A highway slope plant pesticide spraying path planning method and system based on a unmanned aerial vehicle

Abstract

This invention discloses a method and system for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs), relating to the interdisciplinary fields of smart agriculture and UAV autonomous navigation. During operation, the system acquires multi-dimensional spatial geographic information of the slope and constructs a variable-scale three-dimensional digital terrain model; it establishes a dynamic wind field disturbance prediction model by combining meteorological and traffic flow data; it constructs a variable-mass dynamics model of the UAV considering real-time pesticide load loss; and based on the above model, it generates and dynamically corrects a comprehensive spraying path through a multi-objective collaborative optimization algorithm. The system includes a spatial data perception module, a terrain digital modeling module, an environmental wind field analysis module, a variable-mass dynamics solution module, and a path planning and control core module. By adopting the above technical solution, this application can achieve a synergistic improvement in spraying height stability, uniform coverage, flight safety and reliability, and optimal energy efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent agriculture and unmanned aerial vehicle (UAV) autonomous navigation technology, specifically a method and system for planning the path of spraying pesticides on highway slopes based on unmanned aerial vehicles (UAVs). Background Technology

[0002] As a core component of modern logistics and travel infrastructure, highway transportation relies heavily on slope ecological protection and vegetation maintenance to maintain roadbed stability and improve the environmental quality of road sections. With the continuous advancement of aerial remote sensing and automated control technologies, the use of drones for large-scale, high-precision plant protection operations has become an important development direction in the field of transportation infrastructure maintenance.

[0003] Among them, planning the path for drone-based plant spraying in the diverse terrain of highway slopes is a core technical aspect of achieving intelligent and precise slope maintenance. Its basic goal is to optimize the operation trajectory by establishing a high-dimensional flight model, thereby achieving uniform coverage of the pesticide and improving operational efficiency while ensuring flight safety.

[0004] Existing highway slope path planning technologies still face significant challenges in practical applications. First, existing planning models often simplify the three-dimensional terrain features of slopes, frequently using projection planes or contour lines for simplification. This fails to accurately fit steep slopes and irregular cross-sections, leading to frequent and drastic fluctuations in spraying height and inconsistent coverage during drone operations. Second, existing path algorithms lack robust design to handle complex turbulence and lateral wind disturbances along highways. When affected by transient airflow from passing vehicles, flight paths are easily deviated, potentially causing collisions between drones and slope obstacles or vehicles on the road. Third, existing task planning strategies are relatively rigid, failing to effectively integrate the changes in drone dynamics caused by the continuous reduction of pesticide load. This results in difficulty maintaining optimal energy efficiency during operations, easily leading to insufficient pesticide utilization and wasted endurance resources. To address the shortcomings of traditional technologies, such as low terrain fitting accuracy, weak environmental adaptability, and poor operational energy efficiency, the drone-based highway slope vegetation spraying path planning method and system proposed in this invention is particularly important. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs), which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for planning a spraying path for vegetation on a highway slope based on an unmanned aerial vehicle (UAV), comprising the following steps: S1, acquiring multi-dimensional spatial geographic information of the highway slope through multi-source remote sensing equipment, and performing feature extraction and coordinate alignment processing on the acquired multi-dimensional spatial geographic information; S2, constructing a variable-scale three-dimensional digital terrain model of the highway slope based on the processed multi-dimensional spatial geographic information, used to characterize the surface geometric contour and slope aspect variation characteristics of the slope; S3, constructing a dynamic wind field disturbance prediction model under the highway slope environment by combining meteorological monitoring data and traffic flow dynamic parameters along the highway; S4, establishing a variable-mass dynamics model of the UAV considering the real-time loss of pesticide load, and using this to determine the flight attitude constraints and power consumption characteristics of the UAV at different spraying stages; S5, generating a comprehensive spraying path based on the three-dimensional digital terrain model, the dynamic wind field disturbance prediction model, and the variable-mass dynamics model of the UAV through a multi-objective collaborative optimization algorithm, and dynamically correcting the generated path based on real-time feedback data.

[0007] Preferably, step S1 specifically includes the following steps: S11, using an airborne LiDAR scanner to perform high-frequency scanning of the highway slope to obtain raw point cloud data; S12, using a multispectral imaging device to collect image data of the slope vegetation and obtain information on vegetation distribution density and species; S13, using a differential global navigation satellite system to obtain high-precision positioning and attitude information of the sensor at the time of acquisition; S14, fusing the raw point cloud data, image data, and positioning attitude information, removing environmental noise interference, and generating a standardized slope spatial dataset under a unified coordinate system.

[0008] Preferably, step S2 specifically includes the following steps: S21, performing non-uniform sampling on the standardized slope spatial dataset, and automatically adjusting the sampling step size according to the curvature variation of the slope surface; S22, using a non-uniform rational spline interpolation algorithm to reconstruct the surface of the sampling points and generate a continuous slope geometric description function; S23, dividing the reconstructed surface into regional grids according to the vegetation distribution density, and setting different grid fineness for regions with different densities.

[0009] Preferably, step S3 specifically includes the following steps: S31, acquiring real-time meteorological parameters of the highway section, including reference wind direction, reference wind speed, and atmospheric pressure; S32, collecting real-time highway traffic flow information and average vehicle speed, and calculating the intensity of instantaneous induced airflow caused by vehicle movement; S33, superimposing and coupling the reference wind field and instantaneous induced airflow through computational fluid dynamics logic to calculate the local wind vector distribution in the near-surface area of ​​the slope; S34, establishing a wind field uncertainty assessment index to quantify the potential impact of local wind vector distribution on the flight stability of the UAV.

