Plant protection unmanned aerial vehicle intelligent path planning method and system based on multi-objective optimization

Abstract

The application belongs to the technical field of unmanned aerial vehicle operation control, and discloses a plant protection unmanned aerial vehicle intelligent path planning method and system based on multi-target optimization, which comprises the following steps: collecting multi-source data and processing the same to obtain processed data; constructing an energy consumption-pesticide efficacy multi-target optimization model according to the processed data, including a maximum prevention and treatment effect objective function and a minimum energy consumption objective function; solving the energy consumption-pesticide efficacy multi-target optimization model by using an improved NSGA-II algorithm, and generating a Pareto optimal solution set according to a constraint condition; and selecting a plurality of optimal solutions from the Pareto optimal solution set to obtain a three-dimensional flight path and corresponding flight parameters. The application realizes differentiated and accurate pesticide application for plant diseases and insect pests, solves the problems of uneven pesticide application, high energy consumption and poor adaptability of the traditional bow-shaped full coverage path, and significantly improves the accuracy, economy and environmental friendliness of plant protection operation.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) operation control technology, and relates to an intelligent path planning method and system for agricultural UAVs based on multi-objective optimization. Background Technology

[0002] Currently, agricultural drones have become one of the important means of pest and disease control in modern agriculture. They are highly efficient and adaptable, significantly reducing labor costs and improving the uniformity of pesticide application. However, existing agricultural drone operation methods still have many shortcomings in path planning and execution control, mainly reflected in the simplistic path planning methods, rigid pesticide application strategies, and low resource utilization, making it difficult to meet the development needs of precision agriculture and green plant protection.

[0003] Traditional agricultural drones often employ a geometric segmentation-based "bow-shaped" full-coverage scanning method for path planning. Before operation, operators manually set the farmland boundaries using ground station software, and the system automatically generates a reciprocating parallel path covering the entire area. The drone flies along the preset path, maintaining a fixed flight altitude (typically 2-3 meters), flight speed (4-6 m / s), and spray volume. While this method achieves full coverage, it assumes a uniform distribution of pests and diseases within the farmland and fails to consider the influence of actual terrain undulations, environmental wind fields, and crop growth. Therefore, this traditional method reveals the following prominent problems in practical applications:

[0004] (1) Insufficient precision in pesticide application: Due to the disregard for the spatial heterogeneity of pests and diseases, traditional methods apply pesticides insufficiently in areas with severe pests and diseases, while applying excessive amounts in areas with mild or no pests and diseases, resulting in uneven control effects and causing pesticide waste and environmental pollution.

[0005] (2) High energy consumption and low efficiency: The fixed "bow-shaped" path will generate a lot of redundant flight and idle in irregularly shaped farmland, and the climbing strategy is not optimized according to the terrain undulation, resulting in high battery energy consumption and limited area of ​​single operation.

[0006] (3) Fixed flight parameters: The flight altitude, speed and spray volume are fixed throughout the process, and cannot be adaptively adjusted according to the characteristics of the operation area, such as crop canopy height, hotspots of pests and diseases, and distribution of obstacles, which limits the further improvement of the application effect and operation safety.

[0007] (4) Lack of systematic optimization: Existing methods rely heavily on the experience of operators for path design and parameter setting, lack scientific decision support based on multi-source data fusion and intelligent optimization algorithms, and the operation results vary from person to person, with poor repeatability and reliability.

[0008] In recent years, some studies have attempted to introduce path optimization algorithms, such as genetic algorithms and ant colony algorithms, to improve the efficiency of UAV operations. However, these studies mostly focus on optimizing a single objective, such as the shortest path or the shortest time, without comprehensively considering the trade-offs between multiple objectives, such as prevention and control effectiveness, energy consumption, and pesticide utilization. In addition, existing methods often separate path geometry planning from flight parameter control, failing to achieve synergistic optimization between the two, making it difficult to achieve optimal overall performance in complex operating environments.

[0009] Therefore, there is an urgent need for an intelligent path planning method that can integrate multi-source information, collaboratively optimize paths and flight parameters, and take into account both prevention and control effects and operational energy consumption, so as to promote the development of agricultural drones towards precision, intelligence, and greenness. Summary of the Invention

[0010] The purpose of this invention is to provide an intelligent path planning method for plant protection drones based on multi-objective optimization, so as to achieve differentiated and precise application of pesticides to pests and diseases, and solve the problems of uneven application, high energy consumption and poor adaptability of traditional "bow-shaped" full-coverage paths.

[0011] Another objective of this invention is to provide an intelligent path planning system for agricultural drones based on multi-objective optimization.

