Search

CN-122018472-A - Industrial intelligent agent collaborative decision-making system and method

CN122018472ACN 122018472 ACN122018472 ACN 122018472ACN-122018472-A

Abstract

The invention discloses an industrial intelligent agent collaborative decision-making system and a method, which relate to the field of industrial automation and artificial intelligence intersection, wherein the system comprises a data acquisition module for acquiring dynamic working condition signals in real time; the method comprises the steps of receiving dynamic working condition signals by a local intelligent agent module, generating action proposal signals representing self intention and resource requirements based on a built-in decision model, receiving the action proposal signals by a collaborative decision platform module, forming and sending negotiation instruction signals for initiating negotiation to relevant local intelligent agent modules, generating coordination strategy signals and feeding back the coordination strategy signals to the collaborative decision platform module when a global optimization module detects conflicts which cannot be autonomously solved, and storing industrial domain knowledge patterns and historical collaborative cases by a knowledge base module, providing optimized knowledge signals to the global optimization module and providing decision knowledge signals to the local intelligent agent modules. The invention can solve the problems of low industrial multi-agent cooperative efficiency and difficult global optimization.

Inventors

  • FAN YUJUN

Assignees

  • 中科慧智(北京)科技有限公司

Dates

Publication Date
20260512
Application Date
20260402

Claims (10)

