MIMO open-loop uplink precoding optimization system for gigabit transmission

Abstract

The invention discloses an MIMO open-loop uplink precoding optimization system for gigabit transmission, and relates to the technical field of MIMO open-loop precoding, the MIMO open-loop uplink precoding optimization system comprises an MIMO receiving and transmitting module, an MIMO transmitter uplink module, an MIMO receiver uplink module and a simulation platform module, the MIMO receiving and transmitting module comprises components and functions of an MIMO transmitting module and an MIMO receiving module; the MIMO transmitter uplink module is divided into a phase noise test unit, a serial multi-user precoding unit and a system time delay unit; the uplink module of the MIMO receiver is divided into an SC-FDMA (Single Carrier Frequency Division Multiple Access) unit and a terminal polarization multi-antenna unit; the simulation platform module is used for building a simulation platform and accelerating a simulation process through a GPU (Graphics Processing Unit). A phase noise testing technology, a serial multi-user pre-coding technology and a frequency domain resource allocation technology in the method are closely combined with a modern information technology; the method provided by the invention can be ensured to refine the precoding selection standard for the MIMO open-loop uplink precoding optimization system for gigabit transmission in a more accurate range.

Description

Technical Field

[0001] The present invention relates to the technical field of MIMO open-loop precoding, and particularly to an MIMO open-loop uplink precoding optimization system for gigabit transmission. Background Art

[0002] In the development process of the new generation of mobile communication systems, based on the rate requirement of 100M - 1Gbps, under this background, China has successfully achieved a transmission rate of 100Mbps in high-speed mobile scenarios. At the same time, many foreign countries have also achieved the transmission goal of 1Gbps in low-speed mobile scenarios. Most of the current Gbit transmission systems still use the transmission technology of the 100M system. Among them, the open-loop multiple-input multiple-output (MIMO) technology is the core. In practical applications, due to usually harsh channel conditions, in order to ensure the signal transmission quality, a very complex maximum likelihood (ML) detection technology needs to be adopted. With the in-depth research on related technologies, aiming at the innovative ideas of key technologies such as multi-antenna and broadband OFDM, based on the FuTURE technology, further research has been carried out in key technology fields such as precoding, MIMO detection, high-efficiency coding modulation, learning-based environmental parameter acquisition, and environmental adaption, and a set of Gbps wireless transmission test demonstration system based on GE switching has been developed. The Gbps wireless transmission test demonstration system based on GE switching aims to explore the application of Gbps wireless transmission technology in systems such as the fourth-generation mobile communication, wireless local area network, and short-range wireless communication;

[0003] Although the existing technology has made great progress in the direction of MIMO open-loop uplink precoding optimization for gigabit transmission, there are still some problems to be optimized. The open-loop MIMO technology adopted by the current Gbit transmission system faces severe challenges when applied to the gigabit system. The existing technology lacks the test and research on the uplink subsystems of the transmitter and receiver in a multi-antenna environment, which increases the manufacturing cost and operating cost of the equipment, seriously affects the portability and energy efficiency of the equipment, and hinders the wide application and further development of gigabit transmission technology. Summary of the Invention

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: An MIMO open-loop uplink precoding optimization system for gigabit transmission, including an MIMO transceiver module, an MIMO transmitter uplink module, an MIMO receiver uplink module, and a simulation platform module, wherein each module is communicatively connected;

[0005] The MIMO transceiver module includes the components and functions of the MIMO transmitter module and the MIMO receiver module. Among them, the MIMO transmitter module is responsible for signal processing and transmission, framing and frequency conversion, and distribution of service data. The MIMO receiver module is responsible for signal reception and processing, data demodulation and decoding, and phase noise processing;

[0006] The uplink module of the MIMO transmitter is divided into a phase noise test unit, a serial multi-user precoding unit, and a system delay unit, which are used to implement phase noise test, precoding processing, and system delay test;

[0007] The uplink module of the MIMO receiver is divided into an SC-FDMA unit and a terminal polarization multi-antenna unit. Among them, the SC-FDMA unit allocates frequency domain resources based on the improved OFDM technology and combines the reinforcement learning algorithm. The terminal polarization multi-antenna unit is used to obtain the complex correlation coefficient of the MIMO receiver module. The complex correlation coefficient of the MIMO receiver module describes the performance indicators of the MIMO open-loop uplink precoding optimization system for gigabit transmission, and explores the influence of antenna polarization mode and radiation pattern on the performance of the MIMO open-loop uplink precoding optimization system for gigabit transmission;

[0008] The simulation platform module builds a simulation platform and accelerates the simulation process through GPU.

[0009] A further improvement of the technical solution of the present invention is that the components of the MIMO transmitter module and the MIMO receiver module include:

[0010] The components of the MIMO transceiver module are composed of a functional board, an RF unit, a PC, a gigabit Ethernet GE interface, a gigabit Ethernet switch, and a baseband processing cabinet. Among them, the functional board includes an antenna board, a signal processing board, a service board, an IDU sub-board, and a phase noise tracking board;

[0011] A 100M bandwidth below 14.400GHz - 14.500GHz is selected as the test frequency band. The RF unit is composed of an ODU device and a second-stage frequency conversion module in the range of 14.400GHz - 14.500GHz. The corresponding carrier center frequency is at least 14.417GHz and at most 14.483GHz. The baseband signal is subjected to the first-stage frequency conversion, shifted to the intermediate frequency, and the second-stage frequency conversion operations, and the baseband signal is modulated to the target carrier frequency. By changing the clock source settings, the carrier frequency is adjusted between 14.417GHz and 14.483GHz;

[0012] In the MIMO transmission module, the PC sets the parameter configuration of the transmitter, generates service data, and wirelessly transmits the service data to the transmitter; in the MIMO reception module, the PC sets the parameter configuration of the receiver, generates service data, and wirelessly transmits the service data to the receiver, and completes signal processing operations such as timing synchronization, frequency offset estimation, channel estimation, matrix SVD decomposition, and inversion.

[0013] The function board is connected to the PC through a gigabit Ethernet switch. Among them, a phase noise tracking board is built in the baseband processing cabinet, and the daughter board on the baseband processing cabinet provides a gigabit Ethernet GE interface to the outside.

[0014] A further improvement of the technical solution of the present invention lies in that the functions of the MIMO transmission module and the MIMO reception module include:

[0015] The MIMO transmission module consists of an antenna board, a service board, and a signal processing board. Among them, the antenna board receives the encoded data stream from the signal processing board, frames the encoded data stream according to the data transmission frame format, performs digital up-conversion operation, and sends it to the RF unit through the IDU daughter board; the service board receives service data from the PC and distributes it to the corresponding signal processing board; the signal processing board receives service data from the service board and completes the baseband signal transmission processing functions of LDPC coding and 64QAM modulation.

