Methods for the formulation and spraying control of pesticide adjuvants for agricultural drones

Abstract

This invention relates to the field of agricultural aviation intelligent equipment and precision spraying technology, and particularly to a method for the formulation and spraying control of chemical adjuvants in agricultural drones, comprising the following steps: Step S1, acquiring flight status parameters of the agricultural drone and three-dimensional topographic data of the crop canopy below, and constructing a real-time aerodynamic flow field model superimposed with the downwash airflow of the agricultural drone rotor and the environmental wind field; Step S2, predicting the sedimentation trajectory of the pesticide droplets based on the real-time aerodynamic flow field model, calculating the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generating an instantaneous mixing ratio including the main pesticide solution and various adjuvants; In this invention, by constructing a real-time aerodynamic flow field model and directly sending the instantaneous mixing ratio of the main pesticide solution and various adjuvants to a microfluidic mixing device installed at the end of the nozzle for in-situ mixing, the physical space limitations and response lag of traditional agricultural drone pipeline mixing are broken.

Description

Technical Field

[0001] This invention relates to the field of agricultural aviation intelligent equipment and precision spraying technology, and in particular to a method for the formulation and spraying control of chemical adjuvants for crop protection drones. Background Technology

[0002] In recent years, with the rapid development of agricultural aviation technology, plant protection drones have been widely used in pesticide spraying operations due to their advantages such as high operating efficiency and strong terrain adaptability. Existing plant protection drones mainly use fully pre-mixed pesticides in the tank or online mixing in the main pipeline for spraying operations, combined with variable control based on basic flight speed or prescription maps. However, in the actual complex farmland operation environment, this traditional application method has obvious limitations. The existing mixing position is usually located at the front end of the main pipeline or water pump, which is a long physical distance from the nozzle. This means that after the system identifies local differences in pests and diseases or changes in the microenvironment and issues a command to change the pesticide ratio, the newly mixed pesticide still needs to be transported through the dead volume of the pipeline before reaching the nozzle terminal. This physical lag, which is difficult to eliminate, often causes the sprayed pesticide to be spatially misaligned with the actual canopy demand when the drone is flying at high speed.

[0003] Meanwhile, when agricultural drones operate at low altitudes, their high-speed rotating rotors generate strong downwash airflows. These downwash airflows collide with the natural wind field and the undulating crop canopy, forming extremely complex and rapidly changing local micro-meteorological vortices. Current technologies, whether using pre-prepared homogeneous pesticide solutions or slow-responding coarse-grained variable systems, cannot cope with these transient aerodynamic disturbances in real time. This makes it difficult for the physicochemical properties of the droplets to match the transient wind field, easily leading to severe off-target drift of the droplets with the escape vortex, or obstruction by dense canopy layers. In addition, most current plant protection spraying systems rely on unidirectional open-loop control logic. After the system completes the atomization spraying action according to the theoretical flow field model or preset parameters, it lacks an effective means to track the actual spatial trajectory of droplets in the real chaotic micro-meteorological environment. It is impossible to obtain the deviation between the actual sedimentation results and the theoretical prediction. This lack of a closed-loop calibration mechanism makes it impossible for the system to adaptively correct the error of the underlying flow field prediction model. When facing sudden changes in terrain or long-term operation, the accuracy of the system's application is difficult to be continuously and reliably guaranteed. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a method for the formulation and spraying control of chemical adjuvants for agricultural drones, aiming to improve the problems of physical space delay and response lag caused by traditional pipeline mixing.

[0005] In a first aspect, the present invention provides the following technical solution: a method for the formulation and spraying control of chemical adjuvants in agricultural protection drones, comprising the following steps: Step S1: Obtain the flight status parameters of the agricultural drone and the three-dimensional topographic data of the crop canopy below, and construct a real-time aerodynamic flow field model that superimposes the downwash airflow of the agricultural drone rotor and the environmental wind field. Step S2: Predict the sedimentation trajectory of the drug droplets based on the real-time aerodynamic flow field model, calculate the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generate the instantaneous mixing ratio containing the main drug and various adjuvants. Step S3: Send the instantaneous mixing ratio to the microfluidic mixing device installed at the end of the nozzle of the plant protection drone. Inside the microfluidic mixing device, the main drug solution and various adjuvants are mixed in situ according to the instantaneous mixing ratio. During the mixing process, fluorescent tracer adjuvants are periodically injected and then atomized and sprayed. Step S4: Use the multispectral sensor carried by the agricultural drone to obtain the actual settling trajectory of the droplets with the fluorescent tracer after atomization spraying in the real environment. Step S5: Compare the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss. When the trajectory deviation loss exceeds the preset calibration threshold, use the trajectory deviation loss to update the closed-loop parameters of the real-time aerodynamic flow field model.

