A method and system for coordinated scheduling of multi-link data flows of a 5G CPE device

Abstract

The application discloses a kind of 5G CPE device's multi-link data stream cooperative scheduling method and system, it is related to 5G network communication field, including in the different data interaction interfaces of 5G CPE device injection monitoring hook function;Identify service load demand, extract different monitoring factors;With link interference propagation model, heat-power consumption constraint model and domain link capacity benchmark, back-calculate spectrum efficiency and intermodulation interference loss coefficient, online correction link capacity benchmark, generate dynamic link load state;Multi-dimensional coupling analysis locates load bottleneck, and output scheduling decision;Parallel control each domain communication module, implement domain, hierarchical traffic scheduling and load balancing.The application makes 5G CPE device when multiple service data streams concurrent load, provides stable transmission guarantee for each service data stream.

Description

Technical Field

[0001] This application relates to the field of 5G network communication technology, and in particular to a multi-link data stream collaborative scheduling method and system for 5G CPE devices. Background Technology

[0002] With the large-scale deployment of 5G networks, 5G CPE devices, as key access terminals that convert 5G signals into Wi-Fi and wired networks, are increasingly being used in scenarios such as homes, small businesses, and outdoor environments. To meet users' demands for high-definition video, cloud applications, and multi-terminal concurrency, these devices typically integrate multiple communication capabilities, including 5G access, Wi-Fi access, and wired interfaces, and can carry multiple different types of service data streams simultaneously.

[0003] In actual operation, the various communication capabilities of 5G CPE devices are housed within the same compact hardware structure, sharing a power supply and being arranged close to each other. When multiple service data streams are transmitted concurrently and the overall load on the device increases, the actual transmission performance of a certain type of service often does not match its requirements, leading to stuttering, increased latency, or decreased speed in some services, thus affecting the user experience.

[0004] Currently, when facing the aforementioned concurrent transport scenarios, 5G CPE devices primarily process various service data streams based on pre-set fixed rules. However, during operation, the device's internal state continuously changes with variations in load and environment. Fixed rules are difficult to adapt to the device's real-time load conditions, making it challenging for the device to consistently provide transmission guarantees that match the needs of each service data stream during multi-service concurrency.

[0005] Therefore, how to ensure that 5G CPE devices can continuously provide transmission guarantees that match the needs of each service data stream when multiple service data streams are carried concurrently is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] In order to overcome the shortcomings of the existing technology and enable 5G CPE devices to continuously provide transmission guarantees for each service data stream that match its needs when multiple service data streams are carried concurrently, this application provides a multi-link data stream collaborative scheduling method and system for 5G CPE devices.

[0007] Firstly, the objective of this invention is achieved through the following technical solution: A method for multi-link data stream collaborative scheduling of a 5G CPE device, comprising: Inject monitoring hook functions into the data interaction interfaces of the 5G access domain, Wi-Fi access domain, and wired interface domain of the 5G CPE device to intercept the service data streams transmitted through each functional domain and collect the real-time status data of the communication modules to which each functional domain belongs. Identify the intercepted business data streams according to the preset hierarchical matching rules, determine the carrying requirement parameters, and extract the corresponding differentiated monitoring factors from each functional domain; The real-time status data and the differentiated monitoring factors are input into the pre-established link interference propagation model, thermal-power constraint model and domain link capability benchmark. The current spectral efficiency loss coefficient and inter-band intermodulation interference loss coefficient of each functional domain are calculated in reverse. The domain link capability benchmark is then corrected online in combination with the real-time power budget margin to obtain the dynamic link carrying status that characterizes the current available carrying capacity of each functional domain. The load requirement parameters, differential monitoring factors, and dynamic link load status are coupled and analyzed between the link interference propagation model and the thermal-power constraint model to locate the functional domain and bottleneck location parameters of the bottleneck load, and obtain the dynamic link scheduling decision results. Based on the dynamic link scheduling decision results, traffic scheduling and load balancing strategies are generated and triggered for execution by functional domain and service level.

[0008] By adopting the above technical solution, monitoring hook functions are injected into the data interaction interfaces of the three functional domains, enabling the device to simultaneously acquire the real-time status of each communication module while intercepting service data streams, thus associating and collecting "service-side requirements" and "device-side load conditions" at the same time. Next, service data streams are identified and their load requirement parameters are determined through hierarchical matching rules, allowing the device to differentiate the varying transmission needs of different services and avoid treating all services the same. Pre-established link interference propagation models, thermal-power constraint models, and domain-specific link capacity benchmarks are invoked, and the benchmarks are corrected online through back-calculation, ensuring that the load capacity obtained by the device reflects the current real-time situation rather than a fixed nominal value, thus solving the problem of mismatch between fixed rules and real-time load conditions. By performing multi-dimensional coupled analysis of load requirements and dynamic load status to locate load bottlenecks, scheduling decisions simultaneously consider the mutual influence of multiple constraints such as interference, thermal, and power consumption, avoiding misjudgments caused by single-dimensional judgments. Based on this, traffic scheduling and load balancing strategies are generated by functional domain and service level, guiding each service to the functional domain matching its needs, and providing continuous transmission guarantees that match the needs of concurrent multi-service traffic. Thus, this application enables 5G CPE devices to continuously provide stable transmission guarantees for each service data stream that match its needs when multiple service data streams are carried concurrently.

[0009] In a preferred embodiment of this application: the step of identifying the intercepted business data stream according to preset hierarchical matching rules, determining the carrying requirement parameters, and extracting corresponding differentiated monitoring factors from each functional domain includes: In 5G CPE devices, hierarchical matching rules are pre-configured. These rules match service data streams based on service type, functional domain, and QoS level, and support service subclass-level matching through service identifier masks. The carrying requirements of the service data streams matched by the hierarchy are identified, and the minimum effective throughput and the maximum tolerable latency jitter are determined as the carrying requirements parameters. For the 5G access domain, the signal-to-interference-plus-noise ratio and SA / NSA operating status of each member carrier are extracted; for the Wi-Fi access domain, the channel occupancy rate and OFDMA resource unit occupancy ratio are extracted; and for the wired interface domain, the port negotiation rate and VoLTE voice jitter are extracted as differentiated monitoring factors corresponding to each functional domain. Establish correlation monitoring relationships between functional domains based on differentiated monitoring factors of each functional domain, maintain real-time synchronization between the business feature library and external network service deployment, and update the business identifier mask in the hierarchical matching rules based on the change information of the business feature library.

[0010] By adopting the above technical solutions, a hierarchical matching and masking mechanism based on service type, functional domain, and QoS level is used to achieve fine identification at the service sub-category level; by extracting differentiated monitoring factors of each functional domain and establishing associated monitoring relationships, the device can perceive the mutual influence between functional domains; and by real-time synchronization and rule updates of the service feature library, the identification mechanism can be kept consistent with changes in external service deployment.

[0011] In a preferred embodiment, the process of determining the dynamic link bearer state includes: Pre-parse the link configuration files of 5G CPE devices to establish domain-specific link capability benchmarks; Substitute the real-time signal quality and throughput data of key nodes in the multi-link system into the domain-specific link capability benchmark to back-calculate the current spectral efficiency loss coefficient of each functional domain. Substitute the decrease in real-time signal quality under concurrent operation in each functional domain into the link interference propagation model to back-calculate the current inter-band intermodulation interference loss coefficient of each functional domain. The real-time power budget margin is determined based on the real-time voltage data of each power rail, and the available power budget in the domain link capability benchmark is dynamically adjusted. Based on the thermal-power constraint model, the gain attenuation of the power amplifier as the junction temperature increases and the change in insertion loss caused by the switching of the front-end module path are used as correction terms and added to the domain link capability benchmark. Based on the back-calculated loss coefficient, the adjusted available power budget, and the correction term, the domain-specific link capacity benchmark is corrected online to obtain the dynamic link carrying status that characterizes the current available carrying capacity of each functional domain.

[0012] By adopting the above technical solution, this application pre-establishes a domain-specific link capability benchmark, then uses real-time data to back-calculate the loss coefficient, dynamically adjusts the power budget, and superimposes device correction terms, enabling the equipment to obtain a carrying capacity close to the actual operating conditions without rebuilding the model, thus balancing computational overhead and correction accuracy.

