Wireless ad hoc network channel adaptive adjustment method and system based on neural network

Abstract

The invention provides a wireless ad hoc network channel adaptive adjustment method based on a neural network, and the method comprises the following steps: S1, dividing a working frequency band of a wireless ad hoc network into a plurality of sub-channels, scanning the sub-channels to obtain a frequency spectrum scanning result of each sub-channel, selecting the sub-channel with the minimum interference, and obtaining a frequency spectrum scanning result of each sub-channel; taking the center frequency point as the carrier frequency of the OFDM signal; s2, uniformly dividing a subcarrier of the OFDM signal into a plurality of subcarrier blocks; calculating an EVM value and an SNR value of each subcarrier block; s3, inputting the frequency spectrum scanning result, the EVM value and the SNR value into a pre-trained neural network model, and outputting a channel state level; and S4, allocating a modulation pattern for each subcarrier block according to the channel state level, and adjusting the signal bandwidth and the number of the subcarrier blocks according to the channel state level. According to the method, the channel state is detected and tracked in real time, the working frequency of the system and the positions and the number of used subcarrier blocks are determined, and a proper modulation pattern is allocated to each subcarrier.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and more specifically to a method for adaptive adjustment of wireless ad hoc networks based on neural networks. Background Technology

[0002] With the rapid development of wireless communication technology, ad hoc networks, as a communication method that does not rely on fixed infrastructure and can be quickly and flexibly deployed, have been widely used in emergency communication, the Internet of Things, and military communication. In ad hoc network systems, how to efficiently and reliably utilize limited spectrum resources is key to improving the overall network performance. Orthogonal Frequency Division Multiplexing (OFDM) technology, due to its high spectrum utilization and resistance to multipath fading, has become a core physical layer technology for modern broadband wireless communication systems. OFDM technology divides a broadband channel into multiple parallel narrowband subcarrier channels, providing a foundation for achieving refined spectrum resource management.

[0003] However, existing ad hoc network spectrum allocation technologies still face many challenges. First, traditional static or semi-static spectrum allocation strategies are ill-suited to the rapidly changing characteristics of wireless channel environments. Channel fading, interference fluctuations, and other factors can degrade the channel quality of some subcarriers. Continuing to use fixed modulation and coding schemes will lead to decreased communication performance or link interruptions. Second, existing methods often rely on single channel quality metrics for decision-making, such as received signal strength indication (RSI) or signal-to-noise ratio (SNR). While SNR reflects the impact of noise, it cannot comprehensively characterize the degree of signal distortion caused by phase noise, carrier frequency offset, nonlinear distortion, etc. These single-dimensional assessments are often imprecise, leading to biases in resource allocation decisions.

[0004] Furthermore, most existing solutions lack the ability to continuously perceive and dynamically adjust channel states. They typically perform one-time parameter configuration at the initial stage of link establishment or trigger renegotiation only after severe performance degradation, lacking a forward-looking, periodic fine-tuning mechanism. This lag cannot meet the stringent real-time and reliability requirements of ad hoc network communication in highly dynamic channel environments. Therefore, there is an urgent need in the field for a technical solution that can deeply integrate multi-dimensional channel state information, possess intelligent analysis and prediction capabilities, and achieve real-time, accurate, and dynamic allocation of spectrum resources to overcome the shortcomings of existing technologies and improve the overall communication performance of ad hoc networks in complex environments. Summary of the Invention

[0005] The present invention aims to provide a neural network-based adaptive channel adjustment method for wireless ad hoc networks to solve the problems existing in the prior art and improve the reliability, real-time performance and spectral efficiency of ad hoc networks in complex electromagnetic environments.

[0006] To achieve the above objectives, this invention provides a method for adaptive channel adjustment in wireless ad hoc networks based on neural networks, comprising the following steps: S1, the working frequency band of the wireless ad hoc network is divided into multiple parallel sub-channels in a fixed step, the spectrum of the multiple parallel sub-channels is scanned to obtain the spectrum scan result of each sub-channel, and the sub-channel with the least interference is selected from the multiple parallel sub-channels through spectrum scanning, and its center frequency is used as the carrier frequency of the OFDM signal. S2: Divide the subcarriers of the OFDM signal into multiple subcarrier blocks; calculate the error vector amplitude (EVM) and signal-to-noise ratio (SNR) of each subcarrier block; S3: Normalize the spectrum scanning results, EVM values, and SNR values ​​corresponding to the subcarrier block. Use the normalized spectrum scanning results, EVM values, and SNR values ​​as input parameters to the pre-trained neural network model, and output the channel state level. S4: Based on the channel state level output by the neural network model, assign a modulation pattern to each subcarrier block according to the channel state level, and adjust the signal bandwidth and the number of subcarrier blocks according to the channel state level.

