Non-steady-state architecture-based same-type cluster control method and system and storage medium

Abstract

The application relates to the technical field of cluster control and discloses a same-type cluster control method and system based on a non-steady-state architecture and a storage medium, the method comprising the following steps: based on the non-steady-state architecture, performing dynamic consistency analysis on environmental multi-dimensional sensing data to obtain a consistency index and a member role allocation scheme; performing task decomposition on the consistency index and the member role allocation scheme to obtain an optimized subtask sequence; performing path coordination planning on the consistency index and the member role allocation scheme to obtain a coordination path; executing the coordination path on a control strategy and monitoring the coordination behavior of the cluster to generate a behavior consistency log; based on the behavior consistency log, performing predictive evolution analysis on the non-steady-state architecture to obtain architecture reconstruction parameters; and based on the architecture reconstruction parameters, performing multi-objective coordinated optimization on the control strategy to obtain a final control strategy; and the application can improve the efficiency of the same-type cluster control method based on the non-steady-state architecture.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a cluster control method, system, and storage medium of the same type based on an unsteady architecture. Background Technology

[0002] In the field of cluster control under non-steady-state architecture, existing technologies are unable to perform efficient dynamic consistency analysis on multi-dimensional environmental sensing data of the cluster; their data processing lacks effective noise filtering and spatiotemporal alignment mechanisms, making it difficult to accurately obtain cluster consistency indicators, which in turn makes the allocation of member roles lack a scientific basis and cannot fully match the dynamic operating state of the cluster.

[0003] Existing technologies have significant shortcomings in cluster task decomposition, path planning, and control strategy optimization. They cannot achieve reasonable task decomposition and collaborative path planning based on consistency indicators and role allocation schemes, and have weak ability to predict and resolve potential conflicts. The monitoring and analysis of cluster collaborative behavior and the predictive evolution adjustment of the architecture are inadequate, the optimization of control strategies lacks specificity, and it is difficult to adapt to the dynamic changes of non-steady-state architectures, ultimately leading to low cluster control efficiency. Summary of the Invention

[0004] This disclosure provides a cluster control method, system, and storage medium based on an unsteady architecture.

[0005] In the first aspect, this disclosure provides a cluster control method based on a non-steady-state architecture, including: S1, performing dynamic consistency analysis on the multi-dimensional environmental sensing data of the cluster based on the non-steady-state architecture to obtain the consistency index and member role allocation scheme of the cluster; S2. Decompose the consistency index and the member role allocation scheme into tasks to obtain the optimized sub-task sequence of the cluster; S3. Perform path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster; S4. Execute the collaborative path according to the control policy of the cluster, monitor the collaborative behavior of the cluster, and generate the behavior consistency log of the cluster; S5. Based on the behavior consistency log, perform predictive evolution analysis on the non-steady-state architecture to obtain the architecture reconstruction parameters of the cluster; S6. Based on the architecture reconstruction parameters, perform multi-objective collaborative optimization on the control strategy to obtain the final control strategy of the cluster.

[0006] In a preferred embodiment, the step of performing dynamic consistency analysis on the multi-dimensional environmental sensor data of the cluster based on the non-steady-state architecture to obtain the consistency index and member role allocation scheme of the same type of cluster includes: The environmental multidimensional sensing data is subjected to noise filtering and spatiotemporal alignment to obtain the purified environmental data of the cluster. Based on the aforementioned unsteady architecture, a multi-dimensional coordination analysis is performed on the purified environment data to obtain the real-time coordination data of the cluster. Consistency features are extracted from the real-time coordination data to obtain the consistency index of the cluster; According to the dynamic role mapping mechanism of the cluster, the consistency index is assigned to the member roles in the cluster to obtain the member role allocation scheme of the cluster.

[0007] In a preferred embodiment, the step of decomposing the consistency index and the member role allocation scheme into an optimized sub-task sequence for the cluster includes: Based on the member role allocation scheme, a dynamic dependency graph between the cluster's control tasks and the member roles is constructed. Based on the consistency index, critical paths are identified in the dynamic dependency graph to obtain the optimized task flow of the cluster. Based on the optimized task flow, the overall task of the cluster is decomposed into atomic task units of the cluster. The atomic task units are conflict-resolved and dynamically prioritized to obtain the optimized subtask sequence of the cluster.

[0008] In a preferred embodiment, the step of performing path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster includes: A topology analysis is performed on the member role allocation scheme to obtain the dynamic communication topology among the member roles; Based on the consistency index, collaborative conflict prediction is performed on the dynamic communication topology to obtain the set of potential conflict domains of the cluster. Gradient optimization is performed on the conflict paths in the set of potential conflict domains to obtain the priority paths of the cluster; The priority path is subjected to spatiotemporal consistency verification to obtain the collaborative path of the cluster.

[0009] In a preferred embodiment, the control policy for the cluster executes the cooperative path, monitors the cooperative behavior of the cluster, and generates a behavior consistency log for the cluster, including: Collect multi-source behavioral data streams when the cluster executes the collaborative path; Standardized feature extraction is performed on the multi-source behavioral data stream to obtain the standardized behavioral feature vector of the cluster; The behavioral consistency metric of the cluster is calculated based on the standardized behavioral feature vector, wherein the calculation formula for the behavioral consistency metric is as follows: ; in, This represents a measure of the consistency of the behavior. Indicates the number of member roles. Indicates the first The standardized behavioral feature vectors of the aforementioned member roles, This represents the mean of all the standardized behavioral feature vectors. This represents the variance of the standardized behavioral feature vector. Indicates the first The variance between the standardized behavioral feature vector and the mean vector of each member role This indicates the preset adjustment parameters; Anomaly pattern identification is performed on the behavioral consistency metric to obtain the behavioral consistency log of the cluster.

[0010] In a preferred embodiment, the step of standardizing feature extraction from the multi-source behavioral data stream to obtain the standardized behavioral feature vector of the cluster includes: Modal fusion is performed on the multi-source behavioral data streams to obtain the fused behavioral data of the cluster; Extract outliers from the fused behavioral data and perform integrity repair on the fused behavioral data to obtain the purified behavioral data of the cluster; Multi-scale feature extraction is performed on the purified behavioral data to obtain the behavioral feature set of the cluster; By applying dimensional constraints to the behavioral feature set, a standardized behavioral feature set for the cluster is obtained. Based on the standardized behavioral feature set, a standardized behavioral feature vector of the cluster is generated.

[0011] In a preferred embodiment, the step of performing predictive evolution analysis on the non-steady-state architecture based on the behavioral consistency log to obtain the architecture reconfiguration parameters of the cluster includes: Time-series pattern mining is performed on the behavior consistency logs to obtain the behavioral evolution trajectory features of the cluster. Based on the behavioral evolution trajectory characteristics, the vulnerability of key nodes in the non-steady-state architecture is assessed, and a node vulnerability map of the non-steady-state architecture is obtained. Based on the node vulnerability map, a dependency strength analysis is performed on the non-steady-state architecture to obtain a dependency reconstruction strategy for the non-steady-state architecture. Based on the dependency restructuring strategy, the architecture restructuring parameters of the cluster are generated.

