A method for detecting defects in unmanned cluster software

Through the combination of multi-agent modeling and two-state warehouse model, traditional methods are solved, and the problem of complexity and emergence of unmanned cluster software is difficult to deal with, and effective defect analysis and testing verification of unmanned cluster software systems is realized, which improves the System reliability and quality.

CN118069534BActive Publication Date: 2024-11-15INFORMATION CENT OF CHINA NORTH IND GRP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410377728.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-29
Publication Date
2024-11-15
Estimated Expiration
2044-03-29

AI Technical Summary

Technical Problem

Traditional defect analysis and test verification methods cannot effectively deal with the complexity and emergent characteristics of unmanned cluster software systems, resulting in poor defect analysis and test verification results, affecting the reliability and application effect of the system.

Method used

Using multi-agent modeling method and two-state warehouse model, the emergence analysis and defect propagation simulation of unmanned cluster software systems are carried out, and defect propagation simulation analysis is analyzed through multi-agent modeling method and fault propagation SIR model to discover and expose the software system. flaws and weaknesses.

Benefits of technology

It effectively improves the quality and reliability of unmanned clustered software systems, can better discover and expose the defects and weak links of the software system, promotes the improvement and improvement of the software system, and is suitable for the development of a new generation of clustered intelligent system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118069534B_ABST
    Figure CN118069534B_ABST
Patent Text Reader

Abstract

The present invention discloses an unmanned cluster software defect detection method, which relates to the technical field of software detection. Firstly, a system emergence analysis method is proposed, and then a defect propagation simulation analysis method is proposed based on multi-agent modeling and a fault propagation SIR model, and an unmanned cluster software defect propagation and failure mechanism analysis is carried out, which solves the problem that traditional defect analysis and test verification methods cannot effectively cope with the characteristics of unmanned cluster software itself, can effectively discover and expose software system defects and weak links, promote continuous improvement and perfection of software systems, improve the quality and reliability level of software systems, and contribute to the development of a new generation of cluster intelligent systems.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of software detection, and more specifically to an unmanned cluster software defect detection method. Background Art

[0002] With the continuous development of intelligent unmanned equipment, it can replace personnel to complete combat tasks such as reconnaissance perception, patrol monitoring, and fire control. Although the performance of single unmanned equipment has been greatly improved, its ability to complete tasks is limited, and unmanned cluster collaborative combat has come into being. Unmanned clusters are the product of the combination of group intelligence and unmanned systems. They simulate the combat system formed by biological clusters in nature with functional distribution and behavioral emergence characteristics. They are divided into drone clusters, ground unmanned vehicle clusters, surface unmanned boat clusters, underwater unmanned submarine clusters, and cross-domain joint operations of the above unmanned equipment.

[0003] As the core of unmanned swarm control and the carrier of algorithms, the unmanned swarm software system has the characteristics of distribution, collaboration, networking, and emergence, and has gradually become a complex system with hierarchical, open, and nonlinear functions, behaviors, and structures. As the actual battlefield situation environment is full of uncertainty, the unmanned swarm software system needs to have highly secure and reliable decision-making capabilities. However, the current unmanned swarm collaborative control and decision-making algorithms are mostly data-driven methods such as machine learning, which rely on the quality of the basic data set, have poor robustness and poor interpretability, and cannot effectively handle events that have never been encountered. The uncertainty in the scene will cause the recognition accuracy of the machine learning model to drop significantly, which directly affects the practical application of various types of unmanned swarm equipment.

[0004] However, there are few research results on defect analysis and test verification of unmanned cluster software systems. The defect analysis and reliability assessment methods of traditional software systems mainly include state analysis-based, path-based, test-based, system simulation methods, etc. Some scholars at home and abroad have proposed software reliability assessment models based on component structure and complex network, but they lack consideration of software network structure analysis and modeling. As a new type of intelligent distributed complex system, the behavior and results of unmanned cluster software systems are uncertain, and the adaptive effect is constrained by uncertainty. Traditional defect analysis and test verification methods cannot effectively deal with the characteristics of unmanned cluster software systems.