[0007] Preferably, in step S1, the spectrum scanning results include spectrum scanning FFT values, spectrum energy, and spectrum flatness; Among them, the frequency scan is calculated by FFT to obtain the spectrum scan FFT value; Spectral Energy The formula for summing the squares of the FFT values ​​of the spectral scan is as follows: ; in, The FFT result of the spectral scan of the k-th sequence is given, where M is the sequence length; Spectral flatness It can reflect the density of interference in the channel, and the calculation formula is as follows: ; In the formula, mean(·) is the arithmetic mean.

[0008] Preferably, in step S2, the formula for calculating the EVM value is as follows: ; in, For the actual received symbol point of the k-th sequence, Let M be the symbol point of the ideal case of the k-th sequence, and M be the sequence length; The formula for calculating SNR value is: Based on the pilot signal, the SNR is calculated as follows: Y is the pilot signal after channel gain, X is the local pilot signal, H is the channel response matrix, and N is the channel noise. Then we have: ; The channel response matrix is ​​estimated using the LS algorithm. ; Then calculate the noise value. ; The SNR value can be calculated. ; Preferably, in step S3, the neural network model includes an input layer, two hidden layers, and an output layer, with the output layer outputting discrete channel state levels.

[0009] Preferably, in step S4, the modulation pattern is assigned to each subcarrier block according to the channel state level, specifically as follows: There is a preset mapping relationship between channel state level and modulation pattern; the higher the channel state level, the higher the modulation order it is mapped to.

[0010] This invention also provides a neural network-based wireless ad hoc network channel adaptive adjustment system, comprising: The spectrum scanning module divides the working frequency band of the wireless ad hoc network into multiple parallel sub-channels in a fixed step, performs spectrum scanning on the multiple parallel sub-channels to obtain the spectrum scanning result of each sub-channel, and selects the sub-channel with the least interference from the multiple parallel sub-channels through spectrum scanning, and uses its center frequency as the carrier frequency of the OFDM signal. EVM and SNR value calculation module; divides the subcarriers of the OFDM signal into multiple subcarrier blocks; calculates the error vector amplitude (EVM) and signal-to-noise ratio (SNR) value for each subcarrier block; The neural network processing unit receives the spectrum scanning results from the spectrum scanning module and the EVM and SNR values ​​from the EVM and SNR value calculation module, performs normalization processing, and outputs the channel state level based on the normalized spectrum scanning results, EVM values, and SNR values. The parameter configuration and data transmission control module outputs the channel state level according to the neural network model, assigns a modulation style to each subcarrier block according to the channel state level, and adjusts the signal bandwidth and the number of subcarrier blocks according to the channel state level.

[0011] Preferably, in the spectrum scanning module, the spectrum scanning results include spectrum scanning FFT values, spectrum energy, and spectrum flatness; Among them, the frequency scan is calculated by FFT to obtain the spectrum scan FFT value; Spectral Energy The formula for summing the squares of the FFT values ​​of the spectral scan is as follows: ; The FFT result of the spectral scan of the k-th sequence is given, where M is the sequence length; Spectral flatness It can reflect the density of interference in the channel, and the calculation formula is as follows: ; In the formula, mean(·) is the arithmetic mean.

[0012] Preferably, in the EVM and SNR value calculation module, the formula for calculating the EVM value is as follows: ; in, For the actual received symbol point of the k-th sequence, Let M be the ideal sign point of the k-th sequence, and M be the sequence length. The process of calculating SNR value is as follows: Based on the pilot signal, the SNR is calculated as follows: Y is the pilot signal after channel gain, X is the local pilot signal, H is the channel response matrix, and N is the channel noise. Then we have: ; The channel response matrix is ​​estimated using the LS algorithm. ; Then calculate the noise value. ; The SNR value can be calculated. ; Preferably, in the neural network processing unit, the neural network model includes an input layer, two hidden layers, and an output layer, with the output layer outputting discrete channel state levels.

[0013] Preferably, in the parameter configuration and data transmission control module, a modulation pattern is assigned to each subcarrier block according to the channel state level. Specifically, there is a preset mapping relationship between the channel state level and the modulation pattern; the higher the channel state level, the higher the mapped modulation order. Compared with the prior art, the beneficial effects of the present invention are: 1. This invention introduces a neural network model to fuse and analyze multi-dimensional parameters such as spectrum scanning results, EVM value reflecting signal modulation quality, and SNR value reflecting noise level. This enables a more comprehensive and accurate assessment of channel status and provides a reliable basis for resource allocation decisions.

