Multi-agent collaborative management and control method and device

Abstract

The invention discloses a multi-agent collaborative management and control method and device, and relates to the technical field of equipment control. Inputting the management and control state data into an intelligent agent corresponding to the management and control state data to obtain a plurality of primary decisions; calculating a marginal contribution value of each agent; determining candidate decisions based on the marginal contribution values; determining a target decision based on a preset agronomic rule hard constraint and the candidate decision; and generating a control instruction based on the target decision, sending the control instruction to the control equipment, and enabling the control equipment to execute the control instruction. Therefore, the accuracy of determining the target decision is improved.

Description

Technical Field

[0001] This application relates to the field of equipment control technology, and in particular to a method and apparatus for multi-agent collaborative management. Background Technology

[0002] In recent years, the Internet of Things and artificial intelligence technologies have been gradually applied in orchard monitoring, resulting in intelligent decision-making platforms based on sensors, image recognition, and expert systems.

[0003] In related technologies, most systems employ a single decision-making model (such as threshold alarms or fixed rules), which cannot adapt to the dynamic needs of fruit trees at different phenological stages. For example, during the flowering period, the soil needs to be kept moist to protect flowers and fruits, while during the fruit coloring period, water needs to be moderately controlled to improve sugar content and coloring; it is difficult to achieve both simultaneously with the same strategy. Furthermore, although some systems integrate multiple functional modules such as irrigation, fertilization, disease control, and meteorology, each module operates independently and lacks a scientific coordination mechanism. When modules output conflicting instructions (such as "suggest fertilization" and "warning of future heavy rainfall"), the system often uses simple priority overriding or manual intervention, failing to quantify the actual contribution of each module in the current situation. Moreover, while some agricultural question-answering systems based on large language models can provide knowledge services, their generated suggestions often violate national agricultural industry standards (such as recommending spraying before harvesting), and they cannot be linked with physical equipment, making it difficult to form a closed loop of "perception-decision-execution". Summary of the Invention

[0004] In view of this, this application provides a multi-agent collaborative management method and apparatus to improve the accuracy of target determination decisions.

[0005] The objective of this application can be achieved through the following technical solutions: The first aspect of this application is to provide a multi-agent cooperative management method, including: Obtain control status data; Input the control status data into the corresponding intelligent agent to obtain multiple preliminary decisions; Calculate the marginal contribution value of each agent; Candidate decisions are determined based on marginal contribution values; Determine the target decision based on the hard constraints of pre-set agronomic rules and candidate decisions; Control commands are generated based on the target decision and sent to the control equipment to execute the control commands.

[0006] In one optional embodiment, the control status data is input into the agent corresponding to the control status data to obtain multiple preliminary decisions, including: Soil moisture status data is input into the irrigation decision-making agent, and a preliminary irrigation decision is output. Input the fruit tree status data into the pest and disease control decision-making intelligent agent, and output the primary pest and disease control decision; Meteorological status data is input into the meteorological early warning decision-making intelligent agent, which then outputs a preliminary risk early warning decision. Soil nutrient status data is input into the fertility decision-making agent, which then outputs a preliminary fertilization decision. Input fruit tree phenological data into the agricultural planning and decision-making intelligent system, and output basic agricultural reminders.

[0007] In one optional embodiment, the soil moisture status data is the relative soil water content, the fruit tree status data is the fruit tree image, and the meteorological status data includes at least one of the following: air temperature, relative humidity, atmospheric pressure, wind speed, wind direction, rainfall, light intensity, and evaporation; the soil nutrient status data includes soil nitrogen content, soil available phosphorus content, soil available potassium content, and soil pH.

[0008] In one optional embodiment, soil nutrient status data is input into a fertility decision-making agent, and a preliminary fertilization decision is output, including: To obtain soil nitrogen content, available phosphorus content, available potassium content, and soil pH; Calculate the deficits in soil nitrogen content, soil available phosphorus content, and soil available potassium content; A preliminary fertilization decision is calculated based on the deficits in soil nitrogen, available phosphorus, and available potassium, as well as soil pH.

