Channel estimation method and system in communication path environment using reconfigurable intelligent surface

Abstract

A channel estimation method is disclosed in a communication path environment through an intelligent reflective surface. According to one embodiment, the channel estimation method can estimate a spatial frequency () related to the signal arrival angle to the receiver and a spatial frequency () related to the signal reflection angle of the reconfigurable intelligent surface by calculating the reception angle steering vector () and reflection angle steering vector () and atoms () of the atomic norm minimization problem, and then updating the matrix () and Lagrange multiplier () of the atomic norm minimization problem.

Description

Technology Field

[0001] The following embodiments relate to a method and system for channel estimation in a communication path environment through a reconfigurable intelligent surface. Background Technology

[0003] Reconfigurable Intelligent Surfaces (RIS) are emerging as a promising solution for improving wireless coverage beyond 5th generation (5G) networks. For example, reconfigurable intelligent surfaces, which create a favorable communication environment by manipulating electromagnetic waves, are suitable for unmanned aerial vehicle (UAV) supported communications, particularly for extending limited coverage in the millimeter wave (mmWave) frequency band, due to their low-power characteristics.

[0004] While accurate channel estimation is essential for such reconfigurable intelligent surfaces, the cascaded channel, consisting of links between the transmitter and the reconfigurable intelligent surface and between the reconfigurable intelligent surface and the receiver due to reflections from the reconfigurable intelligent surface, acts as a factor that makes channel estimation difficult.

[0005] Accordingly, as a channel estimation technique in a communication path environment through a reconfigurable intelligent surface, a technique applying the atomic norm minimization problem and a technique applying a deep learning model have been proposed.

[0006] However, existing channel estimation techniques that apply the atomic norm minimization problem have the disadvantage of high matrix computational complexity.

[0007] In addition, existing channel estimation techniques using deep learning models have the disadvantage of reduced computational speed and lower computational accuracy when hyperparameters related to the weights of the deep learning model converge improperly.

[0008] Therefore, there is a need to propose a channel estimation technique that addresses the shortcomings of existing technologies. The problem to be solved

[0010] One embodiment proposes a channel estimation method and system that apply the atomic norm minimization problem based on the Alternating Direction Method of Multipliers (ADMM) technique to reduce the matrix computational complexity of the atomic norm minimization problem.

[0011] In addition, some embodiments propose a channel estimation method and system that adjust hyperparameters related to layer-by-layer weights of a deep learning model according to the learning results of the deep learning model in order to improve computation speed and computation accuracy.

[0012] However, the technical problems that the present invention aims to solve are not limited to the above problems and can be expanded in various ways without departing from the technical concept and scope of the present invention. means of solving the problem

[0014] According to one embodiment, a channel estimation method in a communication path environment through an intelligent reflective surface is based on the Alternating Direction Method of Multipliers (ADMM) technique and a spatial frequency related to the signal arrival angle to a receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( AAA( ), BBB( ), CCC( A step of calculating ); using a spectrum shift operation, the matrix of the atomic norm minimization problem ( Step of updating ); Lagrange multiplier of the above atomic norm minimization problem ( A step of updating ); and calculating the atomic norm minimization problem based on the above calculation and update results to calculate the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( It may include a step of estimating ).

[0015] According to one aspect, the calculating step utilizes hyperparameters related to the layer-by-layer weights of a deep learning model for iteratively applying the ADMM (Alternating Direction Method of Multipliers) technique to obtain a receiving angle steering vector of an atomic norm minimization problem ( ), reflection angle steering vector( ), atom( It may be a step that produces ).

[0016] According to another aspect, the channel estimation method may further include a step of adjusting hyperparameters related to the layer-by-layer weights of the deep learning model according to the learning results of the deep learning model.

[0017] According to another aspect, the channel estimation method further comprises the step of acquiring a received signal received at the receiver, and the estimating step comprises a spatial frequency related to the received signal and the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( It may further include a step of estimating the channel in the communication path environment using ).