[0014] 2. This invention determines the system's operating frequency and the location and number of subcarrier blocks by real-time detection and tracking of the channel state, thereby avoiding channel interference. It assigns a suitable modulation pattern to each subcarrier, enabling the use of high-order modulation to increase the data rate when channel conditions are good and low-order modulation to ensure reliability when channel conditions are poor, thereby maximizing spectral efficiency and ensuring the robustness of the communication link.

[0015] 3. This invention establishes a closed-loop optimization mechanism by periodically and proactively sensing the channel state and updating and reallocating resources. This enables the system to respond quickly to channel changes and adjust the allocation and modulation strategies of subcarriers in real time, ensuring the accuracy and real-time nature of resource allocation and fundamentally improving the communication quality and spectrum resource utilization efficiency of ad hoc networks. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of OFDM signal subcarrier block division provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the channel state analysis neural network provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the sub-channel division of the operating frequency band provided in an embodiment of the present invention; Figure 4 This is a flowchart of master-slave station information interaction and channel status update provided in an embodiment of the present invention.

[0017] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1-4 As shown, this embodiment of the invention provides a method for adaptive channel adjustment in wireless ad hoc networks based on neural networks, comprising: S1, the working frequency band of the wireless ad hoc network is divided into multiple parallel sub-channels in a fixed step, the spectrum of the multiple parallel sub-channels is scanned to obtain the spectrum scan result of each sub-channel, and the sub-channel with the least interference is selected from the multiple parallel sub-channels through spectrum scanning, and its center frequency is used as the carrier frequency of the OFDM signal. In step S1, the spectrum scanning results include spectrum scanning FFT values, spectrum energy, and spectrum flatness. Spectral flatness, spectral energy, and spectral scan FFT values ​​are all obtained from the spectral scan results. The spectral scan FFT values ​​are obtained by performing a Fast Fourier Transform (FFT) on the frequency scan. Spectral Energy The formula for summing the squares of the FFT values ​​of the spectral scan is as follows: ; in, The FFT result of the spectral scan of the k-th sequence is given, where M is the sequence length; Spectral flatness It can reflect the density of interference in the channel, and the calculation formula is as follows: ; In the formula, mean(·) is the arithmetic mean.

[0020] In this embodiment, the operating frequency band is first divided into several sub-channels. The carrier frequency of the signal is adjusted based on the spectrum scanning results. The sub-channel with the least interference is selected to transmit the OFDM signal. At the same time, the frequency scanning results are recorded, and the corresponding spectral energy and spectral flatness are calculated. Secondly, the OFDM subcarriers are divided into several subcarrier blocks. While transmitting network maintenance frames, the signal quality indicators EVM value and SNR value are calculated.

[0021] S2: Divide the subcarriers of the OFDM signal into multiple subcarrier blocks; calculate the error vector amplitude (EVM) and signal-to-noise ratio (SNR) of each subcarrier block; In step S2 of this embodiment, EVM is the error vector amplitude, which is the difference between the actual received symbol point and the ideal symbol point. It is a signal loss caused by noise, distortion, phase noise, etc. The EVM value can accurately reflect the signal modulation quality. The lower the EVM value, the better the signal quality. The network maintenance frame data sent by the slave station is repeatedly transmitted on all subcarrier blocks. Calculating the EVM value of each subcarrier block can reflect the state of the corresponding frequency band of each subcarrier block to a certain extent. The calculation formula is as follows: ; in, For the actual received symbol point of the k-th sequence, Let M be the symbol point of the ideal case of the k-th sequence, and M be the sequence length; SNR, or Signal-to-Noise Ratio, is the ratio of signal power to noise power, reflecting the noise level. A higher SNR indicates better signal quality. The formula for calculating SNR is: Based on the pilot signal, the SNR is calculated as follows: Y is the pilot signal after channel gain, X is the local pilot signal, H is the channel response matrix, and N is the channel noise. Then we have: ; The channel response matrix is ​​estimated using the LS algorithm. ; Then calculate the noise value. ; The SNR value can be calculated. ; S3: Normalize the spectrum scanning results, EVM values, and SNR values ​​corresponding to the subcarrier block, and use the normalized spectrum scanning results, EVM values, and SNR values ​​as input parameters to the pre-trained neural network model. In step S3, the neural network model includes an input layer, two hidden layers, and an output layer. The output layer outputs discrete channel state levels.

