Plant protection unmanned aerial vehicle path planning method based on end-side cooperation and plant protection system

Abstract

The plant protection unmanned aerial vehicle path planning method based on end-side cooperation comprises the steps that off-line map data of a to-be-operated area is acquired, a multi-layer planning map is constructed based on the off-line map data, and the multi-layer planning map comprises a plurality of power conversion points and a plurality of chemical adding points; establishing a residual electric quantity model and a residual explosive quantity model based on the characteristic parameters of the plant protection unmanned aerial vehicle; constructing a multi-objective optimization function based on the residual electric quantity model, the residual explosive quantity model and the navigation constraint condition; generating a plurality of initial path planning schemes, and screening the initial path planning schemes by using an improved genetic particle algorithm in combination with a multi-objective optimization function to obtain an optimal path planning scheme; and in the sailing process of the plant protection unmanned aerial vehicle based on the optimal path planning scheme, the optimal path planning scheme is optimized in real time based on the dynamic early warning information. According to the method, the problems of multi-constraint matching, dynamic path adjustment and end-side data collaboration in plant protection unmanned aerial vehicle operation can be solved.

Description

Technical Field

[0001] This invention relates to the field of agricultural drone technology, specifically to a path planning method and agricultural system for agricultural drones based on edge-end collaboration. Background Technology

[0002] Agricultural drones are essential equipment for efficient pest and disease control and crop fertilization in modern agricultural production. Equipped with spraying devices, they can perform precise spraying operations in farmland, orchards, and other areas, significantly improving operational efficiency and reducing labor costs. With the development of smart agriculture, higher demands are being placed on the operational accuracy, efficiency, and safety of agricultural drones. Path planning, as a core component of agricultural drone operations, directly impacts operational quality and cost.

[0003] Existing agricultural drone path planning methods are mostly based on single-dimensional optimization, such as only considering the shortest flight distance or the shortest operation time, without fully taking into account the dynamic changes in drone power consumption and pesticide dosage, or effectively integrating real-time data from edge servers with the drone's local decision-making capabilities. In actual operations, drones often experience mid-operation power outages due to inaccurate power estimations, and operations are interrupted due to unreasonable pesticide dosage planning. Furthermore, when faced with temporary construction areas and dynamic obstacles such as birds, they cannot quickly adjust their paths, which not only affects operational efficiency but may also lead to safety accidents. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a path planning method and system for agricultural drones based on edge-end collaboration. While solving the problems of multi-constraint matching, dynamic path adjustment, and edge-end data collaboration in agricultural drone operations, it optimizes the operation path, thereby improving operational efficiency, reducing operational costs, and ensuring operational safety.

[0005] To achieve the above objectives, the specific solution adopted by this invention is: a path planning method for agricultural drones based on edge-end collaboration, comprising: Obtain offline map data of the area to be worked on, and construct a multi-layer planning map based on the offline map data. The multi-layer planning map contains several battery swapping points and several chemical dosing points. Establish models for remaining battery power and remaining pesticide dosage based on the characteristic parameters of agricultural drones; A multi-objective optimization function is constructed based on the remaining power model, the remaining drug quantity model, and navigation constraints. Multiple initial path planning schemes are generated, and the optimal path planning scheme is obtained by screening the initial path planning schemes using an improved genetic particle algorithm combined with a multi-objective optimization function. During the flight of agricultural drones based on the optimal path planning scheme, the optimal path planning scheme is optimized in real time based on dynamic early warning information.

[0006] As a further optimization of the above-mentioned edge-coordinated agricultural drone path planning method, the method of constructing a multi-layer planning map based on offline map data includes: The offline map data of the work area is denoised, and terrain data, obstacle data, plot boundary data and functional node data are extracted from the offline map data. The functional node data includes several battery swapping points and several chemical dosing points. Construct a local Cartesian coordinate system for the area to be worked on, and convert the offline map data to the local Cartesian coordinate system to obtain the basic planning map; The basic planning map is divided into layers to obtain a multi-layer planning map, which includes a basic terrain layer, an obstacle layer, a functional node layer, and a work site layer.

