An airport cluster collaborative maintenance retransmission analysis method and system based on multiple agents

Abstract

The application relates to the technical field of resource allocation, in particular to an airport cluster cooperative maintenance retransmission analysis method and system based on multiple agents, which specifically comprises the following steps: analyzing the long-term change trend of the packet loss rate and the jitter intensity of the round-trip delay of each airport in each decision cycle, determining the communication quality stability in each decision cycle, determining a risk compensation coefficient according to the difference between the communication quality stability and a preset reference value, predicting the communication quality stability in a future decision cycle according to the communication quality stability in a historical decision cycle, and improving the timeout time by combining the two; and adjusting the retransmission waiting time of each time according to the timeout time improvement result, so that the contract net protocol can effectively adapt to the heterogeneous and high-dynamic network environment of the airport cluster, maintain stable performance in the competition mechanism and the arbitration process, and improve the consistency and timeliness of resource allocation between airports.

Description

Technical Field

[0001] This application relates to the field of resource allocation technology, specifically to a method and system for airport cluster collaborative maintenance retransmission analysis based on multi-agent systems. Background Technology

[0002] Airport cluster maintenance decisions have long relied on independent judgments from individual airports. The allocation of scarce maintenance resources (such as grouting units, inspection vehicles, and experts) is uneven, the cycle is long, and it is easy to conflict with the airport's operating window. Traditional multi-objective optimization algorithms are only applicable to single-airport scenarios and cannot solve the multi-dimensional coupling problem of cross-airport resources, risks, benefits, and lifespan. Although industry-wide models can assist in disease diagnosis and solution generation, they lack cross-airport role-based collaboration mechanisms and cannot form a closed-loop decision-making process.

[0003] In the decision-making process of collaborative maintenance of multi-agent airport clusters, the sensing agents, expert agents, and scheduling agents of each airport interact through the Contract Network Protocol (CNP) to achieve task bidding, competition, and adjudication. The Contract Network Protocol itself assumes reliable communication and a small number of participants. However, the multi-round handshake and timeout retransmission mechanism can lead to a sharp drop in efficiency in high-latency or weak network environments. When the airport cluster is large and the network conditions between airports differ significantly, message delays, packet loss, or inconsistent protocol parsing can easily occur, leading to asynchronous collaborative states or even deadlock. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a multi-agent-based airport cluster collaborative maintenance retransmission analysis method and system, the specific technical solution of which is as follows: In a first aspect, embodiments of this application provide a method for airport group collaborative maintenance retransmission analysis based on multi-agent systems, the method comprising the following steps: Step 1: Collect time-series data on packet loss rate and round-trip delay of each airport network within each decision cycle; Step 2: Analyze the overall packet loss rate of each airport in each decision cycle, as well as the degree of disorder in the distribution of round-trip delays of each airport within each preset sliding window, and determine the stability of communication quality of each airport in each decision cycle. Step 3: Determine the risk compensation coefficient for each airport in each decision cycle based on the difference between the communication quality stability and the preset extreme risk benchmark. Step 4: Based on the communication quality stability of each airport in the historical decision-making cycle, obtain the predicted value of the communication quality stability for the next decision-making cycle. Combine the risk compensation coefficient and the preset benchmark timeout time to determine the timeout time improvement result for each airport in each decision-making cycle. Step 5, the airport group pavement collaborative maintenance decision-making process includes: the airport group pavement defects are detected by the station sensing agent; once a defect is detected, the station sensing agent initiates a tender to the expert agent; the expert agent bids and provides maintenance solutions; if conflicting maintenance solutions exist, they are handled by the arbitration agent, and the arbitration result is written into the federated sharing layer; different airports share cases through the federated sharing layer; the scheduling agent performs cross-airport task merging and resource orchestration; the execution agent issues maintenance tasks to each airport based on the maintenance solutions provided by the expert agents, and performs pavement maintenance according to the maintenance tasks; the agents conduct bidding through a contract network protocol, and the contract network protocol adjusts the waiting time for each retransmission based on the timeout improvement results.

[0005] Furthermore, step 2 specifically includes: The average packet loss rate is calculated after smoothing the time-series data of packet loss rate for each airport and each decision cycle. Calculate the coefficient of variation of round-trip delay in the last sliding window for each airport in each decision cycle; The communication quality stability of each airport in each decision period is determined based on the mean packet loss rate and the coefficient of variation. The communication quality stability is negatively correlated with the mean packet loss rate and the coefficient of variation, respectively.

