APN6 network dynamic switching optimization method based on QoS prediction

Abstract

The invention relates to the technical field of next generation network communication, and particularly provides an APN6 network dynamic switching optimization method and system based on QoS prediction, a terminal and a medium, and the method comprises the steps: collecting QoS indexes of different nodes of a link in an APN6 network, constructing a time dimension prediction model, and enabling the model to output a prediction time result; constructing a spatial dimension prediction model, inputting a QoS index into the model, and outputting a prediction space result; constructing a dynamic dimension prediction model, inputting a QoS index into the model, and outputting a prediction dynamic result used for updating model parameters; and weighting and synthesizing the time prediction result, the space prediction result and the dynamic prediction result to obtain a predicted link quality score, and triggering APN6 path switching when the predicted link quality score is less than a preset threshold. The APN6 network resource scheduling efficiency and the service quality are effectively improved, and the stability and the reliability of service transmission are guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of next-generation network communication, and particularly relates to an APN6 network dynamic handover optimization method, system, terminal and medium based on QoS prediction. Background Art

[0002] APN (Access Point Name) is the access point name that a mobile device (such as a mobile phone, a tablet computer) needs to configure when connecting to a mobile network. Due to changes in network service types, usage scenarios, or device functions, to connect the device to a network channel that better matches the current requirements, APN handover is usually performed.

[0003] With the deployment of 5G networks and the large-scale application of IPv6, APN6 has gradually become one of the core configurations for mobile data services. APN6 (Access Point Name 6) is an access point name designed for the IPv6 network environment, mainly used to support terminals to access the IPv6 Internet through the mobile network.

[0004] Traditional APN6 handover relies on static policies (such as manual selection by users or pre-configuration), and cannot adapt to dynamic network environments (such as sudden congestion scenarios). Moreover, in the IPv6 environment, APN6 has multiple attributes such as bandwidth, delay, and security, and it is difficult to optimize multiple attributes collaboratively to achieve the best network performance. Summary of the Invention

[0005] Aiming at the above deficiencies of the prior art, the present invention provides an APN6 network dynamic handover optimization method, system, terminal and medium based on QoS prediction to solve the above technical problems.

[0006] In the first aspect, the present invention provides an APN6 network dynamic handover optimization method based on QoS prediction, including: Collecting QoS metrics of different nodes on the link in the APN6 network based on a preset sampling strategy, where the QoS metrics include microsecond-level one-way delay, jitter based on the calculation result of IETF RFC 3393, sliding window packet loss rate, current flow rate, and the DSCP field taken from the APN6 header; Constructing a time dimension prediction model based on a time convolutional network combined with a temporal attention mechanism, where the model inputs the QoS metrics and outputs a prediction time result for predicting the QoS trend in the short term in the future; Constructing a spatial dimension prediction model based on a graph neural network, where the model inputs the QoS metrics and outputs a prediction spatial result for predicting the mutual influence intensity of each node in the network topology; Build a dynamic dimension prediction model based on sliding window incremental learning. The model takes QoS metrics as input and outputs prediction dynamic results for updating model parameters. Weightedly synthesize the time prediction result, space prediction result, and dynamic prediction result to obtain a predicted link quality score. When the predicted link quality score is less than a preset threshold, trigger the APN6 path switch.

[0007] In an alternative implementation, the sampling strategy includes a basic sampling period and burst traffic triggering. The basic sampling period is a preset value. Burst traffic triggering means that when the flow rate change rate is greater than 15%, sample once every 1 ms.

[0008] In an alternative implementation, the time dimension prediction model includes a dilated convolutional layer and a multi-head attention layer. The output of the predicted time result for predicting the QoS trend in the short term in the future specifically includes: Preprocess the QoS metrics to obtain a time series data matrix, which includes batch size, time step, and feature dimension. Set different dilation rates, dilate the convolutional layer by inserting holes between the convolutional kernel elements based on the dilation rate, and extract time series features at different time scales to obtain a sequence that fuses multi-scale features. Use the multi-head attention mechanism to project the sequence that fuses multi-scale features onto multiple attention heads, calculate the attention scores of each attention head respectively, and then splice them according to the feature dimension to obtain the feature with time series weights as the predicted time result for predicting the QoS trend in the short term in the future.

[0009] In an alternative implementation, the output of the predicted space result for the mutual influence intensity of each node in the predicted network topology specifically includes: Construct a node feature matrix with the shape of [number of nodes, feature dimension], and store the QoS metrics of each node in the network in the matrix; construct the adjacency matrix of each node. Calculate the feature similarity between all nodes based on the dot product operation combined with the node feature matrix. Calculate the product of the feature similarity of each node and the adjacency matrix, and convert the product to the attention degree of each node to different adjacent nodes through softmax. Each node aggregates the features of adjacent nodes weighted by the attention degree to obtain new features. Based on the attention mechanism of the GAT layer, perform attention calculation and multi-head attention fusion on the new features, and then output the node-to-node weight matrix as the predicted space result for the mutual influence intensity of each node in the predicted network topology.

