Multi-agent cooperative data exchange dynamic routing optimization method and system

Abstract

The application discloses a multi-agent cooperative data exchange dynamic routing optimization method and system, relates to the technical field of network communication, and comprises the following steps: acquiring real-time state information of each node in a network and outputting a node state data set; constructing a historical state sequence of each node, calculating a load change trend, and outputting a routing adjustment signal when the load change trend reaches a load critical value; calculating a cooperation weight according to a network contribution degree of each agent, establishing a task allocation mapping relationship between the agents based on the cooperation weight, and outputting a cooperation scheduling result; determining a candidate path set, performing index evaluation on paths in the candidate path set, and outputting an optimal routing path; and performing data transmission, acquiring a quality difference between an actual transmission effect and an expected effect, and performing parameter correction. The application realizes active optimization and multi-target balance of network routing, and improves routing efficiency and system stability in a dynamic network environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication, in particular to a multi-agent collaborative data exchange dynamic routing optimization method and system. BACKGROUND

[0002] The application of multi-agent systems in network routing optimization is becoming increasingly mature, and existing research mainly focuses on distributed routing algorithms, load balancing mechanisms, and collaborative control strategies. Current multi-agent routing systems generally use reinforcement learning-based methods to achieve path selection and load distribution through Q-learning, deep neural networks, and other technologies. At the same time, the software-defined network (SDN) architecture provides a new technical framework for centralized routing control, making it possible to obtain global network state and plan paths. At the algorithm level, swarm intelligence algorithms such as ant colony optimization and particle swarm optimization are widely used in multi-objective routing optimization problems, laying a theoretical foundation for collaborative decision-making among agents.

[0003] However, existing multi-agent routing technologies still have significant shortcomings in dynamic network environments. First, most existing solutions use a passive response mechanism, which only initiates routing adjustments after network congestion or failures occur, resulting in system response lag and inability to effectively prevent network performance degradation. Second, the collaborative mechanism between agents lacks an effective contribution evaluation system, often using simple polling or random allocation strategies, failing to fully leverage the network resource advantages of each agent, resulting in low collaborative efficiency. Third, existing hierarchical architectures often lack organic connections between levels, with global planning and local execution disjointed, making it difficult to achieve true multi-level collaborative optimization. In addition, traditional routing algorithms focus on single-objective optimization, such as shortest path or minimum latency, lacking comprehensive consideration of transmission quality, energy consumption control, and load balancing, and unable to meet the multi-dimensional performance requirements in complex network environments. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] Therefore, the present application provides a multi-agent collaborative data exchange dynamic routing optimization method and system, which can solve the problems mentioned in the background art.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the application provides a multi-agent cooperative data exchange dynamic routing optimization method, comprising: acquiring real-time state information of each node in the network, and outputting a node state data set; constructing a historical state sequence of each node based on the node state data set, calculating a load change trend of each node through sequence analysis, and outputting a routing adjustment signal when the load change trend reaches a load critical value; in response to the routing adjustment signal, calculating a cooperation weight according to the network contribution degree of each agent, establishing a task allocation mapping relationship between agents based on the cooperation weight, and outputting a cooperative scheduling result; determining a candidate path set through a hierarchical processing architecture according to the cooperative scheduling result, evaluating the paths in the candidate path set according to indicators, and outputting an optimal routing path; performing data transmission based on the optimal routing path, acquiring a quality difference between actual transmission effect and expected effect, and performing parameter correction on routing decision according to the quality difference.

[0008] Preferably, the historical state sequence of each node is constructed based on the node state data set, and the load change trend of each node is calculated through sequence analysis, comprising: extracting state data of each node within a preset time period, and arranging the state data in chronological order to form the historical state sequence; performing change analysis on the historical state sequence to identify the regular characteristics of load change; calculating a load prediction value at a target time based on the regular characteristics; comparing the load prediction value with the load critical value, and generating the routing adjustment signal when the load prediction value exceeds the load critical value.

[0009] Preferably, the cooperation weight of each agent is calculated according to the network contribution degree of each agent, comprising: counting the number of path resources provided by each agent to the network to determine a path contribution value; calculating the corresponding relationship between the network load shared by each agent and the system average load to determine a load balancing contribution value; evaluating the improvement amplitude of system effect after each agent joins the cooperation to determine an effect optimization contribution value; and comprehensively operating the path contribution value, the load balancing contribution value and the effect optimization contribution value to obtain the cooperation weight of each agent.

[0010] Preferably, the candidate path set is determined through a hierarchical processing architecture, comprising: establishing a path basic framework based on the network topology structure at the global planning layer to form an initial path set; updating the initial path set by each agent exchanging local information based on the node state data set at the local coordination layer; selecting paths meeting the quality requirements from the updated path set through a quality index screening mechanism at the execution layer to constitute the candidate path set.

[0011] Preferably, the paths in the candidate path set are evaluated by the following metrics: obtaining the transmission latency metric, energy consumption metric, and load balancing metric of each candidate path; determining the calculation coefficient of each metric according to network application requirements; comprehensively calculating the metric values ​​and the corresponding calculation coefficients to obtain the comprehensive quality score of each candidate path; and selecting the path with the highest comprehensive quality score as the optimal routing path.

[0012] Preferably, the step of correcting the routing decision parameters based on the quality difference includes: calculating the deviation between the actual transmission delay and the expected transmission delay to obtain a first quality deviation; calculating the deviation between the actual energy consumption and the expected energy consumption to obtain a second quality deviation; combining the first quality deviation with a first correction factor, and combining the second quality deviation with a second correction factor to generate a parameter adjustment value; and applying the parameter adjustment value to the baseline value of the corresponding parameter in the routing decision process.

[0013] Preferably, the length of the preset time period is set according to the network status fluctuation frequency. When the network status fluctuation frequency is higher than the fluctuation threshold, the length of the preset time period is reduced, and when the network status fluctuation frequency is lower than the fluctuation threshold, the length of the preset time period is increased.

[0014] Secondly, this application also provides a multi-agent collaborative data exchange dynamic routing optimization system, comprising: a state acquisition module for acquiring real-time state information of each node in the network and outputting a node state data set; an analysis and early warning module for constructing a historical state sequence of each node based on the node state data set, calculating the load change trend of each node through sequence analysis, and outputting a routing adjustment signal when the load change trend reaches a load critical value; a collaborative calculation module for responding to the routing adjustment signal, calculating collaborative weights based on the network contribution of each agent, establishing a task allocation mapping relationship between agents based on the collaborative weights, and outputting a collaborative scheduling result; a hierarchical decision-making module for determining a candidate path set through a hierarchical processing architecture based on the collaborative scheduling result, evaluating the paths in the candidate path set, and outputting the optimal routing path; and a parameter correction module for executing data transmission based on the optimal routing path, acquiring the quality difference between the actual transmission effect and the expected effect, and correcting the routing decision parameters based on the quality difference.

