value between 0 and 1 for the output layer. That is, the preprocessed learning data can be classified by classifying classes according to average occupancy probabilities ranging from 0.1 to 0.9. For example, if the certainty coefficient of the output layer 1 is very close to 1, we can be sure that the average occupancy probability of the input image is 0.1.

의 합을 최소화하기 위한 최적의 가중치와 임계값을 하기의 수학식 6에 기초하여 도출할 수 있다. Also, the learning unit 123 may define an objective function for weight and threshold optimization. More specifically, using linear regression, the squared error of the output layer for each input data defined above

Optimal weights and thresholds for minimizing the sum of can be derived based on Equation 6 below.

[Equation 6]

및

는 데이터의 예측값을 입력하기 위한 변수이고, 우변의

는 설명변수, 좌변의

는 예측 변수로 정의될 수 있다. 이때 k번째 출력층에 주어진 목적변수

와 예측값

사이의 제곱오차

는 하기의 수학식 7에 기초하여 계산될 수 있다. In the above regression

and

is a variable for inputting the predicted value of the data, and

is the explanatory variable,

can be defined as a predictor variable. At this time, the target variable given to the kth output layer

and predicted value

squared error between

Can be calculated based on Equation 7 below.

[Equation 7]

)와 q(=

)를 최적화하기 위한 목적함수

는 전체 T 입력 데이터로부터 계산된

의 합이 최소화될 때 최적의 매개변수를 도출할 수 있다. where the parameters p(=

) and q(=

) objective function for optimizing

is calculated from the total T input data

Optimal parameters can be derived when the sum of is minimized.

[Equation 8]

=

=

일 수 있다. here

can be

Meanwhile, the verification unit 125 may generate result data by analyzing input cases based on the behavior pattern prediction model generated by the learning unit 123 .

The verification unit 125 compares the predicted value predicted through the behavior pattern prediction model with the actual value of the input case to verify the similarity, generates result data based on the similarity verification result, and generates result data for the CR user. It can be delivered to the node 200.

Here, the resulting data may include at least one of a previous solution, which is a solution for a previous case stored in the case database, and a backup channel list, which is a list of idle channels not used by a user at a specific time and location in the CR network.

를 최소화하는 최적의 은닉층 수와 임계값을 사용하여 유사도 검증을 수행한다. In order to verify this degree of similarity, the verification unit 125 performs an objective function when a new case is input.

Similarity verification is performed using the optimal number of hidden layers and a threshold that minimizes .

If the predicted value and the actual value are the same, the verification unit 125 may generate result data such that the CR master node 100 includes both the previous solution and the backup channel list.

Further, if the predicted value and the actual value are not identical, that is, if they are not identical, the verification unit 125 may generate result data such that only the backup channel list is included.

Meanwhile, the storage unit 130 is provided to store information transmitted/received or generated by the CR master node 100, and software (application) for performing an intelligent dynamic real-time spectrum resource management method may be stored.

On the other hand, when the CR user node 200 identifies an idle channel defined as a spectrum hole or white space that is not used at a specific time and location and shares the spectrum in an opportunistic manner, it must not interfere with existing users. .

If an existing user is detected in the process of inferring an idle channel, the CR user node 200 must pause its operation and switch to another idle channel so that communication can be maintained continuously. Such a switching method is defined as a spectrum handoff, and in the present invention, an inference engine operation can be performed through a reactive handoff technique based on a backup channel list.

At this time, if the CR user node 200 using an idle channel with maximum transmission power has an interference effect on an existing user existing in an adjacent channel and location, the CR user node 200 reuses a solution or operates an optimization engine according to the similarity verification result. can be done To this end, the CR user node 200 includes a communication unit 210, a control unit 220, and a storage unit 230.

The communication unit 210 transmits the sensing data to the CR master node 100 when the control unit 220 detects the spectrum for a certain period and generates sensing data, which is channel state information, and transmits the backup channel from the CR master node 100. Receive result data containing a list.

Meanwhile, the controller 220 performs a spectrum handoff operation to another idle channel so that continuous communication is possible without disturbing an existing user.

More specifically, the control unit 220 may include an interference analysis unit 221 configured as an inference engine and determine whether to perform optimization based on result data.

When the interference analysis unit 221 receives result data including a solution from the CR master node 100, it performs verification of the received solution as a verifier, and if the solution verification result satisfies the interference probability criterion, optimization is performed. It is possible to determine whether or not to perform the solution so that the solution received from the CR master node 100 is maintained without performing separate optimization through the optimization unit 223.

The interference analysis unit 221 may determine whether or not to perform optimization when the solution verification result does not satisfy the interference probability criterion, that is, if it is confirmed that the interference probability criterion is not satisfied, so that the optimization unit 223 performs the optimization.

