A multi-resolution modeling method and system based on multi-agent

Abstract

The present invention relates to the technical field of multi-resolution modeling, and proposes a multi-resolution modeling method and system based on multi-agents, which include: constructing a multi-agent system consisting of a perception decision-making agent and multiple model agents; wherein the perception decision-making agent is used to make multi-resolution switching scheduling decisions for any model agent; the model agent is used to perform simulation operations, and use different resolution models according to the decision switching of the perception decision-making agent; obtain resource usage status data of the simulation server, and the resolution information of all model agents currently running; the perception decision-making agent outputs a decision signal including the resolution model that each model agent should select next based on a multi-resolution switching algorithm; each model agent selects to use a target resolution model for performing multi-resolution modeling according to the received decision signal, and outputs the simulation result.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-resolution modeling, and more specifically, to a multi-resolution modeling method and system based on multi-agent. Background Art

[0002] Digital twin is the digital twin of physical entity, which is a real-time dynamic model that can reflect the entire life cycle of physical system. Compared with ordinary simulation models, digital twin has more complete mapping and stronger real-time performance. It can synchronously monitor the real-time status of objects with the help of sensors on physical entities or simulated data, and return the data we want to the physical entity to cause the actual state of the object to change. It is equivalent to realizing a one-to-one modeling mapping between physical entities and digital models. It can integrate data of multiple dimensions and present and feedback the actual state of objects in almost real time. The main components of the basic concept model of digital twin include physical entities in physical space, digital models in virtual space, and data connecting physical entities and digital models. Physical entities and digital models are connected and react to each other through data. Physical entities can transmit data of their own environment to digital models for presentation. We can also manipulate data in digital models to have actual impact on physical entities. At present, digital twin technology has not been widely used. One of the key bottlenecks is the contradiction between the high requirements of digital twin technology for system real-time and realism and the limited resources of simulation platforms. As an effective way to resolve the contradiction between massive data and limited software and hardware resources, multi-resolution modeling is an effective means to break through the bottleneck of digital twins.

[0003] Multi-resolution modeling is one of the frontier topics in current modeling and simulation. With the expansion of simulation scale and the improvement of simulation fidelity, fixed-resolution modeling has been difficult to effectively solve the contradiction between simulation complexity and limited resources in the simulation process. Multi-resolution modeling and simulation technology has become an inevitable trend in the development of system modeling and simulation. Its core idea is to represent data information as a series of different levels from coarse to fine. When users need general information over a large range, large-scale information is provided, and when users need local fine data, small-scale information is provided, thereby meeting user requirements with minimal consumption of software and hardware platform resources. In addition, multi-resolution is now widely used in image and video. Research in the military field mainly focuses on modeling theory and modeling technology for specific business fields, such as force simulation, radar simulation, communication simulation, and combat simulation of a certain type of arms.

[0004] At present, multi-resolution modeling is mainly based on the viewpoint selection method. In the viewpoint selection method, the underlying model always runs at the highest resolution, provides different resolution information at output, and switches based on the set resolution switching rules. This method is relatively simple to implement, but the system complexity and computer resource consumption in this method are relatively high, and the flexibility is poor, which is not conducive to exploratory analysis. Summary of the invention

[0005] In order to overcome the defects of the above-mentioned prior art such as high system complexity, high consumption of computer resources, and high flexibility, the present invention provides a multi-agent based multi-resolution modeling method and system.

[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0007] A multi-resolution modeling method based on multi-agent comprises the following steps:

[0008] S1. Construct a multi-agent system consisting of a perception and decision-making agent and multiple model agents;

[0009] The perception decision agent is used to make a switching scheduling decision for multiple resolutions of any model agent; the model agent is used to perform simulation operations and switch to use different resolution models according to the decision of the perception decision agent;

[0010] S2, obtaining the resource usage status data of the simulation server and the resolution information of all model agents currently running;

[0011] S3, the perception decision agent outputs a decision signal including the resolution model that each model agent should select next based on the multi-resolution switching algorithm according to the resource usage status data of the simulation server and the resolution information of all model agents currently running;

[0012] S4. Each model agent selects a target resolution model to perform multi-resolution modeling based on the received decision signal and outputs the simulation results.

[0013] As a preferred solution, the multi-resolution switching algorithm includes a small-scale multi-resolution switching algorithm and a large-scale multi-resolution switching algorithm.