[0022] Specifically, the input parameters of the neural network model include: normalized spectrum scan results, EVM values, and SNR values. Two hidden layers are used to extract channel features. The number of neurons in the first hidden layer is optimized based on the number of input parameters, while the number of neurons in the second hidden layer is optimized based on the output channel state level requirements. The output layer of the neural network model outputs the channel state level, which reflects the strength of interference and the quality of channel conditions. The three performance indicators—spectrum scan results, EVM values, and SNR values—reflect the state of interference in the channel, the modulation quality of each subcarrier block, and the noise level of the channel, respectively. Normalization of the input parameters reduces the impact of extreme values ​​on the network, accelerates network training convergence, and improves the model's performance and generalization ability. Channel state analysis is performed on a subcarrier block basis, reducing the network size; a two-hidden-layer network is sufficient to meet performance requirements. Hierarchical output of the channel state allows for the selection of the optimal signal modulation pattern based on the analyzed channel conditions.

[0023] S4: Based on the channel state level output by the neural network model, assign a modulation pattern to each subcarrier block according to the channel state level, and adjust the signal bandwidth and the number of subcarrier blocks according to the channel state level.

[0024] In step S4, a modulation pattern is assigned to each subcarrier block according to the channel state level. Specifically, there is a preset mapping relationship between the channel state level and the modulation pattern. The higher the channel state level, the higher the modulation order mapped.

[0025] Specifically, when the channel state level is at its lowest, the channel corresponding to the subcarrier block is severely affected by interference, so the subcarrier block is not used to avoid strong interference. As the channel state level increases, the channel conditions become better, and higher-order modulation methods can be used to improve the communication rate. After completing the channel state analysis of all subcarrier blocks, the number of subcarrier blocks to be used is determined according to the channel state level, and the signal bandwidth is adjusted. At the same time, each subcarrier block uses a modulation order suitable for the signal state to adjust the signal modulation pattern. The location and number of transmission subcarrier blocks are dynamically adjusted according to the real-time channel state to avoid allocating subcarriers in locations with strong interference, and a suitable modulation pattern is assigned to each subcarrier to achieve adjustment of signal bandwidth and modulation pattern.

[0026] This invention also provides a neural network-based wireless ad hoc network channel adaptive adjustment system, comprising: The spectrum scanning module divides the working frequency band of the wireless ad hoc network into multiple parallel sub-channels in a fixed step, performs spectrum scanning on the multiple parallel sub-channels to obtain the spectrum scanning result of each sub-channel, and selects the sub-channel with the least interference from the multiple parallel sub-channels through spectrum scanning, and uses its center frequency as the carrier frequency of the OFDM signal. In the spectrum scanning module, the spectrum scanning results include spectrum scanning FFT values, spectrum energy, and spectrum flatness; Among them, the frequency scan is calculated by FFT to obtain the spectrum scan FFT value; Spectral Energy The formula for summing the squares of the FFT values ​​of the spectral scan is as follows: ; The FFT result of the spectral scan of the k-th sequence is given, where M is the sequence length; Spectral flatness It can reflect the density of interference in the channel, and the calculation formula is as follows: ; In the formula, mean(·) is the arithmetic mean.

[0027] EVM and SNR value calculation module; divides the subcarriers of the OFDM signal into multiple subcarrier blocks; calculates the error vector amplitude (EVM) and signal-to-noise ratio (SNR) value for each subcarrier block; Specifically, in the EVM and SNR value calculation module, the formula for calculating the EVM value is as follows: ; in, For the actual received symbol point of the k-th sequence, Let M be the ideal sign point for the k-th sequence, M be the sequence length, and then the EVM value is normalized.

[0028] The process of calculating SNR value is as follows: Based on the pilot signal, the SNR is calculated as follows: Y is the pilot signal after channel gain, X is the local pilot signal, H is the channel response matrix, and N is the channel noise. Then we have: ; The channel response matrix is ​​estimated using the LS algorithm. ; Then calculate the noise value. ; The SNR value can be calculated. ; The neural network processing unit receives the spectrum scanning results from the spectrum scanning module and the EVM and SNR values ​​from the EVM and SNR value calculation module, performs normalization processing, and outputs the channel state level based on the normalized spectrum scanning results, EVM values, and SNR values. Specifically, in the neural network processing unit, the neural network model includes an input layer, two hidden layers, and an output layer, with the output layer outputting discrete channel state levels.

[0029] The parameter configuration and data transmission control module outputs the channel state level according to the neural network model, assigns a modulation style to each subcarrier block according to the channel state level, and adjusts the signal bandwidth and the number of subcarrier blocks according to the channel state level.

