A multi-network coexistence connection control method and system

Abstract

The present application relates to the technical field of multi-network switching control, in particular to a multi-network coexistence connection control method and system, in the present application, through Dijkstra algorithm, link nodes are taken as vertices and switching time consumption and failure records are taken as edge weights on the basis of statistics, minimum cost search is performed on candidate paths, the path selection process is directly constrained by cumulative switching time consumption and failure number, local decision is avoided according to instantaneous signal strength, bandwidth or time delay, Markov chain is introduced to model time sequence transfer relationship among switching time records, failure marks and throughput changes, transfer probability matrix is used to describe flow transfer intensity from stability to fluctuation and from unblocked to congestion among different links, and then flow transfer fields in the transfer probability matrix are compared with the value of each segment in the low-cost connection sequence, hierarchical adjustment and iterative rewriting are performed on the value according to the rising and falling trend.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-network switching control, and in particular to a multi-network coexistence connection control method and system. BACKGROUND

[0002] The technical field of multi-network switching control aims to establish an executable switching decision mechanism by quantifying network state parameters, so that the terminal maintains connection continuity in multiple coexisting networks, and by controlling the switching trigger condition, switching timing and network selection rules, the terminal avoids communication interruption, avoids repeated switching behavior and maintains stable transmission performance when switching between different networks.

[0003] The purpose of a multi-network coexistence connection control method is to establish executable network selection and switching rules by quantitatively determining the signal strength value, bandwidth measurement value, time delay measurement value and congestion index of multiple available networks, so that the terminal maintains the continuity of the communication process in the multi-network coexistence scenario, and maintains a stable connection state when the network conditions change, thereby reducing the switching failure probability, reducing the number of repeated switches and improving the data transmission efficiency.

[0004] The prior art mainly establishes threshold determination and rule sets around signal strength values, bandwidth measurement values, time delay measurement values and congestion indicators in the multi-network coexistence connection control scenario, and the decision action is mostly based on the parameter state in the current moment or short time window, lacking explicit characterization of long-period residence behavior and historical switching trajectory, making it difficult to distinguish long-term stability differences when short-time parameters between different networks are close. The prior art only records single switching results and success status in the expression of switching cost, does not construct a continuous cost chain structure across nodes, and is difficult to identify high cumulative cost link combinations at the path level. In the state evolution description, it is usually based on simple state markers or single-step feedback, lacking a transition probability matrix structure to quantify the strength of state flow in multiple switching rounds, limiting the ability to specifically constrain fault-prone links and links with large fluctuations. In the cost update strategy, static rules and coarse-grained threshold adjustments are mostly used, without tracking long-term failure records and throughput changes, making it difficult to trigger reasonable convergence actions when the network environment degrades slightly multiple times, and making it prone to come-and-go switching phenomena between multiple candidate links in long-term operation. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and propose a multi-network coexistence connection control method and system.

[0006] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: a multi-network coexistence connection control method, comprising the following steps:

[0007] S1: based on the multi-network link group residence count, the residence time and the switching frequency of the Ethernet link, the Wi-Fi link and the 5G link are arranged in parallel, the three records are presented in the difference of the magnitude, and the steady state count structure is generated;

[0008] S2: based on the steady state count structure, the Dijkstra algorithm is used to check the threshold state of the link switching contact, and the switching time record and the failure record of the three links are concatenated in the order of the link nodes to form a cost chain, and a low cost connection sequence is generated;

[0009] S3: based on the low cost connection sequence, the residence state of the next link node is confirmed to have completed the record, and after the switching action is completed, the time record, the fault mark and the throughput change are used to build a continuous record, the difference between the records is formed into a flow field by using Markov chain, and a transition probability matrix is generated;

[0010] S4: based on the low cost connection sequence and the transition probability matrix, the numerical relationship of each flow field is compared layer by layer, and the link value is rearranged in the upgrading direction and the downgrading direction, so that the new link value is written into the existing cost table, and an updated cost structure is generated;

[0011] S5: based on the updated cost structure, the residence count of the current link is reset, and the link range required for judgment in the next stage is circled in the multi-network link group, so that the range content and the updated cost structure form a corresponding input, and a subsequent link processing set is generated.

[0012] As a further scheme of the application, the steady state count structure includes residence time, switching times, and link difference, the low cost connection sequence includes starting link node, target link node, and link value arrangement order, the transition probability matrix includes state node group, transition amount group, and record interval group, the updated cost structure includes updated link value, updated path value, and updated node value, and the subsequent link processing set includes link selection range, link value corresponding field, and residence count basic value.

[0013] As a further scheme of the application, the specific steps for generating the steady state count structure are as follows:

[0014] Based on the multi-network link group residence count, the residence time and the switching frequency of the Ethernet link, the Wi-Fi link and the 5G link are arranged in parallel, the three values are divided into continuous paragraphs according to the time line, and the contrastable magnitude sequence is arranged according to the paragraph difference, and the residence difference data group is generated;

[0015] Based on the residence difference data group, the difference amplitudes of each paragraph in the sequence are sequentially accumulated, and the interval boundary is drawn according to the accumulation result, the interval content is compressed into a single count framework, and the structure field is written, and the steady state count structure is generated.

