IT Service Collaborative Scheduling Management System Based on Multi-Agent Systems

Abstract

This invention discloses an IT service collaborative scheduling and management system based on a multi-agent system, comprising: a data module for aligning task data, node state data, and resource data to form a state vector; a relationship module for parsing the state vector to generate a node set and dependencies, and identifying agents; a mirroring module for generating shadow policy variables and constructing a mirror policy space; a dual module for generating dual variables based on resource differences and establishing policy coupling relationships; a spectral constraint module for constructing a Jacobian matrix and filtering the policy set based on the spectral radius; a verification module for verifying policy execution consistency and constraints to generate a balanced policy set; and an output module for generating a scheduling sequence based on the policies. This invention achieves multi-agent collaborative scheduling and resource constraint consistency control, significantly improving scheduling accuracy and system stability.

Description

Technical Field

[0001] This invention relates to the field of multi-agent scheduling system technology, and in particular to an IT service collaborative scheduling and management system based on a multi-agent system. Background Technology

[0002] As information systems continue to expand in scale, the IT service operating environment has gradually evolved from a single node to a complex system structure composed of multiple nodes and resource units. The task scheduling process involves the coordinated allocation of computing resources, storage resources, and network resources. In this context, the increased number of scheduling objects, frequent changes in resource status, and complex task dependencies make traditional rule-based or static policy-based scheduling methods difficult to meet actual operational needs.

[0003] In existing technologies, most scheduling methods rely on preset strategies or local optimization mechanisms, lacking a unified modeling structure for resource utilization, task execution order, and node load distribution. This leads to uneven resource allocation, task execution conflicts, or local optima in scheduling results. Furthermore, in multi-node environments, the dependencies and resource constraints between different nodes exhibit dynamic changes. Existing methods typically characterize these relationships using simple mapping or fixed-weight processing, failing to reflect the complex coupling relationships in actual operation.

[0004] On the other hand, with the development of multi-agent methods, abstracting different nodes into independent decision-making units has become a trend. However, existing multi-agent scheduling methods lack a unified constraint mechanism for policy consistency, constraint satisfaction, and system stability during policy coordination, which easily leads to problems such as policy deviation and conflict accumulation. In addition, most technologies still use a single threshold judgment method for the selection and verification of scheduling policies, lacking the ability to comprehensively utilize the dynamic evolution process of policies and historical state information.

[0005] Therefore, how to build a unified scheduling mechanism that includes state modeling, policy mapping, constraint coupling, and stability screening in a multi-agent collaborative environment has become a technical problem that urgently needs to be solved in the current IT service scheduling field. Summary of the Invention

[0006] One objective of this invention is to propose an IT service collaborative scheduling and management system based on a multi-agent system. This invention fully integrates a multi-agent collaborative modeling mechanism, a policy variable mirror mapping structure, a dual variable coupling constraint mechanism, and a spectral radius determination method based on the Jacobian matrix. It describes in detail the entire process of IT service scheduling, from multi-source data alignment, state vector construction, node relationship resolution, policy generation and mapping, constraint coupling adjustment to stability screening and consistency verification. It has the advantages of clear scheduling structure expression, accurate resource constraint characterization, controllable policy evolution process, and strong multi-node collaborative consistency.

[0007] An IT service collaborative scheduling and management system based on a multi-agent system according to an embodiment of the present invention includes: The data module is used to acquire task data, node status data, and resource data, perform alignment at index positions, write values ​​at feature positions, and form a status vector. The relation module is used to parse the state vector, generate a set of nodes at the node positions, write the dependency relationship at the node pair positions, and generate the agent identifier at the index position. The mirror module is used to construct policy variables. It copies and generates shadow policy variables at the agent's identifier position, establishes a mapping relationship at the corresponding position, and forms a mirror policy space. The dual module is used to extract resource capacity and occupancy values ​​at the resource data index location, calculate the difference at the corresponding location, generate constraint data at the difference location, generate dual variables at the policy variable location, establish coupling relationship between the policy location and the dual variable location, and perform calculations at the coupling location to form a policy mapping function. The spectral constraint module is used to perform a linear expansion of the policy mapping function, generate the Jacobian matrix at the derivative position of the policy mapping function, generate the spectral radius at the eigenvalue position, and perform a decision at the spectral radius position to filter the policy set. The verification module is used to perform consistency verification at the index position of the policy variable, calculate the difference between the corresponding positions of the policy variable and the shadow policy variable, perform constraint verification at the constraint position, and form a set of balanced policies. The output module is used to read policy values, perform mapping calculations between task and resource locations, and generate scheduling sequences.

[0008] Optionally, the data module includes the following steps: Receive the task data set at the task index position, the node status data set at the node index position, and the resource data set at the resource index position. Establish a time index sequence at the unified index position. Write the task identifier value and task requirement value in the task data set into the task feature position at the corresponding position of the time index sequence. Write the node load value and node availability value in the node status data set into the node feature position. Write the resource capacity value and resource usage value in the resource data set into the resource feature position. A feature dimension number sequence is established at the feature dimension index position. Dimension alignment calculations are performed on the task feature position, node feature position, and resource feature position at the corresponding positions of the feature dimension number sequence. An aligned feature value sequence is generated at the unified feature dimension position. A concatenation calculation is performed on the task feature value sequence, node feature value sequence, and resource feature value sequence at the unified feature dimension position. A joint feature value sequence is generated at the concatenation position. A state vector index sequence is established at the corresponding position of the joint feature value sequence. The joint feature value sequence is written at the state vector index position to generate a state vector.