[0007] As a further optimization of the above-mentioned edge-to-edge collaborative agricultural drone path planning method: after parsing the functional node data, the first attribute of the battery swapping point and the second attribute of the pesticide application point are determined. The first attribute includes the battery swapping coordinates, battery swapping time period, single battery swapping duration, maximum number of drones that can swap batteries simultaneously, and remaining battery power. The second attribute includes the pesticide application coordinates, pesticide type, single pesticide application duration, and remaining pesticide amount. After parsing the terrain data, the area to be operated on is divided into multiple plots based on the terrain data, and the third attribute of each plot is determined. The third attribute includes plot identifier, boundary coordinate set, plot area, plot type, and application parameters.

[0008] As a further optimization of the above-mentioned edge-cooperative agricultural drone path planning method: the drone's characteristic parameters include inherent parameters, weight parameters, energy consumption-related parameters, and pesticide consumption-related parameters. The inherent parameters include maximum flight speed, application speed, battery capacity, and tank volume. The weight parameters include net weight and pesticide density. The energy consumption-related parameters include unit flight energy consumption, basic hovering power, and payload power coefficient. The pesticide consumption-related parameters include pesticide application rate per unit area and application rate.

[0009] As a further optimization of the above-mentioned edge-cooperative agricultural drone path planning method, the remaining battery power model is as follows: ; in, The number of operation paths corresponding to the land parcel. It is a rounding function. The width of the plot in the first direction. For the flight path spacing of agricultural drones, This represents the total length of the work path. Let be the width of the plot in the second direction, and let the first direction be perpendicular to the second direction. For the duration of medication application, To increase the speed of drug application, Remaining battery power This is the initial charge level. For flight energy consumption, Energy consumption per unit of flight The distance for the drone to return sequentially to the nearest refueling point and the nearest battery swapping point. To reserve a safe amount of power.

[0010] As a further optimization of the above-mentioned edge-cooperative agricultural drone path planning method, the multi-objective optimization function is: ; ; ; ; ; ; in, This represents the total flight distance between functional nodes, where each functional node is either a battery swapping point or a refueling point. This represents the total length of the work path for all plots. For comprehensive flight power consumption, This represents the total electricity consumed during pesticide application. The number of turns between nodes. This represents the total number of turns along the work path within the site. This represents the remaining battery power deviation value, and has... , Let $\frac{ ... This represents the optimal remaining battery power at the battery swapping point. This represents the total number of battery swaps during the entire operation. The total number of times the medicine was applied throughout the entire operation. These are the weighting coefficients.

[0011] As a further optimization of the above-mentioned edge-cooperative agricultural drone path planning method, the method for generating multiple initial path planning schemes includes: The land parcels are clustered based on their location, and the first part of the scheme is generated based on the clustering results. The second part of the scheme is generated based on the first part of the scheme through random perturbation; Integrating the first and second parts of the plan yields the initial path planning scheme. Methods for selecting initial path planning schemes using an improved genetic particle algorithm combined with a multi-objective optimization function include: The initial path planning scheme is screened based on a multi-objective optimization function to obtain multiple alternative path planning schemes; The alternative path planning schemes are encoded, and a population containing multiple particles is generated based on the encoding results; Particles are classified as dominant or suboptimal based on their fitness values. The algorithm uses dynamic particle swarm optimization to iteratively optimize dominant particles and an improved genetic algorithm to iteratively optimize suboptimal particles. During the iterative optimization process, information exchange and cross-mutation are performed on some dominant particles and some suboptimal particles. After the iteration is completed, a set of Pareto optimal paths is generated based on particles; The optimal path planning scheme is obtained by further filtering the Pareto optimal path set.

[0012] As a further optimization of the above-mentioned edge-coordinated path planning method for agricultural drones, the dynamic early warning information includes high-priority early warning information, medium-priority early warning information and low-priority early warning information. Among them, the high-priority early warning information includes obstacle collision warning, low battery warning and low pesticide dosage warning, the medium-priority early warning information includes path deviation warning and wind speed exceeding limit warning, and the low-priority early warning information includes low battery warning and low pesticide dosage warning.