[0012] In a preferred embodiment, the step of performing multi-objective collaborative optimization of the control strategy based on the architecture refactoring parameters to obtain the final control strategy of the cluster includes: The architecture reconstruction parameters are optimized using multiple objectives to obtain the set of optimization objectives for the cluster. Based on the set of optimization objectives, the feasibility and interrelationships of the control strategies are determined, and the set of adaptive strategies for the cluster is obtained. Pareto front analysis is performed on the adaptive policy set to obtain the non-dominated set of the adaptive policy set; Based on preset cluster collaboration criteria, collaborative optimization is performed on the non-dominated set to obtain the final control strategy of the cluster.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention performs dynamic consistency analysis on multi-dimensional sensor data of the cluster environment to accurately obtain consistency indicators and member role allocation schemes, laying a precise foundation for subsequent task processing; at the same time, through scientific task decomposition and path collaborative planning, it effectively improves the orderliness and collaboration of cluster task execution and significantly improves the overall efficiency of cluster control.

[0014] 2. This invention enables predictive evolution analysis of non-steady-state architectures by real-time monitoring of cluster collaborative behavior and generation of behavior consistency logs, providing a reliable basis for architecture reconfiguration; and performs multi-objective collaborative optimization of control strategies based on architecture reconfiguration parameters, enabling cluster control strategies to dynamically adapt to architecture changes, further improving the accuracy and stability of cluster control. Attached Figure Description

[0015] Figure 1 The flowchart of the control method for the same type of cluster based on an unsteady architecture according to Embodiment 1 of the present invention is shown. Figure 2 The diagram shows the functional block diagram of a cluster control system of the same type based on an unsteady architecture according to Embodiment 2 of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.

[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0018] Example 1 Figure 1 This is a flowchart illustrating a cluster control method based on a non-steady-state architecture, provided in an embodiment of this disclosure. Figure 1 As shown, control methods for similar clusters based on non-steady-state architectures include: S1. Based on the non-steady-state architecture, perform dynamic consistency analysis on the multi-dimensional environmental sensing data of the cluster to obtain the consistency index and member role allocation scheme of the cluster. In this embodiment of the invention, the step of performing dynamic consistency analysis on the multi-dimensional environmental sensing data of the cluster based on an unsteady architecture to obtain the consistency index and member role allocation scheme of the same type of cluster includes: The environmental multidimensional sensing data is subjected to noise filtering and spatiotemporal alignment to obtain the purified environmental data of the cluster. Based on the aforementioned unsteady architecture, a multi-dimensional coordination analysis is performed on the purified environment data to obtain the real-time coordination data of the cluster. Consistency features are extracted from the real-time coordination data to obtain the consistency index of the cluster; According to the dynamic role mapping mechanism of the cluster, the consistency index is assigned to the member roles in the cluster to obtain the member role allocation scheme of the cluster.

[0019] Specifically, when performing noise filtering on multi-dimensional environmental sensor data, each set of sensor data is checked one by one to identify abnormal values ​​that significantly exceed the normal data range. These abnormal values ​​are then removed from the dataset. For any gaps that appear after removing abnormal values, the average of the adjacent normal data before and after that position is used to fill the gaps, thus completing the noise filtering. Subsequently, spatiotemporal alignment processing is performed. Based on the cluster's preset unified time standard and spatial coordinate system, the time stamps and spatial location information of all multi-dimensional environmental sensor data are adjusted to ensure that each data point is synchronized in the time dimension and corresponds in the spatial dimension, ultimately obtaining the cluster's purified environmental data.

[0020] Furthermore, based on the dynamic characteristics of the non-steady-state architecture, information from multiple dimensions such as temperature, humidity, air pressure, and pollutant concentration is extracted from the clean environment data. The correlation and synchronicity of changes between data in each dimension are analyzed to determine the collaborative operation of data in different dimensions under non-steady-state conditions. Through quantitative analysis of these collaborative conditions, the real-time collaborative degree data of the cluster is obtained.

[0021] Furthermore, the real-time collaboration data is systematically analyzed to identify key characteristics that reflect the consistency of collaboration among cluster members. These characteristics include the stability of data collaboration changes, the compatibility of data collaboration among different members, and the patterns of collaborative data fluctuations. These key characteristics are integrated and summarized to form quantitative indicators that reflect the overall level of collaboration consistency of the cluster, i.e., the cluster's consistency indicators. Further, the cluster's dynamic role mapping mechanism clarifies the member role functions and responsibilities corresponding to different consistency indicators. Based on this mechanism, and combined with the actual situation of each member in the cluster, such as hardware configuration, task execution capabilities, and historical performance, the obtained consistency indicators are accurately assigned to each member, ensuring that each member's role matches the corresponding consistency indicator, thus ultimately determining the cluster's member role allocation scheme.

[0022] In summary, the control method for similar clusters based on non-steady-state architecture can accurately obtain consistency indicators and scientific member role allocation schemes by processing multi-dimensional sensor data of the cluster environment through noise filtering and spatiotemporal alignment. This solves the problems of inefficient data processing and lack of basis for role allocation in existing technologies. Furthermore, by combining task decomposition and path collaborative planning based on indicators and role schemes, as well as predictive evolution analysis of architecture and multi-objective optimization of control strategies based on behavioral consistency logs, it can effectively adapt to the dynamic changes of non-steady-state architecture and significantly improve the control efficiency and adaptability of similar clusters.

[0023] In summary, specialized algorithms were designed for various typical rescue scenarios, including large-scale dispersed operations, small-scale concentrated operations, and narrow and confined spaces. For example, the Hungarian algorithm was used for low-cost task allocation in large-scale dispersed scenarios, K-means++ and simulated annealing algorithms were used to optimize task planning in small-scale concentrated scenarios, and formation scaling and fast exploration algorithms were used to address spatial constraints in narrow and confined spaces. At the same time, a heterogeneous bee colony system was integrated, combined with real-time obstacle avoidance, multi-objective optimization, and priority evaluation, which not only ensured the accuracy and safety of rescue operations, but also prioritized the rescue of high-risk personnel, significantly improving the response speed, resource utilization, and mission completion quality of maritime rescue.

[0024] S2. Decompose the consistency index and the member role allocation scheme into tasks to obtain the optimized sub-task sequence of the cluster; In this embodiment of the invention, the step of decomposing the consistency index and the member role allocation scheme into an optimized sub-task sequence for the cluster includes: Based on the member role allocation scheme, a dynamic dependency graph between the cluster's control tasks and the member roles is constructed. Based on the consistency index, critical paths are identified in the dynamic dependency graph to obtain the optimized task flow of the cluster. Based on the optimized task flow, the overall task of the cluster is decomposed into atomic task units of the cluster. The atomic task units are conflict-resolved and dynamically prioritized to obtain the optimized subtask sequence of the cluster.

[0025] Specifically, we first clarify the specific responsibilities and execution capabilities of each member role in the member role allocation scheme. For example, a member role is responsible for environmental data transmission, and another member role is responsible for task instruction execution. Then, we sort out the overall control tasks of the cluster, break down the various execution links included in the control tasks, and then analyze the correspondence between each execution link and the member role, as well as the sequential dependencies of different member roles in the execution of task links. For example, member role B can only start the task instruction generation link after member role A completes the data preprocessing link. These task links, member roles and dependencies are presented in the form of nodes and edges to construct a dynamic dependency graph between the cluster's control tasks and member roles.

[0026] Furthermore, we first clarify the cluster collaboration consistency requirements reflected by the consistency indicators, such as the time synchronization of task execution and the accuracy of data interaction. Then, based on these requirements, we traverse all task paths in the dynamic dependency graph, determine the degree of impact of each path on the overall cluster control task completion efficiency and collaborative stability, and screen out the paths that play a key role in meeting the consistency indicators, such as paths that can ensure timely and synchronous task completion and reduce data interaction errors. We then eliminate redundant paths in the graph that contribute little to the consistency indicators, and finally obtain the optimized task flow of the cluster.

