Distributed Agent Ad Hoc Network Method and System Based on Dynamic Node Collaboration

Abstract

The present invention provides a distribution proxy ad-hoc network method and system based on dynamic node cooperation. First, a network status data set of all distribution nodes in the target network area is obtained, including real-time communication load, link stability, and resource occupancy characteristics. Based on this, dynamic cooperation topology information is generated, including a node cooperation relationship topology graph and a communication path weight table. According to the dynamic cooperation topology information, a distribution proxy policy set is generated, including an initial path allocation policy, a dynamic load adjustment rule, and a path switching trigger condition. The target distribution task is subjected to path optimization processing to obtain an optimized distribution path set, which is deployed to the corresponding distribution nodes to perform dynamic cooperation distribution. Moreover, the network status data set is updated according to the real-time feedback data of the distribution, which can improve the distribution efficiency and reliability.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular, to a distributed proxy ad-hoc network method and system based on dynamic node cooperation. Background Art

[0002] In existing network distribution systems, the cooperation methods of distribution nodes are often relatively fixed. Traditional methods usually perform data distribution based on a pre-set static topology structure and a fixed distribution strategy. For example, in some content distribution networks (CDNs), the cooperation relationships and communication paths between nodes are determined at the time of deployment, and it is difficult to make real-time adjustments according to the dynamic changes of the network environment.

[0003] When situations such as a sudden increase in node communication load, a decrease in link stability, or abnormal node resource occupancy occur in the network, the existing static distribution mechanism cannot respond in a timely manner. This leads to low data distribution efficiency, and problems such as packet loss and increased transmission delay may occur, seriously affecting the user experience. Moreover, in the face of a large-scale, complex and changeable network environment, the static distribution strategy cannot make full use of network resources and it is difficult to achieve efficient data distribution. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, embodiments of the present invention provide a distributed proxy ad-hoc network method based on dynamic node cooperation, and the method includes:

[0005] Obtain a network status data set of all distribution nodes in a target network area, where the network status data set includes real-time communication load characteristics of nodes, link stability characteristics between nodes, and node resource occupancy characteristics;

[0006] Generate dynamic cooperation topology information based on the network status data set, where the dynamic cooperation topology information includes a node cooperation relationship topology graph and a communication path weight table;

[0007] Generate a distributed proxy policy set according to the dynamic cooperation topology information, where the distributed proxy policy set includes an initial path allocation policy, a dynamic load adjustment rule, and a path switching trigger condition;

[0008] Perform path optimization processing on a target distribution task based on the distributed proxy policy set to obtain an optimized distribution path set, deploy the optimized distribution path set to corresponding distribution nodes to perform dynamic cooperation distribution operations, and update the network status data set based on real-time feedback data during the distribution process.

[0009] In another aspect, an embodiment of the present invention further provides a distributed proxy ad hoc network system based on dynamic node cooperation, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0010] Based on the above aspects, in the embodiment of the present invention, by obtaining the network status data set of all distribution nodes in the target network area and generating dynamic cooperation topology information and a distribution proxy policy set therefrom, the path optimization processing of the target distribution task is realized, breaking through the limitations of the traditional static distribution mechanism, and being able to dynamically adjust the node cooperation relationship and distribution path according to the real-time state of the network. By updating the network status data set in real time, the distribution system can quickly respond to changes in the network environment, effectively avoiding low data distribution efficiency caused by problems such as too high node communication load, unstable links or abnormal node resource occupation. Thus, the network resources are fully utilized, the reliability and efficiency of data distribution are improved, and a more effective solution is provided for data distribution in a large-scale, complex and changeable network environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a schematic flowchart of the execution process of the distributed proxy ad hoc network method based on dynamic node cooperation provided by an embodiment of the present invention.

[0012] Figure 2 is a schematic diagram of exemplary hardware and software components of the distributed proxy ad hoc network system based on dynamic node cooperation provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0013] The present invention will be specifically described below with reference to the accompanying drawings of the specification. Figure 1 is a schematic flowchart of the distributed proxy ad hoc network method based on dynamic node cooperation provided by an embodiment of the present invention. The distributed proxy ad hoc network method based on dynamic node cooperation will be introduced in detail below.

[0014] Step S110: Obtain the network status data set of all distribution nodes in the target network area, where the network status data set includes the real-time communication load characteristics of the nodes, the link stability characteristics between the nodes, and the node resource occupation characteristics.

[0015] In this embodiment, it is set in an enterprise-level data center network scenario. The data center consists of multiple computer rooms, and a large number of distribution nodes are distributed in each computer room. In order to achieve an efficient data distribution task, it is necessary to comprehensively obtain the network status data set of all distribution nodes first.

[0016] For the real-time communication load characteristics, taking node M as an example, it can be measured by monitoring the total data traffic received and sent within a unit time interval t. Assume that the received data traffic is represented by the letter r and the sent data traffic is represented by the letter s. Then, within the time interval t, the real-time communication load characteristics of node M can be expressed as the sum of r and s within time t. The real-time communication load characteristics of different nodes will vary due to their respective data traffic volumes.

[0017] For the link stability characteristics between nodes, taking the link between node M and its adjacent node N as an example. It can be comprehensively determined by statistically analyzing multiple factors such as the fluctuation of the signal strength and the bit error rate within a continuous time period T of the link. Assume that the signal strength is represented by the letter I, and within the time period T, its fluctuation range can be expressed as from I1 to I2; the bit error rate is represented by the letter e, and the bit error rate within time T is e. These factors are synthesized to characterize the link stability characteristics between nodes M and N. Different links have different stability characteristics due to factors such as physical media and the surrounding electromagnetic environment.

[0018] For the node resource occupancy characteristics, taking node M as an example, it can be obtained by checking aspects such as the usage rate of the central processing unit (CPU) and the memory usage rate of the node. Assume that the CPU usage rate is represented by the letter c and the memory usage rate is represented by the letter m. Then, the resource occupancy characteristics of node M include c and m. Different nodes have different resource occupancy characteristics due to the programs they run and the data tasks they process. Thus, a network status data set of all distribution nodes within the target network area is obtained.

[0019] Step S120: Generate dynamic cooperation topology information based on the network status data set, where the dynamic cooperation topology information includes a node cooperation relationship topology graph and a communication path weight table.

[0020] In the above enterprise-level data center network scenario, generate dynamic cooperation topology information based on the obtained network status data set.

[0021] Step S121: Perform normalization processing on the real-time communication load characteristics in the network status data set to obtain the load balancing coefficient of each node, where the load balancing coefficient is used to characterize the remaining communication capacity characteristics of the node during the current period.

[0022] Taking node A as an example, assume that there are multiple nodes in the network status data set, and each node has its corresponding real-time communication load characteristic. For example, the real-time communication load characteristic of node A is represented by the letter a, that of node B is represented by the letter b, and so on. To obtain the load balancing coefficient of each node, first determine the maximum value among the real-time communication load characteristics of all nodes, assumed to be max(a, b,...), and the minimum value is assumed to be min(a, b,...). For node A, the calculation logic of its load balancing coefficient is to subtract the minimum value min(a, b,...) from the real-time communication load characteristic a of node A, and then divide it by the difference between the maximum value max(a, b,...) and the minimum value min(a, b,...), that is, (a - min(a, b,...)) / (max(a, b,...) - min(a, b,...)). By normalizing the real-time communication load characteristics of each node in this way, the load balancing coefficient of each node is obtained, and this coefficient can be used to characterize the remaining communication capacity characteristics of the node in the current period. The larger the coefficient, the more remaining communication capacity of the node compared to other nodes in the current period.

[0023] Step S122: Perform segmented weighted processing on the link stability characteristics, and extract the communication stability score between nodes. The communication stability score is used to characterize the anti-interference ability characteristics and historical communication success rate characteristics of the link between nodes.

[0024] Taking the link between node C and node D as an example, the link stability characteristics include multiple factors such as signal strength and bit error rate. Assume that the link stability characteristics are divided into n segments according to different factors. For example, the signal strength is the first segment, and the bit error rate is the second segment, etc. Assign corresponding weights to each segment. Assume that the weight of the signal strength segment is w1, the weight of the bit error rate segment is w2, etc. For the signal strength, assume that its score under a certain predefined standard is s1, and the score of the bit error rate under the corresponding standard is s2, etc. Then the calculation method of the communication stability score of the link between node C and node D is s1 multiplied by w1 plus s2 multiplied by w2 plus... until sn multiplied by wn, that is, s1×w1 + s2×w2 +... + sn×wn. Through such segmented weighted processing, the communication stability score between nodes is extracted. This score can comprehensively characterize the anti-interference ability characteristics and historical communication success rate characteristics of the link between nodes. The higher the score, the stronger the anti-interference ability of the link and the higher the historical communication success rate.

[0025] Step S123: Calculate the resource availability index of the node according to the difference between the node resource occupancy characteristics and the preset resource threshold. The resource availability index is used to characterize the resource reservation characteristics of the node when performing the distribution task.

[0026] Taking node E as an example, assume that the node resource occupancy characteristics include the CPU usage rate c and the memory usage rate m, the preset CPU resource threshold is Cth, and the memory resource threshold is Mth. For the calculation of the CPU resource availability metric, subtract the CPU usage rate c of node E from Cth to obtain a part of the CPU resource availability metric, assumed to be Ca = Cth - c; for the calculation of the memory resource availability metric, subtract the memory usage rate m of node E from Mth to obtain a part of the memory resource availability metric, assumed to be Ma = Mth - m. Combine the normalization results of these two parts to obtain the resource availability metric of node E. In this way, calculate the resource availability metric of the node according to the difference between the node resource occupancy characteristics and the preset resource threshold, and this metric is used to characterize the resource reservation characteristics of the node when performing the distribution task. The higher the metric, the more resources the node can reserve when performing the distribution task.

[0027] Step S124: Construct a dynamic collaboration decision matrix based on the load balancing coefficient, communication stability score, and resource availability metric. The dynamic collaboration decision matrix includes the collaboration weight allocation characteristics and node priority ranking characteristics between nodes.

[0028] In an enterprise-level data center network, assume there are multiple nodes such as nodes F, G, H, etc. Taking nodes F and G as an example, based on the load balancing coefficient f1 (a dimensionless value in the range of 0 - 1 itself), communication stability score f2 (already normalized to a dimensionless value in the range of 0 - 1), resource availability metric f3 (already normalized to a dimensionless value in the range of 0 - 1) of node F obtained previously, and the load balancing coefficient g1, communication stability score g2, resource availability metric g3 (also dimensionless values in the range of 0 - 1) of node G, construct a dynamic collaboration decision matrix.

