A dynamic channel allocation method for communication links based on Hongmeng system

Abstract

The present invention belongs to the field of computer communication technology, and in particular to a method for dynamic channel allocation of communication links based on the Hongmeng system. The implementation steps include: establishing a collaborative communication monitoring module in the Hongmeng communication node, distributing and collecting multi-dimensional link status parameters such as channel idleness, and dynamically adjusting the collection period as the status changes; using the improved LSTM algorithm to build an evaluation model, integrating spatial coordinate frequency domain features and time series parameters to generate a channel status evaluation matrix; dynamically allocating channels based on channel stability, task priority and data type; setting a fast re-evaluation mechanism to reallocate when the channel score deviates from the threshold to ensure communication stability. The present invention improves channel resource utilization and communication stability through dynamic monitoring, intelligent evaluation and adaptive allocation. It is suitable for distributed IoT scenarios of the Hongmeng system and solves the problems of channel competition and real-time scheduling under high-density deployment.

Description

Technical Field

[0001] The present invention belongs to the field of computer communication technology, and in particular relates to a dynamic channel allocation method for communication links based on the Hongmeng system. Background Art

[0002] With the continued advancement of the Internet of Everything (IoE) concept, the HarmonyOS system, with its distributed architecture and soft bus capabilities, has been widely adopted in various scenarios, including home automation, industrial IoT, and the Internet of Vehicles. In the HarmonyOS ecosystem, frequent data communication and state coordination between distributed terminals and edge devices have become the norm. This, especially in high-density deployment environments, places higher demands on the quality and stability of communication links. However, existing communication systems generally use static channel allocation strategies, which pre-set communication channels and device bindings. This makes it difficult to dynamically adjust channel resources based on device status changes, link interference fluctuations, or access requirements. In particular, in scenarios with high concurrent access, multi-source link competition, and demanding real-time task scheduling, static allocation schemes not only lead to low channel resource utilization but can also cause link congestion, communication delays, and even connection interruptions. Furthermore, traditional channel selection mechanisms lack real-time perception and intelligent assessment of channel status, making them unable to support the dynamic scheduling requirements of complex, heterogeneous devices in the HarmonyOS system. Summary of the Invention

[0003] In response to the technical problems existing in the above-mentioned background technology, the present invention proposes a dynamic channel allocation method for communication links based on the Hongmeng system.

[0004] In order to achieve the above object, the technical solution adopted by the present invention includes the following steps:

[0005] S1. Establish a collaborative communication monitoring module in multiple communication nodes running the Hongmeng system to collect link status parameters of the communication channel between the local node and surrounding nodes in real time. The link status parameters include: channel idleness, signal strength, bit error rate, and interference index;

[0006] S2. Construct a channel state assessment model based on the collected multi-source link state parameters. The model is used to generate a state assessment matrix for all current channels. The specific implementation process includes the following steps:

[0007] Assign a unique channel number to each channel to be evaluated, establish a corresponding data input channel, and input the link state parameters of the channel into the channel state evaluation model;

[0008] The evaluation model outputs the score value corresponding to each channel respectively, and all the score values are arranged in the order of channel numbers to form a one-dimensional score vector;

[0009] In each scoring cycle, the step of forming a one-dimensional scoring vector is performed on all channels, and all scoring vectors are spliced vertically along the time dimension to form a two-dimensional channel state evaluation matrix;

[0010] S3. Calculate channel stability through the state evaluation matrix and complete channel allocation based on task priority and data type;

[0011] S4. During the communication process, the real-time status parameters of the allocated channels are continuously monitored. If the difference between the score value and the average value of the corresponding scoring period at the time of allocation exceeds the set threshold, the rapid re-evaluation mechanism is triggered to re-evaluate the channel and perform channel allocation to ensure communication stability.

[0012] Preferably, the collaborative communication monitoring module adopts a distributed deployment architecture, realizes cross-node data synchronization through the soft bus technology of the Hongmeng system, and dynamically matches the acquisition period of each node with the frequency of channel state changes, and is calculated as follows: ,in is the current node collection cycle, Represent the minimum and maximum periods of prediction, is the variance of the internal channel state parameters in the previous cycle.

