MIMO-SCMA decoding method and system used between PMS, terminal and computer medium

Abstract

A MIMO-SCMA decoding method and system used between PMS, a terminal and a computer medium, the method comprising: under a random transmission symbol combination, detecting a received signal of a downlink in a PMS communication system based on MIMO-SCMA to decompose multiple paths of estimation signals; extracting an independent channel feature for each path of estimation signal from the multiple paths of estimation signals, and constructing an N-point broken line graph according to the independent channel features; aiming at a plurality of N-point broken line graphs corresponding to multiple random transmission symbol combinations, constructing an image training set, and training a graph neural network suitable for downlink signal decoding of the PMS communication system by utilizing the image training set; and sequentially adding type labels to the feature points in the N-point broken line graph of the current received signal by using a pre-trained graph neural network, and decoding the current received signal according to the type label corresponding to each feature point and the preset codebook combination corresponding to the current received signal.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and more specifically, to a MIMO-SCMA decoding method, system, terminal, and computer medium for between PMSs. Background Art

[0002] A power management system (PMS) is a key technical device applied to the intelligent scenario of electric energy storage, which realizes the large-scale storage system and the coordinated control of energy storage devices. The control object of this device is the power conversion system (PCS) equipment in various energy storage power stations. By adjusting the active power output and reactive power output as technical means, it realizes the continuous adjustment and coordinated control of the active power and reactive power at the grid connection point. However, the PMS needs to receive the total power instruction from the dispatching side first before calculating and allocating specific power instructions to each PCS connected to it. The traditional communication method between the dispatching and the PMS is wired communication, that is, relying on the wired connection of cables or optical fibers. However, this communication method is affected by the terrain and landform, greatly limiting the communication distance between the dispatching and the PMS, and the maintenance cost is too high.

[0003] Sparse code multiple access (SCMA) is an emerging wireless non-orthogonal multiple access (NOMA) technology, which has attracted much attention due to its advantages such as strong anti-interference ability, high frequency utilization rate, low complexity, and low latency. Multiple-input multiple-output (MIMO) technology has been proven to effectively increase the spectrum utilization rate of communication systems. Setting the MIMO-SCMA technology as the wireless communication technology between the dispatching and the PMS not only has the advantages of low bit error rate (BER), low complexity, and low latency, but also meets the requirement of the energy storage system for the rapid response of the dispatching side instructions, thus comprehensively promoting the application of wireless communication technology in the energy storage engineering scenario.

[0004] In view of this, the present invention provides a MIMO-SCMA decoding method, system, terminal, and computer medium for between the dispatching and the PMS. Summary of the Invention

[0005] To address the deficiencies in the existing technologies, the present invention provides a MIMO-SCMA decoding method, system, terminal, and computer medium for scheduling and PMS. By establishing a downlink PMS communication system based on MIMO-SCMA, a wireless control strategy with low BER, low complexity, and low latency for scheduling to PMS is achieved.

[0006] The present invention adopts the following technical solutions.

[0007] In a first aspect of the present invention, a MIMO-SCMA decoding method for PMS is involved. The method includes the following steps: Under random transmission symbol combinations, detecting the received signal in the downlink of the PMS communication system based on MIMO-SCMA to decompose multiple estimated signals; extracting independent channel features for each estimated signal from the multiple estimated signals, and constructing an N-point broken line graph based on the independent channel features; for multiple N-point broken line graphs corresponding to multiple random transmission symbol combinations, constructing an image training set, and training a graph neural network applicable to decoding the downlink signals of the PMS communication system using the image training set; using the pre-trained graph neural network, sequentially adding type labels to the feature points in the N-point broken line graph of the current received signal, and decoding the current received signal based on the type label corresponding to each feature point and the preset codebook combination corresponding to the current received signal.