[0010] In an alternative embodiment, outputting the predicted dynamic result for updating the model parameters specifically includes: Store the latest QoS metrics based on a fixed-size array. When the array is filled with data and new data arrives, the earliest data is overwritten; When the sliding window is filled with data, obtain the prediction result through forward propagation of the model based on the current batch of data, and calculate the loss function for this batch of data; Based on the loss function, calculate the gradient of the loss with respect to all parameters through the backpropagation algorithm; Accumulate the gradients of the current batch into an accumulated gradient variable. When the accumulated number of batches reaches a preset value, update the model parameters in combination with the learning rate as the predicted dynamic result; Calculate the ratio of the current batch loss to the average loss of the previous N batches, and adjust the learning rate according to the ratio.

[0011] In an alternative embodiment, when the predicted link quality score is less than a preset threshold, the specific steps for triggering the APN6 path switch include: Expand relevant message types in the APN6-CP protocol header for path switch control. The message types include: PRE_SWITCH_REQ, PATH_RESERVE_ACK, and FAST_FAILOVER; When the predicted link quality score is less than the preset threshold, the requester sends a PRE_SWITCH_REQ message to the target node. This message carries the link quality prediction result and a reservation request for the target path resources; After receiving the PRE_SWITCH_REQ message, if the target node's resources meet the reservation requirements of the requester, it returns a PATH_RESERVE_ACK message to confirm the reserved resources; When the requester receives the PATH_RESERVE_ACK message and confirms the successful resource reservation, establish a low-latency backup path based on the target node with the reserved resources; The control plane sends a FAST_FAILOVER fast switch execution instruction to the data plane, instructing the data plane to preferentially forward data to the pre-established low-latency backup path based on the identifier, while maintaining the synchronization of the original path session state; After receiving the PRE_SWITCH_REQ message, if the target receiving node does not have enough resources, it returns a PATH_RESERVE_NACK message to inform the requester to switch to the backup path.

[0012] In an alternative embodiment, the switch of the backup path specifically includes: After receiving the PATH_RESERVE_NACK message, the requester starts to query the local path table for the backup path; After obtaining the backup path information, the requester updates the forwarding rules of the data plane; Migrate the session state of the original path to the alternate path to complete the handover operation, and conduct local tests to confirm that data can be normally transmitted on the alternate path.

[0013] In a second aspect, the present invention provides an APN6 network dynamic handover optimization system based on QoS prediction. When the system is implemented, the above-mentioned APN6 network dynamic handover optimization method based on QoS prediction is executed. The system includes: A data acquisition module that collects QoS metrics of different nodes on the links in the APN6 network based on a preset sampling strategy. The QoS metrics include microsecond-level one-way delay, jitter based on the calculation result of IETF RFC 3393, sliding window packet loss rate, current flow rate, and the DSCP field taken from the APN6 header. A model construction module that constructs a time dimension prediction model based on a temporal convolutional network combined with a temporal attention mechanism. The model takes QoS metrics as input and outputs a prediction time result for predicting the QoS trend in the short term in the future; constructs a spatial dimension prediction model based on a graph neural network. The model takes QoS metrics as input and outputs a prediction spatial result for predicting the mutual influence intensity of each node in the network topology; constructs a dynamic dimension prediction model based on sliding window incremental learning. The model takes QoS metrics as input and outputs a prediction dynamic result for updating the model parameters. A path handover module that obtains a predicted link quality score by comprehensively weighting the time prediction result, spatial prediction result, and dynamic prediction result. When the predicted link quality score is less than a preset threshold, it triggers an APN6 path handover.

[0014] In a third aspect, a terminal is provided, including: A processor and a memory, where The memory is used to store a computer program, The processor is used to call and run the computer program from the memory, so that the terminal executes the above-mentioned method of the terminal.

[0015] In a fourth aspect, a computer-readable storage medium is provided. Instructions are stored in the computer-readable storage medium. When it runs on a computer, it causes the computer to execute the methods described in the above aspects.

[0016] The beneficial effects of the present invention are as follows. The method, system, terminal, and medium for optimizing the dynamic switching of the APN6 network based on QoS prediction provided by the present invention accurately collect the QoS indicators of the APN6 network link through a preset sampling strategy, and use a temporal convolutional network, temporal attention, a graph neural network, and sliding window incremental learning to construct three-dimensional prediction models respectively. The data is analyzed from multiple perspectives of time trend, spatial influence, and dynamic change, and the link quality score is obtained by weighted comprehensive prediction results. When the score is lower than the threshold, path switching is triggered, realizing comprehensive and real-time monitoring and intelligent prediction of the network state, effectively improving the APN6 network resource scheduling efficiency and service quality, and ensuring the stability and reliability of service transmission.

[0017] In addition, the design principle of the present invention is reliable, the structure is simple, and it has a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a schematic flowchart of the method for optimizing the dynamic switching of the APN6 network based on QoS prediction according to an embodiment of the present invention.

[0020] Figure 2 It is the hybrid prediction model architecture in the embodiment of the present application.

[0021] Figure 3 It is a schematic block diagram of the system for optimizing the dynamic switching of the APN6 network based on QoS prediction according to an embodiment of the present invention.