[0013] Preferably, constructing a channel state assessment model based on the collected multi-source link state parameters in step S2 includes:

[0014] First, the link state parameters collected from each channel are normalized;

[0015] Construct the channel feature vector based on the normalized state parameters ,in represents the channel idleness parameter of the i-th channel, represents the signal strength parameter of the i-th channel, represents the bit error rate parameter of the i-th channel, represents the interference index parameter of the i-th channel;

[0016] Finally, the improved LSTM algorithm model is used to establish a channel state assessment model and train it.

[0017] Preferably, the specific implementation of the improved LSTM algorithm model is:

[0018] First, a spatial feature embedding module is added before the LSTM input layer. The geographic coordinates of each communication node are obtained through the Hongmeng soft bus. The coordinates are mapped into frequency domain feature vectors through Fourier transform and then concatenated with the link state parameters of the time series to form a multidimensional input tensor.

[0019] Improve the LSTM forget gate structure and adaptively adjust the weight of the forget gate according to the current channel interference index. The calculation method is: ,in represents the activation function, Represents the weight of the forget gate, The hidden state at the previous moment, is a multi-dimensional input tensor, represents the bias of the forget gate, represents the regulating factor, Represents the channel interference index at the current moment;

[0020] Insert the channel attention module before the LSTM output layer to obtain weighted features;

[0021] Finally, the weighted features are mapped to a single dimension through a fully connected layer, and the normalized channel score value is output.

[0022] Preferably, the specific steps of inserting the channel attention module before the LSTM output layer to obtain the weighted features are:

[0023] First, the feature tensor output by the last hidden layer of LSTM is transformed in dimension, and channel descriptors are generated respectively through the global average pooling layer and the global maximum pooling layer;

[0024] The two channel descriptors are concatenated and input into a two-layer fully connected network. The first layer uses the ReLU activation function for dimensionality reduction, and the second layer uses the Sigmoid activation function to generate channel attention weights.

[0025] Finally, the generated attention weights are multiplied element-by-element with the output features to obtain the weighted features.

[0026] Preferably, each scoring cycle in step S2 consists of 10 collection cycles.

[0027] Preferably, the step of calculating channel stability by a state evaluation matrix and completing channel allocation in combination with task priority and data type includes the following steps:

[0028] First, the score sequence of each channel is extracted from the state evaluation matrix, and the channel stability is calculated according to the fluctuation amplitude of its score value;

[0029] Analyze the priority and data type of the task to be scheduled, determine the sensitive requirements for channel performance and stability based on the communication demand characteristics corresponding to the task, and generate the task scheduling constraint set;

[0030] The score value and channel stability of each channel in the current cycle are matched and analyzed with the task scheduling constraint set to obtain a set of schedulable candidate channels;

[0031] In the candidate channel set, the optimal communication channel is determined and allocated based on the score value and the sorting weight set by the task priority level.

[0032] Compared with the prior art, the advantages and positive effects of the present invention are:

[0033] 1. Real-time collection of multi-dimensional channel state parameters through distributed modules, combined with an improved LSTM model to integrate spatiotemporal features, enables accurate assessment of channel status, breaking through the limitations of traditional static allocation that lacks real-time perception and intelligent analysis.

[0034] 2. Dynamically match channel resources based on channel stability, task priority, and data type to improve channel utilization and ensure communication quality for high-priority tasks, solving the problems of static resource allocation waste and inefficient scheduling.

[0035] 3. Real-time monitoring of the status of allocated channels, triggering rapid reassessment and switching, proactively responding to channel fluctuations, significantly enhancing communication stability, and addressing the shortcomings of traditional solutions that cannot dynamically adapt to environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0037] Figure 1 This is a structural flow chart of a dynamic channel allocation method for communication links based on the Hongmeng system. DETAILED DESCRIPTION

[0038] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0039] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0040] In the Internet of Everything environment supported by HarmonyOS, there are a large number of distributed terminals and devices that need to communicate. Traditional static channel allocation cannot adapt to the complex environment of high concurrency, link fluctuations, and multi-source access. In order to cope with the problems of complex channel status, frequent device access, and severe link fluctuations in the distributed communication scenario of the HarmonyOS system, the present invention provides a method for dynamic channel allocation of communication links based on the HarmonyOS system. The specific implementation steps are as follows: Figure 1 shown.