[0029] For the collaboration weight allocation between nodes F and G, assume it is determined by a set algorithm (for example, comprehensively considering the weighted sum of the load balancing coefficient, communication stability score, and resource availability metric). Assume the weight of the load balancing coefficient is w_f1, the weight of the communication stability score is w_f2, and the weight of the resource availability metric is w_f3. Then the collaboration weight between nodes F and G is f1×w_f1 + f2×w_f2 + f3×w_f3. And so on, determine the collaboration weights between all nodes to construct a dynamic collaboration decision matrix. At the same time, according to these coefficients and scores, rank the nodes according to the set rules (such as giving priority to the resource availability metric, followed by the communication stability score, and finally the load balancing coefficient), so that the dynamic collaboration decision matrix includes the collaboration weight allocation characteristics and node priority ranking characteristics between nodes.

[0030] Step S125: Generate dynamic cooperation topology information including a node cooperation relationship topology graph and a communication path weight table based on the dynamic cooperation decision matrix.

[0031] Based on the previously constructed dynamic cooperation decision matrix, taking the nodes of the enterprise-level data center network as an example. For the node cooperation relationship topology graph, according to the cooperation weight distribution characteristics between nodes in the dynamic cooperation decision matrix, connect the nodes with higher cooperation weights with lines. The thickness of the lines can represent the magnitude of the cooperation weights, thereby constructing the node cooperation relationship topology graph to visually display the tightness of the cooperation relationship between nodes. For the communication path weight table, taking the path between node I and node J as an example, determine the weight of this path according to the relevant information of node I and node J in the dynamic cooperation decision matrix. For example, obtain the path weight by comprehensively calculating the load balancing coefficient, communication stability score, and resource availability index of node I and node J (the calculation method is similar to the calculation of the cooperation weight before). Organize the weights of all paths between nodes into a table to form the communication path weight table. In this way, dynamic cooperation topology information including a node cooperation relationship topology graph and a communication path weight table is generated.

[0032] Step S130: Generate a distribution agent policy set according to the dynamic cooperation topology information, and the distribution agent policy set includes an initial path allocation policy, a dynamic load adjustment rule, and a path switching trigger condition.

[0033] In the enterprise-level data center network scenario, generate a distribution agent policy set based on the generated dynamic cooperation topology information.

[0034] Step S131: Extract the node level characteristics corresponding to the node cooperation relationship topology graph from the dynamic cooperation topology information. The node level characteristics are generated by a dynamic election algorithm and include core node identification characteristics, relay node identification characteristics, and edge node identification characteristics. The dynamic election algorithm periodically updates the node levels based on the resource availability index, historical task success rate, and network topology connectivity characteristics of the nodes.

[0035] Taking the nodes in the enterprise-level data center network as an example, extract the node hierarchy features from the node collaboration relationship topology diagram. Assume that the dynamic election algorithm is set to update the node hierarchy every T1 time period. Taking node K as an example, determine its hierarchy in the network based on its resource availability metric k1, historical task success rate k2 (assuming the historical task success rate is obtained by statistically calculating the ratio of the number of successfully completed tasks to the total number of tasks over a certain period of time), and network topology connectivity feature k3 (for example, the number of connections between node K and other nodes). If the resource availability metric k1 of node K is higher than the preset threshold, the historical task success rate k2 is also high, and the network topology connectivity feature k3 shows that it is connected to more nodes, then this node may be elected as a core node and given the core node identification feature. Similarly, for nodes with relatively moderate resource availability metrics, historical task success rates, and network topology connectivity features, they may be elected as relay nodes and given the relay node identification feature; while nodes with relatively low resource availability metrics, historical task success rates, and network topology connectivity features may be elected as edge nodes and given the edge node identification feature. Through such a dynamic election algorithm, periodically update the node hierarchy to adapt to changes in the network state.

[0036] Step S132: Determine the path selection priority between nodes according to the communication path weight table, and generate an initial path allocation strategy in combination with the node hierarchy features. The initial path allocation strategy is used to define the default communication path features from the core node to the edge node.

[0037] In the enterprise-level data center network, taking node L as the core node and node N as the edge node as an example. According to the communication path weight table, for multiple possible paths between node L and node N, the lower the path weight, the higher the path selection priority. Assume that there are multiple paths such as path 1 and path 2 connecting node L and node N, the weight of path 1 is w1, and the weight of path 2 is w2. If w1 < w2, then the selection priority of path 1 is higher than that of path 2. In combination with the node hierarchy features, preferentially select the path starting from the core node L, passing through the relay node (assumed to be node M), and finally reaching the edge node N as the default communication path in the initial path allocation strategy. Such path selection takes into account both the path weight and the node hierarchy to ensure that the transmission of data from the core node to the edge node has high efficiency and stability in the initial stage. In this way, generate an initial path allocation strategy for defining the default communication path features from the core node to the edge node.

[0038] Step S133: Generate a dynamic load adjustment rule based on the correlation between the real-time communication load feature and the resource availability metric. The dynamic load adjustment rule includes the load transfer threshold feature and resource preemption priority feature between nodes.

[0039] Taking nodes O and P as examples, assume that the real-time communication load characteristic of node O is represented by the letter o, and the resource availability index is represented by the letter Oa; the real-time communication load characteristic of node P is represented by the letter p, and the resource availability index is represented by the letter Pa. First, analyze the correlation between the real-time communication load characteristic and the resource availability index. For example, when the real-time communication load o of node O gradually increases while its resource availability index Oa gradually decreases, it indicates that node O may face a situation of excessive load. Based on this correlation, a load transfer threshold is set. Assume that when the real-time communication load o of node O reaches a certain proportion, such as o accounting for 80% of its theoretical maximum load (this is only an example proportion here), and the resource availability index Oa is lower than a certain threshold, it is considered that load transfer is required. This is the load transfer threshold characteristic between nodes. For the resource preemption priority characteristic, assume that it is determined according to the resource availability index and historical task success rate of the node. Nodes with higher resource availability indexes and higher historical task success rates have higher resource preemption priorities during load transfer. For example, if the resource availability index Oa of node O is higher than the resource availability index Pa of node P, and the historical task success rate of node O is also higher than that of node P, then during load transfer, node O has a higher resource preemption priority. In this way, a dynamic load adjustment rule including the load transfer threshold characteristic and resource preemption priority characteristic between nodes is generated.

[0040] Step S134: Generate a path switching trigger condition according to the comparison result between the communication stability score and a preset stability threshold, where the path switching trigger condition is used to define the path switching threshold of the standby path activation rule when the link is interrupted.

[0041] Taking the link between nodes Q and R as an example, assume that the communication stability score of this link is represented by the letter q (normalized to the range of 0 - 1), and the preset stability threshold is Qth (also in the range of 0 - 1).

[0042] When the communication stability score q is lower than the preset stability threshold Qth, it indicates that there may be unstable factors in this link and the link is likely to be interrupted. At this time, this situation below the threshold is used as the path switching trigger condition.

[0043] For the path switching threshold, when the communication stability score q drops to a certain proportion (such as 80%, this is only an example proportion and is actually determined according to the network situation) of the preset stability threshold Qth, it is considered that the link has become unstable enough to require path switching. This is the path switching threshold defined in the path switching trigger condition. When the communication stability score q of the link reaches or is lower than this path switching threshold, the standby path activation rule is triggered, and a suitable path is selected from the standby paths for data transmission to ensure the stability of data transmission. In this way, according to the comparison result between the communication stability score and the preset stability threshold, a path switching trigger condition for defining the path switching threshold of the standby path activation rule when the link is interrupted is generated.

[0044] Step S135: Integrate the initial path allocation strategy, the dynamic load adjustment rule, and the path switching trigger condition to generate a set of distribution agent strategies containing multi-dimensional decision rules.

[0045] In the enterprise-level data center network scenario, integrate the initial path allocation strategy, the dynamic load adjustment rule, and the path switching trigger condition generated above. Taking a data distribution task as an example, first, determine the default communication path from the core node to the edge node according to the initial path allocation strategy, such as the path from the core node S through the relay node T to the edge node U. During data transmission, according to the dynamic load adjustment rule, monitor the real-time communication load characteristics and resource availability metrics of the nodes in real time. Suppose the real-time communication load of node S exceeds the load transfer threshold, and node T has a higher resource preemption priority, then part of the load is transferred from node S to node T. At the same time, according to the path switching trigger condition, if the communication stability score of the link between node S and node T is lower than the path switching threshold, the standby path is activated, such as the path from node S through node V to node U. By synergistically applying different rules at different stages of the data distribution process in this way, a set of distribution agent strategies containing multi-dimensional decision rules is generated to ensure the efficient and stable progress of the data distribution task in a complex and changeable network environment.

[0046] Step S140: Perform path optimization processing on the target distribution task based on the set of distribution agent strategies to obtain an optimized set of distribution paths, deploy the optimized set of distribution paths to the corresponding distribution nodes to perform dynamic collaborative distribution operations, and update the network state data set based on the real-time feedback data during the distribution process.

[0047] In the enterprise-level data center network scenario, perform a series of operations on the target distribution task based on the generated set of distribution agent strategies.

[0048] Step S141: Analyze the task attribute characteristics of the target distribution task, where the task attribute characteristics include task data volume characteristics, transmission timeliness requirement characteristics, and data security level characteristics.

[0049] Suppose there is currently a target distribution task, taking Task A as an example. For the task data volume characteristics, it can be measured by counting the total amount of data that Task A needs to transmit. Suppose it is represented by the letter D, which may include multiple data files or data sets, such as data file a, data file b, etc., and its total amount is the sum of the sizes of these files. The transmission timeliness requirement characteristics, for example, Task A requires data distribution to be completed within time T2, which is the transmission timeliness requirement. The data security level characteristics. Suppose the data security level is divided into three levels: high, medium, and low. The data security level of Task A is evaluated as high level, which may be due to reasons such as the data containing enterprise sensitive information. In this way, the task attribute characteristics of the target distribution task are comprehensively analyzed, providing a basis for subsequent path optimization processing.

[0050] Step S142: Match the resource availability indicators in the dynamic collaboration decision matrix according to the task data volume characteristics, and filter out the subset of distribution nodes that meet the resource reservation characteristics.