  1. 1. An industrial agent collaborative decision-making system, comprising: the data acquisition module acquires industrial field working condition data at least comprising equipment states, production progress and material information in real time and forms dynamic working condition signals; The local intelligent agent module receives the dynamic working condition signals and generates action proposal signals representing self intention and resource requirements based on a built-in decision model; The collaborative decision-making platform module receives the action proposal signals, aggregates and conflict-detects the signals of the local intelligent agent modules, and forms and sends negotiation instruction signals for initiating negotiation to the relevant local intelligent agent modules; The global optimization module generates an optimization trigger signal and transmits the optimization trigger signal to the global optimization module when the global optimization module detects the conflict which cannot be autonomously solved, and the global optimization module operates a distributed optimization algorithm based on the optimization trigger signal, generates a coordination strategy signal and feeds the coordination strategy signal back to the collaborative decision platform module; And the knowledge base module is used for storing knowledge patterns and historical collaborative cases in the industrial field, providing optimized knowledge signals for the global optimization module and providing decision knowledge signals for the local intelligent agent module.
  2. 2. The industrial agent collaborative decision-making system according to claim 1, wherein the data acquisition module comprises a sensor unit deployed on physical equipment and an edge calculation unit located at the edge of a network, wherein the sensor unit senses multimode original physical signals such as vibration, temperature, current and visual images of the equipment in real time, the edge calculation unit is connected with the sensor unit, receives the multimode original physical signals, performs signal filtering, feature extraction and format standardization processing, fuses heterogeneous original data into uniform dynamic working condition signals with time stamps and data source identifiers, and periodically or event-triggered pushing the dynamic working condition signals to the local agent module through an industrial network protocol so as to provide a real-time, clean and structured data base for subsequent decisions.
  3. 3. The industrial agent collaborative decision-making system according to claim 1, wherein the local agent module is internally integrated with a reaction layer, a collaborative planning layer and a communication interface layer, the reaction layer responds to received emergency alert dynamic working condition signals in millisecond level based on a preset rule base to generate direct control instructions, the collaborative planning layer receives conventional dynamic working condition signals and decision knowledge signals from a knowledge base module, and simulates interaction results generated by self action proposals and other agents by using a model-based reinforcement learning algorithm to generate optimized action proposal signals, the communication interface layer is responsible for packaging the action proposal signals into messages conforming to established communication standards and sending the messages to a collaborative decision-making platform module, and is also responsible for receiving negotiation instruction signals and coordination strategy signals from outside, and the model-based reinforcement learning algorithm is as follows: ; Wherein, the For mathematical expectations, state space S and agent policy are traversed Environmental model ; / Mapping the industrial working condition state of the intelligent agent at the time t/t+1 by a dynamic working condition signal; / Action proposal action for the agent t time/t+1 time; is a prompt bonus function; Weighing the instant benefit and the future benefit as discount factors; Predicting the value of the future optimal action as an action value network; for the interaction penalty coefficient, balancing the autonomous decision of the intelligent agent and the cooperative constraint of multiple intelligent agents; to quantify the current intelligent action proposal for multi-intelligent interaction loss Candidate action set with other agents The greater the conflict, the higher the loss value is; outputting the probability distribution of the action proposal under a given state for the agent policy network; and (3) an environment model for simulating the state transition rule of interaction of the industrial working conditions and the multiple agents.
  4. 4. The industrial agent collaborative decision-making system according to claim 1, wherein a virtual collaborative space and a conflict detection engine are implemented in the collaborative decision-making platform module, the virtual collaborative space is used as a shared information exchange medium to continuously receive and broadcast action proposal signals from local agent modules to form a global situation view, the conflict detection engine analyzes the global situation view in real time, and recognizes competitive and mutually exclusive conflicts existing on equipment, materials or productivity by comparing resource requirements, space occupation and time windows in different action proposal signals, when the recognized conflicts meet preset simple conflict rules, the conflict detection engine directly generates a negotiation command signal containing a conflict party identifier and a proposal solution, and when complex or circularly dependent conflicts are recognized, the conflict detection engine generates the optimization trigger signal.
  5. 5. The industrial agent collaborative decision-making system according to claim 1, wherein the global optimization module, upon receiving an optimization trigger signal, initiates a distributed optimization algorithm based on a game theoretical nash equilibrium search, the distributed optimization algorithm being as follows: ; Wherein, the Parameters (such as resource allocation amount, execution time window, capacity allocation ratio) are adjusted for the action proposals of n conflict agents, The adjustment parameter is the ith agent; as a global objective function, synthesizing industrial core indexes as a minimum objective; As a weight coefficient, dynamically adjusting according to industrial production requirements; The energy consumption of each intelligent agent action proposal is calculated in a superposition way for the total energy consumption of the system; the system is the overall production efficiency of the system, and is a comprehensive quantification index of the utilization rate of equipment, the completion rate of tasks and the material circulation efficiency; In order to conflict with the collection of agents, A feasible domain of action parameters for the ith agent; For the i-th agent's revenue function (e.g. equipment utilization, task completion revenue), The parameters of other agents except i are constrained to be Nash equilibrium core conditions, namely that any agent can not promote the income of the agent by independently adjusting the parameters, namely, all agents accept the current solution; The global optimization module calls the optimized knowledge signals provided by the knowledge base module as algorithm prior information, seeks a pareto improved solution which can be accepted by all conflict parties through multiple iterative computations, finally packages the resource reallocation scheme or execution sequence adjustment scheme obtained by computation into the coordination strategy signals, and sends back to the collaborative decision platform module for issuing, wherein the decision formula of the pareto improved solution is as follows: ; Wherein, the The revenue variance sets for all conflicting agents, A change in revenue for the ith agent; Is an industry tolerance threshold; as the amount of change of the global objective function, <0 Indicates that the global overall cost is reduced from that before optimization, i.e., the global objective is optimized.
  6. 6. The industrial agent collaborative decision-making system according to claim 1, wherein the knowledge base module comprises an industrial knowledge graph sub-base and a collaborative case base, wherein the industrial knowledge graph sub-base stores equipment process association, material conversion relations and production constraint rules in a graph structure, the provided decision knowledge signals are used for guiding a local agent module to generate a proposal conforming to a process route, the provided optimized knowledge signals are used for defining a feasible strategy search space for a global optimization module, the collaborative case base records historical collaborative processes and results in a structured log form, the characteristics of successful cases and the sources of failure cases are included, and the cases are clustered and labeled through a machine learning model, so that the provided knowledge signals can be matched and pushed according to the similarity of current dynamic working conditions, and multiplexing and inheritance of experience knowledge are achieved.
  7. 7. The industrial agent collaborative decision-making system according to claim 1, wherein the action proposal signal generated by the local agent module has a standardized data structure, the structure at least comprises proposal agent identity, proposal target description, required resource list and quantity and time requirements thereof, expected start and end time stamps, proposal priority weights and proposal validity period, the structured signal design enables the collaborative decision-making platform module to mechanically analyze and compare the proposal, and clear optimization variables and constraint conditions are provided for the global optimization module, thereby guaranteeing the disambiguation of information transfer and the high efficiency of processing among different modules.
  8. 8. The industrial agent collaborative decision-making system according to claim 1, further comprising a digital twin simulation sandbox module respectively connected to the collaborative decision-making platform module and the local agent modules, wherein the digital twin simulation sandbox module receives global situation view signals from the collaborative decision-making platform module and decision model copies of each local agent module, performs accelerated deduction and pressure testing on various collaborative strategies in a virtual environment, and feeds back performance evaluation signals and potential risk early warning signals obtained by deduction to the collaborative decision-making platform module and the relevant local agent modules for verification and optimization decision before actual execution, thereby forming a closed loop of "decision-simulation-optimization".
  9. 9. The industrial agent collaborative decision-making system according to claim 1, further comprising a security and authorization management module, wherein the security and authorization management module is responsible for identity authentication, authority control and operation audit of all agents in the system, supports a global-regional-local three-level management architecture, realizes dynamic authority allocation based on three dimensions of resource-operation-scene, and is linked with an agent credit level, builds a full-flow closed-loop authorization mechanism of 'application-approval-execution-monitoring-recovery', the security and authorization management module comprises an identity authentication unit, an authority control unit and an audit tracking unit, the identity authentication unit realizes two-way authentication of the agents based on digital certificates and dynamic tokens, the authority control unit supports a global-regional-local three-level management mode, wherein the global-level responsible system policy makes and cross-regional coordination, the regional-level responsible system policy management, the local-level responsible device-level operation authority management, the authority control unit further realizes a model based on three dimensions of 'resource-operation-scene', the dimensions comprise devices, data, network resources, the operation dimensions comprise an emergency control dimension, the authority-writing, the authority-level is better than the authority level is better than the priority is better than the authority level, and the authority is better than the authority is better controlled by the authority level, and is better than the authority is better controlled by the dynamic, and is better than the authority is better in relation with the real-time with the credit and has.
  10. 10. A method of an industrial agent collaborative decision-making system according to claims 1-9, comprising: s1, acquiring industrial field working condition data at least comprising equipment state, production progress and material information in real time, and forming a dynamic working condition signal; s2, receiving the dynamic working condition signals, and generating action proposal signals representing self intention and resource requirements based on a built-in decision model; S3, receiving the action proposal signals, carrying out aggregation and conflict detection on the signals of a plurality of local intelligent agent modules, forming and sending negotiation instruction signals for initiating negotiation to the relevant local intelligent agent modules; S4, when the conflict which cannot be autonomously solved is detected, generating an optimization trigger signal and transmitting the optimization trigger signal to the global optimization module, wherein the global optimization module operates a distributed optimization algorithm based on the optimization trigger signal, generates a coordination strategy signal and feeds back the coordination strategy signal to the collaborative decision platform module; and S5, storing the knowledge graph and the historical collaborative case of the industrial field, providing an optimized knowledge signal for the global optimization module, and providing a decision knowledge signal for the local agent module.