[0014] As a preferred solution, the model agent includes several resolution models, as well as a model consistency maintenance mechanism, a model resolution information sharing mechanism and a model switching response mechanism; wherein the model consistency maintenance mechanism is used to maintain the consistency of the state information of the model agent when different resolution models are switched for use; the model resolution information sharing mechanism is used to share the model information, usage status information and usage constraint information of the resolution model; and the model switching response mechanism is used to respond to received decision signals.

[0015] As a preferred solution, in the step S3, the step of the perception decision agent outputting a decision signal based on a multi-resolution switching algorithm according to the resource usage status data of the simulation server and the resolution information of all model agents currently running includes:

[0016] S3.1. Prioritize the construction and use of high-resolution models;

[0017] S3.2. According to the resource usage status data of the simulation server, analyze the resource usage of the subsystems in the simulation server one by one to determine whether they need to be replaced with low-resolution models:

[0018] If so, a low-resolution model is selected from the resolution models, and the input and output of each low-resolution model in the subsystem are compared and used, and the resolution models with consistent input and output in the corresponding subsystem are combined to generate a low-resolution mathematical model; at the same time, the high-resolution model is used as the high-resolution mathematical model, and the low-resolution mathematical model and the high-resolution mathematical model are used to simulate multi-resolution modeling in the subsystem respectively, and then the multi-resolution modeling results are compared and verified, and the low-resolution mathematical model that meets the consistency of the multi-resolution modeling results is used as the resolution model that the model intelligent agent should select in the next step to form a decision signal output;

[0019] Otherwise, the output maintains the decision signal using the high-resolution model.

[0020] As a preferred solution, the step S1 further includes the following steps: expanding the model agent based on the DEVS specification to construct a new Agent DEVS atomic model; the expression thereof is as follows:

[0021] AM=(X,Y,P,A,δ int ,δ ext ,λ,ta)

[0022] Where, X = {(ip,im)|ip∈IPorts,im∈Msg} represents the set of input ports ip and their input information im, IPorts represents the set of input ports, Msg is the message set of model agents, including the interaction information between different model agents; Y = {(op,om)|op∈OPorts,om∈Msg} represents the set of output ports op and their output information om, OPorts represents the set of input ports; P is the feature set of model agents; A is the set of model agents; δ int represents the internal transfer function; δ ext represents the external transfer function; λ represents the output function, which is triggered when the internal state is transferred; ta represents the time advancement function.

[0023] As a preferred solution, the model agent set A includes:

[0024] The perceptron of the model agent is used to perceive external input and internal feedback and generate processor input;

[0025] Processors, used to abstract, learn, and reason about input data;

[0026] Effectors, which are used to complete the internal feature transfer and produce responses to the environment;

[0027] and, knowledge base;

[0028] The feature set P of the model agent includes the identity of the model agent, the name of the model agent, the location of the model agent, the state of the model agent and the possible output values ​​of the current features of the model agent;

[0029] The message set Msg of the model agent includes message identification, message sender information, message receiver information, message type, message content, message association time and message validity period.

[0030] Furthermore, the present invention also proposes a multi-agent based multi-resolution modeling system, which applies the multi-agent based multi-resolution modeling method proposed in any of the above technical solutions. It includes:

[0031] Simulation server;

[0032] A perception and decision agent, which is used to make switching scheduling decisions for multiple resolutions of any model agent;

[0033] and, a plurality of model agents for performing simulation operations and switching to different resolution models according to the decisions of the perception decision agent;

[0034] The perception decision agent outputs a decision signal including the resolution model that each model agent should select next based on a multi-resolution switching algorithm according to the resource usage status data of the simulation server and the resolution information of all model agents currently running;

[0035] Each model agent selects the target resolution model to perform multi-resolution modeling based on the received decision signal and outputs the simulation results.

[0036] As a preferred solution, the model agent includes several resolution models, as well as a model consistency maintenance mechanism, a model resolution information sharing mechanism and a model switching response mechanism; wherein the model consistency maintenance mechanism is used to maintain the consistency of the state information of the model agent when different resolution models are switched for use; the model resolution information sharing mechanism is used to share the model information, usage status information and usage constraint information of the resolution model; and the model switching response mechanism is used to respond to received decision signals.

