A method and system for multi-task collaboration based on digital employees

Abstract

The application discloses a kind of multi-task cooperative processing method and system based on digital employee.There is beneficial effect in at:The application combines "ant colony algorithm (initial scheduling) + LSTM prediction (trend perception) + model predictive control (dynamic optimization)" for the first time, realizes "static optimal + dynamic adaptation" of task scheduling, solves the problem that traditional scheduling is low in efficiency and cannot cope with fluctuations;The application fuses AINote "plug-in brain + global perception" ability, realizes the "multimodal interaction + cross-system cooperation + long-term memory" of digital employee, breaks through the skill limitation of single digital employee;The application embeds AI training / advisory / computing operation into the system, forms "tool + service" closed loop, reduces the threshold of enterprise AI landing, solves the problem that "digital employee is disconnected with actual demand of enterprise" in prior art.

Description

Technical Field

[0001] This invention relates to the field of AI artificial intelligence technology, and in particular to a multi-task collaborative processing method and system based on digital employees. Background Technology

[0002] With the penetration of artificial intelligence technology in enterprise digital transformation, digital employees, as tools to replace or assist humans in performing repetitive and standardized tasks, have been widely used in production, office, operation and maintenance scenarios.

[0003] However, existing multitasking collaboration solutions for digital employees still have the following core shortcomings:

[0004] 1. Inefficient task scheduling: Traditional scheduling (such as first-come-first-served and round-robin) adopts a static strategy, which does not take into account the complexity of tasks and the differences in the capabilities of digital employees, resulting in unbalanced load and low execution efficiency.

[0005] 2. Lack of dynamic adaptation: There is no feedback mechanism between scheduling decisions and actual execution, making it impossible to dynamically adjust based on resource usage and task progress, and making it difficult to cope with business fluctuations.

[0006] 3. Limited ability to collaborate: Digital employees are mostly "single-skill tools" and lack the ability to collaborate across tasks and systems, making them unable to handle complex business processes;

[0007] 4. Shortcomings in knowledge and security: Enterprises lack tacit knowledge accumulation and data interaction lacks security control. At the same time, digital employees lack long-term memory and self-optimization capabilities. Summary of the Invention

[0008] The purpose of this invention is to provide a multi-task collaborative processing method and system based on digital employees in order to solve the above-mentioned problems.

[0009] The present invention achieves the above objectives through the following technical solutions:

[0010] A multi-task collaborative processing method and system based on digital employees, comprising a five-layer functional architecture:

[0011] Architecture 1: Perception and Interaction Layer, which includes multimodal input interfaces (OCR / ASR / TTS), business data docking gateways and device status acquisition modules, supports natural language interaction triggered by text / voice, and perceives enterprise personnel flow, logistics, capital flow data and equipment operating status in real time;

[0012] Architecture 2: Skill Platform Layer, which encapsulates standardized skill components for digital employees, integrates the RAG knowledge base, skill registration and management engine, and supports skill reuse, knowledge accumulation and cross-task capability invocation;

[0013] Architecture 3: Task Collaboration Layer, which includes a task decomposer, an intelligent scheduling engine, and a collaborative communication module, to realize the subtask decomposition, dynamic scheduling, and state synchronization of complex tasks;

[0014] Architecture 4: Digital Employee Execution Layer, which includes a multi-role digital employee cluster, an execution monitoring module, and an exception feedback interface, executes sub-tasks and provides real-time status feedback;

[0015] Architecture 5: Service Assurance Layer, which includes AI training module, AI consulting module, and computing power management module, providing AI capability training, enterprise AI application planning, computing power operation and maintenance, and data security management.

[0016] Furthermore, the business data interface gateway of the perception and interaction layer supports OPCUA and RESTful API protocols, enabling interface with existing enterprise PLC equipment and ERP systems, as well as the reuse of computing power from historical IT assets.