[0005] Therefore, how to improve the effectiveness of unmanned cluster software system defect analysis and testing is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0006] In view of this, the present invention provides an unmanned cluster software defect detection method, which implements effective analysis and verification of the unmanned cluster software system based on a multi-agent modeling method and a two-state compartment model, discovers and exposes the defects and weak links of the unmanned cluster software system, and improves the quality and reliability level of the unmanned cluster system.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] An unmanned cluster software defect detection method comprises the following steps:

[0009] Step 1: Perform a multi-agent system emergence analysis on the unmanned cluster software to obtain the system evolution law and group behavior mode;

[0010] Step 2: According to the system evolution law and group behavior mode, the multi-agent modeling method and fault propagation SIR model are used to perform defect propagation simulation analysis on the unmanned cluster software to obtain the defect detection results.

[0011] Preferably, step 1 comprises the following steps:

[0012] Step 11: Micro-single-unit analysis: Using multi-agent method and evolutionary game theory, analyze the transfer trajectory of the monomer adaptive behavior state to obtain the monomer adaptive behavior state; conduct quantitative analysis on deterministic factors and qualitative analysis on non-deterministic factors;

[0013] Step 12: Meso-structure analysis: Based on the state of the monomer adaptive behavior, the system dynamics method is used to analyze the internal mechanism and feedback loop of the unmanned cluster software evolution process, analyze the impact of different factors on the evolution of the entire unmanned cluster software system, and obtain the mechanism of the emergence characteristics as the law of system evolution;

[0014] Step 13: Macro-group analysis: Based on the system evolution law, the complex network self-organization evolution analysis method is used to build an unmanned cluster association network evolution model to obtain the emergence results of group behavior modes such as autonomous cooperation and confrontation game.

[0015] Preferably, implementing step 2 in an Agent modeling environment includes the following steps:

[0016] Step 21: Based on the system evolution law and group behavior mode, multi-agent modeling is performed; a single agent model is constructed for each single agent in the unmanned swarm software using a cognitive agent structure; based on the connection relationship between single agents, a connection agent model is constructed using a reactive agent structure;

[0017] Step 22: Construct a cluster environment operation model based on the system evolution law and group behavior mode; the cluster environment operation model includes the cluster system model, software system task profile, task target distribution information and monomer collaborative interaction information;

[0018] Step 23: Construct a task goal generation model according to the system evolution law and group behavior mode; in the task goal generation model, any single agent model publishes the task goal, or the same single agent model publishes different task goals, or multiple single agent models publish the same task goal;

[0019] Step 24: Create a single task call model based on the system evolution law and group behavior mode; determine the target and establish the participation state based on the single task target requirements and the capabilities of the single Agent model;

[0020] Step 25: According to the system evolution law and group behavior mode, and in accordance with the cluster formation rules, build cluster dynamic collaborative relationships;

[0021] Step 26: Construct an unmanned cluster software model based on the single agent model, the connection agent model, the cluster environment operation model, the task target generation model and the cluster dynamic coordination relationship;

[0022] Step 27: Based on the fault propagation SIR model, an unmanned cluster software defect propagation simulation model is constructed for the unmanned cluster software model, and the Monte Carlo method is used to simulate the unmanned cluster software defect propagation simulation model to obtain defect detection results of the unmanned cluster software. The defect detection results include the error node diffusion ratio and the system average repair time.