[0018] According to one embodiment, in a computer program stored on a non-transient computer-readable recording medium for executing a channel estimation method in a communication path environment through an intelligent reflective surface on a computer device, the channel estimation method comprises a spatial frequency related to the signal arrival angle to a receiver based on the Alternating Direction Method of Multipliers (ADMM) technique. ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) receiving angle steering vector in the atomic norm minimization problem of estimating ) ), reflection angle steering vector( ), atom( A step of calculating ); using a spectrum shift operation, the matrix of the atomic norm minimization problem ( Step of updating ); Lagrange multiplier of the above atomic norm minimization problem ( A step of updating ); and calculating the atomic norm minimization problem based on the above calculation and update results to calculate the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( It may include a step of estimating ).

[0019] According to one embodiment, a computer device for executing a channel estimation method in a communication path environment through an intelligent reflective surface includes at least one processor, and the at least one processor includes a spatial frequency related to the signal arrival angle to a receiver based on the Alternating Direction Method of Multipliers (ADMM) technique. ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) receiving angle steering vector in the atomic norm minimization problem of estimating ) ), reflection angle steering vector( ), atom( Calculate ) and use spectral shift operation to matrix of the above atomic norm minimization problem ( Updating ) and the Lagrange multiplier of the above atomic norm minimization problem ( Update ) and calculate the atomic norm minimization problem based on the above calculation and update results to the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( ) can be estimated.

[0020] According to one aspect, the at least one processor uses hyperparameters related to layer-by-layer weights of a deep learning model for iteratively applying the ADMM (Alternating Direction Method of Multipliers) technique to the receiving angle steering vector of an atomic norm minimization problem ( ), reflection angle steering vector( ), atom( ) can be produced.

[0021] According to another aspect, the at least one processor can adjust hyperparameters related to the layer-by-layer weights of the deep learning model according to the learning results of the deep learning model.

[0022] According to another aspect, the at least one processor acquires a received signal received at the receiver, and a spatial frequency related to the received signal and the angle of signal arrival at the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( The channel in the above communication path environment can be estimated using ). Effects of the invention

[0024] One embodiment can achieve the technical effect of reducing the matrix computational complexity of the atomic norm minimization problem by proposing a channel estimation method and system that apply the atomic norm minimization problem based on the ADMM (Alternating Direction Method of Multipliers) technique.

[0025] In addition, some embodiments can achieve technical effects that improve computational speed and computational accuracy by proposing a channel estimation method and system that adjusts hyperparameters related to layer-specific weights of a deep learning model according to the learning results of the deep learning model.

[0026] However, the effects of the present invention are not limited to the above effects and can be extended in various ways without departing from the technical concept and scope of the present invention. Brief explanation of the drawing

[0027] FIG. 1 is a diagram illustrating a communication path environment through a reconfigurable intelligent surface according to one embodiment. FIG. 2 is a diagram illustrating a deep learning model, which is a channel estimation system according to one embodiment. FIG. 3 is a diagram illustrating a spectrum shift operation in a channel estimation process according to one embodiment. FIG. 4 is a diagram illustrating a channel estimation algorithm according to one embodiment. FIG. 5 is a flowchart illustrating a channel estimation method according to one embodiment. Specific details for implementing the invention

[0028] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited or restricted by the embodiments. Also, the same reference numerals in each drawing indicate the same components.

[0029] Furthermore, the terminology used in this specification is used to appropriately describe preferred embodiments of the present invention, and may vary depending on the intent of the viewer or operator, or the conventions of the field to which the present invention belongs. Accordingly, the definitions of these terms should be based on the content throughout this specification. For example, in this specification, the singular form includes the plural form unless specifically stated otherwise in the text. Also, the terms "comprises" and / or "comprising" used in this specification do not exclude the presence or addition of one or more other components, steps, actions, and / or elements to the mentioned components, steps, actions, and / or elements. Additionally, although terms such as "first," "second," etc., are used in this specification to describe various regions, directions, shapes, etc., these regions, directions, and shapes should not be limited by such terms. These terms are used merely to distinguish one specific region, direction, or shape from another region, direction, or shape. Accordingly, a part referred to as the first part in one embodiment may be referred to as the second part in another embodiment.

[0030] Furthermore, it should be understood that various embodiments of the present invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the technical spirit and scope of the present invention in relation to one embodiment. Additionally, it should be understood that the location, arrangement, or configuration of individual components within each presented category of embodiments may be changed without departing from the technical spirit and scope of the present invention.