[0016] As a further scheme of the present application, the specific steps for generating the low-cost connection sequence are as follows:

[0017] Based on the steady-state count structure, the Dijkstra algorithm is adopted, the threshold field of the link switching contact is read, and the read value is compared with the threshold reference item by item, and then the comparison result is written into the contact sequence field according to the arrangement order of the link to generate the contact column field group;

[0018] Based on the contact column field group, the switching time consumption record and the failure record of the three links are taken out one by one, and are sequentially connected into the continuous generation value column according to the order of the contact sequence to generate the generation value connection column;

[0019] Based on the generation value connection column, the generation values are arranged in order of magnitude, and the arranged link nodes are sequentially connected to form the path column, and then the path column is written into the sequence structure item to generate the low-cost connection sequence.

[0020] As a further scheme of the present application, the Dijkstra algorithm first establishes the cumulative cost record on the multi-network link node, sets the cumulative cost of the current link node to zero, sets the cumulative cost of the remaining link nodes to infinity, and sets the predecessor record bit for all nodes, then selects the node in the unprocessed node according to the cumulative cost, reads the generation value item in the generation value connection column starting from the node, and adds the cumulative cost and the generation value in the item to form a candidate cumulative cost, and then compares the candidate cumulative cost with the cumulative cost currently recorded in the adjacent node, when the candidate value is smaller, the candidate value is used to overwrite the original record and write into the predecessor record bit, then continue to select the node in the unprocessed node according to the cumulative cost and repeat the above cost update process until the cumulative cost record update of all candidate nodes is completed, then start from the candidate node with the lowest cumulative cost and backtrack the link node in reverse order according to the predecessor record bit, and write the node sequence obtained by backtracking into the path column field to form the link node sequence sorted by the cumulative cost.

[0021] As a further scheme of the present application, the specific steps for generating the transition probability matrix are as follows:

[0022] Based on the low-cost connection sequence, the switching time record, the failure marker and the throughput change are sorted according to the time point, and the sorted multiple records are continuously spliced according to the order to form the change segment sequence to generate the residence record column;

[0023] Based on the residence record column, the switching time record, the failure marker and the throughput change are time-indexed and aligned, and the aligned multiple records are spliced according to the time sequence to form the continuous change segment sequence to generate the change segment column;

[0024] Based on the change segment column, a Markov chain is used to perform splitting on the difference between adjacent change segments and rearrange the split difference values according to the change direction as a flow field column, and then the field column is written into a structure entry to generate a transition probability matrix.

[0025] As a further scheme of the present application, the Markov chain first divides each change segment in the change segment column according to the bandwidth, time delay, packet loss rate and change direction of the throughput to obtain a state level sequence number, converts the continuous change segment sequence into a state sequence number sequence, then reads the state sequence numbers of adjacent two segments in sequence from the first segment, takes the state of the previous segment as a starting state and the state of the next segment as a target state, performs counting plus one on the corresponding starting state row and target state column position in the state transition counting table, repeats the counting process for the complete change segment column and traverses all adjacent segment combinations, then sums the transition counts of each row in the counting table, divides the counts of each target state in the row by the sum to obtain the transition probability from the state to each target state, writes the state levels corresponding to the row number and column number into the matrix index, and writes the obtained transition probability into the matrix unit in row-column order to form a transition probability matrix corresponding to the change segment column.

[0026] As a further scheme of the present application, the specific steps for generating the updated cost structure are as follows:

[0027] Based on the low-cost connection sequence and the transition probability matrix, the values of the flow field are split in sequence according to the paragraphs, the split change amounts are compared with the cost values of the corresponding link nodes item by item, the ascending and descending trends are collected into field sets respectively, then the set contents are arranged into replaceable field item groups in the order of the links to generate a cost adjustment item set;

[0028] Based on the cost adjustment field group, the existing cost values between the links are taken out one by one, the taken-out items are replaced with the field items according to their respective positions, and then the replaced items are written back into the table in the original cost table order to form a continuous and connected cost value structure, thereby generating an updated cost structure.

[0029] As a further scheme of the present application, the specific steps for generating the subsequent link processing set are as follows:

[0030] Based on the updated cost structure, the resident count of the current link is emptied, then the links that need to participate in the judgment in the next stage are screened according to the link state after being emptied, and the screened links are rearranged into a sequence in the processing order to generate a next-stage link set;

[0031] Based on the next-stage link set, the link sequence is compared with the updated cost structure item by item, and the compared fields are re-collected into an input group, then written into a set item to generate a subsequent link processing set.

[0032] A multi-network coexistence connection control system for executing the multi-network coexistence connection control method, the system comprising:

[0033] The residence counting module: based on the multi-network link group, the residence time of Ethernet, WiFi and 5G is sequentially taken out and arranged in the link order to generate a steady-state counting structure;

[0034] The cost construction module: based on the steady-state counting structure, the switching contact threshold is read and compared with the threshold reference and written into the contact sequence, and the switching time and failure record of the three links are concatenated and combined with Dijkstra algorithm to form a cost value chain to generate a low-cost connection sequence;

[0035] The sequence generation module: based on the low-cost connection sequence, the next link node residence record is spliced in time sequence with the time after switching, failure and throughput to form a continuous record, and the difference between records is formed into a change field and written into a matrix entry combined with Markov chain to generate a transition probability matrix;

[0036] The cost update module: based on the low-cost connection sequence and the transition probability matrix, the flow field is compared with the existing cost value, rearranged into a replaceable cost value segment and written into a cost table to generate an updated cost structure;

[0037] The input generation module: based on the updated cost structure, the current link residence count is emptied, the links needed to be judged in the next stage are screened to form a sequence, and the sequence field is collected into an input entry corresponding to the updated cost value to generate a subsequent link processing set.

