Device and method for compensating channel based on artificial neural network

Abstract

A device and method for compensating channels based on artificial neural networks. The device for compensating channels based on artificial neural networks includes: a channel learning device for learning channel characteristics; and a channel compensation device for compensating channel errors based on the learning results of the channel learning device.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of Korean Patent Application No. 10-2019-0071791 filed in the Korean Intellectual Property Office on June 17, 2019, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates to a technique for compensating for channel errors (noise, distortion) based on an artificial neural network. Background Art

[0004] Vehicle-to-Everything (V2X) communication technology is a core technology that provides traffic safety and application services to drivers. Because data loss and transmission delay are directly related to driver safety, it requires high reliability and low latency (HR / LL).

[0005] However, the Wireless Access in Vehicular Environment (WAVE) technology based on the IEEE 802.11p physical layer standard is a communication technology that improves the IEEE 802.11a / g wireless LAN technology to be suitable for the vehicle environment. Since the WAVE technology specifies the number of subcarriers and pilots suitable for the speed of pedestrian movement, it is difficult to accurately estimate the channel due to channel changes (high Doppler effect and multipath) caused by the high-speed movement of vehicles.

[0006] Although a data pilot aided (DPA) scheme, which estimates the channel of a current orthogonal frequency division multiplexing (OFDM) symbol by constructing a virtual data pilot signal based on a previous OFDM symbol, has been proposed as a scheme for estimating a rapidly changing channel in a WAVE V2X communication environment, the DPA channel estimation scheme may cause errors in the demapping process due to distortion and noise of the estimated channel. Summary of the invention

[0007] Aspects of the present disclosure provide a device for compensating channels based on artificial neural networks, which is capable of reducing the packet error rate caused by the high-speed movement of vehicles by learning the characteristics of the channel for vehicle-to-everything communications, and based on this compensation, the distortion and noise of the channel estimated in a data pilot assisted (DPA) scheme, and a method thereof is also provided.

[0008] The technical problems to be solved by the present inventive concept are not limited to the above-mentioned problems, and any other technical problems not mentioned herein will be clearly understood by those skilled in the art to which the present disclosure belongs from the following description.

[0009] According to aspects of the present disclosure, a device for compensating a channel based on an artificial neural network includes: a channel learning device for learning characteristics of a channel; and a channel compensation device for compensating errors of the channel based on a learning result of the channel learning device.

[0010] The channel learning device may learn characteristics of a channel for vehicle-to-everything (V2X) communication.

[0011] The channel learning device may learn channel correlation in the frequency domain.

[0012] The channel learning device may be implemented by an autoencoder. The autoencoder may be implemented using an input layer, three hidden layers, and an output layer. The hidden layers may include: a first hidden layer including 40 hidden units; a second hidden layer including 20 hidden units; and a third hidden layer including 40 hidden units.

[0013] The channel may include a channel estimated by a data pilot assisted (DPA) scheme.

[0014] The channel compensation device may convert the channel estimated by the DPA scheme into a vector in a complex form, and compensate for an error of the channel by applying a pre-optimized learning parameter to the vector.

[0015] The channel compensation device may sequentially compensate for errors of the channel with respect to a plurality of orthogonal frequency division multiplexing (OFDM) symbols.

[0016] According to aspects of the present disclosure, a method for compensating a channel based on an artificial neural network comprises the following steps: a channel learning device learns the characteristics of the channel; and a channel compensation device compensates for errors of the channel based on the learning results of the channel learning device.

[0017] Learning the characteristics of the channel may include learning characteristics of a channel for vehicle-to-everything (V2X) communications.

[0018] Learning the characteristics of the channel may include learning channel correlation in the frequency domain.

[0019] The channel learning device may be implemented by an autoencoder. The autoencoder may be implemented using an input layer, three hidden layers, and an output layer. The hidden layers may include: a first hidden layer including 40 hidden units; a second hidden layer including 20 hidden units; and a third hidden layer including 40 hidden units.