[0009] Optionally, the writing of dependency relationships at node pair positions in the relationship module specifically includes: Extract the numerical sequences of task requirements, node load, resource capacity, and resource usage at the state vector index positions. Establish a node identifier sequence at the node index positions and establish a node pair combination sequence at the corresponding positions of the node identifier sequences. At the corresponding position of the node pair combination sequence, calculate the difference between the resource capacity value and the resource usage value for each node identifier pair, generate a resource difference value at the difference position, calculate the ratio between the node load value and the task requirement value, generate a load ratio value at the ratio calculation position, perform a weighted calculation between the resource difference value position and the load ratio value position, generate a dependency value at the weighted calculation position, and write the dependency value at the corresponding position of the node pair combination sequence.

[0010] Optionally, the mirror module specifically includes: Extract the agent identifier sequence at the node index position, extract the state vector value sequence at the state vector index position, establish an index mapping relationship between the node index position and the state vector index position, establish a mapping number sequence at the state vector index position, perform mapping calculation on the state vector value sequence at the corresponding position of the mapping number sequence, generate the mapping value at the corresponding position, and write the mapping value at the corresponding position of the mapping number sequence to form the original mapping value sequence. Establish a mirror index sequence at the corresponding position of the mapping number sequence, perform copy calculation on the mapping value at the corresponding position of the mirror index sequence, generate a shadow mapping value at the corresponding position, establish a one-to-one correspondence between the mapping value position and the shadow mapping value position, calculate the difference at the corresponding position, and generate an offset value at the difference position. The mapping value position and the offset value position are combined for calculation, and a mirror mapping value is generated at the combined position. The mirror mapping value is written at the corresponding position of the mirror index sequence to form a mirror strategy space.

[0011] Optionally, the dual module's generation of dual variables at the policy location specifically includes: Extract the resource capacity value sequence and resource occupancy value sequence at the resource data index position, calculate the difference between the resource capacity value and the resource occupancy value at the resource data index position, generate the resource difference value sequence at the difference position, extract the state vector value sequence at the state vector index position, extract the agent identifier sequence at the node index position, establish an index mapping relationship between the state vector index position and the resource index position, establish a policy index sequence at the state vector index position, and establish a constraint mapping position at the corresponding position of the policy index sequence. At the corresponding position of the policy index sequence, perform mapping calculation on the resource difference value sequence, generate constraint mapping value at the constraint mapping position, perform product calculation on the state vector value sequence at the constraint mapping value position, generate first coupling value at the product position, perform index matching calculation on the agent identifier sequence at the constraint mapping value position, generate matching weight value at the matching position, perform combination calculation on the first coupling value position and the matching weight value position, generate second coupling value at the combination position, perform normalization calculation on the second coupling value position, and generate intermediate dual value at the normalization position. Establish a dual variable number sequence at the corresponding position of the strategy index sequence. Perform mapping and writing calculation on the intermediate dual values ​​at the corresponding positions of the dual variable number sequence. Generate dual variable values ​​at the corresponding positions. Establish dual variable writing positions at the corresponding positions of the dual variable number sequence. Write dual variable values ​​at the dual variable writing positions. Form a dual variable sequence at the corresponding positions of the strategy index sequence. Establish a one-to-one correspondence between the dual variable value positions and the constraint mapping value positions. Perform difference correction calculation on the dual variable values ​​and constraint mapping values ​​at the corresponding positions. Generate updated dual variable values ​​at the correction positions. Write updated dual variable values ​​at the corresponding positions of the dual variable number sequence to form a dual variable sequence.

[0012] Optionally, the establishment of a coupling relationship between the policy position and the dual variable position in the dual module, and the execution of computation at the coupling position to form the policy mapping function, specifically includes: Read the numerical sequence of the strategy variable at the position of the strategy variable, read the numerical sequence of the dual variable at the position of the dual variable, perform position alignment operation on the position of the strategy variable and the position of the dual variable under the unified indexing rule, and establish a one-to-one coupled index relationship at the aligned position. Write the constraint weight identifier corresponding to the constraint data at the coupling index position, extract the strategy variable value at the strategy variable position, extract the dual variable value at the dual variable position, and perform product calculation at the same coupling index position to generate a sequence of coupled component values. Perform a weighted accumulation operation at the corresponding position of the coupled component numerical sequence, generate the coupled result value at the accumulation position, introduce the coupled result value and the original strategy variable value at the strategy variable position to perform a combination calculation, and generate an updated strategy variable numerical sequence. Establish a function parameter record structure at the corresponding position of the updated strategy variable value sequence, write the updated strategy variable value at the parameter record position, establish a mapping relationship between the input index and the output index at the parameter record position, and perform function generation operation at the mapping relationship position to form a strategy mapping function expression structure; Write the input policy variable index, dual variable index, and constraint data index at the corresponding positions in the policy mapping function expression structure to form a policy mapping function structure containing the input index set and the output index set.

[0013] Optionally, generating the Jacobian matrix at the derivative position of the policy mapping function in the spectral constraint module specifically includes: Extract the numerical sequence of the strategy mapping function at the corresponding position of the strategy index sequence, establish the derivative index sequence at the corresponding position of the strategy index sequence, establish the variable number sequence at the corresponding position of the derivative index sequence, expand the numerical sequence of the strategy mapping function according to the variable dimension at the corresponding position of the variable number sequence to generate the numerical sequence of function components, and establish the index sequence of function components at the corresponding position of the numerical sequence of function components. At the corresponding position of the function component index sequence, perform difference calculation on the value of each function component along the variable number sequence, generate partial derivative values ​​at the difference position, establish a derivative matrix index structure at the corresponding position of the variable number sequence, use the function component index sequence as the row index and the variable number sequence as the column index in the derivative matrix index structure, write the partial derivative values ​​at the corresponding row and column intersection position to form the derivative matrix value structure. A matrix number sequence is established at the corresponding position of the derivative matrix index structure. A matrix rearrangement calculation is performed on the derivative matrix numerical structure at the corresponding position of the matrix number sequence. A Jacobi matrix value is generated at the rearranged position. The Jacobi matrix value is written at the corresponding position of the matrix number sequence to form the Jacobi matrix.