[0015] Thirdly, this application also provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: acquiring real-time status information of each node in the network and outputting a node status data set; constructing a historical status sequence of each node based on the node status data set, calculating the load change trend of each node through sequence analysis, and outputting a routing adjustment signal when the load change trend reaches a load threshold; responding to the routing adjustment signal, calculating a collaborative weight based on the network contribution of each agent, establishing a task allocation mapping relationship between agents based on the collaborative weight, and outputting a collaborative scheduling result; determining a candidate path set through a hierarchical processing architecture based on the collaborative scheduling result, evaluating the paths in the candidate path set, and outputting the optimal routing path; executing data transmission based on the optimal routing path, acquiring the quality difference between the actual transmission effect and the expected effect, and correcting the routing decision parameters based on the quality difference.

[0016] Fourthly, this application also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program performs the following steps: acquiring real-time status information of each node in the network and outputting a node status data set; constructing a historical status sequence of each node based on the node status data set, calculating the load change trend of each node through sequence analysis, and outputting a routing adjustment signal when the load change trend reaches a load threshold; responding to the routing adjustment signal, calculating a collaborative weight based on the network contribution of each agent, establishing a task allocation mapping relationship between agents based on the collaborative weight, and outputting a collaborative scheduling result; determining a candidate path set through a hierarchical processing architecture based on the collaborative scheduling result, evaluating the paths in the candidate path set, and outputting the optimal routing path; executing data transmission based on the optimal routing path, acquiring the quality difference between the actual transmission effect and the expected effect, and correcting the routing decision parameters based on the quality difference.

[0017] Implementing this application has the following beneficial effects: This application provides a dynamic routing optimization method and system for multi-agent collaborative data exchange. When the historical state sequence of each node in the network is constructed, the load change trend can be calculated through sequence analysis, and a routing adjustment signal can be output before load anomalies occur. Therefore, routing adjustment can be transformed from passive response to active prevention, maintaining timely response even under conditions of rapid fluctuations in network load. When the network contribution of each agent is quantified, collaborative weights can be calculated and task allocation mapping relationships can be established, thereby avoiding the inefficiency of traditional polling or random allocation methods, achieving reasonable allocation of agent resources, and significantly improving the collaborative efficiency among agents.

[0018] When the global planning layer, local coordination layer, and execution layer in a layered processing architecture work together, unified decision-making can be achieved through information transfer between layers. This solves the problem of disconnect between global planning and local execution in existing technologies, enabling path selection to have both a global optimization perspective and consideration for local execution efficiency. When the actual transmission effect differs from the expected effect in terms of quality, the routing decision parameters can be adjusted through a parameter correction mechanism, thus forming a complete feedback optimization loop. This allows for continuous monitoring and improvement, achieving comprehensive optimization across multiple dimensions such as transmission latency, energy consumption, and load balancing.

[0019] In summary, this application not only achieves predictive analysis of network states, but also establishes an efficient collaborative mechanism among agents, ensuring both the globality of routing decisions and the local adaptability of the execution process. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is an overall flowchart of a multi-agent collaborative data exchange dynamic routing optimization method involved in this application;

[0022] Figure 2 This is an application environment diagram of a multi-agent collaborative data exchange dynamic routing optimization method involved in this application;

[0023] Figure 3 This is a schematic diagram of the overall structure of a multi-agent collaborative data exchange dynamic routing optimization system involved in this application;

[0024] Figure 4 This is a computer device diagram of a multi-agent collaborative data exchange dynamic routing optimization method involved in this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0026] In this application, an intelligent agent refers to a software program module deployed on network nodes, with one intelligent agent deployed for each network node participating in route optimization. The intelligent agent possesses autonomous decision-making and cooperative communication capabilities, enabling it to perceive the state information of its node, exchange information with other intelligent agents, and execute cooperative routing algorithms. Intelligent agents establish connections with other intelligent agents through pre-defined communication protocols, forming a distributed cooperative network. Therefore, the network contribution of an intelligent agent in this application essentially refers to the network contribution of the network node where the intelligent agent is deployed, and the task allocation among intelligent agents essentially refers to the task allocation among network nodes.

[0027] Multi-agent collaborative data exchange dynamic routing optimization technology has extremely wide applications in modern network communication. For example, it can be applied to various fields such as intelligent transportation systems, industrial internet, data center networks, and edge computing networks. When there are a large number of data exchange needs in the network or when facing dynamic load changes, multi-agent collaborative routing systems can achieve efficient path planning and load distribution in complex network environments.

[0028] In related technologies, traditional network routing systems select paths based on static routing tables or simple load balancing algorithms, outputting relatively fixed routing schemes. However, in dynamic network environments, network load, node states, and topology change frequently, making it a challenging problem to enable routing systems to possess predictive analysis and collaborative optimization capabilities.

[0029] To address the problem that related technologies cannot effectively predict network state changes and achieve efficient collaboration among agents, a dynamic routing optimization method for multi-agent collaborative data exchange is proposed. This method constructs a historical state sequence based on the real-time state information of each node and calculates the load change trend. It establishes a collaborative weight and task allocation mapping relationship based on the network contribution of each agent, determines the candidate path set through a hierarchical processing architecture and evaluates the indicators, and adjusts the routing decision parameters based on the quality difference between the actual transmission effect and the expected effect.

[0030] The multi-agent collaborative data exchange dynamic routing optimization method provided in this application can be applied to, for example... Figure 2In the application environment shown, network node 102 is connected to control center 104 via a communication link. A distributed database stores network status data and routing information that control center 104 needs to process. The distributed database can be integrated into control center 104 or deployed in the cloud or other distributed servers. Each network node acquires real-time status information and outputs a node status data set; based on the node status data set, it constructs a historical status sequence for each node, calculates the load change trend of each node through sequence analysis, and outputs a routing adjustment signal when the load change trend reaches a load threshold; in response to the routing adjustment signal, it calculates collaborative weights based on the network contribution of each agent, establishes a task allocation mapping relationship between agents based on the collaborative weights, and outputs a collaborative scheduling result; based on the collaborative scheduling result, it determines a candidate path set through a hierarchical processing architecture, evaluates the paths in the candidate path set, and outputs the optimal routing path; based on the optimal routing path, it executes data transmission, obtains the quality difference between the actual transmission effect and the expected effect, and adjusts the routing decision parameters based on the quality difference.