[0009] In one alternative embodiment, calculating the marginal contribution value of each agent includes: Calculate the marginal contribution value of each agent using the following formula: ; in, This represents the marginal contribution value of the i-th agent. Let represent any subset of agents other than the i-th agent. Let n represent the set including all agents, and n represent the number of agents. This represents the utility value of subset S. This indicates that the utility value of the subset includes the i-th agent's subset added to the subset S; ; in, This represents the utility value of subset S. Let t represent the initial decision output by the i-th agent in subset S, and t represent the phenological period. ( ) indicates fusion.

[0010] In one alternative embodiment, determining candidate decisions based on marginal contribution values ​​includes: Primary decisions whose marginal contribution value is greater than the preset marginal contribution threshold are identified as candidate decisions.

[0011] In one optional embodiment, determining the target decision based on preset agronomic rule hard constraints and candidate decisions includes: If the candidate decision does not meet the hard constraints of the preset agronomic rules, the candidate decision is modified based on the hard constraints of the preset agronomic rules to obtain the target decision; If a candidate decision meets the hard constraints of the preset agronomic rules, then the candidate decision is determined as the target decision.

[0012] A second aspect of this application is to provide a multi-agent collaborative management and control device, comprising: The acquisition module is used to acquire control status data; The input module is used to input the control status data into the intelligent agent corresponding to the control status data to obtain multiple primary decisions; The calculation module is used to calculate the marginal contribution value of each agent; The first determination module is used to determine candidate decisions based on marginal contribution values; The second determination module is used to determine the target decision based on preset agronomic rule hard constraints and candidate decisions; The generation module is used to generate control commands based on target decisions, send the control commands to the control device, and enable the control device to execute the control commands.

[0013] A third aspect of this application is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the method as described in the first aspect.

[0014] A fourth aspect of this application is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the method as described in the first aspect.

[0015] Compared with existing technologies, the multi-agent collaborative management and control method provided in this application acquires management and control status data; inputs the management and control status data into the corresponding agents to obtain multiple preliminary decisions; calculates the marginal contribution value of each agent; determines candidate decisions based on the marginal contribution values; determines the target decision based on preset agronomic rule hard constraints and candidate decisions; generates control commands based on the target decision, and sends the control commands to the control equipment to execute the control commands. This improves the accuracy of determining the target decision. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a multi-agent collaborative management and control method provided in an embodiment of this application; Figure 2 A structural block diagram of a multi-agent collaborative management and control device provided in an embodiment of this application; Figure 3 This is a structural block diagram of an electronic device for implementing a multi-agent collaborative management method, as provided in an embodiment of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] It should be understood that in the embodiments of this application, "at least one" means one or more, and "more than one" means two or more. "And / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the related objects before and after it are in an "or" relationship. "Contains A, B and / or C" means containing any one, two, or three of A, B, and C.

[0021] It should be understood that in the embodiments of this application, "B corresponding to A", "B corresponding to A", "A corresponds to B" or "B corresponds to A" means that B is associated with A, and B can be determined based on A. Determining B based on A does not mean that B is determined solely based on A; B can also be determined based on A and / or other information.

[0022] To address the technical problems existing in related technologies, this application provides a multi-agent collaborative management method and apparatus.

[0023] The multi-agent collaborative management and control method provided in this application can be executed by an electronic device, which can be a terminal or a server. The terminal can be a smartphone, tablet, laptop, or other similar device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. It is understood that this application does not limit the specific entity executing the multi-agent collaborative management and control method.

[0024] The technical solution of this application will be described in detail below through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments described below are used to explain the technical solution of this application and are not intended to limit actual use.

[0025] To address the technical problems existing in related technologies, embodiments of this application provide a multi-agent collaborative management and control method, such as... Figure 1 As shown, Figure 1 This is a flowchart illustrating a multi-agent cooperative management method provided in an embodiment of this application. It should be noted that the steps shown may be executed in a different logical order than those shown in the flowchart. The method may include the following steps S101 to S106.

[0026] Step S101: Obtain control status data.

[0027] It should be noted that the management status data is used for the management of fruit trees.