[0027] Furthermore, each task in the optimized task flow is first broken down in detail to determine the minimum execution unit standard for each step. That is, the unit cannot be further divided into smaller tasks that can be executed independently by a single member role. Then, according to this standard, the overall cluster task covered by the optimized task flow is gradually broken down. For example, the "cooperative path execution" step is broken down into units such as "path reception", "path resolution", "execution instruction generation" and "instruction sending". This ensures that each broken-down unit has clear execution content and execution subject, and finally, the atomic task units of the cluster are obtained.

[0028] Furthermore, first check whether there are conflicts between all atomic task units, including resource conflicts and time conflicts. If a conflict is found, adjust the execution role or execution time of the conflicting unit according to the spare resources and functional substitution capabilities of each role in the member role allocation scheme. For example, assign the unit that is competing for the same device to other roles that have the same device usage rights. Then, combined with the consistency index, dynamically prioritize the non-conflicting atomic task units, set the units that have a greater impact on the cluster's collaborative consistency and are more critical to the execution results to higher priority, and arrange all atomic task units in descending order of priority to finally obtain the optimized subtask sequence of the cluster.

[0029] In summary, in the embodiments of the present invention, when decomposing the consistency index and member role allocation scheme into a cluster optimization sub-task sequence, a dynamic dependency graph between control tasks and member roles is first constructed based on the member role allocation scheme. This operation can clearly define the association logic between different roles and corresponding tasks, avoid execution chaos caused by ambiguous matching of roles and tasks, lay a precise association foundation for subsequent task decomposition, and solve the problem of insufficient correspondence between tasks and roles in the prior art.

[0030] In summary, identifying critical paths in dynamic dependency graphs based on consistency metrics to obtain optimized task flows can accurately locate core task paths that play a decisive role in overall task efficiency by relying on the cluster collaboration status reflected by consistency metrics, eliminating redundant or non-critical task associations, ensuring that the task flow revolves around the cluster collaboration goal, and improving the directionality and efficiency of task execution.

[0031] In summary, by decomposing the overall cluster task into atomic task units based on optimized task flow, the complex overall task can be broken down into the smallest executable unit, making the goal of each unit clearer and the execution difficulty lower. This facilitates subsequent allocation to cluster members with corresponding roles and provides a refined task carrier for subsequent conflict handling and priority sorting, thereby enhancing the operability of task execution.

[0032] In summary, resolving conflicts and dynamically prioritizing atomic task units can eliminate potential resource contention and timing conflicts that may occur during the execution of different task units. At the same time, adjusting the task execution order according to the actual operation requirements of the cluster ensures that critical tasks are prioritized. The resulting optimized sub-task sequence enables cluster members to execute tasks efficiently in sequence, significantly improving the overall coordination and efficiency of cluster task execution. This makes up for the shortcomings of existing technologies in task decomposition, which lack conflict handling and priority planning.

[0033] S3. Perform path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster; In this embodiment of the invention, the step of performing path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster includes: A topology analysis is performed on the member role allocation scheme to obtain the dynamic communication topology among the member roles; Based on the consistency index, collaborative conflict prediction is performed on the dynamic communication topology to obtain the set of potential conflict domains of the cluster. Gradient optimization is performed on the conflict paths in the set of potential conflict domains to obtain the priority paths of the cluster; The priority path is subjected to spatiotemporal consistency verification to obtain the collaborative path of the cluster.

[0034] Specifically, we first analyze the specific functional positioning, task execution scope, and required communication objects of each member role in the member role allocation scheme, clarifying the data interaction needs of different member roles during task execution. For example, the member role responsible for environmental monitoring needs to transmit monitoring data to the member role responsible for task decision-making, and the member role responsible for task execution needs to receive instructions from the decision-making member role. Then, we construct an initial communication topology framework with member roles as nodes and communication relationships between member roles as edges. At the same time, we supplement the adjustment rules of member role communication relationships under different task stages by combining the dynamic change characteristics of member role tasks. For example, after a member role completes a stage task, its communication object switches from the decision-making role to the collaborative execution role, ultimately forming a dynamic communication topology structure that can reflect the real-time communication relationship between member roles.

[0035] Furthermore, we first clarify the requirements related to communication coordination in the consistency indicators, including communication delay thresholds, data transmission bandwidth limits, and information synchronization frequency standards. These indicators are the core basis for judging whether there are communication conflicts. Then, we traverse each communication path in the dynamic communication topology, analyze the resource consumption and time scheduling of each path during data transmission, and identify communication path combinations with resource contention or time overlap. For example, two communication paths may occupy the same communication frequency band and the total bandwidth may exceed the limit, or the transmission times of two paths may completely overlap, resulting in data synchronization delay. These communication paths with conflict risks, along with the member roles and conflict types involved, are collectively defined as a conflict domain. All conflict domains are then aggregated to form a set of potential conflict domains for the cluster.

[0036] Furthermore, we first analyze the importance of the cluster tasks supported by each conflict path in the potential conflict domain set. For example, the path supporting the data transmission of the core rescue mission is more important than the path supporting the data transmission of auxiliary monitoring missions. Combining the implicit requirements of task priority in the consistency index, we determine the priority ranking rules for each conflict path. Then, according to the order of priority from high to low, we adjust the resource allocation and time scheduling of the conflict paths. For high-priority paths, we prioritize ensuring the resources such as communication bandwidth and transmission time required by them. For example, we allocate independent communication frequency bands to core mission paths to avoid bandwidth competition, or adjust the transmission time of low-priority paths to avoid the transmission period of high-priority paths. We gradually resolve the conflict problem of each conflict path, and finally form priority paths that meet the resource and time requirements and are ranked by importance.

[0037] Furthermore, priority paths are first validated from a time perspective. The preset transmission time of each priority path is compared with the task execution time of the corresponding member role to check for any overlap between the path transmission time and the time when the member role is performing other tasks, which could prevent the member role from processing data or receiving instructions in a timely manner. For example, if a member role needs to receive data from a high-priority path while performing a task, the path transmission time needs to be adjusted to the interval between the member role's task operations. Then, validation is performed from a spatial perspective. If the member role is an execution unit with spatial mobility capabilities, it is necessary to check whether the spatial movement trajectories involved in different priority paths intersect or overlap, to avoid collisions between member roles in space due to path overlap. For example, if the navigation trajectory corresponding to the communication path of an unmanned surface vessel intersects with the trajectory of other unmanned surface vessels, the spatial direction of one of the trajectories needs to be adjusted. After dual validation in both time and spatial dimensions, it is ensured that all paths are synchronized in time and free from spatial conflicts, ultimately resulting in the collaborative path of the cluster.

[0038] In summary, in the embodiments of the present invention, when performing path collaborative planning for consistency indicators and member role allocation schemes, the member role allocation scheme is first subjected to topology analysis to obtain a dynamic communication topology. This operation can clearly sort out the communication connection relationship and data interaction logic between different roles based on the determined member role division, forming a dynamic communication framework that adapts to the role division. Its dynamic characteristics are just right to meet the needs of non-steady-state architecture, providing a clear communication relationship basis for subsequent collaborative path planning among cluster members, and avoiding collaborative chaos caused by ambiguous communication relationships.