[0037] As a preferred solution, the model agent is expanded based on the DEVS specification, and its expression is as follows:

[0038] AM=(X,Y,P,A,δ int ,δ ext ,λ,ta)

[0039] Where, X = {(ip,im)|ip∈IPorts,im∈Msg} represents the set of input ports ip and their input information im, IPorts represents the set of input ports, Msg is the message set of model agents, including the interaction information between different model agents; Y = {(op,om)|op∈OPorts,om∈Msg} represents the set of output ports op and their output information om, OPorts represents the set of input ports; P is the feature set of model agents; A is the set of model agents; δ int represents the internal transfer function; δ ext represents the external transfer function; λ represents the output function, which is triggered when the internal state is transferred; ta represents the time advancement function.

[0040] As a preferred solution, the system further includes an interactive interface, and data is exchanged between the perception decision agent and the model agent via the interactive interface.

[0041] Compared with the prior art, the beneficial effects of the technical solution of the present invention are: the present invention combines a perception decision-making agent and multiple model agents into a multi-agent system, solves the problems of work scheduling and information interaction among multiple agents, and incorporates the multi-agent system into a unified resource description framework of a multi-resolution model, which can reduce system complexity, reduce computer resource consumption through resource scheduling, and improve usage flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flowchart of a multi-agent based multi-resolution modeling method according to an embodiment of the present invention.

[0043] Figure 2 Schematic diagram of the perception decision-making agent according to an embodiment of the present invention.

[0044] Figure 3 Schematic diagram of the model agent of an embodiment of the present invention.

[0045] Figure 4 The present invention is a flowchart of outputting a decision signal based on a multi-resolution switching algorithm according to an embodiment of the present invention.

[0046] Figure 5 This is a flow chart of the overall simulation of arms combat according to an embodiment of the present invention.

[0047] Figure 6 Schematic diagram of a multi-agent based multi-resolution modeling system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0049] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;

[0050] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0051] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0052] Example 1

[0053] This embodiment proposes a multi-resolution modeling method based on multi-agents, such as Figure 1 As shown, it is a flowchart of the multi-agent based multi-resolution modeling method of this embodiment.

[0054] The multi-agent based multi-resolution modeling method proposed in this embodiment includes the following steps:

[0055] S1. Construct a multi-agent system consisting of a perception and decision-making agent 2 and multiple model agents 4.

[0056] In this embodiment, the perception decision agent 2 is used to make switching scheduling decisions for multiple resolutions of any model agent 4.

[0057] The model agent 4 is used to perform simulation operations and switch to different resolution models according to the decision of the perception decision agent 2.

[0058] S2. Obtain resource usage status data of the simulation server 1 and resolution information of all model agents 4 currently running.

[0059] S3. The perception decision agent 2 outputs a decision signal including the resolution model that each model agent 4 should select next based on the multi-resolution switching algorithm according to the resource usage status data of the simulation server 1 and the resolution information of all model agents 4 currently running.

[0060] S4. Each model agent 4 selects to use the target resolution model for performing multi-resolution modeling according to the received decision signal, and outputs the simulation results.

[0061] In this embodiment, adaptive multi-resolution scaling is achieved through autonomous learning of the perception decision agent 2 and the model agent 4, thereby coordinating the resource utilization efficiency of the simulation server 1 and the realism of the model operation while ensuring the real-time performance of the simulation system, so that server resources can be fully utilized.

[0062] The perception decision agent 2 is responsible for perceiving information such as the server status and the model agent 4 status, making decisions on the resolution of each model agent 4, and sending decision signals to control the resolution switching of the model agent 4, thereby optimizing the computing efficiency of the simulation server 1. The perception decision agent 2 contains the multi-resolution status of the model agent 4, the constraint relationship between the model agents 4, the status of the simulation server 1, etc. Based on this information, the resolution that the decision model agent 4 should choose is selected to achieve the purpose of optimizing the resource utilization efficiency of the simulation server 1.

[0063] like Figure 2 The figure shows the architecture diagram of the perception decision-making agent of this embodiment.

[0064] In an optional embodiment, the multi-resolution switching algorithms preset in the perception decision-making agent 2 include a small-scale multi-resolution switching algorithm (Small Multi-Resolution Switching, SMRS) and a large-scale multi-resolution switching algorithm (Large Multi-Resolution Switching, LMRS) to realize the automatic switching of multiple resolutions of the model agent 4 and the adaptive multi-resolution scaling function, so as to achieve the purpose of balancing the high resource utilization efficiency of the simulation server 1 and the high simulation capability of the model operation.

[0065] The model agent 4 in this embodiment is responsible for performing simulation calculations, transmitting the state of the model to the perception decision agent 2, and responding to the model resolution indication from the perception decision agent 2, maintaining and resolving its own state consistency problems caused by resolution switching.

