5G message sending failure retransmission optimization method and system

Abstract

The invention provides a 5G message sending failure retransmission optimization method and system, and the method comprises the steps: constructing a multi-dimensional failure reason label for a message which fails to be sent, and enabling the label to comprise a network state dimension, a user state dimension and a base station load dimension; calculating a dynamic weight based on the multi-dimensional failure reason label to obtain a failure feature vector for the message; according to the failure feature vector, adaptively selecting a retransmission strategy, including delay retransmission, fragmentation retransmission, path switching retransmission and cross-protocol degradation retransmission; a failure heat index (FHI) is used in the retransmission process, and the FHI is obtained through calculation of the failure frequency, the failure concentration ratio and the network area failure density; performing priority scheduling on the message according to the failure popularity index, and determining whether the message is immediately retransmitted or queued in a delayed manner; a 5G message is split, a progressive retransmission mechanism is adopted, a message digest and key metadata are preferentially sent, and then complete content is gradually retransmitted after network conditions are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of 5G communication, in particular to a 5G message sending failure retransmission optimization method and system. BACKGROUND

[0002] With the popularity of 5G network and the development of rich media message service, the communication between user terminals has been expanded from traditional short text short message SMS to graphic text, voice and video multi-modal information transmission. Compared with traditional short message, 5G message relies on multi-protocol stack cooperation in the transmission layer, including HTTP / 2, SIP and RCS Session, and has the characteristics of high bandwidth, high latency sensitivity and dynamic link dependence. However, in complex wireless environment, message sending failure still exists, especially in network fluctuation, base station overload or user state abnormality including weak signal and mobile switching, the message may appear multiple retransmission failure or delayed delivery.

[0003] Traditional message retransmission mechanism mostly adopts fixed retransmission interval or simple acknowledgement and response mechanism (ACK / NACK), which lacks multi-dimensional recognition of failure causes and dynamic adaptive ability. When the network state is complex and the failure mode is diverse, such mechanism often leads to repeated retransmission, bandwidth waste and message congestion, thereby reducing the overall delivery rate and user experience. Although some systems introduce delayed retransmission or backup path sending mechanism, it is still difficult to realize the global optimization of message content, network characteristics and protocol channel.

[0004] The present application proposes a 5G message sending failure retransmission optimization method, which constructs a failure cause label system, quantitatively generates a failure feature vector, and realizes message priority scheduling combined with failure heat index (FHI), so as to adaptively select the optimal retransmission strategy under different network states. The present application further introduces a gradual fragmentation retransmission mechanism and a cross-protocol dynamic degradation mechanism, which transmits message content in layers according to importance, preferentially sends abstract and key metadata in weak network environment, ensures that core information is delivered first, and simultaneously intelligently switches between RCS, IMS, SMS and other protocols according to real-time network evaluation results, ensuring that the message can still be stably delivered through the degraded path when the protocol link is partially disabled. The present application can significantly improve the delivery rate and retransmission success rate of 5G rich media messages without increasing additional signaling burden, realizing the evolution from "single-dimensional retransmission" to "adaptive intelligent retransmission", and providing a reliable, flexible and intelligent transmission guarantee scheme for 5G message distribution system. SUMMARY

[0005] The present application aims to provide a 5G message sending failure retransmission optimization method and system.

[0006] The present application aims to solve the following technical problems existing in the existing 5G message sending process: first, the single failure cause identification dimension lacks targeted retransmission strategy, and the existing message sending system relies on simple ACK / NACK or timeout mechanism to judge whether to retransmit, which cannot identify the specific failure type, such as different failure scenarios caused by network fluctuation, high base station load or abnormal user state, so that the retransmission strategy cannot be accurately matched, resulting in repeated retransmission or resource waste; second, the static and rigid retransmission mechanism lacks dynamic adaptability, and the traditional system usually adopts fixed interval retransmission or preset path retransmission, which cannot dynamically adjust the retransmission window, fragmentation strategy or delay strategy according to real-time network characteristics and historical failure mode, resulting in that the message may still fail multiple times in weak network or congested environment, reducing the overall delivery rate; third, the message content transmission lacks hierarchical design, resulting in delayed delivery of important information, in the network unstable scene, the transmission of complete rich media message (including pictures, videos, files, etc.) is prone to failure, and the existing mechanism does not distinguish the importance of message content, which cannot realize the priority transmission of abstract, metadata and other key information, causing user experience delay or information loss, therefore, the present application realizes intelligent retransmission and cross-protocol dynamic degradation of 5G message through multi-dimensional failure feature recognition and adaptive retransmission algorithm, thereby significantly improving the message delivery rate and transmission efficiency.

[0007] A 5G message sending failure retransmission optimization method, comprising: S1: constructing a multi-dimensional failure cause label for the 5G message that fails to send, the label comprising network state dimension, user state dimension, and base station load dimension; S2: calculating a dynamic weight based on the multi-dimensional failure cause label to obtain a failure feature vector for the 5G message; S3: adaptively selecting a retransmission strategy according to the failure feature vector, including delay retransmission, fragmentation retransmission, path switching retransmission, and cross-protocol degradation retransmission; S4: using a failure heat index FHI in the retransmission process, the index being calculated from failure frequency, failure concentration, and network area failure density; S5: performing priority scheduling on the message according to the failure heat index to determine whether to retransmit immediately or delay queuing; S6: splitting the 5G message and using a progressive retransmission mechanism to preferentially send message abstract and key metadata, and then gradually supplement the complete content after the network condition improves.

[0008] A 5G message sending failure retransmission optimization system realized based on a 5G message sending failure retransmission optimization method, comprising: Failure analysis and multi-dimensional label module: receive message sending feedback information, including success and failure, network log, extract and generate multi-dimensional failure cause label, including signal strength, packet loss rate, network type in network state dimension, user online state, device type in user state dimension, current load, congestion situation in base station load dimension, provide original data basis for subsequent strategy selection; Failure feature vector generation module: quantize and weight the multi-dimensional failure label, generate failure feature vector using dynamic weight formula, reflect the failure characteristics of each message in different dimensions, convert the original label into a numerical vector for algorithm decision; Retransmission strategy selection module: adaptively select the optimal retransmission strategy according to the failure feature vector, the strategy includes fragmentation retransmission PFRT, delay retransmission, path switching retransmission, cross-protocol dynamic degradation retransmission CPADR, form a specific retransmission scheme for each message; Failure heat index module: calculate the failure heat index according to the failure frequency, failure concentration and network area failure density, which is used to measure network congestion and failure concentration area, and provide indicators for message priority sorting and retransmission rhythm; Priority scheduling module: determine the message retransmission order according to FHI and user historical behavior prediction, decide to retransmit immediately or delay queuing, reasonably allocate retransmission order in resource-limited and high failure environment, and improve core content delivery rate; Gradual retransmission module: split the message into abstract, key metadata and complete content, preferentially send abstract or key fragments, and gradually supplement complete content after network conditions improve, combined with the fragmentation strategy and protocol selection output by the retransmission strategy selection module.

[0009] Further, the step S2 based on the failure cause label obtains a failure feature vector for the 5G message, including: S21: Network state dimension construction, collect network real-time state information from message sending path, including packet loss rate , average delay , bandwidth occupation and network slice availability , after normalizing each index, calculate the network state score: , wherein , , , is an adjustable coefficient, which is optimized according to historical failure log; S22: User state dimension construction, collect user online state, terminal type and reliability, user activity and historical message receiving success rate, quantize each index, and calculate user state score: , wherein representing the user online probability, representing the terminal reliability score, representing the user recent activity, representing the user historical receiving success rate, 、 、 、 is an adjustable weight coefficient; S23: Base station load dimension construction, obtain the current connection number of the base station where the message is sent , base station failure rate and regional failure density information, the specific calculation method of the base station load score is: , wherein is the maximum number of connections that the base station can bear, is an adjustable weight coefficient; S24: Multi-dimensional label vector generation and fusion, integrating three-dimensional scores into a multi-dimensional failure cause label vector of the message: , add a dynamic weight to each dimension to form the final failure feature vector: ; S25: Update each 5G message result of the sending failure to the corresponding dimension index, including network packet loss rate, user online probability and base station load state, regularly analyze failure logs, and adjust the weight through the way of minimizing prediction error and reinforcement learning ; S26: The failure feature vector is directly used as the input of the retransmission strategy selection module, providing the basis for subsequent priority scheduling and progressive retransmission.

