Distributed virtual robot AI system
By using edge computing nodes and center management nodes to cooperate in distributed virtual robot AI systems, the problems of node coordination and uneven resource allocation are solved, cross-node business processing is optimized, system efficiency and reliability are improved, and modern Business requirements.
Patent Information
- Application Number
- CN202410966754.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-18
AI Technical Summary
In the prior art, the distributed node deployment scheme has problems such as node coordination and management complexity, uneven resource allocation, and cross-node service processing complexity.
Adopt a distributed virtual robot AI system to deploy virtual robots through edge computing nodes, and create, replicate or migrate virtual robots on the central management nodes to realize dynamic resource allocation and cross-node business processing optimization.
Reduces management complexity, ensures optimal utilization of resources, improves system efficiency and reliability, and adapts to changes in business needs.
Smart Images

Figure CN118921373A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a distributed virtual robot AI system. Background Art
[0002] Business audit is to verify the integrity, consistency and authenticity of business acceptance documents. Traditional audit is mainly manual, with huge labor costs and low audit efficiency. Take a provincial company of an operator as an example. The annual audit work order exceeds 40 million, and the cost of manual audit of all businesses is nearly 100 million. Through the field of intelligent image recognition, with the help of AI technology, combined with the business characteristics of the communications industry, we can create two capabilities: market business AI audit and government business AI audit, replace manual judgment operations, and realize comprehensive intelligent audit of market and government and enterprise basic businesses.
[0003] The AI audit business process is to transmit the input file to the application system, and return the AI recognition result after the combined service of AI audit preprocessing + AI audit rules + AI core algorithm.
[0004] By focusing on the audit business needs in the government, enterprise and market sectors, and applying AI technology to solve the bottleneck of audit intelligence, AI audit services currently cover all highly sensitive business scenarios such as account opening, card replacement, password reset, identity confirmation, transfer, account cancellation and reopening, significantly improving the digitalization level of government, enterprise and market sectors.
[0005] With the rapid growth of corporate business, especially the popularity of online transactions and services, the number of business documents that enterprises need to process has increased dramatically. These documents include but are not limited to account opening, transaction records, identity verification, etc. They need to go through a strict audit process to ensure compliance and accuracy. Traditional manual audit methods are not only time-consuming and labor-intensive, but also prone to errors and cannot meet the needs of rapid response and large-scale processing.
[0006] To meet this challenge, enterprises have begun to explore and implement automated audit solutions based on artificial intelligence (AI). AI audit business processing uses advanced technologies such as machine learning, natural language processing (NLP), and image recognition to automatically verify the integrity, consistency, and authenticity of business documents. In this way, enterprises can improve audit efficiency, reduce labor costs, and reduce audit risks caused by human errors.
[0007] In order to achieve efficient AI audit business processing, a distributed node deployment solution has been proposed and widely used. In this solution, the AI audit system is deployed on nodes in multiple geographical locations, and each node is responsible for processing business documents in a specific area or type. This distributed architecture provides multiple advantages:
[0008] Load balancing: By distributing the processing load to different nodes, the system can more effectively manage a large number of business requests and avoid single point overload.
[0009] Response speed: The geographical proximity of nodes can reduce data transmission delays and improve the response speed to local business requests.
[0010] Reliability and fault tolerance: Distributed architecture improves the stability of the system. Even if a node fails, other nodes can continue to provide services.
[0011] Scalability: As the volume of business grows, the system's processing capacity can be expanded by adding new nodes without making large-scale changes to the existing architecture.
[0012] Data security and compliance: Distributed nodes can better comply with data privacy and compliance requirements because data can be processed where it is generated, reducing cross-regional transmission.
[0013] By implementing a distributed node deployment solution for AI audit business processing, enterprises can not only improve the efficiency and quality of business processing, but also better adapt to the ever-changing market environment and business needs.
[0014] However, the distributed node deployment solution in the prior art has the following problems:
[0015] 1. Node coordination and management complexity:
[0016] Although the distributed node deployment solution improves processing power and response speed, as the number of nodes increases, the coordination and management between nodes becomes more complicated, and an efficient central management node is needed to maintain the consistency and coordination of the entire system.
[0017] 2. Uneven resource allocation and efficiency issues:
[0018] Existing technologies may not be able to dynamically perform optimal resource allocation based on business needs and node status, resulting in some nodes being overloaded and other node resources being idle.
[0019] 3. Complexity of cross-node business processing:
[0020] When business processing needs to be performed across multiple nodes, existing technologies may lack effective mechanisms to manage and optimize these cross-node business processes. Summary of the invention
[0021] To this end, the technical problem to be solved by the present invention is to overcome the problems of complex node coordination and management, cross-border point business processing and uneven resource allocation in the prior art.
[0022] In order to solve the above technical problems, the present invention provides a distributed virtual robot AI system, comprising:
[0023] Edge computing nodes, used to deploy virtual robots, which are a collection of one or more AI audit capabilities;
[0024] The central management node is used to create, copy or migrate virtual robots based on user needs and in combination with multiple computing resource information of edge computing nodes.
[0025] Preferably, the creating the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes:
[0026] Get the user's virtual robot creation request message;
[0027] Sending a resource query request to one or more edge computing nodes at the location or area where the virtual robot is deployed according to the creation request message;
[0028] Determine one or more AI data models of the virtual robot according to the resource message returned by the edge computing node, and send the AI data models to one or more edge computing nodes for deployment;
[0029] Send a creation success notification to the user.
[0030] Preferably, determining one or more AI data models of the virtual robot according to the resource message returned by the edge computing node, and sending the AI data models to one or more edge computing nodes for deployment includes:
[0031] According to the resource messages returned by the edge computing nodes and the business needs of the virtual robot, an AI data model is selected from the predefined AI model library and sent to the edge computing nodes that match its resources for deployment;
[0032] After deployment is complete, the AI data model is performance verified to ensure that it meets business requirements.