[0038] Compared with the prior art, the advantages and positive effects of the present application are:

[0039] In the present application, Dijkstra algorithm is used to perform minimum cost search on candidate paths based on the statistics, taking link nodes as vertices and switching time and failure record as edge weights, so that the path selection process is directly constrained by cumulative switching time and failure number, avoiding local decision making based only on instantaneous signal strength, bandwidth or delay;

[0040] In the present application, Markov chain is introduced to model the time sequence transition relationship between the time after switching, failure mark and throughput change, the transition probability matrix is used to describe the flow transition intensity from stability to fluctuation and from smooth to congestion between different links, and the flow field in the transition probability matrix is compared with the cost value of each segment in the low-cost connection sequence, the cost value is adjusted and iteratively rewritten according to the rising and falling trend, so that the updated cost structure simultaneously reflects long-term residence stability, cumulative switching cost of multi-hop path and state transition trend in numerical distribution, and the residence count is reset and the next stage link is screened based on the updated cost structure.

[0041] In the present application, in the multi-round judgment, the participation opportunity of the high failure rate path is compressed, the short cycle round trip switching phenomenon is inhibited, the number of sudden interruptions caused by the accumulation of implicit faults is reduced, and the residence time of the better link is prolonged under the premise of maintaining the flexibility of network selection, which brings comprehensive gains to the connection continuity and transmission stability in the multi-network coexistence environment. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 The workflow diagram of the present application is shown in the figure.

[0043] Figure 2 The system flowchart of the present application is shown in the figure. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0045] Example:

[0046] Please refer to Figure 1 The present application provides a technical scheme: a multi-network coexistence connection control method, comprising the following steps:

[0047] S1: based on the residence count of multi-network link group, the residence time and switching frequency of Ethernet link, Wi-Fi link and 5G link are arranged side by side, so that the three records show the difference in magnitude, and a steady-state count structure is generated;

[0048] S2: based on the steady-state count structure, using Dijkstra algorithm, checking the threshold state of link switching contact, and connecting the switching time record and failure record of the three links in sequence according to the link node order to form a cost chain, and generating a low-cost connection sequence;

[0049] S3: based on the low-cost connection sequence, confirming that the residence state of the next link node has been recorded, and constructing a continuous record with time record, fault mark and throughput change after the switching action is completed, using Markov chain to make the difference between records form a flow field, and generating a transition probability matrix;

[0050] S4: based on the low-cost connection sequence and the transition probability matrix, comparing the numerical relationship of each flow field layer by layer, and reorganizing the link value with the upgrade trend and the downgrade trend, so that the new value is written into the existing cost table to generate an updated cost structure;

[0051] S5: resetting the residence count of the current link based on the updated cost structure, and circling the range of links that need to be involved in the next stage of judgment in the multi-network link group, so that the range content corresponds to the input of the updated cost structure, and a subsequent link processing set is generated.

[0052] The steady-state count structure includes a residence duration amount, a switching frequency amount, and a link difference amount, the low-cost connection sequence includes a starting link node, a target link node, and a cost value arrangement order, the transition probability matrix includes a state node group, a transition amount group, and a record interval group, the updated cost structure includes an updated link cost value, an updated path cost value, and an updated node cost value, and the subsequent link processing set includes a link selection range, a cost value corresponding field, and a residence count basis amount.

[0053] The specific steps for generating the steady-state count structure are as follows:

[0054] Based on the residence count of the multi-network link group, the residence duration and switching frequency of the Ethernet link, WiFi link, and 5G link are synchronized and sorted, the three values are divided into continuous paragraphs according to the timeline, and the magnitude sequence that can be compared is arranged according to the paragraph difference, and a residence difference data group is generated;

[0055] Based on the residence difference data group, the difference amplitudes of each paragraph in the sequence are sequentially accumulated, and the interval boundaries are drawn according to the accumulation results, the interval content is compressed into a single count framework, and written into a structure field, and a steady-state count structure is generated;

[0056] Based on the multi-network link group residence count process, the K-means clustering algorithm is used to perform clustering processing on the link residence duration data and switching frequency data, the clustering operation is completed by setting the parameter k to three, the initialization method to k-means++, the maximum iteration number to three hundred, and the distance measurement method to Euclidean distance, the input data is arranged in time order and organized in three column structures, the first column is the time point sequence, the second column is the residence duration value, and the third column is the switching frequency value, the cluster number of each time point is obtained after clustering execution, the paragraph division action is performed by scanning the cluster number sequence in time order, the number change position is used as the paragraph boundary, then the difference amount calculation is performed on adjacent paragraphs, the difference calculation is completed by performing absolute value summation on the residence duration difference value and the switching frequency difference value, and the magnitude sequence is generated by sorting all paragraph difference amounts according to the numerical value, and the residence difference data group is formed by the sorted paragraph difference amounts;