[0010] The multi-dimensional failure cause label is not only composed of numerical features, but also uses a hierarchical semantic mapping structure: Base layer tag (Base Tag): Physical indicators directly extracted from raw data, such as packet loss rate, delay, online rate, load value, etc. Fusion layer tag (Fusion Tag): Convert the base layer tag into a semantic tag through clustering or rule engine, such as "high delay", "area congestion", "terminal high risk", etc. Behavior layer tag (Behavior Tag): Dynamic tags generated by combining historical behavior data, such as "periodic failure user", "instant overload base station", "short-time high failure area".

[0011] When the system generates failure labels, a sliding time window The message sending failure data in a continuous time period is aggregated and analyzed, when a dimension index abnormally fluctuates in a window, the system will automatically increase the weight of the dimension label, when the dimension performance recovers to be stable, a weight attenuation mechanism is triggered, and the mathematical expression is wherein is the ith dimension weight, including network, user, base station, is the change rate of the dimension index in the current window, is the change threshold, is a weight adjustment factor, the dynamic migration mechanism enables the label system to continuously reflect the real changes of the current network environment, rather than static feature description.

[0012] In order to prevent the same geographical area or the same base station from forming a "failure hotspot" in a short time, the system further introduces a spatial clustering algorithm, which identifies potential "fault clusters" by modeling the geographical location attribute of the failure label, and the clustering algorithm is , represents the geographical distance between two failure records, is a spatial distance threshold, is a label similarity threshold, if the number of failure labels in a cluster exceeds the threshold, a new "regional failure label (Regional Failure Tag)" will be generated, and the base station load dimension weight in the region will be dynamically adjusted, and the mechanism enables the label system to have spatial awareness, and prevent concentrated retransmission avalanches at the macro level.

[0013] Further, a label evolution mechanism is used to periodically (such as every hour) retrain and optimize the label feature vector based on failure samples and retransmission success samples: When a certain type of label is associated with a high failure rate for a long time, the risk weight of the label is automatically increased; When a certain label is associated with successful retransmission for a long time, the risk weight is reduced; For new unknown labels (such as new terminal models, temporary regional interference), the system will first assign a low confidence label, and gradually correct it in subsequent learning.

[0014] Further, the failure heat index is used in the retransmission process in step S4, including: S31: Multidimensional failure data collection, automatically collect multidimensional data related to the failure event after the message sending failure, the data includes: network layer parameters, specifically containing signal strength, delay, packet loss rate; user layer state, specifically containing terminal online rate, network switching frequency, moving speed; base station layer parameters, specifically containing base station load, congestion degree, switching frequency; message layer characteristics, specifically containing message type, message length, retransmission times, real-time requirements; geographical layer information, specifically containing failure occurrence position coordinates, area identifier, failure concentration degree; the data is divided into time windows with fixed window size Sampling is performed to form a time-series multidimensional sample set ; S32: Failure density matrix construction, divide the spatial coordinate area into a fixed grid, and count the number of failure events of each grid in each time window, and use Gaussian kernel function to smooth the local area to eliminate the discreteness of the data, and obtain the failure density matrix , which is defined as, Wherein, x, y are the horizontal and vertical coordinates of the geographical position, t is the time, is the geographical coordinates when the ith failure event occurs, is the failure event index set in the current time window , used to limit the time range, is the time stamp when the ith failure event occurs, is a spatial distance function, representing the distance between point and evaluation point , is the geographical position of , The specific expression of is , is a kernel function, which is used to convert spatial distance into weight; S33: Dimension weight dynamic calculation, the system calculates the information entropy value of each dimension Reflecting the contribution of each dimension to the overall failure trend, and generating a weight vector , wherein is the network state dimension, is the user state dimension, is the base station load dimension, is the message characteristic dimension, is the geographical distribution dimension, and the dimension calculation method is, , is the real-time information entropy value of the ith dimension, The sum of information entropy of all dimensions, the higher the information entropy, the greater the volatility and uncertainty of the dimension, the stronger the explanation of the system failure mode, and therefore the higher the weight. The system dynamically adjusts the weight according to real-time data to ensure that the model is self-adaptive to environmental changes; S34: Local failure feature vector generation, after obtaining the weight vector, the system extracts the local failure features in the current time window to form a vector , where each component represents the failure rate and statistical indicators of the corresponding dimension, including base station failure rate, message timeout rate, and regional density. This vector is the "failure feature fingerprint" under the current network state; S35: Time decay factor calculation, the influence of historical failure events will decrease over time, and a time decay factor a(t) is used, where the specific calculation method of a(t) is, , where λ is the decay coefficient, used to control the contribution of past data to the current FHI. By introducing a(t), it is avoided that old data causes excessive interference to the current decision; S36: Geographical aggregation correction, the spatial statistical method is used to evaluate the aggregation of failure events, and the Ripley's K function is used to calculate the aggregation coefficient , where The specific calculation method is, , where K(d) is the spatial distribution function of actual failure events, representing the average number of other failure event points within a radius of d around any failure event point. After area normalization, the spatial aggregation function is obtained, and the specific calculation method is , where A is the observation area, N is the total number of failure events, is the distance between the i th and j th failure event points, is the indicator function, which takes 1 when , otherwise it takes 0, where is the reference function of theoretical random distribution, representing the expected average number of events within a radius of d under completely random distribution. It is used as a control baseline to help the system judge whether the current failure distribution is abnormally aggregated, and its calculation method is , when >0, it indicates that the failure events in a certain area show a significant aggregation trend, and the system will increase the priority of message retransmission in this area; S37: The system integrates the weight of each dimension, local features, time decay, and geographical aggregation to calculate the failure heat index FHI: wherein κ is the aggregation amplification factor, used to adjust the influence of the geographical aggregation effect, the calculated FHI is normalized, with a range of [0, 1], divided into five levels, 0 to 0.2 is normal, 0.2 to 0.4 is slight failure, 0.4 to 0.6 is moderate failure, 0.6 to 0.8 is high-risk failure, and 0.8 or more is serious failure, the scheduling module automatically determines the retransmission strategy according to the FHI level, FHI low, merge queue, delay retransmission; FHI medium, retransmit normally; FHI high, enable path switching, cross-protocol retransmission; FHI extremely high, suspend batch message sending and trigger network diagnosis; the value of FHI reflects the comprehensive risk level of the failure event, when the value of FHI is high, it means that the failure probability of the current region or user group is on the rise, network resources should be allocated preferentially or the retransmission path should be adjusted; S38: Finally, the final priority score of a single message m FHI is determined by the FHI and the message's own failure feature vector jointly determined: , is the region to which the message m belongs, w is a linear weight vector of the message features to the score, is the number of seconds that the message has been waiting for retransmission, is a normalization function, , , are the coefficients.

[0015] The specific calculation method of the information entropy value of each dimension is as follows: Discretization of dimension features and construction of probability distribution: based on the collected failure-related parameters, the indicators of this dimension are segmented and clustered to form a set of discrete states: Then the frequency of failure events in each state is counted to form a probability distribution: ; Noise smoothing calculation, the specific calculation method is , is the minimum noise smoothing coefficient, which is automatically calculated according to the data volatility of the recent time window , is the standard deviation of the dimension feature, is a system adjustment coefficient, with a value range of 0.01 to 0.05; Dimension information entropy calculation When the state distribution of a certain dimension is extremely single, it means that the entropy is low, indicating that this dimension contributes weakly to the failure, when the state distribution of a certain dimension is very scattered and unpredictable, it means that the entropy is high, indicating that this dimension is a major failure factor; Multi-window fusion, , is the current window entropy, is the previous window entropy, For the first two windows entropy, For the window weight; Entropy value normalization, , the normalized entropy can be directly used to calculate the weight: .

[0016] Time complexity: The calculation of FHI is mainly the summation of the failure events in the window and the time decay. The sliding window and incremental update can be used (the incremental update can reduce the processing of each new event to O(1)), which is suitable for high concurrency scenarios; Space: The event summary (count, time statistics, weighted sum) of each region needs to be maintained, rather than storing all original events, saving storage; Robustness: The fallback value and confidence level are designed for missing indicators (such as no slice state at a certain time), so that FHI can still work stably under incomplete information.

[0017] If the FHI of a base station exceeds The system puts the messages in the delay queue that are in the area of the base station and have a Priority(m) lower than a certain threshold, and slowly releases them from low to high according to the multi-scale FHI, avoiding a large number of retransmissions at the moment; for FHI that is low but the Priority(m) of a single message is high (such as OTP), immediate cross-protocol parallel retransmission is allowed to ensure real-time performance; after the FHI drops significantly (lower than ), the system quickly retransmits the delayed summaries and remaining slices (progressive retransmission) in the priority queue.