[0017] Furthermore, the RAG knowledge base of the skill platform layer supports the incremental import and real-time update of enterprise tacit knowledge, and automatically accumulates the task execution experience of digital employees (including sub-task rules and exception handling strategies).

[0018] Furthermore, the intelligent scheduling engine of the task collaboration layer uses the ant colony algorithm to generate an initial scheduling scheme, and combines the resource trend prediction and model prediction control algorithm of the LSTM neural network to achieve dynamic scheduling optimization.

[0019] Furthermore, the multi-role digital employees in the digital employee execution layer include production quality inspectors, administrative assistants, equipment controllers, and data analysts. Each role of the digital employee executes sub-tasks by calling the corresponding skill components in the skill platform layer.

[0020] Furthermore, the computing power management module of the service assurance layer supports local deployment and private cloud leasing modes, and configures data anonymization, RBAC permission control and encryption confidentiality protocols to ensure that the enterprise's core data does not flow out of the controllable scope.

[0021] Furthermore, the processing method of this multi-task collaborative processing system based on digital employees includes the following steps:

[0022] Step 1: Task reception and decomposition. Receive user text / voice tasks and decompose them into a set of subtasks with complexity labels using a task decomposition tool.

[0023] Step 2: Initial schedule generation. Based on the resource requirements of subtasks and the performance parameters of digital employees, an ant colony algorithm is used to generate an initial task allocation scheme.

[0024] Step 3: Subtask execution. Multi-role digital employees call skill components to execute subtasks, and collect monitoring data such as CPU / memory / bandwidth in real time.

[0025] Step 4: Dynamic scheduling optimization, predicting resource trends through LSTM neural networks and adjusting the task allocation scheme using model predictive control algorithms;

[0026] Step 5: Results integration and knowledge accumulation. Integrate the results of sub-tasks and provide feedback to users, while storing the task experience in the RAG knowledge base.

[0027] Furthermore, the resource requirements of the subtasks in step 2 are obtained through a task complexity knowledge base, which stores the CPU, memory, and bandwidth requirement vectors corresponding to different task types.

[0028] Furthermore, the dynamic scheduling optimization in step 4 supports canary releases and retains the rollback entry for the old version's task allocation scheme.

[0029] Furthermore, it also includes a human-machine collaborative review step: after the digital employee performs standardized sub-tasks, the human employee reviews the low-confidence results, and the review results are synchronized to the RAG knowledge base to optimize the digital employee's capabilities.

[0030] The beneficial effects of this invention are as follows:

[0031] 1. This invention is the first to combine "ant colony algorithm (initial scheduling) + LSTM prediction (trend perception) + model predictive control (dynamic optimization)" to achieve "static optimization + dynamic adaptation" of task scheduling, solving the problems of low efficiency and inability to cope with fluctuations in traditional scheduling.

[0032] 2. This invention integrates AINote's "external brain + global perception" capabilities to achieve "multimodal interaction + cross-system collaboration + long-term memory" for digital employees, breaking through the skill limitations of a single digital employee;

[0033] 3. This invention embeds AI training / consulting / computing power operation and maintenance into the system, forming a closed loop of "tools + services", which lowers the threshold for enterprises to implement AI and solves the problem of "digital employees being out of touch with the actual needs of enterprises" in existing technologies. Attached Figure Description

[0034] Figure 1 This is a flowchart of a multi-task collaborative processing method and system based on digital employees as described in this invention. Detailed Implementation

[0035] A multi-task collaborative processing method and system based on digital employees, comprising a five-layer functional architecture:

[0036] Architecture 1: Perception and Interaction Layer, which includes multimodal input interfaces (OCR / ASR / TTS), business data docking gateways and device status acquisition modules, supports natural language interaction triggered by text / voice, and perceives enterprise personnel flow, logistics, capital flow data and equipment operating status in real time;