[0016] The MIMO reception module consists of an antenna board, a signal processing board, a service board, and a phase noise tracking board. Among them, the antenna board receives data from the IDU daughter board, performs digital down-conversion, FFT transformation, and frame splitting processing, and sends the processed data to the PC and the signal processing board; the signal processing board performs matrix multiplication for MIMO detection, 64QAM demodulation, and LDPC decoding processing on the received data, sends the processed data to the service board for aggregation, and transmits it to the PC for service demonstration; a phase noise tracking board is built in the baseband processing cabinet, estimates and tracks the real-time change of phase noise according to the reserved subcarriers in the data frame, sends the common phase noise correction parameters back to the signal processing board for forward demodulation constellation rotation, and at the same time transmits the real-time change data of the phase noise to the antenna board for real-time tracking correction.

[0017] A further improvement of the technical solution of the present invention lies in that the MIMO transmitter uplink module is divided into a phase noise test unit, a serial multi-user precoding unit, and a PCMA unit:

[0018] Among them, the phase noise test unit is used to test phase noise, collect phase noise at different temperatures using a signal analyzer, obtain the influence degree of different temperatures on phase noise testing, and combine with a neural network algorithm to construct a temperature difference phase noise monitoring model.

[0019] The serial multi-user precoding unit is divided into linear precoding and non-linear precoding, which preprocesses the transmitted signals.

[0020] The system delay unit is used to measure the system delay in real time and reduces the system delay by referring to terahertz communication.

[0021] A further improvement of the technical solution of the present invention lies in that: the phase noise test unit tests the phase noise, and the process of constructing the temperature difference phase noise monitoring model includes:

[0022] The phase noise test unit consists of a hardware part and a software part. The software part consists of a temperature difference phase noise monitoring model, a CPE estimation algorithm, and LabVIEW. The hardware part consists of a MIMO transmission module, a channel emulator, and a MIMO reception module;

[0023] In the software part, the temperature difference phase noise monitoring model is used to monitor the influence degree of different temperatures on the phase noise test; the CPE estimation algorithm is used to estimate the carrier phase error caused by the phase noise during the signal transmission process, evaluate the signal quality and the performance of the MIMO open-loop uplink precoding optimization system for gigabit transmission; LabVIEW is used to integrate the temperature difference phase noise monitoring model and the CPE estimation algorithm to control the phase noise test process;

[0024] In the hardware part, the MIMO transmission module is equipped with two transmission antennas to transmit the signals generated by the software part; the MIMO reception module is equipped with two reception antennas to receive the signals transmitted through the channel; the channel emulator sets the channel parameters and simulates the signal transmission in different channels;

[0025] Set the ambient temperature to 23 °C, 55 °C, and 80 °C, and 23 °C, 55 °C, and 80 °C correspond to low temperature, medium temperature, and high temperature respectively. Use a signal analyzer to collect the phase noise of 10 1 、10 2 、10 3 、10 4 、10 5 and 10 6 frequency offsets respectively. Set 10 1 and 10 2 frequency offsets as the low-frequency part, 10 3 and 10 4 frequency offsets as the medium-frequency part, and 10 5 and 10 6 as the high-frequency part;

[0026] Taking the frequency offset as the abscissa and the phase noise as the ordinate, a purple line is selected to represent the relationship curve between the frequency offset and the phase noise at 23 degrees Celsius, a green line is selected to represent the relationship curve between the frequency offset and the phase noise at 55 degrees Celsius, and a red line is selected to represent the relationship curve between the frequency offset and the phase noise at 80 degrees Celsius. The phase noise curves at different temperatures are plotted and analyzed. The phase noise in the intermediate frequency part and the high frequency part is not affected by temperature. When the temperature rises from 23 degrees Celsius to 55 degrees Celsius, the phase noise in the low frequency part is slightly affected by temperature. When the temperature rises from 55 degrees Celsius to 80 degrees Celsius, the phase in the low frequency part is greatly affected by temperature;

[0027] At 10 1 frequency offset, the phase noises corresponding to 23 degrees Celsius and 80 degrees Celsius are respectively selected, and the process of calculating the influence degree of temperature on the phase noise test is as follows:

[0028] 00%

[0029] where I is the influence degree of temperature on the phase noise test, is the phase noise corresponding to 80 degrees Celsius at 10 1 frequency offset, is the phase noise corresponding to 23 degrees Celsius at 10 1 frequency offset;

[0030] A neural network model is constructed. The phase noises corresponding to T1 and T3 and the influence degree of temperature on the phase noise test at 10 1 frequency offset are used as the data set, which is divided into a training set and a test set according to a ratio of 7:3. The MLP is selected as the neural network structure. The input layer includes two neurons, which receive the phase noises corresponding to 23 degrees Celsius and 80 degrees Celsius at 10 1 frequency offset. The hidden layer configures the MSE function, and the output layer includes one neuron, which outputs the influence degree of temperature on the phase noise test;

[0031] The training set data is input into the neural network model. The learning rate is set to 0.01, and the number of iterative training times is 1000. The training process includes forward propagation and backward propagation. Among them, forward propagation is used to calculate the predicted output data, and backward propagation is used to update the weights and biases of the model. By repeating the iterative training, the nonlinear relationship between the phase noises corresponding to T1 and T3 and the influence degree of temperature on the phase noise test is learned until the set number of iterative training times is reached, and the trained neural network model is obtained;

[0032] The test set data is input into the trained neural network model. Using the MSE function, the error between the output value and the actual value of the neural network model is evaluated. According to the evaluation results, the parameters of the neural network model are adjusted to optimize the performance of the neural network model, and the temperature difference phase noise monitoring model is obtained.

[0033] A further improvement of the technical solution of the present invention lies in that: the serial multi-user precoding unit is divided into linear precoding and non-linear precoding, and the process of preprocessing the transmitted signal includes:

[0034] Linear precoding is applied to the linear processing of the transmitted signal at the base station transmitter. Based on the zero-forcing criterion ZF precoding, the multi-user transmitted signal is preprocessed. The multi-user original transmitted signals sequentially pass through the precoding matrix, the channel matrix, and multiplication by the scaling factor to obtain the preprocessed multi-user transmitted signals;

[0035] Non-linear precoding is THP precoding based on the Costa loop. THP precoding cancels interference through hierarchical serial processing. Each layer of the transmitted vector is the data stream of that layer minus the interference of all previous layers plus a modulo operation, and the signal is moved to the original constellation mapping point to obtain the preprocessed transmitted signal;

[0036] Combining linear precoding and non-linear precoding forms a hybrid precoding scheme. Under different channel conditions and signal transmission requirements, the two precoding methods are switched. In the case of high signal-to-noise ratio and low interference, linear precoding is adopted, and in the case of low signal-to-noise ratio and high interference, it is switched to non-linear precoding.