[0020] The channel may include a channel estimated by a data pilot assisted (DPA) scheme.

[0021] Compensating for the error of the channel may include: converting the channel estimated by the DPA scheme into a vector in a complex form; and compensating for the error of the channel by applying a pre-optimized learning parameter to the vector.

[0022] Compensating for the error of the channel may include sequentially compensating for the error of the channel with respect to a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols.

[0023] According to aspects of the present disclosure, a non-transitory computer-readable medium contains program instructions executed by a processor, the computer-readable medium including: program instructions for learning characteristics of a channel based on an artificial neural network; and program instructions for compensating for errors of the channel based on learning results of the channel learning device. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and other objects, features and advantages of the present disclosure will become more apparent through the following detailed description in conjunction with the accompanying drawings:

[0025] Figure 1 is a diagram showing a device for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure;

[0026] Figure 2 is a view showing a structure of an auto encoder provided in a device for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure;

[0027] Figure 3 is a view showing the function of a channel compensation device provided in a device for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure;

[0028] Figure 4 is a performance analysis diagram of a device based on an artificial neural network compensation channel according to an embodiment of the present invention;

[0029] Figure 5 is another performance analysis diagram of a device based on an artificial neural network compensation channel according to an embodiment of the present invention;

[0030] Figure 6 is a flow chart showing a method for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure; and

[0031] Figure 7 is a diagram showing a computing system that executes a method for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0032] It should be understood that the term "vehicle" or "vehicle-related" or other similar terms used herein generally include motor vehicles, such as passenger cars, including sport utility vehicles (SUVs), buses, trucks, various commercial vehicles, water vehicles including various boats, aircraft, etc., and include hybrid vehicles, electric vehicles, plug-in hybrid vehicles, hydrogen-powered vehicles and other alternative fuel vehicles (e.g., fuels produced from resources other than petroleum). As referred to herein, a hybrid vehicle is a vehicle having two or more power sources, such as a gasoline-powered vehicle and an electric-powered vehicle.

[0033] The terms used herein are only for the purpose of describing specific embodiments, and are not intended to limit the present disclosure. As used herein, the singular forms "one", "an" and "the" are also intended to include plural forms, unless the context clearly indicates otherwise. It will be further understood that when the terms "including" and / or "comprising" are used in this specification, it specifies the existence of the features, integers, steps, operations, elements and / or components, but does not exclude the existence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups. As used herein, the term "and / or" includes any and all combinations of one or more associated listed items. Throughout the specification, unless explicitly described to the contrary, the word "including" and variants such as "having" or "comprising" will be understood to imply the inclusion of stated elements, rather than excluding any other elements. In addition, the terms "unit", "... device", "... person" and "module" described in the specification refer to units for processing at least one function and operation, and can be implemented by hardware components or software components and combinations thereof.

[0034] In addition, the control logic of the present disclosure may be embodied as a non-transitory computer-readable medium on a computer-readable medium containing executable program instructions executed by a processor, a controller, etc. Examples of computer-readable media include, but are not limited to, ROM, RAM, compact disk (CD)-ROM, magnetic tape, floppy disk, flash drive, smart card, and optical data storage device. The computer-readable medium may also be distributed in a network-coupled computer system so that the computer-readable medium is stored and executed in a distributed manner, for example, by a telematics server or a controller area network (CAN).

[0035] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to the exemplary drawings. When adding reference numerals to the components of each drawing, it should be noted that even if the same or equivalent components are shown on other drawings, they are represented by the same reference numerals. In addition, when describing the embodiments of the present disclosure, detailed descriptions of well-known features or functions will be excluded so as not to unnecessarily make the subject matter of the present disclosure unclear.

[0036] When describing components according to embodiments of the present disclosure, terms such as first, second, "A", "B", (a), (b), etc. may be used. These terms are intended only to distinguish one component from another, and these terms do not limit the nature, sequence, or order of the constituent components. Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meanings as those generally understood by those skilled in the art to which the present disclosure belongs. Terms defined in general dictionaries should be interpreted as having the same meanings as the contextual meanings associated with the relevant art, and should not be interpreted as having ideal or overly formal meanings unless explicitly defined as having such meanings in this application.