[0030] Specifically, when the channel state level is at its lowest, the channel corresponding to the subcarrier block is severely affected by interference, so the subcarrier block is not used to avoid strong interference. As the channel state level increases, the channel conditions become better, and higher-order modulation methods can be used to improve the communication rate. After completing the channel state analysis of all subcarrier blocks, the number of subcarrier blocks to be used is determined according to the channel state level, and the signal bandwidth is adjusted. At the same time, each subcarrier block uses a modulation order suitable for the signal state to adjust the signal modulation pattern. The location and number of transmission subcarrier blocks are dynamically adjusted according to the real-time channel state to avoid allocating subcarriers in locations with strong interference, and a suitable modulation pattern is assigned to each subcarrier to achieve adjustment of signal bandwidth and modulation pattern.

[0031] In this embodiment of the invention, an OFDM signal is divided into multiple subcarrier blocks on an average basis. In the subcarrier block design for channel state estimation and adaptive modulation based on subcarrier blocks, on the one hand, dividing the signal into as many subcarrier blocks as possible enables more refined channel state estimation and allows for the selection of an appropriate modulation order for each subcarrier; on the other hand, the information of the network maintenance frame is designed to be repeatedly transmitted in units of subcarrier blocks. Therefore, the number of subcarriers contained in a subcarrier block should be sufficient to carry all the information of a network maintenance frame, so that each subcarrier block can completely transmit the data of the network maintenance frame.

[0032] In this invention, the autonomous network channel adaptive adjustment system also introduces frame types, dividing the system frame types into two main categories: network maintenance frames and data frames.

[0033] The main functions of network maintenance frames are time synchronization, real-time time slot alignment calibration, transmission of time slot allocation tables, maintenance of network topology, and sending control commands. Network maintenance frames include synchronization frames, remote control frames, neighbor frames, and status frames. Synchronization frames are primarily used for coarse synchronization, including time slot numbers and clock levels. Remote control frames transmit configuration parameters, including time slot allocation tables, rate settings, transmit power, and response time requests. Neighbor frames broadcast neighbor information to the entire network, contributing to the network topology structure, including a list of neighbor nodes, neighbor node data latency, and location information. Status frames facilitate the exchange of node status information, including TOD (Time of Decision) information and time slot numbers, used for synchronization, and uploading their own status parameters. All network maintenance frames have a uniform and fixed frame length.

[0034] The primary function of a data frame is to encapsulate and transmit user data from the upper layer, i.e., to carry service data. The length of a data frame is not fixed; it is determined by the channel's transmission capacity. When the channel quality is good, all subcarrier blocks and the highest-order modulation scheme can be used, allowing more service data to be carried per unit time, resulting in longer data frame lengths. Conversely, when the channel quality is poor, only a limited number of subcarrier blocks and lower-order modulation schemes can be used to ensure correct data reception, resulting in less service data being carried per unit time slot and relatively shorter data frame lengths.

[0035] In this invention, the autonomous network channel adaptive adjustment system also introduces time slot classification, dividing time slots into network maintenance time slots and data time slots. Network maintenance time slots are used to transmit network maintenance frames, and data time slots are used to transmit service data.

[0036] Network maintenance time slots are dedicated to transmitting network maintenance frames, ensuring network time synchronization, parameter configuration updates, network topology maintenance, and control information exchange. Data time slots, on the other hand, are allocated to carry service data transmission, ensuring effective communication of user information. By dynamically allocating these two types of time slots, the system can achieve real-time tracking and response to channel conditions while ensuring network stability and efficiency.

[0037] The allocation of network maintenance time slots and data time slots primarily considers the real-time response of the system to changes in the network and channel environment, as well as the effectiveness of data transmission. In dynamic mobile wireless networking scenarios, the channel environment between nodes changes over time and distance. When the channel environment changes, to maintain reliable and stable transmission of service data, it is necessary to adjust the currently used frequency, power, and modulation pattern in a timely manner. This requires a sufficient number of network maintenance time slots to accomplish this task. Furthermore, when a node needs to transmit user data in a burst, a data time slot needs to be allocated to that node. Network maintenance time slots dynamically allocate data time slot resources by sending control frames. Only a sufficiently dense network maintenance time slot can dynamically allocate and reclaim data time slots in a timely manner, ensuring sufficiently low latency and higher real-time performance.

