Dual-time-scale asymmetric multi-link channel binding method

Abstract

The invention provides a dual-time-scale asymmetric multi-link channel binding method, which comprises the following steps of: initializing millisecond-level short-term and minute-level long-term dual time scales, and acquiring a channel state; integrating the network state, the user demand and the service flow in real time to generate a feature vector; periodically predicting resource requirements and dividing service priorities based on a long-term timer, and triggering dynamic channel binding adjustment when the sudden change of a network load exceeds a threshold value; high / low frequency band resources are distributed in a differentiated manner by adopting an asymmetric strategy, and key services are guaranteed preferentially; a cross-link channel combination is dynamically optimized in combination with periodic flow and instantaneous interference; and continuously iterating the strategy by monitoring the performance index. According to the method, the limitation of traditional static symmetric binding and single time scale decision making is broken through, a dual-time scale cooperation mechanism and event-driven response are utilized, dynamic adaptation of channel resources is achieved, the network throughput is effectively improved, time delay is reduced, and the problems of dynamic service demand mismatching and cross-link difference perception deficiency are solved.

Description

Technical Field

[0001] The present invention relates to the field of mobile communication technologies, and specifically relates to a dual-time-scale asymmetric multi-link channel bonding method, which is applicable to dynamic spectrum sharing scenarios in unlicensed frequency bands such as WAPI, WiFi, LTE-U, and NR-U. Background Art

[0002] With the rapid development of wireless communication technologies, the number of wireless devices has increased explosively, and users' requirements for network bandwidth and communication quality have also risen steadily. New-generation wireless communication standards such as WiFi 6E have emerged, providing richer wireless channel spectrum resources for alleviating network capacity pressure and improving network performance. However, in multi-link scenarios, how to achieve efficient dynamic allocation of channel resources still faces severe challenges. The current mainstream symmetric channel bonding technologies have three core defects:

[0003] 1. Mismatch between static symmetric bonding and dynamic service requirements

[0004] Traditional channel bonding methods usually adopt a symmetric bundling (such as 160MHz + 160MHz) mode, that is, multiple channels with the same bandwidth are mechanically bundled for use. Although this method broadens the network bandwidth to a certain extent, it ignores the asymmetric characteristics of actual traffic flows. For example, in typical application scenarios, the downlink / uplink traffic ratio of VR services is as high as 10:1, while industrial Internet of Things devices exhibit periodic uplink burst characteristics. This "one-size-fits-all" allocation mode results in an uplink channel utilization rate of less than 35% (measured data), while high-priority services face downlink bandwidth bottlenecks. In addition, different services have significantly different requirements for performance indicators such as bandwidth, latency, and reliability, and coupled with the dynamic changes of the network environment over time and space, it is difficult for symmetric channel bonding methods to meet diverse service requirements.

[0005] 2. Adaptability limitations of single-time-scale decision-making

[0006] Existing solutions mostly make decisions based on a single time scale and cannot take into account both short-term fluctuations and long-term trends of network states. Short-term strategies (millisecond level) can quickly respond to bursty traffic, but it is difficult to capture periodic service patterns, resulting in frequent channel reconfiguration overhead; long-term strategies (hour level) can optimize resource allocation, but they cannot cope with bursty traffic fluctuations. For example, in an intelligent factory scenario, the 10ms cycle control instructions of robotic arms and the timed upload of production reports form a mixed traffic flow, and static allocation on a single time scale will result in a delay jitter of more than 30%. This limitation makes it impossible to guarantee short-term real-time performance and long-term stability in channel resource allocation.

[0007] 3. Lack of cross-link difference awareness

[0008] In an asymmetric multi-link environment, significant characteristic differentiations are presented among channels in different frequency bands. For example, although the 6GHz frequency band has a large bandwidth of 320MHz, its penetration ability is weak; the 5GHz frequency band has stable coverage but is vulnerable to interference. The existing technologies fail to establish a dynamic mapping mechanism between channel characteristics and service requirements, resulting in key services possibly being wrongly allocated to links with high packet loss rates. For example, in a medical AR scenario, if the transmission of key images is allocated to a link with a high packet loss rate, it will seriously affect the service quality and user experience.

