A Method and System for Intelligent Switching of 2M Unframed Dual-Path Protection Based on Redundancy Check

Abstract

This invention provides a method and system for intelligent handover of 2M unframed dual-path protection based on redundancy check, relating to the field of communication technology. The method includes acquiring and aligning dual 2M unframed bitstreams, generating check results using cyclic redundancy check, statistically analyzing error feature vectors within a sliding time window, constructing a link quality trajectory, extracting quality degradation warning signals, constructing a phased handover strategy based on the warning signals, and determining the handover timing sequence and executing link handover using a neural network evaluator. This achieves adaptive optimization of the handover strategy, improving the intelligence and reliability of link handover.

Description

Technical Field

[0001] This invention relates to communication technology, and more particularly to a method and system for intelligent switching of 2M unframed dual-path protection based on redundancy check. Background Technology

[0002] In modern communication networks, 2M unframed digital signals are widely used as the basic transmission unit in critical business systems of industries such as telecommunications, finance, and power. To ensure service continuity, a dual-path protection transmission architecture is typically adopted, where the same data stream is transmitted simultaneously through two physically independent links. When the primary link experiences quality degradation or interruption, the system can promptly switch to the backup link to ensure uninterrupted service.

[0003] 2M unframed dual-path protection technology mainly relies on mechanisms such as signal loss detection, signal quality threshold judgment, and fixed delay protection to achieve link switching. These traditional switching technologies typically use simple signal level detection or a fixed bit error rate threshold as the switching criterion and perform protection operations based on a preset switching time window.

[0004] Traditional switching mechanisms lack the ability to predict link quality and mainly rely on passively detecting link faults before performing switching. This makes it difficult to cope with scenarios of gradual degradation in link quality, which may lead to service interruptions or incorrect switching in actual operation.

[0005] Existing handover technologies use static and fixed decision thresholds, which cannot flexibly adjust handover strategies according to different network environments and service requirements. Especially in complex and ever-changing network environments, fixed threshold strategies often lead to inappropriate handover timing, affecting service continuity.

[0006] Traditional handover methods lack effective quality assessment tools, making it difficult to accurately assess the actual transmission quality of unframed signals. In particular, for link states that do not show obvious bit errors but already have potential quality risks, it is impossible to identify and take preventive handover measures in a timely manner, which increases the risk to system reliability. Summary of the Invention

[0007] The present invention provides a method and system for intelligent switching of 2M unframed dual-path protection based on redundancy check, which can solve the problems in the prior art.

[0008] A first aspect of the present invention provides a 2M unframed dual-path protection intelligent handover method based on redundancy check, comprising:

[0009] The 2M unframed bitstreams carried by each of the two transmission links are obtained, and clock recovery and frame boundary alignment are performed on the 2M unframed bitstreams to obtain dual bitstream data.

[0010] The dual-stream data are verified in real time based on the Cyclic Redundancy Check (CRC) algorithm to generate verification results for each link. Based on the verification results, the frequency and duration of bit error events are counted within a sliding time window to form a bit error feature vector characterizing the link quality status.

[0011] The bit error feature vector is mapped to a multi-dimensional quality space and a link quality trajectory is established. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal that can indicate changes in link quality is extracted.

[0012] Based on the quality degradation warning signal and the service status of the current working link, a phase switching strategy is constructed, and the switching decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory during the pre-switching preparation phase.

[0013] The handover decision threshold and the quality degradation warning signal are input into the neural network evaluator. The handover timing is determined based on the output of the neural network evaluator, and the link handover operation is executed. The neural network evaluator is trained online in combination with the handover execution effect to achieve adaptive optimization of the handover strategy.

[0014] The dual-stream data are verified in real time using a cyclic redundancy check (CRC) algorithm, generating verification results for each link. Based on these verification results, the frequency and duration of bit error events are statistically analyzed within a sliding time window to form a bit error feature vector characterizing the link quality status, including:

[0015] The dual-stream data is segmented according to a preset data block granularity, and a verification operation is performed on each data block based on the characteristic polynomial of the cyclic redundancy check algorithm. The verification pass status and verification failure status of each data block are identified according to the verification result. Based on the transformation relationship between the verification pass status and the verification failure status in the time series, the start time and end time of the bit error event are extracted.

[0016] Based on the start time and the end time, a time distribution sequence of bit error events is constructed within a sliding time window. By calculating the time interval between adjacent bit error events in the time distribution sequence and combining the reliability of the verification results of the cyclic redundancy check algorithm, continuous bit error intervals and discrete bit error intervals are identified.

[0017] Based on the ratio of the continuous error interval to the discrete error interval, a concentration feature value reflecting the error distribution pattern is calculated. Combined with the frequency of the error event within the sliding time window and the duration of the error event from the start time to the end time, an error feature vector containing the frequency of occurrence, the duration, and the concentration feature value is formed.

[0018] By calculating the time interval between adjacent bit error events in the time distribution sequence and combining the reliability of the verification results of the cyclic redundancy check algorithm, continuous bit error intervals and discrete bit error intervals are identified, including:

[0019] For each bit error event in the time distribution sequence, the reliability of the verification result used to evaluate the credibility of the bit error event is calculated based on the verification polynomial characteristics of the cyclic redundancy check algorithm and the bit stream transmission rate.

[0020] In the time distribution sequence, the time interval between adjacent bit error events is calculated, and the time interval is compared with a preset time clustering judgment benchmark. When the time interval is less than the time clustering judgment benchmark, the corresponding adjacent bit error events are marked as time clustering state. When the time interval is greater than or equal to the time clustering judgment benchmark, the corresponding adjacent bit error events are marked as time discrete state, thus obtaining a preliminary bit error event state distribution.

[0021] The reliability of the verification result is used as a weighting factor for the time clustered state and the time discrete state. Adjacent bit error events marked as time clustered states in the preliminary bit error event state distribution are filtered by reliability weighting to form a set of clustered events after reliability correction.

[0022] Traverse the time distribution sequence, merge the time-continuous bit error events in the reliability-corrected set of clustered events into continuous bit error intervals, and classify the bit error events in the time distribution sequence that do not belong to the continuous bit error intervals into discrete bit error intervals.

[0023] The bit error feature vector is mapped to a multi-dimensional quality space and a link quality trajectory is established. By analyzing the evolution trend of the link quality trajectory, quality degradation early warning signals that can indicate changes in link quality are extracted, including:

[0024] Based on the occurrence frequency dimension, duration dimension, and concentration feature value dimension contained in the bit error feature vector, a coordinate system of multidimensional quality space is constructed, and the bit error feature vector corresponding to each sampling time is used as the quality state point in the multidimensional quality space.

[0025] The quality state points at multiple consecutive sampling times are sequentially connected to form a link quality trajectory that reflects the evolution of link quality over time. The link quality trajectory is then segmented to identify the stationary evolution segment and the variable evolution segment within it.

[0026] For the change and evolution segment in the link quality trajectory, calculate the movement direction vector and movement rate of the quality state point in the multi-dimensional quality space, and determine whether the link quality trajectory is moving towards the quality degradation area based on the movement direction vector;

[0027] When the movement direction vector points to the quality degradation area and the movement speed exceeds a preset speed warning threshold, a quality degradation warning signal containing the movement direction vector and the movement speed is generated.

[0028] Based on the quality degradation early warning signal and the service status of the current working link, a phased handover strategy is constructed, and the handover decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory during the pre-handover preparation phase, including:

[0029] The mobile rate is extracted from the quality degradation warning signal, and combined with the service interruption tolerance in the current working link service status, the duration of the preparation phase and the decision threshold benchmark value corresponding to the phase handover strategy are determined; the pre-handover preparation phase is started according to the duration of the preparation phase, and the decision threshold benchmark value is set as the current handover decision threshold;

[0030] During the pre-switching preparation phase, the real-time evolution rate is obtained by calculating the displacement and time interval between the quality state points corresponding to adjacent sampling times of the link quality trajectory. Based on the change amplitude of the evolution rate benchmark value when entering the pre-switching preparation phase, the handover decision threshold is dynamically corrected.