[0037] Figure 1 is a view showing an apparatus for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure.

[0038] like Figure 1 As shown, the apparatus 100 for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure may include a channel learning device 10 and a channel compensating device 20. Depending on the scheme of implementing the apparatus 100 for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure, the components may be combined with each other and implemented as a control device (not shown), and some components may be omitted.

[0039] In this case, the control device can perform overall control so that each component performs normal functions. In addition, the control device can be implemented in the form of hardware or software, and of course can be implemented in the form of a combination of hardware and software. Preferably, the control device can be implemented as a microprocessor, but the embodiment is not limited thereto.

[0040] The control device may include a storage device (not shown). The storage device may include at least one type of storage medium of a memory of a flash memory type, a hard disk type, a micro type, a card type (e.g., a secure digital (SD) card or an extreme digital (XD) card), etc., as well as a random access memory (RAM), a static RAM, a read-only memory (ROM), a programmable ROM (PROM), an electrically erasable PROM (EEPROM), a magnetic memory (MRAM), a magnetic disk, and an optical disk type memory.

[0041] The channel learning device 10 may be implemented as, for example, an autoencoder that receives channels of vehicle-to-everything (V2X) communication as learning data and learns the characteristics of each channel. That is, the channel learning device 10 learns the correlation between channels in the frequency domain.

[0042] In this case, V2X communication (i.e., vehicle communication) refers to vehicle-to-vehicle (V2V) wireless communication between vehicles, vehicle-to-infrastructure (V2I) wireless communication wired between vehicles and infrastructure, intra-vehicle wired and wireless networks (IVN), vehicle-to-pedestrian (V2P) communication, etc.

[0043] In the following, reference will be made to Figure 2 Describes the configuration of an autoencoder.

[0044] Figure 2 is a view showing a structure of an auto encoder provided in an apparatus for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure.

[0045] In general, deep learning (or deep neural networks) is a type of machine learning that involves multiple layers of artificial neural networks (ANNs) between input and output.

[0046] In case of multi-layer artificial neural networks, learning may not be done correctly. To address this problem, an autoencoder has been proposed as a machine learning solution.

[0047] An autoencoder is a learning scheme designed to copy an input to an output, such as Figure 2 As shown, the optimal solution is searched by finding correlations between features in similar data.

[0048] When an artificial neural network is configured, the number of hidden units HI in each layer of the hidden layer is configured to be smaller than the dimension h of the input layer. 0 , so that the features of the input data can be extracted and the noise can be removed.

[0049] In order to utilize the characteristics of such an autoencoder, according to an embodiment of the present disclosure, the autoencoder is first learned by using the channel data in the frequency domain generated based on the V2X channel model in an offline mode as learning data. In this case, since the channel in the frequency domain is generated by a small number of channel taps in the time domain, there is a high correlation between the channels.

[0050] Therefore, the autoencoder can learn channel features including the correlation between channels in the frequency domain through unsupervised learning with a small number of hidden layers and hidden units.

[0051] The autoencoder thus learned can be used to remove the errors (distortion and noise) of the channel estimated by the DPA scheme.

[0052] Specifically, when considering an autoencoder with “L” layers, with d I The dimension of the lth hidden layer hl, It can be expressed by the following equation 1.

[0053] [Equation 1]

[0054]

[0055] Specifically, represents the weight matrix, denotes the deviation vector, and φ l Activation function.

[0056] As an example, the autoencoder can be implemented using one input layer, three hidden layers, and one output layer. In this case, the number of hidden units in each hidden layer can be set to 40 hidden units in the first hidden layer, 20 hidden units in the second hidden layer, and 40 hidden units in the third hidden layer.

[0057] In the following, the learning process of the autoencoder will be described.

