A multi-agent management system and method for collaborative work of robots and unmanned aerial vehicles

Abstract

The application discloses a kind of multi-agent management system and method of robot and unmanned aerial vehicle collaborative work, it is related to multi-agent management technical field, solve the technical problem that prior art lacks collaborative work capability, and the identification result of plant disease and insect pest is not accurate enough;The application collects the image data of crop and the corresponding environmental data by unmanned aerial vehicle;Real-time collected crop image data is identified based on the plant disease and insect pest identification model obtained by training;Based on environmental data, a correction model is constructed, and the initial plant disease and insect pest coverage obtained by identification is corrected to obtain the target plant disease and insect pest coverage;The pesticide spraying robot is calculated based on the target plant disease and insect pest coverage The amount of pesticide spraying;The optimal path of pesticide spraying robot is obtained by path planning algorithm;Based on the optimal path, move to the plant disease and insect pest coverage area, and based on the amount of pesticide spraying, the plant disease and insect pest coverage area is sprayed with pesticide, solve the above technical problems.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of multi-agent management, and particularly relates to a multi-agent management system and method for collaborative work of robots and unmanned aerial vehicles. BACKGROUND

[0002] In modern agriculture, disease and pest control is an important link to ensure crop health and high yield. Traditional disease and pest control methods rely on manual inspection and manual pesticide spraying, which have significant problems: manual inspection is time-consuming and labor-intensive and difficult to fully cover large areas of farmland, and manual spraying can easily lead to uneven spraying, resulting in over-spraying or under-spraying in some areas; manual identification of diseases and pests is low in accuracy and is prone to missed detection or misjudgment.

[0003] Existing technologies mainly use image recognition technology to identify diseases and pests, and use a single agent (such as an unmanned aerial vehicle or a pesticide spraying robot) to monitor and spray pesticides on crops; however, a single agent (such as an unmanned aerial vehicle or a pesticide spraying robot) can usually only complete a specific task (such as image acquisition or spraying), lacks collaborative work capability, and cannot fully utilize the advantages of each agent. At the same time, although existing technologies can identify diseases and pests, they do not adequately consider the influence of environmental factors such as light intensity and air quality, resulting in inaccurate identification results and affecting the development of subsequent control strategies.

[0004] Therefore, the present application proposes a multi-agent management system and method for collaborative work of robots and unmanned aerial vehicles to solve the above problems. SUMMARY

[0005] The present application aims to at least solve one of the technical problems existing in the prior art; for this purpose, the present application proposes a multi-agent management system and method for collaborative work of robots and unmanned aerial vehicles to solve the technical problem that a single agent can usually only complete a specific task, lacks collaborative work capability, and cannot fully utilize the advantages of each agent. At the same time, although existing technologies can identify diseases and pests, they do not adequately consider the influence of environmental factors such as light intensity and air quality, resulting in inaccurate identification results and affecting the development of subsequent control strategies.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application provides a multi-agent management system for collaborative work of robots and unmanned aerial vehicles, comprising: a data acquisition module, a data analysis module, and a multi-agent management module.

[0007] The data acquisition module is used to acquire image data of crops and corresponding environmental data in real time.

[0008] The data analysis module trains an artificial intelligence model based on historical image data to obtain a disease and pest identification model; identifies the real-time acquired image data of crops based on the disease and pest identification model to obtain an initial disease and pest coverage rate; and

[0009] construct a correction model based on the environment data, and correct the initial pest coverage to obtain a target pest coverage;

[0010] a multi-agent management module: calculating a pesticide spraying amount of the pesticide spraying robot based on the target pest coverage; obtaining an optimal path of the pesticide spraying robot to the pest coverage area through a path planning algorithm; and

[0011] moving to the pest coverage area based on the optimal path, and spraying pesticides in the pest coverage area based on the pesticide spraying amount.

[0012] Preferably, the real-time image data of the crops and the corresponding environment data are collected, including:

[0013] dividing the farmland into a plurality of sub-areas according to a preset proportion; wherein each of the plurality of sub-areas is a rectangle.

[0014] real-time image data of the crops in the sub-areas are collected by an image collection device installed on the unmanned aerial vehicle.

[0015] environment data of the farmland are collected by a data sensor in real time; wherein the environment data include: light intensity and air quality index.

[0016] It should be noted that the preset proportion is set according to the resolution and definition of the image collection device to ensure that the collected sub-area crop images are clear.