[0031] The handover decision threshold and the quality degradation warning signal are input into a neural network evaluator. Based on the output of the neural network evaluator, the handover timing is determined and the link handover operation is executed. Online training of the neural network evaluator, considering the handover execution effect, includes:

[0032] The handover decision threshold and the quality degradation warning signal are input into the neural network evaluator. Based on the neural network evaluator, cluster analysis is performed on the feature combinations in the historical handover records to extract the state transition probability distribution that characterizes the evolution of link quality. The handover decision confidence level reflecting the rationality of the current handover timing is calculated based on the state transition probability distribution.

[0033] Based on the state transition probability distribution, the link quality evolution trend after handover is predicted. The optimal handover sequence is determined by combining the handover decision confidence. When the handover decision confidence exceeds a preset confidence threshold and the predicted quality evolution trend meets the optimization objective, the link handover operation is executed, and the execution effect of the link handover operation is recorded. The execution effect includes handover response latency and quality improvement.

[0034] An exponential forgetting factor is used to perform time-series weighting on the switching response delay and the quality improvement magnitude to obtain long short-term memory features that reflect the historical switching performance. The long short-term memory features and the prediction accuracy of the state transition probability distribution are used together as training criteria to train the neural network evaluator online.

[0035] Based on the state transition probability distribution, the evolution trend of link quality after handover is predicted, and the optimal handover sequence is determined by combining the handover decision confidence, including:

[0036] Based on the state transition probability distribution, the link quality status after the handover is predicted in multiple steps. Starting from the current link quality status, the quality status distribution at subsequent time moments is calculated sequentially according to the transition probability between states, so as to obtain an evolution sequence that characterizes the trend of link quality change after the handover.

[0037] The quality state transition paths in the evolution sequence are traversed and calculated. The quality state improvement at each moment is combined with the state transition cost. The expected benefit value corresponding to each switching moment is calculated in combination with the switching decision confidence. The moment with the maximum expected benefit value is selected as the optimal switching sequence.

[0038] A second aspect of the present invention provides a 2M unframed dual-path protection intelligent handover system based on redundancy check, comprising:

[0039] The first unit is used to acquire the 2M unframed bitstream carried by each of the two transmission links, and to perform clock recovery and frame boundary alignment processing on the 2M unframed bitstream to obtain dual bitstream data.

[0040] The second unit is used to perform real-time verification on the dual-stream data based on the cyclic redundancy check algorithm, generate verification results for each link, and statistically analyze the frequency and duration of bit error events within a sliding time window based on the verification results to form a bit error feature vector characterizing the link quality status.

[0041] The third unit is used to map the bit error feature vector to a multi-dimensional quality space and establish a link quality trajectory. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal that can indicate changes in link quality is extracted.

[0042] The fourth unit is used to construct a phase switching strategy based on the quality degradation warning signal and the service status of the current working link, and to dynamically adjust the switching decision threshold according to the evolution rate of the link quality trajectory during the pre-switching preparation phase.

[0043] The fifth unit is used to input the switching decision threshold and the quality degradation warning signal into the neural network evaluator, determine the switching timing based on the output of the neural network evaluator and execute the link switching operation; and perform online training on the neural network evaluator in combination with the switching execution effect to achieve adaptive optimization of the switching strategy.

[0044] A third aspect of the present invention provides an electronic device, comprising:

[0045] processor;

[0046] Memory used to store processor-executable instructions;

[0047] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0048] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0049] The beneficial effects of this application are as follows:

[0050] By performing clock recovery and frame boundary alignment on dual 2M unframed bitstreams, effective processing of unframed signals is achieved, solving the technical problem that traditional methods cannot effectively process unframed signals. An innovative cyclic redundancy check (CRC) algorithm is used to perform real-time verification of the bitstream data and statistically analyze error characteristics within a sliding time window to form an error feature vector. This vector can more comprehensively and accurately reflect the link quality status and has higher reliability compared to the traditional single bit error rate indicator.

[0051] By mapping the bit error feature vector to a multi-dimensional quality space and establishing a link quality trajectory, and by analyzing the evolution trend of the quality trajectory, link quality degradation can be predicted in advance, providing early warning signals, thus realizing a technological upgrade from passive response to proactive prediction.

[0052] A phased handover strategy is constructed based on the quality degradation early warning signal, and the handover decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory, which improves the accuracy and flexibility of handover decision-making and effectively reduces unnecessary handover operations. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the intelligent switching method for 2M non-framed dual-path protection based on redundancy check, according to an embodiment of the present invention.

[0054] Figure 2 This is a flowchart of the adaptive clustering and quality feature extraction of time-domain bit error events according to an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the intelligent handover method for 2M unframed dual-path protection based on redundancy check, as described in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0058] The 2M unframed bitstreams carried by each of the two transmission links are obtained, and clock recovery and frame boundary alignment are performed on the 2M unframed bitstreams to obtain dual bitstream data.

[0059] The dual-stream data are verified in real time based on the Cyclic Redundancy Check (CRC) algorithm to generate verification results for each link. Based on the verification results, the frequency and duration of bit error events are counted within a sliding time window to form a bit error feature vector characterizing the link quality status.

[0060] The bit error feature vector is mapped to a multi-dimensional quality space and a link quality trajectory is established. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal that can indicate changes in link quality is extracted.

[0061] Based on the quality degradation warning signal and the service status of the current working link, a phase switching strategy is constructed, and the switching decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory during the pre-switching preparation phase.

[0062] The handover decision threshold and the quality degradation warning signal are input into the neural network evaluator. The handover timing is determined based on the output of the neural network evaluator, and the link handover operation is executed. The neural network evaluator is trained online in combination with the handover execution effect to achieve adaptive optimization of the handover strategy.

[0063] In one optional implementation, the dual-stream data are checked in real time using a cyclic redundancy check (CRC) algorithm to generate check results for each link. Based on these check results, the frequency and duration of error events are statistically analyzed within a sliding time window to form an error feature vector characterizing the link quality status, including:

[0064] The dual-stream data is segmented according to a preset data block granularity, and a verification operation is performed on each data block based on the characteristic polynomial of the cyclic redundancy check algorithm. The verification pass status and verification failure status of each data block are identified according to the verification result. Based on the transformation relationship between the verification pass status and the verification failure status in the time series, the start time and end time of the bit error event are extracted.

[0065] Based on the start time and the end time, a time distribution sequence of bit error events is constructed within a sliding time window. By calculating the time interval between adjacent bit error events in the time distribution sequence and combining the reliability of the verification results of the cyclic redundancy check algorithm, continuous bit error intervals and discrete bit error intervals are identified.

[0066] Based on the ratio of the continuous error interval to the discrete error interval, a concentration feature value reflecting the error distribution pattern is calculated. Combined with the frequency of the error event within the sliding time window and the duration of the error event from the start time to the end time, an error feature vector containing the frequency of occurrence, the duration, and the concentration feature value is formed.

[0067] The cyclic redundancy check (CRC) processing for the dual-stream data employs a parallel architecture, with an independent check processing unit configured for each stream. The check processing unit receives the raw stream data, which is a continuous binary sequence, with a typical transmission rate ranging from 100 Mbps to 10 Gbps. The CRC-16-CCITT standard is used as the CRC algorithm, with a characteristic polynomial of x. 16 +x 12 +x 5 +1 generates a 16-bit check sequence. The check processing unit has a built-in hardware accelerator, supporting pipelined check operations with a processing latency of less than 10 microseconds.

[0068] The data block segmentation strategy performs precise segmentation based on preset granularity parameters. The data block granularity setting supports four standard options: 128 bytes, 256 bytes, 512 bytes, and 1024 bytes, with a default configuration of 512 bytes. The segmentation processor maintains a double-buffering mechanism: the front-end buffer receives real-time bitstream data, while the back-end buffer provides complete data blocks for verification operations. When the accumulated data in the front-end buffer reaches the preset granularity, a buffer switching operation is triggered, with a switching delay of no more than 1 microsecond. Segment boundaries are byte-aligned to ensure that each data block contains a complete byte sequence.