[0028] In one optional embodiment, the control status data includes: soil moisture status data, fruit tree status data, meteorological status data, soil nutrient status data, and fruit tree phenological period data.

[0029] In one specific embodiment, the soil moisture status data is the relative soil water content, the fruit tree status data is the fruit tree image, and the meteorological status data includes at least one of the following: air temperature, relative humidity, atmospheric pressure, wind speed, wind direction, rainfall, light intensity, and evaporation; the soil nutrient status data includes soil nitrogen content, soil available phosphorus content, soil available potassium content, and soil pH.

[0030] In a more specific embodiment, a soil moisture sensor is connected via LoRa (Long Range) to obtain soil moisture status data. For example, TDR (Time Domain Reflectometry) soil moisture sensors are deployed at depths of 20cm and 40cm around the roots of fruit trees to collect relative soil moisture content in real time.

[0031] In a more specific embodiment, a weather station is connected via RS-485 (Recommended Standard 485) to obtain meteorological status data. For example, detailed weather forecast data for the next 72 hours, including temperature, precipitation, and wind speed, is obtained every 2 hours via a cloud interface.

[0032] In a more specific embodiment, a camera is connected via a USB (Universal Serial Bus) interface or a MIPI (Mobile Industry Processor Interface) to acquire data representing the state of the fruit trees, such as leaf images and fruit images. For example, images of fruit tree leaves can be captured by cameras deployed in the orchard. The fruit tree leaf images can also be uniformly cropped and scaled to 224×224 pixels. The fruit tree leaf images can be in RGB (Red-Green-Blue) format.

[0033] In a more specific embodiment, the nitrogen content in the soil is obtained by connecting an optical nitrate nitrogen sensor via LoRa. The available phosphorus content and available potassium content in the soil are obtained by detecting the soil. The soil pH can also be obtained by deploying a soil pH electrode.

[0034] In a more specific embodiment, time is recorded using a real-time clock, and the effective accumulated temperature is calculated in real time based on the GDD (Growing Degree Days) model. The current phenological period is then output based on the effective accumulated temperature and a preset phenological period threshold table. It should be noted that a phenological period refers to a specific developmental stage that a fruit tree experiences during its one-year growth cycle, such as: bud break, leaf expansion, inflorescence separation, initial flowering, full bloom, flowering end, young fruit stage, fruit enlargement, coloring, and maturity.

[0035] For example, obtain the starting point of the time frame, which is a key point in the previous growing season, such as the date of leaf fall last year or the date of bud break this year; obtain the daily average temperature from the weather station, set the lower biological limit temperature, for example, the lower biological limit temperature for apple trees can be 5℃, and calculate the effective accumulated temperature using the following formula: (1); in, This represents the effective accumulated temperature, and n represents the total number of days. This represents the average daily temperature on day i. This indicates the biological lower limit temperature.

[0036] For example, Table 1 shows the accumulated temperature range required for each phenological stage of apple trees in a certain region. Table 1

[0037] In a more specific embodiment, voice / text can also be received via a microphone or application to obtain control status data.

[0038] Step S102: Input the control status data into the agent corresponding to the control status data to obtain multiple primary decisions.

[0039] In one optional embodiment, the control status data is input into the agent corresponding to the control status data to obtain multiple preliminary decisions, including: Soil moisture data is input into the irrigation decision-making intelligent body, and a basic irrigation decision is output; fruit tree status data is input into the pest and disease control decision-making intelligent body, and a basic pest and disease control decision is output; meteorological status data is input into the meteorological early warning decision-making intelligent body, and a basic risk early warning decision is output; soil nutrient status data is input into the fertility decision-making intelligent body, and a basic fertilization decision is output; fruit tree phenological period data is input into the agricultural planning decision-making intelligent body, and a basic agricultural reminder is output.

[0040] In one specific embodiment, soil moisture status data is input into an irrigation decision-making agent, and a primary irrigation decision is output. This includes: the irrigation decision-making agent determines a primary irrigation decision based on the relative soil moisture content, fruit tree phenological period, harvest plan, and regional preset irrigation rules, and outputs the primary irrigation decision. The primary irrigation decision may include whether to irrigate and the amount of irrigation. Irrigation can be represented by 1, and not irrigating can be represented by 0. Of course, other values ​​can also be used. The amount of irrigation is represented by mm (millimeters).