[0017] Preferably, the artificial intelligence model is trained based on the historical image data, including:

[0018] extracting the crop images of the sub-areas in the historical image data;

[0019] obtaining the coverage of the pests in the sub-areas in the corresponding crop images based on the crop images of the sub-areas in the historical image data by artificial marking, and marking the coverage of the pests in the sub-areas as the pest coverage;

[0020] integrating the crop images of the sub-areas in the historical image data into standard input data; integrating the pest coverage of the sub-areas in the historical image data into standard output data; training an artificial intelligence model based on the standard input data and the standard output data to obtain a pest recognition model; wherein the artificial intelligence model includes: a YOLO model or a deep belief network.

[0021] Preferably, the real-time collected image data of the crops is identified based on the pest recognition model, including:

[0022] extracting the images of the crops in the sub-areas collected in real time;

[0023] The image of the crops in the sub-region collected in real time is identified by the pest and disease identification model to obtain the pest and disease coverage corresponding to the image of the crops in the sub-region collected in real time, and the pest and disease coverage corresponding to the image of the crops in the sub-region collected in real time is marked as an initial pest and disease coverage.

[0024] Preferably, the initial pest and disease coverage is corrected based on the environmental data, and the correction model is constructed, and the correction model comprises:

[0025] The light intensity is marked as G, and the air quality index is marked as K;

[0026] The light intensity and the air quality index are used as independent variables, the corrected pest and disease coverage is used as a dependent variable, and the corrected pest and disease coverage is marked as a target pest and disease coverage MP;

[0027] The independent variables and the dependent variables are fitted by a polynomial fitting method to establish a correction model, and the correction model is specifically: MP=CP×[1+α / [e^(G / (G+1))]+β×e^(K / (K+1))],α and β are proportional coefficients, and CP is the initial pest and disease coverage;

[0028] For the proportional coefficient α: if G<Gmin or G>Gmax, then 0<α<1; if Gmin≤G≤Gmax, then -1<α<0; wherein Gmin is the minimum value of the optimal light intensity range of crops, and Gmax is the maximum value of the optimal light intensity range of crops;

[0029] For the proportional coefficient β: if K≤YK, then -1<β<0; if K>YK, then 0<β<1; wherein YK is a preset air quality index threshold;

[0030] The initial pest and disease coverage output by the pest and disease identification model is extracted, and the initial pest and disease coverage is input into the correction model to obtain the target pest and disease coverage.

[0031] It should be noted that the proportional coefficient is set by experts in the field according to experience; and the preset air quality index threshold is set by experts in the field according to the average value of the air quality index corresponding to the image data in the training data of the pest and disease identification model at the collection time;

[0032] The light intensity has a significant impact on the health of crops and the coverage of pests and diseases; when the light intensity is lower than the optimal light intensity, the photosynthetic efficiency of plants decreases, the growth is slow and the resistance is weakened, and the plants are easily attacked by fungi, bacteria and pests, resulting in an increase in the coverage of pests and diseases; when the light intensity exceeds the optimal light intensity, the plants will suffer from photoinhibition and photodamage, the leaf temperature rises to cause heat stress response, the damaged tissues are more susceptible to pathogenic bacteria and attract pests, which also leads to an increase in the coverage of pests and diseases; the higher the air quality index, the more serious the air pollution, and the pollutants will cause harm to plants, reduce their photosynthetic efficiency and disease resistance, and make the plants more susceptible to pests and diseases, thereby leading to an increase in the coverage of pests and diseases; on the contrary, fresh air is conducive to the normal growth and metabolism of plants and enhances the disease resistance of plants.

[0033] Preferably, the pesticide spraying robot calculates the pesticide spraying amount based on the target pest and disease coverage, comprising:

[0034] Extracting the total area of the sub-region and the corresponding target pest and disease coverage;

[0035] By calculating the product of the total area of the sub-region and the corresponding target pest and disease coverage, the actual pesticide spraying area is obtained;

[0036] By calculating the product of the pesticide spraying amount per unit area and the actual pesticide spraying area, the pesticide spraying amount of each sub-region containing pests and diseases is obtained.