[0012] To achieve the above objectives, the technical solution adopted by this invention is as follows:

[0013] A method for intelligent path planning of agricultural drones based on multi-objective optimization includes the following steps:

[0014] S1: Collect and process multi-source data to obtain processed data;

[0015] S2: Based on the processed data, construct an energy consumption-drug efficacy multi-objective optimization model, including an objective function for maximizing prevention and control effects and an objective function for minimizing energy consumption;

[0016] S3: The improved NSGA-II algorithm is used to solve the energy consumption-drug efficacy multi-objective optimization model, and Pareto optimal solution set is generated according to the constraints.

[0017] S4: Select several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters.

[0018] As a limitation, the processed data obtained in step S1 includes a distribution map of pest and disease levels, three-dimensional topography of farmland, UAV performance parameters, and real-time meteorological data.

[0019] The process of obtaining the pest and disease level distribution map is as follows: acquire multispectral images of farmland by aerial photography with a multispectral camera, input the multispectral images of farmland into a trained pest and disease detection model, output the pest and disease level, and obtain the pest and disease level distribution map based on the pest and disease level results.

[0020] The three-dimensional topography of farmland is constructed by obtaining a farmland digital elevation model from a geographic information system;

[0021] Real-time meteorological data is obtained by using wind speed and direction information acquired in real time by IoT meteorological sensors and then performing standardized processing.

[0022] As a further limitation, in step S1, the resolution of the aerial images taken by the multispectral camera is ≥0.1m / pixel, and the quantification range of the pest and disease level in the pest and disease level distribution map is [0, 1].

[0023] The accuracy of the digital elevation model for farmland is ≤0.5m, and the grid size is 1m×1m;

[0024] Drone performance parameters include battery capacity, canister capacity, speed range, and altitude range;

[0025] The battery capacity is 15000-30000mAh, the medicine box capacity is 10-30L, the speed range is 3-8m / s, and the height range is 1.5-5m.

[0026] The wind speed measurement range of the IoT weather sensor is 0-15m / s, with an accuracy of ±0.5m / s.

[0027] As a second limitation, in step S2, the energy consumption-drug efficacy multi-objective optimization model includes an objective function that maximizes the prevention and control effect and an objective function that minimizes energy consumption.

[0028] The objective function for maximizing the prevention and control effect is:

[0029] ;

[0030] in, This is the value used to evaluate the effectiveness of prevention and control. This refers to the set of all grids covered by the drone's flight path, i.e., the effective area for pesticide application; For the first Line number The pest and disease levels of the grid. ; For the first Line number The area of ​​the column grid; A model for drug deposition efficiency; The flight speed of the drone; This refers to the flight altitude of the drone. Wind speed;

[0031] Drug deposition efficiency model for:

[0032] ;

[0033] in, The baseline drug deposition efficiency; This is the speed influence coefficient; To achieve the optimal flight speed; The coefficient represents the high degree of influence. To achieve the optimal flight altitude; This refers to the wind speed influence coefficient.

[0034] The objective function for minimizing energy consumption is:

[0035] ;

[0036] in, This is an energy consumption assessment value; Energy consumption coefficient per unit flight distance; Energy consumption coefficient per unit turning angle; To increase the energy consumption coefficient for potential energy; Total flight distance; Number of turns; This represents the average turning angle. For the basic quality of drones; For the quality of the medicine solution; It is the acceleration due to gravity; This represents the cumulative altitude climbed.

[0037] As a third limitation, in step S3, the improved NSGA-II algorithm includes:

[0038] a. Latin hypercube sampling was used for population initialization, with a population size of 100-200 individuals;

[0039] b. Each individual is represented using chromosome coding as follows:

[0040] ;

[0041] in, For the first Coordinates of each waypoint; For the first Flight speed to each waypoint; For the first Flight altitude of each waypoint; Number of waypoints;

[0042] c. Set crossover probability Probability of mutation Simulated binary crossover and polynomial mutation operations are used to maintain population diversity;

[0043] d. Combine fast non-dominated sorting with crowding calculation to select an environment and verify the constraints;

[0044] e. Set the maximum number of iterations to 200-500 generations, and optimize the model's changes using a multi-objective approach focusing on energy consumption and drug efficacy. As a convergence threshold.

[0045] As a further limitation, in step b, the number of waypoints... The number of waypoints is adaptively determined based on farmland complexity. The calculation formula is:

[0046] .

[0047] As a further limitation, in step d, the constraints include battery capacity, medicine tank capacity, and flight boundary.

[0048] As the fourth type of limitation, the specific process of step S4 is as follows:

[0049] S41. Based on the Pareto optimal solution set, conduct an effect-energy trade-off analysis and select 3-5 representative optimal solutions from the Pareto optimal solution set;

[0050] S42. Based on the representative optimal solution, generate a three-dimensional flight path using three-dimensional path visualization technology. The three-dimensional flight path includes waypoint coordinates and corresponding flight parameters.