[0057] Based on the resident difference data set, the prefix and accumulation algorithm is used to generate the difference amplitude accumulation sequence, the accumulation action is completed by establishing a sequential array consistent with the number of paragraphs and taking the first paragraph difference as the first item, and the subsequent items are written by adding the current difference to the previous accumulation item, after the accumulation is completed, the equal interval interval division algorithm is used to divide the accumulation sequence into five intervals, the interval division action is completed by calculating the interval width with the interval span from the minimum accumulation value to the maximum accumulation value and the fixed interval number, then the interval number is determined according to the position of the accumulation value in the interval range, and then the compression processing is performed on the paragraph content in the interval, the compression processing adopts the Hash mapping compression method and generates a 64-bit integer with a fixed seed value 14652 to write into the count framework field, and a unified count framework is formed by continuous writing and a stable count structure is generated.

[0058] The specific steps of generating the low-cost connection sequence are as follows:

[0059] Based on the stable count structure, the Dijkstra algorithm is used to read the threshold field of the link switching contact, and the read value is compared with the threshold reference item by item, then the comparison result is written into the contact sequence field according to the arrangement order of the link, and the contact column field group is generated;

[0060] Based on the contact column field group, the switching time records and failure records of the three links are taken out one by one, and are sequentially connected into the continuous generation value column according to the order of the contact sequence, to generate the generation value connection column;

[0061] Based on the generation value connection column, the generation value is arranged in order of magnitude, and the arranged link nodes are sequentially connected to form the path column, then the path column is written into the sequence structure item to generate the low-cost connection sequence;

[0062] Based on the stable count structure, the Dijkstra algorithm is used to read the threshold field of the link switching contact, the reading action is performed in a sequential scanning manner and is positioned with a fixed step length, then the threshold reference array is used to perform item-by-item comparison, the comparison action is completed by establishing a comparison array with a length of three, and the difference between the read value and the threshold reference is taken as the comparison item, the difference is written into the comparison array in array order, and the comparison result is written into the contact sequence field as an integer value, the writing action is completed by sequential appending, to generate the contact column field group;

[0063] Based on the contact column field group, the sequential connection processing method is used to take out the switching time records and failure records of the three links one by one, the taking-out action is completed by performing index scanning on the record array, the scanning step length is one and the record position is used as the index, then the connection action is performed according to the order of the contact sequence field, the connection action is completed by concatenating the time value and the failure value into a single generation value record with the string concatenation command concat, and writing the generation value record into the connection array according to the original sequence number, the array writing is completed with the append command, and the generation value connection column is generated.

[0064] Based on the generation value connection column, the fixed sorting processing algorithm is used to perform magnitude sorting on the generation value. The sorting action is completed by constructing a sorting key array and calling the sort command in ascending order. The corresponding link nodes after sorting are spliced in sequence using the sequential connection method. The node splicing action is written into the path column field in the form of a one-way chain structure. The write action is performed using the push command. Then the path column is written into the sequence structure item as a whole, generating a low-cost connection sequence.

[0065] Dijkstra algorithm, first establish cumulative cost records on multi-network link nodes, set the cumulative cost of the current link node to zero, set the cumulative cost of the remaining link nodes to infinity, and set the predecessor record bit for all nodes. Then select a node from the unprocessed nodes based on the cumulative cost, read the generation value entries in the generation value connection column starting from the node, and add the cumulative cost and the generation value in the entry to form a candidate cumulative cost. Then compare the candidate cumulative cost with the cumulative cost currently recorded in the adjacent node. When the candidate value is smaller, overwrite the original record with the candidate value and write the predecessor record bit. Then continue to select a node from the unprocessed nodes based on the cumulative cost and repeat the above cost updating process until all candidate nodes complete the cumulative cost record updating. Then start from the candidate node with the lowest cumulative cost and backtrack the link nodes in reverse order according to the predecessor record bit. Write the nodes obtained by backtracking in sequence to the path column field, so that the path column forms a link node sequence sorted by cumulative cost.

[0066] Dijkstra algorithm, according to the formula:

[0067]

[0068] Wherein: represents the minimum switching cost from the source contact node to the contact node in a multi-network coexistence connection control method, represents the minimum switching cost from the source contact node to the contact node in a multi-network coexistence connection control method, represents the basic link switching weight from the contact to the contact in a multi-network coexistence connection control method, represents the influence coefficient of the threshold deviation ratio in a multi-network coexistence connection control method, represents the influence coefficient of the steady-state fluctuation index in a multi-network coexistence connection control method, represents the service priority factor in a multi-network coexistence connection control method Influence coefficient, This refers to a contact in a multi-network coexistence connection control method. The threshold deviation ratio, This refers to a contact in a multi-network coexistence connection control method. The threshold, This refers to a contact in a multi-network coexistence connection control method. The count read value, This refers to a contact in a multi-network coexistence connection control method. The steady-state fluctuation index, This refers to a contact in a multi-network coexistence connection control method. Within the observation window Item count read value, This refers to a contact in a multi-network coexistence connection control method. The average count reading within the observation window. This represents the observation window length in a multi-network coexistence connection control method. This refers to a contact in a multi-network coexistence connection control method. Business priority factor, This represents the source contact node for link switching in a multi-network coexistence connection control method. This represents the contact node with the lowest handover cost that has been determined in a multi-network coexistence connection control method. This represents the contact node with the minimum handover cost to be updated in a multi-network coexistence connection control method. This refers to an operator used in a multi-network coexistence connection control method to select the smaller value from two inputs;