[0010] Preferably, step S4 specifically includes the following steps: S41, real-time monitoring of the liquid level or weight change in the agent tank to obtain the real-time value of the remaining agent quantity; S42, based on the principle of mass conservation and the spray flow rate setting, establishing an evolutionary relationship equation between the total mass of the UAV and time or travel distance; S43, calculating the inertial torque generated by the sloshing of the liquid agent and adding it as a disturbance term to the attitude control equation of the UAV; S44, adjusting the thrust distribution scheme of the UAV motors according to the real-time mass change to determine the optimal cruise speed and maximum tilt angle limit under the current mass.

[0011] Preferably, step S5 specifically includes the following steps: S51, setting a cost function for path planning, the cost function including a coverage cost term, an energy consumption cost term, and a safety risk cost term; S52, based on the slope function of the three-dimensional digital terrain model, calculating the normal distance constraint between the UAV spraying device and the slope to ensure that the spraying height is maintained within a preset range; S53, using a multi-objective heuristic search algorithm to search for the optimal path point sequence on the gridded slope model, the search process considering the wind direction compensation provided by the dynamic wind field disturbance prediction model; S54, smoothing the optimal path point sequence to generate an executable flight trajectory that satisfies the kinematic constraints of the UAV.

[0012] Preferably, the non-uniform sampling process includes: calculating the angle between the normal vectors of adjacent data points in the standardized slope spatial dataset; when the angle between the normal vectors is greater than a preset angle threshold, it is determined to be a region of drastic terrain change, and the sampling point distribution density is increased; when the angle between the normal vectors is less than or equal to the preset angle threshold, it is determined to be a gentle region, and the sampling point distribution density is reduced to reduce the computational burden of subsequent modeling.

[0013] Preferably, the surface reconstruction process of the slope also includes the identification and removal of obstacles on the slope. Specifically, deep learning object detection logic is used to identify utility poles, guardrails, and isolated rocks in the point cloud data, and these areas are marked as geometric boundaries that are prohibited from being traversed during surface reconstruction.

[0014] Preferably, in the dynamic wind field disturbance prediction model, the calculation of the instantaneous induced airflow intensity takes into account the vehicle's cross-sectional shape coefficient and the horizontal offset of the slope along the travel distance. When a large truck passes by, the wind field disturbance weight for the corresponding time period is automatically increased, and a pre-command is triggered for the UAV to increase lateral compensation thrust.

[0015] Preferably, in the UAV variable mass dynamics model, the monitoring of the remaining amount of the agent is combined with acceleration compensation provided by the inertial measurement unit to eliminate the influence of false readings on the sensors caused by liquid sloshing during the acceleration or deceleration of the UAV.

[0016] Preferably, during the generation of the coverage spray path, the atomization pressure of the nozzle is automatically adjusted according to the slope's inclination angle. When the slope increases, the spray flow rate towards the uphill direction is increased to counteract the downward offset of the pesticide droplets due to gravity, ensuring uniform deposition of the pesticide on the slope surface.

[0017] Preferably, the safety risk cost term in the cost function is positively correlated with the wind field uncertainty assessment index. In areas with severe wind field fluctuations, the path planning algorithm automatically guides the UAV away from steep edges or narrow areas with dense obstacles, prioritizing flight paths with relatively stable airflow.

[0018] A drone-based system for planning spraying paths for vegetation on highway slopes includes: a spatial data sensing module for acquiring and processing multi-dimensional spatial geographic information of the slope through multi-source remote sensing equipment; a terrain digitization modeling module connected to the spatial data sensing module, configured to construct a variable-scale three-dimensional digital terrain model; an environmental wind field analysis module configured to output dynamic wind field disturbance prediction results based on meteorological and traffic flow data; a variable mass dynamics calculation module for real-time monitoring of pesticide consumption and updating the dynamic state parameters of the drone; and a path planning and control core module connected to the terrain digitization modeling module, the environmental wind field analysis module, and the variable mass dynamics calculation module, respectively, for generating and correcting the spraying path according to a multi-objective collaborative optimization algorithm.

[0019] The spatial data perception module integrates a point cloud filtering algorithm to automatically identify and remove non-vegetation interference signals mentioned in the background technology, thereby improving the purity of feature extraction.

[0020] The terrain digital modeling module has an adaptive subdivision function, which can dynamically generate multi-resolution grid structures based on the local curvature of the slope, achieving an optimal balance between computational efficiency and terrain fitting accuracy.

[0021] The environmental wind field analysis module is connected to a wireless sensor network deployed along the highway to receive real-time data streams from ground weather stations and traffic monitoring systems.

[0022] The variable mass dynamics solution module has a built-in adaptive filter to smooth the instantaneous shift signal of the center of gravity caused by fluctuations in the liquid surface of the reagent.

[0023] The path planning and control core module includes a trajectory smoother, which converts discrete path points into continuous control commands through high-order polynomial fitting or spline curve processing.

[0024] This invention provides a method and system for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs), which has the following beneficial effects: (1) During system operation, this invention integrates multi-source remote sensing information, dynamic wind field simulation, and variable mass dynamics modeling to construct an intelligent path planning system for the special environment of highway slopes. First, the introduction of a variable-scale three-dimensional digital terrain model completely changes the traditional simplified projection plane model. Through non-uniform sampling and surface reconstruction technology, this invention can accurately capture the subtle features of steep slopes and irregular cross-sections. This effectively solves the problem of frequent fluctuations in spraying height caused by insufficient terrain fitting during UAV operation, ensuring that the spraying device always maintains a specific distance from the slope, thereby achieving a high degree of consistency in the coverage of the pesticide and significantly improving the precision of slope maintenance. This invention introduces a dynamic wind field disturbance prediction model, which fully considers the traffic-induced airflow unique to the highway environment. By coupling the baseline wind field with the instantaneous turbulence generated by vehicles through computational fluid dynamics logic, this invention enables the path planning algorithm to have a robust design against complex airflow disturbances. In actual operation, the system can predictively adjust the flight attitude and lateral compensation thrust of the UAV, which greatly reduces the risk of flight trajectory deviation caused by instantaneous airflow impact, effectively prevents collisions between the UAV and slope obstacles or road vehicles, and significantly enhances the safety and reliability of the operation process.