[0009] It can be seen that some channel binding methods based on static policies in the existing technologies cannot adapt to the dynamic changes of network states and user requirements in real time. At the same time, due to the lack of a prediction and dynamic adjustment mechanism for service requirements, the network performance will drop significantly under high network loads. Therefore, there is an urgent need for a new type of channel binding method that can dynamically adjust channel allocation on different time scales, thereby optimizing the network performance. Summary of the Invention

[0010] The objective of the present invention is to provide a dual-time-scale asymmetric multi-link channel binding method. The present invention can solve problems existing in the prior art, such as the mismatch between static symmetric binding and dynamic service requirements, the adaptability limitations of single-time-scale decision-making, and the lack of cross-link difference perception, and achieve efficient dynamic allocation of channel resources and optimize network performance.

[0011] The technical solution of the present invention: A dual-time-scale asymmetric multi-link channel binding method includes the following steps:

[0012] Step 1, system initialization settings: Define the short-term time scale as a millisecond-level response cycle, the long-term time scale as a minute-level decision cycle, and collect the initial channel state matrix;

[0013] Step 2, real-time collection of multi-dimensional data, including network state, user requirements, and service traffic, to generate a standardized feature vector;

[0014] Step 3, based on a long-term timer and an event-driven mutation threshold trigger strategy adjustment: When the long-term timer reaches the preset cycle, execute Step 4; when the network load mutation amount exceeds the event-driven mutation threshold, execute Step 5;

[0015] Step 4, long-term decision-making stage: Use the standardized feature vector to predict future resource requirements, and determine service priorities according to the prediction results;

[0016] Step 5, real-time scheduling stage: Dynamically adjust the number of channel bindings and the scheduling cycle for burst traffic or delay-sensitive services;

[0017] Step 6, dynamically allocate high-frequency band or low-frequency band channels, formulate an asymmetric channel binding strategy, and ensure that high-priority services obtain sufficient resources;

[0018] Step 7: Cross-link resource optimization, dynamically switch the channel combination by combining periodic traffic peaks and instantaneous interference;

[0019] Step 8: Issue channel binding instructions and monitor the network status, record key performance indicators;

[0020] Step 9: Evaluate the network performance. If it does not meet the expectation, return to Step 4 to readjust the strategy.

[0021] In the above dual-time-scale asymmetric multi-link channel binding method, in Step 1, the channel state matrix is represented as follows:

[0022]

[0023] where c ij represents the j-th state index of the i-th channel;

[0024] The channel state matrix includes the bandwidth occupancy rate, signal strength, and packet loss rate of each channel, and is stored as a structured data table through a state database.

[0025] In the aforementioned dual-time-scale asymmetric multi-link channel binding method, the network state includes the channel interference index I and the device load rate L, the user requirements include the service type QoS label Q and the delay sensitivity D, and the service traffic includes the burst traffic peak P and the periodic traffic pattern F;

[0026] The standardized eigenvector is represented as follows:

[0027] V = [I, L, Q, D, P, F].

[0028] In the aforementioned dual-time-scale asymmetric multi-link channel binding method, when the long-term timer reaches the preset long-term decision period, use the periodic traffic pattern F in the standardized eigenvector to predict the future load and trigger long-term network prediction and global decision-making;

[0029] If it is detected that the network load mutation amount ΔL exceeds the event-driven mutation threshold, immediately interrupt the current period and start to trigger short-term real-time scheduling and dynamic challenge decision-making; where the network load mutation amount is expressed as:

[0030] ΔL = L current - L previous ;

[0031] In the formula: L current and L previous are the network load amounts at the current moment and the historical average respectively.

[0032] In the above-mentioned dual-time-scale asymmetric multi-link channel binding method, in step 4, the periodic traffic patterns F and QoS labels Q of the standardized eigenvectors are used to predict the future resource requirements of services.

[0033] The calculation formula for the service priority is as follows:

[0034]

[0035] In the formula: Priority represents the service priority, Q is the QoS label of the service type, D is the delay sensitivity, I is the channel interference index, and α, β, and γ are the weight coefficients respectively.