[0031] Hereinafter, with reference to the drawings, a channel estimation method and system are described that apply the atomic norm minimization problem based on the Alternating Direction Method of Multipliers (ADMM) technique in order to reduce the matrix computation complexity of the atomic norm minimization problem. In addition, a channel estimation method and system are proposed that adjust hyperparameters related to the weights of each layer of a deep learning model according to the training results of the deep learning model in order to improve computation speed and computation accuracy.

[0033] FIG. 1 is a diagram illustrating a communication path environment through a reconfigurable intelligent surface according to one embodiment, FIG. 2 is a diagram illustrating a deep learning model which is a channel estimation system according to one embodiment, FIG. 3 is a diagram for explaining a spectrum shift operation in a channel estimation process according to one embodiment, FIG. 4 is a diagram illustrating a channel estimation algorithm according to one embodiment, and FIG. 5 is a flowchart illustrating a channel estimation method according to one embodiment.

[0034] Referring to the drawings, as shown in FIG. 1, the communication path environment through a reconfigurable intelligent surface (UAV-RIS assisted mmWave communication system) includes a single antenna transmitter, N RX A multi-antenna receiver with N passive reflection elements (Receiver with N RX antennas) and N R Reconfigurable intelligent surface with N reflective elements (UAV-RIS with N R It can include elements).

[0035] Reconfigurable intelligent surfaces are, N e Row and N a N having a column R N passive reflection elements (N R =N eN e It may include a uniform planar array (UPA) of ).

[0036] A multi-antenna receiver is, N RX It may include a uniform linear array (ULA) composed of antennas.

[0037] In a communication path environment through such a reconfigurable intelligent surface, M P - Following the symbol pilot transmission step (MM P Block transmission is performed with M symbols per block, consisting of a )-symbol data transmission step.

[0038] In this case, the signal received at the receiver at the t-th symbol time is equal to Equation 1 below.

[0040] <Equation 1>

[0041]

[0043] Here is power constraint In transmission symbol, and represents the channel between the transmitter and the RIS, and the channel between the RIS and the receiver, respectively, and Is It refers to the RIS reflection pattern, The mean is 0 and the variance is It refers to additive white Gaussian noise.

[0044] For RIS-supported systems operating at mmWave frequencies, the Saleh-Valenzuela channel model is adopted as shown in Equation 2 below.

[0046] <Equation 2>

[0047]

[0049] of each channel For the nth path These represent path gain, Doppler shift, spatial frequency associated with the horizontal signal reflection angle of the reconfigurable intelligent surface, and spatial frequency associated with the vertical signal reflection angle of the reconfigurable intelligent surface, respectively. represents the number of paths, and represents the spatial frequency related to the signal arrival angle at the receiver. The array response vector of the UPA is two ULA response vectors and It is expressed by the Kronecker product of. The ULA response vector is It is defined as, where Igo It is assumed that adjacent antennas and reflective elements are separated by a half-wavelength interval.

[0050] Using the relationship and channel model Equation 2, the received signal of Equation 1 can be expressed as Equation 3 below.

[0052] <Equation 3>

[0053]

[0055] In Equation 3, the effective path gain and path angle are and When and It is defined as.

[0056] Equation 2 can be expressed in a simplified matrix form as shown in Equation 4 below.

[0058] <Equation 4>

[0059]

[0061] Here , the array response matrix is , and It is defined as, represents the Hamadd product.

[0062] On / off RIS patterns in the vertical and horizontal directions can be used to utilize knowledge of the array manifold for channel estimation.

[0064] <Equation 5>

[0065]

[0067] Here and the number of pilot symbols is for low pilot overhead. When it is an integer multiple of the sum of the bidirectional RIS elements, i.e., It is assumed that...

[0068] This RIS pattern is In the case of, in the vertical direction, In this case, since the incident signal is reflected in the horizontal direction, the received signals from each direction can be combined individually. Specifically, the pilot symbol is If set to, The combined signal at the receiver for can be calculated as follows.

[0070] <Equation 6>

[0071]

[0072] <Equation 7>

[0073]

[0075] Here And, is, am.