[0051] Taking the task data volume characteristic D of Task A as an example, in the dynamic collaboration decision matrix, traverse the resource availability indicators of each node. Suppose the resource availability indicator of node X is Xa, the resource availability indicator of node Y is Ya, etc. Determine whether the resource availability indicator can meet the resource requirements of the task data volume characteristic. For example, if the task data volume D is large, it is required that the resource availability indicator of the node reaches a preset value or above to meet the resource reservation characteristics of the task. Suppose it is set that the resource availability indicator needs to be greater than a certain threshold Xd to meet the resource reservation of Task A. Then, filter out the nodes whose resource availability indicators are greater than Xd, such as node X, node Z, etc. These nodes form a subset of distribution nodes that meet the resource reservation characteristics, providing a basis for subsequent selection of a suitable distribution path.

[0052] Step S143: Based on the default communication path characteristics from the core node to the edge node defined in the initial path allocation strategy, combined with the transmission timeliness requirement characteristics, extract a set of candidate paths that meet the communication stability score from the subset of distribution nodes. The set of candidate paths includes multiple node collaboration sequences and corresponding path weight allocation characteristics.

[0053] Taking the enterprise-level data center network as an example, assume that the core node is node A, the edge node is node B, and the initial path allocation policy defines the default communication path from node A through node C and node D to node B. Among the subset of distribution nodes that meet the resource reservation characteristics, combined with the transmission timeliness requirement characteristics, such as the task requires to complete the transmission within time T2. For each possible path, such as the path from node A through node E to node B, calculate the data transmission time of this path under the current network state. Assume that the data transmission time is related to factors such as communication stability score and path weight, and is calculated through a certain algorithm (for example, considering factors such as bandwidth and delay between nodes). If the estimated transmission time of this path can meet the transmission timeliness requirement T2, and its communication stability score meets the preset standard (assuming the communication stability score needs to be greater than a certain threshold Yd), then add this path to the candidate path set. The candidate path set may contain multiple paths, such as the path from node A through node E to node B, the path from node A through node F and node G to node B, etc. Each path has a corresponding node cooperation sequence and path weight allocation characteristics, preparing for subsequent encryption processing and final path selection.

[0054] Step S144: Dynamically select an encryption algorithm according to the data security level characteristics, encrypt the candidate path set, generate a secure transmission path set containing encrypted node identification characteristics and data sharding rules, and correct the weights of the secure transmission path set to obtain an optimized distribution path set.

[0055] In the enterprise-level data center network scenario, continue to take task A as an example, and its data security level characteristics have been determined. Assume that there is a preset corresponding relationship between the data security level and the selection of the encryption algorithm. Use letters to represent different data security levels. For example, the data security level is represented by the letter L, and different values represent different levels, corresponding to different encryption algorithms. When L is in a certain range, for example, L is greater than the preset value L1, select encryption algorithm A; when L is less than or equal to L1 and greater than another value L2, select encryption algorithm B, and so on.

[0056] For each path in the candidate path set, taking path P as an example, it contains a series of node collaboration sequences, assumed to be node M - node N - node O. According to the selected encryption algorithm, path P is encrypted. If encryption algorithm A is selected, this algorithm may stipulate encrypting the data transmitted by each node in a preset manner. For the data transmitted by node M, assuming according to the rules of encryption algorithm A, certain parts of the data are transformed according to preset logic, such as rearranging the first half and the second half of the data in a preset order, and adding a specific encryption identifier, which constitutes part of the encrypted node identifier feature. Similar processing according to the rules of encryption algorithm A is also performed on the data transmitted by node N and node O, thereby generating an encrypted path containing the encrypted node identifier feature.

[0057] Meanwhile, to ensure the integrity and security of data during transmission, it is also necessary to determine the data fragmentation rule. Assume that the data fragmentation rule is determined based on factors such as the data volume of task A, the bandwidth of the path, and the transmission timeliness requirement. Taking the data volume represented by the letter D, the total bandwidth of path P represented by the letter B, and the transmission timeliness requirement time represented by the letter T as an example. First, calculate the data volume that can be transmitted by path P per unit time under ideal conditions, assumed to be B divided by a coefficient E related to the transmission efficiency (this coefficient takes into account factors such as network overhead), and the data volume that can be transmitted per unit time is obtained as B / E. Then, according to the transmission timeliness requirement T, calculate the total data volume that can be transmitted by path P within T time as (B / E)×T. Then compare the data volume D of task A with this transmissible total. Assume D is greater than (B / E)×T, indicating that data fragmentation transmission is required. Divide the data D into data fragments of preset sizes, assumed to be data fragments of sizes d1, d2, d3, etc., and this is the data fragmentation rule. Each data fragment corresponds to different node transmission stages on path P. For example, data fragment d1 is transmitted by node M, d2 is transmitted by node N, etc., thereby generating a secure transmission path containing the data fragmentation rule.

[0058] After completing the encryption process and determining the data fragmentation rule, a set of secure transmission paths is obtained. Since the encryption process may affect the transmission performance of the path, it is necessary to correct the weights of the set of secure transmission paths. Assume that the path weight was originally represented by the letter W. After encryption, the weight is adjusted according to factors such as the complexity of the encryption algorithm and the number of data fragments. Taking the complexity of the encryption algorithm represented by the letter C and the number of data fragments represented by the letter n as an example, assume there is a weight adjustment function f(C, n), then the corrected weight W' = W + f(C, n) (this is only an example here, and the actual function may be more complex and needs to consider multiple factors comprehensively). Such weight correction is performed on each path in the set of secure transmission paths, and finally an optimized set of distribution paths is obtained.

[0059] Step S145: Split the node cooperation sequence in the optimized distribution path set into multiple subtask units. The unit features of each subtask unit include the target node identification feature, the data transmission volume feature, and the path weight execution feature.

[0060] Still taking a path R in the optimized distribution path set of task A in the enterprise-level data center network scenario as an example, the node cooperation sequence of path R is assumed to be node P1 - node P2 - node P3. Split this node cooperation sequence into multiple subtask units.

[0061] For the first subtask unit, the target node identification feature is node P1. The data transmission volume feature, according to the previously determined data sharding rule, assuming the data shard size assigned to node P1 is d1, this is the data transmission volume feature of this subtask unit. The path weight execution feature, assuming the weight of path R is corrected to Wr, according to a preset ratio, for example, the importance ratio of node P1 in the entire path transmission is p1 (this ratio can be comprehensively determined according to factors such as the resource availability and communication stability of the node), then the path weight execution feature of this subtask unit is Wr multiplied by p1, that is, Wr×p1.

[0062] For the second subtask unit, the target node identification feature is node P2. Assuming the data shard size assigned to node P2 is d2, as its data transmission volume feature. The path weight execution feature, assuming the importance ratio of node P2 in the entire path transmission is p2, then the path weight execution feature of this subtask unit is Wr×p2.

[0063] Similarly, for the third subtask unit, the target node identification feature is node P3, the data transmission volume feature is the data shard size d3 assigned to it, and the path weight execution feature is Wr×p3.

[0064] In this way, each node cooperation sequence in the optimized distribution path set is split into multiple subtask units, and each subtask unit has clear target node identification features, data transmission volume features, and path weight execution features, preparing for the subsequent assignment of subtask units to corresponding distribution nodes.

[0065] Step S146: Assign the subtask units to the corresponding distribution nodes according to the unit features, and trigger the cooperation handshake protocol between nodes to establish a communication link.

[0066] Taking the subtask unit of task A as an example, assuming the target node identification feature of one of the subtask units is node Q, the data transmission volume feature is dq, and the path weight execution feature is Wq. First, determine the distribution node to be assigned according to the target node identification feature as node Q.

[0067] Then, trigger the collaborative handshake protocol between nodes. Taking node Q and the previous node (assumed to be node P) as an example, node P sends a handshake request message containing information related to the subtask unit to node Q, which includes the data transfer volume feature dq, the path weight execution feature Wq, and some information for authentication and synchronization. After receiving the handshake request message, node Q checks whether its own resources can handle the subtask unit. Assume that the resource availability metric of node Q is represented by the letter Qa. According to the previously determined rules regarding resource reservation, judge whether Qa meets the requirement for processing the data volume of dq. If it meets the requirement, node Q sends an acknowledgment response message to node P, which also contains some acknowledgment information, such as the confirmation of its own processing ability, the readiness status for receiving data, etc. After receiving the acknowledgment response message, node P establishes a communication link with node Q for transmitting the data of the subtask unit.

[0068] For each subtask unit, in this way, according to its unit characteristics, it is assigned to the corresponding distribution node, and a communication link is established by triggering the collaborative handshake protocol between nodes to ensure that data can be accurately and orderly transmitted between the distribution nodes.

[0069] Step S1461: Analyze the unit characteristics in the subtask unit, and extract the weight allocation ratio characteristics and data transmission timing characteristics between nodes.

[0070] Taking the subtask unit assigned to node Q in task A as an example, analyze its unit characteristics. Assume that the path weight execution feature of the subtask unit is Wq. By analyzing the roles and mutual relationships of each node in path R during the entire transmission process, determine the weight allocation ratio characteristics between nodes. For example, if path R involves node P, node Q, and node S, and the allocation relationship of the path weight execution feature among these three nodes is that node P accounts for p1, node Q accounts for p2, node S accounts for p3, and p1 + p2 + p3 = 1, then the weight allocation ratio characteristic of node Q here is p2.

[0071] For the data transmission timing characteristics, they are determined according to the data fragmentation rules and the transmission timeliness requirements of the entire task. Assume that the data fragment dq needs to be transmitted within a specific time interval, for example, starting from time t1 and ending at time t2. This time interval and the order of data fragment transmission among nodes and other information constitute the data transmission timing characteristics. Through such analysis, the weight allocation ratio characteristics and data transmission timing characteristics between nodes in the subtask unit are extracted, preparing for subsequent data fragmentation and transmission scheduling.

[0072] Step S1462: Split the data volume of the subtask unit into multiple data fragments according to the weight allocation ratio characteristics, and each data fragment corresponds to a distribution node and a transmission time window feature.

[0073] Continuing with the example of the subtask unit assigned to node Q, whose data transfer volume characteristic is dq. Assuming that the weight allocation ratio characteristic is determined based on the bandwidth ratio of the nodes, and the bandwidths of node P, node Q, and node S are Bp, Bq, and Bs respectively, then the weight allocation ratios of node P, node Q, and node S are Bp / (Bp + Bq + Bs), Bq / (Bp + Bq + Bs), and Bs / (Bp + Bq + Bs) respectively, denoted as p1, p2, and p3.