[0013] As a further optimization of the above-mentioned edge-cooperative agricultural drone path planning method, methods for real-time optimization of the optimal path planning scheme based on dynamic early warning information include: When a high-priority warning is triggered, the optimal path planning scheme is regenerated; when a medium-priority warning is triggered, the optimal path planning scheme is locally adjusted; and when a low-priority warning is triggered, the optimal path planning scheme is maintained.

[0014] The plant protection system includes multiple edge servers and multiple plant protection drone terminals. The edge servers are used to plan the optimal path planning scheme based on the above-mentioned edge-based collaborative plant protection drone path planning method. The plant protection drone terminals are used to navigate based on the optimal path planning scheme and spray pesticides on the plants.

[0015] Beneficial effects: This invention constructs accurate planning maps through offline map processing, providing basic data support for path planning; the established multi-constraint model can accurately match the dynamic changes in drone battery power and pesticide dosage, avoiding operation interruption; the multi-objective optimization function achieves synergistic optimization of operation efficiency, cost, and safety; the improved algorithm enhances the accuracy and convergence speed of path solving; the edge-end collaborative architecture ensures real-time path adjustment, enabling this invention to solve the problems of multi-constraint matching, dynamic path adjustment, and edge-end data collaboration in agricultural drone operations, while simultaneously optimizing the operation path, thereby improving operation efficiency, reducing operation costs, and ensuring operation safety. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram showing the result of the land parcel division. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] like Figure 1 and Figure 2 As shown, the present invention first provides a path planning method for agricultural drones based on edge-end collaboration, including S1 to S5.

[0019] S1. Obtain offline map data for the area to be worked on, and construct a multi-layer planning map based on the offline map data. The multi-layer planning map includes several battery swapping points and several chemical dosing points. The offline map data can be obtained through satellite remote sensing mapping, ground lidar scanning, or publicly available industry geographic information databases, and includes topography, obstacle distribution, plot boundaries, and coordinate information of preset functional nodes.

[0020] Furthermore, methods for constructing multi-layer planning maps based on offline map data include S11 to S13.

[0021] S11. Noise reduction is performed on the offline map data of the work area, and terrain data, obstacle data, plot boundary data, and functional node data are extracted from the offline map data. The functional node data includes several battery swapping points and several chemical dosing points. After extracting the functional node data, the first attribute of the battery swapping points and the second attribute of the chemical dosing points are determined.

[0022] The first attribute includes the battery swapping coordinates. Battery swapping time period Duration of a single battery swap Number of batteries Maximum number of drones that can swap batteries simultaneously and remaining battery power. By organizing the coordinates and first attribute of the i-th battery swapping point into a set, we can obtain the battery swapping information set of the i-th battery swapping point. By collecting and organizing the battery swapping information from all battery swapping points, we can obtain a set of battery swapping points. ,in This represents the total number of battery swapping points.

[0023] The second attribute includes the drug dosing coordinates. Pesticide types Duration of a single drug administration and remaining dosage After that, a set of dosing information for each dosing point can be obtained. By collecting and organizing the dosing information from all dosing points, we can obtain a dosing point set. ,in This represents the total number of drug dispensing points.

[0024] After parsing the terrain data, the area to be operated on is divided into multiple plots based on the terrain data, and the third attribute of each plot is determined, including the plot identifier. The set of boundary coordinates, plot area S, plot type, and application parameters are given. The application parameters include the application dose D and the application height range H. This allows the generation of the plot set. ,in This refers to the number of land parcels.

[0025] S12. Construct a local Cartesian coordinate system for the area to be worked on, and convert the offline map data to the local Cartesian coordinate system to obtain the basic planning map. The specific coordinate transformation method is a conventional technique in this field and will not be described in detail here.

[0026] S13. The basic planning map is divided into layers to obtain a multi-layer planning map, which includes a basic terrain layer, an obstacle layer, a functional node layer, and a work site layer. In one embodiment of the present invention, the multi-layer planning map can store structured data in SHP format and export a PNG format visualization map.