[0031] Network node 102 can be, but is not limited to, various routers, switches, base stations, edge computing devices, and IoT gateway devices. IoT gateway devices can be industrial gateways, smart home gateways, vehicle-to-everything (V2X) gateways, sensor gateways, etc. Edge computing devices can be edge servers, MEC devices, fog computing nodes, etc. Base station devices can be 5G base stations, 4G base stations, Wi-Fi access points, LoRa gateways, etc. Control center 104 can be an independent network control server, a distributed control system composed of multiple control servers, or a cloud control platform providing software-defined networking services.

[0032] In one exemplary embodiment, such as Figure 2 As shown, a dynamic routing optimization method for data exchange through multi-agent cooperation is provided, which is then applied to... Figure 1 The following steps, 202 to 210, are used as an example to illustrate the application environment.

[0033] Step 202: Obtain the real-time status information of each node in the network and output the node status data set.

[0034] The real-time status information includes the node's current load rate, link bandwidth usage, processor utilization, cache utilization, number of network connections, and packet transmission latency. This node status data set provides the foundational data support for subsequent load prediction and routing optimization.

[0035] Network nodes can be routers, switches, servers, base stations, or other network devices with data forwarding capabilities. Each node has status monitoring capabilities, enabling it to collect its own operating parameters and network performance indicators in real time.

[0036] For example, a status monitoring agent deployed on each network node periodically collects the node's operational status data. The monitoring agent can be a lightweight software module that samples the node status at preset time intervals (e.g., 1 second, 5 seconds, or 10 seconds). The collected status information is then standardized to form a uniformly formatted status data record.

[0037] The storage format of the node status data set can be represented as: {node_id, timestamp, load_rate, bandwidth_usage, CPU_usage, memory_usage, connection_count, delay}. Here, the node_id is used to uniquely identify a single node in the network, the timestamp records the specific time of data collection, and each performance indicator represents the node's current operating status in numerical form.

[0038] In practice, status information can be collected through Simple Network Management Protocol (SNMP), network device API interfaces, or dedicated monitoring protocols. The collection methods may differ for different types of network devices, but ultimately, the raw data must be converted into a standardized node status data format to ensure data consistency and processability.

[0039] The collected node status data is organized according to time series to form a dataset containing status information of all network nodes. This dataset not only includes a snapshot of the current state but also retains historical state records within a certain time window, providing the necessary data foundation for subsequent trend analysis and predictive calculations.

[0040] Step 204: Construct the historical state sequence of each node based on the node state data set, calculate the load change trend of each node through sequence analysis, and output the routing adjustment signal when the load change trend reaches the load critical value.

[0041] The historical state sequence is a combination of node state data recorded in chronological order, containing state information such as load rate and bandwidth utilization at consecutive time points. Load change trends are determined by calculating the rate and direction of change of load rate in the historical state sequence, reflecting the growth or decline of node load. The load threshold is a pre-set load threshold based on the node's processing capacity; when the predicted load exceeds this threshold, a routing adjustment mechanism is activated.

[0042] For example, state records of the same node at different times are extracted from the node state data set and arranged in ascending order of timestamps to form a historical state sequence for the node. By calculating the difference in load rate between adjacent time points in the sequence, load change rate data is obtained, and then the direction and speed of load change are analyzed. Based on the change rate data and historical load values, the load prediction value at the target time is calculated and compared with the load critical value.

[0043] Furthermore, in step 204, a historical state sequence of each node is constructed based on the node state data set, and the load change trend of each node is calculated through sequence analysis, including steps A1 to A4:

[0044] Step A1: Extract the status data of each node within a preset time period and arrange them in chronological order to form a historical status sequence.

[0045] The preset time period is a fixed time window used for trend analysis, such as the most recent 60 seconds, 120 seconds, or 300 seconds. The length of the time period affects the amount of data analyzed and the accuracy of the prediction, and needs to be selected according to the stability of the network environment.

[0046] For example, all status records of a single node within a preset time period are filtered from the node status data set. The filtered records are then sorted in ascending order by timestamp from earliest to latest to ensure the temporal continuity of the data. The sorted data constitutes the historical status sequence of the node, where each data point contains a timestamp and a corresponding load rate value. When data is missing in the historical status sequence, the missing data points are supplemented using a linear interpolation method, and the load rate value of the missing time point is calculated using two valid data points before and after the missing time point according to the time interval ratio.

[0047] The historical state sequence data is organized as follows: each node identifier corresponds to a time series array, and each element in the array contains two fields: timestamp and load rate. By traversing all records in the node state data set and grouping them according to the node identifier, an independent historical state sequence is constructed for each node.

[0048] Step A2 involves performing a change analysis on the historical state sequence to identify the regular characteristics of load changes.

[0049] The change analysis is achieved by calculating the characteristics of load rate changes over time, including determining the direction of change and calculating the rate of change. Regular characteristics include whether the load is trending upwards, downwards, or remaining stable, and the speed of the change.

[0050] For example, the difference in load rate between every two adjacent time points in the historical state sequence is calculated to obtain a load change rate sequence. The values ​​in the change rate sequence are summed. If the sum is positive, it indicates that the overall load is on an upward trend; if the sum is negative, it indicates that the overall load is on a downward trend; if the sum is close to zero, it indicates that the load remains relatively stable.

[0051] Furthermore, by performing linear fitting on the historical state sequence, the linear trend of load change is determined. The linear fitting calculates the slope and intercept of the fitted line using the least squares method, where the slope value reflects the rate of load change, a positive slope indicates an upward trend, a negative slope indicates a downward trend, and the absolute value of the slope reflects the speed of change.

[0052] Step A3: Calculate the predicted load value at the target time based on the regularity characteristics.

[0053] The target time is a point in time after the current time, usually set as the time after the next data acquisition cycle or several cycles later. The load prediction value is calculated by extending the change pattern identified in step A2 to the target time.

[0054] For example, using the linear fitting parameters obtained in step A2, the target time value is substituted into the fitted line equation to calculate the predicted load value at the target time. The calculation method is as follows: multiply the target time by the slope of the fitted line, and add the intercept value of the fitted line to obtain the predicted load rate. Linear fitting requires at least three data points in the historical state sequence. When there are fewer than three data points, the load rate at the most recent time is used as the predicted value.

[0055] To enhance the reliability of the forecast, the results of multiple forecasting methods are comprehensively evaluated. In addition to linear fitting, a simple forecast value based on the load change rate at the most recent time point is also calculated. The simple forecast value is obtained by calculating the average load change rate at the three most recent time points and adding this average value to the current load rate. The linear fitting forecast value and the simple forecast value are then weighted and averaged in a 7:3 ratio to obtain the final load forecast value.