[0023] Preferably, the unmanned cluster software defect propagation simulation model adopts a two-state compartment model, wherein each monomer Agent model in the unmanned cluster software model serves as a node, and the balance equation of each compartment is expressed as:

[0024]

[0025] Among them, S(t) represents the number of healthy nodes at time t, I(t) represents the number of implicit error nodes at time t; α represents the infection intensity;

[0026] s(t)=c(t)(N-1)(1-I(t) / N)

[0027] Where c(t) represents the connection probability of the implicit error node i at time t; c(t)(N-1) represents the nodes connected to the implicit error node i; 1-I(t) / N represents the proportion of nodes that are not infected by the error among the nodes connected to the implicit error node i; N represents the total number of nodes;

[0028]

[0029] Among them, μ(t) represents the interaction frequency between nodes; β represents the transmission probability between nodes; the probability of a healthy node being infected is βμ(t)S(t)I(t) / N;

[0030] The infection rate a is expressed as:

[0031]

[0032] Therefore

[0033] It can be inferred that the unmanned cluster software defect propagation simulation model is expressed as:

[0034]

[0035] When N is large, N-1≈N, the unmanned cluster software defect propagation simulation model can be simplified as:

[0036]

[0037] Preferably, when simulating the unmanned cluster software defect propagation simulation model, the general solution is obtained as follows:

[0038]

[0039] Where t represents time t, C is a constant, and assuming that there is an implicit error node at the initial time t = 0, that is, I(0) = 1, then The final solution is:

[0040]

[0041] The error node diffusion ratio is calculated as:

[0042]

[0043] Preferably, the Monte Carlo method is used to perform simulation based on the unmanned cluster software defect propagation simulation model, and the number and proportion of faulty nodes, resistant nodes, and nodes that may fail in the simulation process, as well as the simulation step time and the probability of defective node recovery are used to calculate the single repair time for all faulty nodes in a simulation process. After multiple cycles of simulation, the average value is calculated based on multiple groups of single repair times to obtain the system average repair time.

[0044] Preferably, the monomer agent model constructed using the cognitive agent structure is represented by a multi-group as follows:

[0045] Mono_Agent::=<Mono_N,Mono_ST,Mono_P,Mono_Per,Mono_Trans,Mono_KB,Mono_PS,Mono_GS,Mono_AS>

[0046] Among them, Mono_N represents the serial number of the single agent; Mono_ST represents the state set; Mono_P represents the perception set; Mono_Per represents the EQ→Mono_P mapping function, and EQ is the environment state set.

[0047] Preferably, the connection agent model constructed using the reactive agent structure is represented by a quaternary structure:

[0048] Con_Agent::=<Con_N,Con_ST,Con_KB,Con_AS>

[0049] Where Con_N represents the serial number of the connection agent; Con_ST represents the state set of the connection agent; Con_KB represents the knowledge base of the connection agent; Con_AS represents the action set of the connection agent.

[0050] It can be seen from the above technical solution that, compared with the prior art, the present invention discloses a method for detecting defects in unmanned cluster software, which solves the problem that traditional defect analysis and test verification methods cannot effectively cope with the characteristics of unmanned cluster software itself. Aiming at the emergence and evolution characteristics of unmanned cluster software, a method for analyzing the emergence of software systems is proposed, and then defect propagation simulation analysis is performed based on multi-agent modeling and fault propagation SIR model, and unmanned cluster software defect propagation and failure mechanism analysis is carried out, which can effectively discover and expose software system defects and weak links, promote continuous improvement and perfection of software systems, improve the quality and reliability level of software systems, and contribute to the development of a new generation of clustered intelligent systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0052] Figure 1 A schematic diagram of the structure of the emergent analysis method for the unmanned cluster software system provided by the present invention;

[0053] Figure 2 A schematic diagram of the structure of the unmanned cluster software system defect simulation analysis process provided by the present invention;

[0054] Figure 3 A schematic diagram of the SI model compartment diagram structure provided by the present invention;

[0055] Figure 4 A schematic diagram of the index change structure of the model simulation process provided by the present invention;

[0056] Figure 5 A schematic diagram of the system state structure after the simulation provided by the present invention is completed;

[0057] Figure 6 A schematic diagram of the statistical structure of the cycle simulation results provided by the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example 1