[0041] First, in order to realize the channel status collection and synchronization across communication nodes, a collaborative communication monitoring module is established in multiple communication nodes running the Hongmeng system to collect the link status parameters of the communication channel between the local node and the surrounding nodes in real time. The link status parameters include: channel idleness, signal strength, bit error rate, and interference index. The interference index is calculated by measuring the proportion of unexpected frequency band energy in the received signal and combining it with the noise baseline. The collaborative communication monitoring module adopts a distributed deployment architecture and realizes cross-node data synchronization through the soft bus technology of the Hongmeng system. The collection period of each node is dynamically matched with the frequency of channel status changes. The calculation method is: ,in is the current node collection cycle, Represent the minimum and maximum periods of prediction, is the variance of the internal channel state parameters in the previous cycle. Through the distributed deployment architecture and Hongmeng soft bus technology, collaborative monitoring data synchronization between multiple communication nodes is achieved, and the collection cycle of each node is dynamically adjusted based on the frequency of channel state changes, thereby optimizing system resource utilization while ensuring real-time monitoring, and avoiding resource waste or state lag problems caused by fixed-cycle collection. Specifically, the collection cycle is dynamically calculated through the parameter variance of the previous cycle, so that the node automatically shortens the cycle to sample at a high frequency when the channel state fluctuates violently, and extends the cycle to reduce power consumption when the state is stable, thereby improving the monitoring efficiency and adaptability of the Hongmeng system in the distributed communication environment.

[0042] Then, in order to convert the multi-source heterogeneous link state parameters into a unified evaluation index, a channel state evaluation model is constructed based on the collected multi-source link state parameters. The construction method of the channel state evaluation model is to first normalize the link state parameters collected by each channel, and then construct a channel feature vector based on the normalized state parameters. ,in represents the channel idleness parameter of the i-th channel, represents the signal strength parameter of the i-th channel, represents the bit error rate parameter of the i-th channel, Represents the interference index parameter of the i-th channel. Finally, the improved LSTM algorithm model is used to establish a channel state evaluation model and train it. Specifically, first of all, in the actual communication environment, the link state parameters collected by each channel have problems such as inconsistent physical magnitudes and large differences in value ranges. Directly inputting the original data will cause model learning deviation and affect the overall prediction accuracy. Therefore, this step first normalizes the link state parameters of all channels to ensure that the features are comparable in the numerical dimension. Specifically, the system calculates the historical maximum and minimum values of each parameter, and uses the Min-Max normalization method to map it to the [0,1] interval. After normalization, the four core link state parameters are spliced in a predetermined order to construct a channel feature vector in a unified format to characterize the overall operating status of each channel in the current time slice. For the i-th channel, its feature vector can be defined as ,in, represents the channel idleness parameter of the i-th channel, represents the signal strength parameter of the i-th channel, represents the bit error rate parameter of the i-th channel, represents the interference index parameter for the i-th channel. This vector, serving as the model's input feature, possesses a clear semantic structure and physical meaning, enabling complete channel state representation in the subsequent LSTM model. The LSTM algorithm model is then improved to establish a channel state assessment model. Specifically, a spatial feature embedding module is added before the LSTM input layer. The geographic coordinates of each communication node are acquired via the HarmonyOS soft bus. These coordinates are then mapped into frequency-domain feature vectors via a Fourier transform, and then concatenated with the time series link state parameters to form a multidimensional input tensor. Specifically, the system first uses the HarmonyOS soft bus technology to acquire the location information of each communication node in real time, including latitude and longitude or relative coordinates, and normalizes them into two-dimensional vectors in a unified coordinate system. Given the difficulty in expressing periodicity and similarity in spatial location information in the original coordinate system, a Fourier transform is used to map the coordinate vectors into the frequency domain to extract their spatial spectrum features. Frequency-domain vectors can more effectively represent the spatial coupling relationships and distribution patterns between nodes, making them suitable for feature fusion with time series models. Finally, the resulting frequency-domain spatial feature vectors are concatenated with the link state parameters collected by each channel at consecutive moments, forming a unified multidimensional input tensor. This tensor incorporates both temporal dynamics and spatial structural information, providing richer contextual semantic input for the LSTM model. The LSTM forget gate structure is then improved, adaptively adjusting the forget gate weight based on the current channel interference index. This is calculated as follows: ,in represents the activation function, Represents the weight of the forget gate, The hidden state at the previous moment, is a multi-dimensional input tensor, represents the bias of the forget gate, represents the regulating factor, represents the channel interference index at the current moment. A channel attention module is then inserted before the LSTM output layer to obtain weighted features. Specifically, the feature tensor output by the last hidden layer of the LSTM is first transformed to adapt it to the input format of the channel attention module. Global average pooling and global max pooling are then applied to the feature tensor to extract the average and maximum responses of each channel over the entire temporal dimension, respectively. This generates two channel descriptors, representing the overall contribution and local significance of the channel, respectively. These two descriptors are concatenated and input to a weight generator consisting of two fully connected layers. The first layer uses the ReLU activation function for dimensionality reduction to extract high-order semantic features, while the second layer uses the Sigmoid activation function to output normalized channel attention weights. Finally, these weights are element-wise multiplied with the original feature tensor using a broadcast mechanism to obtain the weighted feature output. This mechanism enables the model to automatically focus on feature channels that are more critical for channel state judgment, effectively suppressing redundant information interference, thereby improving the accuracy and generalization of channel state scores. Finally, the weighted features are mapped to a single dimension through a fully connected layer, outputting a normalized channel score. Specifically, the weighted feature tensor is first flattened into a one-dimensional vector, which serves as the input to the fully connected layer. A linear transformation is then performed to obtain a single scalar output. To ensure that the output value has a uniform scale range and facilitate subsequent normalization and horizontal comparison, the output value is mapped to the interval [0, 1] using a sigmoid activation function to obtain the final channel score. This score provides the core quantitative basis for subsequent channel stability assessment, task priority matching, and dynamic channel allocation.