[0056] Step A4: Compare the load forecast value with the load threshold. When the load forecast value exceeds the load threshold, generate a routing adjustment signal.

[0057] The load threshold is determined based on the node's maximum processing capacity and quality of service requirements, and is typically set to 70% to 90% of the node's maximum load capacity. When the predicted load exceeds this threshold, it indicates that the node will face insufficient processing capacity and needs to distribute some data traffic to other nodes through routing adjustments.

[0058] For example, the load forecast value calculated in step A3 is numerically compared with the load threshold. If the load forecast value is greater than the load threshold, a routing adjustment signal is created that includes the node identifier, current load, forecast load, threshold, and degree of overrun. The degree of overrun is represented by calculating the percentage of the difference between the forecast load and the threshold relative to the threshold.

[0059] The routing adjustment signal includes a signal type identifier, the identifier of the node that exceeded the limit, the node's current load rate, the predicted load rate, the preset threshold, the percentage of exceeding the limit, and the timestamp of signal generation. This signal will be transmitted to the collaborative computing module as a trigger condition for initiating agent-based collaborative scheduling.

[0060] It should be noted that the length of the preset time period is set based on the frequency of network state fluctuations. The frequency of network state fluctuations is quantified by calculating the standard deviation of node state data within a fixed time period. A large standard deviation indicates frequent changes in network state, while a small standard deviation indicates relatively stable network state.

[0061] Specifically, when the frequency of network status fluctuations exceeds a preset fluctuation threshold, the length of the preset time period is set to a smaller value, such as adjusting it from the default 120 seconds to 60 seconds, to improve the responsiveness to rapid changes in network status. When the frequency of network status fluctuations is below the preset fluctuation threshold, the length of the preset time period is set to a larger value, such as adjusting it from the default 120 seconds to 240 seconds, to obtain more sufficient historical data to support predictive analysis. The preset fluctuation threshold is set to 1.5 times the standard deviation under normal network operation conditions, and the time period length is adjusted to ensure that it includes no fewer than 5 and no more than 20 valid data points.

[0062] By adjusting the length of a preset time period based on the frequency of network status fluctuations, the analysis time window is matched with the dynamic characteristics of the network. This ensures both prediction accuracy and system response speed, thereby improving the adaptability and effectiveness of the load prediction mechanism.

[0063] Step 206: Respond to the routing adjustment signal, calculate the collaborative weight based on the network contribution of each agent, establish the task allocation mapping relationship between agents based on the collaborative weight, and output the collaborative scheduling result.

[0064] The network contribution is a comprehensive evaluation index calculated by weighted summation of three components: path contribution, load balancing contribution, and effect optimization contribution. The collaboration weight is obtained by standardizing the network contribution, with a value between 0 and 1; a higher value indicates stronger collaboration ability of the agent. The task allocation mapping establishes the correspondence between agent identifiers and assigned tasks, clarifying the specific responsibilities of each agent in the collaboration process.

[0065] For example, after receiving the routing adjustment signal output in step 204, the node identifier and load warning information contained in the signal are identified. Based on the historical performance data and current state information of each agent in the network, the contribution value of each agent in the three dimensions of path provision, load sharing, and effect improvement are calculated respectively. The contribution values ​​of the three dimensions are weighted and summed according to preset weights to obtain the network contribution of each agent. Then, the network contribution is divided by the maximum value of the contribution values ​​of all agents to obtain the collaborative weight. According to the magnitude of the collaborative weight and the specific requirements of routing adjustment, corresponding scheduling tasks are assigned to each agent.

[0066] Furthermore, step 206 calculates the collaborative weights based on the network contribution of each agent, including steps B1 to B4:

[0067] Step B1: Count the number of path resources provided by each agent to the network and determine the path contribution value.

[0068] The number of path resources refers to the number of available paths that an agent can provide for data transmission, including primary paths and backup paths. The path contribution value is calculated by dividing the number of available paths provided by an agent by the total number of paths provided by all agents in the network, and its value ranges from 0 to 1.

[0069] For example, the network topology is traversed to identify the number of neighboring nodes and link states connected to each agent node. The total number of reachable paths from the agent node to all other nodes in the network is counted, including direct paths and multi-hop paths. The reachable paths are calculated using a path search algorithm from graph theory, with the search depth limited to 3 hops to ensure the usability of the paths.

[0070] The path contribution value is calculated by dividing the number of available paths provided by the agent by the total number of paths provided by all agents in the network. When an agent node fails or the link is interrupted, the number of available paths for that agent decreases accordingly, and the path contribution value is recalculated.

[0071] Step B2: Calculate the correspondence between the network load shared by each agent and the average system load, and determine the load balancing contribution value.

[0072] Network load balancing refers to the amount of data traffic handled by an agent during data forwarding, including the number of data packets forwarded and the bandwidth resources used. The system average load is the arithmetic mean of the loads handled by all agents. The load balancing contribution is determined by calculating the deviation between the actual load of an agent and the system average load; the smaller the deviation, the greater the contribution.

[0073] For example, load statistics for each agent over a recent time period are collected, including metrics such as the number of data packets processed, bandwidth usage, and CPU utilization. The average load of all agents in the system is calculated as a baseline load. The actual load of each agent is compared with the baseline load to calculate the load deviation.

[0074] The specific calculation method for the load balancing contribution value is as follows: divide the baseline load by the absolute value of the difference between the agent's actual load and the baseline load, plus 1. When the difference is zero, the contribution value is the baseline load. This calculation method ensures that agents with loads close to the average level receive higher contribution scores.

[0075] Step B3: Evaluate the improvement in system performance after each agent joins the collaboration, and determine the contribution value for performance optimization.

[0076] The system performance is evaluated based on metrics such as network transmission latency, packet loss rate, and throughput. The improvement is measured by comparing the system performance before and after agent participation in collaboration. The contribution value of the performance optimization is calculated based on a weighted composite score of performance metric improvements.

[0077] For example, baseline system performance data is recorded before the agent joins the cooperative scheduling process, including key metrics such as average transmission latency, packet loss rate, and network throughput. After the agent participates in the cooperative scheduling, the same performance metrics are remeasured, and the magnitude of change in each metric is calculated. The greater the magnitude of performance improvement, the more significant the agent's contribution to the optimization.

[0078] The contribution value of performance optimization is calculated using a weighted comprehensive scoring method, which weights the improvement in transmission latency, reduction in loss rate, and increase in throughput in a 4:3:3 ratio. When the participation of a certain agent leads to a decrease in system performance, the performance optimization contribution value of that agent is set to zero to avoid negative contributions affecting the overall evaluation.