[0060] The embodiment of the present invention discloses a method for detecting defects in unmanned cluster software. As a multi-agent system, the software monomer of unmanned cluster software is an intelligent entity with certain autonomous capabilities, corresponding to biological individuals and groups in real society. The agent monomer model is a container for knowledge in different fields to play a role. It is customized according to the application problem, including the structure of the agent, whether it has learning ability, and the interaction mechanism between them. The emergence analysis of unmanned cluster software adopts bottom-up analysis, such as Figure 1 As shown in the figure, the analysis goal is to analyze the microscopic and local behavior mechanisms of Agents and the behavior of Agent monomers to obtain the expected macroscopic and global system characteristics and reveal the internal logic of the software system from microscopic to macroscopic, from monomer behavior to swarm intelligence. The emergence analysis of multi-agent systems requires modeling and analysis of complex systems from three levels: microscopic, mesoscopic and macroscopic. The emergence analysis framework of unmanned cluster software systems adopts a bottom-up analysis method to analyze the evolution laws of software systems from three levels. The details are as follows:

[0061] Step 1: Micro-monomer analysis; the monomers in the multi-agent system can adjust their behavior states according to their environment, making the entire system adaptable and self-organizing; therefore, based on the multi-agent method and evolutionary game theory, we can start from the micro perspective of the software monomer and analyze the transfer trajectory of the monomer's adaptive behavior state; in the process of executing the task, the agent formulates the next action plan and executes the response action based on individual experience, environmental information and behavioral feedback, and calls the corresponding resource services to respond; based on the micro elements in the multi-agent system, quantitative analysis should be conducted for deterministic elements. For uncertain elements, such as limited rationality and learning ability, qualitative analysis should be conducted to judge the adaptive behavior and system robustness of the entire software system;

[0062] Step 2: Meso-structure analysis; The monomers of unmanned swarm software systems will emerge complex structural characteristics during the evolution process, including small-world characteristics and power-law characteristics; The interaction between software monomer agents is the main reason for the emergence of software group intelligence; In order to understand the generation mechanism of these emergent characteristics, based on theories such as system dynamics, the internal mechanism and feedback loop of the system evolution process are analyzed according to the adaptive behavior of the software system; Based on data analysis, the learning method of model parameters is studied, and the influence of different factors on the evolution of the entire system is analyzed, and then the evolution process of the entire system is predicted to obtain the law of system evolution; Through reasonable evolution mechanism design, on the basis of ensuring individual interests, behaviors such as autonomous cooperation can be generated between software monomer agents;

[0063] Step 3: Macro-group analysis; at the macro level, the process of the emergence, development and cyclic feedback of the system group intelligence emergence phenomenon is similar to the self-organization evolution process of complex networks; according to the system evolution law, a correlation network evolution model is formed between the micro-monomer and the macro-whole, and the organizational structure, interaction mechanism and other relationships of the group system are reflected in the self-organization evolution iterative mechanism of the network, which is manifested in the emergence of group behavior modes such as autonomous cooperation and confrontation game at the overall level of the unmanned cluster software system;

[0064] The actions and interactive behaviors of individual software in the micro-level of the unmanned cluster software system generate local organizations through evolution at the meso-level, and finally generate pattern emergence in the swarm network at the macro-level, which is specifically manifested as the self-organizing evolution mechanism of the network, and emerges as group behaviors such as game confrontation or autonomous collaboration on the whole;

[0065] Step 4: Multi-agent modeling is performed according to the system evolution law and group behavior mode. The defect propagation simulation analysis based on multi-agent modeling is to establish a simulation model in the agent modeling environment platform from the aspects of operating environment, task generation, task measurement, etc., and analyze the defect propagation process of the unmanned cluster software system, such as Figure 2 As shown;