[0033] Preferably, the edge computing node is also used to combine multiple computing resource information of the node, create a virtual robot, and register it in the central management node.
[0034] Preferably, the combining of multiple computing resource information of the node to create a virtual robot and registering it in the central management node includes:
[0035] Determine the type and capabilities of virtual robots to be created based on business needs;
[0036] Determine whether its own computing resources can support the operation of the virtual robot;
[0037] Select an AI data model from the predefined AI template library based on the business needs analysis and resource assessment results, and train the model locally or in collaboration with other edge computing nodes.
[0038] Integrate the trained AI data model into the virtual robot and assign a unique identifier to the newly created virtual robot;
[0039] Send the registration information of the newly created virtual robot to the central management node.
[0040] Preferably, the step of replicating the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes:
[0041] Get the virtual robot copy request message;
[0042] According to the ID information of the copied virtual robot, a query request for the copied virtual robot information is sent to the edge computing node at the location or area where the copied virtual robot is deployed;
[0043] Send a resource query request to the edge computing node in the location or area where the new virtual robot is deployed according to the location or area where the new virtual robot is deployed and the ID information of the edge computing node where the new virtual robot is deployed;
[0044] Determine whether the copy operation can be performed based on the returned information. If so, copy it and send a copy success notification to the user.
[0045] Preferably, the migrating the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes:
[0046] Get the virtual robot migration request message;
[0047] Sending a query request for information of the migrated virtual robot to the edge computing node at the location or area where the migrated virtual robot is deployed according to the ID information of the migrated virtual robot, the source network location or area, and the edge computing node ID information;
[0048] Send a resource query request to the edge computing node in the location or area where the new virtual robot is deployed according to the location or area where the new virtual robot is deployed and the ID information of the edge computing node where the new virtual robot is deployed;
[0049] Determine whether the migration operation can be performed based on the returned information. If yes, perform the migration.
[0050] Sending a virtual robot information deletion request to an edge computing node at a source location or region where the migrated virtual robot is deployed;
[0051] After the deletion is successful, a migration success notification is sent to the user.
[0052] Preferably, the virtual robot is used for:
[0053] According to the service request message sent by the service client, the service is parsed and processed. If the processing is unsuccessful, a virtual robot capability query message is sent to the central management node;
[0054] Sending a service analysis processing request to a backup virtual robot recommended by the central management node that meets the capability requirements of the current service virtual robot;
[0055] The processed data after parsing by the backup virtual robot is obtained and sent to the business client.
[0056] Preferably, the parsing of the service includes:
[0057] Perform preliminary verification on received business data to ensure data integrity and format correctness, and pre-process image or document data;
[0058] Automatically match applicable audit rules from the predefined audit rule engine based on business type and data content, and conduct preliminary verification of business data to check consistency and compliance;
[0059] Use the trained AI data model to identify anomalies, risk points, and inconsistencies in business data;
[0060] Summarize the rule engine matching results and AI data model analysis results to form an audit report.
[0061] Preferably, the central management node selects a backup virtual robot that meets the current business virtual robot capability requirements and has the highest priority from the alliance domain for recommendation, wherein the alliance domain is: a virtual robot alliance generated by message instructions during the virtual robot creation phase or a virtual robot alliance generated automatically and intelligently.
[0062] The above technical solution of the present invention has the following advantages compared with the prior art:
[0063] The distributed virtual robot AI system described in the present invention introduces a more advanced central management node, uses a machine learning algorithm to optimize the coordination and resource management between nodes, establishes a virtual robot to process the business, and reduces the management complexity; the present invention implements a dynamic resource allocation mechanism based on real-time load and capacity evaluation to ensure that resources are optimally utilized, and realizes the replication and migration of virtual robots to improve the overall system efficiency; the present invention provides an optimization strategy for cross-node business processing, and improves processing efficiency and reliability through the introduction of a virtual robot alliance. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:
[0065] Figure 1 This is a system architecture diagram of a distributed virtual robot AI system provided by the present invention;
[0066] Figure 2 It is a schematic diagram of the process of the central management node creating a virtual robot;
[0067] Figure 3 It is a schematic diagram of the process of creating a virtual robot by an edge computing node;
[0068] Figure 4 Schematic diagram of the virtual robot replication process;
[0069] Figure 5 It is a schematic diagram of the virtual robot migration process;
[0070] Figure 6 This is a schematic diagram of the process in which virtual robots back up each other. DETAILED DESCRIPTION
[0071] The core of the present invention is to provide a distributed virtual robot AI system, which effectively improves the processing capacity of AI auditing business, reduces costs, enhances the flexibility and security of the system, and better meets the needs of modern business.
[0072] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0073] Please refer to Figure 1 , Figure 1 This is a system architecture diagram of a distributed virtual robot AI system provided by the present invention; specifically including:
[0074] Edge computing nodes, used to deploy virtual robots, which are a collection of one or more AI audit capabilities;
[0075] The central management node is used to create, copy or migrate virtual robots based on user needs and in combination with multiple computing resource information of edge computing nodes.
[0076] Wherein: the user is: a distributed virtual robot AI system manager;
[0077] The central management node is a node in the distributed virtual robot AI system that is used to uniformly manage the virtual robots deployed on each network edge computing node in the system.