[0014] Optionally, the spectral constraint module generates a spectral radius at the eigenvalue location and performs a determination at the spectral radius location. The specific set of filtering strategies includes: Extract the Jacobian matrix numerical sequence at the corresponding position of the matrix number sequence, establish the eigenvalue index sequence at the corresponding position of the matrix number sequence, perform eigenvalue decomposition calculation on the Jacobian matrix numerical sequence at the corresponding position of the eigenvalue index sequence, generate the eigenvalue numerical sequence at the corresponding position, perform modulus calculation on the eigenvalue numerical sequence at the corresponding position of the eigenvalue index sequence, generate the eigenvalue modulus sequence at the corresponding position, perform maximum value calculation on the eigenvalue modulus sequence at the corresponding position of the eigenvalue index sequence, and generate the spectral radius value at the maximum value position; Establish a spectral radius mapping position at the corresponding position of the strategy index sequence, map the spectral radius value to the corresponding strategy position at the spectral radius mapping position, establish a decision index sequence at the corresponding position of the strategy index sequence, perform a comparison calculation between the spectral radius value and the preset threshold value at the corresponding position of the decision index sequence, write a filter identifier at the position that meets the comparison condition, extract the corresponding strategy value at the filter identifier position, and generate a filtered strategy set at the corresponding position of the strategy index sequence.

[0015] Optionally, the constraint verification performed at the constraint location in the verification module to form a set of equilibrium strategies specifically includes: Extract the filtered strategy set at the corresponding position of the strategy index sequence, extract the resource capacity value sequence and resource usage value sequence at the resource index position, perform difference calculation on the resource capacity value and resource usage value at the resource index position, generate the resource difference value sequence at the difference position, establish the constraint mapping position at the corresponding position of the strategy index sequence, map the resource difference value sequence to the corresponding strategy position at the constraint mapping position, and establish the constraint verification index sequence at the corresponding position of the strategy index sequence. At the corresponding position in the constraint verification index sequence, perform difference calculation on the strategy value and the constraint mapping value, generate constraint deviation value at the difference position, perform cumulative calculation on the constraint deviation value corresponding to each strategy at the constraint deviation value position, generate constraint cumulative value at the cumulative position, perform comparison calculation on the constraint cumulative value position, write a valid identifier at the position that meets the comparison condition, extract the corresponding strategy value at the valid identifier position, and generate a set of balanced strategies at the corresponding position in the strategy index sequence.

[0016] The beneficial effects of this invention are: (1) This invention constructs a hierarchical linkage structure among data modules, relationship modules, mirror modules, dual modules, spectral constraint modules, and verification modules. Under a unified indexing system, it achieves collaborative processing of task data, node state data, and resource data, enabling state vectors, node relationships, and policy variables to be expressed and transmitted within the same structural space. By introducing a mapping structure between policy variables and shadow policy variables, as well as a coupling computation mechanism between dual variables and constraint data, resource constraint relationships in the scheduling process can be directly embedded into the computation path during the policy generation stage. This allows resource capacity and occupancy differences to be reflected synchronously during policy formation, reducing the structural disconnect between policy generation and constraint verification.

[0017] (2) This invention constructs a Jacobian matrix based on the policy mapping function, performs screening and judgment based on the spectral radius generated by eigenvalues, and combines constraint deviation recursive calculation and consistency verification mechanism to perform multi-level screening processing on the policy set, so that the scheduling result has structural stability and consistency while satisfying resource constraints. By introducing constraint deviation accumulation and mapping relationship in the verification process, the policy maintains a dynamic matching relationship with the resource state during the evolution process, avoiding local conflicts or policy deviation problems, thereby forming a scheduling sequence output structure with continuity and coordination. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 The flowchart below shows the modules of the IT service collaborative scheduling and management system based on a multi-agent system proposed in this invention. Figure 2 This is a schematic diagram of the mirror strategy space of the IT service collaborative scheduling and management system based on a multi-agent system proposed in this invention.

[0019] Figure 3 This is a schematic diagram of the duality and spectral constraints of the IT service collaborative scheduling and management system based on a multi-agent system proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figures 1-3 An IT service collaborative scheduling and management system based on a multi-agent system includes: The data module is used to acquire task data, node status data, and resource data, perform alignment at index positions, write values ​​at feature positions, and form a status vector. The relation module is used to parse the state vector, generate a set of nodes at the node positions, write the dependency relationship at the node pair positions, and generate the agent identifier at the index position. The mirror module is used to construct policy variables. It copies and generates shadow policy variables at the agent's identifier position, establishes a mapping relationship at the corresponding position, and forms a mirror policy space. The dual module is used to extract resource capacity and occupancy values ​​at the resource data index location, calculate the difference at the corresponding location, generate constraint data at the difference location, generate dual variables at the policy variable location, establish coupling relationship between the policy location and the dual variable location, and perform calculations at the coupling location to form a policy mapping function. The spectral constraint module is used to perform a linear expansion of the policy mapping function, generate the Jacobian matrix at the derivative position of the policy mapping function, generate the spectral radius at the eigenvalue position, and perform a decision at the spectral radius position to filter the policy set. The verification module is used to perform consistency verification at the index position of the policy variable, calculate the difference between the corresponding positions of the policy variable and the shadow policy variable, perform constraint verification at the constraint position, and form a set of balanced policies. The output module is used to read policy values, perform mapping calculations between task and resource locations, and generate scheduling sequences.