[0008] Preferably, under random transmission symbol combinations, detecting the received signal in the downlink of the PMS communication system based on MIMO-SCMA to obtain multiple estimated signals includes: Collecting the received signal in the downlink of the PMS communication system based on MIMO-SCMA, and obtaining each estimated signal after MIMO detection n t is the number of transmit antennas in the MIMO system; for each estimated signal performing complex decomposition, and mapping the estimated signal into the complex space according to the real and imaginary part values of the estimated signal to obtain the constellation coordinates of each estimated signal

[0009] Preferably, extracting independent channel features for each estimated signal from the multiple estimated signals, and constructing an N-point broken line graph based on the independent channel features includes: For the number of each estimated signal , sequentially performing the following coordinate transformation on the constellation coordinates to obtain the N-point broken line coordinates:

[0010]

[0011] [X 1 ,Y1 = [x 1 , y 1

[0012] Wherein, X, Y, x, and y are all N-dimensional vectors;

[0013] is the k-th decoded component of each estimated signal Among them, the number of decoded components of each estimated signal is N;

[0014] RF is the radio frequency in the direction of signal power.

[0015] Preferably, extract the independent channel features for each estimated signal from the multiple estimated signals, and construct an N-point broken line graph based on the independent channel features, including: the color of the k-th broken line coordinate in the N-point broken line graph is related to the channel characteristics of the communication channel where the k-th broken line coordinate is located; the transmitting antenna n t and the receiving antenna n r The channel characteristics between them are Then calculate the reference quantity:

[0016]

[0017] Determine the color of the k-th broken line coordinate according to the value range of the reference quantity.

[0018] Preferably, extract the independent channel features for each estimated signal from the multiple estimated signals, and construct an N-point broken line graph based on the independent channel features, including: the current estimated signal The type label corresponding to the current N-point broken line graph is:

[0019]

[0020] Wherein, and Are respectively the decoded components The N-point broken line coordinates of and The encoded values after pixelization operation through 32×32 channels;

[0021] Is the color label of the feature point.

[0022] Preferably, for the multiple N-point broken line graphs corresponding to multiple random transmission symbol combinations, construct an image training set, and use the image training set to train a graph neural network applicable to the downlink signal decoding of the PMS communication system, including: for the current random transmission symbol combination Code t , obtain the corresponding type label corresponding to the N-point broken line graph​ Randomly generate multiple combinations of random transmission symbols and obtain multiple corresponding type tags; construct an N-point polyline image pixelated input image with each combination of random transmission symbols, and use a simplified MobileNet neural network to train the input image; optimize the simplified MobileNet neural network with the goal of minimizing the cross-entropy loss between the target type tag and the type tag pre-computed through the N-point polyline graph to obtain the pre-trained graph neural network.

[0023] Preferably, using the pre-trained graph neural network, sequentially add type tags to the feature points in the N-point polyline graph of the currently received signal once, and perform decoding on the currently received signal based on the type tag corresponding to each feature point and the preset codebook combination corresponding to the currently received signal once, including: input the currently received signal once into the pre-trained graph neural network to obtain the type tag of the currently received signal once; according to the one-to-one correspondence between the type tag and the preset MIMO-SCMA codebook, respectively perform decoding on the received signals divided under the same type tag, so as to obtain the decoding result of the currently received signal once.

[0024] In the second aspect of the present invention, it relates to a MIMO-SCMA decoding system for between PMSs; the system is implemented by using a MIMO-SCMA decoding method for between PMSs described in the first aspect of the present invention; the system includes a decomposition module, an extraction module, a training module, and a decoding module; wherein, the decomposition module is used to detect the received signal in the downlink of the PMS communication system based on MIMO-SCMA under a random transmission symbol combination to decompose multiple estimated signals; the extraction module is used to extract the independent channel features for each estimated signal from the multiple estimated signals and construct an N-point polyline graph based on the independent channel features; the training module is used to construct an image training set for multiple N-point polyline graphs corresponding to multiple random transmission symbol combinations, and use the image training set to train a graph neural network applicable to decoding the downlink signal of the PMS communication system; the decoding module is used to use the pre-trained graph neural network to sequentially add type tags to the feature points in the N-point polyline graph of the currently received signal once, and perform decoding on the currently received signal once based on the type tag corresponding to each feature point and the preset codebook combination corresponding to the currently received signal once.