[0013] In a preferred embodiment, the process of determining the dynamic link scheduling decision result of the 5G CPE device includes: Based on the bottleneck location parameters, the spectral interference propagation path of the functional domain where the bottleneck is located to adjacent frequency bands and adjacent functional domains is extracted from the link interference propagation model, and the comprehensive impact assessment value of the functional domain where the bottleneck is located on the link quality of adjacent functional domains is obtained. Based on the bottleneck location parameters and comprehensive impact assessment values, the junction temperature and power consumption status data of key components in the bottleneck module are extracted from the thermal-power constraint model to obtain the thermal throttling warning level under the current load. Based on the bottleneck location parameters, the dynamic link carrying status is queried to obtain the real-time available bandwidth and real-time effective throughput of the functional domain where the bottleneck is located. Combined with the thermal throttling warning level and the comprehensive impact assessment value, the predicted value of the link transmission performance under the current load is calculated to obtain the dynamic link scheduling decision result.

[0014] By adopting the above technical solutions, the interference propagation effect and thermal derating warning level are extracted step by step between the link interference propagation model, the thermal-power constraint model and the bearer status, and the performance prediction value is calculated. This allows the scheduling decision to take into account the coupling effect across functional domains and physical constraints, resulting in more accurate and reliable bottleneck location and decision results.

[0015] In a preferred embodiment of this application, the method further includes: when the deviation between the actual transmission performance of the service data flow after scheduling and the bearer requirement parameters exceeds a preset performance deviation threshold, triggering iterative correction of the link scheduling strategy, including: The actual transmission performance of the scheduled service data stream is evaluated based on multi-dimensional transmission indicators, and the performance deviation between the actual transmission performance and the carrying requirement parameters is determined. When the performance deviation value is greater than the preset performance deviation threshold, fine-grained channel probing is enabled for the bottleneck functional domain to obtain the carrier-by-carrier signal-to-interference-plus-noise ratio, and the domain-specific link capability benchmark is corrected at the first level to obtain the first-level correction scheduling strategy. When the difference in performance deviation before and after the first-level correction is less than the preset deviation change threshold, the equivalent junction temperature of the bottleneck module and the power amplifier gain attenuation are calculated based on the measured throughput and power consumption before and after the business data flow scheduling. The second-level correction scheduling strategy is then obtained by performing a second-level correction on the domain link capability benchmark. When the difference in performance deviation before and after the second-level correction is less than the preset deviation change threshold, the voltage drop and actual power budget margin of the corresponding power rail are calculated based on the measured throughput and equivalent junction temperature before and after the second-level correction. The third-level correction is then applied to the domain link capability benchmark to obtain the third-level correction scheduling strategy, and traffic scheduling and load balancing strategies for functional domains and service levels are regenerated.

[0016] By adopting the above technical solution, this application uses a progressive multi-level closed-loop correction method of "deflection threshold trigger detection, and then using the deviation change value to determine whether to go deeper". It calculates physical quantities such as carrier signal-to-interference-plus-noise ratio, equivalent junction temperature and gain attenuation, and power rail voltage drop at each level, so that the correction can be deepened layer by layer as needed. This avoids the excessive overhead caused by uniform deep correction, and can accurately locate deeper physical causes when shallow correction is ineffective, thus significantly improving the effectiveness of correction.

[0017] In a preferred embodiment of this application, the preset deviation change threshold is determined by the following method: Obtain the key business proportion of the business data stream and the current idle bearer capacity proportion of the 5G CPE device; When the proportion of key business is greater than or equal to the preset proportion of key business, the calculation formula for the preset deviation change threshold is: preset deviation change threshold = (1 - proportion of key business) × benchmark deviation change value; When the proportion of key business is less than the preset proportion of key business, and the proportion of idle capacity is greater than or equal to the preset proportion of idle capacity, the calculation formula for the preset deviation change threshold is: preset deviation change threshold = (1 + idle capacity proportion) × baseline deviation change value.

[0018] By adopting the above technical solution, the deviation change threshold is dynamically determined based on the proportion of critical business and idle capacity, so that the trigger sensitivity of the correction can be adaptively adjusted according to the importance of the business and the equipment margin. It is more sensitive when there are many critical businesses or the margin is tight, and more lenient when the margin is sufficient, thus achieving a match between the correction intensity and the actual needs.

[0019] Secondly, the objective of this invention is achieved through the following technical solution: A multi-link data stream collaborative scheduling system for 5G CPE devices includes: The monitoring and acquisition module is used to inject monitoring hook functions into the data interaction interfaces of the 5G access domain, Wi-Fi access domain and wired interface domain of the 5G CPE device to intercept the service data stream transmitted through each functional domain and collect the real-time status data of the communication module to which each functional domain belongs. The identification and extraction module is used to identify the intercepted business data streams according to the preset hierarchical matching rules, determine the carrying requirement parameters, and extract the corresponding differentiated monitoring factors from each functional domain. The bearer status correction module is used to input the real-time status data and the differentiated monitoring factors into the pre-established link interference propagation model, thermal-power constraint model and domain link capability benchmark, back-calculate the current spectral efficiency loss coefficient and inter-band intermodulation interference loss coefficient of each functional domain, and combine the real-time power budget margin to correct the domain link capability benchmark online, so as to obtain the dynamic link bearer status that characterizes the current available bearer capability of each functional domain. The coupling analysis module is used to perform coupling analysis between the load requirement parameters, differential monitoring factors and dynamic link load status in the link interference propagation model and the thermal-power constraint model, locate the functional domain of the bottleneck load and the bottleneck location parameters, and obtain the dynamic link scheduling decision results. The scheduling control module is used to generate and execute traffic scheduling and load balancing strategies based on the dynamic link scheduling decision results, according to functional domains and service levels.

[0020] Thirdly, the objective of this invention is achieved through the following technical solution: A 5G CPE device includes a 5G communication module, a Wi-Fi communication module, a wired interface module, a clock module, a hierarchical power supply module, and a processor and a memory connected to each module. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the multi-link data stream collaborative scheduling method of the 5G CPE device described above.

[0021] Fourthly, the objective of this invention is achieved through the following technical solution: A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described multi-link data stream collaborative scheduling method for a 5G CPE device.

[0022] In summary, this application includes at least one of the following beneficial technical effects: 1. By injecting monitoring hook functions into the interfaces of each functional domain of the 5G CPE device, the service carrying requirements are synchronously correlated with the real-time status of the device. Based on the pre-established link-related model and the domain-specific link capability benchmark, online back-calculation correction and multi-dimensional coupling analysis are performed to generate scheduling and load balancing strategies by functional domain and service level. This ensures that the device's scheduling decisions are always based on the current real-time carrying status, rather than fixed nominal rules. Therefore, when multiple services are carried concurrently, it can continuously provide transmission guarantees for each service data stream that match its needs, effectively solving the problem of mismatch between fixed rules and real-time carrying status. 2. Hook processing chains and refined business identification improve the efficiency and accuracy of bottleneck perception; the back-calculation correction of domain capability benchmarks controls computational overhead while closely reflecting the actual load conditions; cross-model coupling analysis allows decision-making to take into account the multiple interactions of interference, heat, and power consumption; multi-level closed-loop correction driven by deviation change values, combined with adaptive thresholds, allows corrections to deepen layer by layer as needed, with the intensity matching the importance of the business and equipment margin; the quantitative calculation and verification of performance prediction values ​​make bottleneck identification and congestion relief prediction more reliable. Attached Figure Description

[0023] Figure 1 This is a flowchart of a multi-link data stream collaborative scheduling method for a 5G CPE device according to an embodiment of this application; Figure 2 This is a flowchart of step S2 in a multi-link data stream collaborative scheduling method for a 5G CPE device according to an embodiment of this application; Figure 3 This is a schematic diagram of a device according to one embodiment of this application. Detailed Implementation

[0024] The present application will be further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, such as Figure 1 As shown, this application discloses a multi-link data stream collaborative scheduling method for 5G CPE devices, which specifically includes the following steps: S1: Inject monitoring hook functions into the data interaction interfaces of the 5G access domain, Wi-Fi access domain, and wired interface domain of the 5G CPE device to intercept the service data streams transmitted through each functional domain and collect the real-time status data of the communication modules to which each functional domain belongs.