[0038] The design of data time slots primarily considers parameters such as node data transmission rate and latency. Data time slots are dynamically allocated to each node, and obtaining a sufficient number of data time slots is a necessary condition for a network node to achieve a high-speed data transmission rate. Under the condition that the necessary inter-node control parameter interaction for network operation is met, the lower the proportion of network maintenance time slots and the higher the proportion of data time slots per unit time, the higher the network transmission efficiency.

[0039] In this invention, the wireless ad hoc network nodes include a time reference station, a master station, and slave stations. When there is no external time reference, all nodes in the network elect a certain node as the time reference station; the node with the highest time reference priority is called the master station, and the other nodes are called slave stations. The master station can send synchronization frames, remote control frames, neighbor frames, and data frames. It is responsible for responding to time synchronization requests from lower clock-level nodes, unifying the network clock, acting as a relay node to perform multi-hop relay transmission of data frames, coordinating the allocation and reclamation of data time slots, and maintaining network topology.

[0040] The slave station can send remote control frames, neighbor frames, status frames, and data frames. Its main functions include responding to time synchronization requests from lower clock-level nodes, acting as a relay node to perform multi-hop relay transmission of data frames, coordinating and reclaiming data time slots from lower clock-level nodes, and maintaining network topology.

[0041] This invention discloses a wireless ad hoc network channel adaptive adjustment system based on neural networks. The working process of the system is described below.

[0042] Step 1: Operating Frequency Band Spectrum Scanning First, the system's operating frequency band is divided into N sub-channels with fixed bandwidth increments. A schematic diagram of the operating frequency band sub-channel division is shown below. Figure 3 As shown, after the master station is powered on, it scans the working frequency band in units of sub-channels, detects the channel status, sets the center frequency of the sub-channel with the least interference as the OFDM signal carrier frequency, and broadcasts a synchronization frame. Step 2: After the slave station is powered on, it scans the spectrum and searches for synchronization frames in a polling manner according to the channel conditions to complete the coarse time synchronization. The slave station sends a status frame to the master station to request fine time synchronization and to report its own status. Step 3: The master station receives the status frame from the slave station, calculates the EVM and SNR values ​​of each subcarrier block, and uses a neural network model to perform multi-dimensional fusion analysis on the channel status of each subcarrier block based on the spectrum scanning results, the EVM and SNR values ​​of each subcarrier block, and determines whether each subcarrier block in the data frame transmits data and the modulation style used. After the system determines the parameters of the data frame, it completes the interaction of parameter configuration of each node through the remote control frame. Step 4: The master station sends a synchronization frame and a remote control frame to the slave station. The synchronization frame completes precise time synchronization, and the remote control frame enables the master station to configure the parameters of the slave station. Step 5: The slave station configures the data frame to be sent according to the specified parameters of the received remote control frame. For subcarrier blocks with less interference, a higher-order modulation style is selected for mapping; for subcarrier blocks with more severe interference, a lower-order modulation style is selected for mapping, or no mapping is performed for data transmission. Step 6: The master station receives data frames according to the specified parameters and processes the data in the appropriate way based on the number of subcarrier blocks used in the data frame and the mapping method on each subcarrier block configured in the link layer. Step 7: After a preset number of data frame transmission time slots, the slave station retransmits a status frame. The master station recalculates the EVM and SNR values ​​based on this frame, and analyzes and updates the channel status information through a neural network model. In turn, it dynamically adjusts the position of the transmitted subcarrier block and the corresponding modulation style, updates the spectrum resource allocation, and completes steps 3-6 to continue data transmission and reception. Step 8: At the current operating frequency, if all subcarrier blocks within the signal bandwidth are interfered with, network maintenance frames will be unable to be transmitted normally, and all nodes will lose network access. The master station will re-execute the operation in Step 1, perform a spectrum scan, select a carrier frequency less affected by interference, and then proceed according to Steps 2-6.

[0043] This embodiment is used in a scenario where multiple slave stations transmit collected information back to the master station while in motion. The ad hoc network transmission channel is a multipath time-varying channel. The master station schedules the entire ad hoc network system and performs data collection and analysis. To more clearly describe the parameter configuration adjustment process of the wireless ad hoc network channel adaptive adjustment method, a more specific embodiment is illustrated by the information interaction between a master station and a slave station.

[0044] In this embodiment, the system operates at a bandwidth of 200MHz, and the OFDM signal bandwidth used is 20MHz. The operating frequency band is divided into 19 sub-channels in 10MHz increments. Each OFDM signal has 1024 subcarriers, which are divided into 8 subcarrier blocks.