[0051] This invention also provides a multi-objective optimization-based intelligent path planning system for agricultural drones, used to implement the above-mentioned multi-objective optimization-based intelligent path planning method for agricultural drones, comprising:

[0052] The data acquisition module is used to collect and process multi-source data to obtain processed data. The processed data includes a distribution map of pest and disease levels, three-dimensional topography of farmland, performance parameters of UAVs, and real-time meteorological data.

[0053] The multi-objective optimization algorithm engine module is used to construct an energy consumption-drug efficacy multi-objective optimization model based on the processed data, including an objective function to maximize the prevention and control effect and an objective function to minimize energy consumption.

[0054] The algorithm solution module is used to solve the energy consumption-drug efficacy multi-objective optimization model using the improved NSGA-II algorithm, and generate a Pareto optimal solution set based on the constraints.

[0055] The path generation module is used to select several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters.

[0056] The present invention, by adopting the above-described technical solution, achieves the following technical advancements compared to existing technologies:

[0057] (1) This invention obtains a distribution map of pest and disease levels, three-dimensional topography of farmland, performance parameters of UAVs and real-time meteorological data by collecting multi-source data. Based on the processed data, an energy consumption-pesticide efficacy multi-objective optimization model is constructed, which transforms the original extensive mode of assuming uniform application of pesticides into a precise operation mode of variable application of pesticides based on the actual pest and disease levels. This solves the problem of insufficient application of pesticides in areas with severe pest and disease and excessive application of pesticides in areas with mild pest and disease, which significantly improves the accuracy of targeted pesticide application and greatly reduces the total amount of pesticides used while improving the control effect. By establishing an energy consumption-pesticide efficacy multi-objective optimization model, the dual optimization of minimizing energy consumption under the premise of ensuring the control effect is achieved.

[0058] (2) The present invention uses the improved NSGA-II algorithm to solve the energy consumption-drug efficacy multi-objective optimization model. According to the constraints, the Pareto optimal solution set is generated. The improved NSGA-II algorithm adopts chromosome coding to synchronously encode waypoint coordinates, flight speed and altitude as optimization variables, realizing the coordinated optimization of path geometry and flight control parameters. This enables the UAV to dynamically adjust its flight attitude according to terrain undulations, wind field environment and distribution of pests and diseases, further improving the uniformity, safety and environmental adaptability of pesticide application.

[0059] (3) The present invention selects several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters. By selecting several optimal solutions from the Pareto optimal solution set, it provides users with a variety of optional optimization schemes, thereby enhancing the system's flexibility and decision support capabilities in practical applications.

[0060] In summary, this invention is applicable to planning pest control routes for agricultural drones, taking into full account factors such as speed, altitude, and wind speed, thereby achieving automated and intelligent pest control operations and reducing operating costs. Attached Figure Description

[0061] Figure 1 The diagram shown is a flowchart of the intelligent path planning method for agricultural drones based on multi-objective optimization according to Embodiment 1 of the present invention.

[0062] Figure 2 The diagram shown is a flowchart of the improved NSGA-II algorithm in Embodiment 1 of the present invention;

[0063] Figure 3 The image shown is a heatmap of the optimized path in Embodiment 1 of the present invention;

[0064] Figure 4 The image shown is a heatmap of the conventional path in Embodiment 1 of the present invention;

[0065] Figure 5 The figure shown is the Pareto front plot when the iteration number is 0 in Embodiment 1 of the present invention;

[0066] Figure 6 The image shown is a heatmap of the optimized path when the iteration number is 0 in Embodiment 1 of the present invention.

[0067] Figure 7 The figure shown is a distribution diagram of flight parameters when the iteration number is 0 in Embodiment 1 of the present invention;

[0068] Figure 8 The figure shown is the Pareto front plot when the iteration number is 12 in Embodiment 1 of the present invention;

[0069] Figure 9 The image shown is a heatmap of the optimized path when the iteration number is 12 in Embodiment 1 of the present invention.

[0070] Figure 10 The figure shows the convergence curve of the best prevention and control effect when the iteration number is 12 in Embodiment 1 of the present invention;

[0071] Figure 11 The figure shown is a distribution diagram of flight parameters when the iteration number is 12 in Embodiment 1 of the present invention;

[0072] Figure 12 The figure shown is the Pareto front plot when the iteration number is 26 in Embodiment 1 of the present invention;

[0073] Figure 13 The image shown is the optimized path heatmap for Embodiment 1 of the present invention when the iteration number is 26.

[0074] Figure 14 The figure shows the convergence curve of the best prevention and control effect when the iteration number is 26 in Embodiment 1 of the present invention;

[0075] Figure 15 The figure shown is a distribution diagram of flight parameters when the iteration number is 26 in Embodiment 1 of the present invention;

[0076] Figure 16 The figure shown is the Pareto front plot when the iteration number is 41 in Embodiment 1 of the present invention;

[0077] Figure 17 The image shown is the optimized path heatmap for Embodiment 1 of the present invention when the iteration number is 41.