[0036] In the above-mentioned dual-time-scale asymmetric multi-link channel binding method, in step 5, based on real-time data analysis, it is judged whether the current network state meets the burst service requirements: (1) If the burst traffic peak is greater than the threshold, the number of channel bindings is increased to ensure that the system can quickly respond to burst traffic; (2) For delay-sensitive services, to ensure the delay requirements, the scheduling period D is shortened.

[0037] In the above-mentioned dual-time-scale asymmetric multi-link channel binding method, in step 6, the rules of the asymmetric channel binding strategy are as follows:

[0038] High-priority services are bound to high-frequency low-interference channels;

[0039] Large-traffic services are bound to low-frequency wide-coverage channels;

[0040] Delay-sensitive services are preferentially allocated low-delay channels.

[0041] In the above-mentioned dual-time-scale asymmetric multi-link channel binding method, in step 7, the cross-link resource optimization includes:

[0042] Allocate low-frequency channels during periodic traffic peaks;

[0043] Switch to high-frequency channels when instantaneous interference is detected;

[0044] Dynamically aggregate idle channel resources to meet burst demands.

[0045] In the above-mentioned dual-time-scale asymmetric multi-link channel binding method, in step 8, the key performance indicators include average delay, packet loss rate, channel utilization rate, and burst response time, and the long-term policy parameters are adjusted through a feedback mechanism.

[0046] In the above-mentioned dual-time-scale asymmetric multi-link channel binding method, in step 2, a sliding window mechanism is used to perform timestamp alignment and normalization processing on the collected raw data.

[0047] In the aforementioned dual-time-scale asymmetric multi-link channel bonding method, in the long-term decision phase of step 4, when it is predicted that the traffic of a specific service will continue to grow and exceed a certain threshold within a certain time period in the future, some channel resources are reserved in advance to ensure the service quality of the service.

[0048] In the aforementioned dual-time-scale asymmetric multi-link channel binding method, during the cross-link resource optimization process in step 7, when it is detected that the utilization rate of a certain frequency band channel is continuously lower than the set utilization threshold, the idle channel resources of the frequency band are temporarily allocated to other services with urgent needs, and the original channel resource allocation is restored after the service demand ends.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] 1. The present invention uses asymmetric channel bonding to flexibly allocate channel resources based on the needs of different services and changes in network status. Compared with traditional symmetric channel bonding methods, the present invention avoids the waste of channel resources, especially for business scenarios with asymmetric uplink and downlink traffic (such as VR, industrial Internet of Things, etc.), significantly improving channel utilization and solving the problem of insufficient uplink channel utilization in traditional methods.

[0051] 2. This invention introduces a dual-timescale decision-making mechanism that can quickly respond to short-term fluctuations in network status (such as traffic bursts) while adapting to long-term changes in network usage patterns (such as periodic services). This mechanism effectively reduces the overhead caused by frequent channel reorganization, while also reducing latency jitter, improving network stability and reliability, and reducing service delays and packet loss.

[0052] 3. This invention fully considers the differentiated requirements of different services for performance indicators such as bandwidth, latency, and reliability, and provides users with higher-quality network services by dynamically adjusting channel allocation. For example, in medical AR scenarios, critical image transmission can be prioritized to links with low packet loss rates, avoiding the service quality degradation caused by the lack of cross-link differential perception in traditional methods, significantly improving user satisfaction.

[0053] 4. The present invention can effectively cope with the characteristic differentiation of channels in different frequency bands in multi-link scenarios (such as the large bandwidth but weak penetration of the 6GHz band, and the stable coverage but susceptibility to interference of the 5GHz band). By establishing a dynamic mapping mechanism between channel characteristics and business needs, it ensures that key businesses are allocated to the most suitable links, further optimizing network performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic flow diagram of the present invention; DETAILED DESCRIPTION

[0055] The present invention will be further described below in conjunction with the accompanying drawings and embodiments, but it shall not be used as a basis for limiting the present invention.