[0041] In a more specific embodiment, Table 2 shows the irrigation rules for apple growth stages in a certain region, as shown in Table 1: Table 2

[0042] In a more specific embodiment, the expression for the primary irrigation decision is:

[0043] in, Let S represent the current phenological stage of the fruit tree, and let S represent the decision variable. Indicates the relative moisture content of the soil. Indicates the current date. This indicates the number of days remaining until the harvest plan is completed.

[0044] In one specific embodiment, fruit tree status data is input into a pest and disease control decision-making agent, and a preliminary pest and disease control decision is output. This includes: inputting pre-processed fruit tree images into a pre-trained pest and disease identification model through the pest and disease control decision-making agent, and outputting the confidence level of the pest and disease category; when the pest and disease control decision-making agent determines that the maximum confidence level of the pest and disease category is not lower than a preset confidence threshold, the pest and disease category identification result and the preliminary pest and disease control decision are returned; otherwise, a message is displayed: "Identification uncertain, manual review recommended." For example, the preset confidence threshold can be 80%, 89%, or 95%, etc., and this application does not limit the preset confidence threshold.

[0045] It should be noted that primary pest and disease control decisions are determined based on agronomic knowledge rules.

[0046] In a more specific embodiment, the pest and disease identification model can be a lightweight convolutional neural network model, for example, MobileNetV3-Small.

[0047] In another, more specific embodiment, the pest and disease categories include, but are not limited to, the following categories: healthy, brown spot, leaf spot, powdery mildew, aphids, spider mites, leafrollers, and anthracnose.

[0048] In another specific embodiment, the training process of the pre-trained pest and disease identification model is as follows: Obtain sample fruit tree images and corresponding pest and disease category labels; input the sample fruit tree images into the pest and disease identification model to obtain the predicted pest and disease categories corresponding to the sample fruit tree images; calculate the target loss based on the pest and disease category labels and predicted pest and disease categories corresponding to the sample fruit tree images; update the model parameters of the pest and disease identification model based on the target loss.

[0049] In a more specific embodiment, the loss function can be either cross-entropy loss or focus loss.

[0050] In a more specific embodiment, a stochastic gradient descent or Adam (Adaptive Moment Estimation) optimizer is used to update the model parameters based on backpropagation of the target loss until the validation set accuracy converges or reaches the preset training epoch limit.

[0051] In another specific embodiment, meteorological state data is input into the meteorological early warning decision-making intelligent agent, and a preliminary risk early warning decision is output. This includes: inputting meteorological state prediction data for a future preset time period and the current phenological period into the meteorological early warning decision-making intelligent agent; the meteorological early warning decision-making intelligent agent generates a risk warning level for the current phenological period and corresponding agricultural response suggestions based on the agricultural meteorological risk rule base; and sends the risk warning level for the current phenological period and corresponding agricultural response suggestions to the control terminal.

[0052] In another specific embodiment, data cleaning and logical judgment can be performed on meteorological state prediction data and current phenological period based on the agricultural meteorological risk rule base to remove outliers and missing interpolation periods.

[0053] It should be noted that the agricultural meteorological risk rule base is based on public databases.

[0054] In a more specific embodiment, the risk warning level can be divided into high risk, medium-high risk and low risk based on experience, and agricultural response suggestions include starting frost protection fans, spraying protective fungicides in advance, postponing pesticide application to a preset time after rain, increasing drip irrigation and harvesting in advance.

[0055] In another specific embodiment, soil nutrient status data is input into a fertility decision-making agent, which outputs a preliminary fertilization decision, including: Soil nutrient status data is input into the fertility decision-making agent, which determines the primary fertilization decision based on scientific fertilization technology guidelines and outputs the primary fertilization decision.

[0056] In a more specific embodiment, the guidelines for scientific fertilization recommend that the ratio of nitrogen (N), phosphorus pentoxide (P2O5), and potassium oxide (K2O) is typically 1:0.5:1.2. The Loess Plateau soils are generally deficient in nitrogen and phosphorus, with moderate potassium content and a relatively high pH (7.5–8.5), therefore water-soluble fertilizers can be applied.