[0069] Execution process: First, select the source contact node. And initialize the minimum switching cost for multiple touch nodes, corresponding Set to zero and the remaining multiple terms Initialize to positive infinity, and then select the set with the smallest value from the undetermined set. Contact nodes As the iteration base point, for and Multiple contact nodes with linked connections Process sequentially, reading the contacts from the steady-state counting structure. threshold and count reading value Calculate the threshold deviation ratio Read the contact points from the observation window Multiple historical count values Calculate the average value and further calculate the steady-state fluctuation index , then read the service priority level of the contact and map to the service priority factor , then read three initial impact inputs from the configuration unit , , , , , , then read the basic link switching weight and , , , , , , , calculate the comprehensive link switching weight, and add the comprehensive weight to to obtain the candidate path cost and compare it with the current record , select the smaller value from the two through operator to update , and then move the updated contact node into the determined set and continue to select new Repeat the above calculation steps until the undetermined set is empty, and finally obtain the minimum switching cost of multiple contacts based on the formula in the multi-network coexistence environment.

[0070] The specific steps of generating the transition probability matrix are as follows:

[0071] Based on the low-cost connection sequence, sort the switching time record, fault mark and throughput change by time point, and continuously splice the sorted multiple records in the order to form a change segment sequence to generate a residence record column;

[0072] Based on the residence record column, align the switching time record, fault mark and throughput change by time index, and splice the aligned multiple records in time sequence to form a continuous change segment sequence to generate a change segment column;

[0073] Based on the change segment column, use Markov chain to perform splitting on the difference between adjacent change segments, rearrange the split difference values into a flow field column according to the change direction, and then write the field column into a structure entry to generate a transition probability matrix;

[0074] Based on the low-cost connection sequence, the fixed time sorting algorithm is used to sort the post-switch time length record, fault mark and throughput change by time point, and the sorting action is completed by establishing three column arrays, the first column stores the time point value, the second column stores the corresponding time length record, and the third column stores the corresponding fault mark and throughput change, then the sort command is called for the array and the time point column is used as the sorting key to perform ascending sorting, the sorting command is implemented by the compare function and the compare(a, b) returns a.time minus b.time as the judgment basis, after sorting, the continuous splicing action is performed on the sorted array, the splicing action is completed by establishing an empty sequence and appending the time point corresponding time length record, fault mark and throughput change in a fixed order by the append command, then the change segment sequence is formed and written into the record column field, and the resident record column is generated;

[0075] Based on the resident record column, the fixed index alignment method is used to perform time index alignment on the post-switch time length record and fault mark and throughput change, the alignment action is completed by establishing a time index array, each item in the array takes the time point as the key and the corresponding record as the value, then the time index array is scanned in a traversal manner and three types of records are taken out in index order, the getItem command is called for the taking out operation and the time point is used as the parameter, then the aligned multiple records are written into the newly created sequence in a fixed order splicing manner, the splicing action is completed by the string concatenation command concat, the continuous change segment sequence is formed and written into the structure unit, and the change segment column is generated;

[0076] Based on the change segment column, the Markov chain is used to perform splitting on the difference between adjacent change segments, the splitting action is completed by establishing a difference array diff and diff[i] is the current segment value minus the last segment value, the difference calculation is performed by the subtract command and stored as an integer, then the difference value is rearranged in the change direction, the rearrangement action is completed by performing direction judgment on the difference value, the direction judgment is performed by the sign command and the greater than zero is assigned to the uplink mark, the less than zero is assigned to the downlink mark, and the equal to zero is assigned to the flat mark, then the difference value and the direction mark are written into the flow field column in the pair form, the writing action is performed by the push command, and the flow field column is written into the structure body entry, and the transition probability matrix is generated.

[0077] The Markov chain, first, in the change section column, for each change section, according to the bandwidth, the time delay, the packet loss rate and the throughput change direction, the state level sequence number is divided, the continuous change section sequence is converted into the state sequence number sequence, then from the first section, the state sequence number of the adjacent two sections is read in turn, the state of the previous section is taken as the starting state, the state of the next section is taken as the target state, the corresponding starting state row and the target state column position in the state transition count table are executed count plus one, the count process is repeated for the complete change section column and all adjacent section combinations are traversed, then the transition count of each row in the count table is summed up, the count of each target state in the row is divided by the count sum respectively, the transition probability from the indicated state to each target state is obtained, the state level corresponding to the row number and the column number is written into the matrix index, the obtained transition probability is written into the matrix unit in the row and column order, and the transition probability matrix corresponding to the change section column is formed;

[0078] The Markov chain, according to the formula:

[0079]