[0027] S2. Establish models for remaining battery power and remaining pesticide dosage based on the characteristic parameters of the agricultural drone. The characteristic parameters of the drone include inherent parameters, weight parameters, energy consumption-related parameters, and pesticide consumption-related parameters. Inherent parameters include maximum flight speed. Application speed Battery capacity Medicine box capacity Weight parameters include net weight. and pesticide density Energy consumption-related parameters include unit flight energy consumption. Basic hovering power and load power coefficient The parameters related to pesticide consumption include the amount of pesticide applied per unit area. and application rate .

[0028] The remaining power model is as follows: ; in, The number of operation paths corresponding to the land parcel. It is a rounding function. The width of the plot in the first direction. For the flight path spacing of agricultural drones, This represents the total length of the work path. Let be the width of the plot in the second direction, and let the first direction be perpendicular to the second direction. For the duration of medication application, To increase the speed of drug application, Remaining battery power This is the initial charge level. For flight energy consumption, Energy consumption per unit of flight The distance for the drone to return sequentially to the nearest refueling point and the nearest battery swapping point. To reserve a safe amount of power.

[0029] The model for remaining drug dosage is: ; like If the remaining amount of medicine is sufficient, the operation can continue; otherwise, it is necessary to go to the medicine dosing point to add more medicine.

[0030] Furthermore, a dynamic application energy consumption model can be introduced to monitor the application process in real time. The instantaneous dosage is The instantaneous total weight is Instantaneous power is The total energy consumption for pesticide application is After converting to milliampere-hours, we get ,in This refers to the operating voltage of the agricultural drone.

[0031] S3. Construct a multi-objective optimization function based on the remaining power model, remaining drug quantity model, and navigation constraints. More specifically, the multi-objective optimization function is: ; ; ; ; ; ; in, This represents the total flight distance between functional nodes, where each functional node is either a battery swapping point or a refueling point. This represents the total length of the work path for all plots. For comprehensive flight power consumption, This represents the total electricity consumed during pesticide application. The number of turns between nodes. This represents the total number of turns along the work path within the site. This represents the remaining battery power deviation value, and has... , Let $\frac{ ... This represents the optimal remaining battery power at the battery swapping point. This represents the total number of battery swaps during the entire operation. The total number of times the medicine was applied throughout the entire operation. These are the weighting coefficients.

[0032] Further, definition This represents the total flight distance between nodes. The function that minimizes the total flight distance by summing the operation path lengths of all plots is: ,have .definition This represents the total power consumption during flight. The total electricity consumed for pesticide application is given by [the relevant authority]. .definition This represents the number of turns on the flight path between nodes. The number of turns on the work path within the plot. .definition Let $\frac{ ... The optimal value for remaining battery capacity after battery swapping is determined by the battery capacity. 20%-30% of , The smaller the value, the closer the remaining battery capacity is to the optimal value after battery swapping. , The total number of battery swaps during the entire operation is calculated, with the optimal remaining battery power after swapping. , ,in The total number of chemical dosings throughout the operation must meet the following requirements. ,in Let be the remaining amount of medicine before the k-th dosing, and Minimize the number of times medication is added. .

[0033] In the above multi-objective optimization function, the weight coefficients The weighting can be flexibly adjusted according to the actual situation of the area to be worked on. For example, in large, contiguous areas of farmland, priority should be given to ensuring operational efficiency, and the weighting value can be... In scattered plots of land in hilly areas, safety and cost should be prioritized, and the weighting values ​​can be... .

[0034] Furthermore, in the multi-objective optimization function, path smoothness constraints can be added to adjust the turning angles of adjacent segments in the operational path. This means the turning radius should not exceed 30° to avoid instability caused by sharp turns in agricultural drones; the pesticide coverage rate for a single plot is constrained to ≥98%, determined by the flight path spacing d and the pesticide coverage width. The matching is achieved. In one embodiment of the present invention, This allows for a 20% overlap area.

[0035] S4. Generate multiple initial path planning schemes, and use an improved genetic particle algorithm combined with a multi-objective optimization function to select the optimal path planning scheme. When generating the initial path planning scheme, the land parcels can be encoded to obtain parcel codes. A parcel code includes three parts: parcel ID, functional node type, and path parameters. The parcel code can be directly converted into chromosomes in the genetic particle algorithm. For example, chromosomes... The order of operations is plots. Battery swapping points Plots and dosing point Based on this, methods for generating multiple initial path planning schemes include S41 to S43.