[0056] Embodiment 1: A dual-time-scale asymmetric multi-link channel bonding method, as Figure 1 shown, includes the following steps:

[0057] Step 1: System initialization settings, defining the short-term time scale T short as a millisecond-level response period (e.g., 10 - 50 ms), and the long-term time scale T long as a minute-level decision period (e.g., 1 - 5 min), and collecting the initial channel state matrix; using a network probe to collect the initial channel state matrix C, including the bandwidth occupancy rate, signal strength (such as RSSI ≥ -70 dBm), and historical packet loss rate (such as < 5%) of each frequency band channel, and storing the data in the state database. This step provides benchmark parameters and the initial network topology for subsequent multi-dimensional data collection. The channel state matrix C is expressed as follows:

[0058]

[0059] where, c ij represents the j-th state index of the i-th channel;

[0060] Store the initial channel state matrix in the state database, and the data structure is shown in Table 1 below:

[0061] Channel ID Bandwidth Occupancy Rate (%) Signal Strength (dBm) Packet Loss Rate (%) 1 30 -65 2 2 45 -70 4 … … … …

[0062] Table 1

[0063] Step 2: Real-time collection of multi-dimensional data, including network status, user requirements, and traffic, to generate a standardized feature vector; this step uses distributed sensors to capture three main types of data in real time: network status (channel interference index I, device load rate L), user requirements (service type QoS label Q, delay sensitivity D), and traffic (burst traffic peak P, periodic traffic pattern F). Real-time monitoring of multi-dimensional data ensures that the system can comprehensively understand the current network environment and user behavior.

[0064] Among them, the channel interference index I is used to monitor the interference intensity of each channel in real time (for example, when the signal-to-noise ratio SNR < -70 dBm, it is defined as high interference); the device load rate L (%) is used to calculate the current usage rate of the device's CPU, memory, or bandwidth. For example, L = (current bandwidth usage / total bandwidth) × 100%; the QoS label Q is the encoding of the user's service type (such as Q = 1 for voice calls, Q = 2 for video streaming media, Q = 3 for IoT sensor data); the delay sensitivity D (ms) is the maximum delay threshold allowed for the service (such as D = 20 ms for ultra-low delay, D = 100 ms for normal delay); the peak burst traffic P (Mbps) represents the magnitude of traffic surge within a statistical unit time window. For example, P = max(traffic[t], traffic[t - Δt]); the periodic traffic pattern F: extracts the periodic characteristics of traffic through Fourier transform (such as F = 0.5 Hz indicating a peak every 2 seconds). The generated normalized feature vector is denoted as V = [I, L, Q, D, P, F]. The sliding window mechanism is used to align the timestamps and normalize the original data. The window size is W,

[0065]

[0066] where x is the original data, x min and x max are the minimum and maximum values within the window respectively.

[0067] Step 3: Policy cycle trigger determination: Based on the timer T long and the event-driven mutation threshold δ th a dual-mode start policy adjustment is performed: If the long-term timer T long reaches the preset long-term decision cycle, the periodic traffic pattern F in the historical data is used to predict the future load, triggering long-term network prediction and global decision-making, and entering Step 4; if it is detected that the network load mutation amount ΔL exceeds the threshold δ th [[ID=2)]], then the current cycle is immediately interrupted to start triggering short-term real-time scheduling and dynamic challenge decision-making, and enter Step 5; among them, the network load mutation amount is expressed as:

[0068] ΔL = L current -L previous

[0069] where L current and are the network load amounts at the current moment and the historical average respectively.

[0070] Step 4, Long-term Decision-making Phase: Use the periodic traffic pattern F and QoS label Q in the eigenvector V to predict the future resource requirements of the service. For example, if the periodic traffic pattern F shows that the traffic peak is from 18:00 to 20:00 every day and Q shows that the proportion of video services increases, then allocate high-bandwidth channels in advance. According to the prediction results, determine the service priority, formulate a long-term channel binding strategy for the service, ensure the reasonable allocation of resources, optimize network performance, and meet diverse service requirements. The calculation formula for service priority Priority comprehensively considers the QoS label Q, delay sensitivity D, and channel interference index I, and is specifically expressed as follows:

[0071]

[0072] Among them, the coefficients α, β, and γ can be adjusted according to service requirements.