[0037] Preferably, the pesticide spraying robot obtains the optimal path to the pest and disease coverage area through a path planning algorithm, comprising:

[0038] A two-dimensional model of the farmland is constructed by modeling software, and a rectangular coordinate system is constructed with the center point of the farmland as the origin in the two-dimensional model of the farmland;

[0039] The positions of the pesticide spraying robot and the target sub-region are obtained and marked as position information; wherein the target sub-region refers to the sub-region containing pests and diseases;

[0040] The optimal path of the pesticide spraying robot to the target sub-region is obtained based on the position information through a path planning algorithm; wherein the path planning algorithm includes A* algorithm or Dijkstra algorithm.

[0041] Preferably, the positions of the pesticide spraying robot and the target sub-region are obtained, comprising:

[0042] The position of the current pesticide spraying robot is obtained through the GPS device installed on the pesticide spraying robot, and is mapped into the rectangular coordinate system on the two-dimensional model of the farmland to obtain the position coordinates of the pesticide spraying robot;

[0043] Extract the coordinates of the four vertices of the target sub-region in the rectangular coordinate system on the planar two-dimensional model of the farmland, and mark them as the vertex coordinates of the target sub-region;

[0044] Integrate the position coordinates of the pesticide spraying robot and the vertex coordinates of the target sub-region into position information.

[0045] Preferably, the pesticide spraying robot moves to the target sub-region based on the optimal path, and sprays pesticides on the disease and pest coverage area based on the pesticide spraying amount.

[0046] The pesticide spraying robot moves to the vertex position of the nearest target sub-region based on the optimal path.

[0047] Obtain the disease and pest coverage area in the target sub-region based on computer vision technology; wherein the computer vision technology includes edge detection or color segmentation.

[0048] Extract the calculated pesticide spraying amount of the target sub-region, and spray pesticides on the disease and pest coverage area in the target sub-region based on the pesticide spraying amount of the target sub-region.

[0049] The second aspect of the present application provides a multi-agent management method for cooperative work of a robot and a drone, comprising:

[0050] Step one: real-time acquisition of image data and corresponding environment data of crops;

[0051] Step two: training an artificial intelligence model based on historical image data to obtain a disease and pest identification model; identifying the real-time acquired image data of crops based on the disease and pest identification model to obtain an initial disease and pest coverage rate;

[0052] Step three: constructing a correction model based on the environment data and correcting the initial disease and pest coverage rate to obtain a target disease and pest coverage rate;

[0053] Step four: calculating the pesticide spraying amount of the pesticide spraying robot based on the target disease and pest coverage rate;

[0054] Step five: obtaining the optimal path of the pesticide spraying robot to the disease and pest coverage area through a path planning algorithm;

[0055] Step six: moving to the disease and pest coverage area based on the optimal path, and spraying pesticides on the disease and pest coverage area based on the pesticide spraying amount.

[0056] Compared with the prior art, the present application has the following advantages:

[0057] 1.The present application obtains high-resolution images and environmental data of crops in real time by using unmanned aerial vehicles, and ensures comprehensive coverage of large areas of farmland and reduces the time and cost of manual inspection by precise positioning through the GPS device installed on the pesticide spraying robot; secondly, the collected images are automatically identified by an artificial intelligence model to obtain the initial disease and pest coverage rate, and a correction model is constructed combining environmental factors to further improve the identification accuracy, solving the problem that although the prior art can identify diseases and pests, it does not consider the influence of environmental factors such as light intensity and air quality, resulting in inaccurate identification results and affecting the subsequent prevention strategy; based on the target disease and pest coverage rate, the actual pesticide spraying amount of each sub-area is calculated to ensure accurate spraying and reduce resource waste and environmental pollution.

[0058] 2.The present application calculates the pesticide spraying amount of the pesticide spraying robot based on the target disease and pest coverage rate, calculates the actual pesticide spraying area by extracting the total area of the sub-area and the corresponding target disease and pest coverage rate, and accurately determines the pesticide spraying amount of each sub-area by combining the pesticide spraying amount per unit area, ensuring accurate pesticide application and avoiding excessive or insufficient spraying, thereby reducing resource waste and environmental pollution; secondly, the optimal path of the pesticide spraying robot to the disease and pest coverage area is obtained through a path planning algorithm, a two-dimensional model of the farmland is constructed using modeling software, and a rectangular coordinate system is established with the center point of the farmland as the origin to accurately locate the position information of the pesticide spraying robot and the target sub-area, ensuring the accuracy of the path planning; the path planning algorithm optimizes the path to improve the moving efficiency of the pesticide spraying robot and reduce unnecessary moving time and energy consumption; in addition, the real-time position of the pesticide spraying robot is obtained through the GPS device and mapped onto the two-dimensional model, and combined with the vertex coordinates of the target sub-area, accurate position information integration is realized, further enhancing the reliability and operation of the system, not only greatly improving the accuracy and efficiency of disease and pest control, but also effectively reducing labor costs and environmental burden, providing an efficient and intelligent solution for modern agriculture. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, brief descriptions will be given below of the drawings needed to be used in the embodiments or prior art descriptions. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0060] Fig. 1 The system module schematic diagram of the embodiment of the present application;