[0036] S41. Cluster the land parcels based on their location, and generate the first part of the solution based on the clustering results. In essence, this involves clustering particles based on chromosomes to form the first part of the solution. The first part of the solution primarily considers the scope of the work path, avoiding round trips across large areas.

[0037] S42. Based on the first part of the scheme, generate the second part of the scheme through random perturbation. After generating the second part of the scheme, it is necessary to check whether the first part of the scheme and the second part of the scheme meet the requirements of the remaining power model and the remaining drug quantity model. or The particles are removed. In one embodiment of the present invention, both the first and second part schemes include 50 operation paths, forming a population of size S=100.

[0038] S43. Integrate the first part of the scheme and the second part of the scheme to obtain the initial path planning scheme.

[0039] Methods for selecting initial path planning schemes by using an improved genetic particle algorithm combined with a multi-objective optimization function include S44 to S49.

[0040] S44. Based on the multi-objective optimization function, the initial path planning scheme is screened to obtain multiple alternative path planning schemes.

[0041] S45. Encode the alternative path planning schemes and generate a population containing multiple particles based on the encoding results. More specifically, the encoding results used when generating the initial path planning scheme can be used directly.

[0042] S46. Based on the fitness values ​​of particles, classify particles into dominant or suboptimal particles.

[0043] S47. Iterative optimization of dominant particles is performed using a dynamic particle swarm optimization algorithm, and iterative optimization of inferior particles is performed using an improved genetic algorithm. During the iterative optimization process, information exchange and cross-mutation are carried out between some dominant particles and some inferior particles.

[0044] During the iteration process of the dominant particle, a dynamic inertia weight is introduced. The calculation method for the dynamic inertia weight is as follows: ; Dynamic inertia weight is used to control the ability of dominant particles to maintain their original path direction and optimization trend. Furthermore, as the number of iterations increases, the dynamic inertia weight gradually decreases from its maximum value. Reduce to minimum value This enhances the global search capability of dominant particles in the early stages of iteration, avoids missing potential optimal operation paths across plots or functional nodes, and weakens the autonomous direction of dominant particles after iteration, pushing them toward the optimal operation path and improving convergence accuracy. Let be the current iteration number, and have ,in This is the preset maximum number of iterations. In one embodiment of the invention, the maximum value of the dynamic inertia weight... The value is 0.9, which is the minimum value of the dynamic inertia weight. The value is set to 0.4, and the maximum number of iterations is set to 100~200.

[0045] On the other hand, the learning factor of the dynamic particle swarm optimization algorithm and Adjust gradually with the number of iterations, in the early stages of iteration, i.e. During the process, Take the larger value. Taking a smaller value prioritizes strengthening the dominant particle's learning of its own historical best path, in the later stages of iteration, i.e. During the process, Gradually decrease, The number of particles gradually increases, thereby driving the dominant particles to learn the optimal path of the population, ensuring a balance between convergence accuracy and path diversity, and adapting to the multi-constraint path planning needs of agricultural drones.

[0046] It should also be noted that, in order to ensure the safety and reliability of the operation path, the position of the dominant particle is checked after each iteration. If it falls into the obstacle area, the path parameters are adjusted. The specific adjustment method can be the translation correction method, that is, the dominant particle is translated along the boundary normal direction of the obstacle area.

[0047] In the optimization process for inferior particles, the selection operator employs a tournament selection method, randomly selecting five particles as a group and choosing the two particles with the lowest fitness values ​​to advance to the next generation, thereby increasing the probability of retaining superior genes. The crossover operator uses a two-point crossover method, for example, for chromosomes. With chromosomes The intersection points are selected from segments 2 and 3 to generate offspring. and Crossover probability The mutation operator employs site mutation, which randomly selects one plot ID or functional node type from the chromosome for replacement, and the mutation probability... This is to avoid premature convergence.

[0048] Furthermore, every 5 iterations, the top 10 dominant particles with higher fitness values ​​are selected to interact with the top 5 inferior particles with higher fitness values, thereby achieving [the goal of] ... S48. After the iteration is completed, a set of Pareto optimal paths is generated based on particles.