[0022] Figure 4 It is a schematic structural diagram of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0024] KEY TERMS: QoS (Quality of Service) is a key concept in the field of network communication, used to measure the service guarantee ability provided by the network for specific traffic flows, involving multiple indicators such as bandwidth, latency, packet loss rate, etc. Its core goal is to ensure that different types of services (such as voice, video, data, etc.) obtain differential service guarantees under the condition of limited network resources.

[0025] Unless otherwise defined, all technical and scientific terms used in this article have the same meaning as those commonly understood by those skilled in the technical field to which this invention belongs. The terms used in the specification of this invention herein are only for the purpose of describing specific embodiments, and are not intended to limit this invention.

[0026] The APN6 network dynamic switching optimization method based on QoS prediction provided by the embodiments of this invention is executed by a computer device. Correspondingly, the APN6 network dynamic switching optimization system based on QoS prediction runs in the computer device.

[0027] Figure 1 It is a schematic flowchart of the APN6 network dynamic switching optimization method based on QoS prediction according to an embodiment of this invention. Among them, Figure 1 The execution subject can be an APN6 network dynamic switching optimization system based on QoS prediction. According to different requirements, the order of the steps in this flowchart can be changed, and some can be omitted.

[0028] As Figure 1 and Figure 2 shown, this method includes: Step S1, collect QoS metrics of different nodes on the link in the APN6 network based on a preset sampling strategy. The QoS metrics include microsecond-level one-way latency, jitter based on the calculation result of IETF RFC 3393, sliding window packet loss rate, current flow rate, and the DSCP field taken from the APN6 header; After collecting the QoS metrics, preprocessing of the metric data is required: use the NTPv4 protocol to calibrate the system clock with an error less than 1 ms to achieve the accuracy of the sampling period. If an outlier with a large difference from the previous sampling period is encountered, record the log and take the average value of the values of the adjacent two sampling periods as the value of this sampling period. If continuous sampling anomalies occur, an anomaly alarm needs to be reported.

[0029] Step S2, construct a time dimension prediction model based on a temporal convolutional network combined with a temporal attention mechanism. The model inputs the QoS metrics and outputs a prediction time result that predicts the QoS trend in the short term in the future; By capturing the dynamic rules of QoS indicators (such as microsecond-level delay and jitter), we can accurately predict the trend of network performance in a short period of time. For example, we can find potential risks such as sudden increase in link delay and increased jitter in advance, provide "time advance" for path switching, and avoid business interruption caused by delayed real-time response. This is especially suitable for real-time services that are sensitive to delay (such as industrial control and video conferencing).

[0030] A spatial dimension prediction model is built based on a graph neural network. The model inputs QoS indicators and outputs the prediction spatial results of the mutual influence intensity of each node in the predicted network topology. Using the network topology as a graph structure, the model models the mutual influence strength between nodes and explores the spatial dependency of link quality (such as the cascading impact of an increase in the packet loss rate of a node on surrounding nodes). Different from the isolated analysis of a single node, this model can determine the spread of network congestion from a global perspective, avoid misjudgment due to abnormal local indicators, improve the accuracy of switching decisions, and is suitable for network resource scheduling under complex topologies.

[0031] A dynamic dimension prediction model is built based on sliding window incremental learning. The model inputs QoS indicators and outputs dynamic prediction results for updating model parameters. By continuously receiving new QoS data and updating model parameters, it adapts to dynamic changes in network traffic (such as burst traffic and business tidal effects) to avoid model failure due to data distribution drift. For example, when network load fluctuates frequently, the model can adjust the prediction logic in real time to ensure long-term prediction accuracy, reduce the risk of misjudgment caused by "outdated model", and improve the robustness of the system in dynamic scenarios.

[0032] Step S3, weighted integration of the time prediction result, the space prediction result and the dynamic prediction result to obtain a predicted link quality score, and when the predicted link quality score is less than a preset threshold, triggering the APN6 path switching.

[0033] Optionally, as an embodiment of the present invention, before step S1, a dedicated hardware embedded probe module is integrated in the network processing unit of the router interface chip to support parallel processing of 8*100Gbps interfaces, and a P4 programmable pipeline is used to achieve zero-copy data acquisition, reducing CPU intervention.

[0034] Optionally, as an embodiment of the present invention, in the time dimension prediction model: Microsecond one-way delay: As a timing-sensitive indicator, the time series fluctuation of delay directly reflects the change in link transmission efficiency. The model captures the historical trend of delay (such as continuous rise / fall) and sudden jumps, predicts the risk of delay growth in a short period of time, and provides a basis for real-time business scheduling.

[0035] Jitter based on IETF RFC 3393: Jitter is essentially the rate of change of latency. The model analyzes the temporal distribution of jitter (such as variance, peak interval) to predict the deteriorating trend of latency stability. For example, it can identify in advance the increase in jitter caused by network congestion to avoid affecting the quality of streaming services such as voice and video.

[0036] Current flow rate: The time series of the flow rate reflects the change in traffic load. The model combines historical flow rate peaks and periodic patterns (such as morning and evening rush hours) to predict the congestion that may be caused by a short-term traffic surge, and assist in judging the changing trends of latency and packet loss rate.