[0025] In the third aspect of the present invention, it relates to a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1-8.

[0026] In a fourth aspect of the present invention, there is provided a computer-readable storage medium having stored thereon a computer program, characterized in that when the program is executed by a processor, the steps of the method in the first aspect of the present invention are implemented.

[0027] In a third aspect of the present invention, there is provided a terminal, including a processor and a storage medium; the storage medium is used for storing instructions; the processor is used for operating according to the instructions to execute the steps of the method described in the first aspect of the present invention.

[0028] In a fourth aspect of the present invention, there is provided a computer-readable storage medium having stored thereon a computer program, and when the program is executed by a processor, the steps of the method described in the first aspect of the present invention are implemented.

[0029] The beneficial effects of the present invention are as follows. Compared with the prior art, a MIMO-SCMA decoding method, system, terminal and computer medium for scheduling with PMS in the present invention establish a downlink PMS communication system based on MIMO-SCMA, and achieve a wireless control strategy with low BER, low complexity and low latency for scheduling with PMS. The present invention proposes a decoding algorithm with low BER, low complexity and low latency, realizes the wireless control strategy for scheduling with PMS, and effectively promotes the application of wireless communication technology in the energy storage engineering scenario.

[0030] The beneficial effects of the present invention further include:

[0031] 1. The present invention adds channel characteristics to the N-point broken line graph, can fully consider the real-time influence of channels such as channel fading and multipath effects during the signal classification and decoding process, obtain better decoding results, and reduce the decoding error rate. Since the communication channel exhibits different decoding characteristics when the data content is different. For this reason, the present invention randomly generates a plurality of different random transmission symbol combinations and image data sets to try to traverse all possible data transmission situations, and accordingly obtains a more stable pre-trained graph neural network.

[0032] 2. The present invention decodes the computational complexity of different MIMO-SCMA decoding methods in a 2×2 channel. The decoding method of the present invention has the lowest computational complexity. As the number of antennas increases, the computational complexity advantage becomes more obvious. The decoding method of the present invention successfully realizes a wireless control strategy with low BER, low complexity and low latency for scheduling with PMS, avoids problems such as high maintenance costs and limited communication distance brought by traditional wired communication methods, and promotes the application of wireless communication technology in the energy storage engineering field. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic flowchart of a MIMO-SCMA decoding method for PMS in the present invention;

[0034] Figure 2 Schematic diagram of a MIMO-SCMA communication system for PMSs according to the present invention;

[0035] Figure 3 Schematic diagram of implementing decoding through an N-dimensional broken line graph and MobileNet in a MIMO-SCMA decoding method for PMSs according to the present invention;

[0036] Figure 4 Schematic diagram of the structural information of MobileNet in a MIMO-SCMA decoding method for PMSs according to the present invention;

[0037] Figure 5 BER schematic diagram of the decoding method under different channel numbers and different noise power spectral densities in a MIMO-SCMA decoding method for PMSs according to the present invention;

[0038] Figure 6 Comparison graph of the algorithm complexity between a MIMO-SCMA decoding method for PMSs according to the present invention and existing methods. Detailed implementation manners

[0039] To make the objectives, technical solutions and advantages of the present invention clearer and more accurate, the technical solutions of the present invention are described in detail below through multiple specific implementation manners. The embodiments adopted in the present invention are only used to explain the present invention and do not limit the content of the present invention.

[0040] Figure 1 Flow schematic diagram of a MIMO-SCMA decoding method for PMSs according to the present invention. In the first aspect of the present invention, a MIMO-SCMA decoding method for PMSs is involved, which is characterized in that the method includes Step 1 to Step 4.

[0041] Step 1, under a random transmission symbol combination, detect the received signal of the downlink in a PMS communication system based on MIMO-SCMA to decompose multiple estimated signals.