[0066] Step 41: Multi-agent modeling; decompose the unmanned cluster software system into a system composed of multiple monomers, identify each monomer agent, and assemble all the identified monomer agents into an agent; the characteristics of the software agent include responsiveness, goal-drivenness, etc., which can effectively describe the dynamic evolution characteristics of the structure and behavior of the unmanned cluster software system, and study how the software entity responds to the external dynamic environment from the internal mechanism; by studying the static structure and interaction relationship between different types of agent entities, analyze the changes in the operation mechanism of the intelligent software system; the monomer agent sets the task that a single software entity needs to complete, inputs the required information from the operating environment, integrates the information for the internal state, formulates a plan based on the information in the knowledge base, and finally generates a corresponding series of actions according to the goal to exert an impact on the cluster environment; the connecting agent is an agent that connects the software monomer agents, and evolves and operates with a perception-reaction mechanism, does not require logical reasoning, and directly calls their respective knowledge bases to feed back the processed information to the cluster operating environment;

[0067] The single agent is modeled using the cognitive agent structure, and its formal description is represented by a multi-tuple:

[0068] Mono_Agent::=<Mono_N,Mono_ST,Mono_P,Mono_Per,Mono_Trans,Mono_KB,Mono_PS,Mono_GS,Mono_AS>

[0069] Where Mono_N represents the serial number of the single agent; Mono_ST represents the state set; Mono_P represents the perception set; Mono_Per represents the EQ→Mono_P mapping function, and EQ is the environment state set;

[0070] The connection agent is modeled using a reactive agent structure, which is formally described as a four-tuple:

[0071] Con_Agent::=<Con_N,Con_ST,Con_KB,Con_AS>

[0072] Where Con_N represents the serial number of the connection agent; Con_ST represents the state set of the connection agent; Con_KB represents the knowledge base of the connection agent; Con_AS represents the action set of the connection agent;

[0073] Step 42: Establish a cluster environment operation model according to the system evolution law and group behavior mode; use the Agent modeling environment platform to build an unmanned cluster software system environment operation model, and the specific information includes cluster system model, software system task profile, task target distribution information, monomer collaborative interaction information, etc.;

[0074] Step 43: Construct a task target generation model based on the system evolution law and group behavior mode; the monomer collaborative interaction in the unmanned cluster software system is complex. When the software task call cannot be centralized to a monomer agent with a coordination and control role, the task target release is performed by any monomer agent participating in the system, and different targets can be released by the same monomer agent or by multiple monomer agents together;

[0075] Step 44: Create a single task call model based on the system evolution law and group behavior mode; according to the needs of the single task goal, each single agent decides whether to participate in the goal establishment based on its own functions and capabilities; when reaching a single agent, if this single agent can complete the goal alone, it will move on to the next goal; if it cannot complete the goal alone, it will start to analyze other related single agents to determine whether to solve this goal together;

[0076] Step 45: Generate dynamic collaborative relationships of clusters; analyze the cross-linking relationship information of different monomer agents in the monomer agent set of the unmanned cluster software system according to the system evolution law and group behavior mode, and build collaborative and interactive relationships of different monomer agents in the agent modeling environment platform according to cluster formation rules (including proximity rules, alignment rules, separation rules, etc.) to jointly complete the cluster task objectives;

[0077] Step 46: Construct a defect propagation simulation model. In the Agent modeling environment platform, according to the fault propagation SIR model, construct a defect propagation simulation model for the unmanned cluster software system. By setting the software defect propagation law and intensity, the Monte Carlo method is used to simulate the defect propagation process inside the software system. The simulation process includes defect propagation and defect repair, and the parameters include defect propagation probability, single node defect inspection probability, defect node recovery probability, etc.

[0078] According to the characteristics of unmanned cluster software system, the fault propagation SIR model is used to establish the unmanned cluster software defect propagation simulation model, accurately characterize the software fault propagation behavior, and find the weak links of software fault propagation; the two-state compartment model is used to analyze the propagation of software fault behavior. Figure 3 As shown in the figure; Due to the large number of nodes and complex interactive behaviors of the unmanned cluster software system, a deterministic compartment propagation method is adopted, and the balance equation of each compartment is:

[0079]