[0037] Architecture 2: Skill Platform Layer, which encapsulates standardized skill components for digital employees, integrates the RAG knowledge base, skill registration and management engine, and supports skill reuse, knowledge accumulation and cross-task capability invocation;

[0038] Architecture 3: Task Collaboration Layer, which includes a task decomposer, an intelligent scheduling engine, and a collaborative communication module, to realize the subtask decomposition, dynamic scheduling, and state synchronization of complex tasks;

[0039] Architecture 4: Digital Employee Execution Layer, which includes a multi-role digital employee cluster, an execution monitoring module, and an exception feedback interface, executes sub-tasks and provides real-time status feedback;

[0040] Architecture 5: Service Assurance Layer, which includes AI training module, AI consulting module, and computing power management module, providing AI capability training, enterprise AI application planning, computing power operation and maintenance, and data security management.

[0041] In this embodiment, the business data interface gateway of the perception interaction layer supports OPCUA and RESTfulAPI protocols, enabling interface with existing enterprise PLC equipment and ERP systems, as well as the reuse of computing power of historical IT assets.

[0042] In this embodiment, the RAG knowledge base of the skill platform layer supports incremental import and real-time updates of enterprise tacit knowledge, and automatically accumulates the task execution experience of digital employees (including sub-task rules and exception handling strategies).

[0043] In this embodiment, the intelligent scheduling engine of the task collaboration layer uses the ant colony algorithm to generate an initial scheduling scheme, and combines the resource trend prediction and model prediction control algorithm of the LSTM neural network to achieve dynamic scheduling optimization.

[0044] In this embodiment, the multi-role digital employees in the digital employee execution layer include production quality inspector, administrative assistant, equipment control, and data analyst. Each role of digital employee executes sub-tasks by calling the corresponding skill components of the skill platform layer.

[0045] In this embodiment, the computing power management module of the service guarantee layer supports local deployment and private cloud leasing modes, and is configured with data anonymization, RBAC permission control and encryption confidentiality protocols to ensure that the enterprise's core data does not flow out of the controllable scope.

[0046] In this embodiment, the processing method of the multi-task collaborative processing system based on digital employees includes the following steps:

[0047] Step 1: Task reception and decomposition. Receive user text / voice tasks and decompose them into a set of subtasks with complexity labels using a task decomposition tool.

[0048] Step 2: Initial schedule generation. Based on the resource requirements of subtasks and the performance parameters of digital employees, an ant colony algorithm is used to generate an initial task allocation scheme.

[0049] Step 3: Subtask execution. Multi-role digital employees call skill components to execute subtasks, and collect monitoring data such as CPU / memory / bandwidth in real time.

[0050] Step 4: Dynamic scheduling optimization, predicting resource trends through LSTM neural networks and adjusting the task allocation scheme using model predictive control algorithms;

[0051] Step 5: Results integration and knowledge accumulation. Integrate the results of sub-tasks and provide feedback to users, while storing the task experience in the RAG knowledge base.

[0052] In this embodiment, the resource requirements of the subtasks in step 2 are obtained through a task complexity knowledge base, which stores the CPU, memory, and bandwidth requirement vectors corresponding to different task types.

[0053] In this embodiment, the dynamic scheduling optimization in step 4 supports canary releases and retains the rollback entry for the old version task allocation scheme.

[0054] In this embodiment, a human-machine collaborative review step is also included: after the digital employee performs the standardized sub-task, the human employee reviews the low-confidence results, and the review results are synchronized to the RAG knowledge base to optimize the digital employee's capabilities.