[0045] The master station is initially powered off. After powering on, it performs a spectrum scan on each of the 19 sub-channels. Each sub-channel is calculated using a 1024-point FFT. Based on the spectrum scan results, the spectrum energy and spectrum flatness are calculated. The center frequency of the sub-channel with the least interference is selected as the carrier frequency for OFDM signal transmission. The master station then begins broadcasting synchronization frames.

[0046] After the slave station is powered on, it performs a spectrum scan operation on each sub-channel. According to the order of the spectrum scan results, it searches for and receives synchronization frames on the center frequency of the sub-channel in turn. After receiving the master station synchronization frame, it completes coarse time synchronization. The slave station sends a status frame to report its own status parameters and makes a time synchronization request.

[0047] The master station receives the status frame, receives the slave station's time synchronization request, calculates the SNR and the EVM values ​​of the 8 subcarrier blocks, and selects the subcarrier block with the smallest EVM value for reception processing as reliable information.

[0048] A neural network model was used to perform fusion analysis on the EVM value, SNR value, spectral energy, spectral flatness, and 128-point FFT value of the spectrum scan for each subcarrier block. The network had a total of 132 inputs, and the number of hidden layer neural nodes was set to 64 and 32, respectively. The modulation schemes used in this embodiment were QPSK, 16QAM, and 64QAM, and subcarriers were allocated to avoid strong interference locations. The number of channel state levels output by the network was 4.

[0049] In this embodiment, for each subcarrier block, the four channel state levels from bad to good represent no transmission, QPSK, 16QAM, and 64QAM, respectively, and the corresponding relationship is shown in Table 1 below.

[0050] After the master station completes channel assessment and subcarrier dynamic allocation decision, it sends synchronization frames and remote control frames to the slave station. The synchronization frames are used to complete time synchronization, and the remote control frames enable the master station to configure the parameters of the slave station.

[0051] The slave station receives the remote control frame and transmits data frames according to the subcarrier transmission position and modulation pattern specified in the control frame information. The master station receives the data frames according to the corresponding transmission parameters and completes the subsequent signal processing.

[0052] Network maintenance frames are sent at fixed time slot intervals. Based on the time-varying characteristics of the channel state in this embodiment, a network maintenance frame is sent every 10 data frames to achieve dynamic adjustment of parameters while minimizing the additional overhead of network maintenance.

[0053] While receiving network maintenance frames to complete network maintenance, the master station updates the EVM and SNR values ​​accordingly, updates channel state information through neural network analysis, and dynamically adjusts subcarrier allocation and modulation patterns. It also sends synchronization and remote control frames to the slave station to complete network time maintenance and spectrum resource allocation instructions. Both the master and slave stations continue data transmission and reception, awaiting the next control frame uploaded by the slave station.

[0054] During data transmission, if all subcarrier blocks in the current operating frequency are interfered with, both data frames and network maintenance frames cannot be transmitted, and the network connection is lost. The master station re-executes a spectrum scan, sets the center frequency of the carrier in the least interfered frequency band as the operating frequency, and broadcasts a synchronization frame. Similarly, the slave station performs a spectrum scan after the network disconnection, searching for the master station's synchronization frame using a frequency conversion method; this process is consistent with the initial power-on process. After the master station receives the slave station's status frame, completes channel condition assessment, uses a neural network model to output a data frame to generate a configuration, and sends it to the slave station via a remote control frame, both master and slave stations continue data frame transmission and reception.

[0055] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for adaptive channel adjustment in wireless ad hoc networks based on neural networks, characterized in that, Includes the following steps: S1, the working frequency band of the wireless ad hoc network is divided into multiple parallel sub-channels in a fixed step, the spectrum of the multiple parallel sub-channels is scanned to obtain the spectrum scan result of each sub-channel, and the sub-channel with the least interference is selected from the multiple parallel sub-channels through spectrum scanning, and its center frequency is used as the carrier frequency of the OFDM signal. S2: Divide the subcarriers of the OFDM signal into multiple subcarrier blocks; calculate the error vector amplitude (EVM) and signal-to-noise ratio (SNR) of each subcarrier block; S3: Normalize the spectrum scanning results, EVM values, and SNR values ​​corresponding to the subcarrier block. Use the normalized spectrum scanning results, EVM values, and SNR values ​​as input parameters to the pre-trained neural network model, and output the channel state level. S4: Based on the channel state level output by the neural network model, assign a modulation pattern to each subcarrier block according to the channel state level, and adjust the signal bandwidth and the number of subcarrier blocks according to the channel state level.