[0006] Furthermore, step 3 involves calculating the difference between the communication quality stability of each airport in each decision cycle and the extreme risk benchmark, calculating the sum of the extreme risk benchmark and a preset minimum positive number, and determining the risk compensation coefficient for each airport in each decision cycle based on the difference and the sum.

[0007] Furthermore, the risk compensation coefficient is specifically as follows: Calculate the ratio of the difference to the sum; use the maximum value between the ratio and 0 as the risk compensation coefficient for each airport in each decision cycle.

[0008] Furthermore, step 4 involves calculating an exponential function with the negative of the risk compensation coefficient for each airport in each decision cycle as the exponent, calculating the product of the exponential function and the predicted value, and determining the timeout improvement result for each airport in each decision cycle based on the product and the preset benchmark timeout time.

[0009] Furthermore, the improvement result of the timeout period is as follows: The product of the product and the baseline timeout time is used as the timeout time adjustment amount; the difference between the baseline timeout time and the timeout time adjustment amount is calculated, and the maximum value between the difference and the preset minimum timeout time is used as the timeout time improvement result for each airport in each decision cycle.

[0010] Furthermore, the process of obtaining the waiting time in step 5 is as follows: Based on the timeout improvement results and the preset baseline backoff factor, an adaptive dynamic backoff factor is determined for each airport in each decision cycle; the waiting time for each retransmission is adjusted based on the dynamic backoff factor.

[0011] Furthermore, the dynamic backoff factor specifically refers to: Obtain an exponential function with the negative of the normalized value of the improvement result of the timeout time of each airport in each decision cycle as the exponent. Use the product of the exponential function and the preset benchmark backoff factor as the backoff factor adjustment amount. Use half of the benchmark backoff factor and the sum of the backoff factor adjustment amount as the adaptive dynamic backoff factor of each airport in each decision cycle.

[0012] Furthermore, the specific waiting time for each retransmission is as follows: Calculate an exponential function with the dynamic backoff factor as the base and the difference between the current retransmission count and the integer 1 as the exponent. The product of this exponential function and the preset initial retransmission interval is used as the waiting time for the current retransmission.

[0013] Secondly, embodiments of this application also provide a multi-agent-based airport cluster collaborative maintenance retransmission analysis system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0014] Compared with the prior art, the technical solution adopted in this application has the following advantages: (1) This application determines the communication quality stability in each decision cycle by analyzing the long-term trend of packet loss rate and the severity of round-trip delay jitter in each decision cycle. The risk compensation coefficient is determined based on the difference between the communication quality stability and the preset benchmark value. At the same time, the communication quality stability of future decision cycles is predicted based on the communication quality stability of historical decision cycles. The timeout time of the decision cycle is improved by combining the two. When the communication is stable, the timeout time is reduced to improve the communication response efficiency. When the communication is unstable, the timeout time is extended. This avoids the problem of incorrect bidding judgment and asynchronous cooperation status caused by message delay or packet loss in traditional contract network protocols. (2) The improvement results of dynamic timeout time are applied to the improvement of the backoff factor of the binary exponential backoff algorithm, so that the retransmission waiting time can be adjusted according to the actual situation. In the case of weak network, the backoff factor is reduced so that retransmission occurs faster to avoid missing short communication opportunities. In the case of strong network, the normal backoff factor is used to continue congestion control, so that the contract network protocol can compete effectively and make correct decisions in the heterogeneous and highly dynamic network environment of airport clusters, thereby improving the consistency and timeliness of resource allocation between airports. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the steps of a multi-agent-based airport cluster collaborative maintenance retransmission analysis method provided in one embodiment of this application; Figure 2 This is a block diagram of a collaborative model based on five types of intelligent agents and a shared layer; Figure 3 A schematic diagram illustrating the process of obtaining communication quality stability; Figure 4 A schematic diagram illustrating the process of obtaining the timeout improvement results. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details the specific scheme of the airport group collaborative maintenance retransmission analysis method and system based on multi-agent systems provided in this application.