[0026] In this embodiment, the software system of the 5G CPE device is built on an embedded Linux kernel. The 5G access domain corresponds to the baseband modulation and demodulation module and its radio frequency front-end, the Wi-Fi access domain corresponds to the Wi-Fi 6 / 6E wireless chip, and the wired interface domain corresponds to the gigabit / 2.5G Ethernet switching chip. The data interaction interface is the data transceiver interface between the drivers of each functional domain and the kernel protocol stack, including the PDCP / RLC layer transceiver interface of the 5G access domain, the MAC layer transceiver interface of the Wi-Fi access domain, and the network card driver transceiver interface of the wired interface domain. The monitoring hook function adopts a hybrid injection of the hook point of the kernel Netfilter framework and the callback registration mechanism of each functional domain driver to bypass and capture the service data stream passing through each functional domain.

[0027] Specifically, while intercepting service data streams, the monitoring hook function synchronously collects real-time status data through the register read interfaces and sensor interfaces of each communication module: for the 5G access domain, it collects RSRP, RSRQ, SINR, the modulation and coding scheme (MCS) of each component carrier, and instantaneous throughput; for the Wi-Fi access domain, it collects channel utilization, retransmission rate, and the number of associated terminals; for the wired interface domain, it collects port link status, negotiation rate, and transmit / receive queue depth; and simultaneously, it collects the junction temperature of the power amplifier, the voltage of each power rail, and the overall power consumption through the device's built-in temperature sensor and power management chip (PMIC). The collected real-time status data and the timestamps of the corresponding service data streams are written to the kernel shared memory area for subsequent modules to read.

[0028] S2: Identify the intercepted business data streams according to the preset hierarchical matching rules, determine the carrying requirement parameters, and extract the corresponding differentiated monitoring factors from each functional domain.

[0029] In this embodiment, the hierarchical matching rules are pre-stored in the non-volatile storage area of ​​the device in the form of a rule table. The rule table entries sequentially record the service type, the functional domain, the QoS level, and the corresponding service identifier mask. During identification, the five-tuple and DPI features are first extracted from the intercepted service data stream. The service type field, functional domain field, and QoS level field are then compared sequentially. Once a rule table entry is matched, the service category of the service data stream is determined, and the bearer requirement parameters corresponding to the service category are read. The bearer requirement parameters include the minimum effective throughput and the maximum tolerable latency jitter. For example, cloud gaming services require a minimum effective throughput of 30Mbps and a maximum tolerable latency jitter of 8ms; 4K video services require a minimum effective throughput of 25Mbps and a maximum tolerable latency jitter of 50ms; and VoLTE voice services require a minimum effective throughput of 128kbps and a maximum tolerable latency jitter of 20ms.

[0030] While determining the bearer requirements parameters, corresponding differentiated monitoring factors are extracted from each functional domain based on its physical characteristics: for the 5G access domain, the SINR and SA / NSA operating status of each member carrier are extracted; for the Wi-Fi access domain, the channel occupancy rate and OFDMA resource unit (RU) occupancy ratio are extracted; and for the wired interface domain, the port negotiation rate and VoLTE voice jitter are extracted. These extracted differentiated monitoring factors, together with the collected real-time status data, constitute the bearer status description vector for each functional domain, and are categorized and cached according to functional domain identifiers. This enables the equipment to distinguish the differentiated transmission requirements of different services, while simultaneously obtaining fine-grained bearer status reflecting the characteristics of each functional domain.

[0031] Specifically, such as Figure 2 As shown, step S2 includes: S21: Pre-configure hierarchical matching rules in 5G CPE devices. The hierarchical matching rules match service data streams based on service type, functional domain, and QoS level, and support service subclass-level matching through service identifier masks.

[0032] In this embodiment, the hierarchical matching rules are pre-installed in the non-volatile storage area using a three-level matching tree: "service type → functional domain → QoS level". The first level is the service type node, such as cloud gaming, video, voice, and file transfer; the second level is the functional domain node; and the third level is the QoS level node, identified by 5QI with values ​​from 1 to 9. The service identifier mask is a binary mask obtained by comparing the bitwise AND operation of the service identifier field.

[0033] Specifically, during identification, the intercepted service data stream is first processed by extracting the packet 5-tuple and DPI (Deep Packet Inspection) features, followed by a three-level matching process: The first level compares the DPI features with the service type feature library to determine the service type; if a match is found, the process proceeds to the corresponding service type node. The second level determines the functional domain branch based on the current entry functional domain of the service data stream. The third level reads the packet's DSCP field and maps it to a 5QI identifier to complete QoS level matching. When further differentiation of service subclasses is required, the service identifier field of the service data stream (such as port number combinations or application layer feature fields) is bitwise ANDed with the pre-set service identifier mask under that node. A match is determined when the result equals the expected value corresponding to the mask. The three-level matching tree employs a longest prefix priority matching strategy.

[0034] S22: Identify the carrying requirements of the service data streams that are matched by hierarchical classification, and determine the minimum effective throughput and the maximum tolerable latency jitter as carrying requirement parameters.

[0035] In this embodiment, the bearer requirement parameters refer to the link transmission indicators that must be met to ensure a normal experience for a certain service data stream. This application takes two parameters: minimum effective throughput and maximum tolerable latency jitter. Minimum effective throughput refers to the lowest average rate required for the service data stream without stuttering or frame dropping. Maximum tolerable latency jitter refers to the upper limit of the allowable fluctuation in the arrival time interval between adjacent packets in the service data stream. Exceeding this upper limit indicates a degraded experience. Each matching leaf node is pre-associated with a set of calibration values ​​for the above parameters. The calibration values ​​in this embodiment are as follows: cloud gaming video substream: minimum effective throughput 30Mbps, maximum tolerable latency jitter 8ms; cloud gaming control substream: minimum effective throughput 2Mbps, maximum tolerable latency jitter 5ms; 4K video service: minimum effective throughput 25Mbps, maximum tolerable latency jitter 50ms; VoLTE voice service: minimum effective throughput 128kbps, maximum tolerable latency jitter 20ms; file transfer service: minimum effective throughput 5Mbps, maximum tolerable latency jitter unlimited. The calibration values ​​are preset by those skilled in the art based on the protocol specifications and experience thresholds of various services.

[0036] S23: Extract the signal-to-interference-plus-noise ratio and SA / NSA operating status of each member carrier for the 5G access domain, extract the channel occupancy rate and OFDMA resource unit occupancy ratio for the Wi-Fi access domain, and extract the port negotiation rate and VoLTE voice jitter for the wired interface domain, as differentiated monitoring factors corresponding to each functional domain.

[0037] In this embodiment, differentiated monitoring factors refer to the set of feature quantities selected for each functional domain based on its different physical transmission mechanisms, which can most sensitively reflect changes in the carrying capacity of the functional domain. Differential extraction is because the carrying capacity of the 5G access domain is mainly constrained by the quality of the wireless channel and the network mode; the Wi-Fi access domain is mainly constrained by channel contention and resource unit allocation; and the wired interface domain is mainly constrained by the physical negotiation rate and voice stream jitter. SA stands for Standalone networking, and NSA stands for Non-Standalone networking; the channel occupancy rate is the proportion of time the channel is occupied per unit time; the OFDMA resource unit RU occupancy ratio is the proportion of the number of allocated RUs to the total number of available RUs; and VoLTE voice jitter is the latency fluctuation of adjacent packets in the voice stream transmitted through the wired interface.

[0038] S24: Establish the correlation monitoring relationship between functional domains based on the differentiated monitoring factors of each functional domain, and maintain real-time synchronization between the business feature library and the external network business deployment. Update the business identifier mask in the hierarchical matching rules based on the change information of the business feature library.

[0039] In this embodiment, the correlation monitoring relationship refers to the correlation mapping relationship describing the mutual influence and constraint between the carrying status of different functional domains. It is used to take into account the coupling effect between a certain functional domain and its adjacent functional domains when evaluating a certain functional domain. It is expressed in the form of a correlation coefficient matrix, where the matrix element ρ(i,j) represents the influence weight of the change of the differentiated monitoring factor of the i-th functional domain on the carrying status of the j-th functional domain, and the value ranges from 0 to 1. For example, the 5G access domain n78 band (3.5GHz) is adjacent to the Wi-Fi access domain 5GHz band. When they are used concurrently, there is out-of-band radiation interference, so their correlation coefficient takes a higher value. In this embodiment, it is 0.35. However, the wired interface domain and the 5G access domain have no radio frequency coupling, so their correlation coefficient is 0. The service feature library refers to the data set that stores the DPI identification features and service identifiers of various services.