[0058] First, a random time-domain CIR (channel impulse response) is generated based on the profile of the V2X channel, and the generated CIR is converted into a channel frequency response (CFR) in the frequency domain by fast Fourier transform (FFT). The result after conversion is represented as H.

[0059] After that, H in the transformed frequency domain is converted into an input vector (input vector). In this case, as an example, the input vector x(n) can be expressed as the following equation 2.

[0060] [Equation 2]

[0061]

[0062] Specifically, [] T represents the transpose operation, R{} represents the real part of the complex number, and J{} represents the imaginary part of the complex number.

[0063] In addition, H(n) refers to the data subcarrier (S D ) and pilot subcarrier (S P ) corresponds to the nth learning data. In this case, the guard band and the center subcarrier may be excluded from the data subcarrier.

[0064] For example, the learning data refers to 52 subcarriers (4 pilot subcarriers and 48 data subcarriers) excluding 11 guard bands and 1 DC subcarrier among a total of 64 OFDM subcarriers.

[0065] H (n) It can be expressed, for example, according to Equation 3.

[0066] [Equation 3]

[0067]

[0068] On the other hand, the output vector of the autoencoder It can be expressed by the following equation 4.

[0069] [Equation 4]

[0070]

[0071] The autoencoder optimizes the learning parameters based on N learning data. To minimize the mean square error, as shown in Equation 5 below.

[0072] [Equation 5]

[0073]

[0074] For reference, during the learning process, the learning rate can be set to 0.0001, the Adam optimizer can be used to calculate the MSE, and the parameters can be initialized by the Xavier technique.

[0075] Thereafter, the learned autoencoder can be utilized in the channel estimation process of the IEEE 802.11p system.

[0076] Next, the channel compensation device 20 compensates for noise and distortion of the channel estimated by the channel estimation device 200 based on the result learned by the channel learning device 10. In this case, the channel estimation device 200 may estimate the channel in the DPA scheme.

[0077] In the following, reference will be made to Figure 3 The function of the channel compensation device 20 is described.

[0078] Figure 3 1 is a diagram showing the function of a channel compensation device provided in an apparatus for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure, wherein an error compensation process is performed from a preamble to a last OFDM symbol. (Data I)

[0079] like Figure 3 As shown, data 1 to data L represent OFDM symbols, and an initial channel estimation value estimated through an initial channel estimation process can be used to estimate a channel of DATA 1 .

[0080] In this example, the initial channel can be estimated by using two long learning symbols of the preamble. In this case, the initial channel estimate is It can be expressed by the following equation 6.

[0081] [Equation 6]

[0082]

[0083] Specifically, and represents two long-term learning symbols, and X 0 (k) represents the long-term learned symbol known at the receiving end.

[0084] Then, the data recovery device 300 uses the i-th received signal Y i (k) and the channel H of the i-1th symbol i-1 (k) to recover the data of the data subcarrier of the i-th OFDM symbol This can be expressed by the following equation 7.

[0085] [Equation 7]

[0086]

[0087] Thereafter, the channel estimation apparatus 200 uses the data recovered by the data recovery apparatus 300 to and four pilot subcarriers S P The corresponding X i (k) is used to estimate the channel of the current OFDM symbol. In this case, the estimated channel However, it is expressed by the following equation 8.

[0088] [Equation 8]

[0089]

[0090] Thereafter, the channel compensation device 20 converts the channel estimated by the channel estimation device 200 into As an example, the converted vector can be expressed by the following equation 9.

[0091] [Equation 9]

[0092]

[0093] Then, the channel compensation device 20 applies the optimized learning parameters to the complex type vector based on the result learned by the channel learning device 10 to the channel estimated by the channel estimation device 200. The compensation result can be expressed by the following equation 10.

[0094] [Equation 10]

[0095]

[0096] Afterwards, After the integration is in complex form, the channel compensation device 20 can be used to estimate the channel of the next OFDM symbol.