[0017] Please see Figure 1 The diagram illustrates a flowchart of a multi-agent-based airport cluster collaborative maintenance retransmission analysis method according to an embodiment of this application. The method includes the following steps: Step 1: Collect time-series data on packet loss rate and round-trip delay of each airport network within each decision cycle.

[0018] This application employs a distributed collaborative model consisting of five types of intelligent agents and a shared layer for collaborative maintenance decisions on airport pavement. For example... Figure 2 As shown, the collaborative model based on five types of intelligent agents and a shared layer specifically includes: (1) The station perception intelligent agent is deployed on the edge server of each airport, and connects with the pavement detection sensors of each airport to obtain airport data information, which can be used to request help and report; (2) The expert intelligent agent is located in the group cloud center and is divided into sub-intelligent agents such as disease diagnosis, materials, equipment, life, cost-effectiveness and safety risks according to their profession. Each sub-intelligent agent is connected to the industry large model and RAG retrieval enhancement generation engine, and then connected to the airport pavement management system, standard library and project case library. (3) The scheduling agent is located in the cloud center and has a built-in joint optimization engine of improved genetic algorithm and Monte Carlo tree search hybrid strategy and task queue manager, which is responsible for cross-airport maintenance tasks and resource orchestration; (4) The execution agent is located at the industrial gateway at each airport site, and connects with construction equipment, inspection drones and handheld terminals of operators. It is responsible for issuing tasks, monitoring status and transmitting process data. (5) The arbitration agent is located in the cloud center and has an embedded hard rule engine. When multiple agents' solutions conflict, they adjudicate according to the pre-given weights and safety red lines, and write the arbitration reasons into the knowledge base to form incremental learning samples.

[0019] (6) Federated sharing layer: The federated learning framework is used to maintain a decision counter library across airports. Data from all airports does not leave the domain. Only model gradients and delabeled statistical features are exchanged to achieve cross-airport knowledge transfer and data compliance.

[0020] The execution cycle of collaborative maintenance decisions for airport clusters is denoted as the decision cycle. The typical duration of a decision cycle is 15-60 minutes, which can be set by the implementer according to actual conditions. This application does not impose specific limitations; in this embodiment, it is set to 30 minutes. Within each decision cycle, if an airport has no data for five consecutive sampling points due to communication interruptions, this decision cycle is marked as a partially missing cycle. Linear interpolation is used to complete the missing data in subsequent processing. A lightweight network probe is installed at the local end of each airport to collect the loss ratio of bidirectional IP packets between the airport and the scheduling agent at a frequency of 10Hz, obtaining the raw packet loss rate data. The time difference between the airport sending probe messages and receiving responses is collected at the same collection frequency to obtain round-trip delay data. It should be noted that the implementer can set the data collection frequency according to actual conditions; this application does not impose specific limitations.

[0021] The time series of packet loss rate data and round-trip delay data for each airport in each decision cycle are denoted as packet loss rate sequence and round-trip delay sequence, respectively.

[0022] Step 2: Based on the overall packet loss rate of each airport in each decision cycle and the degree of disorder in the distribution of round-trip delays within each preset sliding window of each airport, determine the stability of communication quality of each airport in each decision cycle.

[0023] Due to the large scale of the multi-agent airport cluster and the mixed deployment of dedicated networks, 5G networks, and satellite communications among different airports, the network conditions of each airport vary significantly. This leads to frequent message delays and packet loss in the Contract Network Protocol (CNP) during the multi-agent collaborative maintenance decision-making process. Furthermore, it causes some agents to fail to receive task announcements on time or to send the same plan repeatedly, which seriously affects the timeliness and security of cross-site resource scheduling.

[0024] To solve the above problems, such as Figure 3 As shown, the process of determining the communication quality stability of each airport in each decision cycle, based on the overall packet loss rate of each airport in each decision cycle and the degree of disorder in the distribution of round-trip delay within each preset sliding window of each airport, is as follows: First, the packet loss rate sequence of each airport within each decision period is used as input to the Exponentially Weighted Moving Average (EWMA) algorithm to smooth the packet loss rate data. Specifically, the duration and step size of the sliding window in the EWMA algorithm are set to 1 minute and 10 seconds, respectively, with a smoothing factor of 0.3. The mean of the smoothed packet loss rate for each airport within each decision period is calculated as the average packet loss rate for that airport within that decision period. This average packet loss rate represents the stable trend of the packet loss rate for each airport node within each decision period after removing random burst interference, quantifying the unreliability of the communication link for each airport node. The Exponentially Weighted Moving Average algorithm is a well-known technique, and its specific process will not be elaborated further.