[0022] In this embodiment, the data module includes the following steps: Receive the task data set at the task index position, the node status data set at the node index position, and the resource data set at the resource index position. Establish a time index sequence at the unified index position. Write the task identifier value and task requirement value in the task data set into the task feature position at the corresponding position of the time index sequence. Write the node load value and node availability value in the node status data set into the node feature position. Write the resource capacity value and resource usage value in the resource data set into the resource feature position. A feature dimension number sequence is established at the feature dimension index position. Dimension alignment calculations are performed on the task feature position, node feature position, and resource feature position at the corresponding positions of the feature dimension number sequence. An aligned feature value sequence is generated at the unified feature dimension position. A concatenation calculation is performed on the task feature value sequence, node feature value sequence, and resource feature value sequence at the unified feature dimension position. A joint feature value sequence is generated at the concatenation position. A state vector index sequence is established at the corresponding position of the joint feature value sequence. The joint feature value sequence is written at the state vector index position to generate a state vector.

[0023] In this embodiment, performing dimension alignment calculation specifically includes: In the feature dimension numbering sequence, a unified dimension number is assigned to the task feature position, node feature position, and resource feature position respectively. At the same dimension number position, feature values ​​from different sources are rearranged and filled in, so that various feature values ​​form a corresponding relationship under a unified dimension coordinate.

[0024] In this embodiment, writing dependency relationships at node pair positions in the relationship module specifically includes: Extract the numerical sequences of task requirements, node load, resource capacity, and resource usage at the state vector index positions. Establish a node identifier sequence at the node index positions and establish a node pair combination sequence at the corresponding positions of the node identifier sequences. At the corresponding position of the node pair combination sequence, calculate the difference between the resource capacity value and the resource usage value for each node identifier pair, generate a resource difference value at the difference position, calculate the ratio between the node load value and the task requirement value, generate a load ratio value at the ratio calculation position, perform a weighted calculation between the resource difference value position and the load ratio value position, generate a dependency value at the weighted calculation position, and write the dependency value at the corresponding position of the node pair combination sequence.

[0025] In this embodiment, the mirroring module specifically includes: Extract the agent identifier sequence at the node index position, extract the state vector value sequence at the state vector index position, establish an index mapping relationship between the node index position and the state vector index position, establish a mapping number sequence at the state vector index position, perform mapping calculation on the state vector value sequence at the corresponding position of the mapping number sequence, generate the mapping value at the corresponding position, and write the mapping value at the corresponding position of the mapping number sequence to form the original mapping value sequence. Establish a mirror index sequence at the corresponding position of the mapping number sequence, perform copy calculation on the mapping value at the corresponding position of the mirror index sequence, generate a shadow mapping value at the corresponding position, establish a one-to-one correspondence between the mapping value position and the shadow mapping value position, calculate the difference at the corresponding position, and generate an offset value at the difference position. The mapping value position and the offset value position are combined for calculation, and a mirror mapping value is generated at the combined position. The mirror mapping value is written at the corresponding position of the mirror index sequence to form a mirror strategy space.

[0026] In this embodiment, the dual module generates dual variables at the policy location, specifically including: Extract the resource capacity value sequence and resource occupancy value sequence at the resource data index position, calculate the difference between the resource capacity value and the resource occupancy value at the resource data index position, generate the resource difference value sequence at the difference position, extract the state vector value sequence at the state vector index position, extract the agent identifier sequence at the node index position, establish an index mapping relationship between the state vector index position and the resource index position, establish a policy index sequence at the state vector index position, and establish a constraint mapping position at the corresponding position of the policy index sequence. At the corresponding position of the policy index sequence, perform mapping calculation on the resource difference value sequence, generate constraint mapping value at the constraint mapping position, perform product calculation on the state vector value sequence at the constraint mapping value position, generate first coupling value at the product position, perform index matching calculation on the agent identifier sequence at the constraint mapping value position, generate matching weight value at the matching position, perform combination calculation on the first coupling value position and the matching weight value position, generate second coupling value at the combination position, perform normalization calculation on the second coupling value position, and generate intermediate dual value at the normalization position. Establish a dual variable number sequence at the corresponding position of the strategy index sequence. Perform mapping and writing calculation on the intermediate dual values ​​at the corresponding positions of the dual variable number sequence. Generate dual variable values ​​at the corresponding positions. Establish dual variable writing positions at the corresponding positions of the dual variable number sequence. Write dual variable values ​​at the dual variable writing positions. Form a dual variable sequence at the corresponding positions of the strategy index sequence. Establish a one-to-one correspondence between the dual variable value positions and the constraint mapping value positions. Perform difference correction calculation on the dual variable values ​​and constraint mapping values ​​at the corresponding positions. Generate updated dual variable values ​​at the correction positions. Write updated dual variable values ​​at the corresponding positions of the dual variable number sequence to form a dual variable sequence.

[0027] In this embodiment, the first coupling value and the second coupling value are divided based on the following: by dividing the coupling process into a resource constraint coupling stage and a weight modulation coupling stage, the hierarchical fusion of multi-source information is realized under a unified index structure, so that constraint information and structural information act on policy variables in different computing levels, forming a coupling computing mechanism with a hierarchical structure.