[0078] Figure 18 The figure shows the convergence curve of the best prevention and control effect when the iteration number is 41 in Embodiment 1 of the present invention;

[0079] Figure 19 The figure shown is a distribution diagram of flight parameters when the iteration number is 41 in Embodiment 1 of the present invention;

[0080] Figure 20 The figure shown is the Pareto front plot of the last generation in Embodiment 1 of the present invention;

[0081] Figure 21 The diagram shown is a structural block diagram of the intelligent path planning system for agricultural drones based on multi-objective optimization, according to Embodiment 2 of the present invention. Detailed Implementation

[0082] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0083] Example 1

[0084] like Figure 1 As shown, this embodiment presents an intelligent path planning method for agricultural drones based on multi-objective optimization, including the following steps:

[0085] S1: Collect and process multi-source data to obtain processed data;

[0086] In this step, the processed data obtained includes a distribution map of pest and disease levels, three-dimensional topography of farmland, UAV performance parameters, and real-time meteorological data.

[0087] The process of obtaining the pest and disease level distribution map is as follows: multispectral images of farmland are acquired by aerial photography using a multispectral camera. The multispectral images of farmland are input into a pest and disease detection model trained by YOLOv5 or Faster R-CNN model, which outputs the pest and disease level. Based on the pest and disease level results, the pest and disease level distribution map is obtained. The resolution of the multispectral aerial images is ≥0.1m / pixel, and the quantization range of the pest and disease level in the pest and disease level distribution map is [0, 1].

[0088] The three-dimensional topography of farmland is constructed by obtaining a farmland digital elevation model from a geographic information system; the accuracy of the farmland digital elevation model is ≤0.5m, and the grid size is 1m×1m.

[0089] The performance parameters of the drone include battery capacity, medical kit capacity, speed range, and altitude range; the battery capacity is 15000-30000mAh, the medical kit capacity is 10-30L, the speed range is 3-8m / s, and the altitude range is 1.5-5m.

[0090] Real-time meteorological data is obtained by using wind speed and direction information acquired in real time by IoT meteorological sensors and then performing standardized processing; the wind speed measurement range of the IoT meteorological sensors is 0-15m / s, with an accuracy of ±0.5m / s.

[0091] S2: Based on the processed data, construct an energy consumption-drug efficacy multi-objective optimization model, including an objective function for maximizing prevention and control effects and an objective function for minimizing energy consumption;

[0092] In this step, the energy consumption-drug efficacy multi-objective optimization model includes an objective function that maximizes the prevention and control effect and an objective function that minimizes energy consumption.

[0093] The objective function for maximizing the prevention and control effect is:

[0094] ;

[0095] in, This is the value used to evaluate the effectiveness of prevention and control. This refers to the set of all grids covered by the drone's flight path, i.e., the effective area for pesticide application; For the first Line number The pest and disease levels of the grid. , A higher value indicates a more severe pest or disease outbreak; For the first Line number The area of ​​the column grid; This is a model for drug deposition efficiency, where the drone's flight speed is the factor. Drone flight altitude Wind speed The function.

[0096] Drug deposition efficiency model for:

[0097] ;

[0098] in, The baseline deposition efficiency is typically the ideal deposition efficiency under conditions of no wind, optimal flight speed, and optimal flight altitude. The velocity influence coefficient reflects the degree to which the flight speed deviates from the optimal flight speed on the drug deposition efficiency; The optimal flight speed is the speed at which the drug deposition efficiency is highest under windless conditions and at a fixed flight altitude. The altitude influence coefficient reflects the degree to which the flight altitude deviates from the optimal flight altitude on the efficiency of drug deposition. The optimal flight altitude is the altitude at which the drug deposition efficiency is highest under windless conditions and a fixed flight speed. The wind speed influence coefficient reflects the degree to which wind speed weakens the efficiency of drug deposition.

[0099] The objective function for minimizing energy consumption is:

[0100] ;

[0101] in, This is an energy consumption assessment value; The energy consumption coefficient per unit flight distance represents the energy consumed by the drone for every meter it flies. The energy consumption coefficient per unit turning angle represents the energy consumed by the drone for each unit radian turn. The potential energy increase energy consumption coefficient represents the energy conversion efficiency coefficient required to increase potential energy by one unit. The total flight distance is in meters (m). Number of turns; The average turning angle is expressed in rad. The basic mass of the drone is in kg; The mass of the medicine solution is expressed in kg. It is the acceleration due to gravity; The cumulative elevation gain is expressed in meters (m).