[0025] However, the average packet loss rate obtained through the above analysis cannot reflect the interference of time delay jitter on the Contract Network Protocol (CNP) in real time. In multi-agent collaboration, the CNP process typically involves the scheduling agent issuing a tender → expert agents bidding → the scheduling agent making a decision. Each step has a timeout period. When there is high latency jitter in the network environment, if the timeout thresholds for each agent are set too short, even slight network latency fluctuations can cause the scheduling agent to fail to receive any tenders within the specified time and thus determine that the bidding has failed. This kind of mechanism error will directly affect the task allocation efficiency and reliability of the multi-agent collaborative system.

[0026] Therefore, within each decision cycle, using the aforementioned sliding window configuration, the coefficient of variation of all round-trip delay data for each airport within each sliding window is calculated. Further, the normalized value of the coefficient of variation for the last sliding window of each airport within each decision cycle is obtained. Specifically, in this embodiment, the coefficient of variation of all sliding windows in each decision cycle is used as input to the maximum-minimum normalization method to obtain the normalized value of the coefficient of variation for each sliding window. If the maximum and minimum values ​​are the same, the normalized value of the coefficient of variation is directly set to 0. In other embodiments of this application, the implementer may also use other normalization methods to normalize the coefficient of variation.

[0027] Based on the above analysis, the communication quality stability A of each airport within each decision cycle is calculated. Taking any decision cycle as the current decision cycle and any airport as the current airport, preferably, in this embodiment, the expression for the communication quality stability A of the current airport in the current decision cycle is: In the formula, A represents the communication quality stability of the current airport in the current decision period; P represents the average packet loss rate of the current airport in the current decision period; and CV represents the normalized value of the round-trip delay data variation coefficient of the current airport in the last window of the current decision period. This is a preset, extremely small positive number to prevent the product result from being 0 due to the normalization value of the coefficient of variation being 1. Implementers can set this value according to their actual situation. The value of is not specifically limited in this application; in this embodiment... Pick .

[0028] The coefficient of variation (CV) measures the relative dispersion of round-trip delay data; a higher CV indicates greater volatility. In the multi-agent collaborative maintenance decision-making process of airport clusters, the mean packet loss rate after smoothing across all sliding windows directly reflects the unreliability of the communication link at each airport node, while the delay fluctuation index of this decision-making cycle expresses the severity of jitter in round-trip time delay. By combining these two factors—data loss and communication response—we analyze the timeout judgments affecting bidding and adjudication in the contract network protocol, and assess the stability of communication quality.

[0029] Step 3: Determine the risk compensation coefficient for each airport in each decision cycle based on the difference between the communication quality stability and the preset extreme risk benchmark.

[0030] Within an airport cluster, the network communication quality of each airport fluctuates dynamically in real time. Traditional contract network protocols typically use fixed timeout periods, which are difficult to adapt to environments with high latency jitter and high packet loss. This leads to frequent misjudgments of non-response due to timeouts during the bidding phase, resulting in repeated sending of the same proposal or abandonment of the bid.

[0031] To address the aforementioned issues, this application introduces the Holt two-parameter exponential smoothing model for predicting communication quality stability. Specifically, taking the current airport in the current decision-making period as an example, the communication quality stability of the airport in the current decision-making period and the previous nine historical decision-making periods are sorted in chronological order to obtain a historical communication stability sequence. It should be noted that for the first ten decision-making periods of system operation, if there are fewer than 10 historical data, the mean-filling method is used to complete the sequence. Specifically, the arithmetic mean of the existing observations is calculated, and this mean is used to fill in the beginning of the historical data until the total sequence length reaches 10. Furthermore, the historical communication stability sequence is used as input to the Holt model. The first element of the sequence is used as the initial smoothing value, and the difference between the second and first elements is used as the initial trend value. The smoothing parameter is set to 0.2, allowing the Holt model to have a longer memory of historical data and avoiding drastic changes in timeout due to short-term network fluctuations. The trend parameter is set to 0.1 to smooth long-term drift. Finally, the predicted value of the communication quality stability of the airport in the next decision cycle is output. The output predicted value represents the change in the communication link of the airport node in the near future. If the predicted value is greater than the actual value in the current decision cycle, it indicates that the communication link of the airport node is developing in a more stable direction; conversely, if the predicted value is less than the actual value in the current decision cycle, it indicates that the communication link of the airport node is developing in a worse direction.