The interference analysis unit 221 may perform solution verification using interference analysis based on the Monte Carlo algorithm.

More specifically, the interference analysis unit 221 reuses the solution received from the CR master node 100 when the predicted value and the actual value are the same, and may use interference analysis based on the Monte Carlo algorithm to verify the solution. To this end, the interference analysis unit 221 obtains radio wave transmission-related samples generated randomly based on input parameters applied to a specific interference environment scenario in the interference analysis based on the Monte Carlo algorithm, and uses the obtained samples to communicate between users. Determine the interference probability. In this interference environment scenario, it is possible to define a propagation loss model according to parameters of interference (antenna height, transmission power, center frequency, etc.) and specific distance, location, and terrain characteristics. And this interference environment refers to an existing user in use at a specific location and time in the licensed band, in an interfering link including an interfering transmitter and an interfering receiver, and in an adjacent channel and location based on CR technology in the licensed band. Any communication that temporarily uses an idle channel not used by an existing user at a specific time may include an interfering link including an anti-interfering receiver and an interfering transmitter (CR user node).

The interference analysis unit 221 may calculate the received power that the interfered receiver receives from the interfered transmitter and the received power that the interfered receiver receives from the interfering transmitter, and determine an interference probability based on the calculated received power.

In addition, upon receiving the resultant data including only the backup channel list, the interference analysis unit 221 determines whether or not to perform the optimization and allows the optimization unit 223 to perform the optimization.

In addition, the solution verification may determine the operation of the solution maintenance or optimization unit 223 according to the 5% interference probability criterion applied in the field test. If the interference probability exceeds the 5% criterion or more during the solution verification process, the interference analysis unit 221 determines that the received solution information is not suitable, and the optimization unit 223 determines that the interference probability satisfies the criterion within 5%. Transmission parameters are derived. Conversely, if the criterion within 5% of the interference probability is satisfied, the received previous solution is maintained. These specific interference probability standards are illustrative for convenience of explanation, but are not limited thereto.

If the predicted value and the actual value are different, the interference analysis unit 221 determines that there is no similar case in the past, calculates the interference probability in the interference analysis unit 221, and then, when interference occurs, through the optimization unit 223, the existing user and An optimal transmission parameter capable of mutual coexistence can be derived.

In addition, when the interference analysis unit 221 determines to perform the optimization, the control unit 220 generates transmission parameters that satisfy the standard of interference probability with existing users in the channel, and controls the transmission power based on the transmission parameters. ) is further included.

Then, the control unit 220 matches the transmission parameters (parameters) generated through the operation of the solution reuse or optimization unit 223 for each time slot unit up to the entire inference cycle as a solution for the input new case, which leads to the next inference. to be updated in the case database.

Meanwhile, the storage unit 230 is provided to store information transmitted/received or generated by the CR user node 200, and software (application) for performing an intelligent dynamic real-time spectrum resource management method may be stored.

Through this, the system of the present invention facilitates classification by feature of history data sensed from a large amount of IoT, and when inferring using a case database containing past optimized solution information, existing optimization algorithms are reused by reusing solutions for similar cases in the past. It is possible to reduce the amount of computation and complexity according to decision-making based on and enable efficient real-time resource allocation.

Meanwhile, FIG. 8 is a diagram illustrating an intelligent dynamic real-time spectrum resource management method according to another embodiment of the present invention, and FIGS. 9 and 10 are diagrams showing the intelligent dynamic real-time spectrum resource management process shown in FIG. 8 in detail. it is a drawing

Since the intelligent dynamic real-time spectrum resource management method according to an embodiment of the present invention is performed on substantially the same configuration as the system shown in FIGS. 2 to 7, the same reference numerals for the same components as the system of FIGS. 2 to 7 , and repeated descriptions will be omitted.

The intelligent dynamic real-time spectrum resource management method according to the present embodiment is an intelligent dynamic real-time spectrum resource management method in a system including at least one CR user node 200 and a CR master node 100 constituting a CR network. , Raw data collection and preprocessing (S110) in the CR master node 100, generation of a behavior pattern prediction model based on the preprocessed data in the CR master node 100 (S120), and case based on the behavior pattern prediction model is analyzed (S130), the CR master node 100 transfers the resulting data to the CR user node 200 (S140), the CR user node 200 determines whether to perform optimization (S150), and the optimization is performed. If determined (S150-Yes), the CR user node 200 generates transmission parameters (S160), and the CR user node 200 controls transmission power based on the generated transmission parameters (S170) and maintains the transmission power. Control to do (S180).

If it is determined that optimization is not performed (S150-No), the CR user node 200 controls transmission power to be maintained according to the existing solution based on the existing solution (S180).