[0069] Cyclic Redundancy Check (CRC) operations employ a lookup table optimization algorithm to improve processing efficiency. The check unit pre-calculates the CRC portion results corresponding to 256 8-bit input values ​​and stores them in fast access memory. The data block verification process is performed byte by byte. For each operation, the current byte value is read as the lookup table index, and the corresponding CRC increment value is XORed with the accumulated checksum. After the complete data block is processed, the final checksum is compared with the check field at the end of the data block, and the comparison result is directly mapped to a Boolean status value.

[0070] The verification status identification module monitors the verification operation results of each data block in real time. A successful verification is defined as a checksum match, marked with a value of 0; a failed verification is defined as a checksum mismatch, marked with a value of 1. The status record includes three fields: timestamp, data block number, and verification status value. The timestamp uses nanosecond precision, and the data block number uses an incrementing integer. The status sequence is stored chronologically in a circular buffer, designed to store the most recent 10,000 status records.

[0071] Error event boundary detection achieves automatic identification through state transition analysis. The boundary detector scans and verifies the state sequence, finding the rising edge from state value 0 to state value 1 as the start marker of the error event, and recording the corresponding timestamp as the start time. Similarly, it detects the falling edge from state value 1 to state value 0 as the end marker of the error event, and records the corresponding timestamp as the end time. To eliminate the influence of transient interference, boundary confirmation requires three consecutive data blocks to maintain the same state change trend. After detecting a complete error event, the time difference between the start time and the end time is calculated as the event duration.

[0072] The sliding time window management employs a fixed window length combined with an overlapping sliding strategy. The time window length can be configured with four options: 30 seconds, 60 seconds, 120 seconds, and 300 seconds, with a default setting of 60 seconds. The window sliding step size is set to 20% of the window length, meaning it slides once every 12 seconds, resulting in 80% time overlap between windows. The window manager maintains an active window queue with a fixed queue length of 5, always ensuring that data from the 5 most recent time windows is available. Each time a window slides, the oldest window data is removed, a new window is created, and the statistics counter is initialized.

[0073] The construction of the time distribution sequence of bit error events is based on the timestamp sorting of events within a window. The distribution analyzer collects all bit error event records within the current window and arranges them in ascending order of their start times to form a time-ordered sequence. After the sequence is constructed, the time interval between adjacent events is calculated. The interval is defined as the time difference between the end time of the previous event and the start time of the next event. The time interval sequence is used for subsequent continuity determination, with the interval value in milliseconds and precision maintained to the integer level.

[0074] The adjacent bit error event interval analysis uses a threshold comparison method to distinguish between continuous and discrete patterns. The interval threshold is calculated based on the statistical characteristics of the duration of all bit error events within the window, and the threshold value is equal to 1.5 times the average duration. When the interval between adjacent events is less than or equal to the threshold, it is determined to be a continuous bit error pattern; when the interval is greater than the threshold, it is determined to be a discrete bit error pattern. Continuous bit error determination requires at least three consecutive events to meet the interval condition; single short-interval events are still classified as discrete patterns.

[0075] The reliability assessment of CRC-16 results combines theoretical detection capability with actual error distribution characteristics. Theoretically, the CRC-16 algorithm has a detection probability of 99.998%, but in practical applications, it is affected by factors such as burst errors and periodic interference. The reliability correction module statistically analyzes the error pattern distribution of the most recent 1000 verification results, calculating the occurrence ratios of single-bit errors, multi-bit errors, and burst errors. Based on the error pattern distribution, the interval judgment threshold is adjusted: the threshold is increased by 20% when the proportion of burst errors is high, and decreased by 10% when the proportion of random errors is high.

[0076] The identification of continuous and discrete error intervals is based on the analysis of time interval sequences. The interval identifier traverses the entire time interval sequence, grouping consecutive short-interval events into continuous error intervals and marking isolated long-interval events as discrete error intervals. Continuous error interval records include four attributes: interval start time, end time, number of included events, and total interval duration. Discrete error interval records include three attributes: event start time, event end time, and event duration.

[0077] The concentration characteristic value of the bit error distribution pattern is calculated using a weighted algorithm based on the proportion of interval duration. The concentration calculator counts the total duration of consecutive bit error intervals and the total duration of discrete bit error intervals, and calculates the proportion of consecutive bit error duration. The concentration characteristic value is defined as the ratio of the proportion of consecutive bit error duration to the proportion of discrete bit error duration, with a ratio ranging from 0 to 10. When consecutive bit errors dominate, the concentration characteristic value is greater than 1; when discrete bit errors dominate, the concentration characteristic value is less than 1; and when only one pattern exists, the characteristic value takes the boundary value of 0 or 10.

[0078] The frequency of bit error events within the sliding time window is counted using an event counting method. The frequency counter maintains a counter representing the total number of bit error events within the current window, incrementing by 1 each time a new bit error event is detected. As the window slides, historical events exceeding the window's range are removed, and the corresponding counter value is decremented. The frequency is expressed in terms of events per minute, calculated as the total number of events within the window multiplied by 60, divided by the window length in seconds. The frequency value is preserved to one decimal place, typically ranging from 0 to 100.

[0079] The duration of bit error events is statistically analyzed using a weighted average algorithm. The duration counter collects the individual durations of all bit error events within a window and calculates the arithmetic mean as the representative duration. Duration is measured in milliseconds, with numerical precision maintained to the integer level. Abnormal duration data is filtered during the calculation process; the filtering condition is a value exceeding twice the standard deviation above or below the average. When there are fewer than three bit error events within the window, the duration is taken as the actual duration of the most recent event.

[0080] The error feature vector assembly uses a triple array structure, containing three numerical elements in a fixed order: occurrence frequency, average duration, and concentration feature value. The vector assembler triggers a vector update operation after each window slide, reading the latest statistical results to construct the vector. The vector elements are data types of floating-point numbers, with occurrence frequency precision to one decimal place, duration precision as an integer, and concentration feature value precision to three decimal places.

[0081] In one optional implementation, by calculating the time interval between adjacent error events in the time distribution sequence and combining the reliability of the verification results of the cyclic redundancy check algorithm, the identification of continuous error intervals and discrete error intervals includes:

[0082] For each bit error event in the time distribution sequence, the reliability of the verification result used to evaluate the credibility of the bit error event is calculated based on the verification polynomial characteristics of the cyclic redundancy check algorithm and the bit stream transmission rate.

[0083] In the time distribution sequence, the time interval between adjacent bit error events is calculated, and the time interval is compared with a preset time clustering judgment benchmark. When the time interval is less than the time clustering judgment benchmark, the corresponding adjacent bit error events are marked as time clustering state. When the time interval is greater than or equal to the time clustering judgment benchmark, the corresponding adjacent bit error events are marked as time discrete state, thus obtaining a preliminary bit error event state distribution.

[0084] The reliability of the verification result is used as a weighting factor for the time clustered state and the time discrete state. Adjacent bit error events marked as time clustered states in the preliminary bit error event state distribution are filtered by reliability weighting to form a set of clustered events after reliability correction.

[0085] Traverse the time distribution sequence, merge the time-continuous bit error events in the reliability-corrected set of clustered events into continuous bit error intervals, and classify the bit error events in the time distribution sequence that do not belong to the continuous bit error intervals into discrete bit error intervals.

[0086] For each bit error event in the time-distributed sequence, the reliability of the verification result needs to be calculated based on the characteristics of the check polynomial of the Cyclic Redundancy Check (CRC) algorithm and the bit stream transmission rate. Specifically, for each data packet marked as erroneous by the CRC algorithm, the reliability value of its verification result can be calculated by analyzing the order of the check polynomial, the Hamming distance, and the signal-to-noise ratio at the current bit stream transmission rate. For example, when using the CRC-32 check algorithm, based on its ability to detect all burst errors less than or equal to 32 bits, combined with the bit error rate estimate under the current transmission environment, a reliability coefficient between 0 and 1 can be assigned to each bit error event. The higher the reliability, the greater the probability that the bit error event has been correctly identified.