[0061] Fig. 2 The specific flowchart of the embodiment of the present application. DETAILED DESCRIPTION

[0062] The technical solutions of the present application will be described clearly and completely below in conjunction with the embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0063] Please refer to Figs. 1-2 The first aspect of the embodiment of the present application provides a multi-agent management system for cooperative work of robots and unmanned aerial vehicles, comprising: a data acquisition module, a data analysis module, and a multi-agent management module.

[0064] The data acquisition module is configured to acquire image data of crops and corresponding environmental data in real time.

[0065] The data analysis module is configured to train an artificial intelligence model based on historical image data to obtain a pest and disease identification model, identify the image data of the crops acquired in real time based on the pest and disease identification model to obtain an initial pest and disease coverage rate, construct a correction model based on the environmental data, and correct the initial pest and disease coverage rate to obtain a target pest and disease coverage rate.

[0066] The multi-agent management module is configured to calculate a pesticide spraying amount of a pesticide spraying robot based on the target pest and disease coverage rate, obtain an optimal path of the pesticide spraying robot to a pest and disease coverage area through a path planning algorithm, and move to the pest and disease coverage area based on the optimal path and spray pesticides in the pest and disease coverage area based on the pesticide spraying amount.

[0067] The image data of the crops and the corresponding environmental data are acquired in real time, including:

[0068] The farmland is divided into a plurality of sub-areas according to a preset proportion, and each of the sub-areas is a rectangle.

[0069] An image acquisition device installed on the unmanned aerial vehicle is configured to acquire images of the crops in the sub-areas in real time.

[0070] A data sensor is configured to acquire environmental data of the farmland in real time, and the environmental data includes illumination intensity and air quality index.

[0071] The artificial intelligence model is trained based on the historical image data, including:

[0072] The crop images of the sub-areas in the historical image data are extracted.

[0073] The coverage rate of the disease and insect pests in the sub-region in the corresponding crop image is obtained by manually marking the crop image in the sub-region in the historical image data, and the coverage rate of the disease and insect pests in the sub-region is marked as the disease and insect pest coverage rate.

[0074] The crop image in the sub-region in the historical image data is integrated as standard input data, and the disease and insect pest coverage rate corresponding to the crop image in the sub-region in the historical image data is integrated as standard output data; an artificial intelligence model is trained based on the standard input data and the standard output data to obtain a disease and insect pest recognition model; wherein the artificial intelligence model includes a YOLO model or a deep belief network.

[0075] Based on the disease and insect pest recognition model, the image data of the real-time collected crop is recognized, including:

[0076] The image of the real-time collected crop in the sub-region is extracted.

[0077] The image of the real-time collected crop in the sub-region is recognized by the disease and insect pest recognition model, and the disease and insect pest coverage rate corresponding to the image of the real-time collected crop in the sub-region is obtained, and the disease and insect pest coverage rate corresponding to the image of the real-time collected crop in the sub-region is marked as the initial disease and insect pest coverage rate.

[0078] A correction model is constructed based on environmental data, and the initial disease and insect pest coverage rate is corrected, including:

[0079] The light intensity is marked as G, and the air quality index is marked as K.

[0080] The light intensity and the air quality index are taken as independent variables; the corrected disease and insect pest coverage rate is taken as the dependent variable and is marked as the target disease and insect pest coverage rate MP;

[0081] The independent variables and the dependent variables are fitted by a polynomial fitting method to establish a correction model; wherein the correction model is specifically: MP=CP×[1+α / [e^(G / (G+1))]+β×e^(K / (K+1))],α、β are proportional coefficients, and CP is the initial disease and insect pest coverage rate.

[0082] For the proportional coefficient α: if G<Gmin or G>Gmax, then 0<α<1; if Gmin≤G≤Gmax, then -1<α<0; wherein Gmin is the minimum value of the optimal light intensity range of the crop, and Gmax is the maximum value of the optimal light intensity range of the crop.