[0042] Figure 2 Schematic diagram of a MIMO-SCMA communication system for PMSs according to the present invention. As Figure 2 , in one embodiment, the number of independent PMSs J = 6, the number of orthogonal resource blocks N = 4, the codebook size M = 4, the number of transmit antennas N t = 2, and the number of receive antennas N r = 2.

[0043] The decoding method (LG-ESDS, Lightweight GNNs Based Enhanced Separated Detection Scheme) in the present invention has a process as follows Figure 1 As shown, the received signal y of the downlink PMS communication system based on MIMO-SCMA j After MIMO detection, the estimated signal is obtained Among them, the received signal y j There are a total of J PMSs, j = 1, 2,..., J, and each PMS has M kinds of transmission symbols. The splitting of the estimated signal is implemented according to the total N t Transmit antennas, numbered n t = 1, 2,..., N t , and the N components describe the N resource blocks divided by channel multiplexing.

[0044] Thus, the received signal is expressed as:

[0045] y j = H j x + n j

[0046] In the formula, H j Is the channel matrix, x is the set of overlapping codewords sent by all transmit antennas, and n j Is complex Gaussian noise, with a set mean of 0 and a variance of σ 2 . MIMO detection selects minimum mean square error (MMSE) detection. The estimated signal obtained by splitting is expressed as:

[0047]

[0048] In the formula, Is decomposed into an N-dimensional vector through non-orthogonal decomposition.

[0049] Preferably, under a random transmission symbol combination, the received signal of the downlink in the PMS communication system based on MIMO-SCMA is detected to obtain multiple estimated signals, including: collecting the received signal of the downlink in the PMS communication system based on MIMO-SCMA, and obtaining each estimated signal after MIMO detection n t Is the transmit antenna number in the MIMO system; for each estimated signal Perform complex decomposition, and map the estimated signal to the complex space according to the real and imaginary parts of the estimated signal to obtain the constellation coordinates of each estimated signal

[0050] Because each PMS in the downlink PMS communication system decodes independently, it is necessary to perform spatial splitting on each estimated signal to obtain the N-point constellation coordinates after spatial splitting of the estimated signal This coordinate is the decomposition result of the estimated signal in the constellation diagram

[0051] Step 2: Extract the independent channel characteristics for each estimated signal from the multiple estimated signals, and construct an N-point broken line diagram based on the independent channel characteristics

[0052] Preferably, extracting the independent channel characteristics for each estimated signal from the multiple estimated signals and constructing an N-point broken line diagram based on the independent channel characteristics includes: for each estimated signal number, perform the following coordinate transformation on the constellation coordinates in sequence to obtain the N-point broken line coordinates

[0053]

[0054] [X 1 ,Y 1 = [x 1 ,y 1

[0055] wherein, X, Y, x, and y are all N-dimensional vectors

[0056] is the k-th decoding component in each estimated signal , and the number of decoding components of each estimated signal is N

[0057] RF is the radio frequency in the signal power direction

[0058] Furthermore, map the real part and the imaginary part of the first component of the estimated signal to the plane rectangular coordinate system to obtain the coordinate point c 1 = [X 1 ,Y 1 . Based on this, the coordinates of the coordinate point c 2 are obtained as follows

[0059]

[0060] wherein, (X 2 ,Y 2 ) are the coordinates of the coordinate point c 2 , and are respectively the second component of the estimated signal ​​ The real and imaginary parts, where RF represents the right shift factor, ensuring that c 2 At c 1 On the right side of

[0061] From this, the coordinates of the remaining K - 2 coordinate points can be obtained, and the process can be expressed as:

[0062]

[0063] In the formula, (X z+1 , Y z+1 ) and (X z , Y z ) respectively represent the coordinates of c z+1 and c z (z = 2, 3,..., N - 1),

[0064] And respectively represent the real and imaginary parts of the (z + 1)-th component of the estimated signal , where RF ensures that c z+1 is on the right side of c z .