[0006] Preferably, in step S1, the step of constructing a real-time aerodynamic flow field model that superimposes the rotor downwash airflow of the agricultural drone with the environmental wind field includes: The spatial position, flight attitude angle, and rotor speed of the agricultural drone are obtained as boundary conditions. Based on the three-dimensional topography data of the crop canopy, the superposition effect of rotor torque and air viscosity in the spatial grid is calculated using fluid dynamics calculation methods, and the gridded flow field velocity vector is output as the real-time aerodynamic flow field model.

[0007] Preferably, in step S2, the step of predicting the sedimentation trajectory of the liquid droplets based on the real-time aerodynamic flow field model includes: Extract the flow field velocity vector from the real-time aerodynamic flow field model; By combining the preset initial mass of a single droplet with the aerodynamic drag coefficient, the force state of the liquid droplet in space is calculated using the dynamic force relationship, and a predicted trajectory coordinate sequence in three-dimensional space is generated as the sedimentation trajectory.

[0008] Preferably, in step S2, the step of calculating the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generating the instantaneous mixing ratio of the main drug solution and various adjuvants, includes: Identify airflow disturbance regions present in the predicted trajectory coordinate sequence; The surface tension and dynamic viscosity required to resist airflow deviation are calculated based on the disturbance intensity of the airflow disturbance region. The instantaneous mixing ratio is generated by distributing the required amounts of the main drug solution, anti-drift agent, and penetration agent in reverse order of surface tension and dynamic viscosity.

[0009] Preferably, in step S3, the step of in-situ mixing of the main drug solution and various adjuvants according to the instantaneous mixing ratio inside the microfluidic mixing device includes: The instantaneous mixing ratio is converted into a driving electrical signal; The microchannel valve of the piezoelectric actuator inside the microfluidic mixing device is controlled by the driving electrical signal to mix the main drug solution, anti-drift agent and penetration agent by ultrasonic vibration in the nozzle end chamber.

[0010] Preferably, in step S3, the step of periodically injecting the fluorescent tracer during the mixing process followed by atomized spraying includes: The fluorescent tracer channel inside the microfluidic mixing device is activated at preset time intervals. A fixed tracer amount of the fluorescent tracer is injected into the current mixing chamber and mixed with the main drug solution and various additives, so that a specific batch of droplets released by the nozzle has fluorescent luminescence properties.

[0011] Preferably, in step S4, the step of using a multispectral sensor mounted on an agricultural drone to obtain the actual settling trajectory of droplets carrying the fluorescent tracer in a real environment after atomized spraying includes: Activate the capture mode in the multispectral sensor that matches the wavelength of the fluorescent tracer; The centroid positions of high-brightness droplets with the fluorescence emission characteristics are continuously extracted in three-dimensional space by image recognition, and the centroid positions are connected in chronological order to generate an actual trajectory centroid coordinate sequence as the actual sedimentation trajectory.

[0012] Preferably, in step S5, the step of comparing the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss includes: Extract the predicted trajectory coordinate sequence corresponding to a specific batch of droplets containing the fluorescent tracer; Align the predicted trajectory coordinate sequence with the actual trajectory centroid coordinate sequence on the time axis; The trajectory deviation loss is obtained by calculating the square integral of the spatial distance difference between the predicted centroid position and the actual centroid position at the same time node.

[0013] Preferably, in step S5, the step of updating the closed-loop parameters of the real-time aerodynamic flow field model using the trajectory deviation loss includes: When the trajectory deviation loss is greater than the preset calibration threshold, the gradient descent algorithm is used to calculate the residual of the trajectory deviation loss relative to the environmental wind field parameters in the real-time aerodynamic flow field model; The residual is added as a feedback compensation value to the flow field calculation weight of the previous cycle to complete the correction and calibration of the real-time aerodynamic flow field model.