[0040] Specifically, the correlation monitoring relationship is established as follows: During the equipment debugging phase, each functional domain is made to work individually and in pairs concurrently. The relative degradation of the carrying capacity of adjacent functional domains is measured when the factors of a certain functional domain change. The correlation coefficient ρ(i,j) is calibrated by the ratio of degradation to factor change, filled into the correlation coefficient matrix and stored permanently. It is called by the link interference propagation model at runtime. Real-time synchronization of the service feature database is achieved through a kernel-level communication mechanism: The device periodically pulls service deployment change information from the external network service management platform through the management channel. When a new service type or service identifier change is detected, the change information is automatically written into the service feature database, and the corresponding service identifier and service identifier mask parameters in the three-level matching tree of the hierarchical matching rules are updated.

[0041] S3: Input real-time status data and differentiated monitoring factors into the pre-established link interference propagation model, thermal-power constraint model and domain-specific link capability benchmark, back-calculate the current spectral efficiency loss coefficient and inter-band intermodulation interference loss coefficient of each functional domain, and combine the real-time power budget margin to correct the domain-specific link capability benchmark online, so as to obtain the dynamic link carrying status that characterizes the current available carrying capacity of each functional domain.

[0042] In this embodiment, the link interference propagation model, thermal-power constraint model, and domain-specific link capability benchmark are all pre-established and stored in the storage area before the device leaves the factory or during the first power-on. They are only called and corrected online during operation and do not need to be regenerated. The domain-specific link capability benchmark records the theoretical available bandwidth and throughput limit of each functional domain under nominal conditions. The link interference propagation model records the interference coupling relationship between adjacent frequency bands and adjacent functional domains. The thermal-power constraint model records the correspondence between junction temperature, power consumption, and performance degradation of key components.

[0043] Specifically, step S3 includes: S31: Pre-parse the link configuration file of the 5G CPE device to establish a domain-specific link capability benchmark that includes carrier aggregation capabilities of each functional domain, physical layer rate table, front-end module gain parameters, and hierarchical power budget.

[0044] In this embodiment, the domain-specific link capability benchmark is a dataset recording the theoretically available carrying capacity and constraint parameters under nominal conditions for each of the three functional domains. Each functional domain includes four types of data: carrier aggregation capability, physical layer rate table, front-end module gain parameters, and tiered power budget. The nominal conditions are: rated power supply, 25°C ambient temperature, and no additional interference. The physical layer rate table is a mapping table between CQI / MCS levels and physical layer rates; the front-end module gain parameters include the rated gain of the power amplifier, the gain of the low-noise amplifier (LNA), and the nominal insertion loss of the RF path. The tiered power budget represents the upper limit of the rated power allocated to each functional domain by each power rail.

[0045] Specifically, the link configuration file is a factory-pre-configured structured file that describes the hardware capabilities and power allocation of each functional domain. During the initial power-on initialization phase, this file is parsed item by item, organized into a benchmark data table by functional domain, and the theoretical available bandwidth limit B0 and theoretical throughput limit T0 of each functional domain are calculated based on the physical layer rate table and carrier aggregation capability and recorded in the benchmark.

[0046] S32: Substitute the real-time signal quality and throughput data of key nodes in multiple links into the domain-specific link capability benchmark, calculate the current spectral efficiency loss coefficient of each functional domain, and substitute the decrease in real-time signal quality under concurrent operation in each functional domain into the link interference propagation model to calculate the current inter-band intermodulation interference loss coefficient of each functional domain.

[0047] In this embodiment, the link interference propagation model is a deterministic relationship model that describes the diffusion of radio frequency transmission energy from a certain functional domain to adjacent frequency bands and adjacent functional domains, causing link quality degradation. The link interference propagation model consists of: (1) a frequency band adjacency table, which records the frequency interval Δf and hardware isolation ISO(Δf) between each pair of working frequency bands of each functional domain. In this embodiment, the isolation of n78 to Wi-Fi 5GHz is recorded as 30dB; (2) an interference diffusion function, whose input is the transmission power P_tx and the frequency interval Δf, and whose output is the interference increment ΔI=P_tx−ISO(Δf) to adjacent functional domains, which describes the law of attenuation of transmission energy as the frequency interval increases; (3) an association weighting unit, which calls the S24 association coefficient matrix ρ to weight and summarize the interference increments of each path.

[0048] Specifically, the spectral efficiency loss coefficient α is calculated as α = SE_real / SE_nom, where SE_real is the actual spectral efficiency obtained from the physical layer rate table by looking up the measured SINR of the key node, and SE_nom is the nominal spectral efficiency. α takes a value of 0~1. The inter-band intermodulation interference loss coefficient β is calculated as follows: the measured SINR of a functional domain working alone is denoted as SINR_solo, and the measured SINR of working concurrently with an adjacent functional domain is denoted as SINR_co, both expressed as linear values; SINR_co is substituted into the link interference propagation model and weighted by the correlation coefficient ρ to obtain the equivalent signal-to-interference-plus-noise ratio (SINNR_eff) under concurrent operation: SINR_solo / (1+ρ·(SINR_solo / SINR_co−1)); then, based on the mapping relationship between spectral efficiency and SINNR, the decrease in SINNR is mapped to the spectral efficiency (capacity) loss ratio by β = 1−log2(1+SINR_eff) / log2(1+SINR_solo), with β taking a value of 0~1.

[0049] S33: Determine the real-time power budget margin based on the real-time voltage data of each power rail, and dynamically adjust the available power budget in the domain link capability baseline.

[0050] In this embodiment, the real-time power budget margin is the remaining power that the device can continue to allocate to each functional domain without triggering power supply protection. The PMIC collects the real-time voltage U_real and current I_real of each power rail and calculates the real-time output power P_real = U_real × I_real. The real-time power budget margin ΔP is obtained by subtracting P_real from the rated upper limit P_max of the power rail in the tiered power budget. When U_real drops below 95% of the rated voltage U_nom (e.g., U_real < 0.95·U_nom), the allocable power is reduced by the actual output power upper limit corresponding to the current voltage, i.e., P_avail = U_real × I_max − P_real, where I_max is the rated maximum output current of the power rail.

[0051] S34: Based on the thermal-power constraint model, the gain attenuation of the power amplifier as the junction temperature increases and the change in insertion loss caused by the switching of the front-end module path are used as correction terms and superimposed on the domain link capability benchmark.

[0052] In this embodiment, the thermal-power constraint model is a deterministic relationship model that describes the dynamic relationship and performance constraints between functional domain load, device power consumption and key device junction temperature. It includes three interconnected parts: (1) Power consumption-junction temperature mapping relationship, which converts power consumption P_dis into junction temperature T_j by using the device thermal resistance θ_JA according to the junction temperature calculation formula T_j=T_amb+P_dis×θ_JA; (2) Junction temperature-gain attenuation curve, which describes the law of power amplifier gain attenuation as junction temperature increases. In this embodiment, the gain temperature coefficient is −0.02dB / ℃ and the reference temperature is 25℃, i.e. ΔG_T=−0.02×(T_j−25); (3) Path insertion loss table, which records the insertion loss values ​​of different RF paths of the front-end module.

[0053] Specifically, the current junction temperature T_j of the power amplifier is collected by the temperature sensor, and ΔG_T is calculated from the junction temperature-gain attenuation curve; the current insertion loss L_cur is obtained from the path insertion loss table of the currently selected path in the front-end module, and subtracted from the reference path insertion loss L_ref to obtain ΔL_S=L_cur−L_ref; ΔG_T and ΔL_S are converted into equivalent spectral efficiency loss and then superimposed as correction terms on the domain link capability reference.

[0054] S35: Based on the back-calculated loss coefficient, the adjusted available power budget, and the correction term, the domain-specific link capacity benchmark is corrected online to obtain the dynamic link carrying status that characterizes the current available carrying capacity of each functional domain.

[0055] Specifically, the online correction uses the upper limit of the theoretical baseline as the benchmark value, and successively incorporates α and β of S32, P_avail of S33, and ΔG_T and ΔL_S of S34: Real-time available bandwidth B_a = B0 × (1 − β); Real-time effective throughput T_a = T0 × α × (1 − β) × k_P × k_G, where k_P is the power budget constraint factor, which is proportionally reduced when P_avail is insufficient to support full power and is taken as 1 when sufficient; k_G is the device correction factor, which is obtained by converting and normalizing ΔG_T + ΔL_S according to 10^((ΔG_T + ΔL_S) / 10). After the calculation of B_a and T_a of each functional domain is completed, the dynamic link bearer state is written into shared memory.