[0087] The process involves calculating the time intervals between adjacent error events in the time distribution sequence and comparing these time intervals with a preset time clustering criterion. The time clustering criterion can be predetermined based on factors such as system transmission characteristics, network environment, and historical error patterns, for example, set to 10 milliseconds. During calculation, the entire time distribution sequence is traversed. For each pair of adjacent error events E_i and E_{i+1} in the sequence, the time difference Δt_i = t_{i+1} - t_i is calculated. If Δt_i is less than the time clustering criterion, the pair of error events is marked as a time-clustered state; if Δt_i is greater than or equal to the time clustering criterion, it is marked as a time-discrete state. This process yields a preliminary error event state distribution containing the state labels for each pair of adjacent error events.

[0088] The reliability of the verification results is used as a weighting factor for the temporal clustered state and the temporal discrete state. Adjacent error events marked as temporal clustered states in the initial error event state distribution are then subjected to reliability-weighted filtering. Specifically, for each pair of adjacent error events marked as temporal clustered states, their average reliability is calculated. If the average reliability exceeds a predetermined reliability threshold (e.g., 0.7), the temporal clustered state label is retained; otherwise, it is remarked as a temporal discrete state. This step effectively filters out error events incorrectly labeled as clustered states due to random interference or misjudgment by the verification algorithm, forming a reliability-corrected set of clustered events.

[0089] The entire time distribution sequence is traversed, and temporally continuous bit error events in the reliability-corrected clustered event set are merged into continuous bit error intervals. Here, "temporally continuous" means that the time interval between a series of adjacent bit error events is less than the time clustering criterion, and their reliability all meets the weighted screening criteria. During the merging process, the start and end times of each continuous bit error interval, as well as the number of bit error events it contains, are recorded. Bit error events in the time distribution sequence that do not belong to continuous bit error intervals are classified as discrete bit error intervals.

[0090] In practical applications, a sliding window can be set to dynamically adjust the time aggregation judgment benchmark. For example, in a high-speed optical fiber communication system, when a change in the system's operating state is detected (such as signal strength fluctuations), the time aggregation judgment benchmark can be adaptively adjusted based on the bit error pattern over a recent period, thereby improving the recognition accuracy.

[0091] The continuous and discrete error intervals identified by the above methods have clear physical meanings. Continuous error intervals usually correspond to persistent channel interference, equipment failure, or instantaneous system instability, requiring targeted repair strategies. Discrete error intervals, on the other hand, are mostly related to random noise and occasional interference, and can be addressed by increasing coding redundancy or adjusting transmission parameters.

[0092] This error interval identification method, which is based on time characteristics and reliability weighting, has higher accuracy and stronger adaptability than the traditional method that only relies on bit error rate statistics. It is particularly suitable for high-reliability communication systems that require precise location and differentiation of different types of bit error causes.

[0093] In one optional implementation, the bit error feature vector is mapped to a multi-dimensional quality space and a link quality trajectory is established. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal indicating changes in link quality is extracted, including:

[0094] Based on the occurrence frequency dimension, duration dimension, and concentration feature value dimension contained in the bit error feature vector, a coordinate system of multidimensional quality space is constructed, and the bit error feature vector corresponding to each sampling time is used as the quality state point in the multidimensional quality space.

[0095] The quality state points at multiple consecutive sampling times are sequentially connected to form a link quality trajectory that reflects the evolution of link quality over time. The link quality trajectory is then segmented to identify the stationary evolution segment and the variable evolution segment within it.

[0096] For the change and evolution segment in the link quality trajectory, calculate the movement direction vector and movement rate of the quality state point in the multi-dimensional quality space, and determine whether the link quality trajectory is moving towards the quality degradation area based on the movement direction vector;

[0097] When the movement direction vector points to the quality degradation area and the movement speed exceeds a preset speed warning threshold, a quality degradation warning signal containing the movement direction vector and the movement speed is generated.

[0098] like Figure 2 As shown, the method includes:

[0099] Obtain bit error feature vector data, which includes the dimensions of occurrence frequency, duration, and concentration feature value. Based on these three dimensions, construct a coordinate system for a multi-dimensional quality space, forming a three-dimensional coordinate system. Each bit error feature vector represents the link quality state at a sampling time and is mapped to a quality state point in the multi-dimensional quality space.

[0100] For example, suppose the error feature vector collected at a certain moment contains: error occurrence frequency of 12 times / minute, error duration of 0.8 seconds, and concentration feature value of 0.65. Then the coordinates of the quality state point at that moment in three-dimensional space are (12, 0.8, 0.65).

[0101] After constructing the coordinate system, the quality state points at consecutive sampling times are sequentially connected according to time order. Specifically, the quality state points corresponding to times t1, t2, t3, ... are connected in sequence to form a link quality trajectory reflecting the evolution of link quality over time. This connection can be achieved using straight line segments or curved lines to clearly demonstrate the change process of link quality.

[0102] The resulting link quality trajectory is segmented to identify stable evolution segments and variable evolution segments. Stable evolution segments refer to the parts of the trajectory where the displacement between adjacent quality state points is small and the changes are not obvious; variable evolution segments refer to the parts of the trajectory where the displacement between adjacent quality state points is large and the changes are obvious.

[0103] The segment type can be determined by calculating the Euclidean distance between adjacent quality state points. If the distance between two points is less than a preset threshold (e.g., 0.5 units), the trajectory is considered to be a stationary evolution segment; if it is greater than the threshold, it is classified as a changing evolution segment. The segmentation threshold can be adjusted according to the actual application scenario and quality requirements.

[0104] For the identified evolutionary segments, the movement direction vector and movement rate of the mass state point in the multidimensional mass space are calculated. The movement direction vector represents the development trend of the trajectory and can be obtained by calculating the average of the direction vectors of several consecutive mass state points. For example, if the coordinates of the mass state points corresponding to two adjacent times t1 and t2 are P1(x1, y1, z1) and P2(x2, y2, z2) respectively, then the movement direction vector from P1 to P2 can be expressed as V=(x2-x1, y2-y1, z2-z1).

[0105] The rate of movement represents how quickly the state of mass changes, and can be obtained by calculating the displacement distance per unit time. For example, if the state of mass moves from P1 to P2 within a time interval Δt, and the distance moved is |P1P2|, then the rate of movement is |P1P2| / Δt.

[0106] Quality degradation regions are predefined in the multidimensional quality space. These regions typically correspond to spatial areas with high bit error frequency, long duration, and high concentration. Specifically, spatial areas with bit error frequency greater than a certain threshold (e.g., 20 times / minute), duration greater than a certain threshold (e.g., 1.5 seconds), and concentration characteristic value greater than a certain threshold (e.g., 0.8) can be defined as quality degradation regions.

[0107] Based on the calculated movement direction vector, it is determined whether the link quality trajectory is heading towards the quality degradation area. The specific determination method is to calculate the angle between the movement direction vector and the vector pointing to the center of the quality degradation area. If the angle is less than a preset threshold (such as 30 degrees), it is considered that the trajectory is heading towards the quality degradation area.

[0108] Simultaneously, it checks whether the movement speed exceeds a preset speed warning threshold. This threshold can be set according to actual application requirements, for example, to 0.5 units / second. When the movement direction vector is detected to be pointing towards a quality degradation area and the movement speed exceeds the preset speed warning threshold, a quality degradation warning signal is generated.

[0109] The quality degradation warning signal contains the movement direction vector and movement speed information, which can be represented as a data structure: {direction vector: (dx, dy, dz), speed: v}. For example, if the current movement direction vector is (0.8, 0.5, 0.3) and the movement speed is 0.7 units / second, then the generated quality degradation warning signal can be represented as {direction vector: (0.8, 0.5, 0.3), speed: 0.7}.

[0110] In practical applications, link quality trajectories can be visualized to intuitively reflect trends in link quality changes. For example, in communication network maintenance, technicians can observe 3D visualizations of link quality trajectories to promptly identify potential quality degradation risks and take remedial measures in advance.