[0074] Then the data shard size allocated to node Q is dq × p2. Similarly, the data shard size allocated to node P can be calculated as dq × p1, and the data shard size allocated to node S is dq × p3.

[0075] For each data shard, its transmission time window characteristic is determined according to the data transmission timing characteristic. Assuming that the entire task needs to be completed within the time interval T, and the data transmission timing stipulates that node P transmits its data shard first, then node Q transmits, and finally node S transmits. Then the transmission time window of the data shard of node Q may be within a certain time period after node P completes the transmission. Assuming it starts from time t3 and ends at time t4, this time interval is the transmission time window characteristic of the corresponding data shard of node Q. In this way, according to the weight allocation ratio characteristic, the data volume of the subtask unit is split into multiple data shards, and the corresponding distribution node and transmission time window characteristic are determined for each data shard.

[0076] Step S1463: Generate an inter-node transmission scheduling instruction based on the transmission time window characteristic, where the transmission scheduling instruction includes the sending start time characteristic of the data shard and the receiving timeout threshold characteristic.

[0077] Taking the data shard corresponding to node Q as an example, according to its transmission time window starting from time t3 and ending at time t4, an inter-node transmission scheduling instruction is generated. The sending start time characteristic is time t3, which is the start time for node P to send the data shard to node Q. The receiving timeout threshold characteristic is set to ensure the reliability of data transmission. Assuming considering factors such as possible network delays and interferences, a time threshold Tq is set. If node Q does not successfully receive the data shard within Tq time starting from time t3, it is considered that the reception times out. This Tq is the receiving timeout threshold characteristic. In this way, based on the transmission time window characteristic, a transmission scheduling instruction including the sending start time characteristic and the receiving timeout threshold characteristic is generated for each data shard to standardize the data transmission process between nodes.

[0078] Step S1464: Establish a two-way confirmation mechanism between distribution nodes. After the sending node sends a data shard within the window corresponding to the sending start time feature, the receiving node returns a confirmation signal within the receiving timeout threshold feature.

[0079] Still taking the example of node P sending a data shard to node Q, a two-way confirmation mechanism is established between distribution nodes. When the time reaches t3 corresponding to the sending start time feature, node P sends a data shard to node Q. After receiving the data shard, node Q needs to return a confirmation signal to node P within the time Tq corresponding to the receiving timeout threshold feature. This confirmation signal indicates that node Q has successfully received the data shard, and the integrity and correctness of the data have been preliminarily verified. Through this two-way confirmation mechanism, the accuracy and reliability of data transmission between nodes are ensured, avoiding the situation where data is lost or transmission errors are not detected.

[0080] Step S1465: If the receiving node does not return a confirmation signal within the timeout threshold, reallocate the data shard to the node corresponding to the alternate path according to the path switching trigger condition.

[0081] Suppose after node P sends a data shard to node Q, node Q does not return a confirmation signal within the receiving timeout threshold Tq. At this time, reallocate the data shard according to the path switching trigger condition. The path switching trigger condition may be determined based on various factors such as link stability and node load. For example, if the link stability score between node Q and node P is lower than a certain threshold, or node Q's current load is too high, path switching is triggered. Suppose there is an alternate path involving nodes R and S, and the node cooperation sequence and related parameters of the alternate path have been determined previously. Then, reallocate the data shard originally intended for node Q to the node corresponding to the alternate path, such as node R or node S, to ensure that the data can continue to be transmitted without affecting the progress of the entire task.

[0082] Step S147: During data transmission, continuously monitor the change in the communication load feature of the node according to the dynamic load adjustment rule. When it is detected that the real-time communication load of the node exceeds the load transfer threshold feature, start the load migration operation.

[0083] Step S1471: Collect the real-time bandwidth occupancy rate feature, CPU usage rate feature, and memory consumption rate feature of the node within each transmission time window.

[0084] Taking node R in a certain transmission time window as an example, within this transmission time window, the real-time bandwidth occupancy rate feature of node R is collected through the monitoring mechanism, represented by the letter Br, which reflects the proportion of the network bandwidth occupied by node R within this time window. At the same time, the CPU usage rate feature is collected, assumed to be represented by the letter Cr, which reflects the busy degree of the CPU of node R within this time window. And the memory consumption rate feature, assumed to be represented by the letter Mr, shows the consumption proportion of the memory resources of node R within this time window. These features reflect the resource usage situation of node R in this transmission time window from different aspects, providing data support for subsequent judgment of the node load status.

[0085] Step S1472: Calculate the instantaneous load pressure value of the current node according to the real-time bandwidth occupancy rate feature. When the instantaneous load pressure value exceeds the load transfer threshold, mark this node as an overloaded node.

[0086] Still taking node R as an example, assume there is a calculation function h(Br) related to the real-time bandwidth occupancy rate feature Br to calculate the instantaneous load pressure value. For example, the instantaneous load pressure value Lp = h(Br). This function may comprehensively consider factors such as the impact of network bandwidth on the node's processing ability. Compare the calculated instantaneous load pressure value Lp with the previously determined load transfer threshold Tth. If Lp exceeds Tth, then mark node R as an overloaded node. This step, through a quantitative method, based on the real-time collected bandwidth occupancy rate feature, clarifies whether the node is in an overloaded state, so as to take measures such as load migration in a timely manner.

[0087] Step S1473: Extract the list of data shards currently processed by the overloaded node, and determine the data shards to be migrated and the target migration node according to the resource preemption priority feature of the dynamic load adjustment rule.

[0088] Assume that node R is marked as overloaded. Extract its currently processed data shard list Lr. For example, in dynamic load adjustment rules, resource preemption priority is determined by the node's resource availability index and historical task success rate. As previously mentioned, node S's resource availability index is Sa, and its historical task success rate is Sh. Node T's resource availability index is Ta, and its historical task success rate is Th. Compare the product of the resource availability index and the historical task success rate of nodes S and T (Sa × Sh and Ta × Th). If Sa × Sh > Ta × Th, and node S's resources are currently relatively idle (determined by comparing node S's resource usage with a resource threshold St; if node S's current resource usage is less than St, it is considered relatively idle), node S is selected as the target migration node. Then, based on factors such as the size and urgency of the data shards, the data shards to be migrated are determined from the data shard list Lr. For example, larger data shards with lower urgency are prioritized to alleviate the overload pressure on node R while minimizing the impact on the overall task progress.

[0089] Step S1474: Transfer the data shards to be migrated from the overloaded node to the target migration node, and update the path weight distribution characteristics in the distribution path set.

[0090] Assume that node R is determined to be an overloaded node, node S is the target migration node, and some data shards to be migrated are selected from the data shard list Lr of node R. These data shards are transferred from node R to node S through the network transmission mechanism.

[0091] After the transfer is completed, the path weight distribution features in the distribution path set are updated. Assume that in the original path weight distribution features, the weight relationship between node R and subsequent nodes is Wr-next, and the weight relationship between node S and related nodes is Ws-related.

[0092] Due to the transfer of data shards, consider the load changes of nodes R and S (assuming that they are measured by dimensionless factors such as the proportion of data transferred by overloaded nodes and the remaining load capacity of the target migration node). For example, assume that the proportion of data transferred by overloaded node R to its total processed data volume is α, and the proportion of the remaining load capacity of node S to its total load capacity is β.

[0093] Assume that there is a weight adjustment rule related to these dimensionless factors. Path weights are readjusted based on the load change ratio between nodes R and S. For example, if α and β are both large, it means that node R has offloaded a large amount of load to node S, which has a stronger load-bearing capacity. In this case, the weight of node S and related nodes may be appropriately increased, while the weight of node R and subsequent nodes may be reduced.

[0094] Assume that a new weight relationship is calculated through an adjustment function f(α, β). For example, the weight between the new node R and the subsequent nodes is Wr-next' = Wr-next - f(α, β) × Wr-next, and the weight between the new node S and the related nodes is Ws-related' = Ws-related + f(α, β) × Ws-related. In this way, the path weight distribution characteristics in the distribution path set are updated to ensure that the distribution path can adapt to the network state after the node load changes.

[0095] Step S1475: After the migration is completed, recalculate the resource availability metrics of the overloaded node and the target migration node, and synchronously update the dynamic collaboration decision matrix.

[0096] Assume that node R is the overloaded node and node S is the target migration node. After the data shard migration is completed, recalculate the resource availability metrics of node R and node S. For node R, assume that its CPU usage rate becomes Cr' after migration and its memory usage rate becomes Mr'. The total CPU is Cr_total, and the total memory is Mr_total. According to the previous method of calculating the resource availability metrics, the new resource availability metric Ra' = (Cr_total - Cr') / Cr_total + (Mr_total - Mr') / Mr_total (the actual calculation may be more complex and may also involve other resource dimensions). For node S, assume that its CPU usage rate becomes Cs' after receiving the data shard and its memory usage rate becomes Ms'. The total CPU is Cs_total, and the total memory is Ms_total. The new resource availability metric Sa' = (Cs_total - Cs') / Cs_total + (Ms_total - Ms') / Ms_total.

[0097] Then, synchronously update the dynamic collaboration decision matrix. Taking the corresponding items of node R and node S in the dynamic collaboration decision matrix as an example, assume that parameters such as the resource reservation characteristics and collaboration weights related to node R and node S in the matrix need to be adjusted according to the new resource availability metrics. Assume that there is an adjustment function g1(Ra', Sa') for adjusting the collaboration weight. For example, the new collaboration weight Wrs' = g1(Ra', Sa'). At the same time, other related parameters such as the resource reservation characteristics are also adjusted accordingly to ensure that the dynamic collaboration decision matrix can accurately reflect the network collaboration relationship after the node resource state changes.

[0098] Step S148: When it is monitored that the stability score of the current communication link is lower than the path switching threshold corresponding to the path switching trigger condition, activate the standby path from the candidate path set and switch to the standby path to continue the transmission.

[0099] During the data transmission process in an enterprise-level data center network, taking the current communication link L as an example, continuously monitor its stability score. Assume that the stability score of the current communication link L is calculated by synthesizing multiple factors, such as signal strength, bit error rate, delay fluctuation, etc., which are represented by the letters I, e, Dv respectively. The stability score Sc of the current communication link L can be calculated according to the calculation rules of the aforementioned step S122.