[0018] When it is detected that the FHI abnormally increases sharply in a short time (for example, the growth rate in a short time window exceeds the threshold ), the system will trigger the "abnormal alarm mode": All unfinished retransmission requests in the area are temporarily frozen and the network side is requested to intervene (such as switching back to routing or notifying the operator); Increase the FHI monitoring of adjacent areas to determine whether it is a horizontal spread.

[0019] The failure heat index realizes the dynamic evaluation of the message sending failure risk by fusing time, space, network, user and message multi-dimensional features: Realize the dynamic retransmission priority decision: avoid unnecessary repeated retransmission, reduce network congestion, and improve overall bandwidth utilization; Regional level prediction ability: identify potential "network failure hot spot areas" in advance to avoid avalanche sending failure; Fusion of time decay and historical weight mechanism: enhance the real-time responsiveness of the model, so that the system can "respond immediately and converge quickly", and improve the adaptive ability; Cross-dimension cause diagnosis: "failure tracing" within the system, facilitating intelligent operation and maintenance or policy optimization, forming a failure model with causal explainability.

[0020] FHI value as a unified risk indicator of the system, directly driving the sending and scheduling module: When FHI is low, the system delays retransmission or merges the queue; When FHI is high, the system prioritizes retransmission or path switching; When FHI is extremely high, the system automatically suspends batch tasks and triggers network diagnosis.

[0021] Further, the step S6 adopts a progressive retransmission mechanism, including: S41: Sharding the to-be-retransmitted message M to form a shard set and calculating the semantic importance value for each shard , high importance shards include message digest, title, main image, and structural metadata; S42: Calculate the retransmission priority score of each shard , considering shard importance, network conditions, and regional heat, wherein, is the network failure rate standardized value, , is the failure heat index, and shards with higher priority will be scheduled first in the retransmission window; S43: Based on the current network available bandwidth and the average shard size , calculate the progressive sending window size, i.e., the number of shards allowed to be sent simultaneously: wherein is the window compression coefficient, which automatically compresses the window as FHI increases, avoiding secondary congestion in high failure heat areas, is the minimum number of shards allowed in the window, is the maximum number of shards allowed in the window, calculated by to clip the theoretical number of sendable shards to the range of , ] and multiply by , further reducing the window when the failure heat is high, achieving adaptive progressive sending; S44: After sorting by priority score, select the shard set within the window from high to low , and calculate the adaptive redundancy ratio, wherein is the redundancy growth rate, is the base redundancy ratio, is the shard semantic importance, and the redundancy ratio Failure heat varies with fragment importance; S45: Perform redundancy coding on fragments in the window, generate extra check fragments to form an error-correctable progressive sending group; S46: Perform first sending, the receiving end can display the summary part and part of the visual content of the message after receiving part of the high-priority fragments, to realize the "first available and then complete" progressive experience; S47: Monitor the ACK / NACK messages fed back by the receiving end and calculate the local loss rate and delay indicators; S48: According to the feedback and FHI changes, perform hierarchical retransmission decision, when , retransmit all NACK fragments immediately; when , only retransmit the first k fragments of priority, where , delay the retransmission of low-priority fragments and only perform cross-protocol concurrent sending on high-importance fragments.

[0022] The metadata of each fragment includes: frag_id: integer; importance: , summary ≈ 1 size: bytes; state ; send_count: number of times sent; last_send_ts: timestamp; hash: integrity check; p_score: priority score, calculated in real time.

[0023] Assume that the system performs retransmission on a 20KB 5G rich media message: The message is split into 10 fragments, with importance values of [1.0, 0.9, 0.8, …, 0.1]; The current network bandwidth estimate = 8KB, the average fragment size = 2KB, FHI = 0.6; Then the window size ; The system only selects the top three fragments (summary, title, thumbnail) and adds 20% redundancy; After sending, the receiving end can immediately show the core content; The remaining fragments will automatically trigger the reissue after the FHI drops. ​

[0024] Complexity: Most operations are sorting and window selection (O(k log k)), feedback processing can be batched, FHI and MDFRL are incremental updates (constant time / event), making it suitable for high-concurrency scenarios; Robustness: For missing data (without FHI or network estimation), a conservative default is adopted (FHI is set to 0.5, loss_est is set to the historical median), and the missing data is compensated for by the Learning Module; Security: Each shard contains a verification hash, and important shards can be signed to prevent tampering.

[0025] In this embodiment, before performing the fragmentation operation, the system first performs content structure analysis on the original 5G message, dividing the message into "core content fragments" and "auxiliary content fragments". The core fragments include the title, event summary, first image or key information fields, while the auxiliary fragments include secondary images, long text or multimedia additional content. When the network quality is poor or the failure rate index is high, the system prioritizes retransmitting only the core fragments, so that the receiving end can recover the main semantic content of the message in a short time. When the network condition returns to normal, the auxiliary fragments are gradually retransmitted, realizing a progressive message reconstruction of "information usability first, content completeness later".

[0026] FHI also serves as a feedback update factor, participating in the dynamic scheduling of the retransmission phase. When a fragment fails consecutively during retransmission, the system reports the failure event corresponding to that fragment to the FHI module, correcting the failure heat value of the corresponding region in real time. Conversely, when a fragment retransmits successfully, the FHI value will decay according to the weight factor. Through this two-way linkage, the system can achieve "instant correction" of FHI at the fragment level, making the failure heat distribution closer to the real-time network state and avoiding window control imbalance caused by global lag.

[0027] Furthermore, step S3 uses cross-protocol downgrade retransmission, including: S51: Obtain failure information and evaluation conditions; receive the list of fragments that failed to be transmitted by the progressive fragmentation retransmission module. Obtain the failure heat index FHI value H and the multi-dimensional failure feature vector F to determine whether the cross-protocol degradation conditions are met. and Any one of the conditions in, where, For fragment weights, Based on historical success rate, Set a threshold for the system; S52: Protocol Availability and Weight Calculation, defining the set of currently available protocols. Includes RCS, IP, SMS, and MMS for each fragment. In each protocol Compute the weight, , For the importance of the shard, For the shard In the protocol The historical success rate of H is the failure heat index, The protocol delay normalization value is, The system adjustable weight coefficient is; S53: Dynamic protocol selection, for each shard Select the protocol with the highest weight For high-priority shards, include summaries and key metadata, and preferentially select protocols with high weights and low delays for transmission; S54: Probability-driven redundancy scheduling, calculate the redundancy transmission probability for critical shards Repeat the transmission of critical shards on the selected protocol with a probability To enhance reliability in high failure heat environments, non-critical shards are sent with low redundancy according to system strategy, and delayed transmission; S55: Transmission and feedback iteration, send messages according to shard priority and protocol selection results, receive feedback ACK and NACK, and update shard state For failed shards, update the protocol historical success rate Recalculate the weight Select the next backup protocol to form a loop iteration until the shard is successful, reaching the upper limit of retransmission.

[0028] The system first receives input from the progressive shard retransmission module (PFRT), including the current set of shards to be retransmitted and their transmission state information; At the same time, the system extracts the feedback results of the last transmission from the failure analysis module, including network delay, packet loss rate, signal strength, base station load, user online state and other multi-dimensional feature parameters; The system generates a failure feature vector based on these features, and simultaneously calls the FHI (Failure Heat Index) module to calculate the current failure heat index; If the heat index exceeds the preset threshold or consecutive feedback failures are found, trigger the cross-protocol dynamic degradation process.

[0029] The system evaluates the applicability of each protocol for each shard in the current environment, and the indicators include: Historical transmission success rate (learned from logs); Current network delay and bandwidth utilization; Protocol load capacity and compatibility; User terminal support.

[0030] The system selects the optimal protocol for each message shard according to the protocol priority results obtained in the previous stage. During the selection process, the importance of the shard is considered, including: If the shard contains critical information (such as message digest, title, alert content, etc.), a protocol with low delay and high stability is preferred; If the shard belongs to non-critical content (such as pictures, attachments or supplementary description), a protocol with higher throughput but relatively lower reliability can be selected for transmission; In addition, to avoid the additional consumption caused by frequent switching, the system also judges the protocol switching cost, such as signaling reconstruction time, encoding conversion cost, etc., and only performs cross-protocol degradation when necessary.

[0031] The system dynamically calculates the redundancy transmission probability of the shard according to its importance and historical success rate. For important shards, the system may send them simultaneously on two different protocols to prevent information loss due to the failure of a single channel. The core idea of redundant scheduling is "limited redundancy and precise redundancy": the system does not blindly repeat the transmission, but dynamically judges whether to send redundantly based on real-time network status. When the network is congested or resources are limited, the system can automatically reduce the redundancy strength to prevent new transmission pressure.