[0025] (2) The variable mass dynamics flight model constructed in this invention corrects in real time the changes in the UAV's center of gravity shift, mass reduction, and dynamic response characteristics caused by the reduction of the pesticide load. This dynamic adaptation mechanism enables the UAV to maintain the optimal energy efficiency ratio throughout the entire operation cycle, avoiding power overload caused by insufficient load estimation during the pesticide supply phase, and also avoiding control overshoot caused by reduced inertia at the end phase. This precise mass feedback control maximizes pesticide utilization and significantly extends the UAV's endurance resource utilization rate. By executing a multi-objective collaborative optimization algorithm, this invention deeply integrates coverage efficiency, energy loss, and safety risks. The path search process is no longer a simple geometric obstacle avoidance, but a global optimal solution search based on multi-dimensional constraints of terrain, wind field, and load. With the real-time feedback dynamic correction logic, the system can cope with sudden environmental changes, ensuring that the spraying operation can maintain extremely high stability and accuracy under various harsh conditions. By using a non-uniform rational spline interpolation algorithm for slope reconstruction, extremely complex natural geographical features can be described with low parameter complexity. While ensuring fitting accuracy, the processing pressure on the airborne computing platform is greatly reduced, meeting the computational requirements for real-time path generation.

[0026] (3) By introducing a risk cost term based on wind field uncertainty into the path planning, autonomous avoidance of dangerous areas is achieved. In specific terrain areas with large wind shear, the algorithm automatically adjusts the flight path to balance operational coverage and flight safety, demonstrating strong environmental adaptability. By using an adaptive filter to process the interference signal generated by the sloshing of the agent, the authenticity of the sensor feedback data is ensured, making the dynamic control commands smoother and reducing the energy loss and mechanical wear caused by frequent motor speed adjustments.

[0027] (4) The system structure provided by this invention has high modularity and scalability. The modules of spatial data perception, terrain modeling, wind field analysis and dynamic calculation are both independent and tightly coupled, and coordinated through a unified path planning core. This not only improves the system's operating efficiency, but also reserves standard interfaces for the subsequent integration of more types of sensors or optimization algorithms. This invention fundamentally solves the core structural problems of poor terrain fitting, weak environmental adaptability and unstable energy efficiency in highway slope plant protection operations by deeply modeling and logically coupling the three dimensions of terrain, environment and load. Experiments and practical applications show that the method and system described in this invention can significantly improve the uniformity of slope spraying, effectively control operational energy consumption, and greatly reduce the accident rate, providing solid technical support for the intelligent and refined transformation of highway slope ecological protection. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the overall technical solution architecture of a method for planning the path of spraying pesticides on highway slopes based on unmanned aerial vehicles (UAVs) proposed in this invention. Figure 2 This is a schematic diagram of the core principle framework of the multi-objective collaborative optimization algorithm based on terrain, wind field and dynamic constraints in this invention; Figure 3 This is a logical flowchart of the acquisition of multi-dimensional spatial information and three-dimensional digital modeling of highway slopes in this invention; Figure 4 This is a flowchart illustrating the logical flow of the dynamic wind field disturbance prediction and variable mass dynamics model construction in this invention. Figure 5 This is a schematic diagram of the multi-level interaction relationships and data flow between various functional modules within the path planning system of this invention. Detailed Implementation

[0029] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0030] Example 1 This invention provides a method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs). Please refer to [link / reference]. Figure 1 This embodiment provides a method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs). This method is mainly applied to vegetation protection scenarios on slopes with complex terrain, steep slopes, and dynamic interference from highway traffic. Figure 1 As shown, this invention uses multi-dimensional perception to digitally define the slope environment in all aspects, and combines the dynamic evolution characteristics of the UAV itself to achieve high-precision and high-stability operation path generation.

[0031] Specifically, the execution flow of this embodiment strictly follows the following steps: In step S1, multi-dimensional spatial geographic information of the highway slope is acquired using multi-source remote sensing equipment, and feature extraction and coordinate alignment are performed on the acquired multi-dimensional spatial geographic information. This step is the data foundation for the entire route planning. To ensure the comprehensiveness of the data, this embodiment uses multiple heterogeneous sensors for simultaneous acquisition.

[0032] In step S11, an airborne LiDAR scanner is used to perform a high-frequency scan of the highway slope. The LiDAR system sends hundreds of thousands of laser pulses per second. By measuring the time interval from the emission of the laser pulse to its contact with the target surface and its return, and combining this with the speed of light constant, a product operation is performed and divided by 2 to accurately obtain the three-dimensional spatial coordinates of the target point relative to the sensor. This raw point cloud data densely covers every minute fold of the slope, and its sampling frequency is typically set between 100 Hz and 200 Hz to ensure sufficient point cloud density is maintained even when the UAV is flying at high speed.

[0033] In step S12, image data of the slope vegetation is acquired using a multispectral imaging device. The multispectral camera has sensors with multiple independent bands, including red, green, blue, near-infrared, and red-edge bands, enabling it to capture differences in plant reflectance across different spectral frequencies. By adding, subtracting, and rationing the reflectance intensity of the near-infrared band and the red band, numerical indicators reflecting vegetation growth vitality can be calculated, thereby identifying the distribution density, species information, and growth status of the vegetation on the slope. This provides a basis for subsequent determination of spraying intensity.

[0034] In step S13, high-precision positioning and attitude information of the sensor at the time of acquisition is obtained through the differential global navigation satellite system. To eliminate positioning errors caused by ionospheric and tropospheric delays, this embodiment employs real-time dynamic carrier phase differential technology. By setting up a reference station with known coordinates on the ground, the correction value of the satellite signal is calculated in real time and sent to the UAV. By subtracting the phase observation value from the reference station's observation value from the UAV's observation value, most common errors can be eliminated, enabling the UAV's horizontal and vertical positioning accuracy to reach the level of 2 to 5 centimeters. Simultaneously, combined with the pitch, roll, and yaw angle information obtained by the inertial measurement unit, accurate spatial orientation reference is provided for each frame of scan data.

