CN-122024367-A - Self-adaptive visitor reservation management system and method based on multi-agent collaborative decision

CN122024367ACN 122024367 ACN122024367 ACN 122024367ACN-122024367-A

Abstract

The invention discloses a self-adaptive visitor reservation management system and method based on multi-agent collaborative decision-making, comprising the following steps of obtaining visitor management related data to carry out structural processing, analyzing visitor reservation requirements to generate reservation requirement vectors to construct an extended heterogeneous graph structure, inputting the extended heterogeneous graph structure into an input layer of an improved heterogeneous transducer model to carry out corresponding modulation processing, executing message transformation and aggregation updating, obtaining a global agent state representation set through stacking processing of an encoding layer, carrying out screening and aggregation decoding through a flow strategy decoding passage, carrying out feature extraction and decoding of a security view through the strategy decoding passage, inputting a flow strategy vector and an authority strategy vector into a strategy linkage gating fusion module to obtain a combined strategy output result and executing. The invention adopts an improved heterogeneous transducer model to realize visitor reservation management.

Inventors

GUO RUIYUAN
RONG CHAO
LIU BING
LI DABAO
WANG DESHAN
Niu Chunyun

Assignees

泽瑞科技集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260324

Claims (9)

1. The self-adaptive visitor reservation management method based on multi-agent collaborative decision-making is characterized by comprising the following steps: The method comprises the steps of obtaining visitor management related data, carrying out structural processing on the visitor management related data, and constructing multi-agent heterograms to form an initial heterogram representation set; analyzing the reservation demand of the visitor, generating a reservation demand vector, and associating the reservation demand vector with the initial heterogeneous diagram representation set to obtain an extended heterogeneous diagram structure; Inputting the extended heterogeneous graph structure into an input layer of an improved heterogeneous transducer model, and carrying out corresponding modulation processing one by one to obtain an event modulation attention result set; Based on the event modulation attention result set and the extended iso-composition structure, performing message transformation and aggregation update, and obtaining a global agent state representation set through stacking processing of the coding layers; Inputting the global agent state representation set into a flow strategy decoding path, screening, aggregating and decoding to generate a flow strategy vector; Inputting the global intelligent agent state representation set into a right strategy decoding path, extracting and decoding characteristics of a security view angle, and generating a right strategy vector; and inputting the flow strategy vector and the authority strategy vector into a strategy linkage gating fusion module, obtaining a combined strategy output result and executing.
2. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the obtaining visitor management related data, performing a structuring process on the visitor management related data, and constructing a multi-agent heterogram, the forming an initial heterogram representation set specifically includes: The method comprises the steps of obtaining visitor management related data, and performing structural processing on the visitor management related data to form an input data set; Constructing a node set of multi-agent heterograms and an edge set connecting the nodes based on the input data set; organizing the node set and the edge set according to the node identification and the start-stop node reference relation of the edge to form a multi-agent heterogram; Acquiring node characteristic information, node type embedded data and role embedded data, and combining to form node representation data; Determining the edge type of each edge in the edge set, and generating corresponding edge type embedded data, event type embedded data and time difference coded data; and organizing the node representation data, the edge type embedded data, the event type embedded data and the time difference coding data to form an initial heterograph representation set.
3. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the parsing the visitor reservation requirement, generating a reservation requirement vector, and associating the reservation requirement vector with an initial heterogeneous graph representation set, the obtaining an extended heterogeneous graph structure specifically includes: Receiving visitor reservation request data submitted by a visitor, and performing field analysis and format specification processing on the visitor reservation request data to obtain standard visitor reservation request data; Performing coding processing on standard visitor reservation request data, and combining various coded features according to a preset sequence to generate a reservation demand vector corresponding to a visitor reservation request; creating reservation request nodes in the multi-agent heterogeneous graph, and generating corresponding reservation relation edges between the reservation request nodes and the visited person nodes, conference room resource nodes, access control path nodes and approval nodes; and adding the reservation request node and the reservation relation edge into the initial heterogeneous diagram representation set to obtain an extended heterogeneous diagram structure.
4. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the inputting the extended heterogeneous graph structure into the input layer of the improved heterogeneous transducer model, performing a corresponding modulation process one by one, and obtaining an event modulation attention result set specifically includes: Inputting the extended heterogeneous graph structure to an input layer of an improved heterogeneous transducer model to obtain an input node sequence vector; Generating basic attention weights for any node pair based on node type embedded vectors contained in the node representation data and edge type embedded vectors contained in the edge type embedded data; Generating an event gating factor for modulating the basic attention weight based on an event type embedded vector corresponding to the edge in the event type embedded data and a time difference coded vector corresponding to the edge in the time difference coded data; performing corresponding modulation processing on the basic attention weight and the corresponding event gating factor one by one to generate an event modulation attention result; and collecting the event modulation attention results of all the node pairs according to the node indexes to form an event modulation attention result set.
5. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the performing message transformation and aggregation update based on the event-modulated attention result set and the extended heterograph structure, obtaining the global agent state characterization set through stacking processing of the encoding layers specifically comprises: In the improved heterogeneous transducer model, based on event modulation attention results and edge type embedded vectors, performing relationship-specific message transformation processing on any node pair to generate a relationship-specific message vector; aggregating all relation specific message vectors of the same target node to obtain an aggregated message vector of the node; Splicing the node representation data of the target node with the aggregate message vector of the target node, and performing transformation to generate an updated state representation of the target node; And forming a new node representation set by the updated state representations of all the nodes, inputting the new node representation set into a coding layer of a next-layer improved heterogeneous transducer model for repeated message transformation, message aggregation and node updating processing, and generating a global agent state representation set through multi-layer stacking.
6. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the inputting the global agent state representation set into the flow policy decoding path for screening and aggregate decoding, generating a flow policy vector specifically comprises: Inputting the global intelligent agent state representation set into a process strategy decoding path of the improved heterogeneous transducer model, and selecting update state representations corresponding to visitor nodes, visited person nodes and meeting room resource nodes from the global intelligent agent state representation set to form a process related node state subset; performing feature aggregation processing on the updated state representations in the state subsets of the related nodes of the flow to obtain a flow context vector; Inputting the flow context vector into a multi-layer decoding network, and generating a flow strategy hidden representation vector through full-connection transformation and nonlinear activation processing; Decomposing the flow strategy hidden representation vector into an access time window adjustment vector, a path planning vector, a resource allocation vector and a flow jump as probability vectors through linear mapping; and combining the access time window adjustment vector, the path planning vector, the resource allocation vector and the flow jump operation probability vector to form a flow strategy vector.
7. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the inputting the global agent state representation set into the authority policy decoding path, performing feature extraction and decoding of a security view, generating an authority policy vector specifically comprises: Inputting the global intelligent agent state representation set into a right strategy decoding path of an improved heterogeneous transducer model, and selecting updated state representations corresponding to security nodes, access control path nodes and regional nodes from the global intelligent agent state representation set to form a security related node state subset; Performing feature aggregation processing on the security-related node state subsets to obtain security context vectors, inputting the authority context vectors into an authority matrix generation structure, and generating an access area time matrix; Inputting the authority context vector into an authorization instruction generating structure to generate an authorization instruction set vector, inputting the authority context vector into a withdrawal instruction generating structure to generate a withdrawal instruction set vector, inputting the authority context vector into a risk constraint generating structure to generate a risk constraint vector; and combining the access area time matrix, the authorization instruction set vector, the withdrawal instruction set vector and the risk constraint vector to form a permission policy vector.
8. The adaptive visitor reservation management method based on multi-agent collaborative decision-making according to claim 1, wherein the inputting the flow policy vector and the authority policy vector into the policy linkage gating fusion module, obtaining a joint policy output result and executing specifically comprises: inputting the flow strategy vector and the authority strategy vector into a strategy linkage gating fusion module, and performing splicing treatment on the flow Cheng Celve vector and the authority strategy vector to form a strategy combination vector; inputting the strategy combination vector into a gating coefficient generation structure, and generating a gating coefficient vector through linear mapping and activation processing; Performing weighted fusion processing on the flow strategy vector and the authority strategy vector based on the gating coefficient vector to obtain a combined strategy output result; Executing visitor reservation flow scheduling processing based on the combined policy output result, and adjusting a visitor access time window according to time-related content in the combined policy output result; executing resource state updating processing based on the combined strategy output result, and updating the occupation states of the meeting room and the station according to the resource related content in the combined strategy output result; Executing notification distribution processing of the visitor and the interviewee based on the combined policy output result, and converting action content contained in the combined policy output result into a corresponding notification instruction; And executing access right control processing based on the combined policy output result, and mapping the authorized or removed right content in the combined policy output result into an access right adjustment result.
9. The multi-agent collaborative decision-based adaptive guest reservation management system of claim 1, performing the multi-agent collaborative decision-based adaptive guest reservation management method of any one of claims 1 to 8, comprising the following modules: The data processing module is used for acquiring visitor management related data, and carrying out structural processing on the visitor management related data to form an input data set; the different composition construction module is used for constructing a multi-agent different composition based on the input data set to form an initial heterogeneous diagram representation set; The reservation demand analysis module is used for analyzing the reservation demand of the visitor, generating a reservation demand vector, and associating the reservation demand vector with the initial heterogeneous diagram representation set to obtain an extended heterogeneous diagram structure; the event gating attention module is used for inputting the extended heterogeneous graph structure into an input layer of the improved heterogeneous transducer model, and carrying out corresponding modulation processing one by one to obtain an event modulation attention result set; The heterogeneous relation updating module is used for executing message transformation and aggregation updating based on the event modulation attention result set and the extended iso-composition structure, and obtaining a global intelligent agent state representation set through stacking processing of the coding layers; The flow strategy generation module is used for inputting the global intelligent agent state representation set into a flow strategy decoding path, screening, aggregating and decoding to generate a flow strategy vector; the authority policy generation module is used for inputting the global intelligent agent state representation set into an authority policy decoding path, extracting and decoding characteristics of a security view angle and generating an authority policy vector; and the strategy linkage gating fusion module is used for processing the flow strategy vector and the authority strategy vector, obtaining a combined strategy output result and executing the combined strategy output result.