[0073] Channel Allocation Principle: The higher the service priority, the more resources the system tends to allocate to this service. For example:

[0074] High QoS service (Q = 1):

[0075] Bind high-frequency band channels (such as 5 GHz) to reduce interference (low I value).

[0076] Ensure that high-priority services obtain high-quality channel resources.

[0077] Large traffic service (Q = 2):

[0078] Bind low-frequency band channels (such as 2.4 GHz) to improve the coverage range (stable F value).

[0079] Ensure stable transmission of large traffic services within the coverage range.

[0080] In addition, when it is predicted that the traffic of a specific service will continue to grow and exceed a certain threshold within a certain period in the future, in addition to adjusting the channel allocation, a certain amount of channel resources are reserved in advance to ensure the quality of service of this service.

[0081] Step 5, Real-time Scheduling Phase: Dynamically adjust the number of channel bindings and the scheduling period for burst traffic or delay-sensitive services; in this step, based on real-time data analysis, determine whether the current network state meets the requirements of burst services: (1) If the peak value of burst traffic is greater than the threshold (P > P threshold ), then increase the number of channel bindings to ensure that the system can quickly respond to burst traffic; (2) For delay-sensitive services, to ensure the delay requirements, shorten the scheduling period (delay sensitivity D), for example, update it to D / 2;

[0082] Step 6: Dynamically allocate high-frequency or low-frequency band channels, formulate an asymmetric channel binding strategy to ensure that high-priority services obtain sufficient resources. In this step, according to the QoS label Q and the device load rate I, dynamically allocate high-frequency or low-frequency band channels, formulate an asymmetric channel binding strategy to ensure that high-priority services obtain sufficient resources. Send the channel allocation instruction to the network device to implement channel binding, ensuring that different traffic flows can obtain corresponding bandwidth and resources according to their needs. Among them, the rules of the asymmetric channel binding strategy are: high-priority services are bound to high-frequency band low-interference channels; large-traffic services are bound to low-frequency band wide-coverage channels; delay-sensitive services are preferentially allocated low-delay channels.

[0083] Step 7: Cross-link resource optimization, combining periodic traffic peaks and instantaneous interference, dynamically switch the channel combination. In this step, cross-link resource optimization includes: allocating low-frequency band channels during periodic traffic peaks; switching to high-frequency band channels when instantaneous interference is detected; dynamically aggregating idle channel resources to meet bursty demands. Ensure that the characteristics of different frequency band channels can be fully utilized to improve the overall network performance. Dynamically adjust resource allocation according to the real-time network state to ensure the efficient use of resources and avoid resource waste or shortage. In addition, when it is detected that the utilization rate of a certain frequency band channel continues to be lower than the set utilization rate threshold, temporarily allocate the idle channel resources of this frequency band to other services with urgent needs, and restore the original channel resource allocation after the demand of this service ends.

[0084] Step 8: Instruction execution and status monitoring: Send channel binding and resource allocation instructions to the network device to ensure the timely execution and effective implementation of the instructions. Real-time monitor the channel status and service performance, record the key performance indicators (KPIs) to ensure that the system operates in the best state, and promptly discover and solve potential problems. The key performance indicators include average delay, packet loss rate, channel utilization rate, and burst response time, and adjust the long-term policy parameters through a feedback mechanism.

[0085] Step 9: Performance evaluation and feedback: Evaluate the current network performance, compare the actual performance with the expected performance, analyze the reasons for the deviation to ensure the network service quality. If the actual performance meets the expectations, maintain the current policy; if not, return to the long-term decision-making stage, re-evaluate and adjust the policy to optimize the network performance.

[0086] End: Complete the whole process of channel binding and resource allocation, the system enters the monitoring state, and prepares for a new round of periodic adjustment.