[0111] When a link's quality trajectory shows a rapid trend towards a deteriorating area, the system automatically generates a quality degradation warning signal to alert technicians to the affected link. The movement direction vector and speed information contained in the warning signal help technicians assess the severity and urgency of the degradation trend, enabling them to take targeted intervention measures.

[0112] The link quality trajectory and early warning mechanism established by the above methods can effectively monitor and warn of changes in communication link quality, thereby improving the reliability and service quality of the communication system.

[0113] In one optional implementation, a phase handover strategy is constructed based on the quality degradation early warning signal and the service status of the current working link, and the handover decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory during the pre-handover preparation phase, including:

[0114] The mobile rate is extracted from the quality degradation warning signal, and combined with the service interruption tolerance in the current working link service status, the duration of the preparation phase and the decision threshold benchmark value corresponding to the phase handover strategy are determined; the pre-handover preparation phase is started according to the duration of the preparation phase, and the decision threshold benchmark value is set as the current handover decision threshold;

[0115] During the pre-switching preparation phase, the real-time evolution rate is obtained by calculating the displacement and time interval between the quality state points corresponding to adjacent sampling times of the link quality trajectory. Based on the change amplitude of the evolution rate benchmark value when entering the pre-switching preparation phase, the handover decision threshold is dynamically corrected.

[0116] The quality degradation early warning signal data parsing employs a structured information extraction method to process the input signal. The early warning signal parser locates the movement rate field from the data packet of the quality degradation early warning signal. This field is stored at a fixed offset of 16 bytes in the data packet, occupying an 8-byte length. The movement rate data is encoded in IEEE 754 double-precision floating-point format, representing the movement speed of the quality status point in the multi-dimensional quality space, in units of mass units per second. The parser verifies the validity of the movement rate values, ensuring that the values ​​are within the range of 0.001 to 10.000. Values ​​outside this range are marked as abnormal and trigger a re-acquisition.

[0117] The current working link service status is obtained through real-time data collection via the link status monitoring interface. The service status monitoring module maintains a link operation parameter table, including fields such as the current service type identifier, service priority level, service quality requirements, and real-time traffic statistics. Service interruption tolerance, as a core parameter of service status, is standardized according to service type. The tolerance for real-time audio and video calls is set to 50 milliseconds, for high-definition video streaming to 200 milliseconds, for file transfer to 1000 milliseconds, and for web browsing to 3000 milliseconds. Tolerance values ​​are stored in the service configuration table and can be dynamically adjusted according to specific application requirements.

[0118] The phased handover strategy is constructed using a two-factor decision algorithm based on mobility rate and service interruption tolerance. The strategy builder receives two input parameters: mobility rate and service interruption tolerance. It determines the optimal handover strategy through a predefined mapping matrix. The mapping matrix is ​​divided into 25 strategy intervals, with rows divided into 5 levels based on mobility rate and columns divided into 5 levels based on tolerance. Each interval corresponds to a specific combination of preparation phase duration and decision threshold baseline value. The strategy selection process first determines the mobility rate level, then the tolerance level, and the cross-determination of these two levels yields the corresponding strategy parameters.

[0119] The preparation phase duration calculation employs a piecewise linear interpolation method to ensure parameter continuity. The duration calculation module performs linear interpolation based on the specific position of the movement speed value within the corresponding range. Movement speed ranges are divided into: extremely slow (0.001-0.010), slow (0.010-0.050), medium (0.050-0.100), fast (0.100-0.500), and extremely fast (0.500-10.000). Each range corresponds to a base duration of 60 seconds, 30 seconds, 15 seconds, 8 seconds, and 3 seconds, respectively. The actual duration is adjusted based on the base duration according to business tolerance; the smaller the tolerance, the larger the adjustment coefficient, ensuring that sensitive services receive sufficient preparation time.

[0120] The decision threshold baseline value is set using a risk level assessment strategy. The baseline value setter comprehensively considers the severity of degradation reflected by mobility rate and the business sensitivity reflected by business tolerance. Lower baseline values ​​are set for scenarios with high urgency and business sensitivity to facilitate rapid handover triggering; higher baseline values ​​are set for scenarios with low urgency and low business sensitivity to avoid unnecessary handover operations. The baseline value ranges from 0.60 to 0.95, with lower values ​​indicating easier handover triggering. The baseline value calculation process includes safety margin considerations to ensure the accuracy of handover timing even under boundary conditions.

[0121] The pre-switching preparation phase startup control employs a state machine management mechanism. The phase control state machine includes four state nodes: idle, ready, execution, and completion. Upon receiving the startup command, the state machine transitions from the idle state to the ready state, simultaneously starting the ready phase timer. The timer is configured in decrementing counting mode, with an initial value set to the calculated duration of the ready phase and a counting precision of 10 milliseconds. During the ready state, the state machine continuously monitors link quality changes, transitioning to the execution state when the timer reaches zero or an emergency degradation signal is detected.

[0122] Switching the decision threshold setting takes effect immediately by writing to the threshold register. The decision threshold baseline value is written to a dedicated 32-bit threshold register, which supports direct storage of floating-point numbers. The threshold write operation is guaranteed to be atomic to avoid inconsistencies caused by concurrent access. The threshold register is connected to the quality comparator hardware module, enabling nanosecond-level real-time decision response. Changes to the threshold value trigger system event logging, including detailed information such as the change timestamp, the values ​​before and after the change, and the trigger reason.

[0123] Real-time monitoring of link quality trajectories employs high-precision coordinate tracking technology. The quality status point coordinate acquisition unit measures the coordinates of quality status points in the multi-dimensional quality space at fixed 100-millisecond intervals. Each status point coordinate includes three components: X-axis occurrence frequency, Y-axis duration, and Z-axis concentration characteristic value, with coordinate accuracy maintained to four decimal places. Coordinate data is stored in a circular buffer with a length of 20 status points, automatically overwriting the oldest data. The displacement vector is calculated by vector subtraction between status point coordinates at adjacent sampling times, with the magnitude of the displacement vector representing the displacement distance.

[0124] The time interval measurement achieves microsecond-level accuracy based on a hardware clock counter. A timestamp generator assigns a unique timestamp to each coordinate acquisition operation, sourced from the system's high-precision timer. The time interval between adjacent sampling moments is directly calculated using the timestamp difference, measured in microseconds. When converting the time interval value to seconds, it retains 6 decimal places with a typical range between 0.100000 seconds and 0.100100 seconds. An anomaly detection mechanism monitors the interval value, triggering sampling frequency correction when it exceeds the normal range.

[0125] The real-time evolution rate is obtained using the standard algorithm of dividing the displacement distance by the time interval. The evolution rate calculator receives two input parameters: displacement distance and time interval, and performs the division operation to obtain the instantaneous evolution rate. The evolution rate value is expressed in mass units per second, and the value range is typically between 0.001 and 1.000. To eliminate the influence of transient fluctuations, the evolution rate is filtered using a 5-point moving average. The filtered value is then output as a stable real-time evolution rate. The evolution rate calculation result is updated in real time to the evolution rate register for use by the subsequent dynamic correction module.

[0126] The evolution rate baseline value is acquired using a rate snapshot technique at the start of the preparation phase. The baseline value collector reads the current real-time evolution rate value at a precise moment at the start of the pre-switch preparation phase and saves this value as the evolution rate baseline value. The baseline value is stored in dedicated memory and remains unchanged throughout the preparation phase. The accuracy of the baseline value is consistent with the real-time evolution rate and is stored in 64-bit floating-point format. The validity of the baseline value is ensured through a numerical range check; baseline values ​​outside the reasonable range will trigger a re-acquisition operation.

[0127] The calculation of evolution rate change magnitude is based on the relative difference analysis between the current rate and the baseline rate. The change magnitude analyzer continuously compares the real-time evolution rate with the baseline value, calculating the absolute difference and relative difference between the two. The absolute difference represents the absolute change in rate, while the relative difference represents the relative percentage change in rate. The change magnitude uses the relative difference as the primary indicator, calculated by dividing the absolute difference by the baseline value. Positive change magnitudes indicate accelerated evolution, while negative values ​​indicate decelerated evolution; the magnitude reflects the degree of drastic change.