[0035] In step S14, the original point cloud data, image data, and positioning attitude information are fused. Using collinearity equation logic, the pixels of the two-dimensional multispectral image are projected and mapped one-to-one to the points in the three-dimensional point cloud space, thus endowing the point cloud, which originally only contained geometric information, with rich spectral features. During this process, the system uses outlier filtering logic to identify and remove abnormal data points caused by atmospheric dust, water mist, or sensor electronic noise. Finally, all information is transformed into a unified world coordinate system, generating a standardized slope spatial dataset.

[0036] In step S2, a variable-scale three-dimensional digital terrain model of the highway slope is constructed based on the processed multi-dimensional spatial geographic information.

[0037] In step S21, non-uniform sampling is performed on the standardized slope spatial dataset. The system iterates through all data points and calculates the angle between the normal vectors of adjacent data points. When the angle between the normal vectors is greater than a preset threshold of 10 degrees, the area is identified as a steep slope, fault, or protrusion with drastic topographic changes. In this case, the system automatically reduces the sampling step size and increases the sampling point density in the area to fully preserve the topographic features. Conversely, when the angle between the normal vectors is less than or equal to 10 degrees, it is identified as a gentle area, and the system reduces data redundancy by increasing the sampling step size. This mechanism of dynamically adjusting the sampling frequency based on curvature changes greatly reduces the burden of subsequent calculations while ensuring modeling accuracy.

[0038] In step S22, a non-uniform rational spline interpolation algorithm is used to reconstruct the surface of the sampling points. The system transforms discrete spatial sampling points into a continuous mathematical surface representation by defining a series of control points, weighting factors, and basis functions. During the reconstruction process, the positions of the control points and the influence weights corresponding to each point are adjusted to ensure that the generated continuous slope surface fits the original terrain with minimal deviation. This reconstruction method generates a continuous slope geometric description function, enabling the UAV to obtain the slope, aspect, and normal direction at any coordinate point in real time during path planning.

[0039] In step S23, the reconstructed surface is divided into regional grids based on vegetation density. For areas with dense vegetation cover, the system sets a finer grid size, such as a 0.5m x 0.5m fine grid; while for areas with sparse vegetation or bare ground, a 2m x 2m coarse grid is used. This multi-resolution gridding strategy allows the path planning algorithm to allocate more computational resources to key operational areas.

[0040] In step S3, a dynamic wind field disturbance prediction model for highway slope environments is constructed. Highway environments differ from open farmland; their unique terrain-induced winds and traffic-induced airflows have a significant impact on the flight stability of UAVs.

[0041] In step S31, real-time meteorological parameters are acquired through wireless weather stations deployed around the slope. These parameters include background wind direction, wind speed, and current atmospheric pressure.

[0042] In step S32, real-time traffic flow information and average vehicle speed are collected. The system uses video surveillance equipment or microwave radar deployed along the highway to count the total number of vehicles passing per unit time in real time, distinguishing between large trucks, medium-sized buses, and small private cars. Based on the vehicle's cross-sectional shape coefficient and speed, the system calculates the instantaneous induced airflow intensity caused by the vehicle's rapid movement. For example, when a large truck passes at 80 km / h, it creates a significant negative pressure zone and a strong turbulent wake behind it.

[0043] In step S33, the reference wind field and the instantaneous induced airflow are superimposed and coupled using computational fluid dynamics logic. The system uses the three-dimensional geometric profile of the slope as the boundary condition for fluid motion, simulating the acceleration, deceleration, and vortex formation processes of the airflow as it sweeps across the slope. By performing vector synthesis operations on the background wind vector and the vehicle-induced wind vector, the local wind vector distribution of each grid cell in the near-surface region of the slope is calculated.

[0044] In step S34, a wind field uncertainty assessment index is established. The system quantifies the lift disturbance and lateral drift moment generated by the local wind field on the UAV flight by analyzing the variance and frequency of change of the wind vector over a short period of time. If the uncertainty index exceeds the preset safety limit, the system will reserve a larger safety margin during the path planning stage.

[0045] In step S4, a variable mass dynamics model of the UAV considering the real-time loss of the drug load is established.

[0046] In step S41, the status inside the reagent tank is monitored in real time. Using a pressure sensor or a high-precision ultrasonic level gauge installed at the bottom of the reagent tank, the system obtains the real-time value of the remaining reagent at a frequency of 50 Hz.

[0047] In step S42, an evolution model of the total mass of the UAV is established based on the principle of mass conservation. As the spraying operation progresses, the flow output of the nozzles will directly lead to a decrease in the total mass of the system. The system calculates the function curve of mass decrease over time by recording the nozzle opening time and the preset flow coefficient.

[0048] In step S43, the inertial torque generated by the sloshing of the liquid agent is calculated. When the UAV performs acceleration, deceleration, or turning maneuvers, the liquid in the agent tank will generate complex surface fluctuations. The system calculates the offset of the liquid's center of gravity relative to the UAV's geometric center by acquiring real-time acceleration and angular velocity values ​​provided by the inertial measurement unit. Based on the physical relationship that torque equals force multiplied by lever arm, the additional disturbance torque generated by this center of gravity offset is calculated and added as a compensation term to the attitude control algorithm.

[0049] In step S44, the thrust distribution scheme is adjusted based on real-time mass changes. As the mass decreases, the total thrust required for the UAV to maintain hover decreases accordingly. The system automatically corrects the motor output power reference and redetermines the optimal cruise speed for the current state. Simultaneously, due to the reduced inertia caused by the decreased mass, the system tightens the attitude angle limits to prevent the UAV from oscillating due to excessive control sensitivity.

[0050] In step S5, a coverage spray path is generated.