[0065] By calculating the channel characteristics and using the matplotlib.colors module and colormap class in the matplotlib library of PyToch, set the color of the k-th coordinate point obtained in the first step. The matplotlib.colors module and colormap class can map the values between 0 and 1 to colors, and the value corresponding to the color of the k-th coordinate point can be expressed as:

[0066]

[0067] From this, set the colors and positions of the multiple coordinate points obtained in the above steps, connect the multiple coordinate points in the plane rectangular coordinate system in sequence, generate an N-point line graph with three RGB channels, and determine the one-to-one correspondence between the input symbol combination sent at the transmitting antenna and the N-point line graph in terms of category.

[0068] Channel characteristics can be obtained based on the physical characteristics of the wireless channels between multiple PMSs, such as channel fading, multipath effects, etc. of each wireless channel. Channel fading refers to the phenomenon that the signal power attenuates due to various reasons during the transmission process. And multipath effect refers to the phenomenon that when the signal is transmitted, due to the existence of multiple propagation paths, time delay and phase distortion occur when the signal reaches the receiving end. Channel fading can be divided into large-scale fading and small-scale fading. Large-scale fading refers to the fading caused by factors such as the blockage between transmission paths and the occlusion of objects, and its fading characteristics are mainly determined by environmental factors. Small-scale fading refers to the rapid power attenuation of the signal within a short period of time, mainly affected by the multipath effect. The multipath effect will cause time delay and phase distortion when the signal reaches the receiving end. When the signal reaches the receiving end through multiple propagation paths, the signals on different propagation paths will reach the receiving end with different time delays. These signals with different time delays are superimposed together, which will cause signal interference and aliasing, thus affecting the demodulation and recovery of the signal at the receiving end. In wireless transmission, it is necessary to consider the impact of channel fading and multipath effects on signal transmission to improve the reliability and performance of the system. For channel fading, power compensation technology can be used to compensate for the signal power loss caused by fading; for the multipath effect, equalization technology and multi-carrier technology can be used to reduce time delay and phase distortion.

[0069] Therefore, adding channel characteristics to the N-point line graph can fully consider characteristics such as channel fading and multipath effects during the signal classification and decoding process, obtain better decoding results, and reduce the decoding error rate.

[0070] Connect c 1 ,c 2 ,…,c N in sequence. The connecting lines are set to black, and an N-point line graph can be obtained.

[0071] Figure 3 This is a schematic diagram of implementing decoding through an N-point line graph and MobileNet in a MIMO-SCMA decoding method for PMSs according to the present invention. As Figure 3 shown, in one embodiment, the output format of the line graph is set to 3*32*32, where "3" represents the number of color channels (i.e., red, green, and blue), and "32*32" specifies the width and height of the image in pixels. According to the resolution of the N-point line graph, the type label corresponding to each path of signal can be obtained.

[0072] Preferably, extracting independent channel characteristics for each path of estimated signal from the multiple paths of estimated signals, and constructing an N-point line graph based on the independent channel characteristics, includes: the current path of estimated signal The type label corresponding to the current N-point line graph corresponding to is:

[0073] In the formula, and are respectively the N - point broken - line coordinates of the decoding component and the encoded value after pixelization operation through 32×32 channels; is the color label of the feature point.

[0074] Step 3, for multiple N - point broken - line graphs corresponding to multiple randomly transmitted symbol combinations, construct an image training set, and use the image training set to train a graph neural network applicable to the downlink signal decoding of the PMS communication system.

[0075] In one embodiment, the input of the graph neural network is an N - point broken - line image pixelated into a 3*32*32 pixel map. In other embodiments, the input of the neural network can also be the type label mentioned above.

[0076] Preferably, for the current randomly transmitted symbol combination Code t , obtain the corresponding type label Randomly generate multiple randomly transmitted symbol combinations, and obtain multiple corresponding type labels; construct an input image after pixelization of an N - point broken - line image for each randomly transmitted symbol combination, and use a simplified MobileNet neural network to train the input image; optimize the simplified MobileNet neural network with the goal of minimizing the cross - entropy loss between the target type label and the type label pre - calculated through the N - point broken - line graph, so as to obtain the pre - trained graph neural network.