[0037] In the spatial dimension prediction model: Sliding window packet loss rate: The packet loss rate is a direct manifestation of node or link failures. The graph neural network uses the packet loss rate of each node as a graph node feature, and calculates the spatial propagation effect of the packet loss rate through the edge weights (topological connection relationships) between nodes. For example, when the packet loss rate of a core node increases, the model can predict its cascading impact on adjacent nodes to avoid misjudging local failures as global problems.

[0038] DSCP field (service differentiation identifier in the APN6 header): The DSCP identifies the service priority (such as real-time service vs. non-real-time service). The model uses it as a node attribute and combines the topological structure to analyze the resource competition relationship of different priority services in the network. For example, the DSCP field of high-priority services can affect the model's calculation of the resource allocation weights between nodes, thereby more accurately predicting the QoS degradation risk of low-priority services.

[0039] In the dynamic dimension prediction model: Incremental learning continuously ingests the latest data of metrics such as latency, jitter, packet loss rate, flow rate, and DSCP based on a sliding window, and updates the model parameters in real time. For example, when a sudden network traffic causes a sharp increase in the flow rate, the model can quickly adjust the prediction weights for the packet loss rate and latency to avoid prediction deviations caused by "outdated" historical data, and ensure robustness in scenarios of traffic fluctuations.

[0040] The model can capture the real-time coupling relationship between QoS metrics (such as the chain reaction of flow rate surge → latency increase → packet loss rate increase), and continuously optimize the correlation weights between metrics through incremental learning. For example, when the distribution of DSCP changes due to the launch of a new service, the model can dynamically adjust the prediction logic of QoS metrics corresponding to different priority services to improve the prediction accuracy in dynamic scenarios.

[0041] Optionally, as an embodiment of the present invention, the sampling strategy includes a basic sampling period and burst traffic triggering; The basic sampling period is a preset value (in this embodiment, 10 ms is used, and this value can be dynamically adjusted); The burst traffic trigger is that when the flow rate change rate is greater than 15%, high-frequency sampling at the millisecond level is started, sampling once every 1 ms.

[0042] Optionally, as an embodiment of the present invention, in step S2, the construction process of the three prediction models is as follows: Determine the prediction task type (objective function) and define the evaluation metrics; Collect historical QoS metric data and perform feature extraction, and divide the data into a training set, a validation set, and a test set; After selecting a suitable model architecture, train the model based on the training set, and continuously correct the model through the validation set to obtain the trained model; After testing the trained model based on the test set, evaluate the performance of the model and correct the parameters of the model to obtain the final prediction model.

[0043] Optionally, as an embodiment of the present invention, the time dimension prediction model includes a dilated convolutional layer and a multi-head attention layer, and the time dimension prediction model algorithm is implemented as follows: class TemporalPredictor(nn.Module): def __init__(self): self.tcn = TCNBlock(dilation=[1,2,4,8]) self.attention = MultiHeadAttention(heads=4) def forward(self, x): x = self.tcn(x) # [B, T, C] return self.attention(x, x, x) The specific prediction time results for outputting the predicted QoS trend in the short term in the future include: After preprocessing the QoS metrics, a time series data matrix is obtained, and the time series data matrix includes the batch size, the time step, and the feature dimension; Set different dilation rates, and based on the dilation rates, insert holes between the convolutional kernel elements to dilate the convolutional layer, and extract time series features at different time scales to obtain a sequence that fuses multi-scale features; assuming the input feature map is X, taking dilation (dilation rate) = 2 as an example, the specific calculation steps are as follows: When the convolution kernel size is k = 3 and dilation = 2, the actual convolution stride is k+(k - 1)×(dilation - 1)=3 + 2×1 = 5 time steps, but sampling is only performed at the original kernel positions (with an interval of 1 time step).

[0044] For each time step t, the convolution operation aggregates the features at three positions: t - 2, t, and t + 2 (with a stride of 5 time steps, but only taking 3 points). The formula is: , where W is the convolution kernel weight and b is the bias.

[0045] The four - layer dilated convolution with dilation = [1, 2, 4, 8] processes the input in sequence, capturing the dependencies at 1, 2, 4, and 8 time steps respectively, and finally outputs a sequence that fuses multi - scale features.

[0046] Using the multi - head attention mechanism, project the sequence of fused multi - scale features onto multiple attention heads, calculate the attention scores for each attention head respectively, and then concatenate them according to the feature dimension to obtain the features with temporal weights as the prediction time result for predicting the QoS trend in the short future.

[0047] Let the input features be Q, K, V (Q = query vector, K = key vector, V = value vector, usually Q = K = V = X). Map the input to query, key, and value vectors: , , , where, 、 、 , / number of attention heads; Calculate the correlation between time steps through dot - product to obtain the attention scores: , where, is the scaling factor to avoid the softmax gradient vanishing caused by large values. (For example, if there is a sudden increase in flow rate at a certain time step t, the corresponding score will be significantly higher than other time steps. After softmax, the weight is close to 1, thus strengthening the features of this time step in V (such as the delay mutation corresponding to high flow rate)).