[0079] Step B4 involves performing a comprehensive calculation on the path contribution value, load balancing contribution value, and effect optimization contribution value to obtain the collaborative weight of each agent.

[0080] The collaborative weight is calculated by weighting and summing the path contribution value, load balancing contribution value, and effect optimization contribution value according to preset weight coefficients, and then standardizing the result so that the final result ranges from 0 to 1.

[0081] For example, the three contribution values ​​obtained in steps B1, B2, and B3 are weighted according to a weight ratio of 3:4:3 to obtain the original network contribution value of each agent. The maximum value among the original network contribution values ​​of all agents is used as the standardization benchmark, and the collaborative weight of each agent is equal to its original network contribution value divided by the standardization benchmark. The weight ratio is set based on network performance priority, with the highest weight for load balancing contribution reflecting the importance of load sharing in collaboration.

[0082] After the collaborative weight calculation is completed, the agents are sorted according to their weights, with agents having higher weights participating in collaborative scheduling tasks first. When multiple agents have similar weights, the agent whose network location is closer to the node with abnormal load is selected to reduce the number of data transmission hops and latency.

[0083] Based on collaborative weights, a task allocation mapping relationship is established between agents. The route adjustment task is decomposed into sub-tasks such as path search, load sharing, and performance monitoring. The corresponding sub-tasks are assigned according to the collaborative weights and capabilities of each agent. The collaborative scheduling result includes a list of participating agents, the specific tasks undertaken by each agent, the expected load sharing ratio, and the time schedule for collaborative execution.

[0084] It should be noted that the calculation cycle of the collaborative weights is synchronized with the state information collection cycle in step 202 to ensure that the weights reflect the current state of the agents. When the network topology changes or the state of the agents changes significantly, the collaborative weights of each agent are recalculated to maintain the timeliness and accuracy of task allocation. The historical records of the collaborative weights are used to analyze the performance trends of the agents, providing data support for long-term network optimization strategies.

[0085] Step 208: Based on the collaborative scheduling results, determine the candidate path set through a hierarchical processing architecture, evaluate the paths in the candidate path set, and output the optimal routing path.

[0086] The layered processing architecture consists of three layers: a global planning layer, a local coordination layer, and an execution layer. Each layer undertakes different path processing tasks. The candidate path set comprises combinations of paths that meet quality requirements after being filtered through these three layers. Evaluation metrics are calculated by taking values ​​from three dimensions: transmission latency, energy consumption, and load balancing, and then summing them using a weighted average to arrive at a comprehensive score. The optimal route is the single path with the highest comprehensive score from the candidate path set.

[0087] For example, the cooperative scheduling result output in step 206 is received, and a list of participating agents and task allocation information are obtained. A path search space for the network topology is established at the global planning layer, generating all possible transmission paths. At the local coordination layer, agents exchange current node state information and update path availability based on real-time load. At the execution layer, paths are quality-screened, retaining paths that meet basic transmission requirements to form a candidate set. Performance metrics are calculated for each candidate path and weighted summed, selecting the path with the highest score as the optimal routing scheme.

[0088] Furthermore, step 208 determines the candidate path set through a hierarchical processing architecture, including steps C1 to C3:

[0089] Step C1: Establish a basic path framework based on the network topology at the global planning layer to form an initial path set.

[0090] The network topology includes a graphical representation of all network nodes and their connections. The basic path framework is the set of all theoretically reachable paths from the source node to the target node. The initial path set is generated through graph theory search and contains all path combinations with a hop count not exceeding a preset upper limit.

[0091] For example, based on the coordinated scheduling results, the source nodes that need load balancing and the target nodes that can receive load balancing are determined. A breadth-first search algorithm is used to traverse the network topology, expanding layer by layer from the source nodes to search for all paths to the target nodes. During the search, a maximum hop count limit of 5 hops is set to avoid excessively long paths that would reduce transmission efficiency. All searched paths are recorded according to their hop count and node sequence to form an initial path set.

[0092] The path record format is: {path_id, source_node, target_node, middle_nodes, total hop count}. When there are multiple different paths from the source node to the target node, each path is assigned a unique path identifier for distinction.

[0093] In step C2, each agent in the local coordination layer exchanges local information based on the node state data set and updates the initial path set.

[0094] Local information includes the current load rate of the node where the agent resides, link bandwidth occupancy, and connectivity status of neighboring nodes. Information exchange is achieved through an inter-agent communication protocol, where each agent sends status query requests to other agents along the path and receives responses. The path set is updated based on the acquired real-time status information, removing unavailable paths and marking the current quality status of each path.

[0095] For example, for each path in the initial path set, the agent at the starting node of the path sends a status query message to all intermediate nodes on the path. Upon receiving the query, the agents at the intermediate nodes reply with the current node's load rate, available bandwidth, and link connectivity status. When the load rate of an intermediate node exceeds 90% or the link is interrupted, that path is removed from the initial path set.

[0096] For paths retained in the set, the current path quality metric is calculated based on the load status of each node on the path. The path quality metric is equal to the arithmetic mean of the load rates of all nodes on the path, and is used for subsequent path optimization. The updated path set contains both the path's topology information and its current quality metric.

[0097] Step C3: At the execution layer, a quality index screening mechanism is used to select paths that meet the quality requirements from the updated path set to form a candidate path set.

[0098] The quality indicator screening mechanism filters paths based on preset quality thresholds, including constraints such as maximum allowable latency, minimum available bandwidth, and maximum load rate. Paths that meet the quality requirements are those that simultaneously satisfy all constraints. The candidate path set consists of the remaining path combinations after quality screening, used for subsequent detailed evaluation.

[0099] For example, the constraints for quality screening are set as follows: the average path load rate does not exceed 80%, the available path bandwidth is not less than 100 Mbps, and the estimated path transmission delay does not exceed 50 milliseconds. Each path in the updated path set is checked one by one to see if it meets all the constraints. The average path load rate is calculated using the result of step C2, the available path bandwidth is taken as the minimum of the available bandwidth of all links on the path, and the estimated path transmission delay is calculated by summing the hop count and link delay.

[0100] Only paths that simultaneously satisfy all three constraints are retained to form a candidate path set. When the number of paths in the candidate path set exceeds 10, they are sorted according to their quality indicators, and the top 10 paths with the best quality are retained to avoid excessive computational burden in subsequent evaluations. The candidate path set will serve as input for subsequent detailed evaluations.