[0078] The virtual robot is: a collection of one or more AI audit capabilities deployed on edge computing nodes in a distributed virtual robot AI system. The virtual robot can be implemented by one or more AI data models. One or more AI audit capabilities of the virtual robot can be distributedly deployed on one or more physical edge computing nodes in the same network area. The virtual robot provides a unified input and output interface, or a unified input and output management module, for external nodes (such as a central management node, a business client, or other virtual robots). The virtual robot uses a unique ID in the distributed virtual robot AI system as an identifier in the distributed virtual robot AI system. Multiple virtual robots can share the same AI audit capability, that is, an edge computing node that deploys an AI audit capability can be accessed by multiple virtual robots at the same time. Multiple virtual robots can form a virtual robot alliance, that is, within the virtual robot alliance, the capabilities of the virtual robots can be fully shared or the list of capabilities that can be shared can be determined by the virtual robot alliance. The virtual robot alliance is a logical organization of multiple virtual robots, which may not be bound to the physical deployment location of the virtual robot, that is, multiple virtual robots that are not deployed in adjacent physical network locations can form a virtual robot alliance according to the negotiated rules and determine the list of capabilities that can be shared within the alliance. The same virtual robot can be counted in different virtual robot alliances.
[0079] Based on the above embodiments, in this distributed virtual robot AI system architecture, the central management node maintains one or more virtual robot information, including virtual robot ID information, AI audit capability information, resource information, network topology information, etc. The virtual robot needs to deploy an AI data model that supports AI audit capabilities, and the central management node may not need to deploy an AI data model that supports AI audit capabilities. When the virtual robot actively updates the AI audit capability or AI data model, it needs to report the AI audit capability release / update notification to the central management node. When the virtual robot reports the AI audit capability release / update notification to the central management node, it can indicate to the central management node its AI audit capability list disclosed within different virtual robot alliances, as well as its AI audit capability list disclosed outside the virtual robot alliance. When a virtual robot needs to interact with other virtual robots, it can query the virtual robot information through the central management node.
[0080] The distributed virtual robot AI system involved in the present invention includes the following stages and edge computing node interaction processes: creation of virtual robots, replication and migration of virtual robots, and collaborative services such as mutual assistance and backup provided by virtual robots through alliance domains.
[0081] Based on the above embodiments, Figure 2 As shown, the creation of a virtual robot is described in detail:
[0082] The step of creating the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes:
[0083] Obtain a user virtual robot creation request message; the request message carries the type of virtual robot and / or the type of business it serves, business volume demand information, the location / area where the virtual robot is deployed / ID of the edge computing node where the virtual robot is deployed, the virtual robot capability list information, the ID information of the virtual robot alliance to which the virtual robot applies to join, the capability list information shared by the virtual robot in each virtual robot alliance, the capability list information shared by the virtual robot outside the virtual robot alliance, etc. Among them, the types of virtual robots include network detection robots (for network management and maintenance), business service robots (for customer business services), etc.; the business types served by virtual business service robots include account opening, card replacement, password reset, identity confirmation, account transfer, account cancellation and reopening, etc.; business volume demand information such as x single A business / day or y bps business transmission; the location / area where the virtual robot is deployed includes network area ID information, network device ID information, location GPS information, etc. A virtual robot creation request message can apply to create one or more virtual robots.
[0084] According to the creation request message, a resource query request is sent to one or more edge computing nodes in the location or area where the virtual robot is deployed; if the virtual robot creation request message carries the edge computing node IDs of one or more virtual robot deployments indicated by the user, the central management node sends a resource request message to the edge computing nodes of one or more virtual robot deployments carried in the virtual robot creation request message. If the virtual robot creation request message does not carry the edge computing node ID information of the virtual robot deployment indicated by the user, the central management node sends a resource request message to one or more edge computing nodes in the location / area required for the deployment of the virtual robot carried in the virtual robot creation request message. Among them, the resource request message carries GPU resource information, CPU resource information, memory information, resource cost information, etc.
[0085] According to the resource message returned by the edge computing node, one or more AI data models of the virtual robot are determined, and the AI data models are sent to one or more edge computing nodes for deployment; each AI data model can support the provision of capabilities of one or more virtual robots. An AI data model is downloaded and installed by an edge computing node. The set of capabilities provided by the AI data models installed by one or more edge computing nodes constitutes the total set of capabilities of this virtual computing person.
[0086] A creation success notification is sent to the user. The message carries the virtual robot ID, capabilities provided by the virtual robot, and AI data model deployment topology information for each capability of the virtual robot.
[0087] Based on the above embodiment, determining one or more AI data models of the virtual robot according to the resource message returned by the edge computing node, and sending the AI data models to one or more edge computing nodes for deployment includes:
[0088] According to the resource messages returned by the edge computing nodes and the business needs of the virtual robot, an AI data model is selected from the predefined AI model library and sent to the edge computing nodes that match its resources for deployment;
[0089] After deployment is complete, the AI data model is performance verified to ensure that it meets business requirements.
[0090] Specifically, the central management node determines the AI data model as follows:
[0091] 1) Demand analysis: The central management node first analyzes the virtual robot creation request submitted by the user, including the business type, business volume demand, expected performance indicators, etc.
[0092] 2) Resource evaluation: The central management node sends a resource query request to the corresponding edge computing node based on the deployment location / area specified in the request to collect available computing resource information, such as GPU, CPU, memory, etc.
[0093] 3) Model selection: Based on the results of demand analysis and resource evaluation, the central management node selects the most suitable AI data model from the predefined AI model library. These models may be optimized for specific business scenarios or general models that can be customized according to needs.
[0094] 4) Model customization: If necessary, the central management node can customize the selected AI data model to meet specific business needs. This may include adjusting model parameters, selection of training data sets, or modification of model structure.
[0095] 5) Model deployment: The central management node sends the selected and customized AI data model to the designated edge computing node for deployment. The deployment process may include model installation, configuration, and initialization.
[0096] 6) Model testing: After deployment, the central management node will run a series of test cases to verify whether the AI data model meets the expected performance indicators and ensure its stability and reliability in the actual environment.
[0097] 7) Feedback and optimization: The test results will be fed back to the central management node. If the model does not achieve the expected results, the central management node may make further optimization adjustments.