[0014] Secondly, the present invention provides the following technical solution: a control system for the proportioning and spraying of pesticide adjuvants by agricultural drones, the system comprising: The flow field modeling module is used to acquire flight status parameters of agricultural drones and three-dimensional topographic data of the crop canopy below, and to construct a real-time aerodynamic flow field model superimposed with the downwash airflow of the agricultural drone rotor and the environmental wind field. The proportioning calculation module is used to predict the sedimentation trajectory of the drug droplets based on the real-time aerodynamic flow field model, calculate the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generate the instantaneous mixing ratio including the main drug and various adjuvants. The mixing and spraying module is used to send the instantaneous mixing ratio to a microfluidic mixing device installed at the end of the nozzle of the plant protection drone. Inside the microfluidic mixing device, the main drug solution and various adjuvants are mixed in situ according to the instantaneous mixing ratio. During the mixing process, fluorescent tracer adjuvants are periodically injected and then atomized and sprayed. The trajectory acquisition module is used to acquire the actual settling trajectory of droplets carrying the fluorescent tracer after atomized spraying in the real environment using the multispectral sensor carried by the agricultural drone. The model update module is used to compare the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss. When the trajectory deviation loss exceeds a preset calibration threshold, the closed-loop parameters of the real-time aerodynamic flow field model are updated using the trajectory deviation loss.

[0015] The present invention has the following beneficial effects: 1. This invention constructs a real-time aerodynamic flow field model and directly sends the instantaneous mixing ratio of the main drug solution and various adjuvants to a microfluidic mixing device installed at the end of the nozzle for in-situ mixing. This breaks through the physical space limitations and response lag of traditional plant protection drone pipeline mixing. This feature enables the system to directly change the drug solution formula in real time according to the changes in the eddy current of the underlying environment at the moment the drug solution is sprayed, effectively solving the problem of drug solution drift or uneven deposition caused by system delay and micro-meteorological changes.

[0016] 2. This invention predicts the sedimentation trajectory of drug droplets based on real-time aerodynamic flow field, and calculates the surface tension and dynamic viscosity required to resist airflow disturbance accordingly. It then dynamically adjusts the distribution ratio of anti-drift agent and penetration agent. This feature transforms the traditional open-loop static drug application into adaptive drug application based on the flow field environment, enabling the sprayed droplets to actively utilize or resist the local rotor downwash airflow, ensuring accurate sedimentation and deep penetration of the drug solution under complex canopy structures.

[0017] 3. This invention periodically injects fluorescent tracer additives during the microfluidic mixing process and uses an airborne multispectral sensor to track their actual sedimentation trajectory, thereby calculating the deviation to update the closed-loop parameters of the aerodynamic flow field model. This feature establishes a low-level bidirectional calibration mechanism between the physical spraying action and the digital prediction model, overcoming the defect that traditional open-loop prediction models are prone to accumulating errors in chaotic farmland environments, and ensuring the continuous accuracy of UAVs in long-endurance and complex terrain operations. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of a method for the formulation and spraying control of chemical adjuvants for plant protection drones proposed in this invention. Figure 2 This is a schematic diagram of the system modules of a method for controlling the formulation and spraying of pesticide adjuvants by a plant protection drone, as proposed in this invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: In the first embodiment of the present invention, the present invention provides a method for the formulation and spraying control of chemical adjuvants for agricultural drones, such as... Figure 1 As shown, it includes the following steps: Step S1: Obtain the flight status parameters of the agricultural drone and the three-dimensional topographic data of the crop canopy below, and construct a real-time aerodynamic flow field model that superimposes the downwash airflow of the agricultural drone rotor and the environmental wind field. In step S1, the steps for constructing a real-time aerodynamic flow field model that superimposes the rotor downwash airflow of the agricultural drone with the ambient wind field include: The spatial position, flight attitude angle, and rotor speed of the agricultural drone are obtained as boundary conditions. By combining three-dimensional topographic data of crop canopy, and using fluid dynamics calculation methods, the superposition effect of rotor torque and air viscosity in the spatial grid is calculated, and the gridded flow field velocity vector is output as a real-time aerodynamic flow field model.