[0080] Among them: Indicates the weighted transition probability of the state corresponding to index Transition to index Corresponding state, Indicates the transition count number from index Occurrence to index , Indicates the comprehensive weight factor for correcting the transition influence degree from index To index , Indicates the total result of multiple transition counts with index As the starting point, Indicates the upper limit of the number of states that can be taken in the multi-network coexistence connection control method, Indicates the state index number as the transition starting point in the multi-network coexistence connection control method, Indicates the state index number as the transition target in the multi-network coexistence connection control method, Indicates the intermediate number for traversing all potential target indexes corresponding to index In the multi-network coexistence connection control method, Indicates the weight coefficient for adjusting the difference intensity index, Influence degree, Indicates the weight coefficient for adjusting the direction stability index weighting coefficient of influence degree, representing a network load index for adjusting in a multi-network coexistence connection control method, weighting coefficient of influence degree, representing a timing level index for adjusting in a multi-network coexistence connection control method, weighting coefficient of influence degree, representing an index to an index characteristic quantity of transition difference intensity, representing an index to an index characteristic quantity of transition direction stability, representing an index characteristic quantity of state corresponding network load degree, representing an index to an index characteristic quantity of transition timing level;

[0081] Execution process: after obtaining the state index sequence after splitting the change section, multiple adjacent index pairs are traversed one by one, and the number of transitions from index m to index n is recorded each time to update the count , then the count corresponding to the index n of the multiple index is summed up under the condition of fixing the index m to obtain , then the difference intensity index of each index pair is calculated according to the change section difference intensity and the direction stability index of each index pair is calculated according to the direction coding stability , then the network load index is formed according to the network load corresponding to each index extracted from the multi-network operation monitoring data , and the timing level index corresponding to the time delay level record of multiple transitions is calculated , then the initial evaluation quantity is read and normalized to obtain four weighting coefficients , , , for adjusting the relative influence degree of multiple indexes, and then and the four weighting coefficients are substituted into the calculation, and then the weighted normalization base is calculated for the given index , and then each item is substituted into the formula to obtain the weighted transition probability , and finally the weighted transition probability of all indexes The above steps are completed to obtain a complete weighted transition probability matrix for representing state transition behavior under multi-network coexistence connection.

[0082] The specific steps for generating the updated cost structure are as follows:

[0083] Based on the low-cost connection sequence and the transition probability matrix, the values of the flow field are sequentially split by paragraph, and the split change amount is compared with the cost value of the corresponding link node item by item. Then, the ascending and descending directions are used to form field sets, and the set contents are arranged into replaceable field item groups in link order to generate the cost adjustment item set.

[0084] Based on the cost adjustment field group, the existing cost values between links are taken out one by one, and the taken-out items are replaced with the field items according to their respective positions. Then, the replaced full items are written back to the table in the original cost table order to form a continuous and connected cost value structure, generating the updated cost structure.

[0085] Based on the low-cost connection sequence and the transition probability matrix, the flow field values are sequentially split by paragraph using the segmented splitting processing algorithm. The splitting action is completed by establishing a paragraph index array, which is scanned with a fixed step size. Then, the corresponding change amount is taken out from the flow field column for each paragraph using the slice command, and the taken-out change amount is compared with the link node cost value item by item. The comparison action is completed by the compare command, which takes the change amount and the cost value as input and returns the change direction mark and the difference amount mark. Then, the field sets are established according to the ascending and descending directions. The uplink records are written into the set with the pushUp command, the downlink records are written into the set with the pushDown command, and the flat records are written into the set with the pushFlat command. Then, the set contents are arranged according to the link order. The arrangement action is used to construct the field item group with the link index as the key and write it into the field item group sequence with the append command, generating the cost adjustment item set.

[0086] Based on the cost adjustment field group, the existing cost values between links are taken out one by one using the sequential replacement method. The taking-out action is completed by establishing a read pointer and scanning the cost value array with a fixed step size. Then, the taken-out items are replaced with the corresponding field items in the field item group using their position index as the key. The replacement action is completed with the set command and the field item value covers the original cost value item. After the replacement is completed, the write-back action is performed on all items in the original cost table order. The write-back action is completed with the writeBack command and the index is written back to the cost table structure in an incremental manner. Then, a continuous and connected cost value structure is formed, generating the updated cost structure.

[0087] The specific steps for generating the subsequent link processing set are as follows:

[0088] Based on the update cost structure, the resident count of the current link is emptied, and then the link to be involved in the next stage is screened according to the emptied link state, and the screened link is rearranged into a sequence according to the processing order to generate a next stage link set;

[0089] Based on the next stage link set, the link sequence is compared with the update cost structure item by item, and the compared fields are reassembled into an input group, and then written into the set item to generate a subsequent link processing set.

[0090] Based on the update cost structure, the reset processing algorithm is used to perform the emptying action on the current link resident count, and the emptying action is completed by establishing a count pointer and writing each count unit to zero value with the reset command, and then the screening is performed on the emptied link state, and the screening action is completed by constructing a state array and reading the state value item by item with the scan command, the scan command takes the link index as input and returns the corresponding state quantity, and then the state quantity is compared with the preset condition value, the comparison action is performed with the compare command, and the value greater than the condition value is recorded as the participating item, and the value less than the condition value is recorded as the non-participating item and written into the screening array, and then the rearrangement action is performed on the screened link, the rearrangement action is completed by using the fixed order rearrangement method and written into the newly created sequence with the reorder command in the ascending order of link number, to generate the next stage link set.