In the step of collecting and pre-processing the raw data in the CR master node 100 (S110), the CR user node 200 collects and pre-processes history data including sensing data obtained by sensing the channel state of the CR network as raw data. can

9 is a flow of data mining using history data. Referring to FIG. 9, in the step of generating a behavior pattern prediction model based on the data pre-processed in the CR master node 100 (S120), the pre-processed data It is possible to learn behavioral patterns of CR network users and create behavioral pattern prediction models.

Further, in step S130 in which the CR master node 100 analyzes cases based on the behavior pattern prediction model, result data may be generated by analyzing input cases based on the generated behavior pattern prediction model.

The step 130 of generating result data by analyzing these cases is a step of verifying similarity by comparing the predicted value predicted by the behavior pattern prediction model with the actual value of the input case by the CR master node 100, and the CR master node 100. The node 100 may generate result data based on the similarity verification result, and the CR master node 100 may include transmitting the generated result data to the CR user node 200.

Here, the resulting data includes at least one of a previous solution, which is a solution for a previous case stored in the case database, and a backup channel list, which is a list of idle channels not used by a user at a specific time and location in the CR network.

In addition, if the predicted value predicted through the behavior pattern prediction model for the input case and the actual value are the same, the CR master node 100 generates result data so that the list of previous solutions and backup channels are all included, and the predicted value and actual value are different. If they are the same, result data may be generated so that only the backup channel list is included.

Also, in the step of determining whether the CR user node 200 performs optimization (S150), whether to perform optimization is determined based on the result data.

In the step of determining whether or not to perform such an optimization (S150), when the result data including the solution is received, the CR user node 200 verifies the received solution as a verifier solution, and the CR user node 200 As a result of solution verification, if the interference probability criterion is satisfied, determining whether or not to perform optimization and maintaining the received solution, and if the CR user node 200 does not satisfy the interference probability criterion as a result of solution verification, whether to perform optimization It may include the step of determining as.

In the step of performing solution verification, solution verification may be performed using interference analysis based on the Monte Carlo algorithm.

In the step of determining whether to perform optimization (S150), when the CR user node 200 receives the result data including only the backup channel list from the CR master node 100, the step of determining whether or not to perform optimization is further performed. may also include

When optimization is determined (S150-Yes), in the step of generating transmission parameters by the CR user node 200 (S160), the CR user node 200 determines the transmission parameters that satisfy the interference probability criterion with the existing user in the channel can create

Thereafter, the CR user node 200 may perform a step of controlling transmission power based on the generated transmission parameters (S170).

Then, the CR user node 200 controls to maintain transmission power (S180).

If it is determined that optimization is not performed (S150-No), the CR user node 200 controls transmission power to be maintained according to the existing solution based on the existing solution (S180).

In addition, the intelligent dynamic real-time spectrum resource management method may further include a step of allowing the CR user node 200 to update the case database by matching the generated transmission parameter with a solution for the input case.

As shown in FIG. 10, after learning is completed in the CR master node 100, the spectrum resource management method using case-based reasoning is performed by classifying the roles of the CR master node 100 and the CR user node 200. , similarity verification, solution reuse and verification, solution maintenance, and update operations can be performed using the case database created through case-by-case data mining.

Such an intelligent dynamic real-time spectrum resource management method of the present invention may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer readable recording medium may include program instructions, data files, data structures, etc. alone or in combination.

Program instructions recorded on the computer-readable recording medium may be those specially designed and configured for the present invention, or those known and usable to those skilled in the art of computer software.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like.

Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter or the like as well as machine language codes such as those produced by a compiler. The hardware device may be configured to act as one or more software modules to perform processing according to the present invention and vice versa.

Although various embodiments of the present invention have been shown and described above, the present invention is not limited to the specific embodiments described above, and is commonly used in the technical field to which the present invention pertains without departing from the gist of the present invention claimed in the claims. Of course, various modifications are possible by those with knowledge of, and these modifications should not be individually understood from the technical spirit or prospect of the present invention.