[0076] Spatial frequency related to the horizontal signal reflection angle of a reconfigurable intelligent surface Also, spatial frequency related to the vertical signal reflection angle of a reconfigurable intelligent surface Since it is combined similarly to, below, in order to estimate the channel in a communication path environment through an intelligent reflective surface, the spatial frequency related to the signal arrival angle at the receiver and spatial frequency related to the vertical signal reflection angle of reconfigurable intelligent surfaces The explanation focuses on estimating.

[0077] That is, the spatial frequency related to the vertical signal reflection angle of the reconfigurable intelligent surface described below. The process of estimating the spatial frequency related to the horizontal signal reflection angle of a reconfigurable intelligent surface By performing the same for, the channel estimation method is based on the spatial frequency related to the signal arrival angle at the receiver. spatial frequency related to the signal reflection angle of and reconfigurable intelligent surfaces , After estimating, the spatial frequency related to the signal arrival angle at the estimated receiver spatial frequency related to the signal reflection angle of and reconfigurable intelligent surfaces , Null can be estimated using .

[0079] The spatial frequency related to the signal arrival angle at the receiver in a noise model such as Equation 7. spatial frequency related to the signal reflection angle of and reconfigurable intelligent surfaces , Estimating is conditional It can be formulated as an atomic norm minimization problem. Here, the atomic norm is It is defined as such. This atomic norm can be characterized by the minimization of ranks as in Equation 8.

[0081] <Equation 8>

[0082]

[0084] Here Is Represents the Hermitian Toeplitz matrix with as the first column, and the matrix Is It can be expressed as such. To estimate the parameters of interest around observation Y, the low-rank matrix recovery problem can be reconfigured into a Semidefinite Programming (SDP)-based minimization problem.

[0086] <Equation 9>

[0087]

[0089] Here, the normalization parameter , It is designed to balance scarcity and observational fidelity.

[0090] Equation 9 provides a manageable method for calculating the atomic norm using the CVX toolbox. However, there is a limit to how computational complexity increases significantly in large-scale problems as the size of the positive definite constraint matrix scales with the number of antenna and reflection elements.

[0091] Accordingly, the channel estimation method according to one embodiment may apply the atomic norm minimization problem based on the ADMM (Alternating Direction Method of Multipliers) technique to reduce the matrix computation complexity of the atomic norm minimization problem.

[0092] By applying the ADMM technique, the augmented Lagrangian of Equation 9 can be derived as shown in Equation 10.

[0094] <Equation 10>

[0095]

[0097] Here is the Lagrange multiplier, is the penalty parameter, and and represent the inner product function and the indicator function, respectively, in the appropriate space. The i-th ADMM iteration to minimize the augmented Lagrange is given as follows.

[0099] <Equation 11>

[0100]

[0101] <Equation 12>

[0102]

[0104] Here, the multiplier is the step size It can be updated through a double gradient ascent step with . This ADMM iteration can be implemented as a closed-loop solution, but Updates to require computationally expensive eigenvalue decomposition (ED), which can place a burden on high-dimensional matrices, and hyperparameters , , If it is set improperly, the computational speed, which is the convergence speed, and the computational accuracy of the ADMM algorithm may decrease.

[0105] Accordingly, the channel estimation method according to one embodiment can adjust and change hyperparameters related to layer-by-layer weights of a deep learning model (Deep alternating direction unfolded network; DADU-Net) as shown in FIG. 2 according to the learning results of the deep learning model.

[0106] In deep learning models, ADMM iterations are It can be performed for each of the layers. Accordingly, hyperparameters related to layer-specific weights It can be treated as a trainable parameter and adjusted and changed to improve adaptability.

[0107] Each layer of a deep learning model can be composed of five blocks that process five equations such as Equations 13 to 17 below.

[0109] <Equation 13>

[0110]

[0111] <Equation 14>

[0112]

[0113] <Equation 15>

[0114]

[0115] <Equation 16>

[0116]

[0117] <Equation 17>

[0118]

[0120] Here represents a model-based ADMM iteration block. , , and The basic update for is given Solve the augmented Lagrange of Equations 11 and 12 using [the method], and the double update is based on the obtained solution. It can be improved.