[0128] The dynamic correction of the handover decision threshold employs an adaptive adjustment algorithm to achieve intelligent response. The threshold correction controller dynamically adjusts the handover decision threshold value based on the magnitude of changes in the evolution rate. The adjustment strategy determines the correction direction and magnitude based on the sign and value of the change. When evolution accelerates, the threshold value decreases to improve handover sensitivity; when evolution slows down, the threshold value increases to reduce handover sensitivity. The correction magnitude is proportional to the change magnitude, with a proportionality coefficient set to 0.15 to ensure a moderate and stable correction response. The corrected threshold value undergoes boundary checks, limiting it to the range of 0.50 to 0.98.

[0129] In one optional implementation, the handover decision threshold and the quality degradation warning signal are input into a neural network evaluator; the handover timing is determined based on the output of the neural network evaluator, and a link handover operation is performed; online training of the neural network evaluator is performed in conjunction with the handover execution effect, including:

[0130] The handover decision threshold and the quality degradation warning signal are input into the neural network evaluator. Based on the neural network evaluator, cluster analysis is performed on the feature combinations in the historical handover records to extract the state transition probability distribution that characterizes the evolution of link quality. The handover decision confidence level reflecting the rationality of the current handover timing is calculated based on the state transition probability distribution.

[0131] Based on the state transition probability distribution, the link quality evolution trend after handover is predicted. The optimal handover sequence is determined by combining the handover decision confidence. When the handover decision confidence exceeds a preset confidence threshold and the predicted quality evolution trend meets the optimization objective, the link handover operation is executed, and the execution effect of the link handover operation is recorded. The execution effect includes handover response latency and quality improvement.

[0132] An exponential forgetting factor is used to perform time-series weighting on the switching response delay and the quality improvement magnitude to obtain long short-term memory features that reflect the historical switching performance. The long short-term memory features and the prediction accuracy of the state transition probability distribution are used together as training criteria to train the neural network evaluator online.

[0133] The handover decision threshold and quality degradation warning signal are passed as input parameters to the neural network evaluator. The handover decision threshold includes multi-dimensional parameters such as signal strength threshold, bit error rate threshold, and latency threshold, while the quality degradation warning signal contains real-time monitoring data of the current link status, such as the signal strength decline rate, the bit error rate increase trend, and latency fluctuations. The neural network evaluator adopts a multilayer perceptron structure, including an input layer, multiple hidden layers, and an output layer, to process this input data.

[0134] After receiving the data, the neural network evaluator performs cluster analysis on the feature combinations in the historical handover records. The K-means clustering algorithm is used to group similar historical handover scenarios into the same category, with each category representing a typical link quality change pattern. For each cluster, the state transition probability distribution of link quality over time is extracted. This state transition probability distribution records the probability of transitioning from the current state to various subsequent states, forming a Markov chain model. For example, when the link quality is in a "good" state, there is a 70% probability of remaining "good," a 20% probability of decreasing to "average," and a 10% probability of decreasing to "poor."

[0135] Based on the extracted state transition probability distribution, the confidence level of the current handover decision is calculated. This confidence level reflects the rationality of performing a link handover in the current state. The calculation method is as follows: The currently monitored link state parameters are matched with historical cluster centers to find the closest historical scenario category. Then, based on the state transition probability distribution of that category, the probability that performing the handover will improve link quality is calculated. For example, if historical data shows that performing a handover in a similar current state has an 85% probability of improving link quality, then the confidence level of the current handover decision is 0.85.

[0136] The evolution trend of link quality after handover is predicted using the state transition probability distribution. Based on a Markov chain prediction model, the state distribution of link quality at several future time points is simulated. Simultaneously, the optimal handover sequence is determined by combining the handover decision confidence level. When the handover decision confidence level exceeds a preset confidence threshold (e.g., 0.8) and the predicted quality evolution trend meets the optimization objective (e.g., predicting continuous improvement in link quality over the next 30 seconds), the link handover operation is executed.

[0137] After the link handover operation is executed, two key metrics are recorded: handover response latency and quality improvement. Handover response latency refers to the time from issuing the handover command to the establishment and stabilization of the new link, while quality improvement refers to the degree of improvement in link quality parameters (such as throughput, latency, and packet loss rate) before and after the handover. These metrics are used to evaluate the actual effectiveness of the handover operation.

[0138] To enable the neural network evaluator to learn from historical handover performance, an exponential forgetting factor is used to time-weight the handover response latency and quality improvement. The exponential forgetting factor λ (ranging from 0 to 1, with a typical value of 0.95) ensures that recent handover performance has higher weight. It is calculated as follows: for the performance metric En of the nth handover, its weighting value is λ. (n-1) ×En, the final weighted result is the sum of all weighted values. This processing method obtains long short-term memory features that reflect the historical switching performance, preserving long-term historical experience while focusing on recent switching results.

[0139] The prediction accuracy of long short-term memory features and state transition probability distributions was used as the training basis for online training of the neural network evaluator. Prediction accuracy was evaluated by comparing the degree of matching between the evolution trend of predicted quality and the actual measurement results. Online training employed stochastic gradient descent to dynamically adjust the neural network's weight parameters with the goal of minimizing prediction error. During training, the neural network's judgment strategy for similar scenarios was adjusted based on the actual effects after link switching, thereby improving its prediction and decision-making capabilities.

[0140] In practical applications, when a communication system detects a decline in the quality of the current link, the above method can intelligently determine the optimal handover time, avoiding premature handover (leading to unnecessary system overhead) or late handover (leading to communication interruption). For example, in vehicular communication scenarios, when the signal strength gradually weakens due to the high-speed movement of the vehicle, the system will switch to a base station with a stronger signal at the optimal time based on historical handover experience and the current quality change trend, ensuring the continuity and quality of communication services.

[0141] By continuously accumulating switching experience and conducting online training, the neural network evaluator can continuously optimize its decision-making capabilities, adapt to different communication environments and usage scenarios, and achieve intelligent and adaptive performance improvement of the link switching mechanism.

[0142] In one optional implementation, predicting the post-handover link quality evolution trend based on the state transition probability distribution and determining the optimal handover sequence in conjunction with the handover decision confidence includes:

[0143] Based on the state transition probability distribution, the link quality status after the handover is predicted in multiple steps. Starting from the current link quality status, the quality status distribution at subsequent time moments is calculated sequentially according to the transition probability between states, so as to obtain an evolution sequence that characterizes the trend of link quality change after the handover.

[0144] The quality state transition paths in the evolution sequence are traversed and calculated. The quality state improvement at each moment is combined with the state transition cost. The expected benefit value corresponding to each switching moment is calculated in combination with the switching decision confidence. The moment with the maximum expected benefit value is selected as the optimal switching sequence.

[0145] The state transition probability distribution data is acquired using historical quality state transition statistical analysis. The probability distribution builder extracts the quality state records for the most recent 30 days from the link quality historical database. Quality states are represented using discretized encoding, categorized into five levels: Excellent (Code 1), Good (Code 2), Average (Code 3), Poor (Code 4), and Very Poor (Code 5). The state transition statistician traverses the historical record sequence, calculating the transition frequency between adjacent quality states, constructing a 5×5 state transition frequency matrix. The transition probability is obtained through frequency normalization, with the sum of elements in each row equal to 1, forming the complete state transition probability distribution matrix. The probability values ​​are maintained to four decimal places, ranging from 0.0001 to 0.9999.

[0146] The current link quality status identification employs a real-time quality assessment and status mapping mechanism. The quality status evaluator receives performance indicators such as bit error rate, latency, jitter, and packet loss rate of the current link and calculates a comprehensive link quality score using a weighted comprehensive scoring method. The score ranges from 0 to 100, with the scoring rules being: bit error rate weighted at 0.3, latency at 0.25, jitter at 0.25, and packet loss rate at 0.2. The mapping relationship between the comprehensive score and the quality status level is as follows: 90-100 corresponds to excellent status, 75-89 to good status, 60-74 to average status, 40-59 to poor status, and 0-39 to very poor status. The status identification result serves as the initial state input for multi-step prediction.