[0057] For example, inputting soil nutrient status data into a fertility decision-making agent and outputting a preliminary fertilization decision includes the following steps: To obtain soil nitrogen content, available phosphorus content, available potassium content, and soil pH; Calculate the deficits in soil nitrogen content, soil available phosphorus content, and soil available potassium content; The primary fertilization decision is calculated based on the deficits in soil nitrogen content, available phosphorus content, available potassium content, and pH.

[0058] An alternative example is to calculate the deficits in soil nitrogen content, available phosphorus content, and available potassium content using the following formulas: (3); (4); (5); in, Indicates the nitrogen deficit in the soil. This indicates the deficit in available phosphorus content in the soil. This indicates the deficit in available potassium content in the soil.

[0059] It should be noted that, , and It is based on experience, for example, It can take any value between 80 and 150. It can take any value between 20 and 40. It can take any value between 150 and 250.

[0060] Based on the deficits in soil nitrogen, available phosphorus, and available potassium, the primary fertilization decision is calculated using the following formula: (6); (7); (8); in, This indicates the amount of nitrogen that needs to be added without a proportional calculation. This indicates the amount of phosphorus pentoxide that needs to be supplemented without a proportional calculation. This indicates the amount of potassium oxide that needs to be added without a proportional calculation. , and This is an empirical coefficient, which is related to soil pH.

[0061] In another, more specific embodiment, the following formula can also be used to ensure that the primary fertilization decision satisfies the ratio 1:0.5:1.2: = (9); = (10); = (11); in, This indicates the amount of nitrogen that needs to be added for the proportional calculation. This indicates the amount of phosphorus pentoxide that needs to be added to achieve the desired ratio. This indicates the amount of potassium oxide that needs to be added to achieve the desired ratio.

[0062] In another specific embodiment, the agricultural planning and decision-making intelligent agent constructs a structured agricultural rule base based on a public database. After the current phenological period is input into the agricultural planning and decision-making intelligent agent, it can output primary agricultural reminders based on the current phenological period to achieve precise agricultural scheduling.

[0063] In a more specific embodiment, the primary farming reminder can be a structured list of farming reminders, including recommended actions and action windows. It should be noted that the action window refers to the optimal time period for performing a specific action during farming activities.

[0064] For example, basic agricultural tips include: soil moisture has dropped to 58%, the current period is the fruit enlargement stage, it is recommended to start drip irrigation for 2 hours; there will be heavy rain tomorrow, it is not advisable to apply fertilizer today to avoid nutrient loss; the flowering period has begun, please suspend mechanical operations to avoid damaging the flowers, etc.

[0065] Step S103: Calculate the marginal contribution value of each agent.

[0066] In one alternative embodiment, calculating the marginal contribution value of each agent includes: Calculate the marginal contribution value of each agent using the following formula: (12); in, This represents the marginal contribution value of the i-th agent. Let represent any subset of agents other than the i-th agent. Let n represent the set including all agents, and n represent the number of agents. This represents the utility value of subset S. This represents adding the utility value of the subset with the i-th agent to the subset S; (13); in, This represents the utility value of subset S. Let t represent the initial decision output by the i-th agent in subset S, and t represent the phenological period. ( ) indicates fusion.

[0067] In a more specific embodiment, the utility function for calculating the utility value is designed based on agricultural technical specifications, which are not limited herein.

[0068] In another alternative embodiment, the marginal contribution value of each agent can be normalized using the following formula to obtain a weight vector: (14); in, This represents the weight vector corresponding to the i-th agent.

[0069] Step S104: Determine candidate decisions based on marginal contribution values.

[0070] In one alternative embodiment, determining candidate decisions based on marginal contribution values ​​includes: Primary decisions whose marginal contribution value is greater than the preset marginal contribution threshold are identified as candidate decisions.

[0071] In a more specific embodiment, the preset marginal contribution threshold is set according to the actual situation, and this application does not limit it.