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-task collaborative processing system based on digital employees, characterized in that: It includes a five-layer functional architecture: Architecture 1: Perception and Interaction Layer, which includes multimodal input interfaces (OCR / ASR / TTS), business data docking gateways and device status acquisition modules, supports natural language interaction triggered by text / voice, and perceives enterprise personnel flow, logistics, capital flow data and equipment operating status in real time; Architecture 2: Skill Platform Layer, which encapsulates standardized skill components for digital employees, integrates the RAG knowledge base, skill registration and management engine, and supports skill reuse, knowledge accumulation and cross-task capability invocation; Architecture 3: Task Collaboration Layer, which includes a task decomposer, an intelligent scheduling engine, and a collaborative communication module, to realize the subtask decomposition, dynamic scheduling, and state synchronization of complex tasks; Architecture 4: Digital Employee Execution Layer, which includes a multi-role digital employee cluster, an execution monitoring module, and an exception feedback interface, executes sub-tasks and provides real-time status feedback; Architecture 5: Service Assurance Layer, which includes AI training module, AI consulting module, and computing power management module, providing AI capability training, enterprise AI application planning, computing power operation and maintenance, and data security management.

2. The multi-task collaborative processing system based on digital employees according to claim 1, characterized in that: The business data interface gateway of the perception and interaction layer supports OPCUA and RESTfulAPI protocols, enabling it to interface with existing PLC equipment and ERP systems of enterprises, as well as reuse computing power of historical IT assets.

3. The multi-task collaborative processing system based on digital employees according to claim 1, characterized in that: The RAG knowledge base of the skill platform layer supports the incremental import and real-time update of enterprise tacit knowledge, and automatically accumulates the task execution experience of digital employees (including sub-task rules and exception handling strategies).

4. A multi-task collaborative processing system based on digital employees according to claim 1, characterized in that: The intelligent scheduling engine of the task collaboration layer uses the ant colony algorithm to generate an initial scheduling scheme, and combines the resource trend prediction and model prediction control algorithm of LSTM neural network to achieve dynamic scheduling optimization.

5. A multi-task collaborative processing system based on digital employees according to claim 1, characterized in that: The multi-role digital employees in the execution layer include production quality inspectors, administrative assistants, equipment controllers, and data analysts. Each role of the digital employee executes sub-tasks by calling the corresponding skill components in the skill platform layer.

6. A multi-task collaborative processing system based on digital employees according to claim 1, characterized in that: The computing power management module of the service guarantee layer supports local deployment and private cloud leasing modes, and is configured with data anonymization, RBAC permission control and encryption confidentiality protocols to ensure that the enterprise's core data does not flow out of the controllable scope.

7. A processing method for a multi-task collaborative processing system based on digital employees according to any one of claims 1-6, characterized in that: Includes the following steps: Step 1: Task reception and decomposition. Receive user text / voice tasks and decompose them into a set of subtasks with complexity labels using a task decomposition tool. Step 2: Initial schedule generation. Based on the resource requirements of subtasks and the performance parameters of digital employees, an ant colony algorithm is used to generate an initial task allocation scheme. Step 3: Subtask execution. Multi-role digital employees call skill components to execute subtasks, and collect monitoring data such as CPU / memory / bandwidth in real time. Step 4: Dynamic scheduling optimization, predicting resource trends through LSTM neural networks and adjusting the task allocation scheme using model predictive control algorithms; Step 5: Results integration and knowledge accumulation. Integrate the results of sub-tasks and provide feedback to users, while storing the task experience in the RAG knowledge base.

8. A multi-task collaborative processing system based on digital employees according to claim 7, characterized in that: The resource requirements of the subtasks in step 2 are obtained through a task complexity knowledge base, which stores the CPU, memory, and bandwidth requirement vectors corresponding to different task types.

9. A multi-task collaborative processing system based on digital employees according to claim 7, characterized in that: The dynamic scheduling optimization in step 4 supports canary releases and retains the rollback entry for the old version's task allocation scheme.

10. A multi-task collaborative processing system based on digital employees according to claim 7, characterized in that: It also includes a human-machine collaborative review step: after the digital employee performs the standardized sub-task, the human employee reviews the low-confidence results, and the review results are synchronized to the RAG knowledge base to optimize the digital employee's capabilities.

Images