100: CR Masternode 200: CR user node

Claims (15)

An intelligent dynamic real-time spectrum resource management method in an intelligent dynamic real-time spectrum resource management system including a CR master node constituting a CR network, at least one CR user node, and a case database,
collecting, by the CR master node, history data including sensing data obtained by sensing a channel state of the CR network at the CR user node as raw data and pre-processing;
learning, by the CR master node, a behavior pattern of the user of the CR network using the preprocessed data, and generating a behavior pattern prediction model;
generating result data by analyzing input cases based on the generated behavior pattern prediction model by the CR master node;
determining, by the CR user node, whether to perform optimization based on the result data;
generating, by the CR user node, a transmission parameter that satisfies an interference probability criterion with an existing user within a channel, if the determining step determines to perform the optimization; and
and controlling, by the user node, transmission power based on the transmission parameter. According to claim 1,
The step of generating the result data,
verifying a degree of similarity by comparing a predicted value predicted through the behavior pattern prediction model with an actual value for the case where the CR master node is input;
generating, by the CR master node, result data based on the similarity verification result; and
The intelligent dynamic real-time spectrum resource management method comprising the step of transmitting the resultant data generated by the CR master node to the CR user node. According to claim 2,
The resulting data,
At least one of a previous solution that is a solution for a previous case stored in the case database and a backup channel list that is a list of idle channels not used by a user at a specific time and location in the CR network,
In the step of generating the result data based on the similarity verification result,
If the predicted value and the actual value are identical, the CR masternode generates the result data so that both the previous solution and the backup channel list are included, and if the predicted value and the actual value are not identical, only the backup channel list is included An intelligent dynamic real-time spectrum resource management method, characterized in that for generating the resultant data. According to claim 3,
The step of determining whether to perform the optimization,
If result data including the solution is received, performing solution verification as a verifier on the received solution by the CR user node;
if the CR user node satisfies the interference probability criterion as a result of the solution verification, determining whether or not to perform the optimization and maintaining the received solution; and
and determining whether to perform the optimization if the CR user node does not satisfy the interference probability criterion as a result of the solution verification. According to claim 4,
In the step of performing the solution verification,
An intelligent dynamic real-time spectrum resource management method characterized in that the solution verification is performed using Monte Carlo algorithm-based interference analysis. According to claim 3,
The step of determining whether to perform the optimization,
and determining whether or not to perform the optimization when receiving resultant data including only the backup channel list. According to claim 1,
The intelligent dynamic real-time spectrum resource management method further comprising the step of matching the generated transmission parameter with the solution for the input case by the CR user node and updating the case database. According to claim 1,
In the preprocessing step,
An intelligent dynamic real-time spectrum resource management method, characterized in that for pre-processing the raw data using at least one of data cleaning, data reduction, outlier detection and data scaling. An intelligent dynamic real-time spectrum resource management system including a CR master node constituting a CR network, at least one CR user node, and a case database,
The CR masternode,
a pre-processing unit for collecting and pre-processing history data including sensing data obtained by sensing a channel state of the CR network at the CR user node as raw data;
a learning unit for learning a behavioral pattern of the CR network user using the preprocessed data and generating a behavioral pattern prediction model; and
And a verification unit for generating result data by analyzing input cases based on the generated behavior pattern prediction model;
The CR user node,
an interference analysis unit determining whether to perform optimization based on the resultant data; and
An intelligent dynamic real-time spectrum resource management system comprising an optimization unit for generating transmission parameters that satisfy a standard of interference probability with an existing user in a channel when it is determined to perform the optimization, and controlling transmission power based on the transmission parameters. According to claim 9,
The verification unit,
The similarity is verified by comparing the predicted value predicted through the behavior pattern prediction model with the actual value of the input case, the result data is generated based on the similarity verification result, and the generated result data is transmitted to the CR user node. An intelligent dynamic real-time spectrum resource management system, characterized in that. According to claim 10,
The resulting data,
At least one of a previous solution that is a solution for a previous case stored in the case database and a backup channel list that is a list of idle channels not used by a user at a specific time and location in the CR network,
The verification unit,
If the predicted value and the actual value are identical, the CR masternode generates the result data so that both the previous solution and the backup channel list are included, and if the predicted value and the actual value are not identical, only the backup channel list is included An intelligent dynamic real-time spectrum resource management system, characterized in that for generating the resultant data. According to claim 11,
The interference analysis unit,
When the result data including the solution is received, the CR user node performs verification of the verifier solution for the received solution,
If the solution verification result satisfies the interference probability criterion, it is determined whether or not to perform the optimization so that the received solution is maintained,
An intelligent dynamic real-time spectrum resource management system, characterized in that if the solution verification result does not satisfy the interference probability criterion, it is determined whether to perform the optimization. According to claim 12,
The interference analysis unit,
An intelligent dynamic real-time spectrum resource management system, characterized in that the solution verification is performed using Monte Carlo algorithm-based interference analysis. According to claim 11,
The interference analysis unit,
Intelligent dynamic real-time spectrum resource management system, characterized in that when receiving the result data containing only the backup channel list, it is determined whether or not to perform the optimization. According to claim 9,
The CR user node,
An intelligent dynamic real-time spectrum resource management system characterized in that the generated transmission parameter is matched with a solution for the input case and updated in the case database.

Images