[0072] Step S105: Determine the target decision based on the preset agronomic rules hard constraints and candidate decisions.

[0073] In one optional embodiment, the target decision is determined based on preset agronomic rule hard constraints and candidate decisions, specifically including the following steps: if a candidate decision does not conform to the preset agronomic rule hard constraints, the candidate decision is modified based on the preset agronomic rule hard constraints to obtain the target decision; if a candidate decision conforms to the preset agronomic rule hard constraints, the candidate decision is determined as the target decision.

[0074] In one specific embodiment, the preset agronomic rule hard constraint can be expressed as: , where m is the number of rules. It can be represented as ,in, This includes current weather conditions, soil moisture, and pest and disease status. This includes mandatory actions and / or prohibited agricultural actions.

[0075] In this step, if a candidate decision violates agronomic rules, the candidate decision can be corrected to ensure the rationality of the final target decision.

[0076] Step S106: Generate control commands based on the target decision, and send the control commands to the control device so that the control device executes the control commands.

[0077] In one specific embodiment, an embedded artificial intelligence computing platform is used as the main control unit. This platform has the following capabilities: it supports the TensorFlow Lite / ONNX Runtime inference framework; it can run lightweight disease identification models (such as MobileNetV3, <10MB) in real time; it provides 4 relay outputs (rated current 10 amps, AC voltage 250 volts) and an RS-485 industrial interface; and it supports LoRa (470–510 MHz), 4G and Wi-Fi multimode communication.

[0078] In a more specific embodiment, taking a Fuji apple orchard in Luochuan County, Shaanxi Province as an application scenario, this application is described in detail, specifically including the following steps: Step 1: Automatic Multi-Source Data Acquisition: Real-time acquisition of orchard data through multiple interfaces. Soil data: The relative soil moisture content was obtained through wireless sensors and recorded as 45%. Image data: Photos of the leaves of fruit tree No. 3 in the East District taken using a high-definition camera; Meteorological data: The 24-hour weather forecast obtained from the weather station shows no rainfall; Phenological stage identification: Based on the accumulated temperature model, it automatically determines that the apple fruit is currently in the fruit enlargement stage.

[0079] Step 2: Task Distribution and Multi-Agent Decision Generation: Activate four functional agents, each outputting structured decision suggestions: The irrigation decision-making agent outputs: d1 = "Drip irrigation for 10 minutes"; Disease identification AI output: d2 = "Spray Mancozeb"; Weather warning intelligent agent output: d3 = "Suitable for spraying"; Agricultural planning agent output: d4 = "No agricultural bans currently in effect".

[0080] Step 3: Construct an agronomic utility rating function: Based on the current fruit enlargement stage, different built-in agronomic utility scoring functions v(S) are loaded as shown in Table 3 to evaluate any subset of intelligent agents. The joint decision-making effect.

[0081] Table 3

[0082] Combined decision scoring rule: For any subset S, the score of its joint decision DS is the sum of the scores of each action.

[0083] Step 4: Calculate the weight vector for each agent: Calculate the marginal contribution of different agents (taking the A1 irrigation agent as an example).

[0084]

[0085] After complete calculation (or sampling approximation), the contributions of different agents are obtained as follows: =1.25, =1.15, =0.15, =0 Normalized weights: Total contribution = 1.25 + 1.15 + 0.15 = 2.50

[0086]

[0087] It can be seen that: irrigation intelligence contributes the most (50%) because it solves the most pressing water stress problem; disease identification is the second most important (46%) because it accurately detects diseases; meteorological intelligence plays only an auxiliary role (4%).

[0088] Step 5: Integrate decision-making and conduct agronomic compliance verification: 1. Weighted fusion decision-making: First, set a minimum contribution threshold τ = 0.1. For each agent's suggested action Ai, if its corresponding weight w1 ≥ τ and Ai is not an empty operation (such as "no operation"), it is included in the candidate action set.

[0089] Candidate actions are checked for agronomic conflicts (such as whether irrigation and spraying can be performed simultaneously). If there is no conflict, they are executed together. If there is a conflict, they are sorted by weight priority, with the higher priority taking precedence.