[0083] For the proportional coefficient β: if K≤YK, then -1<β<0; if K>YK, then 0<β<1; wherein YK is a preset air quality index threshold.

[0084] The initial pest coverage rate output by the pest identification model is extracted and input into the correction model to obtain a target pest coverage rate.

[0085] The pesticide spraying amount of the pesticide spraying robot is calculated based on the target pest coverage rate, including:

[0086] The total area of the sub-region and the corresponding target pest coverage rate are extracted;

[0087] The actual pesticide spraying area is obtained by calculating the product of the total area of the sub-region and the corresponding target pest coverage rate;

[0088] The pesticide spraying amount of each sub-region containing pests is obtained by calculating the product of the pesticide spraying amount per unit area and the actual pesticide spraying area.

[0089] The optimal path of the pesticide spraying robot to the pest coverage area is obtained through a path planning algorithm, including:

[0090] A two-dimensional planar model of the farmland is constructed by modeling software, and a two-dimensional planar coordinate system is constructed with the center point of the farmland as the origin in the two-dimensional planar model of the farmland;

[0091] The positions of the pesticide spraying robot and the target sub-region are obtained and marked as position information; wherein the target sub-region refers to a sub-region containing pests;

[0092] The optimal path of the pesticide spraying robot to the target sub-region is obtained based on the position information through a path planning algorithm; wherein the path planning algorithm includes A* algorithm or Dijkstra algorithm.

[0093] The positions of the pesticide spraying robot and the target sub-region are obtained, including:

[0094] The position of the current pesticide spraying robot is obtained through the GPS device installed on the pesticide spraying robot and mapped into the rectangular coordinate system on the two-dimensional planar model of the farmland to obtain the position coordinates of the pesticide spraying robot;

[0095] The coordinates of the four vertices of the target sub-region in the rectangular coordinate system on the two-dimensional planar model of the farmland are extracted and marked as the vertex coordinates of the target sub-region;

[0096] The position coordinates of the pesticide spraying robot and the vertex coordinates of the target sub-region are integrated as position information.

[0097] Based on the optimal path, move to the pest coverage area, and based on the pesticide spraying amount, spray pesticides on the pest coverage area, including:

[0098] The pesticide spraying robot moves to the vertex position of the nearest target sub-region based on the optimal path;

[0099] obtaining the area covered by the pests in the target sub-region based on computer vision technology; wherein the computer vision technology includes edge detection or color segmentation;

[0100] extracting the spraying amount of the target sub-region calculated, and spraying pesticides in the area covered by the pests in the target sub-region based on the spraying amount of the target sub-region.

[0101] For example: assuming there is a large farmland of 1 hectare that needs to be treated for pests and diseases; in order to achieve efficient and accurate pest and disease control, a multi-agent management system and method for cooperative work of robots and unmanned aerial vehicles is used, which is as follows:

[0102] 1. Data acquisition module;

[0103] Step 1: Divide the farmland sub-region;

[0104] Divide the farmland into several rectangular sub-regions according to the preset proportion, and the area of each sub-region is 10 square meters.

[0105] Use an unmanned aerial vehicle to carry a high-resolution camera to collect image data of crops in each sub-region in real time, and collect environmental data (such as light intensity and air quality index) in real time through sensors.

[0106] Example data:

[0107] Total number of sub-regions: 1000;

[0108] Area of each sub-region: 10 square meters;

[0109] Light intensity: 60000 lux;

[0110] Air quality index: 100;

[0111] 2. Data analysis module;

[0112] Step 2: Train artificial intelligence model;

[0113] Based on the historical image data (including pest and disease marks) and corresponding environmental data accumulated in the past few years, the YOLO model is trained to obtain a preliminary pest and disease recognition model;

[0114] Extract the sub-region crop images in the historical image data, and obtain the corresponding pest and disease coverage rate through artificial marking.

[0115] Integrate the historical image data into standard input data, and integrate the pest and disease coverage rate into standard output data, and train the YOLO model as a pest and disease recognition model.

[0116] Step 3: Identify pests and diseases and correct coverage rate;

[0117] The initial disease and pest coverage is obtained by recognizing the real-time collected sub-area crop image using the trained YOLO model.

[0118] A correction model is constructed by combining the light intensity and air quality index to further improve the recognition accuracy.