[0102] In this step, the objective function for maximizing the prevention and control effect is... With the energy consumption minimization objective function Together, these constitute a multi-objective optimization problem that requires trade-offs to be solved under multiple constraints. The core function of the improved NSGA-II algorithm in step S3 is to use these two functions as parallel optimization objectives to perform a global search and iterative evolution in the solution space composed of variables such as waypoint coordinates, UAV flight speed, and UAV flight altitude, thereby finding a series of non-dominated solutions that achieve the best balance between prevention and control effects and energy consumption, i.e., the Pareto optimal solution set.

[0103] S3: The improved NSGA-II algorithm is used to solve the energy consumption-drug efficacy multi-objective optimization model, and Pareto optimal solution set is generated according to the constraints.

[0104] like Figure 2 As shown, the improved NSGA-II algorithm includes:

[0105] a. Latin hypercube sampling was used for population initialization, with a population size of 100-200 individuals;

[0106] b. Each individual is represented using chromosome coding as follows:

[0107] ;

[0108] in, For the first Coordinates of each waypoint; For the first Flight speed to each waypoint; For the first Flight altitude of each waypoint; Number of waypoints The number of waypoints is adaptively determined based on farmland complexity. The calculation formula is:

[0109] ;

[0110] ;

[0111] in, This represents the average level of pest and disease severity across all grid cells in the farmland. The standard deviation of pest and disease severity levels across all grids in the farmland;

[0112] c. Set crossover probability Probability of mutation Simulated binary crossover and polynomial mutation operations are used to maintain population diversity;

[0113] d. Combine fast non-dominated sorting with congestion calculation to select an environment and verify the constraints, including battery capacity, tank capacity and flight boundary.

[0114] e. Set the maximum number of iterations to 200-500 generations, and optimize the model's changes using a multi-objective approach focusing on energy consumption and drug efficacy. As a convergence threshold.

[0115] S4: Select several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters.

[0116] The specific process for this step is as follows:

[0117] S41. Based on the Pareto optimal solution set, conduct an effect-energy trade-off analysis and select 3-5 representative optimal solutions from the Pareto optimal solution set;

[0118] S42. Based on the representative optimal solution, a three-dimensional flight path is generated using three-dimensional path visualization technology. The three-dimensional flight path includes waypoint coordinates and corresponding flight parameters. Users can make the final selection according to actual operational needs (such as prioritizing performance or energy saving).

[0119] S5: Perform feasibility verification and performance evaluation of three-dimensional flight paths in a digital twin simulation environment.

[0120] To verify the effectiveness of this embodiment, the feasibility of the three-dimensional flight path was verified and its performance evaluated in a digital twin simulation environment. First, the generated three-dimensional flight path underwent a rigorous feasibility check to ensure that the path planning did not cross obstacles and met the minimum turning radius requirements of the UAV. Then, a comprehensive performance index evaluation was carried out, and key indicators such as prevention and control effect, energy consumption, and operation time were accurately calculated. Finally, through comparative analysis with traditional methods, the performance advantages of the method in this embodiment in terms of operation efficiency and resource utilization were quantitatively demonstrated, providing a reliable basis for the practical application of the solution.

[0121] Taking a 100m×100m rectangular farmland as an example, aerial photography was conducted using a MicaSense RedEdge-MX multispectral camera to generate a 0.1m resolution multispectral image. A YOLOv5 model was constructed, and the acquired farmland multispectral image dataset was used to train the YOLOv5 model to obtain a pest and disease detection model based on the YOLOv5 model. The farmland multispectral image to be detected was input into the trained pest and disease detection model, and the pest and disease level was output. Based on the pest and disease level results, a pest and disease level distribution map was obtained.

[0122] The distribution of pests and diseases is as follows: 20% in severely affected areas (pest and disease severity levels [0.7, 1.0]), 30% in moderately affected areas (pest and disease severity levels [0.3, 0.7]), and 50% in lightly affected areas (pest and disease severity levels [0, 0.3]). The agricultural drone used is a DJI MG-1P, and the basic weight of the drone is [not specified]. The weight is 25 kg. The severity of pests and diseases is determined by Table 1.

[0123] Table 1. Disease and pest severity levels

[0124]

[0125] The parameters for the agricultural drone are set as follows: battery capacity of 20000mAh and tank capacity of 30L. The model coefficients and constraint parameters are as follows: It is 0.85; It is 0.1; It is 4.0 m / s; It is 0.15; It is 2.5m; It is 0.08; It is 15 J / m; It is 50 J / rad; The value is 0.12; according to IoT weather sensors, the wind speed... It is 2 m / s.

[0126] Among them, the baseline drug deposition efficiency Speed ​​influence coefficient High influence coefficient Wind speed influence coefficient Determined based on Table 2. Energy consumption coefficient per unit flight distance. Energy consumption coefficient per unit turning angle Potential energy increases energy consumption coefficient The determination is based on Table 3.

[0127] Table 2. Values ​​and determination basis of each coefficient in the drug deposition efficiency model.