[0080] Each single agent is a node; at a certain time t, S(t) represents the number of healthy nodes, and I(t) represents the number of nodes with hidden errors; when the connection rate c(t) is known, the node i with hidden errors is connected to at most c(t)(N-1) nodes in the unmanned cluster connection graph formed by connecting all single agents in the unmanned cluster, and the proportion of these nodes that are not infected by errors is 1-I(t) / N. Then, c(t)(N-1)(1-I(t) / N) is the number of healthy nodes s(t) connected to node i at time t, the interaction frequency between nodes is μ(t), and the propagation probability between nodes is β. By definition, β uses different probability calculation methods for nodes of different granularities; therefore, the ratio of healthy nodes to be infected is: βμ(t)S(t)I(t) / N, and the infection intensity α is expressed as:

[0081]

[0082] The infection rate a is expressed as:

[0083]

[0084] Therefore

[0085]

[0086] Among them, s(t) = c(t)(N-1)(1-I(t) / N), and the unmanned cluster software defect propagation simulation model equation can be obtained:

[0087]

[0088] When N is large, N-1≈N, the above formula can be simplified to:

[0089]

[0090] The general solution is obtained as:

[0091]

[0092] Where C is a constant. Assuming that at the initial time t = 0, there is only one node with an implicit error in the unmanned cluster connection graph, that is, I(0) = 1, there is

[0093]

[0094] So the final solution is:

[0095]

[0096] The diffusion ratio of the error node can be obtained, which shows the propagation law and intensity of software defects and is expressed as:

[0097]

[0098] In the fault propagation SIR model, it is necessary to analyze whether a single software fault will break out in the cluster system when the initial fault ratio is very low, and whether it will affect the reliability of the unmanned cluster software system;

[0099] Step 47: Perform single simulation and cyclic simulation on the unmanned cluster software defect propagation simulation model to obtain the system average repair time, etc., and evaluate the system's ability to withstand and repair software defects based on the system average repair time. When the software performs various tasks to achieve the goal, simulate the internal single node failure and calculate the system average repair time of the unmanned cluster software, and implement defect detection based on the system average repair time.

[0100] Example 2

[0101] Based on the above embodiments, in a specific embodiment, unmanned cluster software defect detection is implemented based on the Netlogo simulation environment, multi-agent modeling is performed on the software system structure, a fault propagation simulation model is designed and implemented, and software fault propagation modeling simulation analysis is completed.

[0102] Fault propagation model single simulation:

[0103] First, select the fault propagation model and set the model parameters. The system status interface initially displays the initial state of the system before simulation, which is used to display the system structure. Start simulation to perform a single simulation of the fault propagation model, output the node name injected with the fault at the start of the simulation, and the node name in the fault state in each simulation time step.

[0104] During the model simulation process, the changes in various statistical indicators can be displayed, including the number and proportion of faulty nodes, the number and proportion of resistant nodes, and the number and proportion of nodes that may fail. The simulation ends when the proportion of faulty nodes drops to 0. Figure 4 In the statistical graph shown, the red line shows the proportion of faulty nodes, the green line shows the proportion of resistant nodes, and the blue line shows the proportion of nodes that may fail, which can show how the proportions of these three types of nodes change with the simulation process.

[0105] After the simulation is finished, the system status is as follows Figure 5 As shown, green shows nodes that are resistant and blue shows nodes that may still fail.

[0106] Fault propagation model loop simulation:

[0107] Similar to a single simulation, during the cyclic simulation of the fault propagation model, the data output area displays the node name where the fault is injected at the beginning of the simulation, the time step and real time from the injection of the fault to the completion of the repair in each round of simulation, so as to calculate the real time required for each time step on average. Figure 6 As shown, the average repair time of the system changes with the increase of the number of cycles. The simulation ends after completing the specified number of cycles.