[0100] Meanwhile, determine the path switching threshold Sth according to the path switching trigger condition. The path switching threshold Sth is determined according to the aforementioned step S134.

[0101] When it is monitored that the stability score Sc of the current communication link L is lower than the path switching threshold Sth, it is determined that the current link is in an unstable state. Screen for standby paths in the candidate path set whose average stability score is higher than the path switching threshold Sth. Assume there are paths P1, P2, P3, etc. in the candidate path set, and calculate their average stability scores respectively. Taking path P1 as an example, assume its stability scores in different time windows in the past period are Sc11, Sc12, Sc13, etc. Calculate the average of these scores (for example, assume the average stability score is (Sc11 + Sc12 + Sc13) / 3, this is just an example, and actually different weights may be assigned according to the importance of different time windows for calculation), and obtain the average stability score M1 of path P1. Similar calculations are performed on other paths in the candidate path set, and standby paths with an average stability score higher than the path switching threshold Sth are screened out. Assume path P2 is screened out.

[0102] Step S1481: Obtain the historical stability score curve of the current communication link, and extract the average stability score features within the most recent N time windows.

[0103] Taking the current unstable communication link L as an example, obtain its historical stability score curve, which records the stability scores of link L in different time windows in the past period. Assume that the stability scores within the most recent N time windows to be extracted are ScL1, ScL2,..., ScLN respectively. Calculate the average of these scores to obtain the average stability score feature. For example, the average stability score M = (ScL1 + ScL2 +... + ScLN) / N. This average feature can more comprehensively reflect the recent stability status of the current communication link and provide a more reliable basis for judging whether the link really needs to be switched.

[0104] Step S1482: When the average stability score feature is lower than the path switching threshold, it is determined that the current link is in an unstable state.

[0105] Compare the mean stability score M obtained from the previous calculation with the path switching threshold Sth. If M is lower than Sth, it is determined that the current communication link L is in an unstable state. This determination is based on a quantitative analysis of the recent stability of the link. Compared with simply relying on the stability score at the current moment, considering the mean feature can reduce the impact of instantaneous fluctuations on the judgment and more accurately identify whether the link is in an unstable state where the path really needs to be switched.

[0106] Step S1483: Screen the alternative paths with a mean stability score higher than the path switching threshold from the candidate path set, and extract the node cooperation sequence features of the alternative paths.

[0107] Assume that it has been determined that the current link L is unstable, and screen the alternative paths with a mean stability score higher than the path switching threshold Sth from the candidate path set. Taking the paths P1, P2, P3, etc. in the candidate path set as examples, calculate their mean stability scores respectively. For example, the calculation method of the mean stability score of path P1 is (Sc11 + Sc12 + Sc13) / 3 (in fact, different weights may be assigned according to the importance of different time windows for calculation. Assume the weights are w1, w2, w3 respectively, and w1 + w2 + w3 = 1, then the mean stability score is w1×Sc11 + w2×Sc12 + w3×Sc13). Perform such calculations for all paths in the candidate path set, and screen out the alternative paths with a mean stability score higher than the path switching threshold Sth. Assume that path P2 is screened out.

[0108] Extract the node cooperation sequence features of the alternative path P2. Assume that the node cooperation sequence of path P2 is node A - node B - node C. This node cooperation sequence feature clarifies the nodes involved and their connection order when data is transmitted on the alternative path, providing key information for subsequent link switching operations.

[0109] Step S1484: Send a link switching request to the distribution nodes involved in the alternative path, and terminate the data transmission of the current link after receiving the ready confirmations from all nodes.

[0110] After determining the alternative path P2 and its node cooperation sequence as node A - node B - node C, send a link switching request to nodes A, B, and C. The request message contains detailed information about the current transmission task, such as the uncompleted data sharding situation, transmission progress, relevant attributes of the data, etc., so that the nodes on the alternative path can understand the task status and make preparations for taking over.

[0111] After Node A receives a link switching request, it checks whether its own resources and status can undertake the transmission task. Assume that the resource availability index of Node A is Aa, and it is compared and judged with Aa according to the data volume of the task, the transmission rate requirement, etc. For example, the data volume that the task requires Node A to process per unit time is Da, and according to the current CPU usage rate Ca and the total CPU amount Ca_total of Node A, the data volume that can be processed per unit time is Ca_total×(1 - Ca)×processing efficiency coefficient Ea (the processing efficiency coefficient Ea takes into account factors such as the efficiency difference of the CPU in processing different types of data). If Ca_total×(1 - Ca)×Ea ≥ Da, and other key resources of Node A such as memory also meet the task requirements (assuming that the memory resources also go through a similar judgment, such as the total memory amount Ma_total, the current memory usage rate Ma, and the memory space requirement for the task is Da_mem, if Ma_total×(1 - Ma) ≥ Da_mem), then Node A returns a ready confirmation message to the node that sent the request, indicating that it is ready to undertake the task.

[0112] Similarly, Node B and Node C also perform similar resource and status checks. Node B makes a judgment based on its own resource availability index Ba and the requirements of the task for its processing ability and resources. If the conditions are met, it returns a ready confirmation message. Node C also makes a judgment based on its own resource availability index Ca and the task requirements and returns the corresponding message.

[0113] When the node that sent the request receives the ready confirmations from all nodes Node A, Node B, and Node C, it terminates the data transmission on the current unstable link L. This process ensures that the nodes on the alternate path have the ability to undertake the task, avoiding data transmission interruption or errors caused by insufficient node resources, etc., and ensuring the continuity and stability of data transmission.

[0114] Step S1485: Redistribute the untransmitted data shards to the node cooperation sequence corresponding to the alternate path, and adjust the transmission timing according to the path weight distribution characteristics of the alternate path.

[0115] Assume that the untransmitted data shards on the current unstable link L are d1, d2, d3, etc. Redistribute these data shards according to the node cooperation sequence Node A - Node B - Node C of the alternate path P2 and the path weight distribution characteristics.

[0116] Assume that the path weight distribution characteristics of the alternate path P2 determine the distribution ratio of the data shards among the nodes. For example, the weights of Node A, Node B, and Node C in path P2 are wa, wb, and wc respectively, and wa + wb + wc = 1. Distribute the data shards according to factors such as the size and urgency of the data shards, combined with the path weight distribution characteristics.

[0117] Suppose the size of data shard d1 is size1, the urgency level is urgency1, the size of data shard d2 is size2, the urgency level is urgency2, and the size of data shard d3 is size3, and the urgency level is urgency3. First, calculate the priority of each data shard. Suppose the priority calculation function is priority = size × urgency. Then the priority of data shard d1 is priority1 = size1 × urgency1, the priority of data shard d2 is priority2 = size2 × urgency2, and the priority of data shard d3 is priority3 = size3 × urgency3.

[0118] Allocate according to the order from high to low priority, combined with the path weight allocation characteristics. For example, if priority1 > priority2 > priority3, and wa > wb > wc, then give priority to allocating d1 to node A. Suppose the amount of data allocated to node A is size1 × wa, the amount of data allocated to node B is size2 × wb, and the amount of data allocated to node C is size3 × wc (the actual allocation may need to be fine-tuned considering factors such as the integrity of data shards).

[0119] After the allocation is completed, adjust the transmission timing according to the path weight allocation characteristics of the alternate path. Nodes with larger weights may transmit data shards first, or the transmission time window is relatively longer. For example, the weight of node A, wa, is larger, and its corresponding transmission time window may start from time t1 and end at time t2; the weight of node B, wb, is the second largest, and its transmission time window may start from time t2 and end at time t3; the weight of node C, wc, is the smallest, and its transmission time window may start from time t3 and end at time t4. In this way, adjust the transmission timing according to the path weight allocation characteristics of the alternate path to ensure that the data can continue to be transmitted efficiently and orderly on the alternate path.

[0120] Step S149: After the distribution task is completed, collect the feedback data of each node. The feedback data includes the actual transmission delay characteristics, resource consumption characteristics, and the number of link interruptions characteristics, and update the real-time communication load characteristics and link stability characteristics in the network state data set based on the feedback data.

[0121] In the enterprise-level data center network scenario, after the distribution task of task A is completed, collect the feedback data for each node participating in the task. Taking node Z as an example, collect its actual transmission delay characteristics, resource consumption characteristics, and the number of link interruptions characteristics.

[0122] Regarding the actual transmission delay characteristics, during the transmission of Task A, the time intervals from when Node Z starts receiving data to when it finishes transmitting the data to the next node are respectively recorded as multiple time intervals such as t1, t2, t3, etc. (corresponding to the transmission of different data shards). Assuming that different data shards have different importance levels, corresponding weights are assigned to each time interval. For example, the importance weights of data shards d1, d2, and d3 are wz1, wz2, and wz3 respectively, and wz1 + wz2 + wz3 = 1. The actual transmission delay characteristic is obtained through weighted calculation, that is, Tz = wz1×t1 + wz2×t2 + wz3×t3. These weights can be determined according to factors such as the data type carried by the data shard and the criticality to the service. For example, if data shard d1 contains critical service data, then its weight wz1 can be set relatively high.

[0123] Regarding the resource consumption characteristics, assume that the resources of Node Z include CPU, memory, etc. Taking the CPU as an example, the CPU utilization rate of Node Z before Task A starts is Cz1, and after the task is completed is Cz2. A part of the CPU resource consumption characteristic is obtained by calculating the difference between the two, that is, Cz_diff = Cz2 - Cz1. For the memory resource, the memory utilization rate of Node Z before Task A starts is Mz1, and after the task is completed is Mz2. The memory resource consumption characteristic is Mz_diff = Mz2 - Mz1. Considering the resource consumption characteristics of the CPU and memory comprehensively, assume that there is a comprehensive function g(Cz_diff, Mz_diff) to obtain the overall resource consumption characteristic Resource_diff = g(Cz_diff, Mz_diff) of Node Z. This comprehensive function may consider factors such as the importance differences of different resources in the node operation. For example, if the CPU resource has a greater impact on the data processing speed of the node, then the weight of the CPU resource consumption characteristic Cz_diff in the function g may be relatively high.

[0124] Regarding the number of link interruptions characteristic, assume that during the transmission of Task A, the number of link interruptions between Node Z and its adjacent nodes is recorded as n.