[0032] The sending end listens to the confirmation information (ACK) or failure reply (NACK) from each protocol channel and dynamically updates the internal state table according to the results. For successful shards, the system marks them as completed. For failed shards, the system records the failure reason of the current protocol (such as network timeout, terminal unresponsive, format not supported, etc.) and adjusts the historical success rate parameter of the protocol. Then, the system re-evaluates the protocol weight and selects the next alternative protocol for retransmission. If multiple attempts still fail, the system delays the shard according to the priority and tries again after the network recovers.

[0033] The application realizes message priority scheduling by constructing a multi-dimensional failure cause label, quantifying a failure feature vector and combining a failure heat index (FHI), introduces a gradual fragmentation retransmission mechanism and a cross-protocol dynamic degradation retransmission strategy on the basis of a traditional 5G message retransmission mechanism, and thus forms a complete, adaptive and intelligent message retransmission optimization system; the system can effectively identify multi-dimensional causes of message sending failure in a complex and changeable wireless network environment, including network state fluctuation, user online state anomaly and high base station load factors, and quantifies the multi-dimensional information into a failure feature vector that can be used for algorithm decision-making through a dynamic weight formula, so that the system can fully consider various influencing factors when selecting a retransmission strategy and realize accurate and personalized retransmission decision-making; by introducing the failure heat index, the application can quantify the frequency, concentration and density of failures in a network area, so that message retransmission not only depends on the failure state of a single message, but also can refer to the overall network condition for priority sorting, so as to reasonably arrange the message sending sequence in a congested environment and improve resource utilization efficiency; further, the application innovatively designs a gradual fragmentation retransmission mechanism to split a complete message into an abstract, key metadata and non-key content for hierarchical transmission; when the network condition is poor or congested, the system preferentially sends the abstract and key metadata to ensure that the user can receive the core information as soon as possible; after the network condition improves, the complete content is gradually supplemented, thereby effectively alleviating the high delay and low success rate problems caused by traditional single full retransmission; the fragmentation mechanism not only improves the timeliness of core information delivery, but also significantly reduces the occupation of network bandwidth by redundant data, and realizes optimized utilization of network resources; at the same time, the application proposes a cross-protocol dynamic degradation retransmission strategy, which can be used for various transmission protocols that the 5G message may rely on, including RCS, IMS, SMS, MMS or a low-speed data channel; the system can dynamically calculate the weight of each fragment under different protocols according to the failure feature vector and the FHI, and select the optimal protocol for sending; in a high failure heat environment, for key fragments, the system will perform multi-protocol sending according to a probability-driven redundancy strategy, thereby improving the reliability of message delivery in a complex network condition; for non-key fragments, low redundancy or delayed sending is selected according to the priority and network condition, thereby balancing resource consumption and transmission reliability; through an iterative feedback mechanism, the system can update the historical success rate of each protocol in real time, recalculate the weight and adjust the protocol selection, and realize an adaptive and multi-round optimized dynamic degradation retransmission process; the strategy breaks through the limitation of traditional retransmission mechanisms that cannot guarantee message reliability in the case of single-point protocol failure, so that the system can still ensure the successful transmission of key messages when the protocol link is partially disabled.The application can significantly improve the overall delivery rate of 5G messages in a complex network environment, especially the success rate of core information transmission, provide users with stable and reliable communication experience, support various types of rich media messages including text, image, audio and video files, and have wide application value in enterprise communication, financial transaction notification, public information push and emergency message service scenarios; through the method of the application, even in the environment of frequent network fluctuations, high peak of base station load or frequent user mobile switching, the reliable and timely delivery of messages can be ensured, and the user experience and system service quality are improved; the application realizes the technical upgrading from the traditional single-dimensional and static message retransmission mechanism to the multi-dimensional, dynamic adaptive and intelligent retransmission system, solves the problems of low efficiency and insufficient reliability caused by network fluctuations, protocol single point failure and full message repeated retransmission in the prior art, and has important value for performance optimization and application promotion of the 5G message system. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 It is a 5G message sending failure retransmission optimization method flow chart. Figure 2 It is a 5G message sending failure retransmission optimization system module diagram. DETAILED DESCRIPTION

[0035] The application will be further clearly and completely explained below, but the protection scope of the application is not limited thereto.

[0036] As shown in the figure, it is a 5G message sending failure retransmission optimization method flow chart. Figure 1

[0037] A 5G message sending failure retransmission optimization method, comprising: S1: constructing a multi-dimensional failure reason label for the 5G message sending failure, the label comprising network state dimension, user state dimension and base station load dimension; S2: calculating a dynamic weight based on the multi-dimensional failure reason label to obtain a failure feature vector for the 5G message; S3: adaptively selecting a retransmission strategy according to the failure feature vector, including delayed retransmission, fragmented retransmission, path switching retransmission and cross-protocol degradation retransmission; S4: using a failure heat index FHI in the retransmission process, the index being calculated from failure frequency, failure concentration and network area failure density; S5: performing priority scheduling on the message according to the failure heat index to determine whether to retransmit immediately or delay queuing; S6: splitting the 5G message and using a progressive retransmission mechanism to preferentially transmit message digest and key metadata, and then gradually supplement the complete content after the network condition improves. ​

[0038] As Figure 2 shown, it is a 5G message sending failure retransmission optimization system module diagram.

[0039] A 5G message sending failure retransmission optimization system is realized based on a 5G message sending failure retransmission optimization method, comprising: Failure analysis and multi-dimensional label module: receiving message sending feedback information, including success and failure, network log, extracting and generating multi-dimensional failure reason label, including signal strength, packet loss rate, network type in network state dimension, user online state, device type in user state dimension, current load, congestion situation in base station load dimension, providing original data basis for subsequent strategy selection; Failure feature vector generation module: quantizing and weighting the multi-dimensional failure label, using a dynamic weight formula to generate a failure feature vector reflecting the failure characteristics of each message in different dimensions, and converting the original label into a numerical vector for algorithm decision; Re-transmission strategy selection module: according to the failure feature vector, the optimal retransmission strategy is adaptively selected, the strategy includes fragmented retransmission PFRT, delayed retransmission, path switching retransmission, cross-protocol dynamic degradation retransmission CPADR, forming a specific retransmission scheme for each message; Failure heat index module: according to the failure frequency, failure concentration and network area failure density to calculate the failure heat index, which is used to measure network congestion and failure concentration area, and provide indicators for message priority sorting and retransmission rhythm; Priority scheduling module: according to the FHI and user historical behavior prediction to determine the message retransmission order, decide to retransmit immediately or delay in queue, reasonably allocate retransmission order in resource-limited and high failure environment, and improve core content delivery rate; Progressive retransmission module: split the message into abstract, key metadata and complete content, preferentially send abstract or key fragments, and gradually supplement complete content after network conditions improve, combined with the fragmentation strategy and protocol selection output by the retransmission strategy selection module.

[0040] Further, the step S2 based on the failure reason label obtains a failure feature vector for the 5G message, comprising: S21: Network state dimension construction, collecting network real-time state information from message sending path, including packet loss rate , average delay , bandwidth occupation and network slice availability , after normalizing each index, calculate the network state score: , wherein , , , is an adjustable weight coefficient, optimized according to historical failure logs; S22: User state dimension construction, collect user online state, terminal type and reliability, user activity and historical message receiving success rate, quantize each index, and calculate user state score: , wherein represents the user online probability, represents the terminal reliability score, represents the user recent activity, represents the user historical receiving success rate, , , , is an adjustable weight coefficient; S23: Base station load dimension construction, obtain the current connection number of the base station where the message is sent , base station failure rate and regional failure density information, the specific calculation method of base station load score is: , wherein is the maximum number of connections that the base station can bear, is an adjustable weight coefficient; S24: Multi-dimensional label vector generation and fusion, integrate three-dimensional scores into a multi-dimensional failure reason label vector of the message: , add dynamic weight to each dimension to form the final failure feature vector: ; S25: Update each sent 5G message result to the corresponding dimension index, including network packet loss rate, user online probability and base station load state, regularly analyze failure logs, and adjust weights through minimizing prediction error and reinforcement learning ; S26: The failure feature vector is directly used as the input of the retransmission strategy selection module, providing basis for subsequent priority scheduling and progressive retransmission.