[0039] In summary, predicting collaborative conflicts in dynamic communication topologies based on consistency indices to obtain a set of potential conflict domains allows for accurate identification of collaborative issues such as path crossings, resource contention, and timing conflicts that may occur in dynamic communication topologies, relying on the cluster collaboration status reflected by the consistency indices. This enables the early identification of potential conflict areas. Compared to existing technologies with weak conflict prediction capabilities, this step provides clear direction for subsequent conflict resolution, ensuring the stability of cluster path collaboration and meeting the goal of improving cluster control efficiency under non-steady-state architectures.

[0040] In summary, gradient optimization of conflict paths in the potential conflict domain set to obtain priority paths allows for prioritization of the importance and urgency of different paths based on the cluster's task requirements and member role characteristics. This prioritizes ensuring the smooth execution of core tasks by key roles, while eliminating or adjusting non-critical conflict paths, making path planning more targeted. This aligns with the idea of ​​prioritizing critical paths in various scenarios, ensuring that the cluster can prioritize core objectives and improve path utilization efficiency when executing tasks.

[0041] In summary, performing spatiotemporal consistency verification on priority paths to obtain collaborative paths allows for verification of the compatibility of priority paths in both time and space dimensions. This ensures that when each member role in the cluster executes according to the path, they can coordinate synchronously in time and avoid interference in space. The resulting collaborative path fully adapts to the dynamic operation requirements of the cluster under non-steady-state architecture, effectively compensating for the shortcomings of existing technologies that lack spatiotemporal verification in path planning and are prone to collaborative disorder. This further improves the coordination and reliability of cluster control, which is consistent with the core requirements of optimizing cluster control strategies.

[0042] S4. Execute the collaborative path according to the control policy of the cluster, monitor the collaborative behavior of the cluster, and generate the behavior consistency log of the cluster; In this embodiment of the invention, the control policy for the cluster executes the cooperative path, monitors the cooperative behavior of the cluster, and generates a behavior consistency log for the cluster, including: Collect multi-source behavioral data streams when the cluster executes the collaborative path; Standardized feature extraction is performed on the multi-source behavioral data stream to obtain the standardized behavioral feature vector of the cluster; The behavioral consistency metric of the cluster is calculated based on the standardized behavioral feature vector, wherein the calculation formula for the behavioral consistency metric is as follows: ; in, This represents a measure of the consistency of the behavior. Indicates the number of member roles. Indicates the first The standardized behavioral feature vectors of the aforementioned member roles, This represents the mean of all the standardized behavioral feature vectors. This represents the variance of the standardized behavioral feature vector. Indicates the first The variance between the standardized behavioral feature vector and the mean vector of each member role This indicates the preset adjustment parameters; Anomaly pattern identification is performed on the behavioral consistency metric to obtain the behavioral consistency log of the cluster.

[0043] In this embodiment of the invention, the step of performing standardized feature extraction on the multi-source behavioral data stream to obtain the standardized behavioral feature vector of the cluster includes: Modal fusion is performed on the multi-source behavioral data streams to obtain the fused behavioral data of the cluster; Extract outliers from the fused behavioral data and perform integrity repair on the fused behavioral data to obtain the purified behavioral data of the cluster; Multi-scale feature extraction is performed on the purified behavioral data to obtain the behavioral feature set of the cluster; By applying dimensional constraints to the behavioral feature set, a standardized behavioral feature set for the cluster is obtained. Based on the standardized behavioral feature set, a standardized behavioral feature vector of the cluster is generated.

[0044] Specifically, during the process of the cluster executing the control strategy according to the collaborative path, the behavior data of each member role is collected in real time through the sensors carried by each member role of the cluster. This includes the real-time location data of the member role, task execution progress data, and interaction response data with other members. At the same time, the execution time sequence data and resource usage data of the overall task are collected through the central control interface of the cluster. These behavioral data from different collection sources and of different types together constitute the multi-source behavioral data stream of the cluster.

[0045] Furthermore, the collected multi-source behavioral data streams are first preprocessed, and missing and outlier values ​​are checked one by one. For missing data, they are supplemented based on the average historical behavioral data of the member role in the same task phase. Outliers that exceed the normal behavioral range are directly removed. Next, key information reflecting the behavioral characteristics of the member role is extracted from the preprocessed data, including task execution efficiency characteristics, collaborative interaction characteristics, and path following characteristics. Then, all extracted feature information is converted into numerical values ​​with uniform dimensions to ensure that each feature is comparable under the same dimension. Finally, all standardized features of each member role are arranged in a preset order to form the standardized behavioral feature vector of that member role. The standardized behavioral feature vectors of all member roles together constitute the standardized behavioral feature vector of the cluster.

[0046] Furthermore, the total number of member roles in the cluster is first determined, and then the mean vector of the standardized behavioral feature vectors of all member roles is calculated. This mean vector reflects the average behavioral level of the cluster as a whole. Then, the degree of difference between the standardized behavioral feature vector of each member role and the mean vector is calculated, including the magnitude of the difference between the corresponding feature values ​​in the vector and the deviation trend of the overall vector. Combined with the preset adjustment rules, the degree of behavioral difference of all member roles is comprehensively calculated, and finally a value that can quantitatively reflect the overall behavioral consistency level of the cluster is obtained, that is, the behavioral consistency measure of the cluster.

[0047] Furthermore, based on the historical behavior data of the cluster during normal execution of the collaborative path, the normal range of behavioral consistency metrics is determined, and the judgment criteria for different types of abnormal modes are clarified. Then, the currently calculated behavioral consistency metrics are compared with the normal range. If the metric value exceeds the normal range, the type of abnormal mode is determined according to the judgment criteria, and the specific time of the abnormality, the roles of the members involved, and the behavioral data at the time of the abnormality are recorded. Finally, the abnormal judgment results, abnormal related information, and the trend of changes in normal behavioral consistency metrics during the cluster's execution process are organized into a structured record in chronological order to form the cluster's behavioral consistency log.

[0048] Specifically, the standardized behavior feature vector is obtained by modal fusion of multi-source behavior data streams collected when the cluster executes the collaborative path to obtain fused behavior data, extracting outliers from the fused behavior data and performing integrity repair to obtain purified behavior data, extracting multi-scale features from the purified behavior data to obtain a behavior feature set, and generating the behavior feature set after applying dimensional constraints.

[0049] Furthermore, the mean of all standardized behavioral feature vectors is obtained by summing all the generated standardized behavioral feature vectors and then dividing by the number of member roles.

[0050] Furthermore, the variance of the standardized behavioral feature vector is obtained by first calculating the difference between each standardized behavioral feature vector and the mean vector, squaring each difference, summing all the squared differences, and then dividing by the number of member roles.

[0051] Furthermore, the first The method for calculating the variance between the standardized behavioral feature vector and the mean vector of each member role is the same as the method for calculating the variance of the standardized behavioral feature vector, except that it only applies to the first member role. The standardized behavioral feature vector and mean vector of each member role are calculated.

[0052] Furthermore, the number of member roles is the specific number of member roles divided in the cluster, which is determined when formulating the member role allocation scheme for the cluster.

[0053] Furthermore, the preset adjustment parameters are fixed values ​​that are pre-set based on the actual needs and experience of cluster control before calculating the behavioral consistency metric.

[0054] Furthermore, this calculation process is to quantify the degree of fit between the behavior of all member roles in the cluster and the overall average behavior. The higher the value, the stronger the behavioral consistency among cluster members, and vice versa.