[0108] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0109] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An unmanned cluster software defect detection method, characterized in that: The following steps are involved: Step 1: Conduct a multi-agent system emergence analysis on the unmanned cluster software to obtain the system evolution law and group behavior mode; Step 2: According to the system evolution law and group behavior mode, the multi-agent modeling method and fault propagation SIR model are used to perform defect propagation simulation analysis on the unmanned cluster software to obtain defect detection results; Step 1 includes the following steps: Step 11: Using multi-agent method and evolutionary game theory, analyze the transfer trajectory of the monomer adaptive behavior state to obtain the monomer adaptive behavior state; Step 12: Based on the state of the monomer adaptive behavior, the system dynamics method is used to analyze the internal mechanism and feedback loop of the unmanned cluster software evolution process, analyze the impact of different factors on the evolution of the unmanned cluster software, and obtain the mechanism of the emergence characteristics as the law of system evolution; Step 13: Based on the system evolution law, the complex network self-organization evolution analysis method is used to build an unmanned cluster software association network evolution model to obtain the group behavior mode; Step 2 includes the following steps: Step 21: According to the system evolution law and group behavior mode, a single agent model is constructed for each single agent in the unmanned swarm software using a cognitive agent structure; according to the connection relationship between single agents, a connection agent model is constructed using a reactive agent structure; Step 22: Construct a cluster environment operation model according to the system evolution law and group behavior mode; the cluster environment operation model includes the cluster system model, software system task profile, task target distribution information and monomer collaborative interaction information; Step 23: Construct a task goal generation model according to the system evolution law and group behavior mode; in the task goal generation model, any single agent model publishes the task goal, or the same single agent model publishes different task goals, or multiple single agent models publish the same task goal; Step 24: Create a single task call model based on the system evolution law and group behavior mode; determine the target and establish the participation state based on the single task target requirements and the capabilities of the single Agent model; Step 25: According to the system evolution law and group behavior mode, and in accordance with the cluster formation rules, build cluster dynamic collaborative relationships; Step 26: Construct an unmanned cluster software model based on the single agent model, the connection agent model, the cluster environment operation model, the task target generation model, the single task call model and the cluster dynamic collaborative relationship; Step 27: Based on the fault propagation SIR model, an unmanned cluster software defect propagation simulation model is constructed for the unmanned cluster software model, and the Monte Carlo method is used to simulate the unmanned cluster software defect propagation simulation model to obtain defect detection results of the unmanned cluster software. The defect detection results include the error node diffusion ratio and the system average repair time.

2. The unmanned cluster software defect detection method according to claim 1, characterized in that: The unmanned cluster software defect propagation simulation model adopts a two-state compartment model, which is expressed as: Among them, each single Agent model in the unmanned cluster software model is regarded as a node; I(t) represents the number of implicit error nodes at time t; a represents the infection rate; c(t) represents the connection probability of implicit error node i at time t; N represents the total number of nodes.

3. The unmanned cluster software defect detection method according to claim 2, characterized in that: Solve the unmanned cluster software defect propagation simulation model to calculate the error node diffusion ratio i(t), which is expressed as: Among them, I(t) represents the number of implicit error nodes at time t; a represents the infection rate; c(t) represents the connection probability of implicit error node i at time t; N represents the total number of nodes; t represents time t.

4. The unmanned cluster software defect detection method according to claim 1, characterized in that: The Monte Carlo method is used to simulate the unmanned cluster software defect propagation simulation model. The number and proportion of faulty nodes, resistant nodes, and nodes that may fail in the simulation process, as well as the simulation step time and the probability of defective node recovery are used to calculate the single repair time for all faulty nodes in a simulation process. After multiple cycles of simulation, the average value is calculated based on multiple groups of single repair times to obtain the system average repair time.

5. The unmanned cluster software defect detection method according to claim 2, characterized in that: The single agent model constructed with cognitive agent structure is represented by a multi-tuple, which includes the sequence number of the single agent, state set, perception set, EQ→Mono_P mapping function and environment state set.

6. The unmanned cluster software defect detection method according to claim 1, characterized in that: The connection agent model constructed with the reactive agent structure is represented by a quaternary group, which includes the serial number of the connection agent, the state set of the connection agent, the knowledge base of the connection agent and the action set of the connection agent.