[0125] Update the real-time communication load characteristics and link stability characteristics in the network status data set based on this feedback data. For the real-time communication load characteristics, taking node Z as an example, assume that the previous real-time communication load characteristics were mainly determined by factors such as the total data traffic received and sent per unit time and the rate of processing data packets. Now, update it by combining the actual transmission delay characteristics and resource consumption characteristics. Assume there is an adjustment function k(Tz, Resource_diff) related to the actual transmission delay and resource consumption, where Tz is the actual transmission delay characteristic and Resource_diff is the overall resource consumption characteristic. For example, the possible calculation logic of the adjustment function k(Tz, Resource_diff) could be: Assume there are two coefficients α and β, representing the influence degrees of the actual transmission delay and resource consumption on the real-time communication load respectively. Then k(Tz, Resource_diff) = α×Tz + β×Resource_diff. These coefficients α and β can be determined through methods such as analyzing historical data and network performance testing to accurately reflect the influence of the actual transmission delay and resource consumption on the real-time communication load. Then the updated real-time communication load characteristic Lz' = Lz + k(Tz, Resource_diff) (where Lz is the previous real-time communication load characteristic).

[0126] For the link stability characteristic, assume that the stability score of the link between node Z and its adjacent nodes was previously represented by the letter Sz. Since the number of link interruptions n may affect the link stability, assume there is an adjustment coefficient m(n) related to the number of link interruptions. For example, assume there is a piecewise function. When n ≤ n1, m(n) = a×n; when n1 < n ≤ n2, m(n) = b×n + c; when n > n2, m(n) = d×n + e (where a, b, c, d, e are constants determined according to historical link stability data and network performance requirements, and n1, n2 are thresholds of the number of link interruptions determined based on experience or data analysis). Through such a piecewise function, the link stability score can be adjusted more carefully according to the number of link interruptions to accurately reflect the actual stability change of the link. Then the updated link stability score Sz' = Sz - m(n).

[0127] Integrate the updated average latency feature (i.e., the actual transmission latency feature Tz), the resource consumption rate curve (which can be formed by the change of the overall resource consumption feature Resource_diff with factors such as time or task volume), and the stability score (i.e., the updated link stability score Sz') into the network state data set, covering the original real-time communication load feature and link stability feature. Assume that the network state data set is stored in a certain data structure, such as a dictionary structure indexed by nodes, and each node corresponds to a sub-dictionary containing information such as real-time communication load features and link stability features. For node Z, find the corresponding sub-dictionary in this dictionary structure, and replace the original real-time communication load feature and link stability feature in it with the updated Lz' and Sz', so as to complete the update of the network state data set and enable it to accurately reflect the state changes of each node in the network after the task is completed.

[0128] Step S1491: Conduct statistical analysis on the actual transmission latency feature, and extract the average latency feature and the maximum latency fluctuation feature of each node during task execution.

[0129] Taking node Z as an example, its actual transmission latency feature Tz has been obtained, which is weighted by the transmission time intervals of multiple data shards. To understand the transmission latency situation of node Z during task execution more comprehensively, further statistical analysis is carried out on Tz.

[0130] Assume that during task execution, node Z participates in the transmission of multiple data shards, with corresponding time intervals t1, t2,..., tn, and weights wz1, wz2,..., wzn, and ∑(i = 1 to n) wzi = 1. The calculation of the average latency feature is to sum all time intervals weighted by the weights and then divide by the number of data shards (or use other statistical methods according to the actual situation). For example, the average latency feature Average_Tz = (wz1 × t1 + wz2 × t2 +... + wzn × tn) / n (here n is the number of data shards, and the actual calculation may use more complex statistical methods according to factors such as the importance distribution of data).

[0131] For the maximum delay fluctuation feature, first determine the difference between the transmission time interval of each data shard and the average delay feature, i.e., Δt1 = t1 - Average_Tz, Δt2 = t2 - Average_Tz,..., Δtn = tn - Average_Tz. Then find the maximum value among these differences, that is, the maximum delay fluctuation feature Max_Fluctuation_Tz = max(|Δt1|, |Δt2|,..., |Δtn|). By extracting the average delay feature and the maximum delay fluctuation feature, the stability and overall situation of the transmission delay of node Z during task execution can be evaluated more accurately, providing detailed data support for subsequent network state analysis and optimization.

[0132] Step S1492: Calculate the remaining resource availability index of the node after task execution according to the resource consumption feature, and generate a resource consumption rate curve in combination with the initial resource availability index.

[0133] Taking node Z as an example, assume that the CPU usage rate before task execution is Cz1 (expressed as a percentage), and after task execution is Cz2 (expressed as a percentage), and the memory usage rate before task execution is Mz1 (expressed as a percentage), and after task execution is Mz2 (expressed as a percentage).

[0134] Assume that the total CPU is Cz_total (expressed as a percentage, that is, representing 100%), and the total memory is Mz_total (expressed as a percentage, that is, representing 100%).

[0135] Before task execution, the initial resource availability index can be calculated in a certain way. For example, the initial resource availability index Initial_Availability_Z = (1 - Cz1) + (1 - Mz1) (this is only an example, and the actual calculation may involve more resource dimensions and more complex calculation methods to comprehensively reflect the degree to which the node's initial resources can be used for tasks).

[0136] After task execution, calculate the remaining resource availability index according to the resource consumption feature. For example, the remaining resource availability index Remaining_Availability_Z = (1 - Cz2) + (1 - Mz2).

[0137] Generate a resource consumption rate curve by combining the initial resource availability metric Initial_Availability_Z and the remaining resource availability metric Remaining_Availability_Z. Assume that the task execution progress or time is on the horizontal axis and the resource availability metric is on the vertical axis. At the start of the task, it corresponds to the initial resource availability metric Initial_Availability_Z; at the end of the task, it corresponds to the remaining resource availability metric Remaining_Availability_Z. By recording the resource availability metric at different task stages and connecting these points, a resource consumption rate curve is formed. This curve intuitively shows the changing trend of resource availability of node Z during the task execution, which helps to analyze the resource consumption of the task on the node resources and predict the possible resource situation for future tasks.

[0138] Step S1493: Update the stability score of the corresponding link based on the link interruption count feature. If the link interruption count exceeds the historical average, reduce the stability score of the link.

[0139] Taking the link between node Z and its adjacent node as an example, the interruption count n of this link during the execution of task A has been recorded. Assume that through statistical analysis of historical data, the historical average interruption count of this link is Historical_Average_n.

[0140] If n exceeds Historical_Average_n, it indicates that the stability of this link in the current task has decreased compared to the historical situation, and its stability score needs to be reduced. Assume that the previous stability score of the link is represented by the letter Sz, and there is an adjustment function m(n) related to the link interruption count to reduce the stability score. For example, assume the adjustment function m(n) = (n - Historical_Average_n) × adjustment coefficient (the adjustment coefficient is determined according to factors such as the importance of the link and its impact on the overall network performance. For example, for a critical link, the adjustment coefficient may be larger to more significantly reflect its stability change). Then the updated link stability score Sz' = Sz - m(n).

[0141] If n does not exceed Historical_Average_n, it can be decided whether to fine-tune the stability score according to other factors (such as the severity and duration of the link interruption), or keep the stability score Sz unchanged. By updating the link stability score based on the link interruption count feature, the actual stability status of the link during the task execution can be reflected in real-time, providing an important basis for the adjustment of network topology and transmission strategies.

[0142] Step S1494: Integrate the updated average delay feature, resource consumption rate curve, and stability score into the network state data set, overwriting the original real-time communication load feature and link stability feature.

[0143] Assume that the network state data set is stored in a structured manner, such as a dictionary structure indexed by nodes. For node Z, its record in the network state data set contains information such as real-time communication load feature and link stability feature.

[0144] Integrate the updated average delay feature Average_Tz, resource consumption rate curve, and the updated link stability score Sz' into the record of node Z in the network state data set. Specifically, find the sub-dictionary corresponding to node Z in the dictionary structure, replace the original real-time communication load feature with the updated real-time communication load feature Lz' (calculated as in the previous steps) that combines the actual transmission delay feature and resource consumption feature, and replace the original link stability feature with the updated link stability score Sz'. At the same time, store the resource consumption rate curve in the sub-dictionary in a suitable data format (for example, record the resource availability metric values at different task stages in the form of an array) for subsequent analysis and use.

[0145] Through such an integration operation, the network state data set can accurately reflect the latest state of node Z after the task is completed, including the transmission delay situation, resource consumption situation, and link stability situation. For other nodes in the network, data integration is also performed in the same way, so that the entire network state data set comprehensively and accurately reflects the network state changes after the task execution, providing a reliable data basis for subsequent regeneration of dynamic collaboration topology information and optimization of the distribution strategy.

[0146] Step S1495: Re-execute the generation operation of the dynamic collaboration topology information based on the updated network state data set to adapt to the latest network environment state.

[0147] Based on the updated network state data set, re-execute the generation operation of the dynamic collaboration topology information. This is to enable the collaboration topology of the network to adapt to the changes in the states of each node in the network after the task execution, thereby optimizing the subsequent data distribution tasks.

[0148] First, for the real-time communication load characteristics, the real-time communication load characteristics of each node in the network have been updated according to factors such as actual transmission delay and resource consumption. Taking node X as an example, the updated real-time communication load characteristics are represented by the letter Xl'. Based on these updated real-time communication load characteristics, the load balancing coefficients of all nodes are recalculated. Assume that the maximum value among the updated real-time communication load characteristics of all nodes is determined as max(Xl', Yl', Zl',...), and the minimum value is min(Xl', Yl', Zl',...). For node X, the recalculation logic of its load balancing coefficient is to subtract the minimum value min(Xl', Yl', Zl',...) from the updated real-time communication load characteristic Xl' of node X, and then divide by the difference between the maximum value max(Xl', Yl', Zl',...) and the minimum value min(Xl', Yl', Zl',...), that is, (Xl' - min(Xl', Yl', Zl',...)) / (max(Xl', Yl', Zl',...) - min(Xl', Yl', Zl',...)). The load balancing coefficients of all nodes in the network are recalculated in this way.

[0149] Then, the communication stability scores between nodes are regenerated based on the updated link stability characteristics. Taking the link between node X and node Y as an example, the updated link stability characteristics may involve changes in factors such as signal strength, bit error rate, and delay fluctuation, which are represented by the letters I', e', and Dv' respectively. Assume that there is a new function h'(I', e', Dv') for calculating the communication stability score. Through this function, the updated communication stability score Sc' of the link between node X and node Y is calculated. For all the links between nodes in the network, the communication stability scores are recalculated in this way based on the updated link stability characteristics.