[0043] Next, the constructed model is used to generate the state evaluation matrix of all current channels. The process of generating the state evaluation matrix of all current channels through the channel state evaluation model includes the following steps: first, a unique channel number is assigned to each channel to be evaluated, and a corresponding data input channel is established, and the link state parameters of the channel are input into the channel state evaluation model. The evaluation model outputs the score value corresponding to each channel respectively, and all the score values are arranged in the order of the channel number to form a one-dimensional scoring vector. Finally, in each round of scoring cycle, the step of forming a one-dimensional scoring vector is executed for all channels, and all the scoring vectors are vertically spliced according to the time dimension to form a two-dimensional channel state evaluation matrix. Each round of scoring cycle consists of 10 acquisition cycles, and each update of the acquisition cycle is an update of the scoring cycle, that is, the 1st to 10th rounds are scoring cycles, and the 2nd to 11th rounds are new scoring cycles.

[0044] Channel stability is then calculated using the state assessment matrix, and channel allocation is performed based on task priority and data type. First, a score sequence for each channel is extracted from the state assessment matrix, and channel stability is calculated based on the fluctuation of the score values. Next, the priority and data type of the scheduled task are analyzed. Based on the corresponding communication requirements, the sensitivity requirements for channel performance and stability are determined, and a set of task scheduling constraints is generated. Furthermore, the score and channel stability of each channel in the current cycle are matched with the task scheduling constraints to obtain a set of schedulable candidate channels. Finally, within this set of candidate channels, the optimal communication channel is determined based on the ranking weights set based on the score and task priority. Specifically, after constructing the channel state assessment matrix, the system first extracts the score values of each channel over multiple consecutive acquisition cycles to form a score time series. To assess the stability of this channel, the score series is subjected to a volatility analysis, and its standard deviation is calculated as a stability indicator. A small standard deviation indicates low fluctuation in the channel score between cycles, indicating stable channel performance. Conversely, large fluctuations indicate potentially drastic changes in channel quality and low communication reliability. After calculating channel stability, the system analyzes the currently scheduled communication tasks, extracting their priority levels and data type attributes. Based on these attributes, a set of task scheduling constraints is established. Task priorities can be categorized into multiple levels (such as high, medium, and low) based on system configuration, representing the task's level of competition for communication resources and its real-time requirements. Data types refer to the data transmission characteristics of the task, such as video streams, status commands, and heartbeat packets, which have varying sensitivities to channel bandwidth, latency, and bit error rate. The system maps different task types to their required communication characteristics by consulting a task attribute table. For example, high-priority video tasks require high channel bandwidth and low bit error rate, while low-priority logging tasks are more tolerant to latency. Finally, task attributes are formalized into a set of scheduling constraint parameters, including minimum channel score thresholds, maximum allowable score fluctuations, and a lower limit on the channel stability coefficient. This scheduling constraint set serves as a rule template for selecting available channels, ensuring a good match between tasks and channels in terms of performance and risk, and preventing issues such as packet loss, latency, or scheduling failures caused by channel mismatch. Based on the generated channel score and stability, as well as the task scheduling constraint set, all channels in the current cycle are matched and analyzed one by one. Specifically, channels with score values lower than the minimum threshold of the task or stability lower than the set standard are first filtered out; then, based on the task requirements for parameters such as bit error rate and signal strength, a set of channels that meet specific dimensional indicators is further extracted on the premise of meeting the score and stability. The matching process supports the configuration of soft and hard conditions, where key constraints (such as the minimum score) are hard conditions that must be met; and non-key constraints (such as channel idleness) are weighted and only participate in subsequent decision-making as a ranking basis. Finally, the set of channel numbers that meet all hard constraints and give priority to soft constraints are output as a set of schedulable candidate channels.Finally, after obtaining the set of schedulable candidate channels, the candidate channels are ranked and selected based on the task priority level and channel score to determine the final communication channel allocation result. The specific method is: first, a comprehensive scoring index is calculated for each candidate channel. This index is a weighted sum of the channel's current score, stability, and task priority. All candidate channels are sorted in descending order according to this comprehensive scoring index. The channel with the highest score is selected as the target communication channel for binding and allocation.