[0037] A further improvement of the technical solution of the present invention lies in that: the system delay unit, and the process of real-time testing the system delay includes:

[0038] On the platform configured with VLC soft decoding and set-top box hard decoding in the MIMO receiving module, data is collected by the camera to obtain video data, which is transmitted to the MIMO transmitting module through the VGA interface. After being modulated by the MIMO transmitting module, it is transmitted through the transmitting antenna. The soft decoding platform receives the output signal, inputs it into the computer through the GE interface, and performs soft decoding and video playback by the VLC software; the hard decoding platform receives the output signal, transmits it to the set-top box through the GE interface, and after being hard decoded by the set-top box, it is displayed on the display through the VGA interface. By the time difference displayed by the soft decoding platform and the hard decoding platform, the real-time system delays respectively measured by the soft decoding platform and the hard decoding platform are obtained.

[0039] A further improvement of the technical solution of the present invention lies in that: the SC-FDMA unit, based on the OFDM improvement technology, combines the reinforcement learning algorithm, and the process of allocating frequency domain resources includes:

[0040] Based on the improved OFDM technology, in the MIMO transmission module, the bit information of the original signal is subjected to channel coding, interleaving, and modulation processing, grouped into units of every N symbols, then the data in each group is transformed into the frequency domain using an N-point FFT, the data is mapped to the system subcarriers in the frequency domain, after the mapping is completed, the data is transformed back into the time domain using an N-point IFFT, and a guard interval is added to the signal transformed back into the time domain before transmission;

[0041] Based on the improved OFDM technology, a frequency-domain equalizer is equipped in the MIMO receiving module. The MIMO receiving module performs the operation of removing the guard interval on the received digital baseband signal, transforms the signal into the frequency domain using an M-point FFT, extracts and separates the data in the frequency domain, after channel equalization by the frequency-domain equalizer, transforms the data back into the time domain using an N-point IFFT, and performs demodulation, deinterleaving, and decoding operations on it to recover the bit information of the original signal;

[0042] Through the improved OFDM technology, the channel state information is sensed in real time. The built-in core decision-making unit of the reinforcement learning algorithm is the agent, and the action space is predefined. The action space includes the resource allocation method. The agent selects an action from the predefined action space according to the channel state information and executes it. After the agent executes an action, the MIMO open-loop uplink precoding optimization system for gigabit transmission gives feedback to the agent. According to the feedback result, the agent obtains a reward value, evaluates the suitability of the selected action through the reward value, and based on this reward value, the agent updates the Q value based on the Q-learning algorithm, improves the learning decision of the selected action, and obtains the frequency-domain resource allocation algorithm based on reinforcement learning through iterative learning.

[0043] A further improvement of the technical solution of the present invention lies in that: the process of the terminal polarization multi-antenna unit obtaining the complex correlation coefficient of the MIMO receiving module includes:

[0044] Set the acquisition center frequency to 2.3 GHz and the measurement bandwidth to 200 MHz. Using a vector network analyzer and a radio frequency switch, collect data at a frequency point of k and a location of m at a frequency interval of 1 MHz to obtain the channel transmission matrix. The average power gain of the MIMO open-loop uplink precoding optimization system for gigabit transmission with single input and single output is 1. The process of calculating the normalization coefficient is as follows:

[0045]

[0046] Among them, A is the normalization coefficient, K and M are the total number of measured frequency points and the number of locations respectively, Tr() represents taking the trace of the matrix, represents the channel transmission matrix measured at the frequency point k and the location m, is the pseudo-inverse matrix of, and They are the number of transmit antennas and the number of receive antennas respectively;

[0047] The process of calculating the complex correlation coefficient of the MIMO receive module using the normalization coefficient is as follows:

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] Wherein, is the signal-to-noise ratio, is the average signal-to-noise ratio, is the channel capacity, is the average channel capacity, is the complex correlation coefficient of the MIMO receive module, A is the normalization coefficient, is the power of additive white Gaussian noise, K and M are the total number of measured frequency points and the number of locations respectively, Tr() represents taking the trace of a matrix, represents the channel transfer matrix measured at frequency point k and location m, is the pseudo-inverse matrix of, and are the number of transmit antennas and the number of receive antennas respectively, p and q are the p-th row and the q-th row of the channel transfer matrix respectively, det represents taking the value of the determinant, is the identity matrix of, and are the p-th row and the q-th row of the channel transfer matrix respectively, E represents taking the mathematical expectation.

[0055] A further improvement of the technical solution of the present invention lies in that: the process of building a simulation platform by the simulation platform module and accelerating the simulation process through GPU includes:

[0056] Building the simulation platform includes hardware configuration and software configuration. Among them, the hardware configuration includes GPU, CPU, memory and SSD solid-state drive, and the software configuration includes Windows operating system, GPU driver, CUDA toolkit and simulation software;

[0057] During the simulation process, the GPU adopts a memory architecture that includes global memory, shared memory, and registers. Among them, the global memory has a large capacity but slow access speed; the shared memory has a fast access speed and is used for data sharing within a thread block; the register is the storage unit with the fastest access speed among the three memory architectures and is used by a single thread. The GPU achieves data parallelism and simulation task parallelism during the simulation process through parallel computing, and processes multiple simulation tasks simultaneously.

[0058] Compared with the traditional MIMO open-loop uplink precoding optimization system for gigabit transmission, in the MIMO open-loop uplink precoding optimization system for gigabit transmission of the present invention, the phase noise testing technology, serial multi-user precoding technology, and frequency-domain resource allocation technology in the method of the present invention are closely combined with modern information technology. Among them, the phase noise testing technology uses a signal analyzer to collect phase noise data at different temperatures, combines a neural network algorithm to construct a temperature difference phase noise monitoring model, accurately analyzes the influence of different temperatures on phase noise, obtains the variation of phase noise with frequency offset at different temperatures, and achieves real-time and comprehensive testing and research on phase noise; the serial multi-user precoding technology combines linear precoding and non-linear precoding to form a hybrid precoding scheme, and according to different channel conditions and signal transmission requirements, switches between the two precoding methods in real time, accurately analyzes the channel state and interference situation data, obtains the precoding strategy most suitable for the current transmission environment, and achieves real-time and reasonable selection of the precoding method; the frequency-domain resource allocation technology uses the OFDM improvement technology to sense the channel state information in real time, combines the agent in the reinforcement learning algorithm, selects an action to execute from the predefined action space according to the channel state, and improves the learning decision based on the reward value, accurately analyzes the channel state information data, obtains the frequency-domain resource allocation strategy based on reinforcement learning, and achieves real-time and efficient allocation of frequency-domain resources. It solves the problems in the prior art that lack the testing and research on the uplink subsystems of the transmitter and receiver in a multi-antenna environment, increases the manufacturing cost and operating cost of the equipment, seriously affects the portability and energy efficiency of the equipment, and hinders the wide application and further development of gigabit transmission technology, ensuring that the method in the present invention can refine the precoding selection criteria for the MIMO open-loop uplink precoding optimization system for gigabit transmission within a more accurate range. The research and application of this method significantly enhance the degree of intelligence in the MIMO open-loop precoding process. Brief Description of the Drawings