[0101] Furthermore, step 208 involves evaluating the paths in the candidate path set using various metrics, including steps D1 to D4:

[0102] Step D1: Obtain the transmission latency, energy consumption, and load balancing metrics for each candidate path.

[0103] The transmission latency metric is the total time required for a data packet to travel from the source node to the destination node, including node processing latency and link transmission latency. The energy consumption metric is the total energy consumed by all nodes along the path in forwarding data packets. The load balancing metric is derived by calculating the impact of the path on the existing network load distribution; a smaller value indicates a smaller impact on load balancing.

[0104] For example, for each path in the candidate path set, a transmission latency metric is calculated: the average processing latency of each node on the path is summed with the transmission latency of each link. Node processing latency is calculated based on the node's current load rate and processing capacity; a higher load rate results in greater processing latency. Link transmission latency is calculated based on the link length and transmission rate.

[0105] The energy consumption metric is calculated as follows: the energy consumption required for each node on the path to forward a unit of data is calculated based on the expected data transmission volume, and then the total energy consumption of the path is obtained by summing the results. The load balancing metric is obtained by calculating the standard deviation of the load rate of each node after the path is used; the larger the standard deviation, the more uneven the load distribution.

[0106] Step D2: Determine the calculation coefficients for each indicator based on the network application requirements.

[0107] The calculation coefficients represent the weights of each performance indicator in the comprehensive evaluation, determined based on the current network application's sensitivity to latency, energy consumption control requirements, and the importance of load balancing. For latency-sensitive applications, the calculation coefficient for transmission latency is increased; for energy-sensitive applications, the calculation coefficient for energy consumption is increased; and for load-sensitive applications, the calculation coefficient for load balancing is increased.

[0108] For example, the calculation coefficients are determined based on the type of data stream being processed. For real-time audio and video transmission, the transmission latency coefficient is set to 0.6, the energy consumption coefficient to 0.2, and the load balancing coefficient to 0.2, emphasizing the importance of latency performance. For large file transmission, the transmission latency coefficient is set to 0.2, the energy consumption coefficient to 0.3, and the load balancing coefficient to 0.5, emphasizing load balancing and energy consumption control.

[0109] The sum of the calculated coefficients is always equal to 1.0 to ensure the stability of the numerical range of the comprehensive evaluation results. When network application requirements change, the calculated coefficients are reallocated according to the new application characteristics to maintain the match between the evaluation criteria and application requirements.

[0110] Step D3 involves combining the values ​​of each indicator with their corresponding calculation coefficients to obtain the overall quality score for each candidate path.

[0111] The overall quality score is calculated by multiplying each of the three performance indicators by its calculated coefficient and then summing the results. To ensure comparability between the different indicators, each indicator is standardized before calculation, converting the indicator values ​​into standardized values ​​between 0 and 1. The overall quality score ranges from 0 to 1, with a higher value indicating better overall path performance.

[0112] For example, the same metric for all paths in the candidate path set is standardized: the minimum value of the metric is set to 0, the maximum value to 1, and other values ​​are proportionally converted. Transmission latency is inversely standardized, meaning the minimum latency corresponds to 1 and the maximum latency corresponds to 0. Energy consumption is also inversely standardized. Load balancing is inversely standardized, meaning the minimum standard deviation corresponds to 1.

[0113] The overall quality score is calculated as follows: the standardized transmission delay index multiplied by its calculation coefficient, plus the standardized energy consumption index multiplied by its calculation coefficient, and finally the standardized load balancing index multiplied by its calculation coefficient. After calculation, each candidate path receives an overall quality score between 0 and 1.

[0114] Step D4: Select the path with the highest overall quality score as the optimal routing path.

[0115] The optimal route is the single path with the highest overall quality score among the candidate paths. When multiple paths have the same highest score, the path with fewer hops is selected as the optimal path to reduce transmission complexity. After the optimal route is determined, the path information is sent to the relevant agents for route configuration.

[0116] For example, the overall quality scores of all paths in the candidate path set are numerically compared to identify the path with the highest score. If path A scores 0.85, path B scores 0.82, and path C scores 0.85, then the path with the fewer hops between paths A and C with the same score is selected. After the optimal route path is determined, route configuration instructions are generated and sent to each agent node on the path.

[0117] The routing configuration command includes the path identifier, source node, destination node, complete node forwarding sequence, and expected data traffic volume. After receiving the command, each agent adds the corresponding forwarding rules to its local routing table to prepare for the upcoming data traffic.

[0118] It should be noted that step 208, through layered screening and comprehensive evaluation of multiple indicators, ensures that the selected optimal route path not only meets basic transmission requirements but also achieves balanced optimization across multiple performance dimensions. The processing time at each level is controlled within milliseconds to guarantee the real-time requirements of route selection.

[0119] Step 210: Perform data transmission based on the optimal route path, obtain the quality difference between the actual transmission effect and the expected effect, and adjust the parameters of the routing decision according to the quality difference.

[0120] The actual transmission performance is determined by monitoring the actual latency and energy consumption during data transmission. The expected performance is the theoretical value of the transmission latency and energy consumption indicators calculated in step D1. The quality difference is calculated by comparing the actual performance with the expected performance, including latency and energy consumption differences. Parameter correction adjusts the prediction model parameters used in subsequent routing decisions based on the magnitude and direction of the quality difference.

[0121] For example, after data transmission is initiated, various performance indicators are monitored in real time during the transmission process. The actual total latency of data packets being sent from the source node to the target node and received is recorded, and the actual energy consumption of each node along the path is statistically analyzed. The actual measured performance data is compared with the expected values ​​calculated in step D1.

[0122] The quality difference is calculated as follows: Actual latency - Expected latency = Latency difference; Actual energy consumption - Expected energy consumption = Energy consumption difference. A positive difference indicates that the actual performance is lower than expected, while a negative difference indicates that the actual performance is better than expected. The magnitude of the parameter correction is determined by the absolute value of the difference; the larger the difference, the larger the correction.

[0123] Parameter correction specifically includes adjusting the prediction baseline for node processing latency and revising the calculation baseline for energy consumption. The corrected parameters are used in the next routing decision process to gradually improve the accuracy of routing performance prediction and the effectiveness of route selection.

[0124] Furthermore, step 210 involves parameter correction of the routing decision based on the quality difference, including steps E1 to E4:

[0125] Step E1: Calculate the deviation between the actual transmission delay and the expected transmission delay to obtain the first quality deviation.

[0126] The actual transmission delay is the measured time for a data packet to complete transmission on the optimal routing path, starting from when the source node sends the data packet until the target node successfully receives it. The expected transmission delay is the theoretical delay value calculated in step D1 based on the processing capacity of the path nodes and the link transmission rate. The first quality deviation is calculated by subtracting the two; a positive value indicates that the actual performance is lower than expected, and a negative value indicates that the actual performance is better than expected.