[0098] In summary, the specific process of creating a virtual robot by the central management node is as follows:
[0099] 1) Receiving resource information: The central management node receives resource information from the edge computing nodes, including available computing resources and storage resources.
[0100] 2) Resource matching: The central management node evaluates which resources are most suitable for deploying the requested AI data model based on resource information and the business needs of the virtual robot.
[0101] 3) Model deployment decision: Based on the results of resource matching, the central management node decides which AI data models to deploy on which edge computing nodes and formulates a deployment plan.
[0102] 4) Issuance of deployment instructions: The central management node sends deployment instructions to the selected edge computing nodes, including model files, configuration parameters, and deployment guides.
[0103] 5) Monitor the deployment process: The central management node monitors the deployment process of the AI data model, ensures that the deployment proceeds as planned, and handles any problems that may arise.
[0104] 6) Performance verification: After deployment, the central management node will monitor the performance of the virtual robot to ensure that it meets business requirements.
[0105] 7) User notification: Once the AI data model is deployed and tested, the central management node will send a success notification to the user, including information such as the virtual robot’s ID and the capabilities provided.
[0106] Through this series of steps and processes, the central management node ensures that the AI data model of the virtual robot can be efficiently and accurately deployed in the distributed system to meet business needs and provide stable services.
[0107] Based on the above embodiments, the creation of a virtual robot can also be achieved by an edge computing node:
[0108] like Figure 3 , the edge computing node combines the information of multiple computing resources of the node to create a distributed function virtual robot, and then registers it with the central management node. Compared with the above virtual robot creation process, this embodiment is a decentralized process, in which the edge computing node uses its own computing resources to autonomously create a virtual robot and adds it to the system maintained by the central management node through the registration process.
[0109] Edge computing nodes play a vital role in distributed virtual robot AI systems. They are usually located at the edge of the network, close to the data source, such as user devices or sensors. The purpose of edge computing nodes is to provide computing power and storage capabilities where the data is generated, thereby reducing the need for data transmission to central data centers, reducing latency, and improving processing speed and efficiency.
[0110] Features of edge computing nodes:
[0111] Near-user computing: Edge computing nodes are usually deployed near users and can provide fast response and real-time data processing.
[0112] Resource limitations: Compared with central data centers, edge computing nodes may have limited computing power and storage resources.
[0113] Autonomy: Edge computing nodes are able to run and process tasks independently without direct intervention from a central management node.
[0114] Distributed processing: Edge computing nodes can work together with other edge computing nodes to jointly complete tasks or optimize resource usage.
[0115] Combining multiple computing resource information of this node, creating a virtual robot, and registering it in the central management node includes:
[0116] 1. One or more edge computing nodes create virtual robots through model training in the autonomous domain, where the capabilities of the virtual robot can be jointly provided by one or more AI data models, and one AI data model is downloaded and installed by one edge computing node. The set of capabilities provided by the AI data models installed by one or more edge computing nodes constitutes the total set of capabilities of this virtual computing person. The edge computing node determines the ID for the virtual robot.
[0117] The steps to create an edge computing node virtual robot include:
[0118] 1) Requirements analysis: The edge computing node first analyzes business requirements and determines the type and capabilities of the virtual robot that needs to be created.
[0119] 2) Resource evaluation: The edge computing node evaluates its own computing resources, including CPU, GPU, memory, and storage, to determine whether there are sufficient resources to support the operation of the virtual robot.
[0120] 3) Model selection and training: Based on demand analysis, the edge computing node can select a suitable AI model from the AI model library and perform model training locally or in collaboration with other edge computing nodes. The pre-configuration of the AI model library and the training method of the AI model are not limited in this invention.
[0121] 4) Virtual robot construction: The edge computing node integrates the trained AI model into the virtual robot to build a virtual robot with the required capabilities.
[0122] 5) ID allocation: Assign a unique identifier (ID) to the newly created virtual robot so that it can be identified and managed in the system.
[0123] 6) Registration preparation: The edge computing node prepares a registration message, which contains detailed information about the virtual robot, such as type, business type, deployment location, capability list, etc.
[0124] 2. The edge computing node sends a virtual robot registration message to the central management node. The registration message carries the type of the virtual robot and / or the type of business it serves, the location / area / edge computing node ID where the virtual robot is deployed, the virtual robot capability list information, the ID information of the virtual robot alliance that the virtual robot has applied to join, the capability list information shared by the virtual robot in each virtual robot alliance, and the capability list information shared by the virtual robot outside the virtual robot alliance.
[0125] 3. The central management node records and maintains the virtual robot registration information and sends a registration success message to the edge. Once confirmed, the edge computing node receives the registration success message and the virtual robot officially becomes part of the system.
[0126] Through the above steps, edge computing nodes are able to autonomously create and manage virtual robots, thereby improving the performance and efficiency of the entire system while reducing dependence on central data centers.
[0127] like Figure 4 As shown, based on the above embodiment, according to user needs and in combination with multiple computing resource information of edge computing nodes, copying the virtual robot includes:
[0128] Obtain a virtual robot replication request message; the virtual robot replication request message may come from a user's virtual robot replication request message, or the central management node determines a virtual robot that can be deployed through replication for an area where a virtual robot is not deployed based on a big data algorithm. The virtual robot replication request message carries information such as the virtual robot ID information, the location / area where the virtual robot is deployed, and the edge computing node ID where the virtual robot is deployed.
[0129] Based on the copied virtual robot ID information, a query request for the copied virtual robot information is sent to the edge computing node at the location or area where the copied virtual robot is deployed; when the central management node itself maintains all the information required to create this virtual robot as indicated by the virtual robot ID information (including virtual robot AI data model information, the topology of the nodes required for the innovative virtual robot, and the GPU resource information, CPU resource information, memory information, etc. of each point) and the central management node determines that the information it maintains is the latest version of the virtual robot required to create this virtual robot as indicated by the virtual robot ID information, this step can be omitted.