[0055] Furthermore, the smaller the difference between the standardized behavioral feature vector and the mean vector of a member role, the larger the corresponding calculated value, which will result in a higher value for the overall behavioral consistency measure.

[0056] Furthermore, the smaller the dispersion of the standardized behavioral feature vector, the larger the corresponding calculated value, which will drive up the value of the behavioral consistency measure.

[0057] Furthermore, when the preset adjustment parameter increases, the value of the exponential part will increase, thereby improving the value of the overall behavioral consistency measure.

[0058] Specifically, we first identify the different modal types contained in the multi-source behavioral data stream, such as location sensor data of cluster members, task execution progress data, interaction response data with other members, and energy consumption data. Then, based on a unified timestamp, we associate and match data of different modalities at the same time node. For example, we integrate the location data of a member at a certain moment, the task execution progress data at that moment, and the response data with other members at that moment. At the same time, we remove invalid associated data caused by time asynchrony in different modal data, and finally form cluster fusion behavioral data that contains multimodal information and is time-aligned.

[0059] Furthermore, based on the fusion behavior data of the cluster during the historical normal execution of collaborative paths, the normal value range of each type of data is determined. Then, each record in the current fusion behavior data is traversed, and data that exceeds the normal value range is marked as anomalies and directly removed. For data gaps that appear after removing anomalies, the gaps are filled by the average of the same type of data at adjacent time points before and after the gap, or by the corresponding data of other members with the same role at the same time point, to ensure that all data records are complete and without anomalies, and finally the purified behavior data of the cluster is obtained.

[0060] Furthermore, multi-scale extraction standards are set from two dimensions: time and function. The time scale includes real-time scale, short-term scale, and long-term scale, while the function scale includes path following function, task execution function, and collaborative interaction function. Corresponding features are extracted from the purified behavioral data one by one according to these scales. For example, the instantaneous position deviation of each member per second is extracted at the real-time scale, and the average task execution efficiency of each member per ten minutes is extracted at the short-term scale. The extraction results of all different scales and functions are summarized and integrated to form a behavioral feature set of the cluster.

[0061] Furthermore, a correlation analysis is first performed on all features in the behavioral feature set to calculate the degree of correlation between any two features. If the degree of correlation between two features exceeds a preset threshold, they are identified as redundant features, and the features that better reflect the core behavior of the cluster are retained. Then, the values ​​of the remaining features are converted to a unified dimension, for example, by using a linear transformation to map all feature values ​​to the range of 0-1, ensuring that the numerical ranges of different types of features are consistent and comparable. Finally, features that play a key role in judging the consistency of cluster behavior are selected to form a standardized behavioral feature set for the cluster.

[0062] Furthermore, according to the preset feature sorting rules, all features in the standardized behavior feature set are arranged in order, and a one-dimensional data sequence containing all sorted feature values ​​is generated for each member role. Each data sequence is the standardized behavior feature vector of that member role, and the standardized behavior feature vectors of all member roles together constitute the standardized behavior feature vector of the cluster.

[0063] In summary, in this embodiment of the invention, collecting multi-source behavioral data streams during the execution of collaborative paths by the cluster can comprehensively cover various behavioral information of each member of the cluster during the execution of tasks, providing complete and multi-dimensional data support for subsequent analysis of the cluster's collaborative status. This not only meets the need to dynamically grasp the overall operation of the cluster under non-steady-state architecture, but also provides a data foundation for behavioral monitoring in unmanned platform cluster rescue scenarios, avoiding the problem of one-sided collaborative behavior analysis due to data loss. In summary, standardizing feature extraction from multi-source behavioral data streams to obtain standardized behavioral feature vectors can eliminate the differences between different types and dimensions of behavioral data, provide a unified analytical benchmark for the extracted features, facilitate subsequent comparison of the consistency of behavior among cluster members, solve the analytical error problem caused by chaotic data formats in existing technologies, and lay the foundation for accurate quantitative analysis of cluster behavior under non-steady-state architectures, which is in line with the core goal of improving cluster control efficiency.

[0064] In summary, the behavior consistency metric of the cluster is calculated based on standardized behavior feature vectors. It accurately reflects the synchronization degree and collaborative effect of each member's behavior through quantitative methods. Compared with qualitative analysis, it is more objective and can capture the dynamic changes of cluster collaborative behavior in real time under non-steady-state architecture. It provides a clear basis for judging whether the cluster executes the collaborative path as expected and is also suitable for the requirements of cluster collaborative accuracy in rescue scenarios.

[0065] In summary, identifying abnormal patterns in behavioral consistency metrics to obtain behavioral consistency logs can promptly detect abnormal deviations in cluster collaboration and record abnormal information along with normal data. These logs not only provide crucial information for predictive evolution analysis of subsequent non-steady-state architectures but also offer references for cluster fault diagnosis and strategy adjustment in rescue scenarios. This addresses the shortcomings of insufficient cluster behavior anomaly monitoring in existing technologies and ensures the reliability of cluster control.

[0066] In summary, in the embodiments of the present invention, modal fusion of multi-source behavioral data streams to obtain fused behavioral data of the cluster can integrate behavioral data from different sources and of different types when the cluster executes a collaborative path into a unified dataset, eliminating data isolation and allowing subsequent analysis to be carried out based on complete behavioral information. This not only adapts to the multi-dimensional and fragmented characteristics of cluster behavioral data under non-steady-state architecture, but also meets the need for comprehensive behavioral data in unmanned platform cluster rescue scenarios, providing a unified foundation for subsequent data processing.

[0067] In summary, extracting outliers from fused behavioral data and performing integrity restoration to obtain purified behavioral data can remove noise and errors from the data, while filling in missing parts of the data. This avoids interference from abnormal or incomplete data with subsequent feature extraction, ensuring the accuracy and completeness of the behavioral data. This operation is suitable for situations where data is easily affected by environmental factors and may deviate under non-steady-state architectures, and it can also meet the accuracy requirements of behavioral data in rescue scenarios, providing a guarantee for the reliability of subsequent feature extraction.

[0068] In summary, multi-scale feature extraction of purified behavioral data to obtain a cluster behavioral feature set can mine key behavioral features from different dimensions and levels, capturing both real-time behavioral details of cluster members and grasping overall behavioral trends. This adapts to the dynamic changes in cluster behavior under non-steady-state architectures and meets the needs of multi-level analysis of cluster behavior in rescue scenarios, providing rich feature support for subsequent standardized processing.

[0069] In summary, applying dimensional constraints to the behavioral feature set to obtain a standardized behavioral feature set can eliminate redundant and irrelevant features, reduce feature dimensionality, avoid computational redundancy caused by high-dimensional features, and maintain uniformity in feature set in terms of dimensionality and magnitude. This aligns with the real-time requirements of cluster control under non-steady-state architectures and makes subsequent feature set-based computations more efficient. It also meets the need for rapid analysis of behavioral data in unmanned platform cluster rescue scenarios, thereby enhancing the practicality of the feature set.

[0070] In summary, generating standardized behavioral feature vectors for clusters based on standardized behavioral feature sets can transform discrete standardized features into structured, quantifiable vector forms, making them directly usable for subsequent calculations of behavioral consistency metrics. This provides an operable data format for accurate analysis of cluster collaborative behavior. This process not only meets the needs of quantitative analysis of cluster behavior under non-steady-state architectures but also supports the judgment of cluster behavioral consistency in rescue scenarios, laying the foundation for the subsequent generation of behavioral consistency logs.