[0127] For example, at the start of data transmission, the source node records the exact timestamp of the data packet's transmission. The data packet is forwarded through each intermediate node along the optimal routing path, and a receiving timestamp is recorded upon arrival at the destination node. Actual transmission delay = the time difference between the receiving timestamp and the sending timestamp. The actual transmission delay is then numerically compared with the expected transmission delay calculated in step D1.

[0128] The formula for calculating the first quality deviation is: Actual transmission delay - Expected transmission delay. For example, if the actual transmission delay is 45 milliseconds and the expected transmission delay is 40 milliseconds, then the first quality deviation is 5 milliseconds, meaning the actual delay is 5 milliseconds higher than the expected delay. If the actual transmission delay is 35 milliseconds, then the first quality deviation is -5 milliseconds, meaning the actual delay is 5 milliseconds lower than the expected delay.

[0129] Step E2: Calculate the deviation between the actual energy consumption and the expected energy consumption to obtain the second mass deviation.

[0130] The actual energy consumption is the total energy consumed by all nodes participating in data forwarding along the path during transmission, calculated by summing the energy monitoring data of each node. The expected energy consumption is the estimated value calculated in step D1 based on the theoretical energy consumption model of each node. The second quality deviation is calculated by subtracting the expected energy consumption from the actual energy consumption.

[0131] For example, after data transmission is completed, the actual energy consumption data of each node along the path during transmission is collected. Each node records the electrical energy consumed during data processing and forwarding through its built-in energy monitoring module, and the data is statistically analyzed in joules. The total actual energy consumption is obtained by summing the energy consumption values ​​of all nodes along the path.

[0132] The second mass deviation is calculated as: actual total energy consumption - expected total energy consumption calculated in step D1. For example, if the actual energy consumption is 12.8 joules and the expected energy consumption is 10.5 joules, then the second mass deviation is 2.3 joules, indicating that the actual energy consumption exceeds the expected 2.3 joules. The accuracy of energy monitoring is maintained within 0.1 joules to ensure the accuracy of the deviation calculation.

[0133] Step E3: Combine the first quality deviation with the first correction factor, and combine the second quality deviation with the second correction factor to generate parameter adjustment values.

[0134] The first correction factor is a conversion coefficient used to convert time delay deviation into an adjustment amount for time delay prediction parameters, and its value is determined based on statistical analysis of historical deviation data. The second correction factor is a conversion coefficient used to convert energy consumption deviation into an adjustment amount for energy consumption prediction parameters. The parameter adjustment value is calculated by multiplying the quality deviation by the corresponding correction factor and is used for subsequent correction of the parameter baseline value.

[0135] For example, the first correction factor is determined by collecting delay deviation data from the most recent 100 routing decisions, calculating the standard deviation of the deviation data, using the reciprocal of the standard deviation as the base value of the first correction factor, and then multiplying it by an adjustment coefficient of 0.1 to obtain the first correction factor. When the historical standard deviation of the delay deviation is 8 milliseconds, the first correction factor is equal to 1 / 8 × 0.1 = 0.0125.

[0136] The second correction factor is determined using the same method, calculated based on the standard deviation of historical energy consumption deviation data. When the historical standard deviation of energy consumption deviation is 2.5 joules, the second correction factor equals 1 / 2.5 × 0.1 = 0.04. The value of the correction factor is updated periodically, recalculated after every 50 routing decisions to maintain the adaptability of the correction mechanism.

[0137] The parameter adjustment values ​​are calculated as follows: the delay adjustment value is obtained by multiplying the first quality deviation by the first correction factor, and the energy consumption adjustment value is obtained by multiplying the second quality deviation by the second correction factor. Using the aforementioned example data, the delay adjustment value is 5 milliseconds × 0.0125 = 0.0625, and the energy consumption adjustment value is 2.3 joules × 0.04 = 0.092.

[0138] Step E4: Apply the parameter adjustment values ​​to the baseline values ​​of the corresponding parameters during the routing decision process.

[0139] The baseline values ​​for the corresponding parameters include the predicted baselines for node processing latency and node energy consumption, which are used in the performance metric calculations in step D1. The parameter adjustment values ​​are applied by adding the adjustment values ​​to the corresponding baseline values; a positive adjustment value increases the baseline value, and a negative adjustment value decreases it. The adjusted baseline values ​​are used for performance prediction calculations in the next routing decision process.

[0140] For example, the latency adjustment value is applied to the processing latency prediction baseline for each node on the path. Assuming the processing latency prediction baseline for a certain node is 3 milliseconds and the latency adjustment value is 0.0625, the adjusted prediction baseline is 3.0625 milliseconds. This adjusted value will be used in the next path latency calculation involving that node, improving the accuracy of latency prediction.

[0141] The energy consumption adjustment value is allocated according to the proportion of each node in the total energy consumption of the path. If the path contains 3 nodes, and the energy consumption proportions of each node are 40%, 35%, and 25% respectively, and the energy consumption adjustment value is 0.092 joules, then the energy consumption baseline of the three nodes will increase by 0.092 × 40% = 0.0368, 0.092 × 35% = 0.0322, and 0.092 × 25% = 0.023 joules respectively.

[0142] The adjustment range of the baseline value is constrained by upper and lower limits, with a single adjustment not exceeding 20% ​​of the original baseline value to avoid over-adjustment due to abnormal data. The adjusted baseline value is stored in the parameter database of each node, providing a more accurate predictive basis for subsequent routing decisions. When the cumulative adjustment of the baseline value exceeds 50% of the original value, the baseline value reset mechanism is triggered, and the baseline value is recalibrated using the latest standard test data.

[0143] It should be noted that the parameter correction in this application achieves self-optimization of routing performance prediction through continuous experimental feedback and benchmark adjustments. Historical data generated during the correction process is used to analyze network performance trends, providing data support for network capacity planning and performance optimization. Steps 202 to 210 constitute a complete multi-agent collaborative routing optimization closed loop, achieving continuous optimization and improvement of routing decisions through a combination of predictive analysis, collaborative scheduling, hierarchical processing, and parameter correction.

[0144] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0145] Based on the same inventive concept, this application also provides a multi-agent collaborative data exchange dynamic routing optimization system. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the multi-agent collaborative data exchange dynamic routing optimization system provided below can be found in the limitations of the multi-agent collaborative data exchange dynamic routing optimization method described above, and will not be repeated here.