[0045] Finally, during the communication process, the real-time status parameters of the allocated channels are continuously monitored. If the difference between the channel score value and the mean value of the corresponding scoring period at the time of allocation exceeds the set threshold, the rapid re-evaluation mechanism is triggered to re-evaluate the channel and perform channel allocation to ensure communication stability. Specifically, during the execution of the communication task, the state parameters of the allocated channels are continuously monitored in real time, and their current score values are calculated regularly. If it is found that the score value of a certain channel deviates from the historical mean value of the scoring period to which it belongs at the time of allocation by more than the set threshold, it indicates that the channel may be interfered with or its performance is degraded. At this time, the rapid re-evaluation mechanism is immediately triggered, the channel state evaluation model is re-called, and the current available channels are instantly scored and stability calculated. In combination with the scheduling constraint set of the current task, the new optimal channel is screened and allocated. Through this mechanism, active adjustments can be made before the communication quality deteriorates, avoiding communication interruption or data loss due to channel degradation, and significantly improving the system's dynamic adaptability and communication stability.

[0046] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any person skilled in the art may utilize the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes for application in other fields. However, any simple modification, equivalent change, and modification of the above embodiments made in accordance with the technical essence of the present invention without departing from the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for dynamic channel allocation of communication links based on Hongmeng system, characterized in that: The following steps are involved: S1. Establish a collaborative communication monitoring module in multiple communication nodes running the Hongmeng system to collect link status parameters of the communication channel between the local node and surrounding nodes in real time. The link status parameters include: channel idleness, signal strength, bit error rate, and interference index; S2. Construct a channel state assessment model based on the collected multi-source link state parameters. The model is used to generate a state assessment matrix for all current channels. The specific implementation process includes the following steps: Assign a unique channel number to each channel to be evaluated, establish a corresponding data input channel, and input the link state parameters of the channel into the channel state evaluation model; The evaluation model outputs the score value corresponding to each channel respectively, and all the score values are arranged in the order of channel numbers to form a one-dimensional score vector; In each scoring cycle, the step of forming a one-dimensional scoring vector is performed on all channels, and all scoring vectors are spliced vertically along the time dimension to form a two-dimensional channel state evaluation matrix; S3. Calculate channel stability through the state evaluation matrix and complete channel allocation based on task priority and data type; S4. During the communication process, the real-time status parameters of the allocated channels are continuously monitored. If the difference between the score value and the average value of the corresponding scoring period at the time of allocation exceeds the set threshold, the rapid re-evaluation mechanism is triggered to re-evaluate the channel and re-allocate the channel to ensure communication stability. The process of calculating channel stability by the state evaluation matrix and completing channel allocation in combination with task priority and data type in step S3 includes the following steps: First, the score sequence of each channel is extracted from the state evaluation matrix, and the channel stability is calculated according to the fluctuation amplitude of its score value; Analyze the priority and data type of the task to be scheduled, determine the sensitive requirements for channel performance and stability based on the communication demand characteristics corresponding to the task, and generate the task scheduling constraint set; The score value and channel stability of each channel in the current cycle are matched and analyzed with the task scheduling constraint set to obtain a set of schedulable candidate channels; In the candidate channel set, the optimal communication channel is determined and allocated based on the score value and the sorting weight set by the task priority level.