[0130] The edge computing node where the virtual robot is located, indicated by the virtual robot ID information, replies to the central management node with the virtual robot information, including the virtual robot AI data model information, the topology of the nodes required for the innovative virtual robot, and the GPU resource information, CPU resource information, memory information, etc. of each point.
[0131] According to the location or area where the new virtual robot is deployed and the edge computing node ID information of the new virtual robot deployment, a resource query request is sent to the edge computing node of the location or area where the new virtual robot is deployed; if the virtual robot creation request message carries the edge computing node IDs of one or more virtual robot deployments indicated by the user, the central management node sends a resource request message to the edge computing nodes of one or more virtual robot deployments carried in the virtual robot creation request message. If the virtual robot creation request message does not carry the edge computing node ID information of the virtual robot deployment indicated by the user, the central management node sends a resource request message to one or more edge computing nodes in the location / area required for the deployment of the virtual robot carried in the virtual robot creation request message. Among them, the resource request message carries GPU resource information, CPU resource information, memory information, resource cost information, etc.
[0132] Determine whether the copy operation can be performed based on the returned information. If yes, copy it. If the copy condition can be met, the central management node sends one or more AI data models of the virtual robot to one or more edge computing nodes in the target area network. Each AI data model can support the provision of capabilities of one or more virtual robots. An AI data model is downloaded and installed by an edge computing node. The set of capabilities provided by the AI data models installed by one or more edge computing nodes constitutes the total set of capabilities of this virtual computing person.
[0133] The central management node feeds back a virtual robot replication success response message to the user. The message carries the virtual robot ID, the capabilities provided by the virtual robot, and the AI data model deployment topology information of each capability of the virtual robot.
[0134] As shown in 5, based on the above embodiment, the migration of the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes:
[0135] Get a virtual robot migration request message; the virtual robot migration request message can come from the user's virtual robot migration request message, or the central management node migrates a virtual robot from the source network area to the new network area for deployment according to other trigger conditions. The virtual robot migration request message carries virtual robot ID information, source network location / area / edge computing node ID, target network location / area / edge computing node ID and other information.
[0136] Based on the ID information of the migrated virtual robot, the source network location or area, and the edge computing node ID information, a query request for the migrated virtual robot information is sent to the edge computing node at the location or area where the migrated virtual robot is deployed; when the central management node itself maintains all the information required to create this virtual robot as indicated by the virtual robot ID information (including virtual robot AI data model information, the topology of the nodes required for the innovative virtual robot, and the GPU resource information, CPU resource information, memory information, etc. of each point) and the central management node determines that the information it maintains is the latest version of the virtual robot required to create this virtual robot as indicated by the virtual robot ID information, this step can be omitted.
[0137] The edge management node where the virtual robot is located, indicated by the virtual robot ID information, replies to the central management node with the virtual robot information, including the virtual robot AI data model information, the topology of the nodes required for the innovative virtual robot, and the GPU resource information, CPU resource information, memory information, etc. of each point.
[0138] According to the location or area where the new virtual robot is deployed and the edge computing node ID information of the new virtual robot deployment, a resource query request is sent to the edge computing node in the area where the new virtual robot is deployed; if the virtual robot creation request message carries the edge computing node IDs of one or more virtual robot deployments indicated by the user, the central management node sends a resource request message to the edge computing nodes of one or more virtual robot deployments carried in the virtual robot creation request message. If the virtual robot creation request message does not carry the edge computing node ID information of the virtual robot deployment indicated by the user, the central management node sends a resource request message to one or more edge computing nodes in the location / area required for the deployment of the virtual robot carried in the virtual robot creation request message. Among them, the resource request message carries GPU resource information, CPU resource information, memory information, resource cost information, etc.
[0139] According to the returned information, it is determined whether the migration operation can be performed. If yes, the migration is performed; if the migration conditions can be met, the central management node sends one or more AI data models of the virtual robot to one or more edge computing nodes in the target area network. Each AI data model can support the provision of capabilities of one or more virtual robots. An AI data model is downloaded and installed by an edge computing node. The set of capabilities provided by the AI data models installed by one or more edge computing nodes constitutes the total set of capabilities of this virtual computing person.
[0140] A virtual robot information deletion request is sent to the edge computing node of the source location or area where the migrated virtual robot is deployed; the message carries the ID of the virtual robot.
[0141] After the deletion is successful, a migration success notification is sent to the user. The message carries the virtual robot ID (the same ID as the virtual robot before the migration can be used or a new ID can be reassigned), the capabilities provided by the virtual robot, and the AI data model deployment topology information of each capability of the virtual robot.
[0142] like Figure 6 Based on the above embodiments, the virtual robot is used for:
[0143] According to the service request message sent by the service client, the service is parsed and processed (the message carries the service type, service ID, and data content involved in the service (such as files, photos, signatures, etc.). When the service client and the virtual robot are co-located on the physical device, this step can be omitted.)
[0144] If the processing is successful, a business response is made directly. If the processing is unsuccessful (i.e. the business request information is not within the capabilities that virtual robot 1 can provide or virtual robot 1 needs to use a backup virtual robot service due to fault maintenance, current processing volume overload, etc.), a virtual robot capability query message is sent to the central management node; the message carries information such as the business type, business ID, and virtual robot capability requirements.
[0145] A business analysis processing request is sent to the backup virtual robot recommended by the central management node that meets the current business virtual robot capability requirements; the message carries the business type, business ID, and data content involved in the business (such as files, photos, signatures, etc.).
[0146] The processed data after parsing by the backup virtual robot is obtained and sent to the business client.