[0051] In step S51, the cost function for path planning is defined. This function consists of three main parts: first, a coverage cost term, which aims to ensure that all grids marked with vegetation can be covered by the spray fan; second, an energy consumption cost term, which calculates the length of the drone's flight path and the additional work required to overcome wind resistance, so as to minimize the total power consumption; and third, a safety risk cost term, which takes the distance between the drone and the slope surface and the distance to obstacles such as utility poles as inputs, with the cost increasing the closer the distance.

[0052] In step S52, the normal distance constraint between the UAV spraying device and the slope is calculated using the slope geometric description function generated in step S22. The system requires that during flight, the projected distance of the UAV's center of gravity along the slope normal direction must always remain within a preset range of 2 to 3 meters. The ideal flight altitude of the UAV is determined in real time by calculating the partial derivatives of the path point coordinates with the slope function.

[0053] In step S53, a multi-objective heuristic search algorithm is used to search for the optimal path point sequence on the gridded model. The search algorithm references the local wind vector provided in step S33 when expanding at each node. If a path segment is located in a strong crosswind region, the algorithm automatically calculates a offset displacement biased towards the wind source to compensate for the drift of the spray caused by the wind.

[0054] In step S54, the searched discrete path points are smoothed. A high-order polynomial fitting technique is used to connect the discrete points into a continuous, smooth flight trajectory with a radius of curvature that meets the turning limits of the UAV.

[0055] This embodiment also provides a system for implementing the above method, the structure of which includes: a spatial data sensing module, integrating a lidar, a multispectral camera and a differential positioning unit; a terrain digital modeling module, which has an adaptive non-uniform sampling function; an environmental wind field analysis module, which can receive vehicle flow dynamic data and perform flow field coupling; a variable mass dynamics calculation module, which updates mass parameters in real time through sensor feedback; and a path planning and control core module, which is used to perform multi-objective optimization and output control commands.

[0056] Example 2 This embodiment is an explanation based on Embodiment 1. Please refer to it. Figures 1 to 5 Specifically: the spatial data sensing module is used to acquire and process multi-dimensional spatial geographic information of the slope through multi-source remote sensing devices. The spatial data sensing module integrates an airborne lidar scanner, a multispectral imaging device and a differential global navigation satellite system, and includes a high-frequency data synchronization clock for time alignment of point cloud, image and positioning data. The terrain digital modeling module, connected to the spatial data sensing module, is configured to construct a variable-scale three-dimensional digital terrain model through non-uniform sampling and non-uniform rational spline interpolation algorithms. The terrain digital modeling module has an adaptive subdivision function, which dynamically generates a multi-resolution mesh structure based on the local curvature of the slope. The environmental wind field analysis module is configured to acquire meteorological parameters and traffic flow dynamic parameters, and output dynamic wind field disturbance prediction results through computational fluid dynamics logic. The variable mass dynamics solution module is used to monitor the state inside the agent tank in real time and establish a variable mass dynamics model of the UAV that takes into account the real-time loss of agent load. The variable mass dynamics solution module has an adaptive filter built in it to smooth the instantaneous shift signal of the center of gravity caused by fluctuations in agent liquid level. The path planning and control core module is connected to the terrain digital modeling module, the environmental wind field analysis module, and the variable mass dynamics solution module, respectively. It is used to set the path planning cost function, generate the optimal path point sequence on the grid structure of the three-dimensional digital terrain model through a multi-objective collaborative optimization algorithm, and use a trajectory smoother to convert discrete path points into continuous flight trajectories.

[0057] The spatial data sensing module integrates a point cloud filtering algorithm, which is used to automatically identify and remove non-vegetation interference signals in the background environment based on the reflection intensity characteristics of the point cloud. The environmental wind field analysis module is connected to a wireless sensor network deployed along the highway, which is used to receive real-time data streams from ground weather stations and traffic monitoring systems, and to correct the wind field model parameters based on the difference between the actual correction value of the UAV's flight attitude and the predicted wind force. The variable mass dynamics solution module synchronously reduces the proportional gain coefficient of the flight controller according to the decrease in the remaining amount of the agent, so as to prevent the UAV from generating control oscillations due to the decrease in rotational inertia; The path planning and control core module includes a trajectory tracking compensator. When the deviation between the measured position and the preset path exceeds a preset deviation threshold during flight, a correction vector is calculated based on the current mass parameters and wind field parameters. By performing a disturbance search in the control parameter space, a control command sequence to restore the preset path is output.

[0058] Example 3 This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, this embodiment optimizes the path planning method for extremely steep, fractured rock slopes. In this application scenario, the slope angle generally exceeds 60 degrees, and the surface contains numerous isolated rocks and depressions.

[0059] During step S1, this embodiment increases the frequency of UAV oblique photography. Due to the extremely steep slope, vertical downward scanning often results in the loss of a large amount of facade information. Therefore, the spatial data perception module controls the gimbal to perform cross-scanning at a 45-degree tilt angle to obtain a more complete facade point cloud.

[0060] In performing non-uniform sampling in step S21, this embodiment lowers the preset threshold of the angle between the normal vectors to 5 degrees. This means the system is more sensitive to terrain changes and generates an extremely dense sampling network when faced with minute protrusions formed by broken rocks. In the surface reconstruction in step S22, automatic obstacle recognition logic is introduced. Feature extraction is performed on the multispectral image using deep learning algorithms to identify pixel regions in the point cloud that belong to utility poles, guardrails, base station towers, and large isolated rocks. When constructing the continuous slope function, the system marks these regions as "geometric forbidden zones," meaning the values ​​of the surface function at these coordinate points are set to infinity or defined as impenetrable physical boundaries.