[0041] The multi-dimensional failure reason label is not only composed of numerical features, but also uses a hierarchical semantic mapping structure: Base layer tag (Base Tag): Physical indicators directly extracted from raw data, such as packet loss rate, delay, online rate, load value, etc. Fusion layer tag (Fusion Tag): Convert base layer tags into semantic tags through clustering or rule engine, such as "high delay", "regional congestion", "terminal high risk", etc. Behavior Tag: Dynamic tags generated by combining historical behavior data, such as "periodic failure user", "instant overload base station", "short-time high failure area".

[0042] The system uses a sliding time window when generating failure tags , the system will automatically increase the weight of the dimension tag when a certain dimension indicator shows abnormal fluctuations within the window, and when the dimension performance returns to stability, the weight decay mechanism will be triggered, which is mathematically expressed as , where is the weight of the ith dimension, including network, user, base station, is the change rate of the dimension indicator within the current window, is the change threshold, is the weight adjustment factor, this dynamic migration mechanism enables the tag system to continuously reflect the real changes in the current network environment, rather than static feature descriptions.

[0043] To prevent the formation of "failure hotspots" in the same geographic area or the same base station within a short period of time, the system further introduces a spatial clustering algorithm to identify potential "fault clusters" by modeling the geographic location attribute of failure tags. The clustering algorithm is , represents the geographic distance between two failure records, is the spatial distance threshold, is the tag similarity threshold, if the number of failure tags within a cluster exceeds the threshold, the system will generate a new "regional failure tag" and dynamically adjust the base station load dimension weight within the region. This mechanism enables the tag system to have spatial awareness and prevent concentrated retransmission avalanches at the macro level.

[0044] Further use of the tag evolution mechanism, periodically (such as every hour) based on failure samples and retransmission success samples, to retrain and optimize the tag feature vector: When a certain type of tag is associated with a high failure rate for a long time, the risk weight of that type of tag is automatically increased; When a certain tag is associated with successful retransmission for a long time, the risk weight is reduced; For new unknown tags (such as new terminal models, temporary regional interference), the system will first assign a low confidence tag and gradually correct it in subsequent learning.

[0045] Further, the failure heat index used in the retransmission process in step S4 includes: S31: Multidimensional failure data collection, automatically collect multidimensional data related to the failure event after the message sending failure, the data includes: network layer parameters, specifically containing signal strength, delay, packet loss rate; user layer state, specifically containing terminal online rate, network switching frequency, moving speed; base station layer parameters, specifically containing base station load, congestion degree, switching frequency; message layer characteristics, specifically containing message type, message length, retransmission times, real-time requirements; geographical layer information, specifically containing failure occurrence position coordinates, area identifier, failure concentration degree; the data is divided into time windows with fixed window size Sampling to form a time-series multidimensional sample set ; S32: Failure density matrix construction, divide the spatial coordinate area into a fixed grid, and count the number of failure events in each grid within each time window, use Gaussian kernel function to smooth the local area to eliminate the discreteness of the data, and obtain the failure density matrix , which is defined as, Where x and y are the horizontal and vertical coordinates of the geographical position, t is the time, is the geographical coordinates when the ith failure event occurs, is the failure event index set in the current time window , used to limit the time range, is the time stamp when the ith failure event occurs, is a spatial distance function, representing the distance between point and evaluation point , is the geographical position of , is specifically represented as , is a kernel function, which is used to convert spatial distance into weight; S33: Dimension weight dynamic calculation, the system calculates the information entropy value of each dimension, reflecting the contribution of each dimension to the overall failure trend, and generates a weight vector , where is the network state dimension, is the user state dimension, is the base station load dimension, is the message characteristic dimension, is the geographical distribution dimension, and the dimension calculation method is, , is the real-time information entropy value of the ith dimension, The sum of information entropy of all dimensions, the higher the information entropy, the greater the volatility and uncertainty of the dimension, the stronger the explanation of the system failure mode, and therefore the higher the weight. The system dynamically adjusts the weight according to real-time data to ensure that the model is self-adaptive to environmental changes; S34: Local failure feature vector generation, after obtaining the weight vector, the system extracts the local failure features in the current time window to form a vector , where each component represents the failure rate and statistical indicators of the corresponding dimension, including base station failure rate, message timeout rate, and regional density. This vector is the "failure feature fingerprint" under the current network state; S35: Time decay factor calculation, the influence of historical failure events will decrease over time, and a time decay factor a(t) is used, where the specific calculation method of a(t) is, , where λ is the decay coefficient, used to control the contribution of past data to the current FHI. By introducing a(t), it is avoided that old data causes excessive interference to the current decision; S36: Geographical aggregation correction, the spatial statistical method is used to evaluate the aggregation of failure events, and the Ripley's K function is used to calculate the aggregation coefficient , where The specific calculation method is, , where K(d) is the spatial distribution function of actual failure events, representing the average number of other failure event points within a radius of d around any failure event point. After area normalization, the spatial aggregation function is obtained, and the specific calculation method is , where A is the observation area, N is the total number of failure events, is the distance between the i th and j th failure event points, is the indicator function, which takes 1 when , otherwise it takes 0, where is the reference function of theoretical random distribution, representing the expected average number of events within a radius of d under completely random distribution. It is used as a control baseline to help the system determine whether the current failure distribution is abnormally aggregated, and its calculation method is , when >0, it indicates that the failure events in a certain area show a significant aggregation trend, and the system will increase the priority of message retransmission in this area; S37: The system integrates the weight of each dimension, local features, time decay, and geographical aggregation to calculate the failure heat index FHI: wherein K is the aggregation amplification factor, used to adjust the influence of the geographical aggregation effect, the calculated FHI is normalized, with a range of [0, 1], divided into five levels, 0 to 0.2 is normal, 0.2 to 0.4 is slight failure, 0.4 to 0.6 is moderate failure, 0.6 to 0.8 is high-risk failure, and 0.8 or more is serious failure, the scheduling module automatically determines the retransmission strategy according to the FHI level, FHI low, merge queue, delay retransmission; FHI medium, retransmit normally; FHI high, enable path switching, cross-protocol retransmission; FHI extremely high, suspend batch message sending and trigger network diagnosis; the value of FHI reflects the comprehensive risk level of the failure event, when the value of FHI is high, it means that the failure probability of the current region or user group is on the rise, network resources should be allocated preferentially or the retransmission path should be adjusted; S38: Finally, the final priority score of a single message m Determined by FHI and the message's own failure feature vector Jointly determined: , is the region to which the message m belongs, w is a linear weight vector of the message features to the score, is the number of seconds that the message has been waiting for retransmission, is a normalization function, , , are the coefficients of each term.

[0046] The specific calculation method of the information entropy value of each dimension is as follows: Discretization of dimension features and construction of probability distribution: based on the collected failure-related parameters, the indicators of this dimension are segmented and clustered to form a set of discrete states: Then the frequency of failure events in each state is counted to form a probability distribution: ; Noise smoothing calculation, the specific calculation method is , is the minimum noise smoothing coefficient, which is automatically calculated according to the data volatility of the recent time window , is the standard deviation of the dimension feature, is a system adjustment coefficient, with a value range of 0.01 to 0.05; Dimension information entropy calculation When the state distribution of a certain dimension is extremely single, it means that the entropy is low, indicating that this dimension has weak contribution to failure, when the state distribution of a certain dimension is very scattered and unpredictable, it means that the entropy is high, indicating that this dimension is a major failure factor; Multi-window fusion, , is the current window entropy, is the previous window entropy, For the first two windows of entropy, For the window weight; Entropy value normalization, , the normalized entropy can be directly used to calculate the weight: .

[0047] Time complexity: The calculation of FHI is mainly the summation of the failure events in the window and the time decay. The sliding window and incremental update can be used (the incremental update can reduce the processing of each new event to O(1)), which is suitable for high concurrency scenarios; Space: The event summary (count, time statistics, weighted sum) of each region needs to be maintained, rather than storing all original events, saving storage; Robustness: The fallback value and confidence level are designed for missing indicators (such as no slice state at a certain time), so that FHI can still work stably under incomplete information.

[0048] If the FHI of a base station exceeds The system puts the messages in the area of the base station whose Priority(m) is lower than a certain threshold into the delay queue, and slowly releases them from low to high according to the multi-scale FHI, avoiding a large number of retransmissions at the same time; for FHI that is low but the Priority(m) of a single message is high (such as OTP), immediate cross-protocol parallel retransmission is allowed to ensure real-time performance; after the FHI drops significantly (lower than ), the system quickly retransmits the delayed summaries and remaining slices (progressive retransmission) in the priority queue.