[0077] Since different data contents result in different decoding characteristics of the communication channel. Therefore, the present invention randomly generates multiple different randomly transmitted symbol combinations to try to traverse all possible data transmission situations, and accordingly obtains a more stable pre - trained graph neural network.

[0078] Therefore, based on the one - to - one correspondence in category between the transmitted symbol combination sent at the transmitting antenna n t and the N - point broken - line graph, various randomly transmitted symbol combinations are generated proportionally at the transmitting antenna n t to generate an image training set containing various N - point broken - line graphs, and it is used for training the lightweight graph neural network.

[0079] Figure 4 is a schematic diagram of the structure information of MobileNet in a MIMO - SCMA decoding method for PMS of the present invention. The graph neural network adopts a simplified MobileNet model, and the specific structure is as Figure 4As shown in the figure. The simplified MobileNet model successively includes multiple depth convolution layers and point convolution layers connected alternately. The input layer is a fully convolutional layer, and the output layer is an average pooling, fully connected, and Softmax layer. This convolutional layer supports 4096 signal classifications.

[0080] The Stochastic Gradient Descent (SGD) optimizer is used with a learning rate set to 0.002 and a momentum set to 0.9 to minimize the loss function to obtain the network parameters.

[0081] The loss function is:

[0082]

[0083] In the formula, L(·) is the cross-entropy loss function, p is the output of the fully connected layer of the graph neural network, and b is the label assigned to the N-point broken line graph of the specific class. The codebook used for input end encoding is the preset codebook for each PMS.

[0084] Step 4: Using the pre-trained graph neural network, successively add type labels to the feature points in the N-point broken line graph of the currently received signal, and perform decoding on the currently received signal according to the type label corresponding to each feature point and the preset codebook combination corresponding to the currently received signal.

[0085] Input the currently received signal into the pre-trained graph neural network to obtain the type label of the currently received signal; then, according to the one-to-one correspondence between the type label and the preset MIMO-SCMA codebook, perform decoding on the received signals classified under the same type label respectively, so as to obtain the decoding result of the currently received signal.

[0086] At the transmitting antenna n t Generate various random input symbol combinations proportionally to generate an image training set containing various N-point broken line graphs, and use it for training the graph neural network. Thus, the decoding work after MIMO detection in the complex downlink PMS communication system is converted into an image classification task that can be executed by a lightweight graph neural network.

[0087] In one embodiment, after classifying the transmission symbol combinations, the graph neural network also undertakes the decoding work. In this embodiment, a one-to-one correspondence is established in advance between the N-point broken line graph with a determined format and the transmission symbol combination that generates it, and the label of the N-point broken line graph is set to a one-hot vector with a length of M J where M represents that each PMS has M transmission symbols, for example, 0, 1, … M - 1, corresponding to M codewords of the specific codebook of each PMS. The index of the element with a value of 1 in the one-hot vector is set to:

[0088]

[0089] Where m j represents the transmission symbol of the j-th PMS in the transmission compliance combination corresponding to the N-point broken line diagram.

[0090] Figure 5 This is the BER schematic diagram of the decoding method under different channel numbers and different noise power spectral densities in a MIMO-SCMA decoding method for PMSs according to the present invention.

[0091] In the figure, S, M, and B respectively represent the power spectral density ratios of different signals and noises. S is the training model with an E b / N 0 value of 6 dB. M is the training model with an E b / N 0 value of 8 dB. B is the training model with an E b / N 0 value of 10 dB. E b and N 0 are the energy and the power spectral density of the noise respectively.

[0092] After training the model using each of the above strategies, Figure 5 a and Figure 5 b respectively show the BER performance of the decoding method of the present invention under 2×2 and 4×4 channels. The simulation results show that compared with other strategies, M is the optimal training strategy, so the training model with an E b / N 0 value of 8 dB is selected.