[0061] Example 4 This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 Specifically, in performing the wind field coupling calculation in step S33, this embodiment focuses on the strong "narrowing effect" generated by steep slopes. When the background wind blows towards narrow slope gullies, the wind speed will increase significantly. The system simulates the intense friction and turbulence transformation of near-surface airflow by adding a terrain roughness coefficient to the wind field prediction model. For traffic interference, this embodiment sets a dynamic weighting factor. When a large, heavily loaded truck is detected passing by, the system automatically increases the weight of wind field disturbances within the next 5 seconds by 200% and instructs the UAV to execute an "early yaw" strategy, that is, before the airflow impact arrives, it tilts at a small angle towards the direction of the airflow source to use its own dynamic balance to offset the instantaneous impact.

[0062] When performing step S43 regarding the handling of agent sloshing, this embodiment enhances the algorithm to address the large-scale center of gravity shift unique to steep slope flight. During steep slope operations, the UAV needs to maintain a parallel attitude to the slope, resulting in a large fuselage tilt angle, causing liquid in the agent tank to accumulate on one side. This embodiment reduces the free surface area of ​​the liquid by installing longitudinal and lateral anti-sloshing plates inside the agent tank. The dynamics calculation module calculates the static center of gravity position of the liquid in the tilted state based on the current fuselage tilt angle and, combined with the dynamic shift generated by real-time acceleration, accurately compensates for the motor thrust torque.

[0063] In the path generation step S5, the coverage term in the cost function is given the highest weight for steep slope scenarios. Due to the steep slope, the pesticide solution easily flows downwards under gravity. Therefore, in the spraying constraints of step S52, the system not only controls the altitude but also adjusts the spray pressure of the nozzles in real time. When the drone moves uphill, the system instructs to increase the output current of the spray pump, increase the atomization pressure, and increase the kinetic energy of the pesticide droplets, enabling them to overcome the interference of airflow on the slope and be vertically deposited on the vegetation surface. At the same time, the path search algorithm prioritizes lateral reciprocating routes rather than longitudinal climbing routes to reduce the drastic energy consumption fluctuations caused by frequent ascents.

[0064] Example 5 A drone-based vegetation spraying path planning system for highway slopes, please refer to... Figures 2 to 5 Specifically, this embodiment focuses on path planning for the complex slope environment at the junction of a highway tunnel exit and an elevated bridge. This type of environment is characterized by extremely unstable instantaneous wind shear and very strong spatial constraints.

[0065] During step S1, the spatial data perception module integrates an ultrasonic radar array as an auxiliary to the airborne lidar. Due to the drastic changes in light and shadow at the tunnel entrance, traditional optical imaging is easily interfered with. The ultrasonic radar can provide stable obstacle distance feedback at close range through the principle of sound wave reflection, ensuring the safety of the UAV when approaching the tunnel slope.

[0066] During steps S31 and S32, the environmental wind field analysis module connects to a group of wind speed vector sensors deployed inside and outside the tunnel entrance. The system monitors the atmospheric pressure difference inside and outside the tunnel in real time. The "piston wind" generated by the pressure difference produces a huge air thrust at the moment the vehicle exits the tunnel. By establishing a correlation function between the pressure difference and wind speed, the system predicts the sudden change in the wind field in the slope area at the moment the vehicle exits the tunnel.

[0067] In the path search process of step S53, this embodiment introduces a search logic based on risk evolution. In areas with high wind field uncertainty assessment indicators, the cost function calculates a "safety buffer". The size of this buffer is directly proportional to the variance of the current wind speed. If it is predicted that there may be instantaneous wind force exceeding 70% of the drone's wind resistance level in the future, the path planning module will automatically shift the flight path 1.5 meters away from the slope obstacle.

[0068] In the trajectory smoothing process of step S54, this embodiment employs a time-optimal smoothing strategy. Considering the rapid changes in wind field at the tunnel entrance, the UAV needs to have an extremely fast response speed. When generating curves, the trajectory smoother rigorously verifies the centripetal acceleration value of each curve segment. If the centripetal force required for a certain turning trajectory exceeds 80% of the maximum radial thrust that the current motor can provide with its remaining power, the system will forcibly increase the turning radius, sacrificing some operational efficiency in exchange for absolute safety during flight.

[0069] Furthermore, the variable mass dynamics calculation module in this embodiment incorporates an adaptive filter. This filter can monitor the data stream returned by the remaining reagent sensor in real time, and by comparing the numerical fluctuations over 10 consecutive sampling periods, it identifies false liquid level signals caused by severe shaking of the fuselage. The filter employs weighted average logic, giving higher weight to historically stable values, thereby outputting smooth and accurate mass parameters and preventing the power system from making erroneous thrust adjustment actions due to false load fluctuations.

[0070] To support the operation of the above embodiments, the system described in this invention has undergone deep integration in terms of hardware and software logic architecture.

[0071] The spatial data perception module is not simply a stack of sensors; it contains a high-frequency data synchronization clock. This clock timestamps every line of point cloud data from the LiDAR, every frame of image data from the multispectral camera, and every coordinate from the positioning system with nanosecond-level precision. This ensures that data from different sources are precisely aligned in the temporal dimension during UAV movement, eliminating geometric misalignment caused by asynchronous sensor sampling. The module also integrates a point cloud filtering algorithm that can automatically distinguish between concrete slopes, rocks, and green vegetation based on the reflection intensity characteristics of the point cloud.

[0072] The core of the terrain digital modeling module lies in its adaptive subdivision logic. It doesn't just statically build the model; instead, it dynamically adjusts the resolution as the path planning progresses. When the UAV is far from a certain slope area, the module outputs a low-resolution, coarse terrain map; once the UAV enters the operational radius, the module immediately refines the mesh of that area using interpolation algorithms, generating a high-precision 3D surface. This on-demand generation mode significantly saves memory resources on the onboard computing platform.

[0073] The environmental wind field analysis module has real-time learning capabilities. It not only uses preset fluid dynamics logic for prediction, but also corrects the parameters of the wind field model by comparing the actual correction values ​​of the UAV's flight attitude with the predicted wind force. For example, if the lateral force actually felt by the UAV at a certain location is greater than the predicted value, the system will increase the weight of the "terrain roughness" around that coordinate point, thereby outputting a more accurate wind field vector in the next loop.