[0021] Specifically, the system synchronously collects multi-source environmental and status data through multiple sensor arrays mounted on the plant protection drone. It uses an onboard real-time dynamic differential positioning module to obtain the three-dimensional spatial position of the plant protection drone in a three-dimensional geographic coordinate system. It uses an onboard inertial measurement unit to obtain the current flight attitude angle of the plant protection drone, which is composed of three spatial angular components: pitch angle, roll angle, and yaw angle. It uses an electronic speed controller communication interface to read and obtain the real-time motor rotor speed of each independent rotor of the plant protection drone. The above three-dimensional spatial position, flight attitude angle, and rotor speed are extracted and defined as the dynamic input boundary conditions of the flow field calculation domain.

[0022] Using a downward-firing lidar mounted on the underside of an agricultural drone, laser pulses are emitted downwards at a set scanning frequency and reflected echoes are received. The system analyzes the flight time data of the laser echoes to generate three-dimensional point cloud data of the crop canopy surface below. A spatial interpolation algorithm is then used to reconstruct the surface from the three-dimensional point cloud data, generating continuous three-dimensional morphological data of the crop canopy.

[0023] The system uses the geometric center of the agricultural drone as the origin, divides a three-dimensional Euler mesh within a set spatial range, and maps the generated three-dimensional topographic data of the crop canopy to the bottom boundary layer nodes of the three-dimensional Euler mesh, setting them as solid wall boundaries with physical resistance.

[0024] Within a three-dimensional Eulerian grid space, the superposition effect of rotor torque and air viscosity on the spatial grid is calculated using fluid dynamics methods. The calculation process employs the momentum control equation for an unsteady incompressible fluid to solve for the spatial flow field distribution. The specific form of this momentum control equation is as follows: In the formula, This represents the air density constant; The vector represents the velocity vector of the flow field at the node of the three-dimensional Eulerian grid; t represents the time variable. This represents the static pressure distribution in the flow field space; Indicates the dynamic viscosity of air; Represents the gradient operator; Represents the Laplace operator; This represents the torque generated by the rotor of an agricultural drone within a spatial grid, which is the momentum source term. This represents the torque generated by the environmental source term from the external wind field.

[0025] Among them, the momentum source term torque generated by the rotor Quantitative calculations were performed using an excitation disk aerodynamic model, and the calculation equations are as follows: In the formula, This represents the volume of a single 3D Euler mesh element; This indicates the rated thrust coefficient of the rotor of an agricultural drone; This indicates the swept area covered by the rotor's rotation; This indicates the obtained rotor speed; Indicates the physical radius of the rotor; This represents the downwash normal vector of the rotor plane calculated based on the obtained flight attitude angle.

[0026] The system substitutes the boundary conditions into the above calculation equations for iterative solution, obtaining the spatial grid node within the three-dimensional Eulerian grid. axis, axis, The velocity component values ​​in the three mutually perpendicular directions of the axis are aggregated and processed to output a gridded flow field velocity vector. This flow field velocity vector, which contains the velocity distribution data of the entire space grid, serves as a real-time aerodynamic flow field model to characterize the real space airflow distribution state after the physical superposition of the UAV rotor downwash airflow and the environmental wind field.

[0027] Step S2: Predict the sedimentation trajectory of the drug droplets based on the real-time aerodynamic flow field model, calculate the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generate the instantaneous mixing ratio containing the main drug and various adjuvants. In step S2, the step of predicting the sedimentation trajectory of the liquid droplets based on the real-time aerodynamic flow field model includes: Extract the velocity vector of the flow field from the real-time aerodynamic flow field model; By combining the preset initial mass of a single droplet with the aerodynamic drag coefficient, the force state of the liquid droplet in space is calculated using the dynamic force relationship, and a predicted trajectory coordinate sequence in three-dimensional space is generated as the sedimentation trajectory. In step S2, the steps of calculating the surface tension and dynamic viscosity required for droplets to reach the target canopy based on the sedimentation trajectory, and generating the instantaneous mixing ratio of the main drug solution and various adjuvants, include: Identify airflow disturbance regions present in the predicted trajectory coordinate sequence; Calculate the surface tension and dynamic viscosity required to resist airflow deviation based on the disturbance intensity in the airflow disturbance region; The required proportions of the main drug solution, anti-drift agent, and penetration agent are distributed according to the reverse distribution of surface tension and dynamic viscosity to generate an instantaneous mixing ratio.