[0056] Furthermore, a multi-link data flow collaborative scheduling method for 5G CPE devices also includes: when the deviation between the actual transmission performance of the service data flow after scheduling and the bearer requirement parameters exceeds a preset performance deviation threshold, triggering iterative correction of the link scheduling strategy, specifically including the following steps: S351: Based on multi-dimensional transmission indicators, evaluate the actual transmission performance of the scheduled service data stream and determine the performance deviation between the actual transmission performance and the carrying requirement parameters.

[0057] In this embodiment, the performance deviation value refers to the degree of deviation between the actual transmission performance of the service data flow after scheduling and its carrying requirement parameters, denoted as E. The larger E is, the more the actual carrying deviates from the service requirements. Among the multi-dimensional transmission indicators, the delay jitter refers to the fluctuation of the time interval between the arrival of adjacent packets, and the packet loss mode refers to whether the packet loss is random and scattered or continuous and bursty.

[0058] For example, the performance evaluation is performed continuously in 100ms statistical windows after scheduling execution: the monitoring hook function collects the measured throughput T_real, measured latency jitter J_real, and packet loss sequence of the service data stream. The packet loss sequence is classified into packet loss modes according to the length of consecutive packet loss and weighted accordingly, with random type having a weight of 1.0 and burst type having a weight of 1.5. Then, the three types of indicators are normalized and weighted to synthesize the actual transmission performance, which is compared with the service carrying requirement parameters to obtain the performance deviation value E=w1·max(0,(T_d−T_real) / T_d)+w2·max(0,(J_real−J_d) / J_d)+w3·L_w, where L_w is the packet loss impact term, which is obtained by multiplying the packet loss rate by the packet loss mode weight (random type is taken as 1.0, burst type is taken as 1.5) and normalizing. In this embodiment, the throughput weight w1=0.5, the jitter weight w2=0.3, and the packet loss mode weight w3=0.2.

[0059] S352: When the performance deviation value is greater than the preset performance deviation threshold, fine-grained channel probing is enabled for the bottleneck functional domain to obtain the carrier-by-carrier signal-to-interference-plus-noise ratio, and the domain-specific link capability benchmark is corrected at the first level to obtain the first-level correction scheduling strategy.

[0060] In this embodiment, the first-level correction is the first level in the step-by-step correction, which targets the degradation of the bearer at the "channel quality level". That is, it first suspects that the decline in bearer capacity is due to the deterioration of the wireless channel, and corrects the domain link capacity benchmark by measuring the channel quality in a more granular way. The preset performance deviation threshold is the threshold for determining whether to start the correction. In this embodiment, it is set to 0.1, that is, the correction is started when the performance deviation value exceeds 10% of the bearer requirement.

[0061] Specifically, when E>0.1, a fine-grained channel probing command is issued to the functional domain where the bottleneck is located: the current SINR of each component carrier in the 5G access domain is measured. The current SINR is the carrier-by-carrier signal-to-interference-plus-noise ratio, which is different from the carrier-average SINR extracted in S23. This allows the identification of which specific carrier is degraded. Then, the signal-to-noise ratio of each RU in the Wi-Fi access domain is measured. The carrier-by-carrier signal-to-interference-plus-noise ratio is substituted into the domain-specific link capability benchmark to recalculate the actual spectral efficiency of each carrier. The spectral efficiency loss coefficient α in S32 is refreshed, and the real-time effective throughput T_a is recalculated accordingly, completing the first-level correction of the domain-specific link capability benchmark. Then, the traffic scheduling strategy for each functional domain and service level is regenerated with the corrected carrying capacity and issued as the first-level corrected scheduling strategy for execution.

[0062] S353: When the difference between the performance deviation before and after the first-level correction is less than the preset deviation change threshold, the equivalent junction temperature of the bottleneck module and the power amplifier gain attenuation are calculated based on the measured throughput and power consumption before and after the service data flow scheduling. The second-level correction is then applied to the domain link capability benchmark to obtain the second-level correction scheduling strategy.

[0063] In this embodiment, the difference between the performance deviation value E0 before the first-level correction and the performance deviation value E1 after the first-level correction is denoted as ΔE1=E0−E1, which is used to determine whether the first-level correction is effective. The second-level correction is the second level in the step-by-step correction, which is aimed at the "device thermal degradation level". That is, when the improvement of the first-level correction (channel level) is limited, it is suspected that the degradation of the load is caused by the power amplifier gain attenuation due to the increase of the device junction temperature. The equivalent junction temperature correction benchmark is calculated by back calculation. The preset deviation change threshold is the threshold for determining whether the current level correction has converged and whether it should enter the next level correction.

[0064] Specifically, calculate ΔE1 = E0 − E1. When ΔE1 < the preset deviation change threshold, it indicates that the first-level correction has little effect on performance improvement, and the bottleneck is not purely a channel factor. Then, proceed to the second-level correction: take the measured throughput before and after the service data flow scheduling and the measured power consumption of the module. Based on the power consumption-junction temperature mapping relationship in the thermal-power constraint model, calculate the equivalent junction temperature T_j' of the bottleneck module from the measured power consumption. Then, calculate the actual gain attenuation ΔG_T' of the power amplifier from the junction temperature-gain attenuation curve. Update the correction term of S34 with the calculated T_j' and ΔG_T', perform the second-level correction on the domain link capability benchmark, recalculate the carrying capacity, and regenerate the second-level correction scheduling strategy for execution.

[0065] S354: When the difference between the performance deviation values ​​before and after the second-level correction is less than the preset deviation change threshold, the voltage drop and actual power budget margin of the corresponding power rail are calculated based on the measured throughput and equivalent junction temperature before and after the second-level correction. The third-level correction is then applied to the domain link capability benchmark to obtain the third-level correction scheduling strategy, and traffic scheduling and load balancing strategies for functional domains and service levels are regenerated.

[0066] In this embodiment, the difference between the performance deviation value E1 before the second-level correction and the performance deviation value E2 after the second-level correction is denoted as ΔE2=E1−E2; the third-level correction is the third level in the step-by-step correction, which is aimed at the "power supply capacity level". That is, when the improvement of the second-level correction (device thermal level) is still limited, it is finally suspected that the load degradation is caused by the voltage drop of the power rail under heavy load and the insufficient actual available power. The benchmark is corrected by back-calculating the voltage drop and the actual power budget margin.

[0067] Specifically, ΔE2 = E1 − E2 is calculated. When ΔE2 < the preset deviation change threshold, it indicates that the improvement from the second-level correction is minimal, the bottleneck is in the power supply, and the process switches to the third-level correction: the measured throughput before and after the second-level correction and the equivalent junction temperature are back-calculated, combined with the power rail voltage collected by the PMIC, to back-calculate the voltage drop ΔU of the power rail relative to the rated voltage under the current load and the corresponding real power budget margin P_avail'. The real power budget margin here is different from the margin estimated by S33 based on instantaneous voltage; the real value after the load dynamic response is included here. ΔU and P_avail' are used to correct the available power budget of S33, and the third-level correction is performed on the domain link capacity benchmark. The traffic scheduling and load balancing strategies for functional domains and service levels are regenerated and executed according to the carrying capacity after the third-level correction. If the performance deviation value still exceeds the threshold after the third-level correction, it is determined that the bottleneck functional domain has no available carrying margin, triggering cross-functional domain service migration or degradation processing.

[0068] In this embodiment, the preset deviation change threshold is determined by the following method: S71: Obtain the key business proportion of the business data stream and the current idle bearer capacity proportion of the 5G CPE device.

[0069] In this embodiment, critical services refer to latency-sensitive or high-priority services, such as cloud gaming, VoLTE voice, and industrial control. The critical service percentage refers to the proportion of critical services among all currently carried services, denoted as r_k, with a value of 0 to 1. The idle bearer capacity percentage refers to the proportion of the sum of the currently idle and available bearer capacities of each functional domain of the device to its total bearer capacity, denoted as r_idle, with a value of 0 to 1.