[0071] S5. Based on the behavior consistency log, perform predictive evolution analysis on the non-steady-state architecture to obtain the architecture reconstruction parameters of the cluster; In this embodiment of the invention, the step of performing predictive evolution analysis on the non-steady-state architecture based on the behavioral consistency log to obtain the architecture reconfiguration parameters of the cluster includes: Time-series pattern mining is performed on the behavior consistency logs to obtain the behavioral evolution trajectory features of the cluster. Based on the behavioral evolution trajectory characteristics, the vulnerability of key nodes in the non-steady-state architecture is assessed, and a node vulnerability map of the non-steady-state architecture is obtained. Based on the node vulnerability map, a dependency strength analysis is performed on the non-steady-state architecture to obtain a dependency reconstruction strategy for the non-steady-state architecture. Based on the dependency restructuring strategy, the architecture restructuring parameters of the cluster are generated.

[0072] Specifically, we first sort out all the information recorded in the behavior consistency log, including the cluster's behavior consistency metrics at different time points, anomaly pattern types, anomaly occurrence times, involved member roles, and corresponding behavioral data. We then arrange this information in chronological order to form a time series dataset. Next, we analyze the changing trends of the behavior consistency metrics in this dataset, such as whether the metrics continuously rise, fall, or fluctuate periodically within a certain time period. At the same time, we statistically analyze the frequency of anomaly patterns and the patterns of associated member roles, such as specific member roles frequently triggering collaborative anomalies within fixed time intervals. We then identify the transition nodes of behavioral characteristics at different stages, such as the time point when low-frequency anomalies turn into high-frequency anomalies. We integrate these trends, patterns, and transition nodes to obtain the cluster behavior evolution trajectory characteristics that can reflect the process of cluster behavior changing over time.

[0073] Furthermore, we first identify the key nodes in the non-steady-state architecture. These nodes are typically the core units that play a central role in cluster collaborative control, including central control nodes, data interaction nodes, and task scheduling nodes. Then, we associate the behavioral evolution trajectory characteristics with each key node and analyze the performance of each key node in the trajectory characteristics. For example, we examine whether the behavioral consistency metric value of the central control node frequently falls below the normal range within the corresponding time period, and whether the data interaction node triggers collaborative anomalies multiple times. We then set vulnerability assessment criteria based on the degree of influence of key nodes on behavioral evolution. If a key node causes a decrease in behavioral consistency more frequently than a preset number of times, or if the anomaly it triggers has a significant impact on cluster task execution, then the node is judged to have a high vulnerability level; otherwise, it is considered to have a low vulnerability level. Finally, we organize the names, locations, vulnerability levels, and corresponding trajectory characteristics of all key nodes into a structured chart to obtain the node vulnerability map of the non-steady-state architecture.

[0074] Furthermore, the dependencies between all critical nodes are extracted from the node vulnerability map to identify which other nodes each node depends on for data or functional support. For example, a task scheduling node depends on environmental sensor data transmitted by a data interaction node. The strength of these dependencies is then analyzed in conjunction with behavioral evolution trajectory characteristics. If an object that a vulnerable node depends on becomes abnormal, the vulnerable node will immediately trigger a cluster behavior consistency problem, indicating a high dependency strength. If the vulnerable node can still maintain normal functionality through backup paths after the dependent object becomes abnormal, it indicates a low dependency strength. For cases with high dependency strength involving highly vulnerable nodes, adjustment plans are formulated, such as reducing the strong dependency between highly vulnerable nodes and core functional nodes, and adding backup dependent nodes for highly vulnerable nodes. These adjustment plans together constitute the dependency reconstruction strategy for the non-steady-state architecture.

[0075] Furthermore, we first break down the various adjustment requirements in the dependency restructuring strategy. For example, the requirement to "reduce the strong dependency between highly vulnerable communication nodes and task execution nodes" requires clarifying the current dependency strength value and the target dependency strength value of data transmission between communication nodes and task execution nodes, as well as the parameter types required to achieve the strength adjustment. Then, we determine the specific parameter values ​​for each adjustment requirement. For example, we adjust the transmission frequency between communication nodes and task execution nodes from 10 times per second to 5 times per second, configure the identification information of the backup dependency nodes for highly vulnerable nodes, and set the monitoring threshold for dependency stability. All these parameters are then categorized and organized according to node type and dependency type to form a specific set of parameters that can guide the adjustment of the non-steady-state architecture, i.e., the cluster architecture restructuring parameters.

[0076] In summary, in this embodiment of the invention, time-series pattern mining of behavioral consistency logs to obtain the behavioral evolution trajectory features of the cluster can capture the changing patterns and trends of the cluster's collaborative behavior at different stages from the time dimension of the logs, such as fluctuations in behavioral consistency and the temporal patterns of anomalies. This not only meets the core requirement of non-steady-state architectures to dynamically track behavioral changes, but also provides a basis for predicting behavioral trends in unmanned cluster rescue scenarios. This allows subsequent architecture analysis to no longer be limited to static data, but to be based on dynamic evolution patterns, laying a precise trajectory foundation for architecture vulnerability assessment.

[0077] In summary, vulnerability assessment of key nodes in non-steady-state architectures based on behavioral evolution trajectory characteristics yields a node vulnerability map. By combining behavioral anomalies and weak links in coordination reflected in the trajectory characteristics, the vulnerability map can accurately locate nodes in the architecture that play a crucial role in overall coordination but are prone to failure or have insufficient adaptability. For example, in rescue scenarios, this can be the architecture node corresponding to the unmanned platform responsible for core communication. This not only solves the problem of difficult identification of risks of key nodes in non-steady-state architectures but also meets the reliability requirements of architecture nodes in unmanned swarm rescue. The map clearly presents the distribution of vulnerable nodes, providing a clear target for subsequent dependency optimization.

[0078] In summary, by performing dependency strength analysis on non-steady-state architectures based on node vulnerability maps to derive dependency reconstruction strategies, we can analyze the data transmission and task execution dependencies among nodes in the architecture based on the distribution of vulnerable nodes. This allows us to determine which dependencies are high-risk due to the presence of vulnerable nodes and which dependencies are insufficient in strength to affect collaborative efficiency. Targeted reconstruction strategies can then be formulated, such as weakening high-risk dependencies and strengthening key collaborative dependencies. This approach not only meets the need for dynamic adjustment of dependencies in non-steady-state architectures but also supports the resilience of the architecture in unmanned cluster rescue scenarios to cope with sudden failures, preventing the overall collaborative collapse due to the vulnerability of a single node.

[0079] In summary, generating cluster architecture refactoring parameters based on dependency refactoring strategies can transform abstract refactoring strategies into concrete, executable parameters, such as node connection weights and data interaction frequency thresholds. These parameters can be directly used to adjust and optimize non-steady-state architectures, which not only meets the ultimate goal of improving cluster control efficiency through architecture refactoring, but also provides a quantitative basis for the actual adjustment of architecture in unmanned cluster rescue scenarios, ensuring that the refactored architecture can better adapt to dynamic rescue tasks and improve the stability and efficiency of cluster collaboration.

[0080] S6. Based on the architecture reconstruction parameters, perform multi-objective collaborative optimization on the control strategy to obtain the final control strategy of the cluster.