[0090] Generate a comprehensive decision based on weights: d-fusion = "10-minute drip irrigation + spraying mancozeb" 2. Agronomic rule verification The device calls the rule base to perform checks: Rule 1: Soil moisture <60% during the fruit enlargement stage → Irrigation is required → Compliant; Rule 2: Diseases discovered outside the harvest period → Spraying is permitted → Compliant.

[0091] If there are no violations, the decision is passed directly without needing to be amended. If it is 12 days before harvest, even if the d fusion includes "spraying pesticides", it will be forcibly corrected to "manually removing diseased fruit".

[0092] Step Six: Convert the final decision into equipment control signals: Irrigation control command: Send an "open" signal to the water pump solenoid valve. Plant protection control instructions: Issue work commands to the intelligent sprayer, including pesticide name, dilution ratio, and work area; Anomaly Handling: If the device reports an execution failure, an alarm message will be immediately pushed to the fruit farmer's mobile phone via the 4G network.

[0093] Corresponding to the multi-agent cooperative management method provided in the embodiments of this application, the embodiments of this application also provide a multi-agent cooperative management device, such as... Figure 2 As shown, the multi-agent collaborative management and control device includes: Module 201 is used to acquire control status data; Input module 202 is used to input control status data into the intelligent agent corresponding to the control status data to obtain multiple primary decisions; Calculation module 203 is used to calculate the marginal contribution value of each agent; The first determining module 204 is used to determine candidate decisions based on marginal contribution values; The second determination module 205 is used to determine the target decision based on preset agronomic rule hard constraints and candidate decisions; The generation module 206 is used to generate control commands based on the target decision, send the control commands to the control device, and make the control device execute the control commands.

[0094] Corresponding to the multi-agent cooperative management method provided in the embodiments of this application, the embodiments of this application also provide an electronic device for executing the multi-agent cooperative management method, such as... Figure 3 As shown, the electronic device includes: a processor 301; and a memory 302 for storing a program for a multi-agent cooperative management method. After the device is powered on and the processor runs the program for the multi-agent cooperative management method, it performs the following steps: Obtain control status data; Input the control status data into the corresponding intelligent agent to obtain multiple preliminary decisions; Calculate the marginal contribution value of each agent; Candidate decisions are determined based on marginal contribution values; Determine the target decision based on the hard constraints of pre-set agronomic rules and candidate decisions; Control commands are generated based on the target decision and sent to the control equipment to execute the control commands.

[0095] Corresponding to the multi-agent cooperative management method provided in the embodiments of this application, the embodiments of this application also provide a computer-readable storage medium storing a program for the multi-agent cooperative management method, which is executed by a processor to perform the following steps: Obtain control status data; Input the control status data into the corresponding intelligent agent to obtain multiple preliminary decisions; Calculate the marginal contribution value of each agent; Candidate decisions are determined based on marginal contribution values; Determine the target decision based on the hard constraints of pre-set agronomic rules and candidate decisions; Control commands are generated based on the target decision and sent to the control equipment to execute the control commands.

[0096] Corresponding to the multi-agent cooperative management method provided in the embodiments of this application, the embodiments of this application also provide a computer program containing instructions, which, when executed by a computer, cause the computer to perform the following steps: Obtain control status data; Input the control status data into the corresponding intelligent agent to obtain multiple preliminary decisions; Calculate the marginal contribution value of each agent; Candidate decisions are determined based on marginal contribution values; Determine the target decision based on the hard constraints of pre-set agronomic rules and candidate decisions; Control commands are generated based on the target decision and sent to the control equipment to execute the control commands.

[0097] It should be noted that for a detailed description of the multi-agent collaborative management and control device, electronic device, computer-readable storage medium and computer program product provided in the embodiments of this application, please refer to the relevant description of the embodiments of the multi-agent collaborative management and control method provided in the embodiments of this application, which will not be repeated here.

[0098] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

[0099] In a typical configuration, an electronic device includes one or more processors (Central Processing Units), input / output interfaces, network interfaces, and memory.

[0100] Memory may include non-persistent storage in computer-readable media, such as random access memory and / or non-volatile memory, like read-only memory or flash memory. Memory is an example of computer-readable media.