[0074] The variable mass dynamics solution module is a closed-loop control system. It constructs a complete multi-degree-of-freedom dynamic equation by receiving data from the kit sensors, battery level monitor, and inertial measurement unit. It calculates the current rotational inertia matrix of the drone. As the kit depletes, the rotational inertia decreases, meaning the drone becomes more responsive to control commands. To prevent control overload, this module automatically reduces the proportional gain coefficient of the flight controller, ensuring a smooth transition in flight attitude.

[0075] The path planning and control core module is the brain of the entire system. It is responsible not only for path search but also for real-time path correction. This module includes a trajectory tracking compensator. During flight, if the deviation between the measured position and the preset path exceeds 10 centimeters, the compensator immediately intervenes. It calculates an optimal correction vector based on the current mass parameters and wind field parameters. This process does not rely on preset fixed formulas but rather searches for control command sequences that can return to the flight path with minimal energy cost by performing small perturbations within the control parameter space.

[0076] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: S1. Obtain multi-dimensional spatial geographic information of highway slopes through multi-source remote sensing devices, acquire raw point cloud data using airborne lidar scanners, collect image data of slope vegetation using multispectral imaging devices, and acquire high-precision positioning and attitude information of sensors at the time of acquisition using differential global navigation satellite system. Then, fuse the raw point cloud data, image data, and positioning attitude information, remove environmental noise interference, and generate a standardized slope spatial dataset under a unified coordinate system. S2. Non-uniform sampling is performed on the standardized slope spatial dataset. The sampling step size is adjusted according to the curvature change of the slope surface. The non-uniform rational spline interpolation algorithm is used to reconstruct the surface of the sampling points, generate a continuous slope geometric description function, and perform regional grid segmentation on the reconstructed surface according to the vegetation distribution density to construct a variable-scale three-dimensional digital terrain model of the highway slope, which is used to characterize the surface geometric contour and slope aspect change characteristics of the slope. S3. Obtain the reference wind direction, reference wind speed and atmospheric pressure of the highway section, collect highway traffic flow information and average vehicle speed in real time, calculate the intensity of instantaneous induced airflow caused by vehicle movement, superimpose and couple the reference wind field and instantaneous induced airflow through computational fluid dynamics logic, calculate the local wind vector distribution in the near-surface area of ​​the slope, establish wind field uncertainty assessment index, and construct a dynamic wind field disturbance prediction model in the highway slope environment. S4. Real-time monitoring of the liquid level in the agent tank; based on the principle of mass conservation and the spray flow rate setting, establish an evolutionary relationship equation between the total mass and the travel distance of the UAV; obtain real-time acceleration and angular velocity values ​​provided by the inertial measurement unit; calculate the inertial torque generated by the liquid agent sloshing and add it as a disturbance term to the attitude control equation; adjust the thrust distribution scheme of the UAV motor according to the real-time mass change; determine the optimal cruise speed and maximum tilt angle limit under the current mass; and establish a UAV variable mass dynamics model that considers the real-time loss of agent load. S5. Set a path planning cost function that includes coverage cost, energy consumption cost, and safety risk cost. Calculate the normal distance constraint between the UAV spraying device and the slope based on the slope geometric description function. Use a multi-objective heuristic search algorithm to search for the optimal path point sequence on the gridded slope model. In the search process, introduce wind direction compensation provided by the dynamic wind field disturbance prediction model to smooth the optimal path point sequence, generate a coverage spraying path, and dynamically correct the generated path based on real-time feedback data.

2. The method for planning the spraying path for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: Step S1 specifically includes: By using an airborne lidar scanner to measure the time interval from the emission of a laser pulse to its contact with the surface of the target object and its return, and combining this with the speed of light constant, the three-dimensional spatial coordinates of the target point relative to the sensor are obtained. Multispectral imaging equipment is used to capture the differences in reflectance of plants in different spectral bands. By performing logical operations on the reflectance intensity of the near-infrared band and the reflectance intensity of the red band, numerical indicators reflecting the growth vitality of vegetation are calculated, and the distribution density, species information and growth status of vegetation on the slope are identified. By using real-time dynamic carrier phase differential technology through the differential global navigation satellite system, the phase observation values ​​of the ground reference station and the observation values ​​of the UAV are calculated to eliminate common errors and obtain positioning information. The pitch, roll and yaw angle information obtained by the inertial measurement unit is combined as attitude information. By utilizing collinearity equation logic, the pixels of two-dimensional multispectral images are projected and mapped one by one to the points in three-dimensional point cloud space, thus endowing the point cloud with spectral features. Outlier filtering logic is used to identify and remove abnormal data points caused by atmospheric dust, water mist, or sensor electronic noise, and the information is transformed into a unified world coordinate system to generate a standardized slope spatial dataset.

3. The method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: Step S2 involves non-uniform sampling of the standardized slope spatial dataset, including: Traverse the data points in the standardized slope spatial dataset and calculate the angle between the normal vectors of adjacent data points; The angle between the normal vectors is compared with a preset angle threshold. When the angle between the normal vectors is greater than the preset angle threshold, the area where the data point is located is determined to be an area with drastic terrain changes, and the sampling step size is reduced and the sampling point distribution density is increased. When the angle between the normal vectors is less than or equal to a preset angle threshold, the area where the data point is located is determined to be a flat area, and the sampling step size is increased and the sampling point distribution density is reduced.