[0097] Since the estimation process of the initial channel and the channel estimation process performed by the channel estimation apparatus 200 in the above-mentioned channel compensation process are well-known technologies, they are not the gist of the present disclosure and are therefore briefly described.

[0098] Figure 4 is a performance analysis diagram of a device for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure. Figure 4 The packet error rate in an environment where the relative speed between vehicles is 126 km / h is shown.

[0099] like Figure 4 As shown, each graph represents a packet error rate when the embodiments of the present disclosure are not applied and a packet error rate when the embodiments of the present disclosure are applied.

[0100] pass Figure 4 It can be understood that the packet error rate when applying the (proposed) implementation of the present disclosure is much smaller than the packet error rate when the implementation of the present disclosure is not applied, that is, spectrum time averaging, constructed data pilot (CDP), time domain reliable test frequency domain interpolation (TRFI) and iterative channel estimation and decoding (ICED) are respectively applied as a kind of DPA scheme.

[0101] Figure 5 is another performance analysis diagram of the apparatus for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure, wherein a packet error rate in an environment where the relative speed between vehicles is 252 km / h is shown.

[0102] like Figure 5 As shown, each graph represents a packet error rate when the embodiments of the present disclosure are not applied and a packet error rate when the embodiments of the present disclosure are applied.

[0103] pass Figure 5 It can be understood that the packet error rate when applying the (proposed) implementation of the present disclosure is much smaller than the packet error rate when the implementation of the present disclosure is not applied, that is, spectrum time averaging, constructed data pilot (CDP), time domain reliable test frequency domain interpolation (TRFI) and iterative channel estimation and decoding (ICED) are respectively applied as a kind of DPA scheme.

[0104] Meanwhile, the embodiments of the present disclosure may be implemented in an 802.11p chipset structure. In this case, the embodiments may be implemented to perform a channel compensation process in a low-speed driving case rather than a high-speed driving case based on the speed of the vehicle.

[0105] Figure 6Detailed description is a flow chart showing a method for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure.

[0106] First, in operation 601, the channel learning device 10 learns characteristics of a channel.

[0107] Thereafter, in operation 602 , the channel compensation apparatus 20 compensates for an error of a channel estimated by the channel estimation apparatus 200 based on a learning result of the channel learning apparatus 10 .

[0108] Figure 7 is a diagram showing a computing system that executes a method for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure.

[0109] Reference Figure 7 The method for compensating channels based on artificial neural networks according to the embodiments of the present disclosure can be implemented by a computing system. The computing system 1000 may include at least one processor 1100, a memory 1300, a user interface input device 1400, a user interface output device 1500, a memory 1600, and a network interface 1700 connected to each other via a bus 1200.

[0110] The processor 1100 may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in the memory 1300 and / or the storage device 1600. The memory 1300 and the storage device 1600 may include various types of volatile or non-volatile storage media. For example, the memory 1300 may include a ROM (Read Only Memory) and a RAM (Random Access Memory).

[0111] Therefore, the operation of the method or algorithm described in conjunction with the embodiments disclosed herein may be directly embodied in a hardware or software module executed by the processor 1100, or in a combination thereof. The software module may reside on a storage medium such as RAM, flash memory, ROM, EPROM, EEPROM, register, hard disk, removable disk, CD-ROM (i.e., memory 1300 and / or storage manufacturing 1600). The exemplary storage medium may be coupled to the processor 1100, and the processor 1100 may read information from the storage medium and may record information in the storage medium. Alternatively, the storage medium may be integrated with the processor 1100. The processor 1100 and the storage medium may reside in an application specific integrated circuit (ASIC). The ASIC may reside in a user terminal. In another case, the processor 1100 and the storage medium may reside in a user terminal as separate components.

[0112] According to an embodiment of the present disclosure, an apparatus and method for compensating a channel based on an artificial neural network can learn the characteristics of a V2X communication channel, and can compensate for the distortion and noise of the channel estimated by the DPA scheme based on the channel characteristics, thereby reducing the packet error rate caused by the high-speed movement of the vehicle.