[0070] Specifically, the critical service ratio r_k is calculated based on the S21 hierarchical matching results: Critical service ratio = Sum of the throughput requirements of service data flows with a hit rate of 5QI level ≤ 5 (high priority) ÷ Sum of the throughput requirements of all current services. The idle capacity ratio is calculated based on the S35 dynamic link carrying status: Idle capacity ratio = (Sum of real-time effective throughput T_a of each functional domain - Sum of current actual occupied throughput) ÷ Sum of T_a of each functional domain.

[0071] S72: When the proportion of key business is greater than or equal to the preset proportion of key business, the calculation formula for the preset deviation change threshold is: preset deviation change threshold = (1 - key business proportion) × benchmark deviation change value.

[0072] In this embodiment, the reference deviation change value is the calibration reference of the preset deviation change threshold, denoted as δ0, which is 0.03 in this embodiment; the preset key business proportion value is 0.5.

[0073] S73: When the proportion of critical business is less than the preset proportion of critical business, and the proportion of idle capacity is greater than or equal to the preset proportion of idle capacity, the calculation formula for the preset deviation change threshold is: preset deviation change threshold = (1 + idle capacity proportion) × baseline deviation change value.

[0074] Specifically, when r_k is less than the preset critical business ratio and r_idle is less than the preset idle capacity ratio, the preset deviation change threshold is taken as the baseline deviation change value.

[0075] S4: Couple the load demand parameters, differential monitoring factors, and dynamic link load status between the link interference propagation model and the thermal-power constraint model to locate the functional domain of the bottleneck load and the bottleneck location parameters, and obtain the dynamic link scheduling decision results.

[0076] In this embodiment, the coupling analysis is a multi-dimensional coupling analysis. The multi-dimensional coupling analysis takes the carrying requirement parameters of the business data flow as the target constraint, compares the dynamic link carrying status of each functional domain with the carrying requirement parameters one by one. When the current available carrying capacity of a certain functional domain is lower than the sum of the carrying requirement parameters of the services carried on it, it is determined that the functional domain has a carrying bottleneck, and the functional domain identifier, the available carrying capacity gap value and the corresponding differential monitoring factors are recorded as bottleneck location parameters.

[0077] Coupling analysis does not judge a single functional domain in isolation, but evaluates it in conjunction with the link interference propagation model and the thermal-power constraint model: on the one hand, the link interference propagation model evaluates the impact of increasing the transmit power of the bottleneck functional domain or aggregating carriers on the interference propagation of adjacent functional domains; on the other hand, the thermal-power constraint model evaluates whether the changes in device junction temperature and power consumption caused by the adjustment reach the thermal throttling threshold. After combining the two constraints, the predicted link transmission performance of each candidate scheduling scheme is calculated, and the scheme with the highest predicted performance and which does not cause new bottlenecks is selected as the dynamic link scheduling decision result.

[0078] Specifically, step S4 includes: S41: Based on the bottleneck location parameters, extract the spectral interference propagation path of the functional domain where the bottleneck is located to adjacent frequency bands and adjacent functional domains from the link interference propagation model, and obtain the comprehensive impact assessment value of the functional domain where the bottleneck is located on the link quality of adjacent functional domains.

[0079] In this embodiment, the spectrum interference propagation path is the interference transmission link that starts from the bottleneck functional domain and points to each disturbed adjacent functional domain along the frequency band adjacency relationship table recorded in the S32 link interference propagation model. The comprehensive impact assessment value is a single quantitative value obtained by weighting and summing the interference increment of the bottleneck functional domain to all adjacent functional domains according to the correlation coefficient, denoted as I_total. The larger I_total is, the greater the negative impact.

[0080] Specifically, the bottleneck functional domain identifier and its current transmit power P_tx and operating frequency band are read from the bottleneck location parameters; the adjacent functional domains and frequency interval Δf are determined by looking up the frequency band adjacency table; the interference increment ΔI_j=max(0,P_tx−ISO(Δf)−N_floor) for each adjacent functional domain j is calculated using the interference spread function, where N_floor is the noise floor of the adjacent functional domain. When P_tx−ISO(Δf) is lower than the noise floor, the interference of that path can be ignored and ΔI_j is taken as 0; then, the associated weighting unit is used to weight and summarize with ρ to obtain I_total=Σjρ(bottleneck domain,j)×ΔI_j, where I_total≥0.

[0081] S42: Based on the bottleneck location parameters and comprehensive impact assessment values, extract the junction temperature and power consumption status data of the key components in the bottleneck module from the thermal-power constraint model to obtain the thermal throttling warning level under the current load.

[0082] In this embodiment, the thermal throttling warning level is determined based on the degree to which the junction temperature of the key component approaches its throttling protection threshold. The warning level is given by the junction temperature-throttling threshold table in the thermal-power constraint model. In this embodiment, it is divided into 0 to 3 levels: Level 0 (junction temperature <75℃, safe), Level 1 (75~90℃, attention), Level 2 (90~105℃, warning, limit power growth), and Level 3 (>105℃, emergency, load must be reduced immediately).

[0083] Specifically, the current power consumption P_dis and ambient temperature T_amb of the bottleneck module are read from the bottleneck location parameters. The junction temperature is calculated using the power consumption-junction temperature mapping relationship of the thermal-power consumption constraint model according to T_j=T_amb+P_dis×θ_JA. If I_total indicates that the transmit power needs to be reduced to suppress interference, P_dis is corrected accordingly and T_j is recalculated. Finally, the thermal throttling warning level is determined by looking up the junction temperature-throttling threshold table with T_j.

[0084] S43: Query the dynamic link carrying status based on the bottleneck location parameters, obtain the real-time available bandwidth and real-time effective throughput of the functional domain where the bottleneck is located, and calculate the predicted value of the link transmission performance under the current load by combining the thermal throttling warning level and the comprehensive impact assessment value, and obtain the dynamic link scheduling decision result.

[0085] In this embodiment, the predicted link transmission performance is a quantitative prediction of the actual achievable transmission performance of the bottleneck functional domain under a candidate scheduling scheme after comprehensively considering available carrying capacity, interference impact, and thermal constraints. It is denoted as η. The larger η is, the higher the performance of the scheme under the conditions of meeting service requirements and multiple constraints.

[0086] Specifically, based on the functional domain identifier in the bottleneck location parameters, the dynamic link bearer status of S35 is queried, and the real-time available bandwidth B_a and real-time effective throughput T_a of the functional domain are read. For each candidate scheduling scheme, based on B_a and T_a, performance reduction is applied according to the S42 thermal throttling warning level: level 0 1.0, level 1 0.95, level 2 0.85. For candidate schemes with a thermal throttling warning level of 3, since they have triggered the throttling protection and must immediately reduce the load, they are directly judged as invalid schemes and excluded from performance comparison. Interference penalties are applied according to the S41 comprehensive impact assessment value I_total, and the η of each scheme is calculated. All candidate schemes are compared, and the scheme with the highest η that does not cause new bottlenecks and does not raise the thermal throttling warning level of any functional domain to level 3 is selected. Its target bearer functional domain, target power and carrier configuration, and service priority ranking are output as the dynamic link scheduling decision result. Candidate scheduling schemes include service migration, power or carrier adjustment, etc.

[0087] Further, in step S43, the predicted link transmission performance value η is calculated, including: S431: Obtain the RF link insertion loss and spectral efficiency loss coefficient of the functional domain where the bottleneck is located, and calculate the calibrated bandwidth B_d and calibrated throughput T_d required for the corresponding service level.

[0088] In this embodiment, the calibrated bandwidth B_d refers to the bandwidth required to meet the carrying requirements of the service level carried by the bottleneck functional domain, after taking into account the functional domain RF link loss; the calibrated throughput T_d refers to the throughput required to be achieved under the same conditions; the spectrum efficiency loss coefficient along the path refers to the proportion of total spectrum efficiency loss caused by the accumulation of insertion loss of devices at all levels and channel degradation during the transmission of service data stream along the functional domain RF link.

[0089] Specifically, S23 / S32 reads the insertion loss L_path of the bottleneck functional domain RF link and the spectral efficiency loss coefficient γ, where γ = the ratio of the current actual spectral efficiency to the ideal spectral efficiency of the functional domain, with a value of 0~1; then, it takes the minimum effective throughput T_min and the maximum tolerable delay jitter J_max from the carrying requirements parameters of the service level carried by the functional domain, and calculates the calibrated throughput and calibrated bandwidth according to T_d=T_min / γ and B_d=T_min / (SE_nom×γ), that is, the link loss is reflected as the additional demand for bandwidth and throughput.