[0093] Figure 5 c and Figure 5 d respectively compare the BER performance of our decoding method with different MIMO-SCMA decoding methods under 2×2 and 4×4 channels. At different E b / N 0 values, the decoding method of the present invention is always superior to the MMSE + Message Passing Algorithm (MPA), and also achieves a lower BER than the Synthesis MPA (SMPA) and the Improved Maximum Distance MPA (IMDMPA) at high E b / N 0 values.

[0094] Figure 6 This is the comparison diagram of the algorithm complexity between a MIMO-SCMA decoding method for PMSs according to the present invention and existing methods. As Figure 6As shown, the computational complexity of the decoding method of the present invention and different MIMO-SCMA decoding methods under a 2×2 channel is such that the decoding method of the present invention has the lowest computational complexity. As the number of antennas increases, the advantage of the computational complexity of the decoding method of the present invention will become more obvious.

[0095] In summary, the present invention has successfully realized a wireless control strategy for PMS with low BER, low complexity, and low latency in scheduling, avoiding problems such as high maintenance costs and limited communication distance brought by traditional wired communication methods, and promoting the application of wireless communication technology in the energy storage engineering field.

[0096] In a second aspect of the present invention, there is provided a MIMO-SCMA decoding system for between PMSs; the system is implemented by using a MIMO-SCMA decoding method for between PMSs in the first aspect of the present invention; the system includes a decomposition module, an extraction module, a training module, and a decoding module; wherein, the decomposition module is configured to perform detection on the received signal of the downlink in a PMS communication system based on MIMO-SCMA under a random transmission symbol combination to decompose multiple estimated signals; the extraction module is configured to extract independent channel features for each of the multiple estimated signals and construct an N-point broken line graph based on the independent channel features; the training module is configured to construct an image training set for multiple N-point broken line graphs corresponding to multiple random transmission symbol combinations, and train a graph neural network applicable to decoding the downlink signal of the PMS communication system by using the image training set; the decoding module is configured to use the pre-trained graph neural network to sequentially add type labels to the feature points in the N-point broken line graph of the currently received signal, and perform decoding on the currently received signal based on the type label corresponding to each feature point and the preset codebook combination corresponding to the currently received signal.

[0097] In a third aspect of the present invention, there is provided a terminal, including a processor and a storage medium; the storage medium is used for storing instructions; the processor is configured to operate according to the instructions to execute the steps of the method in the first aspect of the present invention.

[0098] In a fourth aspect of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the method in the first aspect of the present invention.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that there are still contents in the technical solutions of the present invention that can be modified or equivalently replaced. Any modification or equivalent replacement without departing from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A MIMO-SCMA decoding method for PMSs, characterized in that: The method comprises the following steps: Under random transmission symbol combinations, the downlink received signal of the PMS communication system based on MIMO-SCMA is detected to decompose the multi-path estimated signals. Extracting independent channel features for each estimated signal from the multi-path estimated signals, and constructing an N-point line graph according to the independent channel features; For multiple N-point line graphs corresponding to multiple random transmission symbol combinations, an image training set is constructed, and a graph neural network suitable for downlink signal decoding of a PMS communication system is trained using the image training set; Using the pre-trained graph neural network, type labels are added to the feature points in the N-point line graph of the current received signal in sequence, and the current received signal is decoded based on the type label corresponding to each feature point and the preset codebook combination corresponding to the current received signal.