[0032] However, for each airport, communication quality can be affected by extreme events such as typhoons and electromagnetic interference, resulting in current communication quality being significantly lower than historical communication quality. The above-mentioned simple trend predictions are difficult to quantify the extent of the impact of such rare extreme events. If this is not compensated for, cross-airport resource scheduling will fail.

[0033] Therefore, this application uses the 10th percentile of the historical communication stability sequence of the current airport in the current decision-making cycle as the extreme risk benchmark to calculate the risk compensation coefficient. Preferably, in this embodiment, the expression for the risk compensation coefficient is: In the formula, The risk compensation coefficient for the current airport during the current decision-making cycle. The stability of communication quality at the airport during the current decision-making cycle; This serves as the benchmark for extreme risks at the current airport during the current decision-making cycle. It is a preset, extremely small positive number to prevent the denominator from being 0; This is a function to find the maximum value. For The implementer can set the value according to the actual situation. This application does not impose specific restrictions. In this embodiment, The value is .

[0034] Calculated risk compensation coefficient This represents the degree to which the current communication quality deviates from the historical extreme lower bound. The larger the value, the more unstable the network state is in the current decision-making cycle. Conversely, a value of 0 indicates that the network state is relatively stable in the current decision-making cycle.

[0035] Step 4: Based on the communication quality stability of each airport in the historical decision-making cycle, obtain the predicted value of the communication quality stability for the next decision-making cycle. Combine the risk compensation coefficient and the preset benchmark timeout time to determine the timeout improvement result for each airport in each decision-making cycle.

[0036] like Figure 4 As shown, the improvement result B of the timeout time for each airport in each decision cycle is calculated. The specific calculation formula can be: In the formula, B represents the improved timeout result for the current decision-making cycle of the current airport; In this embodiment, the inherent minimum timeout time for network layer photoelectric transmission and node hardware computation is preset. Take 0.1s; The preset baseline timeout time typically ranges from 2 to 5 seconds. In this embodiment... The value is 2s; This is a predicted value for the stability of communication quality in the next decision-making cycle of the current airport; This represents the risk compensation coefficient for the current airport during the current decision-making cycle. This is the function for finding the maximum value. Where, This is recorded as the timeout adjustment amount. It should be noted that for... and The value can be set by the implementer according to the actual situation; this application does not impose specific restrictions.

[0037] The output predicted values ​​reflect the local trends of the input sequence and embody the expected state in the short term. The risk compensation coefficient represents the degree to which the current communication situation deviates from historical extreme risks; a larger value indicates a more severe deviation. When the predicted value of the communication quality stability in the next decision cycle is larger or the risk compensation coefficient is smaller, it indicates that the communication situation is relatively stable, and the timeout should be shortened to improve the efficiency of communication response. Conversely, when the predicted value of the communication quality stability in the next decision cycle is lower or the risk compensation coefficient is larger, it indicates that the communication situation is relatively poor, and the timeout should be maintained at the baseline timeout to prevent misjudgments in bidding caused by network jitter.

[0038] This application first introduces a product operation combining predicted values, resulting in a larger timeout period for smaller predicted values. Second, it introduces a risk compensation coefficient and maps it using a negative exponential function to ensure that the contract network protocol maintains consistency and timeliness in collaborative decision-making within heterogeneous and highly dynamic airport cluster environments. Finally, it uses a maximum value function to force a change in time when the estimated adjustment time is less than the actual network interface card (NIC) computational consumption. The fallback output is 0.1 seconds, leaving a basic operational gap for the system to process handshake messages.

[0039] Step 5, the airport group pavement collaborative maintenance decision-making process includes: the airport group pavement defects are detected by the station sensing agent; once a defect is detected, the station sensing agent initiates a tender to the expert agent; the expert agent bids and provides maintenance solutions; if conflicting maintenance solutions exist, they are handled by the arbitration agent, and the arbitration result is written into the federated sharing layer; different airports share cases through the federated sharing layer; the scheduling agent performs cross-airport task merging and resource orchestration; the execution agent issues maintenance tasks to each airport based on the maintenance solutions provided by the expert agents, and performs pavement maintenance according to the maintenance tasks; the agents conduct bidding through a contract network protocol, and the contract network protocol adjusts the waiting time for each retransmission based on the timeout improvement results.