[0066] Furthermore, the model agent 4 includes several resolution models, as well as a model consistency maintenance mechanism, a model resolution information sharing mechanism and a model switching response mechanism.

[0067] Among them, the model consistency maintenance mechanism is used to maintain the consistency of the state information of the model agent 4 when switching between models with different resolutions, so as to ensure the rationality of its switching operation.

[0068] The model resolution information sharing mechanism is used to share the model information, usage status information and usage constraint information of the resolution model, such as how many resolutions the model has, which resolution the current model agent 4 is running, and whether there are constraints and restrictions between the model resolution and other models.

[0069] The model switching response mechanism is used to respond to the received decision signal.

[0070] like Figure 3 The figure shows the architecture diagram of the model agent of this embodiment.

[0071] In an optional embodiment, while constructing a multi-agent system consisting of a perception decision-making agent 2 and multiple model agents 4, this embodiment expands the model agent 4 based on the DEVS specification to construct a new AgentDEVS atomic model.

[0072] In order to encapsulate model agents 4 with multiple resolutions, we need to incorporate the multi-agent system into the unified resource description framework of the multi-resolution model. Since the traditional discrete time system specification (DEVS) does not consider the modeling of model agents 4, we need to expand the DEVS specification and establish a new Agent DEVS atomic model. The main extensions are: expanding the state tuple S to become a tuple P with Agent characteristics; adding Agent model tuple A to reflect individual intelligence; expanding the input and output of the model port to Agent message types to reflect social collaboration.

[0073] Specifically, the new Agent DEVS atomic model obtained after expansion is described as the following 8-tuple structure:

[0074] AM=(X,Y,P,A,δ int ,δ ext ,λ,ta)

[0075] In the formula, X={(ip,im)|ip∈IPorts,im∈Msg} represents the set of input port ip and its input information im, IPorts represents the input port set, and Msg is the message set of model agent 4, including the interaction information between different model agents 4.

[0076] Y={(op,om)|op∈OPorts,om∈Msg} represents the set of output ports op and their output information om, and OPorts represents the set of input ports.

[0077] P is the feature set of model agent 4; A is the set of model agent 4.

[0078] δ int :P→P represents the internal transfer function, which is triggered when e=ta(p), where e is the duration of the current feature p.

[0079] δ ext :Q×X→P represents an external transfer function, which is triggered when there is an external input event, and Q={(p,e)|p∈P,e∈[0,ta(p)]}, where X is the input event.

[0080] λ:P→Y represents the output function, which is triggered when the internal state is transferred, and Y is the output time.

[0081] ta represents the time advancement function.

[0082] The model agent 4 set A, the feature set P of the model agent 4, and the message set Msg of the model agent 4 in the Agent DEVS atomic model are further explained:

[0083] Set A is a tuple that represents the individual agents of the model, and its structure is:

[0084] A=(Ap,Disp,Op,K)

[0085] Ap is the sensor of the model agent, which is responsible for sensing external input and internal feedback and generating processor input; Disp is the processor, which is responsible for abstracting, learning and reasoning the input data and operating the knowledge base; Op is the effector, which is responsible for completing the internal feature transfer and generating responses to the environment; K is the knowledge base owned by the Agent, which is used to store domain knowledge, communication knowledge, control knowledge, etc., to support the agent's calculation and input and output.

[0086] The feature set P of model agent 4 is a tuple that reflects the basic features of the model agent, and its structure is:

[0087] P=(ID,Name,Location,State,FeasibleV)

[0088] Where ID is the identity of the Agent, Name is the name of the Agent, Location is the location of the Agent, State is the state of the Agent, and FeasibleV is the possible output value of the current feature of the Agent.

[0089] The message set Msg of model agent 4 is the interactive information between different models and is the key to reflecting the social collaboration of Agent. Its structure can be described as:

[0090] Msg=(D,Source,Dest,Type,Content,Time,ValidT)

[0091] Where D is the identifier of the message, Source is the message sender, Dest is the message receiver, Type is the message type, Content is the message content, Time is the time associated with the message, and ValidT is the validity period of the message.

[0092] The set A constructs the specific functional framework of the Agent model, the feature set P distinguishes the basic features of different Agent models, and the message set Msg determines the information format of the interaction between different models. The three constitute the main body of the AgentDEVS atomic model.