[0028] In this embodiment, performing difference correction calculations on the dual variable values ​​and constraint mapping values ​​at corresponding positions, and generating updated dual variable values ​​at the correction positions specifically includes: At the corresponding position of the dual variable number sequence, perform difference calculation on the dual variable value and the constraint mapping value, generate constraint deviation value at the difference position, establish update coefficient identifier at the constraint deviation value position, write preset weight coefficient value at the update coefficient identifier position, perform product calculation on the constraint deviation value and weight coefficient value at the constraint deviation value position, and generate proportional correction value at the product position. At the corresponding position in the dual variable number sequence, the dual variable value and the proportional correction value are combined proportionally to calculate the value, and the updated dual variable value is generated at the corresponding position.

[0029] In this embodiment, the establishment of a coupling relationship between the policy position and the dual variable position in the dual module, and the execution of calculations at the coupling position to form the policy mapping function specifically includes: Read the numerical sequence of the strategy variable at the position of the strategy variable, read the numerical sequence of the dual variable at the position of the dual variable, perform position alignment operation on the position of the strategy variable and the position of the dual variable under the unified indexing rule, and establish a one-to-one coupled index relationship at the aligned position. Write the constraint weight identifier corresponding to the constraint data at the coupling index position, extract the strategy variable value at the strategy variable position, extract the dual variable value at the dual variable position, and perform product calculation at the same coupling index position to generate a sequence of coupled component values. Perform a weighted accumulation operation at the corresponding position of the coupled component numerical sequence, generate the coupled result value at the accumulation position, introduce the coupled result value and the original strategy variable value at the strategy variable position to perform a combination calculation, and generate an updated strategy variable numerical sequence. Establish a strategy mapping function parameter record structure at the corresponding position of the updated strategy variable value sequence, write the updated strategy variable value at the parameter record position, establish a mapping relationship between the input index and the output index at the parameter record position, and perform function generation operation at the mapping relationship position to form the strategy mapping function expression structure; Write the input policy variable index, dual variable index, and constraint data index at the corresponding positions in the policy mapping function expression structure to form a policy mapping function structure containing the input index set and the output index set.

[0030] In this embodiment, establishing a one-to-one corresponding coupling index relationship at the alignment position specifically involves: Under the unified indexing rules, the index sequence of strategy variables and the index sequence of dual variables are aligned according to the index number. A correspondence between variables is established at the same index number position, so that each strategy variable value and its corresponding dual variable value form a one-to-one mapping relationship.

[0031] In this embodiment, performing function generation operations at the mapping relationship location specifically includes: Extract the update strategy variable value sequence from the parameter record position, establish a parameter number sequence in the parameter record position, and write the update strategy variable value sequence into the parameter number position in index order to form the parameter value sequence. Establish an input index sequence at the input index position and an output index sequence at the output index position. Establish a connection relationship between the input position and the output position based on the input index sequence and the output index sequence at the mapping relationship position, and write the corresponding parameter value into the connection relationship position. Perform a product calculation on the input value and parameter value corresponding to the input index position at the connection relationship position, generate intermediate calculated values ​​at the product position, perform an accumulation calculation on each intermediate calculated value at the output index position, and generate an output value sequence at the accumulation position. Establish a function expression record structure at the position corresponding to the output value sequence, and write the input index sequence, parameter value sequence, and output value sequence into the function expression record structure to form the strategy mapping function expression structure.

[0032] In this embodiment, generating the Jacobian matrix at the derivative position of the policy mapping function in the spectral constraint module specifically includes: Extract the numerical sequence of the strategy mapping function at the corresponding position of the strategy index sequence, establish the derivative index sequence at the corresponding position of the strategy index sequence, the derivative index sequence is used to identify the index sequence of the partial derivative calculation position of each variable in the strategy mapping function, establish the variable number sequence at the corresponding position of the derivative index sequence, expand the numerical sequence of the strategy mapping function according to the variable dimension at the corresponding position of the variable number sequence to generate the numerical sequence of the strategy mapping function components, establish the index sequence of the strategy mapping function components at the corresponding position of the numerical sequence of the strategy mapping function components, the function component index sequence is used to identify the position number of each function component in the sequence; At the corresponding position of the function component index sequence, perform difference calculation on the value of each function component along the variable number sequence, generate partial derivative values ​​at the difference position, establish a derivative matrix index structure at the corresponding position of the variable number sequence, use the function component index sequence as the row index and the variable number sequence as the column index in the derivative matrix index structure, write the partial derivative values ​​at the corresponding row and column intersection position to form the derivative matrix value structure. A matrix number sequence is established at the corresponding position of the derivative matrix index structure. A matrix rearrangement calculation is performed on the derivative matrix numerical structure at the corresponding position of the matrix number sequence. A Jacobi matrix value is generated at the rearranged position. The Jacobi matrix value is written at the corresponding position of the matrix number sequence to form the Jacobi matrix.

[0033] In this embodiment, performing matrix rearrangement specifically includes: The matrix row indices are sorted according to the function component index sequence, and the matrix column indices are sorted according to the variable number sequence. The corresponding partial derivative values ​​are written into the sorted positions to generate a Jacobian matrix with consistent index order.