[0040] In the multi-agent collaborative maintenance decision-making process of airport clusters, the above steps have been used to adaptively adjust the contract network protocol timeout threshold using dynamic timeout improvement results. However, under conditions of high network packet loss rate or severe latency jitter, even if the baseline timeout is maintained, bidding information may still be lost during network transmission due to congestion or signal attenuation, requiring retransmission. The commonly used binary exponential backoff (BEB) algorithm doubles the backoff time for each retransmission to avoid congestion. However, the backoff factor of the binary exponential backoff algorithm is usually a fixed constant, specifically set to 2 in this embodiment. But in the heterogeneous network environment of airport clusters, the communication status between airports is constantly changing, so a fixed backoff factor cannot adapt to the dynamic communication quality changes of different airports in the heterogeneous network of airport clusters.

[0041] If the communication quality stability A of an airport is small, using a fixed backoff factor will make the retransmission time too long, potentially causing missed communication opportunities within a short period of time. Conversely, if A is large, using a fixed backoff factor will make the retransmission time shorter, increasing unnecessary collisions and contention, thus affecting efficiency.

[0042] To address the aforementioned shortcomings, this application dynamically improves the backoff factor in the binary exponential backoff algorithm based on the timeout improvement result B obtained from the above calculation for each airport. When network quality deteriorates (i.e., B is larger, representing a longer timeout), the backoff factor is appropriately reduced to avoid missing short-term recovery communication opportunities due to excessive waiting time; when network quality is good (i.e., B is smaller), the backoff factor is maintained or slightly increased to maintain the congestion control effect.

[0043] Specifically, firstly, the timeout improvement result B of the current airport's current decision cycle is normalized. In this embodiment, the current airport's current decision cycle and the timeout improvement results of all decision cycles within the previous 24 hours are used as inputs to the maximum-minimum normalization method for normalization, and the normalized value of the timeout improvement result is obtained. Specifically, if the maximum and minimum values ​​of the timeout improvement results are the same during the normalization process, the normal normalization operation is skipped, and the normalized value of the timeout improvement result for the current airport in the current decision cycle is directly set to a fixed constant value. This fixed constant ranges from 0 to 1, and the implementer can set it according to the actual situation. This application does not impose specific restrictions; in this embodiment, the fixed constant value is set to 0.5. Based on the above... The dynamic backoff factor d is calculated using the following expression: In the formula, d is the adaptive dynamic backoff factor for the current airport in the current decision-making cycle; This is the baseline backoff factor for the preset binary exponential backoff algorithm, typically ranging from 1 to 5. Implementers can set it according to their specific circumstances; this application does not impose any specific limitations. In this embodiment... The specific value is 3; This represents the normalized value of the improved timeout result for the current decision-making cycle at the current airport; e is the natural constant. This is the adjustment amount for the retreat factor.

[0044] When network quality deteriorates, that is When the value is close to 1, the retreat factor d decreases to close to 1. This reduces the rate of increase in retransmission waiting time and improves the retransmission success rate in weak network environments; when network quality is good, i.e. When it approaches 0, the retreat factor d approaches It maintains the congestion control characteristics of traditional algorithms.

[0045] For each decision cycle, the waiting time for the improved adaptive retransmission backoff algorithm during each retransmission is: In the formula, The waiting time for the k-th retransmission of the network during the current decision-making cycle at the current airport; The preset initial retransmission interval is specifically set as the base timeout time in this embodiment. Half of; The preset maximum number of retransmissions is typically 3 to 7 times. In this embodiment... 5; d is the adaptive dynamic backoff factor for the current airport in the current decision cycle; k represents the order value of the k-th retransmission; It is an exponential function, representing the waiting time for each retransmission adjusted by the backoff factor. The exponential growth, i.e., the first retransmission wait Waiting for the second retransmission Waiting for the third retransmission And so on. Meanwhile, to avoid infinite retransmissions, when If the remaining time of the current decision cycle is exceeded, the agent will immediately report the bidding failure status to the arbitration agent.