[0093] In another optional embodiment, the step of the perception decision agent 2 outputting a decision signal based on a multi-resolution switching algorithm according to the resource usage status data of the simulation server 1 and the resolution information of all model agents 4 currently running includes:

[0094] S3.1. Prioritize the construction and use of high-resolution models;

[0095] S3.2. According to the resource usage status data of the simulation server 1, the resource usage of the subsystems in the simulation server 1 is analyzed one by one to determine whether it needs to be replaced with a low-resolution model:

[0096] If yes, a low-resolution model is selected from the resolution models, and the input and output of each low-resolution model in the subsystem are compared and used, and the resolution models with consistent input and output in the corresponding subsystem are combined to generate a low-resolution mathematical model;

[0097] At the same time, the high-resolution model is used as the high-resolution mathematical model, and the low-resolution mathematical model and the high-resolution mathematical model are used to simulate multi-resolution modeling in the subsystem respectively, and then the multi-resolution modeling results are compared and verified. The low-resolution mathematical model that meets the consistency of the multi-resolution modeling results is used as the resolution model that the model intelligent agent 4 should select in the next step to form a decision signal output.

[0098] Otherwise, the output maintains the decision signal using the high-resolution model.

[0099] like Figure 4 As shown, it is a flow chart of outputting a decision signal based on a multi-resolution switching algorithm of this embodiment.

[0100] The core goal of this step is to make full use of system resources as much as possible while ensuring the real-time performance of the parallel battlefield simulation system, and to use more high-fidelity and high-resolution models as soon as possible. It can be described in the following optimization language:

[0101] Max{VerFun(hm,lm)}

[0102] St RUrat(cpu,mem)→T1

[0103] RealtimeRatio≥rt

[0104] This embodiment uses the method of "multi-resolution consistent modeling" and constructs a multi-resolution model by deriving different resolution models from each other. In the whole process of "conceptual model-mathematical model-software model-model verification", various indicators such as input and output characteristics at each stage are analyzed for differences and verified against each other to achieve the optimization goal.

[0105] In the specific implementation process, after the simulation is run, the perception decision agent 2 uses the model resolution selection decision algorithm to determine the model resolution that each model agent 4 should choose to run by perceiving the simulation server 1 and the model agent 4, and sends this information to each model agent 4. After receiving the resolution information, the model agent 4 responds to it, while maintaining the data consistency of its own multi-resolution model to advance the simulation process.

[0106] Furthermore, the simulation main process outputs the simulation process and result information feedback to the human-computer interaction interface for simulation personnel to review and perform interactive control.

[0107] In this embodiment, the perception decision-making agent 2 and multiple model agents 4 are combined into a multi-agent system, which solves the problems of work scheduling and information interaction among the multi-agents. The multi-agent system is incorporated into the unified resource description framework of the multi-resolution model, which can reduce the complexity of the system, reduce the consumption of computer resources through resource scheduling, and improve the flexibility of use.

[0108] Furthermore, this embodiment proposes the AgentDEVS atomic model based on the discrete time system specification (DEVS) specification, which provides a hierarchical and modular description mechanism for the modeling and design of multi-agents; at the same time, this embodiment proposes a "multi-resolution consistent modeling" method, which constructs a multi-resolution model by deriving different resolution models from each other, so as to make full use of system resources as much as possible and adopt as many high-fidelity and high-resolution models as possible while ensuring the real-time performance of the parallel battlefield simulation system.

[0109] Example 2

[0110] This embodiment applies the multi-agent based multi-resolution modeling method proposed in Embodiment 1 to a combat simulation of a hypothetical infantry battle between the red and blue sides. Figure 5 As shown, it is the overall simulation flow chart of this embodiment.

[0111] In this embodiment, it is assumed that a reinforced company (15 squads) of the Red side and a company (10 squads) of the Blue side fight until one side is eliminated. The Red and Blue sides have the same forces and use the same high-low score model, except that the Red side has more forces.

[0112] At this time, the model is constructed based on the company level, so the red side has 1.5 companies and the blue side has 1 company. The attack power is calculated in the following way: each attack power attack_low calculation follows a Gaussian distribution with a mean of μ and a remaining number of companies as σ, which is 10 times the number of remaining companies. The specific mathematical formula is:

[0113] attack_low=random.gauss(mu=10*n 连 ,sigma=n连 )

[0114] where n 连 Indicates the number of currently alive links, expressed as a decimal.