[0090] S432: When B_a≥B_d and T_a≥T_d, the predicted link transmission performance η is calculated using formula (1): η=(B_a×T_a) / (K×B_d×T_d) (1) where K represents a multi-link concurrent interference correction coefficient, and K is determined in real time according to the per-carrier signal-to-interference-plus-noise ratio obtained in step S352 and the inter-band intermodulation interference loss coefficient obtained by back calculation in step S32; when either of the conditions B_a<B_d or T_a<T_d is satisfied, η takes the value of min(B_a / B_d,T_a / T_d) / K.

[0091] In this embodiment, the multi-link concurrent interference correction coefficient K is a correction coefficient that accounts for the transmission performance loss caused by inter-band intermodulation interference when multi-function domains operate concurrently, and is defined as K≥1. A larger K indicates more severe concurrent interference and greater performance loss, and is used to convert the ideal performance of a single link into the actual performance of multi-link concurrent operation.

[0092] Specifically, the real-time determination method of K is: taking the minimum value SINR_min of the per-carrier signal-to-interference-plus-noise ratio of the bottleneck function domain obtained by fine-grained channel sounding in S352 (reflecting the worst carrier) and the inter-band intermodulation interference loss coefficient β obtained by back calculation in S32, calculating according to K=1 / ((1−β)×ζ(SINR_min)), where ζ(SINR_min) is the efficiency retention factor mapped by the per-carrier signal-to-interference-plus-noise ratio. The higher the SINR_min, the closer ζ is to 1, and the lower the SINR_min, the smaller ζ, with a value ranging from 0 to 1. When B_a≥B_d and T_a≥T_d, η is calculated according to formula (1) as η=(B_a×T_a) / (K×B_d×T_d), and a larger η indicates more surplus and better performance; when B_a<B_d or T_a<T_d, η takes the satisfaction ratio of the restricted dimension and then divides by K, that is, η=min(B_a / B_d,T_a / T_d) / K, which reflects the restriction of the bottleneck dimension on performance.

[0093] S433: Taking the ratio of the actual reachable throughput to the minimum effective throughput and the ratio of the maximum tolerable delay jitter to the actual delay jitter as auxiliary verification items for η, performing secondary quantitative verification on η, and calculating the load congestion relief time under the current service according to the verified η.

[0094] In this embodiment, the load congestion relief time refers to the estimated time required for the backlog of data to be transmitted in the bottleneck functional domain to be relieved and the bearing status to return to normal under the current scheduling strategy. The secondary quantization verification is as follows: the ratio of the actual achievable throughput T_real to the minimum effective throughput T_min, c1=T_real / T_min, and the ratio of the maximum tolerable latency jitter J_max to the actual latency jitter J_real, c2=J_max / J_real, are taken as auxiliary verification items. Both should be greater than 1 to meet the requirements. When c1<1 or c2<1, η is reduced according to the verification items to η'=η×min(c1,c2,1), so that η is closer to the actual experience. The prediction of the load congestion relief time t is as follows: Take the current backlog of data Q in the bottleneck functional domain, and estimate it according to t=Q / (T_a−T_occupied), where T_occupied is the currently occupied throughput and (T_a−T_occupied) is the remaining throughput that can be used to relieve the backlog; and weight t with the verified η' to obtain the final predicted load congestion relief time: the final predicted load congestion relief time t'=t / η', that is, the larger η' is, the better the performance and the shorter the relief time.

[0095] S5: Based on the dynamic link scheduling decision results, generate traffic scheduling and load balancing strategies by functional domain and service level and trigger their execution.

[0096] In this embodiment, the scheduling control, based on the dynamic link scheduling decision results, issues configuration commands through the control interfaces driven by each functional domain, and coordinates the carrier aggregation and power configuration of the 5G access domain, the RU allocation and rate set of the Wi-Fi access domain, and the queue scheduling weight of the wired interface domain, thereby generating traffic scheduling and load balancing strategies by functional domain and service level: for high-priority and latency-sensitive services such as cloud gaming and VoLTE, functional domains with sufficient bearing capacity and low interference are prioritized for allocation; for services such as downloading and 4K video with high throughput requirements but low latency, the load is distributed among the multi-functional domains according to the load balancing principle.

[0097] Furthermore, after generating traffic scheduling and load balancing strategies by functional domain and service level, and when the deviation between the actual transmission performance of the service data flow after scheduling and the carrying requirement parameters exceeds a preset performance deviation threshold, the link scheduling strategy is iteratively corrected, and the occupied link resources are reclaimed when the service data flow is detected to have ended.

[0098] Specifically, after scheduling is executed, the monitoring hook function continues to collect the actual transmission performance of each service data stream and compare it with its carrying requirement parameters to obtain the deviation value between the actual transmission performance and the requirement. When the deviation value exceeds the preset performance deviation threshold (e.g., 10% of the requirement value), it is determined that the scheduling strategy and the actual carrying status are mismatched, triggering the iterative correction of the link scheduling strategy, and returning to step S3 to re-perform the carrying capacity correction and coupling analysis. When the monitoring hook function detects a connection release message of a certain service data stream or no data transmission within a preset silent period (e.g., 30s), it determines that the service data stream has ended, and then issues a resource reclamation command to the corresponding functional domain to release the carrier, RU and queue resources it occupies, so that the resources are not ineffectively occupied, and overall, it continuously provides transmission guarantee that matches the needs of concurrent multi-channel services.

[0099] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0100] In one embodiment, a multi-link data flow collaborative scheduling system for a 5G CPE device is provided, which corresponds to the multi-link data flow collaborative scheduling method for a 5G CPE device described in the above embodiment.

[0101] A multi-link data stream collaborative scheduling system for 5G CPE devices includes a monitoring and acquisition module, an identification and extraction module, a bearer status correction module, a coupling analysis module, and a scheduling control module. Detailed descriptions of each functional module are as follows: The monitoring and acquisition module is used to inject monitoring hook functions into the data interaction interfaces of the 5G access domain, Wi-Fi access domain and wired interface domain of the 5G CPE device to intercept the service data stream transmitted through each functional domain and collect the real-time status data of the communication module to which each functional domain belongs. The identification and extraction module is used to identify the intercepted business data streams according to the preset hierarchical matching rules, determine the carrying requirement parameters, and extract the corresponding differentiated monitoring factors from each functional domain. The bearer status correction module is used to input real-time status data and differentiated monitoring factors into the pre-established link interference propagation model, thermal-power constraint model and domain-specific link capability benchmark, back-calculate the current spectral efficiency loss coefficient and inter-band intermodulation interference loss coefficient of each functional domain, and combine the real-time power budget margin to correct the domain-specific link capability benchmark online, so as to obtain the dynamic link bearer status that characterizes the current available bearer capability of each functional domain. The coupling analysis module is used to perform coupling analysis between the load demand parameters, differential monitoring factors and dynamic link load status in the link interference propagation model and the thermal-power constraint model, locate the functional domain of the bottleneck load and the bottleneck location parameters, and obtain the dynamic link scheduling decision results. The scheduling control module is used to generate and execute traffic scheduling and load balancing strategies by functional domain and service level based on the dynamic link scheduling decision results.

[0102] For specific limitations regarding the multi-link data flow collaborative scheduling system for 5G CPE devices, please refer to the limitations of a multi-link data flow collaborative scheduling method for 5G CPE devices mentioned above, which will not be repeated here. Each module in the aforementioned multi-link data flow collaborative scheduling system for 5G CPE devices can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the computer device in hardware form, or it can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0103] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores differential monitoring factors, dynamic link scheduling decision results, etc. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a multi-link data flow collaborative scheduling method for a 5GCPE device.

[0104] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the multi-link data stream collaborative scheduling method for a 5G CPE device as described above.

[0105] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the multi-link data stream collaborative scheduling method for a 5G CPE device as described above.