2. The MIMO-SCMA decoding method for PMSs according to claim 1, characterized in that: The method of detecting a received signal of a downlink in a PMS communication system based on MIMO-SCMA under a random transmission symbol combination to obtain a multipath estimation signal includes: Collect the received signal of the downlink in the PMS communication system based on MIMO-SCMA, and obtain the estimated signal of each channel after MIMO detection n t Number the transmitting antennas in the MIMO system; For each estimated signal Perform complex decomposition and map the estimated signal into complex space according to the values ​​of the real and imaginary parts of the estimated signal to obtain each estimated signal Constellation coordinates 3. The MIMO-SCMA decoding method for PMSs according to claim 2, characterized in that: The extracting the independent channel feature for each estimated signal from the multi-path estimated signals and constructing an N-point line graph according to the independent channel feature comprises: For each estimated signal The constellation coordinates are Implement the following coordinate transformation to obtain the coordinates of the N-point polyline: [X1,Y1]=[x1,y1] Where X, Y, x and y are all N-dimensional vectors; For each estimated signal The k-th dimension decoding component in , the number of decoding components of each estimated signal is N; RF is the radio frequency in the direction of signal power.

4. The MIMO-SCMA decoding method for PMSs according to claim 3, characterized in that: The extracting the independent channel feature for each estimated signal from the multi-path estimated signals and constructing an N-point line graph according to the independent channel feature comprises: The color of the kth broken line coordinate in the N-point broken line graph is related to the channel characteristic of the communication channel where the kth broken line coordinate is located; The transmitting antenna n where the kth broken line coordinate is located t With receiving antenna n r The channel characteristics between Then calculate the reference amount: Determine the color of the kth polyline coordinate according to the value range of the reference quantity.

5. The MIMO-SCMA decoding method for PMSs according to claim 4, characterized in that: The extracting the independent channel feature for each estimated signal from the multi-path estimated signals and constructing an N-point line graph according to the independent channel feature comprises: Current estimated signal The type label corresponding to the current N-point line chart is: In the formula, and Decoding components The coordinates of the N-point polyline and The encoded value after pixelation operation of 32×32 channels; is the color label of the feature point.

6. The MIMO-SCMA decoding method for inter-PMSs according to claim 5, characterized in that: For multiple N-point line graphs corresponding to multiple random transmission symbol combinations, an image training set is constructed, and a graph neural network suitable for downlink signal decoding of a PMS communication system is trained using the image training set, including: For the current random transmission symbol combination Code t , get the type label corresponding to the N-point line chart Randomly generate multiple random transmission symbol combinations and obtain multiple corresponding type labels; Constructing an N-point line graph pixelated input image with each random transmission symbol combination, and training the input image using a simplified MobileNet neural network; The simplified MobileNet neural network is optimized with the goal of minimizing the cross entropy loss between the target type label and the type label pre-calculated by the N-point line graph to obtain the pre-trained graph neural network.

7. The MIMO-SCMA decoding method for PMSs according to claim 6, characterized in that: The method of using the pre-trained graph neural network to sequentially add type labels to the feature points in the N-point line graph of the current received signal, and decoding the current received signal according to the type label corresponding to each feature point and the preset codebook combination corresponding to the current received signal, includes: Inputting the currently received signal into the pre-trained graph neural network to obtain a type label of the currently received signal; According to the one-to-one correspondence between the type label and the preset MIMO-SCMA codebook, the received signals classified under the same type label are decoded respectively, so as to obtain the decoding result of the current received signal.

8. A MIMO-SCMA decoding system for PMSs; characterized in that: The system is implemented by using a MIMO-SCMA decoding method for PMSs according to any one of claims 1 to 7; The system includes a decomposition module, an extraction module, a training module and a decoding module; wherein, The decomposition module is used to detect the received signal of the downlink of the PMS communication system based on MIMO-SCMA under the random transmission symbol combination to decompose the multi-path estimation signal; The extraction module is used to extract the independent channel features for each estimated signal from the multi-channel estimated signals, and construct an N-point line graph according to the independent channel features; The training module is used to construct an image training set for multiple N-point line graphs corresponding to multiple random transmission symbol combinations, and use the image training set to train a graph neural network suitable for downlink signal decoding of the PMS communication system; The decoding module is used to use the pre-trained graph neural network to add type labels to the feature points in the N-point line graph of the current received signal in sequence, and decode the current received signal based on the type label corresponding to each feature point and the preset code book combination corresponding to the current received signal.

9. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images