[0049] When it is detected that FHI abnormally increases sharply in a short time (such as the growth rate in a short time window exceeds the threshold ), the system will trigger the "abnormal alarm mode": Temporarily freeze all unfinished retransmission requests in the area and request network intervention (such as switching back to routing or notifying the operator); Increase the monitoring of FHI in adjacent areas to determine whether it is a horizontal spread.

[0050] The failure heat index realizes the dynamic evaluation of the risk of message sending failure by fusing time, space, network, user, and message multi-dimensional features: Realize the dynamic retransmission priority decision: avoid unnecessary repeated retransmission, reduce network congestion, and improve overall bandwidth utilization; Regional prediction ability: identify potential "network failure hotspots" in advance to avoid avalanche-like sending failures; Fusion of time decay and historical weight mechanism: enhance the real-time responsiveness of the model, so that the system can "respond immediately and converge quickly", and improve the adaptive ability; Cross-dimension cause diagnosis: "failure tracing" within the system, facilitating intelligent operation and maintenance or policy optimization, forming a failure model with causal explainability.

[0051] FHI value as a unified risk indicator of the system, directly drives the sending and scheduling module: When FHI is low, the system delays retransmission or merges the queue; When FHI is high, the system prioritizes retransmission or path switching; When FHI is extremely high, the system automatically suspends batch tasks and triggers network diagnosis.

[0052] Further, the step S6 adopts a progressive retransmission mechanism, including: S41: Sharding the to-be-retransmitted message M to form a shard set And calculate the semantic importance value for each shard High importance shards include message digest, title, main image, and structural metadata; S42: Calculate the retransmission priority score of each shard Where the shard importance, network condition, and regional heat are considered comprehensively, Where, is the network failure rate standardized value, , is the failure heat index, and shards with higher priority will be scheduled first in the retransmission window; S43: Based on the current network available bandwidth And the average shard size Calculate the progressive sending window size, i.e., the number of shards allowed to be sent simultaneously: Where is the window compression coefficient, which automatically compresses the window as FHI increases, avoiding secondary congestion in high failure heat areas, is the minimum number of shards allowed in the window, is the maximum number of shards allowed in the window, which is calculated by The theoretical number of sendable shards is clipped to the range of , ] and multiplied by Further reduce the window when the failure heat is high to achieve adaptive progressive sending; S44: After sorting by priority score, select the shard set within the window from high to low And calculate the adaptive redundancy ratio, Where is the redundancy growth rate, is the base redundancy ratio, is the shard semantic importance, and the redundancy ratio Failure heat varies with fragment importance; S45: Perform redundancy coding on fragments in the window, generate extra check fragments to form an error-correctable progressive sending group; S46: Perform first sending, the receiving end can display the summary part and part of the visual content of the message after receiving part of the high-priority fragments, to realize the "first available and then complete" progressive experience; S47: Monitor the ACK / NACK messages fed back by the receiving end, and calculate the local loss rate and delay index; S48: According to the feedback and FHI change, perform hierarchical retransmission decision, when , retransmit all NACK fragments immediately; when , only retransmit the first k fragments of priority, where , when , delay the retransmission of low-priority fragments, and only perform cross-protocol concurrent sending on high-importance fragments.

[0053] The metadata of each fragment includes: frag_id: integer; importance: , summary ≈ 1 size: bytes; state ; send_count: number of times sent; last_send_ts: timestamp; hash: integrity check; p_score: priority score, calculated in real time.

[0054] Suppose the system performs retransmission on a 20KB 5G rich media message: The message is split into 10 fragments, with importance values [1.0, 0.9, 0.8, …, 0.1]; The current network bandwidth is estimated = 8KB, the average fragment size = 2KB, and FHI = 0.6; Then the window size ; The system only selects the top three fragments (summary, title, thumbnail) and adds 20% redundancy; After sending, the receiving end can immediately display the core content; The remaining fragments will automatically trigger reissuance after FHI drops.

[0055] Complexity: Most operations are sorting and window selection (O(k log k)), feedback processing can be batched, FHI and MDFRL are incremental updates (constant time / event), making it suitable for high-concurrency scenarios; Robustness: For missing data (without FHI or network estimation), a conservative default is adopted (FHI is set to 0.5, loss_est is set to the historical median), and the missing data is compensated for by the Learning Module; Security: Each shard contains a verification hash, and important shards can be signed to prevent tampering.

[0056] In this embodiment, before performing the fragmentation operation, the system first performs content structure analysis on the original 5G message, dividing the message into "core content fragments" and "auxiliary content fragments". The core fragments include the title, event summary, first image or key information fields, while the auxiliary fragments include secondary images, long text or multimedia additional content. When the network quality is poor or the failure rate index is high, the system prioritizes retransmitting only the core fragments, so that the receiving end can recover the main semantic content of the message in a short time. When the network condition returns to normal, the auxiliary fragments are gradually retransmitted, realizing a progressive message reconstruction of "information usability first, content completeness later".

[0057] FHI also serves as a feedback update factor, participating in the dynamic scheduling of the retransmission phase. When a fragment fails consecutively during retransmission, the system reports the failure event corresponding to that fragment to the FHI module, correcting the failure heat value of the corresponding region in real time. Conversely, when a fragment retransmits successfully, the FHI value will decay according to the weight factor. Through this two-way linkage, the system can achieve "instant correction" of FHI at the fragment level, making the failure heat distribution closer to the real-time network state and avoiding window control imbalance caused by global lag.

[0058] Furthermore, step S3 uses cross-protocol downgrade retransmission, including: S51: Obtain failure information and evaluation conditions; receive the list of fragments that failed to be transmitted by the progressive fragmentation retransmission module. Obtain the failure heat index FHI value H and the multi-dimensional failure feature vector F to determine whether the cross-protocol degradation conditions are met. and Any one of the conditions in, where, For fragment weights, Based on historical success rate, Set a threshold for the system; S52: Protocol Availability and Weight Calculation, defining the set of currently available protocols. Includes RCS, IP, SMS, and MMS for each fragment. In each protocol Compute weight, , Importance of the shard, Shard In the protocol Historical success rate, H is the failure heat index, Protocol delay normalization value, System adjustable weight coefficient; S53: Dynamic protocol selection, for each shard Select the protocol with the highest weight , for high priority shards, include abstract and key metadata, preferentially select protocols with high weight and low delay for transmission; S54: Probability-driven redundancy scheduling, calculate the redundancy transmission probability for critical shards , with a probability Repeat sending critical shards on selected protocols to enhance reliability in high failure heat environments, and send non-critical shards with low redundancy according to system strategy, delay sending; S55: Transmission and feedback iteration, send messages according to shard priority and protocol selection results, receive feedback ACK and NACK, and update shard state , for failed shards, update the historical success rate of the protocol , recalculate the weight , select the next backup protocol, form a cycle iteration, until the shard is successful, reach the upper limit of retransmission.

[0059] The system first receives input from the progressive shard retransmission module (PFRT), including the current set of shards to be retransmitted and their transmission state information; At the same time, the system extracts the feedback results of the last transmission from the failure analysis module, including network delay, packet loss rate, signal strength, base station load, user online state and other multi-dimensional feature parameters; The system generates a failure feature vector according to these features, and simultaneously calls the FHI (Failure Heat Index) module to calculate the current failure heat index; If the heat index exceeds the preset threshold or continuous feedback failure is found, trigger the cross-protocol dynamic degradation process.

[0060] The system evaluates the applicability of each protocol for each shard in the current environment, and the indicators include: Historical transmission success rate (learned from logs); Current network delay and bandwidth utilization; Protocol load capacity and compatibility; User terminal support.

[0061] The system selects the optimal protocol for each message shard according to the protocol priority results obtained in the previous stage, and in the selection process, the importance of the shard is comprehensively considered, including: If the shard contains critical information (such as message digest, title, alert content, etc.), a protocol with low delay and high stability is preferred; If the shard belongs to non-critical content (such as pictures, attachments or supplementary description), a protocol with higher throughput but relatively lower reliability can be selected for transmission; In addition, in order to avoid the additional consumption caused by frequent switching, the protocol switching cost is also judged, such as signaling reconstruction time, encoding conversion cost, etc., and only when necessary, cross-protocol degradation is performed.

[0062] The system dynamically calculates the redundancy sending probability of the shard according to the importance and historical success rate of the shard, and for the shard with higher importance, the system may send it simultaneously on two different protocols to prevent the failure of a single channel from causing information loss. The core idea of redundant scheduling is "limited redundancy, accurate redundancy": the system does not blindly repeat the sending, but dynamically judges whether to send redundantly in combination with the real-time network state; when the network is congested or the resources are limited, the system can automatically reduce the redundancy strength to prevent new transmission pressure.