[0048] Independently execute the above steps for multiple attention heads. In this embodiment, four are used to obtain Concatenate the multi - head outputs according to the feature dimension: , where, For the output projection matrix, the output after splicing is consistent with the input dimension.

[0049] In summary, dilated convolution is used to ensure that the output of each time step only depends on the inputs of the past and current moments, avoiding the leakage of future information, which conforms to the causal logic of time series prediction. When calculating the attention scores, the influence of future time steps is masked to ensure that the model makes predictions only based on historical data.

[0050] Optionally, as an embodiment of the present invention, the spatial dimension prediction model algorithm is implemented as follows: class TopoGAT(nn.Module): def forward(self, h, adj): # h: node feature matrix [N, F] # adj: adjacency matrix [N, N] attention = torch.matmul(h, h.T) * adj return torch.softmax(attention, dim=1) The predicted spatial results of the mutual influence strength of each node in the output prediction network topology specifically include: Construct a node feature matrix h with the shape [number of nodes, feature dimension], and the matrix stores the QoS indicators of each node in the network; construct the adjacency matrix of each node; Calculate the feature similarity between all nodes based on the dot product operation in combination with the node feature matrix: h·h T , to obtain an N×N similarity matrix, where similarity i,j represents the dot product of the feature vectors of node i and node j.

[0051] Calculate the product of the feature similarity of each node and the adjacency matrix (only keep the similarity of topologically connected nodes, and set the similarity of disconnected nodes to 0), and convert the product through softmax to the attention degree of each connected node to different adjacent nodes; Each node aggregates the features of adjacent nodes by attention degree to obtain new features. For example, there are nodes 1, 2, and 3. Node 1 is connected to nodes 2 and 3, node 2 is connected to node 1, and node 3 is connected to node 1. The new feature of node 1 is the weighted average of the features of its neighbor nodes 2 and 3 according to the attention weights. If node 2 becomes congested (the delay increases and the packet loss rate increases), its node feature matrix will change, resulting in a decrease in its similarity to the adjacent node 1. The attention degree of node 1 to node 2 will decrease, and at the same time, the attention degree of node 1 to node 3 may increase, that is, node 1 pays more attention to the uncongested node 3. When node 1 needs to switch paths, based on the updated attention weights, it preferentially selects a path connected to a low-congestion node (such as node 3).

[0052] After performing attention calculation and multi-head attention fusion on the new features based on the attention mechanism of the GAT layer (using the above calculation method, the difference is only that the node feature matrix is transformed into a feature matrix in a new space and then subsequent calculations are performed), the weight matrix between nodes is output as the prediction space result of the mutual influence intensity of each node in the predicted network topology.

[0053] Optionally, as an embodiment of the present invention, the dynamic dimension prediction model algorithm is implemented as follows: def dynamic_update(new_data): # Sliding window cache buffer.append(new_data) if len(buffer)>= BATCH_SIZE: # Incremental training loss = model.train_on_batch(buffer) # Dynamically adjust the learning rate adjust_lr(based_on=loss) The prediction dynamic result for updating the model parameters specifically includes: Store the latest QoS metrics based on a fixed-size array. When the array is full of data and the latest data arrives, the earliest data is overwritten; (Based on the above strategy for data collection, the flow rate change rate between the current sampling point and the previous sampling point is calculated in real time. When a flow rate mutation is detected, the sampling period is switched and the sliding window is reset.) When the sliding window is full of data, start batch training. Obtain the prediction result through the forward propagation of the model based on the current batch of data, and calculate the loss function of this batch of data; Based on the loss function, calculate the gradient of the loss with respect to all parameters through the backpropagation algorithm; Accumulate the gradients of the current batch into an accumulated gradient variable. When the accumulated number of batches reaches a preset value, update the model's parameters in combination with the learning rate as the prediction dynamic result: θ←θ-α·▽ θ L(θ;batch_data) where α is the learning rate and L is the loss function.

[0054] Calculate the ratio of the current batch loss to the average loss of the previous N batches, and adjust the learning rate according to the ratio.

[0055] Optionally, as an embodiment of the present invention, obtaining the predicted link quality score by weighted integrating the time prediction result, the space prediction result, and the dynamic prediction result specifically includes: , where The final QoS prediction result is mapped to a link quality score from 0 to 100; : The output of the time dimension prediction model, representing the temporal trend (such as the delay change in the next 5 seconds).

[0056] : The output of the space dimension prediction model, quantifying the topological association impact (such as the cascading effect of adjacent node congestion on the current link).

[0057] : The output of the dynamic dimension prediction model, reflecting the ability to adapt to sudden changes (such as the immediate response when the flow rate suddenly increases).

[0058] , , : Dynamic weight coefficients, satisfying that the sum of the three coefficients is 1, and are adaptively adjusted according to the network scenario.