[0147] Based on the above embodiment, the business analysis process includes:
[0148] Perform preliminary verification on received business data to ensure data integrity and format correctness, and pre-process image or document data;
[0149] Automatically match applicable audit rules from the predefined audit rule engine based on business type and data content, and conduct preliminary verification of business data to check consistency and compliance;
[0150] Use the trained AI data model to identify anomalies, risk points, and inconsistencies in business data;
[0151] Summarize the rule engine matching results and AI data model analysis results to form an audit report.
[0152] Specifically:
[0153] 1) Data verification and preprocessing:
[0154] The virtual robot 1 performs preliminary verification on the received business data to ensure the integrity and correctness of the data format. It performs necessary preprocessing on image or document data, such as image cropping, rotation correction, denoising, etc., and OCR recognition and structuring of text data.
[0155] 2) Rule engine matching:
[0156] Apply the predefined audit rule engine to automatically match applicable audit rules based on business type and data content.
[0157] 3) AI model analysis:
[0158] Virtual robot 1 uses trained AI models (such as machine learning, deep learning models) to conduct in-depth analysis of data to identify abnormal patterns, potential risks, and inconsistencies.
[0159] 4) Summary of results and decision-making recommendations:
[0160] Summarize the rule engine matching results and AI model analysis results to form an audit report.
[0161] Provide decision-making suggestions based on the analysis results to help business personnel respond and process quickly.
[0162] Based on the above embodiments, when there are multiple virtual robots that meet the virtual robot capability requirements, the central management node feeds back a list of virtual robot IDs to the virtual robot 1, which can be sorted according to the recommended priority. Different virtual robots can form a virtual robot alliance (the formation conditions of the virtual robot alliance include virtual robots with close physical locations of edge computing nodes or virtual robots with data sharing contracts, etc.).
[0163] When the central management node prioritizes, it needs to comprehensively consider business needs to ensure that the selected virtual robot best meets the specific requirements of the current business. The following are examples of business demand factors that the central management node may consider when prioritizing:
[0164] Business type matching: The central management node will give priority to virtual robots that are designed or optimized to handle similar businesses based on the type of business request.
[0165] Business urgency: For urgent or time-sensitive business requests, the central management node will give priority to virtual robots that can respond and process quickly.
[0166] Business complexity: For businesses that are complex or require specific expertise, the central management node will give priority to virtual robots with expertise in related fields.
[0167] Business data security: For businesses involving sensitive information, the central management node will consider the data security records and security measures of the virtual robots, and give priority to virtual robots with good reputation and high security performance.
[0168] Historical business performance: The central management node will refer to the historical performance of the virtual robot in handling similar businesses in the past, including success rate, customer satisfaction, etc., to predict its future performance.
[0169] Business-specific rules: If the business has specific processing rules or standards, the central management node will give priority to virtual robots that can comply with these rules.
[0170] Business cost-effectiveness: While meeting business needs, the central management node will also consider cost-effectiveness and give priority to virtual robots with high cost-effectiveness to optimize overall costs.
[0171] Through such comprehensive considerations, the central management node can ensure that the selected virtual robot not only has the ability to handle specific business, but also can complete the task efficiently, safely and economically, thereby improving the business processing capability and customer satisfaction of the entire system. This detailed priority sorting mechanism is an important improvement of the present invention over the existing technical solutions, which makes the collaboration of virtual robots more intelligent and business-oriented.
[0172] Based on the above embodiments, the central management node selects a backup virtual robot that meets the current business virtual robot capability requirements and has the highest priority from the alliance domain for recommendation, wherein the alliance domain is: a virtual robot alliance generated by message instructions during the virtual robot creation phase or a virtual robot alliance generated by automated intelligence.
[0173] The method for virtual robots to form an alliance domain can be determined by the method one of stage "Ⅰ, creation of virtual robots" of the present invention, which includes "ID information of the virtual robot alliance that the virtual robot applies to join, and capability list information shared by the virtual robot in each virtual robot alliance" carried in the virtual robot creation request message sent by the user to the central management node, or by the method two, which includes "ID information of the virtual robot alliance that the virtual robot applies to join, and capability list information shared by the virtual robot in each virtual robot alliance" carried in the virtual robot registration message sent by the edge computing node to the central management node.
[0174] In addition to generating through message instructions in the "virtual robot creation" stage, the virtual robot alliance domain can also automatically generate intelligence based on the AI system architecture of distributed virtual robots.
[0175] The alliance domain composed of virtual robots can be managed by the alliance initiator (which may be one or more influential virtual robots) or the central management node. The specific methods are as follows:
[0176] 1) Goal and condition setting
[0177] Automated target identification: Determine the alliance's goals through algorithm analysis, such as optimizing resource utilization, improving task processing speed, capability complementarity, and load balancing.
[0178] Condition definition algorithm: Use preset standards and algorithms to automatically set the conditions for joining the alliance, such as computing power thresholds, storage requirements, etc.
[0179] 2) Automated formulation of policies and rules
[0180] Automatically generate interaction rules: Use machine learning models to automatically generate resource sharing protocols and task scheduling mechanisms based on historical data and performance indicators.
[0181] Dynamic role allocation: Dynamically allocate roles and responsibilities of virtual robots based on real-time performance data and load conditions.
[0182] 3) Automated member recruitment
[0183] Intelligent identification of potential members: Automatically identify virtual robots suitable for joining the alliance by analyzing network topology and performance logs.
[0184] Automated invitation process: The system automatically sends invitations to potential members and provides the necessary alliance information.
[0185] 4) Automated registration and review
[0186] Automatic registration: Potential members automatically submit registration information and proof of ability through the API interface.
[0187] Intelligent Audit: The central management node uses an automated audit algorithm to verify the accuracy and completeness of registration information.
[0188] 5) Alliance establishment and information synchronization
[0189] Automatic confirmation of joining: Once the registration review is passed, the system automatically updates the member status and grants corresponding permissions.