[0091] Based on the next stage link set, the sequence comparison method is used to perform item-by-item comparison between the link sequence and the update cost structure, and the comparison action is completed by establishing a double-pointer structure, the first pointer points to the current position of the link sequence, and the second pointer points to the corresponding position of the update cost structure, and then the difference value of the pointer position is read with the compare command, and the difference value is combined with the index number to record as a field item, and then the field item is written into the input group structure with the append command, and the writing action is performed in the original order of the link number, and then the input group is written into the set item with the writeEntry command, to generate the subsequent link processing set.

[0092] Please refer to Figure 2 A multi-network coexistence connection control system, the multi-network coexistence connection control system is used to execute the multi-network coexistence connection control method, the system comprises:

[0093] Resident count module: based on the multi-network link group, the resident time of Ethernet, WiFi and 5G and the switching frequency are taken out in turn and arranged into a reference field in the order of link, to generate a steady state count structure;

[0094] Cost construction module: based on the steady state count structure, the switching contact threshold value is read and compared with the threshold reference and written into the contact sequence, and then the switching time and failure record of the three links are concatenated according to the contact sequence and combined with Dijkstra algorithm to form a cost value chain, to generate a low-cost connection sequence.

[0095] The sequence generation module: based on the low-cost connection sequence, the next link node residence record is spliced into continuous records in time sequence with the time length after switching, failure and throughput, the difference between records is constituted into the change field and written into the matrix entry combined with Markov chain, and the transition probability matrix is generated;

[0096] The cost update module: based on the low-cost connection sequence and the transition probability matrix, the flow conversion field is compared with the existing cost value, rearranged into replaceable cost value segments and written into the cost table to generate the updated cost structure;

[0097] The input generation module: based on the updated cost structure, the current link residence count is emptied, the link formation sequence required to be judged in the next stage is screened, the sequence field is collected into the input entry corresponding to the updated cost value, and the subsequent link processing set is generated.

[0098] The above is only the preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can use the disclosed technical content to make changes or modifications into equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution content of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A method for controlling multi-network coexistence connections, characterized in that, Includes the following steps: S1: Based on the dwell time count of multiple network link groups, the dwell time and switching frequency of Ethernet links, Wi-Fi links and 5G links are sorted out in parallel to make the three records show comparable magnitude differences and generate a steady-state counting structure. S2: Based on the steady-state counting structure, the Dijkstra algorithm is used to check the threshold status of the link switching contact, and the switching time records and failure records of the three links are connected in the order of the link nodes to form a cost chain to generate a low-cost connection sequence. S3: Based on the low-cost connection sequence, confirm that the dwell status of the next link node has been recorded, and after the handover action is completed, construct a continuous record with duration record, fault mark and throughput change. Use Markov chain to make the difference between records form a flow field and generate a transition probability matrix. S4: Based on the low-cost connection sequence and transition probability matrix, compare the numerical relationships of each flow field layer by layer, and reorganize the cost value between links according to the upgrade direction and the downgrade direction, so that the new cost value is written into the existing cost table and an updated cost structure is generated. S5: Based on the updated cost structure, reset the current link dwell count, and delineate the range of links that need to participate in the judgment in the next stage in the multi-network link group, so that the range content and the updated cost structure form a corresponding input, and generate the subsequent link processing set; The specific steps for generating the low-cost connection sequence are as follows: Based on the steady-state counting structure, the Dijkstra algorithm is used to read the threshold field of the link switching contact, and the read value is compared with the threshold benchmark item by item. Then, the comparison result is written into the contact sequence field according to the arrangement order of the links to generate a contact column field group. Based on the contact point column field group, the switching time records and failure records of the three links are extracted one by one, and connected sequentially according to the contact point sequence to form a continuous cost value column, generating a cost value connection column. Based on the value connection column, the value is arranged in order of magnitude, and the arranged link nodes are connected sequentially to form a path column. Then, the path column is written into the sequence structure item to generate a low-cost connection sequence.

2. The multi-network coexistence connection control method according to claim 1, characterized in that, The steady-state counting structure includes dwell time, number of handovers, and link difference; the low-cost connection sequence includes starting link node, target link node, and cost value order; the transition probability matrix includes state node group, transition quantity group, and record interval group; the updated cost structure includes updated link cost value, updated path cost value, and updated node cost value; and the subsequent link processing set includes link selection range, cost value corresponding field, and dwell count base quantity.

3. The multi-network coexistence connection control method according to claim 1, characterized in that, The specific steps for generating the steady-state counting structure are as follows: Based on the dwell count of multiple network link groups, the dwell time and switching frequency of Ethernet links, WiFi links and 5G links are synchronously sorted out. The three values ​​are divided into continuous segments according to the timeline, and then the magnitude sequence can be compared by segment difference to generate dwell difference data group. Based on the aforementioned resident difference data set, the difference magnitudes of each segment within the sequence are accumulated sequentially, and then interval boundaries are drawn according to the accumulation results. The interval content is compressed into a single counting frame and written into the structure field to generate a steady-state counting structure.