[0150] The resource availability metrics of nodes are updated based on the latest resource consumption rate curves. Taking node Z as an example, according to the resource consumption rate curve, it can be known how the resource usage of node Z changes at different time points or task stages after the task execution. Assume that at a certain time point t, the CPU usage rate of node Z is Czt, the memory usage rate is Mzt, the total CPU is Cz_total, and the total memory is Mz_total. By comprehensively considering these resource usage situations, the updated resource availability metric is calculated. For example, the new resource availability metric Za' = (Cz_total × (1 - Czt) + Mz_total × (1 - Mzt)) / (Cz_total + Mz_total) (the actual calculation may involve more resource dimensions and more complex calculation methods to comprehensively reflect the degree to which the resources of the node at this time point are available for subsequent tasks). The resource availability metrics of all nodes in the network are updated in this way based on the latest resource consumption rate curves.

[0151] Next, adjust the collaborative weight allocation feature in the dynamic collaborative decision matrix based on the recalculated load balancing coefficient. Taking the collaborative weight between node X and node Y as an example, assume the previous collaborative weight was Wxy. Now, considering the change in the load balancing coefficient, assume there is an adjustment function f1(X load balancing coefficient, Y load balancing coefficient) related to the load balancing coefficient, and use this adjustment function to adjust the collaborative weight. For example, the new collaborative weight Wxy' = Wxy + f1(X load balancing coefficient, Y load balancing coefficient). The adjustment function f1 here comprehensively considers the impact of the changes in the load balancing coefficients of node X and node Y on their collaborative relationship. For instance, if the load balancing coefficient of node X increases while that of node Y decreases, the function f1 may increase the weight of node X in the collaboration, thereby adjusting the value of Wxy'. Such adjustments are made to the collaborative weights between all nodes in the dynamic collaborative decision matrix to enable the dynamic collaborative decision matrix to adapt to the updated network state.

[0152] Correct the path connection priority in the node collaboration relationship topology diagram according to the updated communication stability score. Taking the path between node X and node Y in the node collaboration relationship topology diagram as an example, assume the previous path connection priority was Pxy, and now correct it according to the updated communication stability score Sc'. Assume there is an adjustment rule related to the communication stability score. For example, if Sc' is higher than a certain threshold Th1, the path connection priority is increased; if Sc' is lower than a certain threshold Th2, the path connection priority is decreased. Assume there is an adjustment function f2(Sc', Th1, Th2), then the corrected path connection priority Pxy' = Pxy + f2(Sc', Th1, Th2). Such corrections are made to the connection priorities of all paths in the node collaboration relationship topology diagram to enable the topology diagram to accurately reflect the path relationship between nodes based on the updated link stability characteristics.

[0153] Synchronously adjust the resource reservation features in the dynamic collaborative decision matrix. Taking the relationship between node X and node Y in the dynamic collaborative decision matrix as an example, according to the updated resource availability metrics, assume the resource availability metric of node X is Xa' and that of node Y is Ya'. Assume there is an adjustment function f3(Xa', Ya') related to the resource availability metrics, which is used to adjust the relevant parameters regarding resource reservation features between node X and node Y in the dynamic collaborative decision matrix, such as collaborative weights, resource allocation priorities, etc. For example, the new collaborative weight Wxy'' = f3(Xa', Ya') (this is just an example, and the actual adjustment may involve more aspects and more complex functions). Such synchronous adjustments are made to all parameters regarding resource reservation features between all nodes in the dynamic collaborative decision matrix to ensure that the dynamic collaborative decision matrix can accurately reflect the changes in the resource status of nodes in the network.

[0154] Merge the corrected dynamic collaborative decision matrix with the node collaboration relationship topology graph to generate updated dynamic collaborative topology information. The dynamic collaborative decision matrix contains information such as the collaborative weight allocation features between nodes, the node priority ranking features, and the updated resource reservation features, presented in the form of a matrix, where each element corresponds to specific relationship parameters between different nodes. The node collaboration relationship topology graph graphically shows the connection relationships between nodes and the path connection priorities corrected based on the communication stability scores.

[0155] Integrate the information in the dynamic collaborative decision matrix into the node collaboration relationship topology graph. For example, for the connection between node X and node Y, the collaborative weight determined in the dynamic collaborative decision matrix is reflected in the node collaboration relationship topology graph through a certain visualization method (such as the thickness or color depth of the line); the node priority ranking features can be shown in the topology graph by the size or marking method of the nodes. At the same time, information such as the path connection priorities in the topology graph is fed back into the dynamic collaborative decision matrix to ensure the consistency and integrity of the two pieces of information. Through such mutual integration, updated dynamic collaborative topology information is generated, which comprehensively reflects the node collaboration relationships and communication path features based on the latest network state data set, providing an accurate basis for the regeneration of the subsequent distribution agent policy set.

[0156] Trigger the regeneration operation of the distribution agent policy set according to the updated dynamic collaboration topology information. Extract the node hierarchical features corresponding to the node collaboration relationship topology graph from the updated dynamic collaboration topology information. Determine the node hierarchy through a dynamic election algorithm based on the updated resource availability metrics, historical task success rate, and network topology connectivity features of the nodes. Taking node M as an example, its updated resource availability metric is Ma', the historical task success rate has a new statistical value Hs' after the task is completed, and the network topology connectivity may change due to the establishment or disconnection of new links, denoted by the letter Nc'. Assume that the dynamic election algorithm re-determines the hierarchy of node M through a complex calculation and comparison process based on these new feature values (for example, setting weights for different features, such as the weight of the resource availability metric is w1, the weight of the historical task success rate is w2, and the weight of the network topology connectivity is w3, calculating the comprehensive score S = w1×Ma' + w2×Hs' + w3×Nc', and comparing it with the comprehensive scores of other nodes). Perform such operations on all nodes to obtain the updated node hierarchical features, including core node identification features, relay node identification features, and edge node identification features.

[0157] Determine the path selection priority between nodes according to the updated communication path weight table. After the communication path weight table is updated based on link stability features and others, its weight values have changed. Taking the path between node N and node O as an example, according to the new weight value and combining the preset rules (such as the lower the weight, the higher the path selection priority) to determine its path selection priority. Assume that the weight of the path between node N and node O was Wno before and Wno' after the update. If Wno' < a certain threshold Th3, the selection priority of this path is increased. Determine the updated path selection priority by performing such analysis on all paths between nodes.

[0158] Generate dynamic load adjustment rules based on the correlation between the updated real-time communication load characteristics and resource availability metrics. As the network state is updated, both the real-time communication load characteristics and resource availability metrics change. Taking node P as an example, its updated real-time communication load characteristics are represented by the letter Pl', and the updated resource availability metric is Pa'. By analyzing the new correlation between Pl' and Pa', for example, when Pl' increases to a certain proportion and Pa' drops to a certain extent, load transfer is triggered. Assume there is a function k'(Pl', Pa') related to Pl' and Pa' to determine the specific conditions for load transfer. When k'(Pl', Pa') meets a certain condition, it is considered that load transfer is required. At the same time, re-determine the resource preemption priority characteristics according to factors such as the resource availability metric of the node and the historical task success rate. For example, for nodes Q and R, compare their updated resource availability metrics Qa' and Ra' and historical task success rates Qs' and Rs'. Determine the resource preemption priority characteristics through a certain comparison rule (such as when Qa'×Qs'>Ra'×Rs', node Q has a higher resource preemption priority). Integrate this information to generate the updated dynamic load adjustment rules.

[0159] Then generate the path switching trigger condition based on the comparison result between the updated communication stability score and the preset stability threshold. The updated communication stability score is calculated based on the new link stability characteristics. Taking the link between node S and node T as an example, its updated communication stability score is represented by the letter Sst', and the preset stability threshold is assumed to remain unchanged at Th4. When Sst' is lower than Th4, path switching is triggered. Assume there is a judgment function f4(Sst', Th4) related to Sst' and Th4. When f4(Sst', Th4) meets the preset condition, it is determined that the path needs to be switched. At the same time, according to the updated candidate path set (the candidate path set may also change due to network state changes), determine the path switching threshold and the standby path activation rule, etc., to generate the updated path switching trigger condition.

[0160] Finally, integrate the above-updated initial path allocation strategy (generated based on the updated node hierarchy features and path selection priorities), dynamic load adjustment rules, and path switching trigger conditions to generate a new set of distribution agent strategies. This new set of distribution agent strategies fully considers various factors after the network state update and can optimize the path allocation rules for subsequent distribution tasks, enabling more efficient and stable data distribution in the new network environment. For example, in subsequent distribution tasks, based on the new initial path allocation strategy, data can be more reasonably transmitted from the core node to the edge node through relay nodes; during the transmission process, according to the dynamic load adjustment rules, nodes can better handle load changes and avoid overload; when the link becomes unstable, according to the path switching trigger conditions, it can timely switch to a suitable backup path to ensure the continuity of data transmission. By continuously updating the set of distribution agent strategies according to the network state, it can adapt to the complex and changeable environment of the enterprise-level data center network and meet the requirements of data distribution tasks.

[0161] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of a distributed agent ad-hoc network system 100 based on dynamic node cooperation that can implement the ideas of the present application provided by some embodiments of the present application. For example, the processor 120 can be used on the distributed agent ad-hoc network system 100 based on dynamic node cooperation and is used to execute the functions in the present application.

[0162] The distributed agent ad-hoc network system 100 based on dynamic node cooperation can be a general-purpose server or a special-purpose server, both of which can be used to implement the distributed agent ad-hoc network method based on dynamic node cooperation of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0163] For example, the distributed agent ad-hoc network system 100 based on dynamic node cooperation can include a network port 110 connected to the network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the distributed agent ad-hoc network system 100 based on dynamic node cooperation can also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The distributed agent ad-hoc network system 100 based on dynamic node cooperation also includes an I / O interface 150 between the computer and other input / output devices.

[0164] For ease of explanation, only one processor is described in the distributed agent ad hoc network system 100 based on dynamic node cooperation. However, it should be noted that the distributed agent ad hoc network system 100 based on dynamic node cooperation in this application may also include multiple processors. Therefore, the steps performed by one processor described in this application can also be jointly performed or separately performed by multiple processors. For example, if the processor of the distributed agent ad hoc network system 100 based on dynamic node cooperation performs step A and step B, it should be understood that step A and step B can also be jointly performed by two different processors or separately performed in one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0165] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-described distributed agent ad hoc network method based on dynamic node cooperation is implemented.