[0081] In this embodiment of the invention, the step of performing multi-objective collaborative optimization of the control strategy based on the architecture refactoring parameters to obtain the final control strategy of the cluster includes: The architecture reconstruction parameters are optimized using multiple objectives to obtain the set of optimization objectives for the cluster. Based on the set of optimization objectives, the feasibility and interrelationships of the control strategies are determined, and the set of adaptive strategies for the cluster is obtained. Pareto front analysis is performed on the adaptive policy set to obtain the non-dominated set of the adaptive policy set; Based on preset cluster collaboration criteria, collaborative optimization is performed on the non-dominated set to obtain the final control strategy of the cluster.

[0082] Specifically, the architecture refactoring parameters are first comprehensively analyzed to clarify the various adjustment requirements and specific values, such as the target adjustment value of node dependency strength, the configuration information of backup dependent nodes, and the monitoring threshold of dependency stability. Then, the core objectives that need to be optimized through control strategies are extracted from these parameters, including reducing the dependence of highly vulnerable nodes on core functional nodes, improving the stability of communication between cluster members to reduce collaboration anomalies, ensuring that the cluster task execution efficiency is not lower than the preset standard, and balancing the load of each node to avoid overload of a single node. These clear objectives are then compiled and summarized to form a set of optimization objectives for the cluster.

[0083] Furthermore, we first examine the various functional modules included in the current cluster control strategy, such as the task allocation module, path planning module, communication scheduling module, and exception response module. We then check whether each module can meet the requirements of the optimization objective set. For example, we determine whether the task allocation module has the ability to adjust the allocation rules based on node dependencies. If it does, the strategy corresponding to that module is deemed feasible; otherwise, it is marked as needing adjustment. Next, we analyze the interrelationships between the control strategies corresponding to different optimization objectives. For example, we determine whether strategies to improve communication stability conflict with strategies to ensure task execution efficiency. We then develop coordination solutions for conflicts and integrate all feasible and coordinated control strategy modules to obtain the cluster's adaptive strategy set.

[0084] Furthermore, firstly, quantitative evaluation indicators are set for each objective in the optimization target set. For example, "reducing dependence on highly vulnerable nodes" corresponds to the indicator of "interaction frequency between highly vulnerable nodes and core nodes," "improving communication stability" corresponds to the indicator of "communication anomaly rate," and "ensuring task efficiency" corresponds to the indicator of "task completion time." Then, for each strategy in the adaptive strategy set, its performance value on all quantitative indicators is calculated. For example, the interaction frequency of highly vulnerable nodes for strategy 1 is 5 times / hour, the communication anomaly rate is 2%, and the task completion time is 30 minutes. The corresponding values ​​for strategy 2 are 6 times / hour, 1%, and 28 minutes. Then, the indicator performance of every two strategies is compared. If strategy A performs no worse than strategy B on all indicators, and at least one indicator performs better than strategy B, then strategy B is determined to be dominated by strategy A, and the dominated strategy is removed from the set. This comparison process is repeated until there are no more dominated strategies in the set. The remaining strategies constitute the non-dominated set of the adaptive strategy set.

[0085] Furthermore, the preset cluster coordination criteria include core task priority criteria, communication latency threshold criteria, node load balancing criteria, and anomaly response timeliness criteria. Each strategy in the non-dominated set is matched and verified against these criteria one by one. For example, it checks whether the strategy can ensure that the execution priority of core tasks is the highest, whether the communication latency is controlled within the threshold, whether the node load is balanced, and whether the anomaly response meets the timeliness requirements. Strategies that fully comply with all preset criteria are selected. If there are multiple strategies that comply with the criteria, the performance of each strategy on the core indicators is further compared, and the strategy with the best performance is selected as the final control strategy of the cluster.

[0086] In summary, in the embodiments of the present invention, processing the architecture reconfiguration parameters to obtain the set of optimization targets for the cluster allows for the identification of the core dimensions to be optimized for control strategies based on the non-steady-state architecture adjustment requirements reflected by the architecture reconfiguration parameters. This includes cluster collaboration efficiency and member behavior synchronization under non-steady-state architecture, as well as practical needs such as task completion timeliness and resource utilization in unmanned cluster rescue scenarios. This ensures that the optimization targets are closely linked to dynamic changes in the architecture and application scenarios, avoiding a disconnect between targets and architecture adaptation, and providing a precise direction for subsequent strategy optimization.

[0087] In summary, by determining the feasibility and interrelationships of control strategies based on the optimization target set to obtain an adaptive strategy set, feasible strategies that conform to the current architecture state can be selected by combining the dynamic characteristics of the non-steady-state architecture. The complementary or restrictive relationships between different strategies can also be sorted out. At the same time, the dynamic needs of the rescue scenario can be adapted to ensure that the strategies have the flexibility to adjust according to the architecture and scenario. This solves the problems of rigid control strategies and difficulty in adapting to non-steady-state and complex scenarios in the existing technology, and provides a diverse and feasible foundation for the subsequent selection of high-quality strategies.

[0088] In summary, performing Pareto front analysis on the adaptive strategy set to obtain the non-dominated set can find equilibrium solutions across multiple objective dimensions, eliminate inferior solutions that are comprehensively superior to other strategies, retain high-quality strategies with advantages in different objective dimensions, and avoid the disadvantages caused by single-objective optimization. This not only meets the core requirements of multi-objective collaborative optimization under non-steady-state architecture, but also satisfies the requirements of balancing multiple needs in rescue scenarios, thus narrowing down the range of efficient candidates for final strategy selection.

[0089] In summary, the method optimizes the non-dominated set based on preset cluster collaboration criteria to obtain the final control strategy. It emphasizes the overall collaboration and architectural stability of the cluster as core criteria, while incorporating specific criteria for unmanned cluster rescue scenarios. It selects the strategy that best fits the operational needs of the non-steady-state architecture and the actual rescue mission from the non-dominated set, ensuring that the final control strategy is both adaptable to dynamic changes in the architecture and can efficiently support practical applications. This achieves a balance between architectural adaptability and mission practicality, and ultimately improves the control efficiency of similar clusters based on non-steady-state architectures.

[0090] Example 2 like Figure 2 As shown in the figure, this embodiment also provides a functional block diagram of a similar cluster control system based on an unsteady-state architecture.

[0091] The non-steady-state architecture-based cluster control system 100 described in this embodiment can be installed in a storage medium. Depending on the functions implemented, the non-steady-state architecture-based cluster control system 100 may include a data analysis module 101, a task decomposition module 102, a path planning module 103, a behavior monitoring module 104, an architecture reconfiguration module 105, and a strategy optimization module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the storage medium's processor and perform a fixed function, stored in the storage medium's memory.

[0092] In this embodiment, the functions of each module / unit are as follows: The data analysis module 101 is used to perform dynamic consistency analysis on the multi-dimensional environmental sensing data of the cluster based on the non-steady-state architecture, and obtain the consistency index and member role allocation scheme of the cluster. The task decomposition module 102 is used to decompose the consistency index and the member role allocation scheme into tasks to obtain the optimized sub-task sequence of the cluster. The path planning module 103 is used to perform path collaborative planning on the consistency index and the member role allocation scheme to obtain the collaborative path of the cluster. The behavior monitoring module 104 is used to execute the collaborative path according to the control policy of the cluster, monitor the collaborative behavior of the cluster, and generate the behavior consistency log of the cluster. The architecture reconfiguration module 105 is used to perform predictive evolution analysis on the non-steady-state architecture based on the behavior consistency log, and obtain the architecture reconfiguration parameters of the cluster. The strategy optimization module 106 is used to perform multi-objective collaborative optimization of the control strategy based on the architecture reconstruction parameters to obtain the final control strategy of the cluster.