[0101] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable operations, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PCM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital video disc (DMCD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0102] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, compact disc read-only memory, optical storage, etc.) containing computer-usable program code.

[0103] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A multi-agent collaborative management method, characterized in that, The method comprises the following steps: obtaining control state data; inputting the control state data into agents corresponding to the control state data to obtain multiple primary decisions; calculating the marginal contribution value of each agent; determining a candidate decision based on the marginal contribution value; determining a target decision based on a preset agronomic rule hard constraint and the candidate decision; generating a control instruction based on the target decision, and sending the control instruction to a control device to enable the control device to execute the control instruction.

2. The multi-agent collaborative management method according to claim 1, characterized in that, The inputting the control state data into agents corresponding to the control state data to obtain multiple primary decisions comprises the following steps: inputting soil moisture state data into an irrigation decision agent to output a primary irrigation decision; inputting fruit tree state data into a pest control decision agent to output a primary pest control decision; inputting meteorological state data into a meteorological warning decision agent to output a primary risk warning decision; inputting soil nutrient state data into a fertility decision agent to output a primary fertilization decision; inputting fruit tree phenological stage data into a farming planning decision agent to output a primary farming reminder.

3. The multi-agent collaborative management method according to claim 2, wherein, The soil moisture state data is soil relative water content, the fruit tree state data is a fruit tree image, the meteorological state data comprises at least one of the following data: air temperature, relative humidity, atmospheric pressure, wind speed, wind direction, rainfall, light intensity and evaporation amount, and the soil nutrient state data comprises soil nitrogen content, soil available phosphorus content, soil available potassium content and soil pH.

4. The multi-agent collaborative management method according to claim 3, characterized in that, The inputting soil nutrient state data into a fertility decision agent to output a primary fertilization decision comprises the following steps: obtaining the soil nitrogen content, soil available phosphorus content, soil available potassium content and soil pH; calculating the deficiency amount of the soil nitrogen content, soil available phosphorus content and soil available potassium content; calculating a primary fertilization decision based on the deficiency amount of the soil nitrogen content, the deficiency amount of the soil available phosphorus content, the deficiency amount of the soil available potassium content and the soil pH.

5. The multi-agent collaborative management method according to claim 1, wherein, The calculating the marginal contribution value of each agent comprises the following steps: calculating the marginal contribution value of each agent by using the following formula: ； Wherein, the represents the marginal contribution value of the i-th agent, represents an arbitrary subset composed of other agents except the i-th agent, represents a set including all agents, and n represents the number of agents, represents the utility value of the subset S, represents the utility value of the subset including the i-th agent on the basis of the subset S; ； wherein, represents the utility value of the subset S, represents the primary decision of the i-th agent output in the subset S, t represents the phenophase, () represents fusion.

6. The multi-agent collaborative management method according to claim 1, wherein, The determining a candidate decision based on the marginal contribution value comprises the following steps: determining a primary decision with a marginal contribution value greater than a preset marginal contribution threshold as a candidate decision.

7. The multi-agent collaborative management method according to claim 1, wherein, The determining a target decision based on a preset agronomic rule hard constraint and the candidate decision comprises the following steps: if the candidate decision does not conform to the preset agronomic rule hard constraint, modifying the candidate decision based on the preset agronomic rule hard constraint to obtain a target decision; if the candidate decision conforms to the preset agronomic rule hard constraint, determining the candidate decision as a target decision.

8. A multi-agent collaborative management and control device, characterized in that, The method comprises the following steps: an acquisition module for acquiring control state data; an input module for inputting the control state data into agents corresponding to the control state data to obtain multiple primary decisions; a calculation module for calculating the marginal contribution value of each agent; a first determination module for determining a candidate decision based on the marginal contribution value; a second determination module for determining a target decision based on a preset agronomic rule hard constraint and the candidate decision; The generating module is configured to generate a control instruction based on the target decision, and send the control instruction to a control device to enable the control device to execute the control instruction.

9. An electronic device, comprising: The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the multi-agent collaborative control method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the multi-agent collaborative control method according to any one of claims 1-7.

Images