[0115] When attacked, the remaining strength is the current company number minus attack_low / blood_low, that is:

[0116] num_lowt=num_lowt–attack_low / blood_low

[0117] Where blood_low is the initial blood volume defined under the low-score model.

[0118] At this time, the model is constructed with classes as the basic unit, and 15 classes of the red side and 10 classes of the blue side are fighting. The attack power is calculated in the following way: the attack power of each class is calculated as a unit, and finally the sum of the attack power of the remaining classes is calculated, where the attack power of k whole classes follows a Gaussian distribution with a mean μ of 1 and σ of 0.1. The force x of less than one class is generated as a random number according to a Gaussian distribution with a mean μ of x and σ of x / 10. The specific mathematical formula is:

[0119]

[0120] When under attack, the high-resolution model is similar to the low-resolution model, and the remaining force is the current squad number minus attack_high / blood_high, that is:

[0121] num_hight=num_hight–1–attack_high / blood_high

[0122] Where num_hight is the initial blood volume defined under the high-resolution model.

[0123] Different multi-resolution switching algorithms are used according to the size of the multi-agent scale to automatically control the multi-resolution switching of the red and blue model instances. In the case of a small scale, the statistical traversal method can be used to find the optimal model resolution combination, that is, the SMRS algorithm based on statistical learning; in the case of a large scale of agents, the deep reinforcement learning method in machine learning is used to predict and estimate the performance of the simulation server 1 of the model combination, so as to decide the next resolution of each model agent 4, that is, the LMRS algorithm based on deep reinforcement learning.

[0124] During simulation, the model instance automatically adjusts the resolution of each agent according to the usage of the server CPU and memory resources. By using dedicated threads to run the calculation load and open up memory space, the consumption of CPU and memory resources is simulated. During simulation, the resource usage of the simulated load can be adjusted to observe the changes in the resolution status of the agent.

[0125] It can be seen from this embodiment that the multi-resolution modeling method based on multi-agent proposed in Example 1 can effectively reduce the consumption of computer resources by applying multi-resolution modeling application scenarios such as force simulation, radar simulation, communication simulation, and certain arms combat simulation.

[0126] Example 3

[0127] This embodiment proposes a multi-agent based multi-resolution modeling system, and applies the multi-agent based multi-resolution modeling method proposed in Embodiment 1. Figure 6 , which is an architecture diagram of the multi-agent based multi-resolution modeling system of this embodiment.

[0128] The multi-agent-based multi-resolution modeling system proposed in this embodiment includes a simulation server 1, a perception decision agent 2 for making switching scheduling decisions for the multi-resolutions of any model agent 4, and a plurality of model agents 4.

[0129] The model agent 4 in this embodiment is used to perform simulation operations and switch to different resolution models according to the decision of the perception decision agent 2.

[0130] In the specific implementation process, the perception decision agent 2 outputs a decision signal including the resolution model that each model agent 4 should select next based on the multi-resolution switching algorithm according to the resource usage status data of the simulation server 1 and the resolution information of all model agents 4 currently running. Each model agent 4 selects to use the target resolution model for performing multi-resolution modeling according to the received decision signal, and outputs the simulation result.

[0131] Among them, the multi-resolution switching algorithm adopted by the perception decision-making agent 2 includes a small-scale multi-resolution switching algorithm and a large-scale multi-resolution switching algorithm.

[0132] In an optional embodiment, the step of the perception decision agent 2 outputting a decision signal based on a multi-resolution switching algorithm according to the resource usage status data of the simulation server 1 and the resolution information of all model agents 4 currently running includes:

[0133] S3.1. Prioritize the construction and use of high-resolution models;

[0134] S3.2. According to the resource usage status data of the simulation server 1, the resource usage of the subsystems in the simulation server 1 is analyzed one by one to determine whether it needs to be replaced with a low-resolution model:

[0135] If so, a low-resolution model is selected from the resolution models, and the input and output of each low-resolution model in the subsystem are compared and used, and the resolution models with consistent input and output in the corresponding subsystem are combined to generate a low-resolution mathematical model; at the same time, the high-resolution model is used as the high-resolution mathematical model, and the low-resolution mathematical model and the high-resolution mathematical model are used to simulate multi-resolution modeling in the subsystem respectively, and then the multi-resolution modeling results are compared and verified, and the low-resolution mathematical model that meets the consistency of the multi-resolution modeling results is used as the resolution model that the model intelligent agent 4 should select in the next step to form a decision signal output;

[0136] Otherwise, the output maintains the decision signal using the high-resolution model.