[0128]

[0129] Table 3. Values ​​of the coefficients in the energy consumption minimization objective function and the basis for their determination.

[0130]

[0131] Based on the above preconditions, an improved NSGA-II algorithm is used for optimization, and a representative optimal solution is selected from the generated Pareto optimal solution set. The key flight and path parameters corresponding to this solution are output as follows: UAV flight speed The speed is 4.5 m / s, and the drone's flight altitude is... It is 2.8m, and the total flight distance is... The length is 1200m, and the number of turns is... The cumulative elevation gain is 15. The value is 30m. Substituting the parameters of this optimal solution into the drug deposition efficiency model, we obtain:

[0132] Drug deposition efficiency model for:

[0133]

[0134] Based on the objective function of maximizing the prevention and control effect, the evaluation value of the prevention and control effect is calculated as follows:

[0135]

[0136] Based on the energy consumption minimization objective function, the energy consumption assessment value is calculated as follows:

[0137]

[0138] After obtaining this set of flight parameters and path, a digital twin simulation was run on the NVIDIA Jetson Xavier NX edge computing device to verify the feasibility and evaluate the performance of the three-dimensional flight path.

[0139] I. Improved prevention and control effectiveness

[0140] The generated optimal 3D flight path achieved a prevention and control effect of 84.26% in the experimental scenario, with an average of 70.30% and a standard deviation of 8.10%, indicating that the method in this embodiment effectively improves the coverage rate of pest and disease control while ensuring operational stability.

[0141] According to the pest and disease heat map statistics, the coverage rate of hotspot areas has increased to 44.41%, which is about 21% higher than the traditional "bow-shaped" full coverage path. Through methods such as... Figure 3 and Figure 4 The comparison chart shows that the optimized path is more concentrated in areas with high incidence of pests and diseases, achieving precise matching of pesticide application.

[0142] II. Reduced energy consumption and increased resource utilization

[0143] The optimized path length is only 295.88 units, corresponding to a total energy consumption of 3112.70 J, which is about 30%–40% lower than the traditional fixed scanning path (typical value 4000–5000 J). The pesticide usage is 24.85 L, close to the set upper limit of 25 L, but the coverage efficiency is improved, and the control output per unit of pesticide is increased by 33%. Under the same energy and pesticide constraints, this embodiment achieves higher application efficiency and lower operational redundancy.

[0144] Experimental results show that the optimal flight parameters automatically searched by the system are: average speed 4.09 m / s (optimal value 4.0 m / s), average altitude 2.43 m (optimal value 2.5 m), and drug deposition efficiency model η = 0.90.

[0145] The parameter configuration in this validation example demonstrated excellent energy consumption-powder efficacy balance in simulation tests. Compared with the traditional setting of a fixed height of 3m and a speed of 5m / s for a "bow-shaped" full-coverage path, the uniformity of powder deposition was improved by approximately 18%, and the target area hit rate was increased by 22%. The standard deviations of parameter fluctuations were 1.02m / s (speed) and 0.72m (height), respectively, verifying the dynamic stability of the algorithm under complex terrain and wind field conditions.

[0146] III. Enhanced Path Intelligence and Spatial Adaptability

[0147] like Figures 5 to 19The diagram shows the Pareto front, path heatmap, convergence curve of optimal control effect, and flight parameter distribution for iteration numbers of 0, 12, 26, and 41. The Pareto front is crucial for the solution set trade-offs in multi-objective optimization, demonstrating the Pareto optimal solution set between "control effect" and "operation time" for the UAV operation scheme during the nth generation optimization process. Each point in the diagram represents a scheme, and these schemes cannot simultaneously excel in both improving control effect and reducing time, reflecting the trade-off between the two objectives. The path heatmap focuses on the optimal flight trajectory of the UAV operation. The horizontal and vertical axes are coordinates representing the spatial location of the farmland, and the gradient background color corresponds to environmental information such as pest and disease density and crop distribution. It shows the optimal operation path of the UAV in the farmland obtained from the nth generation optimization. Circular points are typically the starting point, and square points are the ending point, visually presenting the UAV's operation route planning results. The optimal prevention and control effect convergence curve is used to represent the changing trend of prevention and control effect during the optimization process; it reflects the change process of the "optimal prevention and control effect" corresponding to each generation as the number of optimization iterations increases, and can reflect the convergence of the optimization algorithm. The flight parameter distribution reflects the combination and optimal selection of UAV operation parameters; it shows the parameter combination distribution of flight speed and flight altitude during UAV operation. The star represents the optimal parameter combination in this set of parameters, which is the flight speed and flight altitude that can match the optimal operation effect.

[0148] Depend on Figures 5 to 19 It can be seen that the optimized UAV flight trajectory exhibits highly adaptive characteristics: the path automatically avoids low-risk areas and areas with large terrain undulations, prioritizing coverage of areas with severe pest and disease distribution. Through digital twin simulation comparison, the traditional path spends 18.5% of its time in severely affected areas, while the optimized path of this invention reaches 42.3%, improving pesticide utilization efficiency and targeted application capability.