[0119] Example data:

[0120] Initial disease and pest coverage: 50%;

[0121] Light intensity: 60000 lux;

[0122] Air quality index: 100;

[0123] Suppose the preset air quality index threshold is 70;

[0124] Since the air quality index K(100) is greater than the preset air quality index threshold YK(70), 0<β<1;

[0125] Suppose the optimal light intensity range of the crop is 40000 lux to 50000 lux; the minimum value Gmin of the optimal light intensity range of the crop is 40000 lux, and the maximum value Gmax of the optimal light intensity range of the crop is 50000 lux;

[0126] Since the light intensity G(60000 lux) is greater than the maximum value Gmax(50000 lux) of the optimal light intensity range of the crop, the value range of the proportion coefficient α is 0<α<1;

[0127] Correction model formula: MP=CP×[1+α / [e^(G / (G+1))]+β×e^(K / (K+1))];

[0128] Suppose the proportion coefficient α=0.5 and β=0.06;

[0129] Calculate the target disease and pest coverage: MP≈67.25% is obtained by substituting the data;

[0130] 3. Multi-agent management module;

[0131] Step 4: Calculate the pesticide spraying amount;

[0132] Extract the total area of the sub-area and the corresponding target disease and pest coverage, and calculate the actual pesticide spraying area.

[0133] Combine the pesticide spraying amount per unit area (assuming 0.5 liters of pesticide per square meter) to calculate the pesticide spraying amount of each sub-area.

[0134] Example data:

[0135] Sub-area area: 10 square meters;

[0136] Target pest coverage: 67.25%;

[0137] Unit area pesticide spraying amount: 0.5 liters / square meter;

[0138] Actual pesticide spraying area: 10 square meters x 67.25% = 6.725 square meters;

[0139] Total pesticide spraying amount: 6.725 square meters x 0.5 liters / square meter = 3.3625 liters;

[0140] Step 5: Path planning;

[0141] Construct a two-dimensional model of the farmland and establish a rectangular coordinate system with the farmland center point as the origin.

[0142] Obtain the position information of the pesticide spraying robot and the target sub-area, and calculate the optimal path using A* algorithm or Dijkstra algorithm.

[0143] Example data:

[0144] Farmland center point coordinates: (0, 0);

[0145] Current position coordinates of the pesticide spraying robot: (500, 500);

[0146] Target sub-area vertex coordinates: (1000, 1000), (1000, 1100), (1100, 1000), (1100, 1100);

[0147] Calculate the optimal path using A* algorithm;

[0148] Step 6: Move and spray;

[0149] The pesticide spraying robot moves to the nearest target sub-area vertex position based on the optimal path.

[0150] Use computer vision technology (such as edge detection or color segmentation) to obtain the specific area covered by pests and diseases in the target sub-area, and perform accurate spraying according to the calculated pesticide spraying amount (3.3625 liters).

[0151] As can be seen from the above examples, the present application automatically identifies the disease and pest coverage rate through a deep learning model (such as YOLO), and constructs a correction model in combination with environmental factors (light intensity and air quality index), thereby improving the identification accuracy and ensuring the accuracy of subsequent prevention strategies; the pesticide spraying amount is accurately calculated according to the target disease and pest coverage rate and the sub-area area, thereby avoiding excessive or insufficient spraying and reducing resource waste and environmental pollution; the movement path of the pesticide spraying robot is optimized through A* or Dijkstra algorithm, thereby reducing unnecessary movement time and energy consumption and improving the overall operation efficiency; the image data and environmental data are collected in real time by the unmanned aerial vehicle, and the GPS device is used for accurate positioning, thereby realizing comprehensive coverage and automatic operation of large-area farmland and greatly reducing the time and cost of manual inspection.

[0152] The second aspect embodiment of the present application provides a multi-agent management method for cooperative operation of a robot and an unmanned aerial vehicle, comprising:

[0153] Step one: collecting image data of crops and corresponding environmental data in real time;

[0154] Step two: training an artificial intelligence model based on historical image data to obtain a disease and pest identification model; identifying the image data of the crops collected in real time based on the disease and pest identification model to obtain an initial disease and pest coverage rate;

[0155] Step three: constructing a correction model based on the environmental data and correcting the initial disease and pest coverage rate to obtain a target disease and pest coverage rate;

[0156] Step four: calculating the pesticide spraying amount of the pesticide spraying robot based on the target disease and pest coverage rate;

[0157] Step five: obtaining the optimal path of the pesticide spraying robot to the disease and pest coverage area through a path planning algorithm;

[0158] Step six: moving to the disease and pest coverage area based on the optimal path, and spraying pesticides in the disease and pest coverage area based on the pesticide spraying amount.