[0190] Real-time information synchronization: The capabilities and resource status of new members are shared with existing members through a real-time data synchronization mechanism.
[0191] 6) Efficient alliance operation
[0192] Automatic resource sharing: The system automatically allocates and adjusts resources based on the preset resource sharing agreement.
[0193] Dynamic task scheduling: Based on real-time load and priority, the system automatically schedules tasks to the most appropriate virtual robot.
[0194] Collaborative execution mechanism: Through API and message queue, efficient collaboration and data backup among members are achieved.
[0195] 7) Maintenance and performance evaluation
[0196] Real-time performance monitoring: Use monitoring tools to track the performance indicators of virtual robots in real time.
[0197] Periodic evaluation: Run the evaluation algorithm regularly and adjust the alliance configuration based on performance data and business needs.
[0198] 8) Expansion and dissolution mechanism
[0199] Intelligent expansion analysis: Based on business growth and market trends, the system automatically analyzes and makes expansion recommendations.
[0200] Condition-driven disbanding: preset disbanding conditions, such as goal achievement or performance degradation, and the system automatically executes the disbanding process.
[0201] Through these optimized steps, the establishment and operation of the virtual robot alliance domain will be more automated and intelligent, reducing human intervention and improving efficiency and reliability.
[0202] Taking AI auditing business as an example, the specific operation steps of "virtual robots provide mutual assistance, backup and other collaborative services through alliance domains" are as follows:
[0203] 1) Business request reception and verification: Virtual robot 1 receives requests from business clients and verifies the validity of the requests, including business type, ID, and data integrity.
[0204] 2) Data preprocessing and feature extraction: Preprocess the received data, such as image enhancement, text structuring, etc., and extract key features for subsequent analysis.
[0205] 3) AI rule matching and preliminary verification: Use the built-in AI audit rule engine to perform preliminary verification of data to check consistency and compliance.
[0206] 4) Deep learning model analysis: Use deep learning models to conduct in-depth analysis of data to identify abnormal patterns, risk points, and potential inconsistencies.
[0207] 5) Alliance domain collaboration request: When virtual robot 1 encounters a processing bottleneck or a specific business demand exceeds its capabilities, it initiates a collaboration request to the central management node.
[0208] 6) Central management node coordination: The central management node selects the most appropriate virtual robot 2 from the alliance domain for assistance based on the capabilities, load, and geographic location of the virtual robot, the alliance open capability list, and other information.
[0209] 7) Task distribution and execution: Virtual Robot 1 forwards the business request to the selected Virtual Robot 2, which performs specific audit tasks based on its professional capabilities.
[0210] 8) Result summary and report generation: After the virtual robot 2 completes the processing, it feeds the results back to the virtual robot 1, which summarizes all the information and generates a detailed audit report.
[0211] 9) Decision support and response: Virtual robot 1 provides decision support based on the summary results and returns the final processing results to the business client in a timely manner.
[0212] Improvements of the present invention:
[0213] 1) Improved collaboration efficiency: Through collaboration within the alliance domain, virtual robots can quickly respond to complex business needs, shorten processing time, and improve overall efficiency.
[0214] 2) Optimal utilization of resources: The dynamic task allocation and resource scheduling mechanism of the alliance domain ensures the optimal utilization of the computing resources of each virtual robot, reducing the waste of system resources.
[0215] 3) Enhanced system reliability: The alliance domain provides a multi-level backup and disaster recovery mechanism, so that the system can maintain stable operation even in the event of partial node failure.
[0216] 4) Flexibility and scalability of business processing: Virtual robots can be dynamically expanded or reduced according to business needs. Resource sharing and capability complementarity within the alliance domain enable the system to flexibly adapt to the ever-changing business environment.
[0217] 5) Depth and breadth of data processing: Virtual robots within the alliance domain can share advanced AI models and processing strategies, enabling the system to handle more complex data and business scenarios.
[0218] In summary, the present invention not only improves the automation level of AI auditing business, but also enhances the intelligent collaboration capability of the system, providing enterprises with a more efficient, reliable and intelligent auditing solution.
[0219] The present invention proposes that in a distributed virtual robot AI system, based on the collaboration of edge computing nodes and central management nodes, the virtual robot installation and upgrade, replication and migration, and mutual backup are iteratively completed, so as to realize the flexible call of system capabilities while avoiding the impact of single point failure on the system. The distributed virtual robot AI system is suitable for AI auditing systems and can also be used in other AI data processing systems.
[0220] The distributed virtual robot AI system described in the present invention introduces a more advanced central management node, uses a machine learning algorithm to optimize the coordination and resource management between nodes, establishes a virtual robot to process the business, and reduces the complexity of management; the present invention implements a dynamic resource allocation mechanism based on real-time load and capacity evaluation to ensure that resources are optimally utilized, and realizes the replication and migration of virtual robots to improve the overall system efficiency; the present invention provides an optimization strategy for cross-node business processing, and improves the processing efficiency and reliability through the introduction of a virtual robot alliance. Through these improvement measures, the scheme of the present invention aims to solve the problems existing in the distributed node deployment scheme of the prior art, and provide a more efficient, reliable and intelligent AI audit business processing capability; the distributed virtual robot system proposed in the present invention can also solve the cost problem in the existing AI audit business: the existing node deployment scheme requires a lot of hardware and maintenance costs. Virtual robots can use existing computing resources to reduce additional hardware investment. Through these improvements, the distributed virtual robot system of the present invention not only improves the processing capacity of AI audit business, but also reduces costs, enhances the flexibility and security of the system, and better meets the needs of modern business.