[0034] In this embodiment, the spectral constraint module generates a spectral radius at the eigenvalue location and performs a determination at the spectral radius location. The specific set of filtering strategies includes: Extract the Jacobian matrix numerical sequence at the corresponding position of the matrix number sequence, establish the eigenvalue index sequence at the corresponding position of the matrix number sequence, perform eigenvalue decomposition calculation on the Jacobian matrix numerical sequence at the corresponding position of the eigenvalue index sequence, generate the eigenvalue numerical sequence at the corresponding position, perform modulus calculation on the eigenvalue numerical sequence at the corresponding position of the eigenvalue index sequence, generate the eigenvalue modulus sequence at the corresponding position, perform maximum value calculation on the eigenvalue modulus sequence at the corresponding position of the eigenvalue index sequence, and generate the spectral radius value at the maximum value position; Establish a spectral radius mapping position at the corresponding position of the strategy index sequence, map the spectral radius value to the corresponding strategy position at the spectral radius mapping position, establish a decision index sequence at the corresponding position of the strategy index sequence, perform a comparison calculation between the spectral radius value and the preset threshold value at the corresponding position of the decision index sequence, write a filter identifier at the position that meets the comparison condition, extract the corresponding strategy value at the filter identifier position, and generate a filtered strategy set at the corresponding position of the strategy index sequence.

[0035] In this embodiment, the constraint verification performed at the constraint location in the verification module to form a set of equilibrium strategies specifically includes: Extract the filtered set of strategies at the corresponding position of the strategy index sequence, extract the resource capacity value sequence and resource usage value sequence at the resource index position, perform difference calculation on the resource capacity value and resource usage value at the resource index position, generate the resource difference value sequence at the difference position, establish the constraint mapping position at the corresponding position of the strategy index sequence, map the resource difference value sequence to the corresponding strategy position at the constraint mapping position, and establish the constraint verification index sequence at the corresponding position of the strategy index sequence. The constraint verification index sequence is used to identify the corresponding verification relationship between the strategy value and the constraint mapping value. At the corresponding position in the constraint verification index sequence, perform difference calculation on the strategy value and the constraint mapping value, generate constraint deviation value at the difference position, perform cumulative calculation on the constraint deviation value corresponding to each strategy at the constraint deviation value position, generate constraint cumulative value at the cumulative position, perform comparison calculation on the constraint cumulative value at the constraint cumulative value position, write a valid identifier at the position that meets the comparison condition, extract the corresponding strategy value at the valid identifier position, and generate a balanced strategy set at the corresponding position in the strategy index sequence, which is the strategy set composed of the strategy values ​​corresponding to the strategy index positions that meet the constraint verification conditions.

[0036] Example 1: To verify the feasibility of this invention in practice, it was applied to a large-scale IT service environment. This environment simultaneously contains numerous task requests, multiple computing nodes, and various resource units. Task processing requires resource allocation and node scheduling. Under traditional scheduling methods, frequent changes in task requirements, significant fluctuations in node load, and dynamic changes in resource occupancy easily lead to scheduling conflicts, uneven resource allocation, and accumulated execution delays, resulting in decreased overall operating efficiency. In this environment, the data module first performs unified indexing and alignment processing on task data, node status data, and resource data, mapping task requirement values, node load values, and resource capacity and occupancy values ​​to a unified feature dimension, forming a state vector. Subsequently, the relationship module parses the state vector, establishing node sets at node locations and generating dependency relationship values ​​at node pairs, creating a correlation structure between task requirements and node load. Based on this, the mirroring module performs replication and mapping processing on policy variables, generating shadow policy variables at corresponding locations. Through difference calculation and combination calculation, a mirrored policy space is formed, giving the policy a dual-expression structure. Furthermore, the dual module calculates the difference between resource capacity and occupancy values ​​to generate resource difference values. This difference is then combined with the state vector to perform product calculations and weighted combination calculations to generate dual variables, thus introducing resource constraint information. After establishing a coupling relationship between the policy variables and dual variables, a policy mapping function expression structure is formed. The spectral constraint module performs derivative calculations on the policy mapping function to generate a Jacobian matrix and performs modulus calculations on the eigenvalues ​​to generate a spectral radius. The spectral radius is used to determine and filter a stable policy set. In the verification phase, the difference between the constraint mapping and the policy values ​​is calculated to generate constraint deviation values. Accumulation and comparison operations are performed on the deviations to filter policies that meet the constraints, forming a balanced policy set. Finally, the output module generates a scheduling sequence. In this implementation, the operating effects of traditional scheduling methods and the method of this invention are compared, and the key indicators are recorded in the table below.

[0037] Table 1: Scheduling Performance Comparison Table

[0038] The data in Table 1 further demonstrates the significant differences between the proposed method and traditional scheduling methods under different task scales. When the number of tasks is 120, the average scheduling latency of the traditional method is 85ms, while the proposed method reduces it to 52ms, a decrease of 33ms, or approximately 38.8%. When the number of tasks increases to 180, the traditional method's latency is 96ms, while the proposed method's is 60ms, a decrease of 36ms, or approximately 37.5%. When the number of tasks reaches 240, the traditional method's latency is 110ms, while the proposed method's is 68ms, a decrease of 42ms, or 38.2%, indicating that the proposed method maintains a stable performance advantage even as the task scale increases. Regarding resource utilization, the traditional method achieves 72%, 75%, and 78% respectively, while the proposed method achieves 86%, 88%, and 90%, representing improvements of 14%, 13%, and 12%, respectively. This indicates that the proposed method can more fully utilize resource capacity and reduce resource idle time during resource allocation. Regarding the number of conflicts, the traditional method reaches 30 conflicts when the number of tasks is 240, while the present invention controls it to within 10, reducing it by 20, a decrease of approximately 66.7%. This demonstrates that the present invention can effectively reduce resource contention conflicts during multi-node scheduling. In terms of the proportion of stable strategies, the traditional method achieves 61%, 63%, and 65% respectively, while the present invention improves these proportions to 82%, 85%, and 87%, representing increases of 21%, 22%, and 22% respectively. This indicates that through spectral constraints and constraint verification mechanisms, the present invention can select a set of strategies that better meet the system's operational requirements. Overall, considering all indicators, as the number of tasks increases from 120 to 240, the present invention demonstrates a sustained advantage in scheduling latency, resource utilization, conflict control, and strategy stability, indicating that the method possesses good adaptability and stable operation capabilities in complex dynamic environments.