4. The method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: Step S2 involves using a non-uniform rational spline interpolation algorithm to reconstruct the surface from the sampling points, including: Define a series of control points, weight factors, and basis functions to transform discrete sampling points into a continuous mathematical surface representation; By adjusting the position of the control points and the influence weight of each control point, the generated continuous slope is made to fit the original terrain. The slope value, aspect value and normal direction at any coordinate point are obtained in real time through the slope geometric description function. Obstacle recognition and removal logic is introduced during the surface reconstruction process. Deep learning object detection logic is used to identify utility poles, guardrails and isolated rocks in point cloud data, and the identified areas are marked as geometric boundaries that are prohibited from being crossed during surface reconstruction.

5. The method for planning the spraying path for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The process of calculating the instantaneous induced airflow intensity in step S3 includes: The total number of vehicles passing through per unit time is counted, and the vehicles are classified according to their size and outline characteristics. The instantaneous induced airflow intensity caused by vehicle movement is calculated based on the vehicle's cross-sectional shape factor, driving speed, and horizontal offset of the slope from the driving distance. When the classification result is a large heavy-duty truck, the wind field disturbance weight in the corresponding time period is increased, and a pre-command is triggered for the UAV to increase lateral compensation thrust.

6. The method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The process of establishing wind field uncertainty assessment indicators in step S3 includes: The variance and frequency of change of local wind vectors within a preset time period are analyzed to quantify the lift disturbance and lateral drift moment generated by the local wind field on the flight of the UAV. The wind field uncertainty assessment index is calculated, and the safety risk cost term is positively correlated with the wind field uncertainty assessment index. In areas with severe wind field fluctuations, the path planning cost function guides the UAV to deviate from steep edges or areas with dense obstacles, and selects flight channels with stable airflow.

7. The method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: Step S4 involves calculating the inertial torque generated by the sloshing of the liquid medicine, including: The real-time value of the remaining amount of medicine is obtained by using a pressure sensor or ultrasonic level gauge installed at the bottom of the medicine tank. Based on the real-time value of the remaining amount of the agent, combined with the current tilt angle of the drone, the static center of gravity of the liquid in the tilted state is calculated. By combining the dynamic offset generated by real-time acceleration, the offset of the liquid center of gravity relative to the geometric center of the UAV is calculated, and the additional disturbance torque generated by the offset of the center of gravity is used as the inertial torque. An adaptive filter is introduced to identify and eliminate false liquid level signals caused by drastic changes in the attitude of the UAV by comparing the numerical fluctuations within a continuous sampling period.

8. The method for planning spraying paths for vegetation on highway slopes based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The process of generating a coverage spray path in step S5 includes: Using the slope geometry description function, the projected distance of the UAV's center of gravity in the slope normal direction is calculated to ensure that the projected distance is maintained within the preset normal distance range; The ideal flight altitude of the UAV is determined by calculating the partial derivatives of the path point coordinates with the slope geometric description function; During the execution of the multi-objective heuristic search algorithm, the local wind vector distribution is referenced. When the search path is in the crosswind region, the offset displacement towards the wind source is calculated to compensate for the drift of the drug mist caused by the wind. Adjust the atomization pressure of the nozzle according to the slope inclination angle. When the slope increases, increase the spray flow rate in the uphill direction to counteract the downward offset of the drug droplets due to gravity. A high-order polynomial fitting technique is used to smooth the optimal path point sequence, connecting discrete points into a continuous, smooth flight trajectory with a radius of curvature that meets the turning limits of the UAV.

9. The UAV-based highway slope vegetation spraying path planning system according to claim 1, applied to the UAV-based highway slope vegetation spraying path planning method according to any one of claims 1 to 8, characterized in that: include: The spatial data sensing module is used to acquire and process multi-dimensional spatial geographic information of slopes through multi-source remote sensing devices. The spatial data sensing module integrates an airborne lidar scanner, multispectral imaging equipment and differential global navigation satellite system, and includes a high-frequency data synchronization clock for time alignment of point cloud, image and positioning data. The terrain digitization modeling module, connected to the spatial data perception module, is configured to construct a variable-scale 3D digital terrain model through non-uniform sampling and non-uniform rational spline interpolation algorithms. The terrain digitization modeling module has an adaptive subdivision function, which dynamically generates a multi-resolution mesh structure based on the local curvature of the slope. The environmental wind field analysis module is configured to acquire meteorological parameters and traffic flow dynamic parameters, and output dynamic wind field disturbance prediction results through computational fluid dynamics logic. The variable mass dynamics solution module is used to monitor the state inside the agent tank in real time and establish a variable mass dynamics model of the UAV that takes into account the real-time loss of agent load. The variable mass dynamics solution module has an adaptive filter built in, which is used to smooth the instantaneous shift signal of the center of gravity caused by fluctuations in agent level. The path planning and control core module is connected to the terrain digital modeling module, the environmental wind field analysis module, and the variable mass dynamics solution module, respectively. It is used to set the path planning cost function, generate the optimal path point sequence on the grid structure of the three-dimensional digital terrain model through a multi-objective collaborative optimization algorithm, and use a trajectory smoother to convert discrete path points into continuous flight trajectories.

10. A drone-based vegetation spraying path planning system for highway slopes according to claim 9, characterized in that: The spatial data sensing module integrates a point cloud filtering algorithm, which is used to automatically identify and remove non-vegetation interference signals in the background environment based on the reflection intensity characteristics of the point cloud. The environmental wind field analysis module is connected to a wireless sensor network deployed along the highway to receive real-time data streams from ground weather stations and traffic monitoring systems, and to correct the wind field model parameters based on the difference between the actual correction values ​​of the UAV's flight attitude and the predicted wind force. The variable mass dynamics solution module synchronously reduces the proportional gain coefficient of the flight controller as the amount of remaining agent decreases, in order to prevent the UAV from experiencing control oscillations due to the decrease in rotational inertia. The path planning and control core module includes a trajectory tracking compensator. When the deviation between the measured position and the preset path exceeds the preset deviation threshold during flight, the correction vector is calculated based on the current mass parameters and wind field parameters. By performing a disturbance search in the control parameter space, the control command sequence that restores the preset path is output.

Images