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0060] Figure 1It is a block diagram of the MIMO open-loop uplink precoding optimization system for gigabit transmission of the present invention;

[0061] Figure 2 It is a component and logical connection diagram of the MIMO transmitter and the MIMO receiver;

[0062] Figure 3 It is a phase noise curve graph at different temperatures. Specific embodiments

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0064] As Figure 1 shown, the present invention provides a MIMO open-loop uplink precoding optimization system for gigabit transmission, including a MIMO transceiver module, a MIMO transmitter uplink module, a MIMO receiver uplink module, and a simulation platform module, wherein each module is communicatively connected;

[0065] The MIMO transceiver module includes the components and functions of the MIMO transmitter module and the MIMO receiver module. Among them, the MIMO transmitter module is responsible for signal processing and transmission, framing and frequency conversion, and distribution of service data, and the MIMO receiver module is responsible for signal reception and processing, data demodulation and decoding, and phase noise processing;

[0066] The MIMO transmitter uplink module is divided into a phase noise test unit, a serial multi-user precoding unit, and a system delay unit, and is used to implement phase noise test, precoding processing, and system delay test;

[0067] The MIMO receiver uplink module is divided into a SC-FDMA unit and a terminal polarization multi-antenna unit. Among them, the SC-FDMA unit, based on the OFDM improvement technology, combines the reinforcement learning algorithm to allocate frequency domain resources; the terminal polarization multi-antenna unit is used to obtain the complex correlation coefficient of the MIMO receiver module, and this complex correlation coefficient of the MIMO receiver module describes the performance index of the MIMO open-loop uplink precoding optimization system for gigabit transmission, and explores the influence of the antenna polarization mode and the radiation pattern on the performance of the MIMO open-loop uplink precoding optimization system for gigabit transmission;

[0068] The simulation platform module builds a simulation platform to accelerate the simulation process through GPU.

[0069] Preferably, asFigure 2 As shown in the figure, the components of the MIMO transmitting module and the MIMO receiving module include:

[0070] The components of the MIMO transceiver module consist of a functional board, an RF unit, a PC, a Gigabit Ethernet GE interface, a Gigabit Ethernet switch, and a baseband processing cabinet. Among them, the functional board includes an antenna board, a signal processing board, a service board, an IDU daughter board, and a phase noise tracking board;

[0071] Select a 100M bandwidth below 14.400GHz - 14.500GHz as the test frequency band. The RF unit consists of an ODU device and a second-stage frequency conversion module in the range of 14.400GHz - 14.500GHz. The corresponding carrier center frequency is at least 14.417GHz and at most 14.483GHz. Perform the first-stage frequency conversion, shift to the intermediate frequency, and the second-stage frequency conversion operations on the baseband signal, modulate the baseband signal to the target carrier frequency, and adjust the carrier frequency between 14.417GHz and 14.483GHz by changing the clock source settings;

[0072] In the MIMO transmitting module, the PC sets the parameter configuration of the transmitting end, generates service data, and wirelessly transmits the service data to the transmitting end; in the MIMO receiving module, the PC sets the parameter configuration of the receiving end, generates service data, and wirelessly transmits the service data to the receiving end, and completes signal processing operations such as timing synchronization, frequency offset estimation, channel estimation, matrix SVD decomposition, and inversion;

[0073] The functional board is connected to the PC through a Gigabit Ethernet switch. Among them, a phase noise tracking board is built in the baseband processing cabinet, and the daughter board on the baseband processing cabinet provides a Gigabit Ethernet GE interface externally.

[0074] Preferably, the functions of the MIMO transmitting module and the MIMO receiving module include:

[0075] The MIMO transmitting module consists of an antenna board, a service board, and a signal processing board. Among them, the antenna board receives the encoded data stream from the signal processing board, frames the encoded data stream according to the data transmission frame format, performs digital up-conversion operations, and sends it to the RF unit through the IDU daughter board; the service board receives service data from the PC and distributes it to the corresponding signal processing board; the signal processing board receives service data from the service board and completes the baseband signal transmission processing functions of LDPC encoding and 64QAM modulation;

[0076] The MIMO receiving module consists of an antenna board, a signal processing board, a service board, and a phase noise tracking board. Among them, the antenna board receives data from the IDU daughter board, performs digital down-conversion, FFT transformation, and frame splitting processing, and sends the processed data to the PC and the signal processing board; the signal processing board performs matrix multiplication for MIMO detection, 64QAM demodulation, and LDPC decoding on the received data, sends the processed data to the service board for aggregation, and transmits it to the PC for service demonstration; the phase noise tracking board is built into the baseband processing cabinet, estimates and tracks the real-time changes of phase noise according to the reserved subcarriers in the data frame, sends the common phase noise correction parameters back to the signal processing board for forward demodulation constellation rotation, and at the same time transmits the real-time change data of phase noise to the antenna board for real-time tracking and correction.

[0077] Preferably, the MIMO transmitter uplink module is divided into a phase noise test unit, a serial multi-user precoding unit, and a PCMA unit:

[0078] Among them, the phase noise test unit is used to test phase noise, collect phase noise at different temperatures using a signal analyzer, obtain the influence degree of different temperatures on phase noise testing, and combine with a neural network algorithm to construct a temperature difference phase noise monitoring model;

[0079] The serial multi-user precoding unit is divided into linear precoding and non-linear precoding, and preprocesses the transmitted signal;

[0080] The system delay unit is used to test the system delay in real time and reduce the system delay by introducing terahertz communication.