[0149] IV. The algorithm exhibits superior convergence performance and a balanced distribution of the solution set.

[0150] The results show that the improved NSGA-II algorithm, after 120 generations, converged in approximately 2611.92 seconds, obtaining 103 Pareto optimal solutions, and rapidly reached the ideal frontier after the 20th generation. Figure 20 As shown, the non-dominated solution set is uniformly distributed on the effect-energy consumption two-dimensional plane, indicating that the improved NSGA-II algorithm has strong global convergence ability while maintaining diversity. This characteristic allows users to flexibly choose schemes that emphasize prevention and control effects or energy saving according to operational requirements.

[0151] Compared with the "bow-shaped" full-coverage path planning method, the overall performance improvement of this embodiment is shown in the table below.

[0152] Table 4 Comparison of Overall Performance Improvements

[0153]

[0154] The data for the traditional method in Table 4 comes from a fixed-parameter "bow-shaped" full-coverage path benchmark model under the same digital twin simulation environment, and is representative and repeatable.

[0155] The experimental data and simulation results confirm that the intelligent path planning method for agricultural drones based on multi-objective optimization in this embodiment achieves a dual balance between control effectiveness and energy consumption under the same resource constraints. Through an intelligent algorithm-driven path and parameter adaptive control mechanism, the drone can perform differentiated operations based on the actual distribution of pests and diseases, thus achieving breakthroughs in precise pest and disease control, improved energy utilization, pesticide conservation, and operational efficiency. This provides a feasible and scalable technical solution for intelligent agricultural plant protection.

[0156] Example 2

[0157] This embodiment describes an intelligent path planning system for agricultural drones based on multi-objective optimization, such as... Figure 21 The diagram shown is a block diagram of an intelligent path planning system for agricultural drones based on multi-objective optimization, provided according to the method of Embodiment 1. This system includes:

[0158] The data acquisition module is used to collect and process multi-source data to obtain processed data. The processed data includes a distribution map of pest and disease levels, three-dimensional topography of farmland, performance parameters of UAVs, and real-time meteorological data.

[0159] The multi-objective optimization algorithm engine module is used to construct an energy consumption-drug efficacy multi-objective optimization model based on the processed data, including an objective function to maximize the prevention and control effect and an objective function to minimize energy consumption.

[0160] The algorithm solution module is used to solve the energy consumption-drug efficacy multi-objective optimization model using the improved NSGA-II algorithm, and generate a Pareto optimal solution set based on the constraints.

[0161] The path generation module is used to select several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters.

[0162] In this embodiment, the intelligent path planning system for agricultural drones based on multi-objective optimization is illustrated by the above-mentioned functional modules when processing data. In actual applications, the above functions can be assigned to different functional modules as needed.

[0163] It should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still modify the technical solutions described in the above embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent path planning of agricultural drones based on multi-objective optimization, characterized in that, Includes the following steps: S1: Collect and process multi-source data to obtain processed data; S2: Based on the processed data, construct an energy consumption-drug efficacy multi-objective optimization model, including an objective function for maximizing prevention and control effects and an objective function for minimizing energy consumption; S3: The improved NSGA-II algorithm is used to solve the energy consumption-drug efficacy multi-objective optimization model, and Pareto optimal solution set is generated according to the constraints. S4: Select several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters; In step S2, the objective function for maximizing the prevention and control effect is: ； in, This is the value used to evaluate the effectiveness of prevention and control. This refers to the set of all grids covered by the drone's flight path, i.e., the effective area for pesticide application; For the first Line number The pest and disease levels of the grid. ; For the first Line number The area of ​​the column grid; A model for drug deposition efficiency; The flight speed of the drone; This refers to the flight altitude of the drone. Wind speed; Drug deposition efficiency model for: ； in, The baseline drug deposition efficiency; This is the speed influence coefficient; To achieve the optimal flight speed; The coefficient represents the high degree of influence. To achieve the optimal flight altitude; This refers to the wind speed influence coefficient. The objective function for minimizing energy consumption is: ； in, This is an energy consumption assessment value; Energy consumption coefficient per unit flight distance; Energy consumption coefficient per unit turning angle; To increase the energy consumption coefficient for potential energy; Total flight distance; Number of turns; This represents the average turning angle. For the basic quality of drones; For the quality of the medicine solution; It is the acceleration due to gravity; This represents the cumulative altitude climbed.