4. The multi-network coexistence connection control method according to claim 1, characterized in that, The Dijkstra algorithm first establishes a cumulative cost record on multiple network link nodes, sets the cumulative cost of the current link node to zero, sets the cumulative cost of the remaining link nodes to infinity, and sets a predecessor record bit for all nodes. Then, it selects nodes from the unprocessed nodes based on the cumulative cost, reads the cost value entries in the cost value connection column starting from the node, and adds the cumulative cost to the cost value in the entry to form a candidate cumulative cost. Then, it compares the candidate cumulative cost with the cumulative cost of the current record of the adjacent node. When the candidate value is smaller, it overwrites the original record with the candidate value and writes it into the predecessor record bit. Then, it continues to select nodes from the unprocessed nodes based on the cumulative cost and repeats the above cost update process until all candidate nodes have completed the cumulative cost record update. Then, starting from the candidate node with the lowest cumulative cost, it backtracks the link nodes in reverse order according to the predecessor record bit, and writes the backtracked nodes sequentially into the path column field, so that the path column forms a link node sequence sorted by cumulative cost.

5. The multi-network coexistence connection control method according to claim 1, characterized in that, The specific steps for generating the transition probability matrix are as follows: Based on the low-cost connection sequence, the duration record, fault marker and throughput change after the switch are sorted by time point, and the sorted multiple records are continuously spliced ​​in chronological order to form a change segment sequence, generating a resident record column. Based on the resident record column, the time index is used to align the switching duration record with the fault mark and throughput change, and the aligned multiple records are spliced ​​together in time order to form a continuous change segment sequence, generating a change segment column; Based on the aforementioned change segment column, a Markov chain is used to split the differences between adjacent change segments and rearrange the split difference values ​​into a transition field column according to the change direction. Then, the field column is written into the structure entry to generate the transition probability matrix.

6. The multi-network coexistence connection control method according to claim 5, characterized in that, The Markov chain first divides each change segment in the change segment column into state level numbers according to the direction of change in bandwidth, latency, packet loss rate, and throughput, converting the continuous change segment sequence into a state number sequence. Then, starting from the first segment, it reads the state numbers of two adjacent segments in sequence, taking the previous segment as the starting state and the next segment as the target state. It increments the count by one at the corresponding starting state row and target state column positions in the state transition count table. The counting process is repeated for the complete change segment column and all adjacent segment combinations are traversed. Then, the transition counts of each row in the count table are summed, and the counts of each target state in the row are divided by the sum of the counts to obtain the transition probabilities from the indicated state to each target state. The row and column numbers corresponding to the state levels are written into the matrix index, and the obtained transition probabilities are written into the matrix cells in row and column order to form a transition probability matrix corresponding to the change segment column.

7. The multi-network coexistence connection control method according to claim 1, characterized in that, The specific steps for generating the update cost structure are as follows: Based on the low-cost connection sequence and transition probability matrix, the values ​​of the flow field are split into segments, and the changes after splitting are compared with the cost value of the corresponding link node item by item. Then, according to the rising and falling direction, the field sets are respectively compiled. Subsequently, the contents of the sets are organized into replaceable field item groups according to the link order to generate cost adjustment item sets. Based on the cost adjustment field group, the existing cost values ​​between links are extracted one by one, and the extracted entries and field items are replaced according to their respective positions. Then, all the replaced entries are rewritten back into the table in the original cost table order, thereby forming a continuous cost value structure and generating an updated cost structure.

8. The multi-network coexistence connection control method according to claim 1, characterized in that, The specific steps for generating the subsequent link processing set are as follows: Based on the updated cost structure, the dwell count of the current link is cleared, and then the links that need to participate in the judgment in the next stage are selected according to the cleared link status. The selected links are rearranged into a sequence according to the processing order to generate the link set of the next stage. Based on the next-stage link set, the link sequence is compared item by item with the update cost structure, and the fields obtained from the comparison are reassembled into an input group, which is then written into the set item to generate the subsequent link processing set.

9. A multi-network coexistence connection control system, used to execute the multi-network coexistence connection control method according to any one of claims 1-8, characterized in that, The system includes: Dwell time counting module: Based on multiple network link groups, the dwell time and switching frequency of Ethernet, WiFi and 5G are extracted sequentially and organized into comparable fields according to the link order to generate a steady-state counting structure; Cost construction module: Based on the steady-state counting structure, read the switching contact threshold and compare it with the threshold benchmark and write it into the contact sequence. Then, the switching time and failure records of the three links are concatenated according to the contact sequence and combined with the Dijkstra algorithm to form a value chain and generate a low-cost connection sequence. Sequence generation module: Based on the low-cost connection sequence, the next link node dwell record and the duration, fault, and throughput after the switch are concatenated into a continuous record in chronological order, and the difference between the records is used to form a change field and combined with the Markov chain to write into the matrix entry to generate the transition probability matrix. Cost update module: Based on the low-cost connection sequence and transition probability matrix, the flow field is compared with the existing cost value segment by segment, rearranged into replaceable cost value segments and written into the cost table to generate an updated cost structure; Input generation module: Based on the update cost structure, clear the current link dwell count and filter out the link formation sequence to be judged in the next stage. Then, aggregate the sequence field and update cost into input entries to generate the subsequent link processing set.

Images