[0113] In addition, the device and method for compensating channels based on artificial neural networks according to an embodiment of the present disclosure can be combined with a DPA channel estimation scheme and an autoencoder, which is one of the unsupervised learning schemes of neural networks, so as to compensate for the weaknesses of the DPA scheme with lower complexity and significantly improve the accuracy of channel estimation.

[0114] In addition, the apparatus and method for compensating a channel based on an artificial neural network according to an embodiment of the present disclosure can learn interference and distortion of a wireless channel that is difficult to mathematically analyze through a learning scheme of an autoencoder.

[0115] Although the present disclosure has been described above with reference to exemplary embodiments and the accompanying drawings, the present disclosure is not limited thereto, and those skilled in the art to which the present disclosure belongs may make various modifications and changes to the present disclosure without departing from the spirit and scope of the present disclosure as claimed in the appended claims.

[0116] Therefore, the exemplary embodiments of the present disclosure are provided to explain the spirit and scope of the present disclosure, but not to limit them, and therefore, the spirit and scope of the present disclosure are not limited by the embodiments. The scope of the present disclosure should be interpreted based on the attached claims, and all technical concepts within the scope equivalent to the claims should be included in the scope of the present disclosure.

Claims

1. A device for compensating a channel based on an artificial neural network, the device include: a channel learning device configured to learn characteristics of a channel; as well as a channel compensation device configured to compensate for an error of a channel estimated by a data pilot aided scheme based on a learning result of the channel learning device, The channel compensation device is configured to convert the channel estimated by the data pilot-aided scheme into a vector in complex form, and compensate for the error of the estimated channel by applying the pre-optimized learning parameters from the channel learning device to the vector.

2. The device according to claim 1, in, The channel learning device is configured to learn characteristics of a channel for vehicle-to-everything communications.

3. The device according to claim 1, in, The channel learning device is configured to learn channel correlation in the frequency domain.

4. The device according to claim 1, in, The channel learning device is implemented by an automatic encoder.

5. The device according to claim 4, in, The autoencoder is implemented with one input layer, three hidden layers, and one output layer.

6. The device according to claim 5, in, The hidden layers include: a first hidden layer including 40 hidden units; a second hidden layer including 20 hidden units; and a third hidden layer including 40 hidden units.

7. The device according to claim 1, in, The channel compensation device is configured to sequentially compensate for errors of the channel with respect to a plurality of orthogonal frequency division multiplexing symbols.

8. A method for compensating a channel based on an artificial neural network, the method The following steps are involved: The channel learning device learns the characteristics of the channel; as well as The channel compensation device compensates for the error of the channel estimated by the data pilot auxiliary scheme based on the learning result of the channel learning device. Wherein, compensating for the error of the channel estimated by the data pilot auxiliary scheme comprises: converting the channel estimated by the data pilot-aided scheme into a vector in complex form; and The error of estimating the channel is compensated by applying the pre-optimized learning parameters from the channel learning device to the vector.

9. The method according to claim 8, in, Learning the characteristics of the channel includes learning the characteristics of the channel for vehicle-to-everything communications.

10. The method according to claim 8, in, Learning the characteristics of the channel includes learning channel correlation in the frequency domain.

11. The method according to claim 8, in, The channel learning device is implemented by an automatic encoder.

12. The method according to claim 11, in, The autoencoder is implemented with one input layer, three hidden layers, and one output layer.

13. The method according to claim 12, in, The hidden layers include: a first hidden layer including 40 hidden units; a second hidden layer including 20 hidden units; and a third hidden layer including 40 hidden units.

14. The method according to claim 8, in, Compensating for the error of the channel includes sequentially compensating for the error of the channel with respect to a plurality of orthogonal frequency division multiplexing symbols.

15. A non-transitory computer-readable medium comprising program instructions executed by a processor, wherein when the program instructions are executed by the processor, the processor executes the method for compensating channels based on an artificial neural network according to any one of claims 8 to 14.

Images