[0046] Through the above improvements, the Contract Network protocol can dynamically adjust the retransmission backoff strategy according to the real-time communication quality in heterogeneous network and highly dynamic airport cluster environments. This avoids excessive backoff caused by fixed backoff factors in weak network environments, which leads to bidding timeouts, and also prevents ineffective congestion in good network environments, thereby significantly improving the consistency and timeliness of cross-airport resource scheduling.

[0047] Building upon the architecture of Step 1, agents collaborate to make decisions via the Contract Network Protocol (CNP). First, the airport perception agent detects pavement defects within the airport cluster and initiates a bidding process with expert agents. Expert agents then submit bids and maintenance plans; conflicting plans are resolved through an arbitration agent. Throughout this process, the communication quality of each airport directly impacts the timeout determination for bidding and adjudication. Therefore, this application dynamically adjusts the timeout period of the Contract Network Protocol and improves the retransmission backoff strategy. When the network quality at a particular airport is poor, the backoff factor is automatically reduced, allowing retransmissions to occur faster and avoiding missed communication opportunities. When the network is good, normal backoff is maintained to prevent ineffective congestion. Thus, even in a heterogeneous and highly dynamic airport cluster network, defect tasks can be completed in a timely manner, ensuring that cross-airport resource scheduling between different airports does not fail due to communication problems, laying a solid communication foundation for collaborative pavement maintenance across the airport cluster.

[0048] Subsequently, the scheduling agent aggregates the bidding proposals from all airports and uses optimization algorithms to merge cross-airport tasks and unify resource allocation. For example, grouting tasks at adjacent airports are grouped together to reduce equipment transfer time and costs, and urgent tasks are prioritized for recent nighttime flight suspension windows. Finally, the execution agent distributes maintenance tasks to each airport site based on the maintenance plan provided by the expert agent, coordinates construction equipment and personnel, and the on-site maintenance personnel carry out pavement maintenance for the airport complex according to the maintenance tasks and transmit progress data back.

[0049] Throughout the process, maintenance needs, resource status, and decision-making outcomes for all airports are transferred through a federated sharing layer. This means that airports only exchange de-identified statistical characteristics and decision-making experience, without disclosing raw data. This ensures data compliance while allowing airports to learn from successful cases of other airports. Arbitration results can also be written into the federated sharing layer for case sharing among different airports. Through this mechanism, this application upgrades the original independent decision-making of a single airport to collaborative decision-making across an airport group, enabling dynamic allocation of scarce maintenance resources within the group and avoiding resource idleness or competition. The improved network timeout and retransmission strategy in this application is a key foundation for ensuring that this complex cross-airport communication is not interrupted by network fluctuations, ensuring that defect bidding, solution competition, and resource adjudication are always timely and consistent.

[0050] Based on the same inventive concept as the above method, this application embodiment also provides a multi-agent-based airport group collaborative maintenance retransmission analysis system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described multi-agent-based airport group collaborative maintenance retransmission analysis methods.

[0051] In summary, this application provides a multi-agent-based airport cluster collaborative maintenance retransmission analysis method. By analyzing the long-term trend of packet loss rate and the severity of round-trip delay jitter within each decision cycle, the method determines the communication quality stability within each decision cycle. A risk compensation coefficient is determined based on the difference between this communication quality stability and a preset benchmark value. Simultaneously, the method predicts the communication quality stability of future decision cycles based on the communication quality stability of historical decision cycles. Combining these two methods improves the timeout time of the decision cycle. When communication is stable, the timeout time is reduced to improve communication response efficiency; when communication is unstable, the timeout time is extended to avoid... This approach avoids the problems of incorrect bidding judgments and asynchronous cooperation states that traditional contract network protocols suffer from due to message delays or packet loss. The improved dynamic timeout is applied to the improvement of the backoff factor in the binary exponential backoff algorithm, allowing the retransmission waiting time to be adjusted according to the actual situation. In weak network conditions, the backoff factor is reduced to make retransmissions occur faster to avoid missing brief communication opportunities. In strong network conditions, the normal backoff factor is used to continue congestion control. This enables the contract network protocol to compete effectively and make correct decisions in heterogeneous and highly dynamic network environments of airport clusters, improving the consistency and timeliness of resource allocation across airports.