[0049] S49. Further filtering is performed on the Pareto optimal path set to obtain the optimal path planning scheme.

[0050] S5. During the flight of the agricultural drone based on the optimal path planning scheme, the optimal path planning scheme is optimized in real time based on dynamic early warning information. The dynamic early warning information includes high-priority early warning information, medium-priority early warning information, and low-priority early warning information. High-priority early warning information includes obstacle collision warning, low battery warning, and low pesticide dosage warning. Medium-priority early warning information includes path deviation warning and wind speed exceeding limit warning. Low-priority early warning information includes low battery warning and low pesticide dosage warning.

[0051] Furthermore, the method for real-time optimization of the optimal path planning scheme based on dynamic early warning information includes: regenerating the optimal path planning scheme when a high-priority early warning information is triggered, locally adjusting the optimal path planning scheme when a medium-priority early warning information is triggered, and maintaining the optimal path planning scheme when a low-priority early warning information is triggered.

[0052] The present invention also provides a plant protection system, including multiple edge servers and multiple plant protection drone terminals. The edge servers are used to plan the optimal path planning scheme based on the above-mentioned edge-based collaborative plant protection drone path planning method. The plant protection drone terminals are used to navigate based on the optimal path planning scheme and spray pesticides on plants.

[0053] Furthermore, once the task is completed, the edge server automatically generates a report containing basic data, performance metrics, and optimization suggestions.

[0054] The basic data includes the total operating area, total duration, number of battery swaps / chemical dosings, actual power consumption, and actual chemical consumption.

[0055] Performance metrics include path deviation rate, pesticide coverage rate, and early warning processing success rate. The total deviation is related to the total flight distance. .

[0056] Optimization suggestions include adjusting the weight coefficients of the multi-objective optimization function based on actual data and correcting the energy consumption model. For example, if the actual power consumption is higher than the estimated value, then it will increase. .

[0057] Next, the data is associated with and stored with the initial planning data to form a work case library. When the case library reaches 50 cases, the gradient descent method is used to iteratively optimize the remaining power model. Compared with the residual drug quantity model This improves the accuracy of model predictions.

[0058] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A path planning method for agricultural drones based on edge-end collaboration, characterized in that, include: Obtain offline map data of the area to be worked on, and construct a multi-layer planning map based on the offline map data. The multi-layer planning map contains several battery swapping points and several chemical dosing points. Establish models for remaining battery power and remaining pesticide dosage based on the characteristic parameters of agricultural drones; A multi-objective optimization function is constructed based on the remaining power model, the remaining drug quantity model, and navigation constraints. Multiple initial path planning schemes are generated, and the optimal path planning scheme is obtained by screening the initial path planning schemes using an improved genetic particle algorithm combined with a multi-objective optimization function. During the flight of agricultural drones based on the optimal path planning scheme, the optimal path planning scheme is optimized in real time based on dynamic early warning information.

2. The path planning method for agricultural drones based on edge-end collaboration as described in claim 1, characterized in that, Methods for constructing multi-layer planning maps based on offline map data include: The offline map data of the work area is denoised, and terrain data, obstacle data, plot boundary data and functional node data are extracted from the offline map data. The functional node data includes several battery swapping points and several chemical dosing points. Construct a local Cartesian coordinate system for the area to be worked on, and convert the offline map data to the local Cartesian coordinate system to obtain the basic planning map; The basic planning map is divided into layers to obtain a multi-layer planning map, which includes a basic terrain layer, an obstacle layer, a functional node layer, and a work site layer.

3. The path planning method for agricultural drones based on edge-end collaboration as described in claim 2, characterized in that, After parsing the functional node data, the first attribute of the battery swapping point and the second attribute of the pesticide application point are determined. The first attribute includes the battery swapping coordinates, battery swapping time period, single battery swapping duration, maximum number of drones that can swap batteries simultaneously, and remaining battery power. The second attribute includes the pesticide application coordinates, pesticide type, single pesticide application duration, and remaining pesticide amount. After parsing the terrain data, the area to be operated on is divided into multiple plots based on the terrain data, and the third attribute of each plot is determined. The third attribute includes plot identifier, boundary coordinate set, plot area, plot type, and application parameters.