[0059] Optionally, as an embodiment of the present invention, when the predicted link quality score is less than a preset threshold, the specific steps for triggering the APN6 path switch include: Relevant message types are extended in the APN6-CP protocol header for path switch control, and the message types are shown in Table 1: Table 1 Message Types Type code Message name Direction Function description 0x0A PRE_SWITCH_REQ Requesting party → Target Path reservation request carrying prediction result 0x0B PATH_RESERVE_ACK Target → Requesting party Resource reservation confirmation 0x0C FAST_FAILOVER Control plane → Data plane Fast switch execution instruction When the predicted link quality score is less than the preset threshold, the requesting party sends a PRE_SWITCH_REQ message to the target node, and this message carries the link quality prediction result and a reservation request for the target path resources; After the target node receives the PRE_SWITCH_REQ message, if its own resources meet the reservation requirements of the requester, it returns a PATH_RESERVE_ACK message to confirm the reserved resources; when the requester receives the PATH_RESERVE_ACK message and confirms the successful resource reservation, it establishes a low-latency backup path based on the target node of the reserved resources; the control plane sends a FAST_FAILOVER fast-switching execution instruction to the data plane, instructing the data plane to preferentially forward data to the pre-established low-latency backup path based on the identifier, while maintaining the session state synchronization of the original path; After the target receiving node receives the PRE_SWITCH_REQ message, if the resources are insufficient, it returns a PATH_RESERVE_NACK message to inform the requester to switch to the backup path.

[0060] Optionally, as an embodiment of the present invention, the switching of the backup path specifically includes: After the requester receives the PATH_RESERVE_NACK message, it starts to query the local path table for the backup path; (in some network architectures, the requester device will query the built-in path table to find a backup path that meets requirements such as service bandwidth and latency. If the local path table information is insufficient, it may send a path query request to a network controller (such as an SDN controller), and the controller calculates and returns a suitable backup path according to the network global topology and resource information.) After obtaining the backup path information, the requester updates the forwarding rules of the data plane; taking the SRv6 (Segment Routing over IPv6) network as an example, the requester will modify the destination address or routing header information of the data packet to the SID (Segment Identifier) list of the backup path, and these SIDs represent a segment of the path in the network, so as to guide the data packet to be transmitted along the backup path.

[0061] Migrate the session state of the original path to the backup path to complete the switching operation. For example, in a TCP connection, the state information such as the connection sequence number and window size needs to be migrated. In some network architectures, through a specific state synchronization mechanism, such as sharing session state information among network nodes, fast switching can be achieved without interrupting the service. And local testing is performed to confirm that the data can be normally transmitted on the backup path. After that, the switching result is notified to relevant network nodes, including the source device, the destination device, and the intermediate nodes on the path, so that these nodes can update their own states to cooperate with the data transmission on the new path.

[0062] This method focuses on optimizing the dynamic switching of the APN6 network and constructs a comprehensive and accurate guarantee system. On the one hand, it establishes a multi-dimensional prediction mechanism. By using a temporal convolutional network combined with temporal attention, a graph neural network, and sliding window incremental learning, it captures QoS metrics from temporal trends, spatial associations, and dynamic changes respectively. At the same time, it finely collects key data such as microsecond-level latency, and outputs a link quality score through weighted fusion, providing a reliable basis for handover decisions.

[0063] On the other hand, it realizes proactive intelligent scheduling. When the score is lower than the threshold, it triggers the APN6 path switch, turning passive response into proactive avoidance, effectively reducing service interruptions. Moreover, the algorithm architecture is flexible, and incremental learning reduces overhead. It has both standardization and forward-lookingness, and can adapt to different network environments, providing an efficient solution for traffic scheduling in future networks such as 6G.

[0064] In some embodiments, the APN6 network dynamic switching optimization system based on QoS prediction may include multiple functional modules composed of computer program segments. The computer programs of each program segment in the APN6 network dynamic switching optimization system based on QoS prediction can be stored in the memory of a computer device and executed by at least one processor to execute (see details Figure 1 described) the functions of the APN6 network dynamic switching optimization based on QoS prediction.

[0065] In this embodiment, the APN6 network dynamic switching optimization system based on QoS prediction can be divided into multiple functional modules according to the functions it performs, such as Figure 3 shown. The functional modules of the system may include: a data acquisition module, a model construction module, and a path switching module. The module referred to in the present invention means a series of computer program segments that can be executed by at least one processor and can complete fixed functions, and are stored in the memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments. The system includes: A data acquisition module that collects QoS metrics of different nodes of the link in the APN6 network based on a preset sampling strategy. The QoS metrics include microsecond-level one-way latency, jitter based on the calculation result of IETF RFC 3393, sliding window packet loss rate, current flow rate, and the DSCP field taken from the APN6 header; The model construction module constructs a time - dimension prediction model based on a temporal convolutional network combined with a temporal attention mechanism. The model takes QoS metrics as input and outputs a predicted time result for predicting the QoS trend in the short future. It constructs a spatial - dimension prediction model based on a graph neural network. The model takes QoS metrics as input and outputs a predicted spatial result for predicting the mutual influence strength of each node in the network topology. It constructs a dynamic - dimension prediction model based on sliding - window incremental learning. The model takes QoS metrics as input and outputs a predicted dynamic result for updating the model parameters. The path switching module comprehensively weights the time prediction result, spatial prediction result, and dynamic prediction result to obtain a predicted link quality score. When the predicted link quality score is less than a preset threshold, it triggers the APN6 path switching.