[0063] The sending end listens to the confirmation information (ACK) or failure return (NACK) from each protocol channel, and dynamically updates the internal state table according to the results; for successful shards, the system marks them as completed; for failed shards, the system records the failure reason of the current protocol (such as network timeout, terminal unresponsive, format not supported, etc.), and adjusts the historical success rate parameter of the protocol; then, the system reevaluates the protocol weight and selects the next alternative protocol for retransmission, if multiple attempts still fail, the system delays the shard according to the priority and tries again after the network recovers.

[0064] The embodiments of the present application disclose the preferred embodiments, but are not limited thereto, and those skilled in the art can easily understand the spirit of the present application according to the above embodiments, and make different inferences and changes, as long as they do not deviate from the spirit of the present application, they are within the protection scope of the present application.

[0065] The application realizes message priority scheduling by constructing a multi-dimensional failure cause label, quantifying a failure feature vector and combining a failure heat index (FHI), introduces a gradual fragmentation retransmission mechanism and a cross-protocol dynamic degradation retransmission strategy on the basis of a traditional 5G message retransmission mechanism, and thus forms a complete, adaptive and intelligent message retransmission optimization system; the system can effectively identify multi-dimensional causes of message sending failure in a complex and changeable wireless network environment, including network state fluctuation, user online state anomaly and high base station load factors, and quantifies the multi-dimensional information into a failure feature vector that can be used for algorithm decision-making through a dynamic weight formula, so that the system can fully consider various influencing factors when selecting a retransmission strategy and realize accurate and personalized retransmission decision-making; by introducing the failure heat index, the application can quantify the frequency, concentration and density of failures in a network area, so that message retransmission not only depends on the failure state of a single message, but also can refer to the overall network condition for priority sorting, so as to reasonably arrange the message sending sequence in a congested environment and improve resource utilization efficiency; further, the application innovatively designs a gradual fragmentation retransmission mechanism to split a complete message into an abstract, key metadata and non-key content for hierarchical transmission; when the network condition is poor or congested, the system preferentially sends the abstract and key metadata to ensure that the user can receive the core information as soon as possible; after the network condition improves, the complete content is gradually supplemented, thereby effectively alleviating the high delay and low success rate problems caused by traditional single full retransmission; the fragmentation mechanism not only improves the timeliness of core information delivery, but also significantly reduces the occupation of network bandwidth by redundant data, and realizes optimized utilization of network resources; at the same time, the application proposes a cross-protocol dynamic degradation retransmission strategy, which can be used for various transmission protocols that the 5G message may rely on, including RCS, IMS, SMS, MMS or a low-speed data channel; the system can dynamically calculate the weight of each fragment under different protocols according to the failure feature vector and the FHI, and select the optimal protocol for sending; in a high failure heat environment, for key fragments, the system will perform multi-protocol sending according to a probability-driven redundancy strategy, thereby improving the reliability of message delivery in a complex network condition; for non-key fragments, low redundancy or delayed sending is selected according to the priority and network condition, thereby balancing resource consumption and transmission reliability; through an iterative feedback mechanism, the system can update the historical success rate of each protocol in real time, recalculate the weight and adjust the protocol selection, and realize an adaptive and multi-round optimized dynamic degradation retransmission process; the strategy breaks through the limitation of traditional retransmission mechanisms that cannot guarantee message reliability in the case of single-point protocol failure, so that the system can still ensure the successful transmission of key messages when the protocol link is partially disabled.The application can significantly improve the overall delivery rate of 5G messages in a complex network environment, especially the transmission success rate of core information, provide stable and reliable communication experience for users, support various types of rich media messages, including text, image, audio and video files, and has wide application value for enterprise communication, financial transaction notification, public information push and emergency message service scene; through the method of the application, even in the environment of frequent network fluctuation, high peak of base station load or frequent user mobile switching, the reliable and timely delivery of messages can be guaranteed, and the user experience and system service quality are improved; the application realizes the technical upgrading from the traditional single-dimensional and static message retransmission mechanism to the multi-dimensional, dynamic adaptive and intelligent retransmission system, solves the problems of low efficiency and insufficient reliability caused by network fluctuation, protocol single point failure and full amount message repeated retransmission in the prior art, and has important value for performance optimization and application promotion of the 5G message system.

Claims

1. A method for optimizing 5G message transmission failure retransmission, characterized in that, include: S1: Construct multi-dimensional failure reason labels for 5G messages that fail to be sent, including network status dimension, user status dimension, and base station load dimension; S2: Calculate dynamic weights based on the multi-dimensional failure reason labels to obtain the failure feature vector for the 5G message; S3: Based on the failure feature vector, adaptively select a retransmission strategy, including delayed retransmission, fragmented retransmission, path switching retransmission, and cross-protocol degradation retransmission; S4: The Failure Heat Index (FHI) is used during the retransmission process. The index is calculated from the failure frequency, failure concentration, and network area failure density. S5: Prioritize messages based on the failure heat index to determine whether to resend them immediately or delay queuing. S6: 5G messages are split and a progressive retransmission mechanism is adopted, prioritizing the transmission of message summaries and key metadata, and then gradually retransmitting the complete content as network conditions improve.

2. The 5G message transmission failure retransmission optimization method according to claim 1, characterized in that, In step S2, based on the failure reason label, a failure feature vector for the 5G message is obtained, including: S21: Network state dimension construction, collecting real-time network state information from the message sending path, including packet loss rate. Average delay Bandwidth usage and network slicing availability After normalizing each indicator, the network state score is calculated: ,in , , , This is an adjustable coefficient, optimized based on historical failure logs; S22: User status dimension construction, collecting user online status, terminal type and reliability, user activity, and historical message reception success rate, quantifying each indicator, and calculating user status score: ,in Indicates the probability of a user being online. This indicates the terminal reliability score. Indicates the user's recent activity level. This indicates the user's historical reception success rate. , , , These are adjustable weighting coefficients; S23: Base station load dimension construction, obtain the current number of connections of the base station where the message was sent. Base station failure rate and regional failure density Information: The specific calculation method for base station load score is as follows: ,in This represents the maximum number of connections a base station can support. These are adjustable weighting coefficients; S24: Multidimensional Tag Vector Generation and Fusion, integrating three-dimensional scores into a multidimensional failure reason tag vector for the message: Add dynamic weights to each dimension This forms the final failure feature vector: ; S25: Update the corresponding metrics for each failed 5G message, including network packet loss rate, user online probability, and base station load status. Periodically analyze failure logs and adjust weights by minimizing prediction error and using reinforcement learning. ; S26: The failure feature vector It serves directly as input to the retransmission strategy selection module, providing a basis for subsequent priority scheduling and progressive retransmission.