[0081] Preferably, as Figure 3 shown, the process of the phase noise test unit testing phase noise and constructing a temperature difference phase noise monitoring model includes:

[0082] The phase noise test unit consists of a hardware part and a software part. The software part consists of a temperature difference phase noise monitoring model, a CPE estimation algorithm, and LabVIEW. The hardware part consists of a MIMO transmission module, a channel emulator, and a MIMO receiving module;

[0083] In the software part, the temperature difference phase noise monitoring model is used to monitor the influence degree of different temperatures on phase noise testing; the CPE estimation algorithm is used to estimate the carrier phase error caused by phase noise during signal transmission, evaluate signal quality, and optimize the system performance of the MIMO open-loop uplink pre-coding system for gigabit transmission; LabVIEW is used to integrate the temperature difference phase noise monitoring model and the CPE estimation algorithm to control the phase noise testing process;

[0084] In the hardware part, the MIMO transmitting module is equipped with two transmitting antennas to transmit the signals generated by the software part; the MIMO receiving module is equipped with two receiving antennas to receive the signals transmitted through the channel; the channel emulator sets channel parameters to simulate the transmission of signals in different channels.

[0085] Set the ambient temperature to 23 degrees Celsius, 55 degrees Celsius, and 80 degrees Celsius, and 23 degrees Celsius, 55 degrees Celsius, and 80 degrees Celsius correspond to low temperature, medium temperature, and high temperature respectively. Use a signal analyzer to collect 10 1 、10 2 、10 3 、10 4 、10 5 and 10 6 phase noises with frequency offsets, set 10 1 and 10 2 frequency offsets as the low-frequency part, 10 3 and 10 4 frequency offsets as the medium-frequency part, 10 5 and 10 6 as the high-frequency part;

[0086] Taking the frequency offset as the abscissa and the phase noise as the ordinate, select a purple line to represent the relationship curve between the frequency offset and the phase noise at 23 degrees Celsius, select a green line to represent the relationship curve between the frequency offset and the phase noise at 55 degrees Celsius, and select a red line to represent the relationship curve between the frequency offset and the phase noise at 80 degrees Celsius. Plot the phase noise curves at different temperatures, analyze the phase noise curves at different temperatures. The phase noises in the medium-frequency part and the high-frequency part are not affected by temperature. When the temperature rises from 23 degrees Celsius to 55 degrees Celsius, the phase noise in the low-frequency part is less affected by temperature. When the temperature rises from 55 degrees Celsius to 80 degrees Celsius, the phase in the low-frequency part is greatly affected by temperature;

[0087] At a frequency offset of 10 1 , respectively select the phase noises corresponding to 23 degrees Celsius and 80 degrees Celsius, and the process of calculating the influence degree of temperature on the phase noise test is as follows:

[0088] 00%

[0089] where I is the influence degree of temperature on the phase noise test, is the phase noise corresponding to 80 degrees Celsius at a frequency offset of 10 1 , is the phase noise corresponding to 23 degrees Celsius at a frequency offset of 10 1 ;

[0090] Construct a neural network model, at a frequency offset of 10 1At a frequency offset, the phase noise corresponding to T1 and T3 and the influence degree of temperature on the phase noise test are used as a data set, which is divided into a training set and a test set according to a ratio of 7:3. The MLP is selected as the neural network structure. The input layer includes two neurons, receiving 10 1 At a frequency offset, the phase noise corresponding to 23 degrees Celsius and 80 degrees Celsius. The hidden layer configures the MSE function, and the output layer includes one neuron, outputting the influence degree of temperature on the phase noise test;

[0091] Input the training set data into the neural network model, set the learning rate to 0.01, and the number of iterative training times to 1000. The training process includes forward propagation and backward propagation. Among them, forward propagation is used to calculate the predicted output data, and backward propagation is used to update the weights and biases of the model. By repeating iterative training, learn the non-linear relationship between the phase noise corresponding to T1 and T3 and the influence degree of temperature on the phase noise test until the set number of iterative training times is reached, and obtain the trained neural network model;

[0092] Input the test set data into the trained neural network model, use the MSE function to evaluate the error between the output value and the actual value of the neural network model, adjust the parameters of the neural network model according to the evaluation results, optimize the performance of the neural network model, and obtain the temperature difference phase noise monitoring model.

[0093] Preferably, the serial multi-user precoding unit is divided into linear precoding and non-linear precoding. The process of preprocessing the transmitted signal includes:

[0094] Linear precoding is applied to the linear processing of the transmitted signal at the base station transmitter. Based on the zero-forcing criterion ZF precoding, preprocess the multi-user transmitted signal. The multi-user original transmitted signal passes through the precoding matrix, channel matrix, and multiplication with the scaling factor in sequence to obtain the preprocessed multi-user transmitted signal;

[0095] Non-linear precoding is THP precoding based on the Costa loop. THP precoding cancels interference through hierarchical serial. Each layer of the transmitted vector is the data stream of this layer minus the interference of all previous layers plus a modulo operation, and the signal is moved to the original constellation mapping point to obtain the preprocessed transmitted signal;

[0096] Combine linear precoding and non-linear precoding to form a hybrid precoding scheme. Under different channel conditions and signal transmission requirements, switch between the two precoding methods. In the case of high signal-to-noise ratio and low interference, use linear precoding. In the case of low signal-to-noise ratio and high interference, switch to non-linear precoding.

[0097] Preferably, the system delay unit. The process of real-time testing the system delay includes:

[0098] In a platform where the MIMO receiving module is configured with VLC soft decoding and set-top box hard decoding, data is collected by a camera to obtain video data, which is transmitted to the MIMO transmitting module through a VGA interface. After modulation by the MIMO transmitting module, it is sent through a transmitting antenna. The soft decoding platform receives the output signal, inputs it into a computer through a GE interface, and performs soft decoding and video playback using VLC software; the hard decoding platform receives the output signal, transmits it to the set-top box through a GE interface, and after hard decoding by the set-top box, it is displayed on a monitor through a VGA interface. By measuring the display time difference between the soft decoding platform and the hard decoding platform, the real-time system latency measured by the soft decoding platform and the hard decoding platform respectively can be obtained.

[0099] Preferably, for the SC-FDMA unit, based on the OFDM improvement technology and combined with the reinforcement learning algorithm, the process of allocating frequency domain resources includes:

[0100] Based on the OFDM improvement technology, in the MIMO transmitting module, channel coding, interleaving, and modulation processing are performed on the bit information of the original signal. The data is grouped with every N symbols as a unit, and then each group of data is transformed to the frequency domain using an N-point FFT. In the frequency domain, the data is mapped to the system subcarriers. After the mapping is completed, the data is transformed back to the time domain using an N-point IFFT. After adding a guard interval to the signal transformed back to the time domain, it is then sent.