[0039] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An IT service collaborative scheduling and management system based on a multi-agent system, characterized in that, include: The data module is used to acquire task data, node status data, and resource data, perform alignment at index positions, write values ​​at feature positions, and form a status vector. The relation module is used to parse the state vector, generate a set of nodes at the node positions, write the dependency relationship at the node pair positions, and generate the agent identifier at the index position. The mirror module is used to construct policy variables. It copies and generates shadow policy variables at the agent's identifier position, establishes a mapping relationship at the corresponding position, and forms a mirror policy space. The dual module is used to extract resource capacity and occupancy values ​​at the resource data index location, calculate the difference at the corresponding location, generate constraint data at the difference location, generate dual variables at the policy variable location, establish coupling relationship between the policy location and the dual variable location, and perform calculations at the coupling location to form a policy mapping function. The spectral constraint module is used to perform a linear expansion of the policy mapping function, generate the Jacobian matrix at the derivative position of the policy mapping function, generate the spectral radius at the eigenvalue position, and perform a decision at the spectral radius position to filter the policy set. The verification module is used to perform consistency verification at the index position of the policy variable, calculate the difference between the corresponding positions of the policy variable and the shadow policy variable, perform constraint verification at the constraint position, and form a set of balanced policies. The output module is used to read policy values, perform mapping calculations between task and resource locations, and generate scheduling sequences.

2. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 1, characterized in that, The data module includes the following steps: Receive the task data set at the task index position, the node status data set at the node index position, and the resource data set at the resource index position. Establish a time index sequence at the unified index position. Write the task identifier value and task requirement value in the task data set into the task feature position at the corresponding position of the time index sequence. Write the node load value and node availability value in the node status data set into the node feature position. Write the resource capacity value and resource usage value in the resource data set into the resource feature position. A feature dimension number sequence is established at the feature dimension index position. Dimension alignment calculations are performed on the task feature position, node feature position, and resource feature position at the corresponding positions of the feature dimension number sequence. An aligned feature value sequence is generated at the unified feature dimension position. A concatenation calculation is performed on the task feature value sequence, node feature value sequence, and resource feature value sequence at the unified feature dimension position. A joint feature value sequence is generated at the concatenation position. A state vector index sequence is established at the corresponding position of the joint feature value sequence. The joint feature value sequence is written at the state vector index position to generate a state vector.

3. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 2, characterized in that, The specific steps in the relation module, such as writing dependencies at node positions, include: Extract the numerical sequences of task requirements, node load, resource capacity, and resource usage at the state vector index positions. Establish a node identifier sequence at the node index positions and establish a node pair combination sequence at the corresponding positions of the node identifier sequences. At the corresponding position of the node pair combination sequence, calculate the difference between the resource capacity value and the resource usage value for each node identifier pair, generate a resource difference value at the difference position, calculate the ratio between the node load value and the task requirement value, generate a load ratio value at the ratio calculation position, perform a weighted calculation between the resource difference value position and the load ratio value position, generate a dependency value at the weighted calculation position, and write the dependency value at the corresponding position of the node pair combination sequence.

4. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 3, characterized in that, The mirror module specifically includes: Extract the agent identifier sequence at the node index position, extract the state vector value sequence at the state vector index position, establish an index mapping relationship between the node index position and the state vector index position, establish a mapping number sequence at the state vector index position, perform mapping calculation on the state vector value sequence at the corresponding position of the mapping number sequence, generate the mapping value at the corresponding position, and write the mapping value at the corresponding position of the mapping number sequence to form the original mapping value sequence. Establish a mirror index sequence at the corresponding position of the mapping number sequence, perform copy calculation on the mapping value at the corresponding position of the mirror index sequence, generate a shadow mapping value at the corresponding position, establish a one-to-one correspondence between the mapping value position and the shadow mapping value position, calculate the difference at the corresponding position, and generate an offset value at the difference position. The mapping value position and the offset value position are combined for calculation, and a mirror mapping value is generated at the combined position. The mirror mapping value is written at the corresponding position of the mirror index sequence to form a mirror strategy space.

5. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 4, characterized in that, The dual module generates dual variables at the policy location, specifically including: Extract the resource capacity value sequence and resource occupancy value sequence at the resource data index position, calculate the difference between the resource capacity value and the resource occupancy value at the resource data index position, generate the resource difference value sequence at the difference position, extract the state vector value sequence at the state vector index position, extract the agent identifier sequence at the node index position, establish an index mapping relationship between the state vector index position and the resource index position, establish a policy index sequence at the state vector index position, and establish a constraint mapping position at the corresponding position of the policy index sequence. At the corresponding position of the policy index sequence, perform mapping calculation on the resource difference value sequence, generate constraint mapping value at the constraint mapping position, perform product calculation on the state vector value sequence at the constraint mapping value position, generate first coupling value at the product position, perform index matching calculation on the agent identifier sequence at the constraint mapping value position, generate matching weight value at the matching position, perform combination calculation on the first coupling value position and the matching weight value position, generate second coupling value at the combination position, perform normalization calculation on the second coupling value position, and generate intermediate dual value at the normalization position. Establish a dual variable number sequence at the corresponding position of the strategy index sequence. Perform mapping and writing calculation on the intermediate dual values ​​at the corresponding positions of the dual variable number sequence. Generate dual variable values ​​at the corresponding positions. Establish dual variable writing positions at the corresponding positions of the dual variable number sequence. Write dual variable values ​​at the dual variable writing positions. Form a dual variable sequence at the corresponding positions of the strategy index sequence. Establish a one-to-one correspondence between the dual variable value positions and the constraint mapping value positions. Perform difference correction calculation on the dual variable values ​​and constraint mapping values ​​at the corresponding positions. Generate updated dual variable values ​​at the correction positions. Write updated dual variable values ​​at the corresponding positions of the dual variable number sequence to form a dual variable sequence.

6. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 5, characterized in that, The dual module establishes a coupling relationship between the policy position and the dual variable position, and performs calculations at the coupling position to form the policy mapping function, specifically including: Read the numerical sequence of the strategy variable at the position of the strategy variable, read the numerical sequence of the dual variable at the position of the dual variable, perform position alignment operation on the position of the strategy variable and the position of the dual variable under the unified indexing rule, and establish a one-to-one coupled index relationship at the aligned position. Write the constraint weight identifier corresponding to the constraint data at the coupling index position, extract the strategy variable value at the strategy variable position, extract the dual variable value at the dual variable position, and perform product calculation at the same coupling index position to generate a sequence of coupled component values. Perform a weighted accumulation operation at the corresponding position of the coupled component numerical sequence, generate the coupled result value at the accumulation position, introduce the coupled result value and the original strategy variable value at the strategy variable position to perform a combination calculation, and generate an updated strategy variable numerical sequence. Establish a function parameter record structure at the corresponding position of the updated strategy variable value sequence, write the updated strategy variable value at the parameter record position, establish a mapping relationship between the input index and the output index at the parameter record position, and perform function generation operation at the mapping relationship position to form a strategy mapping function expression structure; Write the input policy variable index, dual variable index, and constraint data index at the corresponding positions in the policy mapping function expression structure to form a policy mapping function structure containing the input index set and the output index set.

7. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 6, characterized in that, The generation of the Jacobian matrix at the derivative position of the policy mapping function in the spectral constraint module specifically includes: Extract the numerical sequence of the strategy mapping function at the corresponding position of the strategy index sequence, establish the derivative index sequence at the corresponding position of the strategy index sequence, establish the variable number sequence at the corresponding position of the derivative index sequence, expand the numerical sequence of the strategy mapping function according to the variable dimension at the corresponding position of the variable number sequence to generate the numerical sequence of function components, and establish the index sequence of function components at the corresponding position of the numerical sequence of function components. At the corresponding position of the function component index sequence, perform difference calculation on the value of each function component along the variable number sequence, generate partial derivative values ​​at the difference position, establish a derivative matrix index structure at the corresponding position of the variable number sequence, use the function component index sequence as the row index and the variable number sequence as the column index in the derivative matrix index structure, write the partial derivative values ​​at the corresponding row and column intersection position to form the derivative matrix value structure. A matrix number sequence is established at the corresponding position of the derivative matrix index structure. A matrix rearrangement calculation is performed on the derivative matrix numerical structure at the corresponding position of the matrix number sequence. A Jacobi matrix value is generated at the rearranged position. The Jacobi matrix value is written at the corresponding position of the matrix number sequence to form the Jacobi matrix.

8. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 7, characterized in that, The spectral constraint module generates spectral radii at eigenvalue locations and performs judgments at spectral radius locations. The specific set of filtering strategies includes: Extract the Jacobian matrix numerical sequence at the corresponding position of the matrix number sequence, establish the eigenvalue index sequence at the corresponding position of the matrix number sequence, perform eigenvalue decomposition calculation on the Jacobian matrix numerical sequence at the corresponding position of the eigenvalue index sequence, generate the eigenvalue numerical sequence at the corresponding position, perform modulus calculation on the eigenvalue numerical sequence at the corresponding position of the eigenvalue index sequence, generate the eigenvalue modulus sequence at the corresponding position, perform maximum value calculation on the eigenvalue modulus sequence at the corresponding position of the eigenvalue index sequence, and generate the spectral radius value at the maximum value position; Establish a spectral radius mapping position at the corresponding position of the strategy index sequence, map the spectral radius value to the corresponding strategy position at the spectral radius mapping position, establish a decision index sequence at the corresponding position of the strategy index sequence, perform a comparison calculation between the spectral radius value and the preset threshold value at the corresponding position of the decision index sequence, write a filter identifier at the position that meets the comparison condition, extract the corresponding strategy value at the filter identifier position, and generate a filtered strategy set at the corresponding position of the strategy index sequence.

9. The IT service collaborative scheduling and management system based on a multi-agent system according to claim 8, characterized in that, The constraint verification module performs constraint verification at constraint locations to form a set of equilibrium strategies, specifically including: Extract the filtered strategy set at the corresponding position of the strategy index sequence, extract the resource capacity value sequence and resource usage value sequence at the resource index position, perform difference calculation on the resource capacity value and resource usage value at the resource index position, generate the resource difference value sequence at the difference position, establish the constraint mapping position at the corresponding position of the strategy index sequence, map the resource difference value sequence to the corresponding strategy position at the constraint mapping position, and establish the constraint verification index sequence at the corresponding position of the strategy index sequence. At the corresponding position in the constraint verification index sequence, perform difference calculation on the strategy value and the constraint mapping value, generate constraint deviation value at the difference position, perform cumulative calculation on the constraint deviation value corresponding to each strategy at the constraint deviation value position, generate constraint cumulative value at the cumulative position, perform comparison calculation on the constraint cumulative value position, write a valid identifier at the position that meets the comparison condition, extract the corresponding strategy value at the valid identifier position, and generate a set of balanced strategies at the corresponding position in the strategy index sequence.

Images