Description

Self-adaptive visitor reservation management system and method based on multi-agent collaborative decision Technical Field The invention relates to the technical field of intelligent visitor management, in particular to a self-adaptive visitor reservation management system and method based on multi-agent cooperative decision. Background The existing visitor reservation management system generally relies on a rule driving mode, and visitor registration, approval circulation, meeting room occupation registration and access control are realized through preset flow configuration. The system is based on static flow and fixed logic, lacks dynamic modeling capability for visitor behavior characteristics, visited person states, resource occupation conditions and security risks, and is difficult to adjust in real time when facing access peaks, resource conflicts, multi-role collaboration and temporary risk changes. A part of schemes introduce basic graph structures or simple strategy models, but cannot establish complex dependency relationships among multiple entities capable of expressing visitor nodes, visited person nodes, conference room resource nodes, security nodes and access control path nodes at the same time, and cannot combine access time, event triggering information and approval links to perform unified reasoning. In addition, the existing system generally separates the process strategy generation and the authority strategy generation, so that a linkage mechanism is lacked among time coordination, resource scheduling, path planning and access authority control, and scheduling conflict, repeated approval or authority lag can be easily generated. For the scenes of cross-regional access, resource shortage or security requirements, the traditional method cannot generate an executable and landable comprehensive decision strategy in real time based on the states of multiple parties. Therefore, how to provide an adaptive visitor reservation management system and method based on multi-agent collaborative decisions is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention aims to provide a self-adaptive visitor reservation management system and method based on multi-agent collaborative decision, and the self-adaptive visitor reservation management system and method introduce a coding mechanism of a multi-agent heterogeneous graph modeling and improved heterogeneous Transformer model, uniformly express and reason multi-class relations among visitors, interviewees, meeting room resources, security units, access control paths and approval entities, construct a double-channel decoding structure for flow strategy generation and authority strategy generation, and finally realize collaborative decision of access flow scheduling, resource allocation and authority control through strategy linkage gating fusion. The method can process dynamic access requirements, resource conflicts and security risk changes, and has the advantages of high flow flexibility, good strategy consistency and high decision automation degree. According to the embodiment of the invention, the self-adaptive visitor reservation management method based on multi-agent collaborative decision-making comprises the following steps: The method comprises the steps of obtaining visitor management related data, carrying out structural processing on the visitor management related data, and constructing multi-agent heterograms to form an initial heterogram representation set; analyzing the reservation demand of the visitor, generating a reservation demand vector, and associating the reservation demand vector with the initial heterogeneous diagram representation set to obtain an extended heterogeneous diagram structure; Inputting the extended heterogeneous graph structure into an input layer of an improved heterogeneous transducer model, and carrying out corresponding modulation processing one by one to obtain an event modulation attention result set; Based on the event modulation attention result set and the extended iso-composition structure, performing message transformation and aggregation update, and obtaining a global agent state representation set through stacking processing of the coding layers; Inputting the global agent state representation set into a flow strategy decoding path, screening, aggregating and decoding to generate a flow strategy vector; Inputting the global intelligent agent state representation set into a right strategy decoding path, extracting and decoding characteristics of a security view angle, and generating a right strategy vector; and inputting the flow strategy vector and the authority strategy vector into a strategy linkage gating fusion module, obtaining a combined strategy output result and executing. Optionally, the obtaining the visitor management related data, performing a structuring process on the visitor management related data, and constructing a multi-agent heterogram, whe