Claims

1. A multi-agent-based airport cluster collaborative maintenance retransmission analysis method, characterized in that, The method includes the following steps: Step 1: Collect time-series data on packet loss rate and round-trip delay of each airport network within each decision cycle; Step 2: Analyze the overall packet loss rate of each airport in each decision cycle, as well as the degree of disorder in the distribution of round-trip delays of each airport within each preset sliding window, and determine the stability of communication quality of each airport in each decision cycle. Step 3: Determine the risk compensation coefficient for each airport in each decision cycle based on the difference between the communication quality stability and the preset extreme risk benchmark. Step 4: Based on the communication quality stability of each airport in the historical decision-making cycle, obtain the predicted value of the communication quality stability for the next decision-making cycle. Combine the risk compensation coefficient and the preset benchmark timeout time to determine the timeout time improvement result for each airport in each decision-making cycle. Step 5, the airport group pavement collaborative maintenance decision-making process includes: the airport group pavement defects are detected by the station sensing agent; upon detection, the station sensing agent initiates a tender with the expert agent; the expert agent bids and provides maintenance solutions; if conflicting maintenance solutions exist, they are handled by the arbitration agent, and the arbitration result is written into the federated sharing layer; different airports share cases through the federated sharing layer; and the scheduling agent performs cross-airport task merging and resource orchestration. The execution agent issues maintenance tasks to each airport based on the maintenance solutions provided by the expert agents, and performs pavement maintenance according to the maintenance tasks. The station sensing agent and the expert agent conduct bidding through a contract network protocol, which adjusts the waiting time for each retransmission based on the timeout improvement results.

2. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 1, characterized in that, Step 2 specifically involves: The average packet loss rate is calculated after smoothing the time-series data of packet loss rate for each airport and each decision cycle. Calculate the coefficient of variation of round-trip delay in the last sliding window for each airport in each decision cycle; The communication quality stability of each airport in each decision period is determined based on the mean packet loss rate and the coefficient of variation. The communication quality stability is negatively correlated with the mean packet loss rate and the coefficient of variation, respectively.

3. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 1, characterized in that, Step 3 involves calculating the difference between the communication quality stability of each airport in each decision cycle and the extreme risk benchmark, calculating the sum of the extreme risk benchmark and a preset minimum positive number, and determining the risk compensation coefficient for each airport in each decision cycle based on the difference and the sum.

4. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 3, characterized in that, The risk compensation coefficient is specifically as follows: Calculate the ratio of the difference to the sum; use the maximum value between the ratio and 0 as the risk compensation coefficient for each airport in each decision cycle.

5. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 1, characterized in that, Step 4 involves calculating an exponential function with the negative of the risk compensation coefficient for each airport in each decision cycle as the exponent, multiplying the exponential function by the predicted value, and determining the timeout improvement result for each airport in each decision cycle based on the product and the preset baseline timeout.

6. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 5, characterized in that, The specific results of the timeout improvement are as follows: The product of the product and the baseline timeout time is used as the timeout time adjustment amount; the difference between the baseline timeout time and the timeout time adjustment amount is calculated, and the maximum value between the difference and the preset minimum timeout time is used as the timeout time improvement result for each airport in each decision cycle.

7. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 1, characterized in that, The process of obtaining the waiting time in step 5 is as follows: Based on the timeout improvement results and the preset baseline backoff factor, an adaptive dynamic backoff factor is determined for each airport in each decision cycle. The waiting time for each retransmission is adjusted based on a dynamic backoff factor.

8. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 7, characterized in that, The dynamic backoff factor is specifically: Obtain an exponential function with the negative of the normalized value of the improvement result of the timeout time of each airport in each decision cycle as the exponent. Use the product of the exponential function and the preset benchmark backoff factor as the backoff factor adjustment amount. Use half of the benchmark backoff factor and the sum of the backoff factor adjustment amount as the adaptive dynamic backoff factor of each airport in each decision cycle.

9. The airport cluster collaborative maintenance retransmission analysis method based on multi-agent systems as described in claim 7, characterized in that, The specific waiting time for each retransmission is as follows: Calculate an exponential function with the dynamic backoff factor as the base and the difference between the current retransmission count and the integer 1 as the exponent. The product of this exponential function and the preset initial retransmission interval is used as the waiting time for the current retransmission.

10. A multi-agent-based airport cluster collaborative maintenance retransmission analysis system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

Images