[0121] The proposed deep learning model, according to the algorithm shown in Fig. 4, relates hyperparameters to the received signal Y and the layer-specific weights of the deep learning model After obtaining, the receiving angle steering vector of the atomic norm minimization problem ( ), reflection angle steering vector( ), atom( Calculate ) and the matrix of the atomic norm minimization problem( Updating ), the Lagrange multiplier of the atomic norm minimization problem ( By updating ), the atomic norm minimization problem is calculated based on the output and update results to obtain the spatial frequency related to the signal arrival angle at the receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) can be estimated.

[0122] In this case, the main computational complexity of the ADMM iteration is the intermediate solution This can occur in the eigenvalue decomposition (ED) projecting onto the PSD cone, so to solve this, the matrix of the atomic norm minimization problem ( In the process of updating ), a spectrum shift operation that approximates the PSD constraint can be used as shown in Fig. 3.

[0123] Here, is defined in line 3 of Algorithm 1 shown in FIG. 4 It represents a similar transformation that applies a smaller scaling factor to the negative eigenvalues ​​of. The choice is It can depend on the eigenvalues ​​of and can be repeatedly updated based on primary and dual variables.

[0124] Table 1 below shows the CNN within a BS block using a size 2x2 NF filter in terms of end-to-end inference performance. We present a quantitative analysis of the calculations. Empirical results indicate that CNN-based networks improve Mean Squared Error (MSE) performance, whereas, About It shows that using the Frobenius norm achieves a good balance between MSE and runtime performance.

[0125] MethodMetric Matrix norm CNN-based network MSE(dB) -38.38 -38.22 -38.48 -38.50 Runtime(ms) 6.84 15.05 15.20 15.43

[0127] The hyperparameters explained earlier , , Regarding adjusting and changing based on the training results of the deep learning model, first, the learnable hyperparameters included in each layer of the deep learning model can be initialized to a constant value during algorithm execution. The size training dataset is a known parameter Received signal corresponding to It includes. The loss function of the deep learning model is defined by the weighted MSE as shown in Equation 18 below.

[0129] <Equation 18>

[0130]

[0132] Here, the variable with hat represents the estimated parameter, and The term regarding must be in the form of a real value. The deep learning model uses stochastic gradients to minimize the loss and weights You can adjust it.

[0133] As explained above, by calculating the atomic norm minimization problem, the spatial frequency related to the signal arrival angle at the receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( Once ) is estimated, the channel in the communication path environment can be estimated.

[0135] The described channel estimation method can be performed by a computer device implementing a channel estimation system. More specifically, the channel estimation method can be performed by at least one processor included in the computer device through the steps (S510 to S540) of the flowchart illustrated in FIG. 5.

[0136] In step (S510), at least one processor, based on the Alternating Direction Method of Multipliers (ADMM) technique, has a spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) receiving angle steering vector in the atomic norm minimization problem of estimating ) ), reflection angle steering vector( ), atom( ) can be produced.

[0137] For example, at least one processor, using hyperparameters related to the layer-by-layer weights of a deep learning model for iteratively applying the ADMM (Alternating Direction Method of Multipliers) technique, the receiving angle steering vector of the atomic norm minimization problem ( ), reflection angle steering vector( ), atom( ) can be produced.

[0138] In particular, at least one processor can adjust and change hyperparameters related to the layer-by-layer weights of the deep learning model according to the learning results of the deep learning model prior to step (S510).

[0139] In step (S520), at least one processor uses a spectrum shift operation to form a matrix of the atomic norm minimization problem ( Can update ).

[0140] At step (S530), at least one processor is the Lagrange multiplier of the atomic norm minimization problem ( Can update ).

[0141] In step (S540), at least one processor calculates the atomic norm minimization problem based on the output and update results and the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) can be estimated.

[0142] The channel estimation method described above, although not illustrated as a separate step, obtains a received signal received at a receiver, and together with the received signal, through step (S540), a spatial frequency ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( The channel in the communication path environment can be estimated using ).