[0137] In an optional embodiment, the model agent 4 includes several resolution models, as well as a model consistency maintenance mechanism, a model resolution information sharing mechanism and a model switching response mechanism.

[0138] The model consistency maintenance mechanism is used to maintain the consistency of the state information of the model agent 4 when different resolution models are switched for use.

[0139] The model resolution information sharing mechanism is used to share model information, usage status information and usage constraint information of the resolution model.

[0140] The model switching response mechanism is used to respond to the received decision signal.

[0141] Furthermore, the model agent 4 in this embodiment is obtained by expanding the DEVS specification, and its expression is as follows:

[0142] AM=(X,Y,P,A,δ int ,δ ext ,λ,ta)

[0143] In the formula, X={(ip,im)|ip∈IPorts,im∈Msg} represents the set of input port ip and its input information im, IPorts represents the input port set, and Msg is the message set of model agent 4, including the interaction information between different model agents 4.

[0144] Y={(op,om)|op∈OPorts,om∈Msg} represents the set of output ports op and their output information om, and OPorts represents the set of input ports.

[0145] P is the feature set of model agent 4; A is the set of model agent 4.

[0146] δ int :P→P represents the internal transfer function, which is triggered when e=ta(p), where e is the duration of the current feature p.

[0147] δ ext :Q×X→P represents an external transfer function, which is triggered when there is an external input event, and Q={(p,e)|p∈P,e∈[0,ta(p)]}, where X is the input event.

[0148] λ:P→Y represents the output function, which is triggered when the internal state is transferred, and Y is the output time.

[0149] ta represents the time advancement function.

[0150] In an optional embodiment, the multi-resolution modeling system of this embodiment further includes an interaction interface 5 , and data is exchanged between the perception decision agent 2 and the model agent 4 via the interaction interface 5 .

[0151] In actual simulation, the interaction interface 5 , the perception and decision-making agent 2 , and the model agent 4 may be implemented in different programming languages, so it is necessary to define the interface specification for the interaction between the agent and the interaction interface 5 .

[0152] In this embodiment, the decision-making agent interacts with the model agent 4 through the interactive interface 5 to transmit the resolution information of the model agent 4, the model resolution information after the decision, the simulation server 1 information, etc.; the model agent 4 interacts with the decision-making agent through the interactive interface 5 to transmit the resolution information, simulation information and other information.

[0153] Among them, the interactive interface 5 acts as an intermediate bridge between the perception decision agent 2 and the model agent 4, calling the simulation decision agent and the model agent 4 to complete the corresponding tasks respectively, ensuring the circulation and sharing of system data.

[0154] The same or similar reference numerals correspond to the same or similar components;

[0155] The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent;

[0156] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. 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 embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A multi-resolution modeling method based on multi-agent, characterized in that: The following steps are involved: S1. Construct a multi-agent system consisting of a perception and decision-making agent and multiple model agents; The perception decision agent is used to make a switching scheduling decision for multiple resolutions of any model agent; the model agent is used to perform simulation operations and switch to use different resolution models according to the decision of the perception decision agent; S2, obtaining the resource usage status data of the simulation server and the resolution information of all model agents currently running; S3, the perception decision agent outputs a decision signal including the resolution model that each model agent should select next based on the multi-resolution switching algorithm according to the resource usage status data of the simulation server and the resolution information of all model agents currently running; S4. Each model agent selects a target resolution model to perform multi-resolution modeling based on the received decision signal and outputs the simulation results.

2. The multi-agent based multi-resolution modeling method according to claim 1, characterized in that: The multi-resolution switching algorithm includes a small-scale multi-resolution switching algorithm and a large-scale multi-resolution switching algorithm.

3. The multi-agent based multi-resolution modeling method according to claim 1, characterized in that: The model agent includes multiple resolution models, as well as a model consistency maintenance mechanism, a model resolution information sharing mechanism and a model switching response mechanism; The model consistency maintenance mechanism is used to maintain the consistency of state information of the model agent when different resolution models are switched; The model resolution information sharing mechanism is used to share model information, usage status information and usage constraint information of the resolution model; The model switching response mechanism is used to respond to the received decision signal.

4. The multi-agent based multi-resolution modeling method according to claim 3, characterized in that: In the step S3, the step of the perception decision agent outputting a decision signal based on a multi-resolution switching algorithm according to the resource usage status data of the simulation server and the resolution information of all model agents currently running comprises: S3.