[0159] Some data in the above formula are calculated by removing the dimension and taking the numerical value, and the formula is obtained by software simulation of a large amount of collected data to be closest to the actual situation; the preset parameters and the preset threshold in the formula are set by a person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0160] The above embodiments are only used to illustrate the technical method of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A multi-agent management system for collaborative operation of robots and drones, characterized in that, include: Data acquisition module, data analysis module, multi-agent management module; Data acquisition module: used to collect real-time image data of crops and corresponding environmental data; Data analysis module: Trains an artificial intelligence model based on historical image data to obtain a pest and disease identification model; identifies real-time collected crop image data based on the pest and disease identification model to obtain the initial pest and disease coverage rate; as well as, A modified model was constructed based on environmental data, and the initial pest and disease coverage rate was corrected to obtain the target pest and disease coverage rate; among which, A revised model was constructed based on environmental data, and the initial pest and disease coverage rate was revised, including: Light intensity is labeled G; air quality index is labeled K. Light intensity and air quality index were used as independent variables; the modified pest and disease coverage rate was used as the dependent variable and labeled as the target pest and disease coverage rate MP. The independent and dependent variables are fitted using a multinomial fitting method to establish a modified model; the modified model is specifically: MP=CP×[1+α / [e^(G / (G+1))]+β×e^(K / (K+1))], where α and β are proportionality coefficients, and CP is the initial pest and disease coverage rate; For the proportionality coefficient α: if G < Gmin or G > Gmax, then 0 < α < 1; if Gmin ≤ G ≤ Gmax, then -1 < α < 0; where Gmin is the minimum value of the optimal light intensity range for crops, and Gmax is the maximum value of the optimal light intensity range for crops. For the proportionality coefficient β: if K≤YK, then -1<β<0; if K>YK, then 0<β<1; where YK is the preset air quality index threshold. Extract the initial pest and disease coverage rate output by the pest and disease identification model, and input the initial pest and disease coverage rate into the correction model to obtain the target pest and disease coverage rate; Multi-agent management module: Calculates the amount of pesticide sprayed by the pesticide spraying robot based on the target pest and disease coverage rate; obtains the optimal path from the pesticide spraying robot to the pest and disease coverage area through a path planning algorithm; and, Move to the pest-infested area based on the optimal path, and spray pesticides on the pest-infested area based on the amount of pesticide to be sprayed.

2. The multi-agent management system for collaborative operation of robots and drones according to claim 1, characterized in that, The real-time acquisition of crop image data and corresponding environmental data includes: The farmland is divided into several sub-regions according to a preset ratio; each sub-region is rectangular. Images of crops in a sub-region are acquired in real time using image acquisition equipment installed on a drone; Environmental data of farmland is collected in real time through data sensors; the environmental data includes light intensity and air quality index.

3. The multi-agent management system for collaborative operation of robots and drones according to claim 1, characterized in that, The training of the artificial intelligence model based on historical image data includes: Extract crop images from sub-regions of historical image data; Based on the crop images of sub-regions in historical image data, the coverage rate of pests and diseases in the corresponding crop images in the sub-regions is obtained by manual labeling, and the coverage rate of pests and diseases in the sub-regions is labeled as the pest and disease coverage rate. By integrating crop images of sub-regions in historical image data into standard input data; integrating the pest and disease coverage rates corresponding to crop images of sub-regions in historical image data into standard output data; training an artificial intelligence model based on the standard input data and standard output data to obtain a pest and disease identification model; wherein, the artificial intelligence model includes: YOLO model or deep belief network.

4. A multi-agent management system for collaborative operation of robots and drones according to claim 1, characterized in that, The identification of crop image data collected in real time based on the pest and disease identification model includes: Extract images of crops from sub-regions acquired in real time; The pest and disease identification model is used to identify crop images in real-time collected sub-regions, and the pest and disease coverage rate corresponding to the crop images in real-time collected sub-regions is obtained. The pest and disease coverage rate corresponding to the crop images in real-time collected sub-regions is marked as the initial pest and disease coverage rate.