2. A method for dynamic channel allocation of communication links based on Hongmeng system according to claim 1, characterized in that: The collaborative communication monitoring module adopts a distributed deployment architecture and realizes cross-node data synchronization through the soft bus technology of the Hongmeng system. The collection cycle of each node is dynamically matched with the frequency of channel state changes. The calculation method is: Where T t is the current node collection period, T min , T max Represent the minimum and maximum periods of the forecast, var(S t-1 ) is the variance of the internal channel state parameters in the previous cycle.

3. The method for dynamic channel allocation of a communication link based on the Hongmeng system according to claim 1 is characterized in that: The step S2 includes constructing a channel state assessment model based on the collected multi-source link state parameters: First, the link state parameters collected from each channel are normalized; Construct the channel feature vector V according to the normalized state parameters i =[P1(i), P2(i), P3(i), P4(i)], where P1(i) represents the channel idleness parameter of the i-th channel, P2(i) represents the signal strength parameter of the i-th channel, P3(i) represents the bit error rate parameter of the i-th channel, and P4(i) represents the interference index parameter of the i-th channel; Finally, the improved LSTM algorithm model is used to establish a channel state assessment model and train it.

4. A method for dynamic channel allocation of communication links based on Hongmeng system according to claim 3, characterized in that: The specific implementation of the improved LSTM algorithm model is: First, a spatial feature embedding module is added before the LSTM input layer. The geographic coordinates of each communication node are obtained through the Hongmeng soft bus. The coordinates are mapped into frequency domain feature vectors through Fourier transform and then concatenated with the link state parameters of the time series to form a multidimensional input tensor. Improve the LSTM forget gate structure and adaptively adjust the weight of the forget gate according to the current channel interference index. The calculation method is: f t =σ(W f ·[h t-1 ,x t ]+b f )×(1-λ·G t ), where σ represents the activation function, W f Represents the weight of the forget gate, h t-1 The hidden state at the previous moment, x t is a multi-dimensional input tensor, b f Represents the bias of the forget gate, λ represents the adjustment factor, G t Represents the channel interference index at the current moment; Insert the channel attention module before the LSTM output layer to obtain weighted features; Finally, the weighted features are mapped to a single dimension through a fully connected layer, and the normalized channel score value is output.

5. A method for dynamic channel allocation of communication links based on Hongmeng system according to claim 4, characterized in that: The specific steps of inserting the channel attention module before the LSTM output layer to obtain the weighted features are as follows: First, the feature tensor output by the last hidden layer of LSTM is transformed in dimension, and channel descriptors are generated respectively through the global average pooling layer and the global maximum pooling layer; The two channel descriptors are concatenated and input into a two-layer fully connected network. The first layer uses the ReLU activation function for dimensionality reduction, and the second layer uses the Sigmoid activation function to generate channel attention weights. Finally, the generated attention weights are multiplied element-by-element with the output features to obtain the weighted features.

6. A method for dynamic channel allocation of communication links based on Hongmeng system according to claim 1, characterized in that: Each scoring cycle in step S2 consists of 10 collection cycles.

Images