1. Prioritize the construction and use of high-resolution models; S3.

2. According to the resource usage status data of the simulation server, analyze the resource usage of the subsystems in the simulation server one by one to determine whether they need to be replaced with low-resolution models: If so, a low-resolution model is selected from the resolution models, and the input and output of each low-resolution model in the subsystem are compared and used, and the resolution models with consistent input and output in the corresponding subsystem are combined to generate a low-resolution mathematical model; at the same time, the high-resolution model is used as the high-resolution mathematical model, and the low-resolution mathematical model and the high-resolution mathematical model are used to simulate multi-resolution modeling in the subsystem respectively, and then the multi-resolution modeling results are compared and verified, and the low-resolution mathematical model that meets the consistency of the multi-resolution modeling results is used as the resolution model that the model intelligent agent should select in the next step to form a decision signal output; Otherwise, the output maintains the decision signal using the high-resolution model.

5. The multi-agent based multi-resolution modeling method according to claim 3, characterized in that: The step S1 also includes the following steps: expanding the model agent based on the DEVS specification to construct a new Agent DEVS atomic model; the expression is as follows: AM=(X,Y,P,A,δ int ,d ext (( Where, X = {(ip,im)|ip∈IPorts,im∈Msg} represents the set of input ports ip and their input information im, IPorts represents the set of input ports, Msg is the message set of model agents, including the interaction information between different model agents; Y = {(op,om)|op∈OPorts,om∈Msg} represents the set of output ports op and their output information om, OPorts represents the set of input ports; P is the feature set of model agents; A is the set of model agents; δ int represents the internal transfer function; δ ext represents the external transfer function; λ represents the output function, which is triggered when the internal state is transferred; ta represents the time advancement function.

6. The multi-agent based multi-resolution modeling method according to claim 5, characterized in that: The model agent set A includes: The perceptron of the model agent is used to perceive external input and internal feedback and generate processor input; Processors, used to abstract, learn, and reason about input data; Effectors, which are used to complete the internal feature transfer and produce responses to the environment; and,a knowledge base, which is used to support the computation and input and output of the model agent; The feature set P of the model agent includes the identity of the model agent, the name of the model agent, the location of the model agent, the state of the model agent and the possible output values ​​of the current features of the model agent; The message set Msg of the model agent includes message identification, message sender information, message receiver information, message type, message content, message association time and message validity period.

7. A multi-agent based multi-resolution modeling system, characterized in that: include: Simulation server; A perception and decision agent, which is used to make switching scheduling decisions for multiple resolutions of any model agent; and, a plurality of model agents for performing simulation operations and switching to different resolution models according to the decision of the perception decision agent; The perception decision agent outputs a decision signal including the resolution model that each model agent should select next based on a multi-resolution switching algorithm according to the resource usage status data of the simulation server and the resolution information of all model agents currently running; Each model agent selects the target resolution model to perform multi-resolution modeling based on the received decision signal and outputs the simulation results.

8. The multi-agent based multi-resolution modeling system according to claim 7, characterized in that: The model agent includes multiple resolution models, as well as a model consistency maintenance mechanism, a model resolution information sharing mechanism and a model switching response mechanism; The model consistency maintenance mechanism is used to maintain the consistency of state information of the model agent when different resolution models are switched; The model resolution information sharing mechanism is used to share model information, usage status information and usage constraint information of the resolution model; The model switching response mechanism is used to respond to the received decision signal.

9. The multi-agent based multi-resolution modeling system according to claim 8, characterized in that: The model agent is expanded based on the DEVS specification, and its expression is as follows: AM=(X,Y,P,A,δ int ,d ext (( Where, X = {(ip,im)|ip∈IPorts,im∈Msg} represents the set of input ports ip and their input information im, IPorts represents the set of input ports, Msg is the message set of model agents, including the interaction information between different model agents; Y = {(op,om)|op∈OPorts,om∈Msg} represents the set of output ports op and their output information om, OPorts represents the set of input ports; P is the feature set of model agents; A is the set of model agents; δ int represents the internal transfer function; δ ext represents the external transfer function; λ represents the output function, which is triggered when the internal state is transferred; ta represents the time advancement function.

10. The multi-agent based multi-resolution modeling system according to any one of claims 7 to 9, characterized in that: The system also includes an interactive interface, and the perception decision agent and the model agent exchange data through the interactive interface.

Images