[0087] Example 2: In this example, the method of Example 1 is applied to the vehicle-road coordination scenario in a smart city. 200 roadside units (RSUs) are deployed on the main road of a certain smart city, covering a road with two-way 10 lanes and a total length of 5 kilometers. The system needs to process three types of services simultaneously: Among them, V2X communication (Q = 1): Transmit real-time position, speed, and emergency braking signals between vehicles and RSUs (delay requirement < 20 ms, packet loss rate < 0.1%); Traffic monitoring video (Q = 2): 4K / 60fps high-definition video stream (bandwidth requirement ≥ 50 Mbps / road); Environmental sensor data (Q = 3): Environmental indicators such as PM2.5, temperature, and humidity (periodic upload, period 1 s). The parameter settings are shown in Table 2 below.

[0088]

[0089] Table 2

[0090] Table 3 shows the corresponding simulation verification results:

[0091] Indicator Traditional Symmetric Binding Method of the Present Invention Improvement Range Average V2X Delay (ms) 25.4 18.2 ↓28.3％ Video Packet Loss Rate (%) 2.3 0.7 ↓69.6％ Channel Utilization Rate (%) 68.7 82.4 ↑19.9％ Burst Traffic Response Time (ms) 45 22 ↓51.1％

[0092] Table 3

[0093] Analysis of simulation results:

[0094] Improvement in delay performance: In the high-density scenario of 200 vehicles / km during the morning rush hour, the V2X communication delay is reduced from 18.7 ms to 9.2 ms. This benefits from the dual-time-scale mechanism: Long-period prediction reserves 5.9 GHz dedicated channel resources in advance, and short-period dynamic adjustment quickly responds to sudden braking instructions. 2 Enhanced anti-interference ability: The packet loss rate of the video stream decreases by 69.6%, which stems from the cross-link optimization strategy: When the SNR of the 5 GHz channel is detected to be < -75 dBm, the video service is automatically switched to the 2.4 GHz + 5 GHz hybrid binding mode. Simulation data shows that the channel switching response time during the interference period (14:00 - 16:00, the active period of the commercial area) is less than 15 ms, ensuring the smooth transmission of 4K videos.

[0095] Optimization of spectral efficiency: The key to the 17.5% increase in bandwidth utilization lies in the intelligent allocation of asymmetric binding. Among them, for V2X services: The combination of 5.9 GHz (40 MHz) + 5 GHz (80 MHz) makes full use of the low-interference characteristics of the high-frequency band; for sensor data: Single 2.4 GHz channel dynamic aggregation, and on-demand allocation to avoid resource waste.

[0096]

[0097] ​Burst traffic adaptability: In traffic accident simulations (60Mbps+ burst traffic triggered by 3 events), the system completes the following responses within 20ms. The video service bandwidth is dynamically expanded to 160MHz (5GHz+2.4GHz bonding); the sensor data period is temporarily adjusted to a 2s interval; the core service QoS remains stable during the event, verifying the effectiveness of the δ_th = 30% load mutation threshold.

[0098] In summary, the present invention can solve the problems existing in the prior art, such as the mismatch between static symmetric binding and dynamic service requirements, the limitations of single-time-scale decision-making adaptability, and the lack of cross-link difference perception, etc., realize the efficient dynamic allocation of channel resources, and optimize network performance.

Claims

1. A dual-time-scale asymmetric multi-link channel bonding method, characterized in that: It includes the following steps: Step 1: System initialization settings. Define the short-term time scale as a millisecond-level response cycle and the long-term time scale as a minute-level decision cycle, and collect the initial channel state matrix; Step 2: Real-time collection of multi-dimensional data, including network status, user requirements, and traffic volume, to generate a standardized feature vector; Step 3: Based on the long-term timer and event-driven mutation threshold trigger strategy adjustment, when the long-term timer reaches the preset cycle, execute Step 4; when the network load mutation amount exceeds the event-driven mutation threshold, execute Step 5; Step 4: Long-term decision-making stage: Use the standardized feature vector to predict future resource requirements, and determine service priorities according to the prediction results; Step 5: Real-time scheduling stage: Dynamically adjust the number of channel bindings and the scheduling cycle for burst traffic or delay-sensitive services; Step 6: Dynamically allocate high-frequency or low-frequency band channels, formulate an asymmetric channel binding strategy to ensure that high-priority services obtain sufficient resources; Step 7: Cross-link resource optimization, combining periodic traffic peaks and instantaneous interference, dynamically switch channel combinations; Step 8: Issue channel binding instructions and monitor the network status, record key performance indicators; Step 9: Evaluate the network performance. If it does not meet the expectations, return to Step 4 to readjust the strategy.