[0166] It should be noted that, in order to simplify the description of the present invention disclosure and thus help the understanding of one or more embodiments of the present invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes merged into one embodiment, drawing, or description thereof.

Claims

1. A distribution agent ad-hoc network method based on dynamic node collaboration, characterized in that The method includes: Obtaining a network status data set of all distribution nodes within a target network area, where the network status data set includes real-time communication load characteristics of nodes, link stability characteristics between nodes, and node resource occupancy characteristics; Generating dynamic cooperation topology information based on the network status data set, where the dynamic cooperation topology information includes a node cooperation relationship topology graph and a communication path weight table; Generating a distribution proxy policy set according to the dynamic cooperation topology information, where the distribution proxy policy set includes an initial path allocation policy, a dynamic load adjustment rule, and a path switching trigger condition; Performing path optimization processing on a target distribution task based on the distribution proxy policy set to obtain an optimized distribution path set, deploying the optimized distribution path set to corresponding distribution nodes to perform dynamic cooperation distribution operations, and updating the network status data set based on real-time feedback data during the distribution process; The generating dynamic cooperation topology information based on the network status data set includes: Performing normalization processing on the real-time communication load characteristics in the network status data set to obtain a load balancing coefficient for each node, where the load balancing coefficient is used to characterize the remaining communication capacity characteristics of the node in the current period; Performing segmented weighting processing on the link stability characteristics to extract a communication stability score between nodes, where the communication stability score is used to characterize the anti-interference ability characteristics and historical communication success rate characteristics of the link between nodes; Calculating a resource availability index of a node according to the difference between the node resource occupancy characteristics and a preset resource threshold, where the resource availability index is used to characterize the resource reservation characteristics of the node when performing a distribution task; Constructing a dynamic cooperation decision matrix based on the load balancing coefficient, communication stability score, and resource availability index, where the dynamic cooperation decision matrix includes cooperation weight allocation characteristics between nodes and node priority ranking characteristics; Generating dynamic cooperation topology information including a node cooperation relationship topology graph and a communication path weight table according to the dynamic cooperation decision matrix; The generating a distribution proxy policy set according to the dynamic cooperation topology information includes: Extracting node hierarchy characteristics corresponding to the node cooperation relationship topology graph from the dynamic cooperation topology information, where the node hierarchy characteristics are generated by a dynamic election algorithm and include core node identification characteristics, relay node identification characteristics, and edge node identification characteristics; the dynamic election algorithm periodically updates the node hierarchy based on the resource availability index, historical task success rate, and network topology connectivity characteristics of the node; Determining the path selection priority between nodes according to the communication path weight table, and generating an initial path allocation policy in combination with the node hierarchy characteristics, where the initial path allocation policy is used to define the default communication path characteristics from the core node to the edge node; Generating a dynamic load adjustment rule based on the correlation between the real-time communication load characteristics and the resource availability index, where the dynamic load adjustment rule includes load transfer threshold characteristics and resource preemption priority characteristics between nodes; Generate a path switching trigger condition based on the comparison result between the communication stability score and a preset stability threshold, where the path switching trigger condition is used to define the path switching threshold for the standby path activation rule when the link is interrupted; Integrate the initial path allocation strategy, dynamic load adjustment rule, and path switching trigger condition to generate a set of distribution agent strategies containing multi-dimensional decision rules.

2. The method for distributed agent ad hoc network based on dynamic node cooperation according to claim 1, wherein Perform path optimization processing on the target distribution task based on the set of distribution agent strategies to obtain an optimized set of distribution paths, deploy the optimized set of distribution paths to the corresponding distribution nodes to perform dynamic cooperative distribution operations, and update the network state data set based on the real-time feedback data during the distribution process, including: Analyze the task attribute characteristics of the target distribution task, where the task attribute characteristics include task data volume characteristics, transmission timeliness requirement characteristics, and data security level characteristics; Match the resource availability indicators in the dynamic cooperative decision matrix according to the task data volume characteristics, and filter out a subset of distribution nodes that meet the resource reservation characteristics; Based on the default communication path characteristics from the core node to the edge node defined in the initial path allocation strategy, combined with the transmission timeliness requirement characteristics, extract a set of candidate paths that meet the communication stability score from the subset of distribution nodes, where the set of candidate paths includes multiple node cooperation sequences and corresponding path weight allocation characteristics; Dynamically select an encryption algorithm according to the data security level characteristics, encrypt the set of candidate paths, generate a set of secure transmission paths containing encrypted node identification characteristics and data sharding rules, and correct the weights of the set of secure transmission paths to obtain an optimized set of distribution paths; Split the node cooperation sequences in the optimized set of distribution paths into multiple sub-task units, and the unit characteristics of each sub-task unit include target node identification characteristics, data transmission volume characteristics, and path weight execution characteristics; Allocate the sub-task units to the corresponding distribution nodes according to the unit characteristics, and trigger the cooperative handshake protocol between nodes to establish a communication link; Monitor the change of the communication load characteristics of the nodes in real time during data transmission according to the dynamic load adjustment rule, and start the load migration operation when it is detected that the real-time communication load of a node exceeds the load transfer threshold characteristic; When it is monitored that the stability score of the current communication link is lower than the path switching threshold corresponding to the path switching trigger condition, activate the standby path from the set of candidate paths and switch to the standby path to continue the transmission; Collect the feedback data of each node after the distribution task is completed, where the feedback data includes actual transmission delay characteristics, resource consumption characteristics, and link interruption times characteristics, and update the real-time communication load characteristics and link stability characteristics in the network state data set based on the feedback data.

3. The method for a distributed agent ad-hoc network based on dynamic node cooperation according to claim 2, wherein The allocating the sub-task units to the corresponding distribution nodes according to the unit characteristics includes: Analyze the unit characteristics in the sub-task unit, and extract the weight allocation ratio characteristics and data transmission timing characteristics between nodes; Split the data volume of the subtask unit into multiple data shards according to the weight allocation ratio feature, and each data shard corresponds to a distribution node and a transmission time window feature; Generate an inter-node transmission scheduling instruction based on the transmission time window feature, and the transmission scheduling instruction includes a sending start time feature and a receiving timeout threshold feature of the data shard; Establish a two-way confirmation mechanism between the distribution nodes. After the sending node sends the data shard within the window corresponding to the sending start time feature, the receiving node returns a confirmation signal within the receiving timeout threshold feature; If the receiving node does not return a confirmation signal within the timeout threshold, reallocate the data shard to the node corresponding to the alternate path according to the path switching trigger condition.

4. The method for a distributed agent ad hoc network based on dynamic node collaboration according to claim 2, wherein During the data transmission process, the communication load feature change of the node is monitored in real time according to the dynamic load adjustment rule. When it is detected that the real-time communication load of the node exceeds the load transfer threshold feature, start the load migration operation, including: Collect the real-time bandwidth occupancy rate feature, CPU usage rate feature and memory consumption rate feature of the node in each transmission time window; Calculate the instantaneous load pressure value of the current node according to the real-time bandwidth occupancy rate feature. When the instantaneous load pressure value exceeds the load transfer threshold, mark the node as an overloaded node; Extract the data shard list currently processed by the overloaded node, and determine the data shards to be migrated and the target migration node according to the resource preemption priority feature of the dynamic load adjustment rule; Transfer the data shards to be migrated from the overloaded node to the target migration node, and update the path weight allocation feature in the distribution path set; After the migration is completed, recalculate the resource availability metrics of the overloaded node and the target migration node, and synchronously update the dynamic cooperation decision matrix.

5. The method for distributed agent ad hoc network based on dynamic node cooperation according to claim 2, characterized in that, When it is monitored that the stability score of the current communication link is lower than the path switching threshold corresponding to the path switching trigger condition, activate the alternate path from the candidate path set and switch to the alternate path to continue the transmission, including: Obtain the historical stability score curve of the current communication link, and extract the stability score mean feature within the last N time windows; When the stability score mean feature is lower than the path switching threshold, determine that the current link is in an unstable state; Screen the alternate paths with a stability score mean higher than the path switching threshold from the candidate path set, and extract the node cooperation sequence feature of the alternate path; Send a link switching request to the distribution nodes involved in the alternate path, and terminate the data transmission of the current link after receiving the ready confirmation from all nodes; Reallocate the uncompleted data shards to the node cooperation sequence corresponding to the alternate path, and adjust the transmission timing according to the path weight allocation feature of the alternate path.

6. The method for a distributed agent ad-hoc network based on dynamic node collaboration according to claim 2, wherein Update the real-time communication load feature and link stability feature in the network state data set based on the feedback data, including: Perform statistical analysis on the actual transmission delay feature, and extract the average delay feature and the maximum delay fluctuation feature of each node during the task execution; Calculate the remaining resource availability index of the node after task execution according to the resource consumption characteristics, and generate a resource consumption rate curve in combination with the initial resource availability index; Update the stability score of the corresponding link based on the link interruption times characteristic. If the link interruption times exceed the historical average, reduce the stability score of the link; Integrate the updated average delay characteristic, resource consumption rate curve and stability score into the network state data set, covering the original real-time communication load characteristic and link stability characteristic; Re-execute the generation operation of the dynamic cooperation topology information according to the updated network state data set to adapt to the latest network environment state.

7. The method for a distributed agent ad hoc network based on dynamic node cooperation according to claim 6, wherein The re-executing the generation operation of the dynamic cooperation topology information according to the updated network state data set to adapt to the latest network environment state includes: Recalculate the load balancing coefficient of all nodes based on the updated real-time communication load characteristic, and adjust the cooperation weight allocation characteristic in the dynamic cooperation decision matrix; Regenerate the communication stability score between nodes according to the updated link stability characteristic, and correct the path connection priority in the node cooperation relationship topology graph; Update the resource availability index of the node based on the latest resource consumption rate curve, and synchronously adjust the resource reservation characteristic in the dynamic cooperation decision matrix; Merge the corrected dynamic cooperation decision matrix with the node cooperation relationship topology graph to generate updated dynamic cooperation topology information; Trigger the regeneration operation of the distribution agent policy set according to the updated dynamic cooperation topology information to optimize the path allocation rule of the subsequent distribution task.

8. A distribution agent ad-hoc network system based on dynamic node cooperation, characterized in that Comprising a processor and a memory, the memory is connected to the processor, the memory is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the memory to implement the distribution agent ad-hoc network method based on dynamic node cooperation according to any one of claims 1-7 above.

Images