[0147] Multi-step prediction computation employs a Markov chain state evolution algorithm for forward inference. The prediction calculator uses the current link quality state as the initial state vector, a 5-dimensional unit vector where the element corresponding to the current state position is 1, and the elements at other positions are 0. State evolution is achieved through matrix multiplication of the initial state vector and the state transition probability distribution matrix, yielding the state probability distribution vector for the next time step. Repeated matrix multiplication yields the state distribution for multiple subsequent time steps. The prediction step count is set to 10 steps, corresponding to the quality evolution over the next 10 sampling periods. The result of each step is a 5-dimensional probability vector, where each element represents the probability of the corresponding quality state occurring.

[0148] The evolutionary sequence construction employs a time-series storage and indexing management strategy. The sequence builder stores the state probability distribution vectors obtained from multi-step predictions in chronological order within an evolutionary sequence array, with the array length equal to the number of prediction steps. Each array element contains three fields: a time index, a state probability distribution vector, and a maximum probability state identifier. The time index is incremented starting from 1, the state probability distribution vector is stored as a floating-point array, and the maximum probability state identifier records the quality state code with the highest probability at the current time. The evolutionary sequence supports both random access and sequential traversal modes, providing a data foundation for subsequent quality state transition path analysis.

[0149] The quality state transition path traversal employs a depth-first search algorithm to enumerate all paths. The path traverser starts from the beginning of the evolution sequence and recursively explores all quality states at each time step, constructing a complete state transition path tree. Each path records the complete state sequence from the current time step to the prediction termination time, and the path length is equal to the number of prediction steps. The number of paths grows exponentially with the number of prediction steps. To control computational complexity, a probability threshold pruning strategy is used during the traversal process, ignoring state transition branches with a probability less than 0.01. Valid paths are stored in a path set, and each path element contains three attributes: state sequence, path probability, and cumulative transition cost.

[0150] The calculation of the quality status improvement magnitude is based on the numerical difference analysis of the status level. The improvement magnitude calculator compares the quality status levels of adjacent moments in the path and calculates the numerical difference of the status levels. A difference is negative when the status level improves, positive when the status level decreases, and zero when the status level remains unchanged. The cumulative improvement magnitude is calculated by weighting and summing the improvement magnitudes of all moments within the path. The weighting coefficient decreases over time, with a weight of 1.0 for recent moments and gradually decreasing to 0.5 for longer-term moments. The cumulative improvement magnitude reflects the overall quality improvement effect of the path; the smaller the value, the more significant the quality improvement.

[0151] State transition cost calculation is based on resource consumption assessment of transition operations. The cost calculator defines a fixed operation cost for each state transition, and the cost of transitions between states is proportional to the difference in state level. Adjacent state transitions cost 1 unit, transitions across one level cost 2 units, transitions across two levels cost 4 units, and so on. The switching operation itself incurs additional switching costs, the value of which is set according to the switching complexity: a simple switch costs 5 units, and a complex switch costs 10 units. The total path cost is calculated by accumulating the costs of all state transitions and switching operations within the path; a smaller cost value indicates a lower path execution cost.

[0152] The confidence level for handover decisions is obtained using historical handover success rate statistics. A confidence calculator analyzes the execution results of the most recent 100 handover operations, counting the number of successful and failed handovers. A successful handover is defined as an improvement in link quality after the handover, while a failure is defined as no improvement or deterioration in quality after the handover. The confidence level is equal to the number of successful handovers divided by the total number of handovers, ranging from 0.0 to 1.0; a higher value indicates a higher reliability of the handover decision. The confidence level calculation considers time decay, giving higher weight to recent handover operations and lower weight to older operations, ensuring that the confidence level reflects the current decision-making capability.

[0153] The expected return calculation employs a multi-factor weighted evaluation model that integrates returns and risks. The return calculator receives three input parameters: the magnitude of quality improvement, the cost of state transition, and the confidence level of the switching decision. The expected return is calculated through weighted averages. The return calculation method is: magnitude of improvement multiplied by confidence level, minus the cost of transition multiplied by a risk coefficient. The risk coefficient is set to 0.8 to balance the weighting of returns and costs. A positive expected return indicates that the switching operation is expected to be beneficial, while a negative value indicates that the switching operation is expected to be unfavorable; the magnitude reflects the degree of benefit. The return calculation process includes uncertainty compensation, adding a risk penalty for cases with low confidence levels.

[0154] The expected return calculation for each switching moment employs a time-sliding window analysis strategy. The moment-based return calculator calculates the expected return value for each moment in the prediction sequence, sequentially calculating moments from 1 to the prediction step number. The return calculation for each moment considers the complete evolution path from that moment to the end of the sequence, obtaining the expected return through a path probability-weighted average. The moment-based return values ​​are stored in a return array, where the array index corresponds to the moment number, and the array element is the expected return value for that moment. The return calculation process uses dynamic programming optimization to avoid repeatedly calculating the return values ​​for the same sub-path.

[0155] The optimal handover timing selection employs a global maximum search algorithm to determine the best moment. The timing optimizer traverses all moment reward values ​​in the reward array and identifies the moment index with the maximum expected reward value. When multiple moments have the same maximum reward value, the moment with the smaller index, i.e., the earlier handover moment, is prioritized to ensure timely response to quality degradation. The optimal timing result includes three output fields: optimal handover moment, corresponding expected reward value, and recommended execution path. The moment value is converted to actual time by multiplying by the sampling period, typically 1 second.

[0156] A second aspect of the present invention provides a 2M unframed dual-path protection intelligent handover system based on redundancy check, comprising:

[0157] The first unit is used to acquire the 2M unframed bitstream carried by each of the two transmission links, and to perform clock recovery and frame boundary alignment processing on the 2M unframed bitstream to obtain dual bitstream data.

[0158] The second unit is used to perform real-time verification on the dual-stream data based on the cyclic redundancy check algorithm, generate verification results for each link, and statistically analyze the frequency and duration of bit error events within a sliding time window based on the verification results to form a bit error feature vector characterizing the link quality status.

[0159] The third unit is used to map the bit error feature vector to a multi-dimensional quality space and establish a link quality trajectory. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal that can indicate changes in link quality is extracted.

[0160] The fourth unit is used to construct a phase switching strategy based on the quality degradation warning signal and the service status of the current working link, and to dynamically adjust the switching decision threshold according to the evolution rate of the link quality trajectory during the pre-switching preparation phase.

[0161] The fifth unit is used to input the switching decision threshold and the quality degradation warning signal into the neural network evaluator, determine the switching timing based on the output of the neural network evaluator and execute the link switching operation; and perform online training on the neural network evaluator in combination with the switching execution effect to achieve adaptive optimization of the switching strategy.

[0162] A third aspect of the present invention provides an electronic device, comprising:

[0163] processor;

[0164] Memory used to store processor-executable instructions;

[0165] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0166] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0167] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A 2M unframed dual-path protection intelligent handover method based on redundancy check, characterized in that, include: The 2M unframed bitstreams carried by each of the two transmission links are obtained, and clock recovery and frame boundary alignment are performed on the 2M unframed bitstreams to obtain dual bitstream data. The dual-stream data are verified in real time based on the Cyclic Redundancy Check (CRC) algorithm to generate verification results for each link. Based on the verification results, the frequency and duration of bit error events are counted within a sliding time window to form a bit error feature vector characterizing the link quality status. The bit error feature vector is mapped to a multi-dimensional quality space and a link quality trajectory is established. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal that can indicate changes in link quality is extracted. Based on the quality degradation warning signal and the service status of the current working link, a phase switching strategy is constructed, and the switching decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory during the pre-switching preparation phase. The switching decision threshold and the quality degradation warning signal are input into the neural network evaluator. The switching timing is determined based on the output of the neural network evaluator, and the link switching operation is executed. The neural network evaluator is trained online in combination with the switching execution effect to achieve adaptive optimization of the stage switching strategy.