[0146] In one exemplary embodiment, such as Figure 3 As shown, a multi-agent collaborative data exchange dynamic routing optimization system is provided, comprising:

[0147] The status acquisition module is used to acquire real-time status information of each node in the network and output a set of node status data.

[0148] The analysis and early warning module is used to construct the historical status sequence of each node based on the node status data set, calculate the load change trend of each node through sequence analysis, and output a routing adjustment signal when the load change trend reaches the load critical value.

[0149] The collaborative computing module is used to respond to routing adjustment signals, calculate collaborative weights based on the network contribution of each agent, establish task allocation mapping relationships between agents based on collaborative weights, and output collaborative scheduling results.

[0150] The hierarchical decision-making module is used to determine the candidate path set based on the collaborative scheduling results and through the hierarchical processing architecture, evaluate the paths in the candidate path set, and output the optimal routing path.

[0151] The parameter correction module is used to perform data transmission based on the optimal routing path, obtain the quality difference between the actual transmission effect and the expected effect, and correct the routing decision parameters based on the quality difference.

[0152] The modules in the aforementioned multi-agent collaborative data exchange dynamic routing optimization system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can invoke and execute the corresponding operations of each module.

[0153] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a dynamic routing optimization method for multi-agent collaborative data exchange. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0154] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0155] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0156] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0157] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0158] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0159] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0160] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0161] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A dynamic routing optimization method for data exchange in a multi-agent cooperative manner, characterized in that, include: Obtain real-time status information of each node in the network and output a set of node status data; Based on the node status data set, a historical status sequence of each node is constructed. The load change trend of each node is calculated through sequence analysis, and a routing adjustment signal is output when the load change trend reaches the load critical value. In response to the routing adjustment signal, cooperative weights are calculated based on the network contribution of each agent, including: The number of path resources provided by each agent to the network is counted to determine the path contribution value; the number of path resources refers to the number of available paths provided by the agent for data transmission, including the main path and backup paths; the path contribution value is the ratio of the number of available paths provided by the agent to the total number of paths provided by all agents in the network. Calculate the correspondence between the network load shared by each agent and the average system load, and determine the load balancing contribution value; Evaluate the improvement in system performance after each agent joins the collaboration, and determine the contribution value of the performance optimization; The path contribution value, the load balancing contribution value, and the effect optimization contribution value are combined and calculated to obtain the collaborative weight value of each agent. The network contribution is a comprehensive evaluation index calculated by weighted summation of three components: path contribution, load balancing contribution, and effect optimization contribution; the collaborative weight is obtained by standardizing the network contribution. Based on the collaborative weights, a task allocation mapping relationship is established between intelligent agents, and the collaborative scheduling result is output. Based on the collaborative scheduling results, a set of candidate paths is determined through a hierarchical processing architecture. The paths in the set of candidate paths are evaluated by metrics, and the optimal routing path is output. Data transmission is performed based on the optimal routing path, the quality difference between the actual transmission effect and the expected effect is obtained, and the routing decision parameters are adjusted according to the quality difference.

2. The multi-agent collaborative data exchange dynamic routing optimization method as described in claim 1, characterized in that: The process of constructing historical state sequences for each node based on the node state data set, and calculating the load change trend of each node through sequence analysis, includes: Extract the status data of each node within a preset time period and arrange them in chronological order to form the historical status sequence; The historical state sequence is analyzed to identify the regularity and characteristics of load changes; The predicted load value at the target time is calculated based on the aforementioned regularity characteristics; The load prediction value is compared with the load threshold. When the load prediction value exceeds the load threshold, the routing adjustment signal is generated.

3. The dynamic routing optimization method for multi-agent collaborative data exchange as described in claim 2, characterized in that: The process of determining the candidate path set through a hierarchical processing architecture includes: At the global planning layer, a basic path framework is established based on the network topology, forming an initial set of paths; In the local coordination layer, each agent exchanges local information based on the node state data set, and updates the initial path set; At the execution layer, a quality index screening mechanism is used to select paths that meet the quality requirements from the updated path set, forming the candidate path set.

4. The multi-agent collaborative data exchange dynamic routing optimization method as described in claim 3, characterized in that: The paths in the candidate path set are evaluated using metrics, including: Obtain the transmission latency, energy consumption, and load balancing metrics for each candidate path; The calculation coefficients for each indicator are determined based on the requirements of network applications; The comprehensive quality score of each candidate path is obtained by combining the values ​​of each indicator with the corresponding calculation coefficients. The path with the highest overall quality score is selected as the optimal routing path.

5. The multi-agent collaborative data exchange dynamic routing optimization method as described in claim 4, characterized in that: The step of adjusting the routing decision parameters based on the quality difference includes: The deviation between the actual transmission delay and the expected transmission delay is calculated to obtain the first quality deviation; The deviation between actual energy consumption and expected energy consumption is calculated to obtain the second mass deviation; The first quality deviation is combined with the first correction factor, and the second quality deviation is combined with the second correction factor to generate parameter adjustment values; The adjusted parameter values ​​are applied to the baseline values ​​of the corresponding parameters during the routing decision process.

6. The multi-agent collaborative data exchange dynamic routing optimization method as described in claim 5, characterized in that: The length of the preset time period is set according to the network status fluctuation frequency. When the network status fluctuation frequency is higher than the fluctuation threshold, the length of the preset time period is reduced, and when the network status fluctuation frequency is lower than the fluctuation threshold, the length of the preset time period is increased.

7. A multi-agent collaborative data exchange dynamic routing optimization system, employing the multi-agent collaborative data exchange dynamic routing optimization method as described in any one of claims 1 to 6, characterized in that, include: The status acquisition module is used to acquire real-time status information of each node in the network and output a set of node status data. The analysis and early warning module is used to construct the historical state sequence of each node based on the node status data set, calculate the load change trend of each node through sequence analysis, and output a routing adjustment signal when the load change trend reaches the load critical value. The collaborative computing module is used to respond to the routing adjustment signal, calculate the collaborative weight based on the network contribution of each agent, establish a task allocation mapping relationship between agents based on the collaborative weight, and output the collaborative scheduling result. The hierarchical decision-making module is used to determine a set of candidate paths based on the collaborative scheduling results through a hierarchical processing architecture, evaluate the paths in the set of candidate paths, and output the optimal routing path. The parameter correction module is used to perform data transmission based on the optimal routing path, obtain the quality difference between the actual transmission effect and the expected effect, and correct the routing decision parameters based on the quality difference.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the multi-agent cooperative data exchange dynamic routing optimization method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-agent cooperative data exchange dynamic routing optimization method as described in any one of claims 1 to 6.

Images