[0066] Figure 4 FIG. 6 is a schematic structural diagram of a terminal provided by an embodiment of the present invention. The terminal can be used to execute the method for dynamic switching optimization of an APN6 network based on QoS prediction provided by the embodiment of the present invention.

[0067] Among them, the terminal may include: a processor, a memory, and a communication unit. These components communicate through one or more buses. Those skilled in the art can understand that the structure of the server shown in the figure does not constitute a limitation to the present invention. It can be a bus - shaped structure, a star - shaped structure, and may also include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0068] Among them, the memory can be used to store the execution instructions of the processor. The memory can be implemented by any type of volatile or non - volatile storage terminal or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read - only memory (EEPROM), erasable programmable read - only memory (EPROM), programmable read - only memory (PROM), read - only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory are executed by the processor, the terminal can execute some or all of the steps in the above - mentioned method embodiments.

[0069] The processor is the control center of the storage terminal, connecting various parts of the entire electronic terminal through various interfaces and circuits. By running or executing software programs and / or modules stored in the memory, and invoking data stored in the memory, it performs various functions of the electronic terminal and / or processes data. The processor may be composed of an integrated circuit (IC for short), for example, it may be composed of a single packaged IC, or it may be composed of multiple packaged ICs with the same or different functions connected together. For example, the processor may only include a central processing unit (CPU for short). In the embodiments of the present invention, the CPU may be a single arithmetic core or may include multiple arithmetic cores.

[0070] A communication unit, used to establish a communication channel so that the storage terminal can communicate with other terminals. It receives user data sent by other terminals or sends user data to other terminals.

[0071] The present invention also provides a computer storage medium. Among them, the computer storage medium can store a program, and when the program is executed, it may include some or all of the steps in the embodiments provided by the present invention. The storage medium may be a magnetic disk, an optical disc, a read-only memory (ROM for short), a random access memory (RAM for short), etc.

[0072] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be realized by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solutions in the embodiments of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc, etc., which can store program codes, and includes several instructions to enable a computer terminal (which may be a personal computer, a server, or a second terminal, a network terminal, etc.) to execute all or part of the steps of the methods described in the embodiments of the present invention.

[0073] For the same and similar parts among the various embodiments in this specification, reference can be made to each other. In particular, for the terminal embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the descriptions in the method embodiments.

[0074] In several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the system or module can be in electrical, mechanical or other forms.

[0075] The modules described as separate components may or may not be physically separated. The components displayed as modules may or may not be physical modules, that is, they can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0076] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0077] Although the present invention has been described in detail by referring to the drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, those of ordinary skill in the art can make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions should be within the scope of the present invention. / Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered by the protection scope of the present invention.

Claims

1. A dynamic handover optimization method for APN6 network based on QoS prediction, characterized in that Including: Collecting QoS metrics of different nodes in the APN6 network based on a preset sampling strategy. The QoS metrics include microsecond-level one-way delay, jitter based on the calculation result of IETF RFC 3393, sliding window packet loss rate, current flow rate, and the DSCP field taken from the APN6 header; Constructing a time dimension prediction model based on a temporal convolutional network combined with a temporal attention mechanism. The model takes QoS metrics as input and outputs a prediction time result for predicting the QoS trend in the short term in the future; Constructing a spatial dimension prediction model based on a graph neural network. The model takes QoS metrics as input and outputs a prediction spatial result for predicting the mutual influence strength of each node in the network topology; Constructing a dynamic dimension prediction model based on sliding window incremental learning. The model takes QoS metrics as input and outputs a prediction dynamic result for updating the model parameters; Weightedly integrating the time prediction result, spatial prediction result, and dynamic prediction result to obtain a predicted link quality score. When the predicted link quality score is less than a preset threshold, trigger the APN6 path switching; 2. The APN6 network dynamic handover optimization method based on QoS prediction according to claim 1, characterized in that The sampling strategy includes a basic sampling period and burst traffic triggering; The basic sampling period is a preset value; Burst traffic triggering is to sample once every 1ms when the flow rate change rate is greater than 15%; 3. The APN6 network dynamic handover optimization method based on QoS prediction according to claim 1, wherein The time dimension prediction model includes a dilated convolutional layer and a multi-head attention layer. The output of the prediction time result for predicting the QoS trend in the short term in the future specifically includes: Preprocessing the QoS metrics to obtain a time series data matrix, which includes batch size, time step, and feature dimension; Setting different dilation rates, dilating the convolutional layer by inserting holes between the convolutional kernel elements based on the dilation rate, and extracting time series features at different time scales to obtain a sequence integrating multi-scale features; Using the multi-head attention mechanism to project the sequence integrating multi-scale features onto multiple attention heads, calculating the attention scores of each attention head respectively, and then splicing them according to the feature dimension to obtain the feature with temporal weights as the prediction time result for predicting the QoS trend in the short term in the future; 4. The APN6 network dynamic handover optimization method based on QoS prediction according to claim 1, wherein The output of the prediction spatial result for predicting the mutual influence strength of each node in the network topology specifically includes: Constructing a node feature matrix with the shape of [number of nodes, feature dimension], where the matrix stores the QoS metrics of each node in the network; constructing the adjacency matrix of each node; Calculating the feature similarity between all nodes based on the dot product operation combined with the node feature matrix; Calculating the product of the feature similarity of each node and the adjacency matrix, and converting the product through softmax to obtain the attention degree of each node to different adjacent nodes; Each node aggregates the features of adjacent nodes weighted by the attention degree to obtain new features; Based on the attention mechanism of the GAT layer, performing attention calculation and multi-head attention fusion on the new features, and then outputting the node-to-node weight matrix as the prediction spatial result for predicting the mutual influence strength of each node in the network topology; 5. The APN6 network dynamic handover optimization method based on QoS prediction according to claim 1, wherein, The output of the prediction dynamic result for updating the model parameters specifically includes: Storing the latest QoS metrics based on a fixed-size array. When the array is filled with data and the latest data arrives, the earliest data is overwritten; When the sliding window is filled with data, the prediction result is obtained through the forward propagation of the model based on the current batch of data, and the loss function of this batch of data is calculated; based on the loss function, the gradients of the loss with respect to all parameters are calculated through the backpropagation algorithm; The gradients of the current batch are accumulated into an accumulated gradient variable. When the accumulated number of batches reaches a preset value, the parameters of the model are updated in combination with the learning rate as the prediction dynamic result; Calculate the ratio of the current batch loss to the average loss of the previous N batches, and adjust the learning rate according to the ratio.