[0144] The system or device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and component described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a Field Programmable Gate Array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0145] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0146] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available 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; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0147] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0148] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 In a channel estimation method in a communication path environment through an intelligent reflective surface, based on the ADMM (Alternating Direction Method of Multipliers) technique, the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) receiving angle steering vector in the atomic norm minimization problem of estimating ) ), reflection angle steering vector( ), atom( A step of calculating ); using a spectrum shift operation, the matrix of the atomic norm minimization problem ( Step of updating ); Lagrange multiplier of the above atomic norm minimization problem ( A step of updating ); and calculating the atomic norm minimization problem based on the above calculation and update results to calculate the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( The step of estimating ) is included, and the calculation step is to use hyperparameters related to the layer-by-layer weights of a deep learning model for iteratively applying the ADMM (Alternating Direction Method of Multipliers) technique, to obtain the receiving angle steering vector of the atomic norm minimization problem ( ), reflection angle steering vector( ), atom( A channel estimation method in a communication path environment through an intelligent reflective surface, which is a step of calculating ). Claim 2 delete Claim 3 A method for estimating a channel in a communication path environment through an intelligent reflective surface, comprising, in addition to the step of adjusting hyperparameters related to layer-specific weights of the deep learning model, which sequentially passes through a finite number of layers to produce an output value, according to the learning result of the deep learning model. Claim 4 In claim 1, the step of acquiring a received signal received at the receiver is further included, and the estimating step comprises a spatial frequency related to the received signal and the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( A method for estimating a channel in a communication path environment through an intelligent reflective surface, further comprising the step of estimating a channel in the communication path environment using ). Claim 5 In a computer program stored on a non-transient computer-readable recording medium for executing a channel estimation method in a communication path environment through an intelligent reflective surface on a computer device, said channel estimation method is based on the ADMM (Alternating Direction Method of Multipliers) technique and is a spatial frequency related to the signal arrival angle to a receiver ( ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) receiving angle steering vector in the atomic norm minimization problem of estimating ) ), reflection angle steering vector( ), atom( A step of calculating ); using a spectrum shift operation, the matrix of the atomic norm minimization problem ( Step of updating ); Lagrange multiplier of the above atomic norm minimization problem ( A step of updating ); and calculating the atomic norm minimization problem based on the above calculation and update results to calculate the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( The step of estimating ) is included, and the calculation step is to use hyperparameters related to the layer-specific weights of a deep learning model that calculates an output value by sequentially passing through a finite number of layers for iteratively applying the ADMM (Alternating Direction Method of Multipliers) technique, to calculate the receiving angle steering vector of the atomic norm minimization problem ( ), reflection angle steering vector( ), atom( A computer program that is a step of producing ). Claim 6 A computer device for executing a channel estimation method in a communication path environment through an intelligent reflective surface comprises at least one processor, wherein the at least one processor comprises a spatial frequency related to the signal arrival angle to a receiver based on the Alternating Direction Method of Multipliers (ADMM) technique. ) and spatial frequency related to the signal reflection angle of reconfigurable intelligent surfaces ( ) receiving angle steering vector in the atomic norm minimization problem of estimating ) ), reflection angle steering vector( ), atom( Calculate ) and use spectral shift operation to matrix of the above atomic norm minimization problem ( Updating ) and the Lagrange multiplier of the above atomic norm minimization problem ( Update ) and calculate the atomic norm minimization problem based on the above calculation and update results to the spatial frequency related to the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( Estimating ), and using hyperparameters related to the layer-specific weights of a deep learning model that calculates an output value by sequentially passing through a finite number of layers to iteratively apply the above ADMM (Alternating Direction Method of Multipliers) technique, the receiving angle steering vector of the atomic norm minimization problem ( ), reflection angle steering vector( ), atom( A computer device that produces ). Claim 7 delete Claim 8 In claim 6, the at least one processor is a computer device that adjusts hyperparameters related to layer-by-layer weights of the deep learning model according to the learning results of the deep learning model. Claim 9 In paragraph 6, the at least one processor acquires a received signal received at the receiver, and a spatial frequency related to the received signal and the signal arrival angle to the receiver ( ) and spatial frequency related to the signal reflection angle of the reconfigurable intelligent surface ( A computer device that estimates a channel in the above communication path environment using ).

Images