[0221] The present invention proposes to complete the installation and upgrade, replication and migration, and mutual backup of virtual robots in a distributed virtual robot AI system based on the collaboration of edge computing nodes and central management nodes, so as to achieve flexible call of system capabilities and avoid the impact of single point failures on the system. Its technical advantages are as follows:
[0222] 1. In the AI system architecture based on distributed virtual robots, the central management node only needs to maintain the information of one or more virtual robots, without deploying AI data models that support AI audit capabilities. This avoids the data processing load and waste of network transmission resources caused by the deployment of centralized AI audit capabilities.
[0223] 2. Using edge devices to deploy distributed virtual robots can fully integrate and utilize the computing resources of edge computing nodes, and because edge computing nodes are closer to users on the transmission path, data transmission delays can be saved.
[0224] 3. Virtual robots can be updated, upgraded, and maintained independently without affecting the business processing capabilities of the entire system.
[0225] 4. Virtual robots can be flexibly copied and migrated in the distributed virtual robot AI system to achieve large-scale promotion of personalized AI audit capabilities on the network.
[0226] 5. Virtual robots can collaborate with each other for backup to prevent single point failure.
[0227] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.
Claims
1. A distributed virtual robot AI system, characterized in that: include: Edge computing nodes, used to deploy virtual robots, which are a collection of one or more AI audit capabilities; The central management node is used to create, copy or migrate virtual robots based on user needs and in combination with multiple computing resource information of edge computing nodes.
2. The distributed virtual robot AI system according to claim 1, characterized in that: The step of creating the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes: Get the user's virtual robot creation request message; Sending a resource query request to one or more edge computing nodes at the location or area where the virtual robot is deployed according to the creation request message; Determine one or more AI data models of the virtual robot according to the resource message returned by the edge computing node, and send the AI data models to one or more edge computing nodes for deployment; Send a creation success notification to the user.
3. The distributed virtual robot AI system according to claim 2, characterized in that: Determining one or more AI data models of the virtual robot according to the resource message returned by the edge computing node, and sending the AI data models to one or more edge computing nodes for deployment includes: According to the resource messages returned by the edge computing nodes and the business needs of the virtual robot, an AI data model is selected from the predefined AI model library and sent to the edge computing nodes that match its resources for deployment; After deployment is complete, the AI data model is performance verified to ensure that it meets business requirements.
4. The distributed virtual robot AI system according to claim 1, characterized in that: The edge computing node is also used to combine multiple computing resource information of the node, create a virtual robot, and register it in the central management node.
5. The distributed virtual robot AI system according to claim 4, characterized in that: The step of combining multiple computing resource information of the node to create a virtual robot and registering it in the central management node includes: Determine the type and capabilities of virtual robots to be created based on business needs; Determine whether its own computing resources can support the operation of the virtual robot; Select an AI data model from the predefined AI template library based on the business needs analysis and resource assessment results, and train the model locally or in collaboration with other edge computing nodes. Integrate the trained AI data model into the virtual robot and assign a unique identifier to the newly created virtual robot; Send the registration information of the newly created virtual robot to the central management node.
6. The distributed virtual robot AI system according to claim 1, characterized in that: The step of replicating the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes: Get the virtual robot copy request message; According to the ID information of the copied virtual robot, a query request for the copied virtual robot information is sent to the edge computing node at the location or area where the copied virtual robot is deployed; Send a resource query request to the edge computing node in the location or area where the new virtual robot is deployed according to the location or area where the new virtual robot is deployed and the ID information of the edge computing node where the new virtual robot is deployed; Determine whether the copy operation can be performed based on the returned information. If so, copy it and send a copy success notification to the user.
7. The distributed virtual robot AI system according to claim 1, characterized in that: The migrating of the virtual robot according to user needs and in combination with multiple computing resource information of edge computing nodes includes: Get the virtual robot migration request message; Sending a query request for information of the migrated virtual robot to the edge computing node at the location or area where the migrated virtual robot is deployed according to the ID information of the migrated virtual robot, the source network location or area, and the edge computing node ID information; Send a resource query request to the edge computing node in the location or area where the new virtual robot is deployed according to the location or area where the new virtual robot is deployed and the ID information of the edge computing node where the new virtual robot is deployed; Determine whether the migration operation can be performed based on the returned information. If yes, perform the migration. Sending a virtual robot information deletion request to an edge computing node at a source location or region where the migrated virtual robot is deployed; After the deletion is successful, a migration success notification is sent to the user.
8. The distributed virtual robot AI system according to claim 1, characterized in that: The virtual robot is used to: According to the service request message sent by the service client, the service is parsed and processed. If the processing is unsuccessful, a virtual robot capability query message is sent to the central management node; Sending a service analysis processing request to a backup virtual robot recommended by the central management node that meets the capability requirements of the current service virtual robot; The processed data after parsing by the backup virtual robot is obtained and sent to the business client.
9. The distributed virtual robot AI system according to claim 8, characterized in that: The business analysis process includes: Perform preliminary verification of received business data to ensure data integrity and correct format. And pre-process the image or document data; Automatically match applicable audit rules from the predefined audit rule engine based on business type and data content, and conduct preliminary verification of business data to check consistency and compliance; Use the trained AI data model to identify anomalies, risk points, and inconsistencies in business data; Summarize the rule engine matching results and AI data model analysis results to form an audit report.
10. The distributed virtual robot AI system according to claim 8, characterized in that: The central management node selects a backup virtual robot that meets the current business virtual robot capability requirements and has the highest priority from the alliance domain for recommendation, wherein the alliance domain is: a virtual robot alliance generated by message instructions during the virtual robot creation phase or a virtual robot alliance generated automatically and intelligently.
Citation Information
Patent Citations
Deployment and migration system of virtual CDN in intelligent fusion identification network
CN112994937A
Network system, resource processing method and equipment
CN113676512A
Efficient deployment and scheduling policy for shared virtualized network functions in mobile edge networks
CN117395147A
Ai processing distribution method and system, and ai processing node
JP2022057428A
Edge node autonomy
US20240007547A1