5. A multi-agent management system for collaborative operation of robots and drones according to claim 1, characterized in that, The amount of pesticide sprayed by the pesticide spraying robot, calculated based on the target pest and disease coverage rate, includes: Extract the total area of ​​the sub-region and the corresponding target pest and disease coverage rate; The actual spraying area is obtained by multiplying the total area of ​​the sub-region by the corresponding target pest and disease coverage rate. The amount of pesticide to be sprayed in each sub-region containing pests and diseases is obtained by multiplying the amount of pesticide sprayed per unit area by the actual area sprayed.

6. A multi-agent management system for collaborative operation of robots and drones according to claim 1, characterized in that, The process of obtaining the optimal path from the pesticide spraying robot to the pest-covered area using a path planning algorithm includes: A two-dimensional planar model of the farmland is constructed using modeling software, and a Cartesian coordinate system is built in the two-dimensional planar model of the farmland with the center point of the farmland as the origin. The locations of the pesticide spraying robot and the target sub-region are obtained and marked as location information; the target sub-region refers to the sub-region containing pests and diseases. The optimal path from the pesticide spraying robot to the target sub-region is obtained based on the location information through a path planning algorithm; the path planning algorithm includes: A* algorithm or Dijkstra algorithm.

7. A multi-agent management system for collaborative operation of robots and drones according to claim 6, characterized in that, The acquisition of the location of the pesticide spraying robot and the target sub-region includes: The current position of the pesticide spraying robot is obtained by using a GPS device installed on the robot and mapping it to a Cartesian coordinate system on a two-dimensional model of the farmland to obtain the position coordinates of the pesticide spraying robot. Extract the coordinates of the four vertices of the target sub-region in the Cartesian coordinate system on the two-dimensional planar model of the farmland, and mark them as the vertex coordinates of the target sub-region; The position coordinates of the pesticide spraying robot and the vertex coordinates of the target sub-region are integrated into position information.

8. A multi-agent management system for collaborative operation of robots and drones according to claim 1, characterized in that, The process of moving to the pest-infested area based on the optimal path and spraying pesticides on the pest-infested area based on the amount of pesticide applied includes: The pesticide spraying robot moves to the vertex position of the nearest target sub-region based on the optimal path; The area covered by pests and diseases in the target sub-region is obtained based on computer vision technology; among which, computer vision technology includes: edge detection or color segmentation; Extract the calculated pesticide application amount for the target sub-region, and spray pesticides on the areas covered by pests and diseases in the target sub-region based on the pesticide application amount for the target sub-region.

9. A multi-agent management method for collaborative operation of robots and drones, applied to the multi-agent management system for collaborative operation of robots and drones as described in any one of claims 1-8, characterized in that, include: Step 1: Real-time acquisition of crop image data and corresponding environmental data; Step 2: Train an artificial intelligence model based on historical image data to obtain a pest and disease identification model; use the pest and disease identification model to identify real-time collected crop image data to obtain the initial pest and disease coverage rate; Step 3: Construct a revised model based on environmental data and revise the initial pest and disease coverage rate to obtain the target pest and disease coverage rate, including: A revised model was constructed based on environmental data, and the initial pest and disease coverage rate was revised, including: Light intensity is labeled G; air quality index is labeled K. Light intensity and air quality index were used as independent variables; the modified pest and disease coverage rate was used as the dependent variable and labeled as the target pest and disease coverage rate MP. The independent and dependent variables are fitted using a multinomial fitting method to establish a modified model; the modified model is specifically: MP=CP×[1+α / [e^(G / (G+1))]+β×e^(K / (K+1))], where α and β are proportionality coefficients, and CP is the initial pest and disease coverage rate; For the proportionality coefficient α: if G < Gmin or G > Gmax, then 0 < α < 1; if Gmin ≤ G ≤ Gmax, then -1 < α < 0; where Gmin is the minimum value of the optimal light intensity range for crops, and Gmax is the maximum value of the optimal light intensity range for crops. For the proportionality coefficient β: if K≤YK, then -1<β<0; if K>YK, then 0<β<1; where YK is the preset air quality index threshold. Extract the initial pest and disease coverage rate output by the pest and disease identification model, and input the initial pest and disease coverage rate into the correction model to obtain the target pest and disease coverage rate; Step 4: Calculate the amount of pesticide to be sprayed by the pesticide spraying robot based on the target pest and disease coverage rate; Step 5: Obtain the optimal path from the pesticide spraying robot to the pest-covered area using a path planning algorithm; Step 6: Move to the pest-infested area based on the optimal path, and spray pesticides on the pest-infested area based on the amount of pesticide to be sprayed.

Images