[0101] Based on the OFDM improvement technology, a frequency domain equalizer is equipped in the MIMO receiving module. The MIMO receiving module performs an operation to remove the guard interval on the received digital baseband signal, transforms the signal to the frequency domain using an M-point FFT, extracts and separates the data in the frequency domain, performs channel equalization through the frequency domain equalizer, and then transforms the data back to the time domain using an N-point IFFT. Demodulation, deinterleaving, and decoding operations are performed on it to recover the bit information of the original signal.

[0102] By the OFDM improvement technology, the channel state information is sensed in real time. The built-in core decision-making unit of the reinforcement learning algorithm is an agent, and an action space is predefined. This action space includes resource allocation methods. The agent selects an action to execute from the predefined action space according to the channel state information. When the agent executes an action, the MIMO open-loop uplink precoding optimization system for gigabit transmission gives feedback to the agent. According to the feedback result, the agent obtains a reward value, evaluates the suitability of the selected action through the reward value, and based on this reward value and the Q-learning algorithm, the agent updates the Q value to improve the learning decision of the selected action. Through iterative learning, a frequency domain resource allocation algorithm based on reinforcement learning is obtained.

[0103] Preferably, for the terminal polarization multi-antenna unit, the process of obtaining the complex correlation coefficient of the MIMO receiving module includes:

[0104] Set the acquisition center frequency to 2.3 GHz, the measurement bandwidth to 200 MHz. Using a vector network analyzer and a radio frequency switch, data is collected at a frequency point of k and a location of m with a frequency interval of 1 MHz to obtain the channel transmission matrix. For a single-input single-output MIMO open-loop uplink precoding optimization system with a target average power gain of 1, the process of calculating the normalization coefficient is as follows:

[0105]

[0106] Where A is the normalization coefficient, K and M are the total number of measured frequency points and the number of locations respectively, Tr() represents the trace of the matrix, represents the channel transmission matrix measured at frequency point k and location m, is the pseudo-inverse matrix of, and are the number of transmit antennas and the number of receive antennas respectively;

[0107] The process of calculating the complex correlation coefficient of the MIMO receiving module using the normalization coefficient is as follows:

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114] Where, is the signal-to-noise ratio, is the average signal-to-noise ratio, is the channel capacity, is the average channel capacity, is the complex correlation coefficient of the MIMO receiving module, A is the normalization coefficient, is the additive white Gaussian noise power, K and M are the total number of measured frequency points and the number of locations respectively, Tr() represents the trace of the matrix, represents the channel transmission matrix measured at frequency point k and location m, is the pseudo-inverse matrix of, and are the number of transmit antennas and the number of receive antennas respectively, p and q are the p-th row and q-th row of the channel transmission matrix respectively, det represents the value of the determinant, is The identity matrix, and are the p-th row and the q-th row of the channel transmission matrix respectively, and E represents the mathematical expectation.

[0115] Preferably, the process of building a simulation platform by the simulation platform module and accelerating the simulation process through GPU includes:

[0116] Building the simulation platform includes hardware configuration and software configuration. Among them, the hardware configuration includes GPU, CPU, memory and SSD solid-state drive, and the software configuration includes Windows operating system, GPU driver, CUDA toolkit and simulation software;

[0117] During the simulation process, the GPU adopts a memory architecture of global memory, shared memory and registers. Among them, the global memory has a large capacity and slow access speed; the shared memory has a fast access speed and is used for data sharing within a thread block; the register is the storage unit with the fastest access speed among the three memory architectures and is used by a single thread. The GPU realizes data parallelism and simulation task parallelism during the simulation process and processes multiple simulation tasks simultaneously.

[0118] First, the components of the MIMO transceiver module include a functional board, an RF unit, a PC, a Gigabit Ethernet (GE) interface, a Gigabit Ethernet switch, and a baseband processing cabinet. In the MIMO transmit module, the PC sets parameters and generates service data for transmission to the sending end. In the MIMO receive module, the PC completes the corresponding parameter settings, data transmission, and signal processing operations. In terms of its logical connection, in the MIMO transmit module, the antenna board, service board, and signal processing board cooperate to complete data processing and transmission. In the MIMO receive module, each component is responsible for data reception, processing, demodulation and decoding, and phase noise processing. Secondly, the uplink module of the MIMO transmitter is divided into a phase noise test unit, a serial multi-user precoding unit, and a system delay unit. The phase noise test unit collects phase noise at different temperatures and constructs a temperature difference phase noise monitoring model by combining neural network algorithms. The serial multi-user precoding unit includes linear precoding and non-linear precoding, and preprocesses the transmitted signal by switching the precoding method according to the channel conditions. The system delay unit uses a specific platform to measure the system delay in real time and reduces the delay by introducing terahertz communication. Then, the uplink module of the MIMO receiver is divided into an SC-FDMA unit and a terminal polarization multi-antenna unit. The SC-FDMA unit is based on an improved OFDM technology and combines reinforcement learning algorithms to perform signal processing and allocate frequency domain resources at both the MIMO transmit and receive ends. The terminal polarization multi-antenna unit calculates the normalization coefficient through the channel transmission matrix, and then obtains the complex correlation coefficient of the MIMO receive module, which is used to explore the impact of antenna-related factors on system performance. Finally, a simulation platform is built through the simulation platform module, including hardware configuration and software configuration. During the simulation process, the GPU adopts a specific memory architecture and realizes data parallelism and simulation task parallelism through parallel computing to accelerate the simulation process.

[0119] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all of them should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A MIMO open-loop uplink precoding optimization system for gigabit transmission, including a MIMO transceiver module, a MIMO transmitter uplink module, a MIMO receiver uplink module and a simulation platform module, wherein: Each module is connected in communication, characterized by: The MIMO transceiver module includes components and functions of a MIMO transmitting module and a MIMO receiving module; The MIMO transmitter uplink module is divided into a phase noise test unit, a serial multi-user precoding unit and a system delay unit, wherein the phase noise test unit is used to test the phase noise and build a temperature difference phase noise monitoring model; the serial multi-user precoding unit is used to pre-process the transmission signal through linear precoding and nonlinear precoding; the system delay unit is used to test the system delay in real time and use terahertz communication to reduce the system delay; The MIMO receiver uplink module is divided into an SC-FDMA unit and a terminal polarization multi-antenna unit, wherein the SC-FDMA unit allocates frequency domain resources based on an improved OFDM technology combined with a reinforcement learning algorithm; the terminal polarization multi-antenna unit is used to obtain a complex correlation coefficient of the MIMO receiving module; The simulation platform module builds a simulation platform and accelerates the simulation process through GPU.

2. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 1, characterized in that: The components of the MIMO transmitting module and the MIMO receiving module include: The components of the MIMO transmitting module and the MIMO receiving module are composed of a function board, an RF unit, a PC, a Gigabit Ethernet GE interface, a Gigabit Ethernet switch, and a baseband processing cabinet, wherein the function board includes an antenna board, a signal processing board, a service board, an IDU sub-board, and a phase noise tracking board; A 100M bandwidth below 14.400GHz~14.500GHz is selected as the test frequency band. The RF unit is composed of an ODU device of 14.400GHz~14.500GHz and a secondary frequency conversion module. The corresponding carrier center frequency is as low as 14.417GHz and as high as 14.483GHz. The baseband signal is subjected to the first frequency conversion, moved to the intermediate frequency, and the second frequency conversion to modulate the baseband signal to the target carrier frequency. The carrier frequency is adjusted between 14.417GHz and 14.483GHz by changing the clock source setting. In the MIMO transmission module, the PC sets the parameter configuration of the transmission end, generates service data, and transmits the service data wirelessly to the transmission end; in the MIMO reception module, the PC sets the parameter configuration of the reception end, generates service data, and transmits the service data wirelessly to the reception end, and completes the signal processing operations of timing synchronization, frequency offset estimation, channel estimation, matrix SVD decomposition and inversion; The function board is connected to the PC through a Gigabit Ethernet switch, wherein a phase tracking board is built in the baseband processing cabinet, and a sub-board on the baseband processing cabinet provides a Gigabit Ethernet GE interface to the outside.

3. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 2, characterized in that: The functions of the MIMO transmitting module and the MIMO receiving module include: The MIMO transmission module consists of an antenna board, a service board, and a signal processing board. The antenna board receives the encoded data stream from the signal processing board, frames the encoded data stream according to the data transmission frame format, performs digital up-conversion, and sends it to the RF unit through the IDU sub-board; the service board receives service data from the PC and distributes it to the corresponding signal processing board; the signal processing board receives service data from the service board and completes the baseband signal transmission processing function of LDPC coding and 64QAM modulation; The MIMO receiving module consists of an antenna board, a signal processing board, a service board and a phase noise tracking board. The antenna board receives data from the IDU sub-board, performs digital down-conversion, FFT transformation and frame deframing, and sends the processed data to the PC and the signal processing board. The signal processing board performs matrix multiplication, 64QAM demodulation and LDPC decoding on the received data for MIMO detection, sends the processed data to the service board for aggregation, and transmits it to the PC for service demonstration. A phase noise tracking board is built into the baseband processing cabinet. According to the reserved subcarriers in the data frame, it estimates and tracks the real-time changes of phase noise, transmits the common phase noise correction parameters back to the signal processing board, performs forward demodulation constellation rotation, and transmits the real-time change data of phase noise to the antenna board for real-time tracking and correction.

4. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 3, characterized in that: The MIMO transmitter uplink module is divided into a phase noise test unit, a serial multi-user precoding unit and a PCMA unit: The phase noise test unit is used to test the phase noise, collect the phase noise at different temperatures using a signal analyzer, obtain the influence of different temperatures on the phase noise test, and build a temperature difference phase noise monitoring model in combination with a neural network algorithm; The serial multi-user precoding unit is divided into linear precoding and nonlinear precoding to preprocess the transmission signal; The system delay unit is used to test the system delay in real time and use terahertz communication to reduce the system delay.

5. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 4, characterized in that: The phase noise testing unit tests the phase noise and constructs a temperature difference phase noise monitoring model, including: The phase noise test unit consists of hardware and software. The software consists of a temperature difference phase noise monitoring model, a CPE estimation algorithm, and LabVIEW. The hardware consists of a MIMO transmitting module, a channel simulator, and a MIMO receiving module. In the software part, the temperature difference phase noise monitoring model is used to monitor the influence of different temperatures on the phase noise test; the CPE estimation algorithm is used to estimate the carrier phase error caused by phase noise during the signal transmission process, evaluate the signal quality and the performance of the MIMO open-loop uplink precoding optimization system for gigabit transmission; the LabVIEW is used to integrate the temperature difference phase noise monitoring model and the CPE estimation algorithm to control the phase noise test process; In the hardware part, the MIMO transmitting module is equipped with two transmitting antennas to transmit the signals generated by the software part; the MIMO receiving module is equipped with two receiving antennas to receive the signals transmitted by the channel; the channel emulator sets the channel parameters and simulates the transmission of signals in different channels; Set the ambient temperature to T1, T2 and T3, and T1, T2 and T3 correspond to low temperature, medium temperature and high temperature respectively. Use the signal analyzer to collect 10 1 , 10 2 , 10 3 , 10 4 , 10 5 and 10 6 Phase noise of frequency deviation, set to 10 1 and 10 2 The frequency deviation is the low frequency part, 10 3 and 10 4 The frequency deviation is the intermediate frequency part, 10 5 and 10 6 For the high frequency part; With frequency deviation as the horizontal axis and phase noise as the vertical axis, the purple line is used to represent the relationship curve between frequency deviation and phase noise under T1, the green line is used to represent the relationship curve between frequency deviation and phase noise under T2, and the red line is used to represent the relationship curve between frequency deviation and phase noise under T3. The phase noise curves at different temperatures are plotted and analyzed. The phase noise of the intermediate frequency part and the high frequency part is not affected by temperature. When the temperature increases from T1 to T2, the phase noise of the low frequency part is less affected by temperature. When the temperature increases from T2 to T3, the phase of the low frequency part is greatly affected by temperature. In 10 1 Under frequency deviation, the phase noise corresponding to T1 and T3 is selected respectively, and the process of calculating the influence of temperature on the phase noise test is as follows: 00% Among them, I is the influence of temperature on phase noise test, For 10 1 Under frequency deviation, the phase noise corresponding to T3 is: For 10 1 Phase noise corresponding to T1 degree under frequency offset; Construct a neural network model and convert 10 1 The phase noise corresponding to T1 and T3 under frequency deviation and the influence of temperature on phase noise test are taken as data sets, which are divided into training set and test set in a ratio of 7:

3. MLP is selected as the neural network structure. The input layer includes two neurons and receives 10 1 Under frequency offset, the phase noise corresponding to T1 and T3, the hidden layer is configured with MSE function, the output layer includes a neuron, and the output temperature has an impact on the phase noise test; Input the training set data into the neural network model, set the learning rate to 0.01, and the number of iterative training to 1000. The training process includes forward propagation and back propagation, where forward propagation is used to calculate the predicted output data, and back propagation is used to update the weights and biases of the model. Through repeated iterative training, the nonlinear relationship between the phase noise corresponding to T1 and T3 and the degree of influence of temperature on the phase noise test is learned until the set number of iterative training is reached, and the trained neural network model is obtained; The test set data is input into the trained neural network model, and the MSE function is used to evaluate the error between the output value of the neural network model and the actual value. The parameters of the neural network model are adjusted according to the evaluation results, the performance of the neural network model is optimized, and the temperature difference phase noise monitoring model is obtained.

6. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 5, characterized in that: The serial multi-user precoding unit is divided into linear precoding and nonlinear precoding, and the process of preprocessing the transmission signal includes: Linear precoding is applied to the linear processing of the signal transmitted by the base station transmitter. Based on the ZF precoding of the zero forcing criterion, the multi-user transmission signal is preprocessed. The original transmission signal of the multi-user passes through the precoding matrix, the channel matrix and is multiplied by the scaling factor in turn to obtain the transmission signal after multi-user preprocessing; Nonlinear precoding is THP precoding based on Costa loop. THP precoding cancels interference through layered serial processing. The transmission vector of each layer is the data stream of that layer minus the interference of all previous layers plus a modulo operation. The signal is moved to the original constellation mapping point to obtain the preprocessed transmission signal. Linear precoding is combined with nonlinear precoding to form a hybrid precoding scheme. Under different channel conditions and signal transmission requirements, the two precoding methods are switched. When the signal-to-noise ratio is high and the interference is small, linear precoding is used. When the signal-to-noise ratio is low and the interference is large, it is switched to nonlinear precoding.

7. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 6, characterized in that: The process of the system delay unit for real-time testing of the system delay includes: A platform for VLC soft decoding and set-top box hard decoding is configured in the MIMO receiving module. The camera collects data, obtains video data, and transmits it to the MIMO transmitting module through the VGA interface. After being modulated by the MIMO transmitting module, it is sent through the transmitting antenna. The soft decoding platform receives the output signal, inputs it into the computer through the GE interface, and performs soft decoding and video playback by the VLC software; the hard decoding platform receives the output signal, transmits it to the set-top box through the GE interface, and after hard decoding by the set-top box, it is displayed on the monitor through the VGA interface. The time difference is displayed by the soft decoding platform and the hard decoding platform, and the real-time system delay tested by the soft decoding platform and the hard decoding platform respectively is obtained.

8. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 7, characterized in that: The SC-FDMA unit, based on the OFDM improved technology and combined with the reinforcement learning algorithm, allocates frequency domain resources in the following process: Based on the improved OFDM technology, in the MIMO transmission module, the bit information of the original signal is channel coded, interleaved and modulated, and grouped into units of N symbols. Then, each group of data is transformed into the frequency domain using N-point FFT, and the data is mapped to the system subcarrier in the frequency domain. After the mapping is completed, the data is transformed back to the time domain using N-point IFFT, and the signal transformed back to the time domain is sent after adding a guard interval. Based on the improved OFDM technology, the MIMO receiving module is equipped with a frequency domain equalizer. The MIMO receiving module removes the guard interval of the received digital baseband signal, transforms the signal to the frequency domain using M-point FFT, extracts and separates the data in the frequency domain, and transforms the data back to the time domain using N-point IFFT after channel equalization by the frequency domain equalizer. The data is then demodulated, deinterleaved and decoded to restore the bit information of the original signal. Through the improved OFDM technology, the channel state information is perceived in real time. The core decision-making unit of the reinforcement learning algorithm is an intelligent agent. The action space is predefined, which includes the resource allocation method. The intelligent agent selects an action to execute from the predefined action space according to the channel state information. When the intelligent agent executes an action, the MIMO open-loop uplink precoding optimization system for gigabit transmission provides feedback to the intelligent agent. According to the feedback result, the intelligent agent obtains a reward value and evaluates the suitability of the selected action through the reward value. According to the reward value, based on the Q-learning algorithm, the intelligent agent improves the learning decision of the selected action by updating the Q value. Through iterative learning, a frequency domain resource allocation algorithm based on reinforcement learning is obtained.

9. The MIMO open-loop uplink for Gigabit transmission according to claim 8, characterized in that: The process of obtaining the complex correlation coefficient of the MIMO receiving module by the terminal polarization multi-antenna unit includes: Set the acquisition center frequency to 2.3GHz and the measurement bandwidth to 200MHz. Use a vector network analyzer and an RF switch to collect data at a frequency point k and location m with a frequency interval of 1MHz to obtain the channel transmission matrix. Use a single-input single-output MIMO open-loop uplink precoding optimization system for gigabit transmission with an average power gain of 1. The process of calculating the normalization coefficient is as follows: ； Among them, A is the normalization coefficient, K and M are the total number of frequency points and the number of locations measured respectively, and Tr() represents the trace of the matrix. represents the channel transmission matrix measured at frequency point k and location m, for The pseudo-inverse matrix of and are the number of transmitting antennas and the number of receiving antennas respectively; The process of calculating the complex correlation coefficient of the MIMO receiving module using the normalization coefficient is as follows: ； ； ； ； ； ； in, is the signal-to-noise ratio, is the average signal-to-noise ratio, is the channel capacity, is the average channel capacity, is the complex correlation coefficient of the MIMO receiving module, A is the normalization coefficient, is the power of additive white Gaussian noise, K and M are the total number of frequency points and the number of locations measured respectively, Tr() represents the trace of the matrix, represents the channel transmission matrix measured at frequency point k and location m, for The pseudo-inverse matrix of and are the number of transmitting antennas and the number of receiving antennas respectively, p and q are the pth and qth rows of the channel transmission matrix respectively, det represents the value of the determinant, for The identity matrix of and are the p-th and q-th rows of the channel transmission matrix respectively, and E represents the mathematical expectation.

10. The MIMO open-loop uplink precoding optimization system for gigabit transmission according to claim 9, characterized in that: The simulation platform module builds a simulation platform and accelerates the simulation process through GPU, including: The construction of the simulation platform includes hardware configuration and software configuration. The hardware configuration includes GPU, CPU, memory and SSD solid state drive, and the software configuration includes Windows operating system, GPU driver, CUDA toolkit and simulation software. During the simulation process, the GPU adopts a memory architecture of global memory, shared memory and registers. Among them, the global memory has a large capacity and slow access; the shared memory has a fast access speed and is used for data sharing within the thread block; the register is the fastest storage unit to access among the three memory architectures and is used by a single thread. The GPU realizes data parallelism and simulation task parallelism in the simulation process through parallel computing, and can process multiple simulation tasks at the same time.

Images