2. The intelligent path planning method for agricultural drones based on multi-objective optimization according to claim 1, characterized in that, In step S1, the processed data obtained includes a distribution map of pest and disease levels, three-dimensional topography of farmland, UAV performance parameters, and real-time meteorological data. The process of obtaining the pest and disease level distribution map is as follows: acquire multispectral images of farmland by aerial photography with a multispectral camera, input the multispectral images of farmland into a trained pest and disease detection model, output the pest and disease level, and obtain the pest and disease level distribution map based on the pest and disease level results. The three-dimensional topography of farmland is constructed by obtaining a farmland digital elevation model from a geographic information system; Real-time meteorological data is obtained by using wind speed and direction information acquired in real time by IoT meteorological sensors and then performing standardized processing.

3. The intelligent path planning method for agricultural drones based on multi-objective optimization according to claim 2, characterized in that, In step S1, the resolution of the aerial images taken by the multispectral camera is ≥0.1m / pixel, and the quantification range of the pest and disease level in the pest and disease level distribution map is [0, 1]. The accuracy of the digital elevation model for farmland is ≤0.5m, and the grid size is 1m×1m; Drone performance parameters include battery capacity, canister capacity, speed range, and altitude range; The battery capacity is 15000-30000mAh, the medicine box capacity is 10-30L, the speed range is 3-8m / s, and the height range is 1.5-5m. The wind speed measurement range of the IoT weather sensor is 0-15m / s, with an accuracy of ±0.5m / s.

4. The intelligent path planning method for agricultural drones based on multi-objective optimization according to claim 1, characterized in that, In step S3, the improved NSGA-II algorithm includes: a. Latin hypercube sampling was used for population initialization, with a population size of 100-200 individuals; b. Each individual is represented using chromosome coding as follows: ； in, For the first Coordinates of each waypoint; For the first Flight speed to each waypoint; For the first Flight altitude of each waypoint; Number of waypoints; c. Set crossover probability Probability of mutation Simulated binary crossover and polynomial mutation operations are used to maintain population diversity; d. Combine fast non-dominated sorting with crowding calculation to select an environment and verify the constraints; e. Set the maximum number of iterations to 200-500 generations, and optimize the model's changes using a multi-objective approach focusing on energy consumption and drug efficacy. As a convergence threshold.

5. The intelligent path planning method for agricultural drones based on multi-objective optimization according to claim 4, characterized in that, In step b, the number of waypoints The number of waypoints is adaptively determined based on farmland complexity. The calculation formula is: 。 6. The intelligent path planning method for agricultural drones based on multi-objective optimization according to claim 5, characterized in that, In step d, the constraints include battery capacity, medicine tank capacity, and flight boundaries.

7. The intelligent path planning method for agricultural drones based on multi-objective optimization according to claim 1, characterized in that, The specific process of step S4 is as follows: S41. Based on the Pareto optimal solution set, conduct an effect-energy trade-off analysis and select 3-5 representative optimal solutions from the Pareto optimal solution set; S42. Based on the representative optimal solution, generate a three-dimensional flight path using three-dimensional path visualization technology. The three-dimensional flight path includes waypoint coordinates and corresponding flight parameters.

8. A multi-objective optimization-based intelligent path planning system for agricultural drones, used to implement the multi-objective optimization-based intelligent path planning method for agricultural drones as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect and process data from multiple sources to obtain processed data. The processed data obtained includes a distribution map of pest and disease levels, three-dimensional topography of farmland, UAV performance parameters, and real-time meteorological data; The multi-objective optimization algorithm engine module is used to construct an energy consumption-drug efficacy multi-objective optimization model based on the processed data, including an objective function to maximize the prevention and control effect and an objective function to minimize energy consumption. The objective function for maximizing the prevention and control effect is: ； in, This is the value used to evaluate the effectiveness of prevention and control. This refers to the set of all grids covered by the drone's flight path, i.e., the effective area for pesticide application; For the first Line number The pest and disease levels of the grid. ; For the first Line number The area of ​​the column grid; A model for drug deposition efficiency; The flight speed of the drone; This refers to the flight altitude of the drone. Wind speed; Drug deposition efficiency model for: ； in, The baseline drug deposition efficiency; This is the speed influence coefficient; To achieve the optimal flight speed; The coefficient represents the high degree of influence. To achieve the optimal flight altitude; This refers to the wind speed influence coefficient. The objective function for minimizing energy consumption is: ； in, This is an energy consumption assessment value; Energy consumption coefficient per unit flight distance; Energy consumption coefficient per unit turning angle; To increase the energy consumption coefficient for potential energy; Total flight distance; Number of turns; This represents the average turning angle. For the basic quality of drones; For the quality of the medicine solution; It is the acceleration due to gravity; This represents the cumulative altitude climbed. The algorithm solution module is used to solve the energy consumption-drug efficacy multi-objective optimization model using the improved NSGA-II algorithm, and generate a Pareto optimal solution set based on the constraints. The path generation module is used to select several optimal solutions from the Pareto optimal solution set to obtain the three-dimensional flight path and corresponding flight parameters.

Images