Description

Industrial intelligent agent collaborative decision-making system and method Technical Field The invention relates to the crossing field of industrial automation and artificial intelligence, in particular to an industrial intelligent agent collaborative decision-making system and method. Background Currently, with the deep development of intelligent manufacturing and industrial internet, an industrial production system is increasingly complex, and a decision process of the industrial production system is evolving from traditional centralized control to distributed intelligent. In this context, multi-intelligent system technology is introduced into the industry in an effort to increase the flexibility and autonomy of the system by abstracting physical devices or logical units into agents with sensing, computing and decision making capabilities. However, at the practical industrial application level, existing multi-agent based solutions still face serious challenges. Firstly, in the architecture level, many systems do not thoroughly get rid of the centralized thought, and still rely on a powerful central controller to perform task allocation and scheduling, so that the autonomy of an intelligent agent is limited, the expansibility of the system is poor, and a central node becomes a performance bottleneck and a single point of failure source. Secondly, in the cooperative mechanism level, the intelligent agents often lack efficient and standard interactive protocols and negotiation mechanisms, and simple information transmission or response based on fixed rules are often adopted, so that the problems of complex resource competition and target conflict are difficult to process, and local optimum or decision-making dead offices are easy to trap. Again, at the decision quality level, the decision of the agent depends mostly on real-time data and preset models, lacks systematic utilization of deep knowledge (such as process principles, equipment association) and historical collaborative experience in the industrial field, results in that the decision is technically feasible but not necessarily economical or globally optimal, and is difficult to cope with rare working conditions. Finally, in the aspect of dynamic adaptability, when the existing system faces sudden disturbance such as order change, equipment failure and the like, the original plan is generally required to be interrupted for global rescheduling, the response is slow, and the production continuity is affected. Although there have been studies attempting to introduce contractual networks, game theory, etc. to improve synergy, it is often difficult to land due to excessive computational complexity or deviation from industry practical constraints. Therefore, a new collaborative decision-making system architecture and method that can ensure high autonomy and rapid response of each agent, realize effective global coordination and optimization, and integrate industrial knowledge and data is needed in the industry. Disclosure of Invention In view of the above drawbacks of the prior art, the present invention is directed to providing an industrial intelligent agent collaborative decision-making system and method, which are used for solving the problems of low industrial multi-intelligent agent collaborative efficiency and difficult global optimization. The present invention solves this problem by constructing a hierarchical distributed system. The data acquisition module at the bottom layer of the system and the local agent module are responsible for sensing and quick response, the collaborative decision-making platform module at the upper layer provides a virtual collaborative space, action proposals of agents are converged, conflict detection is carried out, and autonomous negotiation among agents is driven. For complex conflicts, a global optimization module starts a distributed algorithm to coordinate. The knowledge base module provides field knowledge and historical experience support for decision making and optimization throughout. By the aid of the mechanism of 'autonomous negotiation is dominant, global optimization is auxiliary and knowledge is continuously driven', the system approaches global optimization while maintaining flexibility, and efficient, self-adaptive and evolutionary industrial collaborative decision-making is realized. The invention provides an industrial intelligent agent collaborative decision-making system, which comprises: The data acquisition module acquires industrial field working condition data at least comprising equipment states, production progress and material information in real time and forms dynamic working condition signals; the local intelligent agent module receives the dynamic working condition signals and generates action proposal signals representing self intention and resource requirements based on the built-in decision model; The collaborative decision-making platform module receives the action proposal signals,