3. The 5G message transmission failure retransmission optimization method according to claim 1, characterized in that, The failure popularity index used in step S4 during the retransmission process includes: S31: Multi-dimensional failure data collection. After a message transmission failure, multi-dimensional data related to the failure event is automatically collected. This data includes: network layer parameters, specifically signal strength, latency, and packet loss rate; user layer status, specifically terminal online rate, network handover frequency, and mobile speed; base station layer parameters, specifically base station load, congestion level, and handover frequency; message layer characteristics, specifically message type, message length, retransmission count, and real-time requirements; and geographic layer information, specifically the coordinates of the failure location, area identifier, and failure concentration. The data is then processed in a fixed time window. Sampling is performed to form a time-series multidimensional sample set. ; S32: Failure density matrix construction. The spatial coordinate region is divided into fixed grids, and the number of failure events in each grid is counted within each time window. A Gaussian kernel function is used to smooth the local region to eliminate data discreteness, resulting in the failure density matrix. Its definition is: Where x and y are the horizontal and vertical coordinates of the geographical location, and t is the time. Let be the geographical coordinates at the time the i-th failure event occurs. For the current time window A set of failure event indexes within the specified time range. Let i be the timestamp of the i-th failure event. It is a spatial distance function, representing a point. With assessment points The distance between them for Geographical location The specific representation is as follows , This is a kernel function used to convert spatial distance into weights; S33: Dynamic calculation of dimension weights; the system calculates the information entropy value for each dimension. This reflects the contribution of each dimension to the overall failure trend, and a weight vector is generated accordingly. ,in For the network state dimension, For the user status dimension, From the perspective of base station load, As a message characteristic dimension, For the geographical distribution dimension, the general calculation method for each dimension is as follows: Where i is the information dimension, and its value ranges from network status dimension, user status dimension, base station load dimension, message characteristic dimension, and geographical distribution dimension. Let be the real-time information entropy value of the i-th dimension. The sum of information entropy of all dimensions. The higher the information entropy, the greater the volatility and uncertainty of that dimension, and the stronger its explanatory power for system failure modes. Therefore, its weight is also higher. The system dynamically adjusts the weights according to real-time data to ensure that the model adapts to changes in the environment. S34: Generation of local failure feature vectors. After obtaining the weight vector, the system extracts the local failure features under the current time window to form a vector. Each component represents the failure rate and statistical indicators of the corresponding dimension, including base station failure rate, message timeout rate, and area density. This vector is the "failure feature fingerprint" under the current network state. S35: Calculation of the time decay factor. The impact of historical failure events weakens over time, so a time decay factor α(t) is used. The specific calculation method for α(t) is as follows: , where λ is the attenuation coefficient, used to control the contribution of past data to the current FHI. By introducing α(t), we can avoid the old data from causing excessive interference to the current decision. S36: Geographic clustering correction, assessing the clustering of failure events using spatial statistical methods, and calculating the clustering coefficient using Ripley's K function. ,in The specific calculation method is as follows: Where K(d) is the spatial distribution function of actual failure events, representing the average number of other failure event points around any failure event point within an observation radius of d. It is a spatial aggregation function obtained after region area normalization, and its specific calculation method is as follows: A is the area of ​​the observation region, and N is the total number of failure events. Let be the distance between the i-th and j-th failure event points. For indicator functions, when The value is 1 if the condition is met, and 0 otherwise. Let be a reference function for the theoretical random distribution, representing the expected average number of events within a radius d under a completely random distribution. It serves as a baseline to help the system determine if there are anomalous clusters in the current failure distribution. Its calculation method is as follows: ,when A value >0 indicates that failure events show a clear clustering trend in a certain area, and the system will increase the priority retransmission level of messages in that area; S37: The system integrates weights across various dimensions, local features, time decay, and geographical clustering to calculate the Failure Heat Index (FHI). κ is the aggregation amplification factor, used to adjust the impact of geographic aggregation effect. The calculated FHI is normalized, and its range is [0,1]. It is divided into five levels: 0 to 0.2 is normal, 0.2 to 0.4 is slight failure, 0.4 to 0.6 is moderate failure, 0.6 to 0.8 is high-risk failure, and above 0.8 is severe failure. The scheduling module automatically determines the retransmission strategy according to the FHI level. If the FHI is low, the queue is merged and the retransmission is delayed; if the FHI is medium, the retransmission is carried out through the normal path; if the FHI is high, path switching is enabled and cross-protocol retransmission is carried out; if the FHI is extremely high, batch message sending is suspended and network diagnosis is triggered. The value of FHI reflects the comprehensive risk level of failure events. When the FHI value is high, it indicates that the failure probability of the current region or user group is on the rise, and network resources should be allocated or the retransmission path should be adjusted first. S38: Finally, the final priority score for a single message m. FHI and the message's own failure feature vector Joint decision: , Let m be the region to which message m belongs, and w be the linear weight vector from message features to score. This is the number of seconds the message has been waiting to be retransmitted. For normalization function, , , These are the coefficients for each item.

4. The 5G message transmission failure retransmission optimization method according to claim 1, characterized in that, Step S6 employs a progressive retransmission mechanism, including: S41: Fragment the message M to be retransmitted, forming a fragment set. And calculate semantic importance values ​​for each piece. High-importance fragments include message digest, title, main image, and structural metadata; S42: Calculate the retransmission priority score for each fragment. This involves comprehensively considering the importance of fragmentation, network conditions, and regional popularity. ,in, This is a standardized value for network failure rate. , As a failure popularity index, the higher the priority of the fragment, the more likely it will be scheduled in the resend window; S43: Based on the current available network bandwidth and average fragment size Calculate the progressive send window size, which is the number of fragments currently allowed to be sent simultaneously: ,in This is the window compression factor. As FHI increases, the window automatically compresses to avoid secondary congestion in high-failure-rate areas of the network. The minimum number of slices allowed for a window. The maximum number of slices allowed for a window, determined by... Theoretically, the number of fragments that can be sent can be calculated, and clipping can be used to trim them to [...]. , Range, then multiply by When the failure rate is high, the window is further narrowed to achieve adaptive progressive sending; S44: After sorting by priority score, select the set of fragments within the window from high to low. And calculate the adaptive redundancy ratio. ,in For redundant growth rate, Based on the basic redundancy ratio, To assess the semantic importance of fragmentation, redundancy ratio It changes in tandem with the popularity of failures and the importance of fragmentation; S45: Perform redundant coding on the fragments within the window to generate additional check fragments, in order to form an error-correctable progressive transmission group; S46: Execute the initial transmission. After receiving some high-priority fragments, the receiving end can display the message summary and some visual content to achieve a gradual experience of "usable first, then complete". S47: Monitor the ACK / NACK messages fed back by the receiver and calculate the local loss rate and latency indicators; S48: Based on feedback and changes in FHI, execute a tiered retransmission decision when... Immediately retransmit all NACK fragments; when Only the top k priority fragments are retransmitted, among which ,when It delays the retransmission of low-priority fragments and only performs cross-protocol concurrent transmission on high-importance fragments.

5. The 5G message transmission failure retransmission optimization method according to claim 1, characterized in that, The use of cross-protocol downgrade retransmission in step S3 includes: S51: Obtain failure information and evaluation conditions; receive the list of fragments that failed to be transmitted by the progressive fragmentation retransmission module. Obtain the failure heat index FHI value H and the multi-dimensional failure feature vector F to determine whether the cross-protocol degradation conditions are met. and Any one of the conditions in, where, For fragment weights, Based on historical success rate, Set a threshold for the system; S52: Protocol Availability and Weight Calculation, defining the set of currently available protocols. Includes RCS, IP, SMS, and MMS for each fragment. In each protocol Calculate the weights above. , Given the importance of fragmentation, For fragmentation In the agreement The historical success rate, where H is the failure popularity index. For the protocol delay normalized value, These are the adjustable weighting coefficients of the system. S53: Dynamic protocol selection for each fragment. Select the protocol with the highest weight. For high-priority fragments, which contain digests and key metadata, protocols with high weight and low latency are prioritized for transmission. S54: Probability-driven redundancy scheduling calculates the redundancy transmission probability for critical fragments. With probability Key fragments are repeatedly sent on the selected protocol to enhance reliability in high failure hot environments, while non-critical fragments are sent with low redundancy or delayed according to the system policy. S55: Iterate through sending and receiving feedback, select the result to send messages according to fragmentation priority and protocol, receive feedback ACK and NACK, and update the fragmentation status. For failed fragments, update the protocol's historical success rate. Recalculate the weights Then, select the next backup protocol and form a loop iteration until the fragmentation is successful and the retransmission limit is reached.

6. A system for optimizing 5G message transmission failure retransmission based on the method of claim 1, characterized in that, include: Failure Analysis and Multidimensional Labeling Module: Receives messages and sends feedback information, including success and failure, network logs, extracts and generates multidimensional failure reason labels, including signal strength, packet loss rate, and network type in the network status dimension, user online status and device type in the user status dimension, and current load and congestion status in the base station load dimension, providing the raw data foundation for subsequent strategy selection; Failure Feature Vector Generation Module: Quantizes and weights multi-dimensional failure labels, uses dynamic weight formula to generate failure feature vectors, reflects the failure characteristics of each message in different dimensions, and transforms the original labels into numerical vectors that can be used for algorithm decision-making. Retransmission strategy selection module: Adaptively selects the optimal retransmission strategy based on the failure feature vector. The strategies include fragmented retransmission PFRT, delayed retransmission, path switching retransmission, and cross-protocol dynamic degradation retransmission CPADR, forming a specific retransmission plan for each message. Failure Heat Index Module: Calculates the failure heat index based on failure frequency, failure concentration, and network area failure density. It is used to measure network congestion and areas of concentrated failure, and provides indicators for message prioritization and retransmission rhythm. Priority scheduling module: Determines the message retransmission order based on FHI and user historical behavior prediction, decides whether to retransmit immediately or delay queuing, and reasonably allocates the retransmission order in resource-limited and high-failure environments to improve the delivery rate of core content; The progressive retransmission module splits messages into digests, key metadata, and the complete content. It prioritizes sending digests or key fragments and gradually retransmits the complete content as network conditions improve. It combines the fragmentation strategy and protocol selection output by the retransmission strategy selection module.

Images