2. The wireless ad hoc network channel adaptive adjustment method based on neural networks according to claim 1, characterized in that, In step S1, the spectrum scanning results include spectrum scanning FFT values, spectrum energy, and spectrum flatness. Among them, the frequency scan is calculated by FFT to obtain the spectrum scan FFT value; Spectral Energy The formula for summing the squares of the FFT values ​​of the spectral scan is as follows: ； The FFT result of the spectral scan of the k-th sequence is given, where M is the sequence length; Spectral flatness The calculation formula is: ； In the formula, mean(·) is the arithmetic mean.

3. The wireless ad hoc network channel adaptive adjustment method based on neural networks according to claim 1, characterized in that, In step S2, the formula for calculating the EVM value is as follows: ； in, For the actual received symbol point of the k-th sequence, Let M be the symbol point of the ideal case of the k-th sequence, and M be the sequence length; The process of calculating SNR value is as follows: Based on the pilot signal, the SNR is calculated as follows: the pilot signal after channel gain is Y, the local pilot signal is X, the channel response matrix is ​​H, and the channel noise is N. ； The channel response matrix is ​​estimated using the LS algorithm. ； Then calculate the noise value. ; ； The formula for calculating SNR value is: 。 4. The wireless ad hoc network channel adaptive adjustment method based on neural networks according to claim 1, characterized in that, In step S3, the neural network model includes an input layer, two hidden layers, and an output layer. The output layer outputs discrete channel state levels.

5. The wireless ad hoc network channel adaptive adjustment method based on neural networks according to claim 1, characterized in that, In step S4, a modulation pattern is assigned to each subcarrier block according to the channel state level. Specifically, there is a preset mapping relationship between the channel state level and the modulation pattern. The higher the channel state level, the higher the modulation order mapped.

6. A wireless ad hoc network channel adaptive adjustment system based on neural networks, characterized in that, include: The spectrum scanning module divides the working frequency band of the wireless ad hoc network into multiple parallel sub-channels in a fixed step, performs spectrum scanning on the multiple parallel sub-channels to obtain the spectrum scanning result of each sub-channel, and selects the sub-channel with the least interference from the multiple parallel sub-channels through spectrum scanning, and uses its center frequency as the carrier frequency of the OFDM signal. The EVM and SNR value calculation module divides the subcarriers of the OFDM signal into multiple subcarrier blocks; and calculates the error vector amplitude (EVM) and signal-to-noise ratio (SNR) value for each subcarrier block. The neural network processing unit receives the spectrum scanning results from the spectrum scanning module and the EVM and SNR values ​​from the EVM and SNR value calculation module, performs normalization processing, and outputs the channel state level based on the normalized spectrum scanning results, EVM values, and SNR values. The parameter configuration and data transmission control module outputs the channel state level according to the neural network model, assigns a modulation style to each subcarrier block according to the channel state level, and adjusts the signal bandwidth and the number of subcarrier blocks according to the channel state level.

7. The wireless ad hoc network channel adaptive adjustment system based on neural networks according to claim 6, characterized in that, In the spectrum scanning module, the spectrum scanning results include spectrum scanning FFT values, spectrum energy, and spectrum flatness; Among them, the frequency scan is calculated by FFT to obtain the spectrum scan FFT value; Spectral Energy The formula for summing the squares of the FFT values ​​of the spectral scan is as follows: ； The FFT result of the spectral scan of the k-th sequence is given, where M is the sequence length; Spectral flatness The calculation formula is: ； In the formula, mean(·) is the arithmetic mean.

8. The wireless ad hoc network channel adaptive adjustment system based on neural networks according to claim 6, characterized in that... In the EVM and SNR value calculation module The formula for calculating the EVM value is: ； in, For the actual received symbol point of the k-th sequence, Let M be the symbol point of the ideal case of the k-th sequence, and M be the sequence length; The process of calculating SNR value is as follows: Based on the pilot signal, the SNR is calculated as follows: the pilot signal after channel gain is Y, the local pilot signal is X, the channel response matrix is ​​H, and the channel noise is N. ； The channel response matrix is ​​estimated using the LS algorithm. ; ； Then calculate the noise value. ; ； The formula for calculating SNR value is: 。 9. The wireless ad hoc network channel adaptive adjustment system based on neural networks according to claim 6, characterized in that, In the neural network processing unit, the neural network model includes an input layer, two hidden layers, and an output layer. The output layer outputs discrete channel state levels.

10. The wireless ad hoc network channel adaptive adjustment system based on neural networks according to claim 6, characterized in that, In the parameter configuration and data transmission control module, a modulation pattern is assigned to each subcarrier block according to the channel state level. Specifically, there is a preset mapping relationship between the channel state level and the modulation pattern. The higher the channel state level, the higher the mapped modulation order.

Images