4. The path planning method for agricultural drones based on edge-end collaboration as described in claim 1, characterized in that, The characteristic parameters of a drone include inherent parameters, weight parameters, energy consumption-related parameters, and pesticide consumption-related parameters. Inherent parameters include maximum flight speed, application speed, battery capacity, and tank volume. Weight parameters include net weight and pesticide density. Energy consumption-related parameters include unit flight energy consumption, basic hovering power, and payload power coefficient. Pesticide consumption-related parameters include pesticide application rate per unit area and application rate.

5. The path planning method for agricultural drones based on edge-end collaboration as described in claim 4, characterized in that, The remaining power model is as follows: ； in, The number of operation paths corresponding to the land parcel. It is a rounding function. The width of the plot in the first direction. For the flight path spacing of agricultural drones, This represents the total length of the work path. Let be the width of the plot in the second direction, and let the first direction be perpendicular to the second direction. For the duration of medication application, To increase the speed of drug application, Remaining battery power This is the initial charge level. For flight energy consumption, Energy consumption per unit of flight The distance for the drone to return sequentially to the nearest refueling point and the nearest battery swapping point. To reserve a safe amount of power.

6. The path planning method for agricultural drones based on edge-end collaboration as described in claim 4, characterized in that, The multi-objective optimization function is: ； ； ； ； ； ； in, This represents the total flight distance between functional nodes, where each functional node is either a battery swapping point or a refueling point. This represents the total operation path length for all plots. For comprehensive flight power consumption, This represents the total electricity consumed during pesticide application. The number of turns between nodes. This represents the total number of turns along the work path within the site. This represents the remaining battery power deviation value, and has... , Let $\frac{ ... This represents the optimal remaining battery power at the battery swapping point. This represents the total number of battery swaps during the entire operation. The total number of times the medicine was applied throughout the entire operation. These are the weighting coefficients.

7. The path planning method for agricultural drones based on edge-end collaboration as described in claim 1, characterized in that, Methods for generating multiple initial path planning schemes include: The land parcels are clustered based on their location, and the first part of the scheme is generated based on the clustering results. The second part of the scheme is generated based on the first part of the scheme through random perturbation; Integrating the first and second parts of the plan yields the initial path planning scheme. Methods for selecting initial path planning schemes using an improved genetic particle algorithm combined with a multi-objective optimization function include: The initial path planning scheme is screened based on a multi-objective optimization function to obtain multiple alternative path planning schemes; The alternative path planning schemes are encoded, and a population containing multiple particles is generated based on the encoding results; Particles are classified as dominant or suboptimal based on their fitness values. The algorithm uses dynamic particle swarm optimization to iteratively optimize dominant particles and an improved genetic algorithm to iteratively optimize suboptimal particles. During the iterative optimization process, information exchange and cross-mutation are performed on some dominant particles and some suboptimal particles. After the iteration is completed, a set of Pareto optimal paths is generated based on particles; The optimal path planning scheme is obtained by further filtering the Pareto optimal path set.

8. The path planning method for agricultural drones based on edge-end collaboration as described in claim 1, characterized in that, Dynamic early warning information includes high-priority, medium-priority, and low-priority early warning information. High-priority early warning information includes obstacle collision warning, low battery warning, and low drug quantity warning. Medium-priority early warning information includes path deviation warning and wind speed exceeding limit warning. Low-priority early warning information includes low battery warning and low drug quantity warning.

9. The path planning method for agricultural drones based on edge-end collaboration as described in claim 8, characterized in that, Methods for real-time optimization of optimal path planning schemes based on dynamic early warning information include: When a high-priority warning is triggered, the optimal path planning scheme is regenerated; when a medium-priority warning is triggered, the optimal path planning scheme is locally adjusted; and when a low-priority warning is triggered, the optimal path planning scheme is maintained.

10. A plant protection system, characterized in that, It includes multiple edge servers and multiple agricultural drone terminals. The edge servers are used to plan the optimal path planning scheme based on the edge-to-edge collaborative agricultural drone path planning method as described in any one of claims 1-9. The agricultural drone terminals are used to navigate based on the optimal path planning scheme and spray pesticides on plants.

Images