2. The method according to claim 1, characterized in that, The dual-stream data are verified in real time using a cyclic redundancy check (CRC) algorithm, generating verification results for each link. Based on these verification results, the frequency and duration of bit error events are statistically analyzed within a sliding time window to form a bit error feature vector characterizing the link quality status, including: The dual-stream data is segmented according to a preset data block granularity, and a verification operation is performed on each data block based on the characteristic polynomial of the cyclic redundancy check algorithm. The verification pass status and verification failure status of each data block are identified according to the verification result. Based on the transformation relationship between the verification pass status and the verification failure status in the time series, the start time and end time of the bit error event are extracted. Based on the start time and the end time, a time distribution sequence of bit error events is constructed within a sliding time window. By calculating the time interval between adjacent bit error events in the time distribution sequence and combining the reliability of the verification results of the cyclic redundancy check algorithm, continuous bit error intervals and discrete bit error intervals are identified. Based on the ratio of the continuous error interval to the discrete error interval, a concentration feature value reflecting the error distribution pattern is calculated. Combined with the frequency of the error event within the sliding time window and the duration of the error event from the start time to the end time, an error feature vector containing the frequency of occurrence, the duration, and the concentration feature value is formed.

3. The method according to claim 2, characterized in that, By calculating the time interval between adjacent bit error events in the time distribution sequence and combining the reliability of the verification results of the cyclic redundancy check algorithm, continuous bit error intervals and discrete bit error intervals are identified, including: For each bit error event in the time distribution sequence, the reliability of the verification result used to evaluate the credibility of the bit error event is calculated based on the verification polynomial characteristics of the cyclic redundancy check algorithm and the bit stream transmission rate. In the time distribution sequence, the time interval between adjacent bit error events is calculated, and the time interval is compared with a preset time clustering judgment benchmark. When the time interval is less than the time clustering judgment benchmark, the corresponding adjacent bit error events are marked as time clustering state. When the time interval is greater than or equal to the time clustering judgment benchmark, the corresponding adjacent bit error events are marked as time discrete state, thus obtaining a preliminary bit error event state distribution. The reliability of the verification result is used as a weighting factor for the time clustered state and the time discrete state. Adjacent bit error events marked as time clustered states in the preliminary bit error event state distribution are filtered by reliability weighting to form a set of clustered events after reliability correction. Traverse the time distribution sequence, merge the time-continuous bit error events in the reliability-corrected set of clustered events into continuous bit error intervals, and classify the bit error events in the time distribution sequence that do not belong to the continuous bit error intervals into discrete bit error intervals.

4. The method according to claim 1, characterized in that, The bit error feature vector is mapped to a multi-dimensional quality space and a link quality trajectory is established. By analyzing the evolution trend of the link quality trajectory, quality degradation early warning signals that can indicate changes in link quality are extracted, including: Based on the occurrence frequency dimension, duration dimension, and concentration feature value dimension contained in the bit error feature vector, a coordinate system of multidimensional quality space is constructed, and the bit error feature vector corresponding to each sampling time is used as the quality state point in the multidimensional quality space. The quality state points at multiple consecutive sampling times are sequentially connected to form a link quality trajectory that reflects the evolution of link quality over time. The link quality trajectory is then segmented to identify the stationary evolution segment and the variable evolution segment within it. For the change and evolution segment in the link quality trajectory, calculate the movement direction vector and movement rate of the quality state point in the multi-dimensional quality space, and determine whether the link quality trajectory is moving towards the quality degradation area based on the movement direction vector; When the movement direction vector points to the quality degradation area and the movement speed exceeds a preset speed warning threshold, a quality degradation warning signal containing the movement direction vector and the movement speed is generated.

5. The method according to claim 1, characterized in that, Based on the quality degradation early warning signal and the service status of the current working link, a phased handover strategy is constructed, and the handover decision threshold is dynamically adjusted according to the evolution rate of the link quality trajectory during the pre-handover preparation phase, including: The mobile rate is extracted from the quality degradation warning signal, and combined with the service interruption tolerance in the current working link service status, the duration of the preparation phase and the decision threshold benchmark value corresponding to the phase handover strategy are determined; the pre-handover preparation phase is started according to the duration of the preparation phase, and the decision threshold benchmark value is set as the current handover decision threshold; During the pre-switching preparation phase, the real-time evolution rate is obtained by calculating the displacement and time interval between the quality state points corresponding to adjacent sampling times of the link quality trajectory. Based on the change of the evolution rate relative to the baseline value of the evolution rate when entering the pre-switching preparation phase, the handover decision threshold is dynamically corrected.

6. The method according to claim 1, characterized in that, The handover decision threshold and the quality degradation warning signal are input into a neural network evaluator. Based on the output of the neural network evaluator, the handover timing is determined and the link handover operation is executed. Online training of the neural network evaluator, considering the handover execution effect, includes: The handover decision threshold and the quality degradation warning signal are input into the neural network evaluator. Based on the neural network evaluator, cluster analysis is performed on the feature combinations in the historical handover records to extract the state transition probability distribution that characterizes the evolution of link quality. The handover decision confidence level reflecting the rationality of the current handover timing is calculated based on the state transition probability distribution. Based on the state transition probability distribution, the link quality evolution trend after handover is predicted. The optimal handover sequence is determined by combining the handover decision confidence. When the handover decision confidence exceeds a preset confidence threshold and the predicted quality evolution trend meets the optimization objective, the link handover operation is executed, and the execution effect of the link handover operation is recorded. The execution effect includes handover response latency and quality improvement. An exponential forgetting factor is used to perform time-series weighting on the switching response delay and the quality improvement magnitude to obtain long short-term memory features that reflect the historical switching performance. The long short-term memory features and the prediction accuracy of the state transition probability distribution are used together as training criteria to train the neural network evaluator online.

7. The method according to claim 6, characterized in that, Based on the state transition probability distribution, the evolution trend of link quality after handover is predicted, and the optimal handover sequence is determined by combining the handover decision confidence, including: Based on the state transition probability distribution, the link quality status after the handover is predicted in multiple steps. Starting from the current link quality status, the quality status distribution at subsequent time moments is calculated sequentially according to the transition probability between states, so as to obtain an evolution sequence that characterizes the trend of link quality change after the handover. The quality state transition paths in the evolution sequence are traversed and calculated. The quality state improvement at each moment is combined with the state transition cost. The expected benefit value corresponding to each switching moment is calculated in combination with the switching decision confidence. The moment with the maximum expected benefit value is selected as the optimal switching sequence.

8. A 2M non-framed dual-path protection intelligent switching system based on redundancy check, used to implement the method of any one of claims 1-7, characterized in that, include: The first unit is used to acquire the 2M unframed bitstream carried by each of the two transmission links, and to perform clock recovery and frame boundary alignment processing on the 2M unframed bitstream to obtain dual bitstream data. The second unit is used to perform real-time verification on the dual-stream data based on the cyclic redundancy check algorithm, generate verification results for each link, and statistically analyze the frequency and duration of bit error events within a sliding time window based on the verification results to form a bit error feature vector characterizing the link quality status. The third unit is used to map the bit error feature vector to a multi-dimensional quality space and establish a link quality trajectory. By analyzing the evolution trend of the link quality trajectory, a quality degradation early warning signal that can indicate changes in link quality is extracted. The fourth unit is used to construct a phase switching strategy based on the quality degradation warning signal and the service status of the current working link, and to dynamically adjust the switching decision threshold according to the evolution rate of the link quality trajectory during the pre-switching preparation phase. The fifth unit is used to input the switching decision threshold and the quality degradation warning signal into the neural network evaluator, determine the switching timing based on the output of the neural network evaluator and execute the link switching operation; and perform online training on the neural network evaluator in combination with the switching execution effect to achieve adaptive optimization of the stage switching strategy.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images