[0106] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0107] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0108] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multi-link data stream collaborative scheduling method for a 5G CPE device, characterized in that, include: Inject monitoring hook functions into the data interaction interfaces of the 5G access domain, Wi-Fi access domain, and wired interface domain of the 5G CPE device to intercept the service data streams transmitted through each functional domain and collect the real-time status data of the communication modules to which each functional domain belongs. Identify the intercepted business data streams according to the preset hierarchical matching rules, determine the carrying requirement parameters, and extract the corresponding differentiated monitoring factors from each functional domain; The real-time status data and the differentiated monitoring factors are input into the pre-established link interference propagation model, thermal-power constraint model and domain link capability benchmark. The current spectral efficiency loss coefficient and inter-band intermodulation interference loss coefficient of each functional domain are calculated in reverse. The domain link capability benchmark is then corrected online in combination with the real-time power budget margin to obtain the dynamic link carrying status that characterizes the current available carrying capacity of each functional domain. The load requirement parameters, differential monitoring factors, and dynamic link load status are coupled and analyzed between the link interference propagation model and the thermal-power constraint model to locate the functional domain and bottleneck location parameters of the bottleneck load, and obtain the dynamic link scheduling decision results. Based on the dynamic link scheduling decision results, traffic scheduling and load balancing strategies are generated and triggered for execution by functional domain and service level.

2. The multi-link data stream collaborative scheduling method for a 5G CPE device according to claim 1, characterized in that, The process involves identifying the intercepted service data streams according to preset hierarchical matching rules, determining the carrying requirement parameters, and extracting corresponding differentiated monitoring factors from each functional domain, including: In 5G CPE devices, hierarchical matching rules are pre-configured. These rules match service data streams based on service type, functional domain, and QoS level, and support service subclass-level matching through service identifier masks. The carrying requirements of the service data streams matched by the hierarchy are identified, and the minimum effective throughput and the maximum tolerable latency jitter are determined as the carrying requirements parameters. For the 5G access domain, the signal-to-interference-plus-noise ratio and SA / NSA operating status of each member carrier are extracted; for the Wi-Fi access domain, the channel occupancy rate and OFDMA resource unit occupancy ratio are extracted; and for the wired interface domain, the port negotiation rate and VoLTE voice jitter are extracted as differentiated monitoring factors corresponding to each functional domain. Establish correlation monitoring relationships between functional domains based on differentiated monitoring factors of each functional domain, maintain real-time synchronization between the business feature library and external network service deployment, and update the business identifier mask in the hierarchical matching rules based on the change information of the business feature library.

3. The multi-link data stream collaborative scheduling method for a 5G CPE device according to claim 1, characterized in that, The process of determining the dynamic link bearer status includes: Pre-parse the link configuration files of 5G CPE devices to establish domain-specific link capability benchmarks; Substitute the real-time signal quality and throughput data of key nodes in the multi-link system into the domain-specific link capability benchmark to back-calculate the current spectral efficiency loss coefficient of each functional domain. Substitute the decrease in real-time signal quality under concurrent operation in each functional domain into the link interference propagation model to back-calculate the current inter-band intermodulation interference loss coefficient of each functional domain. The real-time power budget margin is determined based on the real-time voltage data of each power rail, and the available power budget in the domain link capability benchmark is dynamically adjusted. Based on the thermal-power constraint model, the gain attenuation of the power amplifier as the junction temperature increases and the change in insertion loss caused by the switching of the front-end module path are used as correction terms and added to the domain link capability benchmark. Based on the back-calculated loss coefficient, the adjusted available power budget, and the correction term, the domain-specific link capacity benchmark is corrected online to obtain the dynamic link carrying status that characterizes the current available carrying capacity of each functional domain.

4. The multi-link data stream collaborative scheduling method for a 5G CPE device according to claim 1, characterized in that, The process of determining the dynamic link scheduling decision result of the 5G CPE device includes: Based on the bottleneck location parameters, the spectral interference propagation path of the functional domain where the bottleneck is located to adjacent frequency bands and adjacent functional domains is extracted from the link interference propagation model, and the comprehensive impact assessment value of the functional domain where the bottleneck is located on the link quality of adjacent functional domains is obtained. Based on the bottleneck location parameters and comprehensive impact assessment values, the junction temperature and power consumption status data of key components in the bottleneck module are extracted from the thermal-power constraint model to obtain the thermal throttling warning level under the current load. Based on the bottleneck location parameters, the dynamic link carrying status is queried to obtain the real-time available bandwidth and real-time effective throughput of the functional domain where the bottleneck is located. Combined with the thermal throttling warning level and the comprehensive impact assessment value, the predicted value of the link transmission performance under the current load is calculated to obtain the dynamic link scheduling decision result.

5. A multi-link data stream collaborative scheduling method for a 5G CPE device according to claim 1, characterized in that, The method also includes: when the deviation between the actual transmission performance of the service data flow after scheduling and the bearer requirement parameters exceeds a preset performance deviation threshold, triggering iterative correction of the link scheduling strategy, including: The actual transmission performance of the scheduled service data stream is evaluated based on multi-dimensional transmission indicators, and the performance deviation between the actual transmission performance and the carrying requirement parameters is determined. When the performance deviation value is greater than the preset performance deviation threshold, fine-grained channel probing is enabled for the bottleneck functional domain to obtain the carrier-by-carrier signal-to-interference-plus-noise ratio, and the domain-specific link capability benchmark is corrected at the first level to obtain the first-level correction scheduling strategy. When the difference in performance deviation before and after the first-level correction is less than the preset deviation change threshold, the equivalent junction temperature of the bottleneck module and the power amplifier gain attenuation are calculated based on the measured throughput and power consumption before and after the business data flow scheduling. The second-level correction scheduling strategy is then obtained by performing a second-level correction on the domain link capability benchmark. When the difference in performance deviation before and after the second-level correction is less than the preset deviation change threshold, the voltage drop and actual power budget margin of the corresponding power rail are calculated based on the measured throughput and equivalent junction temperature before and after the second-level correction. The third-level correction is then applied to the domain link capability benchmark to obtain the third-level correction scheduling strategy, and traffic scheduling and load balancing strategies for functional domains and service levels are regenerated.

6. A multi-link data stream collaborative scheduling method for a 5G CPE device according to claim 5, characterized in that, The preset deviation change threshold is determined by the following method: Obtain the key business proportion of the business data stream and the current idle bearer capacity proportion of the 5G CPE device; When the proportion of key business is greater than or equal to the preset proportion of key business, the calculation formula for the preset deviation change threshold is: preset deviation change threshold = (1 - proportion of key business) × benchmark deviation change value; When the proportion of key business is less than the preset proportion of key business, and the proportion of idle capacity is greater than or equal to the preset proportion of idle capacity, the calculation formula for the preset deviation change threshold is: preset deviation change threshold = (1 + idle capacity proportion) × baseline deviation change value.

7. A multi-link data stream collaborative scheduling system for 5G CPE devices, characterized in that, include: The monitoring and acquisition module is used to inject monitoring hook functions into the data interaction interfaces of the 5G access domain, Wi-Fi access domain and wired interface domain of the 5G CPE device to intercept the service data stream transmitted through each functional domain and collect the real-time status data of the communication module to which each functional domain belongs. The identification and extraction module is used to identify the intercepted business data streams according to the preset hierarchical matching rules, determine the carrying requirement parameters, and extract the corresponding differentiated monitoring factors from each functional domain. The bearer status correction module is used to input the real-time status data and the differentiated monitoring factors into the pre-established link interference propagation model, thermal-power constraint model and domain link capability benchmark, back-calculate the current spectral efficiency loss coefficient and inter-band intermodulation interference loss coefficient of each functional domain, and combine the real-time power budget margin to correct the domain link capability benchmark online, so as to obtain the dynamic link bearer status that characterizes the current available bearer capability of each functional domain. The coupling analysis module is used to perform coupling analysis between the load requirement parameters, differential monitoring factors and dynamic link load status in the link interference propagation model and the thermal-power constraint model, locate the functional domain of the bottleneck load and the bottleneck location parameters, and obtain the dynamic link scheduling decision results. The scheduling control module is used to generate and execute traffic scheduling and load balancing strategies based on the dynamic link scheduling decision results, according to functional domains and service levels.

8. A 5G CPE device, characterized in that, The device includes a 5G communication module, a Wi-Fi communication module, a wired interface module, a clock module, a hierarchical power supply module, and a processor and a memory connected to each module. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the multi-link data stream collaborative scheduling method for a 5G CPE device as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-link data stream collaborative scheduling method for a 5G CPE device as described in any one of claims 1 to 6.

Images