2. The dual-time-scale asymmetric multi-link channel bonding method according to claim 1, wherein: In Step 1, the channel state matrix is expressed as follows: Among them, c ij represents the j-th state index of the i-th channel; The channel state matrix includes the bandwidth occupancy rate, signal strength, and packet loss rate of each channel, and is stored as a structured data table through the state database.

3. The dual-time-scale asymmetric multi-link channel binding method according to claim 1, wherein: The network status includes the channel interference index I and the device load rate L. The user requirements include the service type QoS label Q and the delay sensitivity D. The traffic volume includes the burst traffic peak P and the periodic traffic pattern F; The standardized feature vector is expressed as follows: V = [I, L, Q, D, P, F].

4. The dual-time-scale asymmetric multi-link channel binding method according to claim 1, wherein: When the long-term timer reaches the preset long-term decision cycle, use the periodic traffic pattern F in the standardized feature vector to predict future loads, and trigger long-cycle network prediction and global decision-making; If it is detected that the network load mutation amount ΔL exceeds the event-driven mutation threshold, immediately interrupt the current cycle and trigger short-cycle real-time scheduling and dynamic challenge decision-making; among them, the network load mutation amount is expressed as: ΔL = L current -L previous ; where: L current and L previous are the network load at the current moment and the historical average respectively.

5. The dual-time-scale asymmetric multi-link channel bonding method according to claim 1, wherein: In Step 4, use the periodic traffic pattern F and the QoS label Q of the standardized feature vector to predict the future resource requirements of the service; The calculation formula for the service priority is as follows: In the formula: Priority represents the service priority, Q is the service type QoS label, D is the delay sensitivity, I is the channel interference index, and α, β, and γ are weight coefficients respectively.

6. The dual-time-scale asymmetric multi-link channel bonding method according to claim 1, wherein: In Step 5, based on real-time data analysis, determine whether the current network status meets the burst service requirements: (1) If the burst traffic peak is greater than the threshold, increase the number of channel bindings to ensure that the system can quickly respond to burst traffic; (2) For delay-sensitive services, to ensure the delay requirements, shorten the scheduling cycle D.

7. The dual-time-scale asymmetric multi-link channel binding method according to claim 1, characterized in that: In Step 6, the rules of the asymmetric channel binding strategy are: High-priority services are bound to high-frequency band and low-interference channels; Large-traffic services are bound to low-frequency band and wide-coverage channels; Delay-sensitive services are preferentially allocated low-delay channels.

8. The dual-time-scale asymmetric multi-link channel bonding method according to claim 1, characterized in that: In Step 7, the cross-link resource optimization includes: Allocate low-frequency band channels during periodic traffic peaks; Switch to high-frequency band channels when instantaneous interference is detected; Dynamically aggregate idle channel resources to meet bursty demands.

9. The dual-time-scale asymmetric multi-link channel bonding method according to claim 1, wherein: In step 8, the key performance indicators include average delay, packet loss rate, channel utilization rate, and burst response time, and the long-term policy parameters are adjusted through a feedback mechanism.

10. The dual-time-scale asymmetric multi-link channel binding method according to claim 1, characterized in that: In step 2, a sliding window mechanism is adopted to perform timestamp alignment and normalization processing on the collected raw data.

11. The dual-time-scale asymmetric multi-link channel binding method according to claim 1, characterized in that: In the long-term decision-making stage of step 4, when it is predicted that the traffic of a specific service will continue to grow and exceed a certain threshold within a certain future time period, some channel resources are reserved in advance to ensure the quality of service of this service.

12. The dual-time-scale asymmetric multi-link channel binding method according to claim 1, wherein: In the cross-link resource optimization process of step 7, when it is detected that the utilization rate of a certain frequency band channel continues to be lower than the set utilization rate threshold, the idle channel resources of this frequency band are temporarily allocated to other services with urgent needs, and the original channel resource allocation is restored after the demand of this service ends.

Images