[0093] In the several embodiments provided by this invention, it should be understood that the disclosed storage medium, system, and method can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0094] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0095] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0096] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0097] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for controlling a cluster of the same type based on a non-steady state architecture, characterized in that, The method includes: S1. Based on the non-steady-state architecture, perform dynamic consistency analysis on the multi-dimensional environmental sensing data of the cluster to obtain the consistency index and member role allocation scheme of the cluster. S2. Decompose the consistency index and the member role allocation scheme into tasks to obtain the optimized sub-task sequence of the cluster; S3. Perform path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster; S4. Execute the collaborative path according to the control policy of the cluster, monitor the collaborative behavior of the cluster, and generate the behavior consistency log of the cluster; S5. Based on the behavior consistency log, perform predictive evolution analysis on the non-steady-state architecture to obtain the architecture reconstruction parameters of the cluster; S6. Based on the architecture reconstruction parameters, perform multi-objective collaborative optimization on the control strategy to obtain the final control strategy of the cluster.

2. The same type cluster control method based on a non-steady state architecture according to claim 1, wherein, The method, based on an unsteady architecture, performs dynamic consistency analysis on the multi-dimensional environmental sensor data of the cluster to obtain consistency indicators and member role allocation schemes for clusters of the same type, including: The environmental multidimensional sensing data is subjected to noise filtering and spatiotemporal alignment to obtain the purified environmental data of the cluster. Based on the aforementioned unsteady architecture, a multi-dimensional coordination analysis is performed on the purified environment data to obtain the real-time coordination data of the cluster. Consistency features are extracted from the real-time coordination data to obtain the consistency index of the cluster; According to the dynamic role mapping mechanism of the cluster, the consistency index is assigned to the member roles in the cluster to obtain the member role allocation scheme of the cluster.

3. The same type cluster control method based on a non-steady state architecture according to claim 1, wherein, The step of decomposing the consistency metric and the member role allocation scheme into an optimized sub-task sequence for the cluster includes: Based on the member role allocation scheme, a dynamic dependency graph between the cluster's control tasks and the member roles is constructed. Based on the consistency index, critical paths are identified in the dynamic dependency graph to obtain the optimized task flow of the cluster. Based on the optimized task flow, the overall task of the cluster is decomposed into atomic task units of the cluster. The atomic task units are conflict-resolved and dynamically prioritized to obtain the optimized subtask sequence of the cluster.

4. The control method for similar clusters based on an unsteady architecture as described in claim 1, characterized in that, The step of performing path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster includes: A topology analysis is performed on the member role allocation scheme to obtain the dynamic communication topology among the member roles; Based on the consistency index, collaborative conflict prediction is performed on the dynamic communication topology to obtain the set of potential conflict domains of the cluster. Gradient optimization is performed on the conflict paths in the set of potential conflict domains to obtain the priority paths of the cluster; The priority path is subjected to spatiotemporal consistency verification to obtain the collaborative path of the cluster.

5. The control method for similar clusters based on an unsteady architecture as described in claim 1, characterized in that, The control policy for the cluster executes the cooperative path, monitors the cooperative behavior of the cluster, and generates a behavior consistency log for the cluster, including: Collect multi-source behavioral data streams when the cluster executes the collaborative path; Standardized feature extraction is performed on the multi-source behavioral data stream to obtain the standardized behavioral feature vector of the cluster; The behavioral consistency metric of the cluster is calculated based on the standardized behavioral feature vector, wherein the calculation formula for the behavioral consistency metric is as follows: ； in, This represents a measure of the consistency of the behavior. Indicates the number of member roles. Indicates the first The standardized behavioral feature vectors of the aforementioned member roles, This represents the mean of all the standardized behavioral feature vectors. This represents the variance of the standardized behavioral feature vector. Indicates the first The variance between the standardized behavioral feature vector and the mean vector of each member role This indicates the preset adjustment parameters; Anomaly pattern identification is performed on the behavioral consistency metric to obtain the behavioral consistency log of the cluster.

6. The control method for similar clusters based on an unsteady architecture as described in claim 5, characterized in that, The step of standardizing feature extraction from the multi-source behavioral data stream to obtain the standardized behavioral feature vector of the cluster includes: Modal fusion is performed on the multi-source behavioral data streams to obtain the fused behavioral data of the cluster; Extract outliers from the fused behavioral data and perform integrity repair on the fused behavioral data to obtain the purified behavioral data of the cluster; Multi-scale feature extraction is performed on the purified behavioral data to obtain the behavioral feature set of the cluster; By applying dimensional constraints to the behavioral feature set, a standardized behavioral feature set for the cluster is obtained. Based on the standardized behavioral feature set, a standardized behavioral feature vector of the cluster is generated.

7. The control method for similar clusters based on an unsteady architecture as described in claim 1, characterized in that, Based on the behavioral consistency logs, predictive evolution analysis is performed on the non-steady-state architecture to obtain the cluster's architecture reconfiguration parameters, including: Time-series pattern mining is performed on the behavior consistency logs to obtain the behavioral evolution trajectory features of the cluster. Based on the behavioral evolution trajectory characteristics, the vulnerability of key nodes in the non-steady-state architecture is assessed, and a node vulnerability map of the non-steady-state architecture is obtained. Based on the node vulnerability map, a dependency strength analysis is performed on the non-steady-state architecture to obtain a dependency reconstruction strategy for the non-steady-state architecture. Based on the dependency restructuring strategy, the architecture restructuring parameters of the cluster are generated.

8. The control method for similar clusters based on an unsteady architecture as described in claim 1, characterized in that, The step of performing multi-objective collaborative optimization of the control strategy based on the architecture reconstruction parameters to obtain the final control strategy of the cluster includes: The architecture reconstruction parameters are optimized using multiple objectives to obtain the set of optimization objectives for the cluster. Based on the set of optimization objectives, the feasibility and interrelationships of the control strategies are determined, and the set of adaptive strategies for the cluster is obtained. Pareto front analysis is performed on the adaptive policy set to obtain the non-dominated set of the adaptive policy set; Based on preset cluster collaboration criteria, collaborative optimization is performed on the non-dominated set to obtain the final control strategy of the cluster.

9. A cluster control system of the same type based on an unsteady architecture, characterized in that, include: The data analysis module is used to perform dynamic consistency analysis on the multi-dimensional environmental sensor data of the cluster based on the non-steady-state architecture, and obtain the consistency index and member role allocation scheme of the cluster. The task decomposition module is used to decompose the consistency index and the member role allocation scheme into tasks to obtain the optimized subtask sequence of the cluster. The path planning module is used to perform path coordination planning on the consistency index and the member role allocation scheme to obtain the coordination path of the cluster. The behavior monitoring module is used to execute the collaborative path according to the control policy of the cluster, monitor the collaborative behavior of the cluster, and generate the behavior consistency log of the cluster. The architecture refactoring module is used to perform predictive evolution analysis on the non-steady-state architecture based on the behavior consistency log, and obtain the architecture refactoring parameters of the cluster. The strategy optimization module is used to perform multi-objective collaborative optimization of the control strategy based on the architecture reconstruction parameters to obtain the final control strategy of the cluster.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 8.

Images