6. The APN6 network dynamic handover optimization method based on QoS prediction according to claim 1, characterized in that When the predicted link quality score is less than the preset threshold, the specific steps to trigger the APN6 path switch include: Related message types are extended in the APN6-CP protocol header for path switch control, and the message types include: PRE_SWITCH_REQ, PATH_RESERVE_ACK, and FAST_FAILOVER; When the predicted link quality score is less than the preset threshold, the requester sends a PRE_SWITCH_REQ message to the target node, and this message carries the link quality prediction result and the reservation request for the target path resources; After receiving the PRE_SWITCH_REQ message, if the target node's own resources meet the reservation requirements of the requester, it returns a PATH_RESERVE_ACK message to confirm the reserved resources; when the requester receives the PATH_RESERVE_ACK message and confirms that the resource reservation is successful, a low-latency backup path is established based on the target node of the reserved resources; the control plane sends a FAST_FAILOVER fast switch execution instruction to the data plane, instructing the data plane to preferentially forward the data to the pre-established low-latency backup path based on the identifier, while keeping the session state of the original path synchronized; After receiving the PRE_SWITCH_REQ message, if the target receiving node does not have enough resources, it returns a PATH_RESERVE_NACK message to inform the requester to switch to the backup path.

7. The APN6 network dynamic handover optimization method based on QoS prediction according to claim 6, wherein, The specific switch of the backup path includes: After receiving the PATH_RESERVE_NACK message, the requester starts to query the local path table for the backup path; After obtaining the backup path information, the requester updates the forwarding rules of the data plane; Migrate the session state of the original path to the backup path to complete the switch operation, and perform local testing to confirm that the data can be normally transmitted on the backup path.

8. A dynamic handover optimization system for APN6 network based on QoS prediction, characterized in that, When the system is implemented, it executes the APN6 network dynamic switch optimization method based on QoS prediction described in any one of claims 1-7. The system includes: A data collection module that collects QoS metrics of different nodes on the links in the APN6 network based on a preset sampling strategy. The QoS metrics include microsecond-level one-way delay, jitter based on the calculation result of IETF RFC 3393, sliding window packet loss rate, current flow rate, and the DSCP field taken from the APN6 header; The model construction module constructs a time - dimension prediction model based on a temporal convolutional network combined with a temporal attention mechanism. The model takes QoS metrics as input and outputs a predicted time result for predicting the QoS trend in the short - term future. It constructs a spatial - dimension prediction model based on a graph neural network. The model takes QoS metrics as input and outputs a predicted spatial result for predicting the mutual influence strength of each node in the network topology. It constructs a dynamic - dimension prediction model based on sliding - window incremental learning. The model takes QoS metrics as input and outputs a predicted dynamic result for updating the model parameters. The path switching module weighted - synthesizes the time prediction result, the spatial prediction result, and the dynamic prediction result to obtain a predicted link quality score. When the predicted link quality score is less than a preset threshold, it triggers the APN6 path switching.

9. A terminal, characterized in that, It includes: A memory for storing an APN6 network dynamic switching optimization program based on QoS prediction; A processor for implementing the steps of the APN6 network dynamic switching optimization method based on QoS prediction as described in any one of claims 1 - 7 when executing the APN6 network dynamic switching optimization program based on QoS prediction.

10. A computer-readable storage medium, characterized in that, An APN6 network dynamic switching optimization program based on QoS prediction is stored on the readable storage medium. When the APN6 network dynamic switching optimization program based on QoS prediction is executed by the processor, it implements the steps of the APN6 network dynamic switching optimization method as described in any one of claims 1 - 7.

Images