[0028] Specifically, the system extracts the velocity vectors of the flow field corresponding to the nodes of the three-dimensional Eulerian grid in the real-time aerodynamic flow field model. The initial mass of a single droplet released by the agricultural drone nozzle is set to... The initial release rate is The aerodynamic drag coefficient is The system establishes a Lagrange particle tracking model of the droplets in three-dimensional space, with the nozzle center as the origin. Based on Newton's second law, the dynamic force state of the drug droplets in the spatial flow field is calculated, and its governing equations are as follows: In the formula, Indicates that the droplets are in The instantaneous velocity vector at time t; Represents the gravitational acceleration vector; This represents the vector of air drag force acting on the droplets; This represents the air buoyancy vector.

[0029] Among them, the air drag vector The calculation formula is: In the formula, This represents the air density constant; This represents the windward cross-sectional area of ​​a single fog droplet; This represents the velocity vector in the real-time aerodynamic flow field model corresponding to the current spatial location of the droplet.

[0030] The system is set to a fixed time step. The fourth-order Runge-Kutta method was used to numerically integrate and solve the above motion control equations, calculating the three-dimensional coordinates of the droplets in three-dimensional space at each time point. The three-dimensional coordinates of each time point were connected sequentially along the time axis to generate a predicted trajectory coordinate sequence in three-dimensional space, which was then used as the settling trajectory of the liquid droplets.

[0031] Based on the aforementioned settling trajectory, the system calculates the surface tension and dynamic viscosity required for droplets to reach the target canopy. The system iterates through each coordinate point in the predicted trajectory coordinate sequence, extracts the flow field velocity vector difference between adjacent coordinate points, and calculates the flow field velocity gradient. The system internally presets a flow field velocity gradient threshold; when the flow field velocity gradient at a certain coordinate point exceeds the preset threshold, the three-dimensional spatial range containing that coordinate point is marked as an airflow disturbance region. The system calculates the maximum velocity shear rate within this airflow disturbance region and uses it as an indicator of the disturbance intensity of the airflow disturbance region. .

[0032] For the identified airflow disturbance areas, the system uses disturbance intensity indices... Calculate the surface tension and dynamic viscosity required to resist airflow deviation. Set the maximum allowable Weber number for critical droplet breakup as follows. Target surface tension The calculation formula is: In the formula, This represents the maximum relative velocity component between the droplets and the flow field within the airflow disturbance region; This represents the average geometric diameter of the droplets; The first weighting coefficient preset for the system.

[0033] Set the target Reynolds number for maintaining the penetration state of the fog droplets as: Target dynamic viscosity The calculation formula is: In the formula, Indicates the density of the base drug solution; This is the second weighting coefficient preset by the system.

[0034] The system calculates the target surface tension. With target dynamic viscosity The required quantities of the main drug solution, anti-drift agent, and penetration agent are then distributed in reverse order. The system reads the calibration physicochemical parameter tables of the main drug solution, anti-drift agent, and penetration agent from the storage module and constructs a set of mixed physicochemical property equations: In the formula, , , These are the mass percentages of the main drug solution, anti-drift agent, and penetration agent, respectively. , , These are the calibrated surface tension constants of the three independent components mentioned above; , , These are the calibrated dynamic viscosity constants of the three independent components mentioned above; and These are the preset coefficients for the mixed interaction terms.

[0035] The system uses Newton's iterative algorithm to solve the above nonlinear equations and calculates the results that meet the requirements. , , Numerical values. The system converts the obtained mass percentage values ​​into corresponding volume fractions, generating an instantaneous mixing ratio instruction file containing the specific proportions of the main drug solution, anti-drift agent, and penetration agent.

[0036] Step S3: Send the instantaneous mixing ratio to the microfluidic mixing device installed at the end of the nozzle of the plant protection drone. Inside the microfluidic mixing device, the main drug solution and various adjuvants are mixed in situ according to the instantaneous mixing ratio. Fluorescent tracer adjuvants are periodically injected during the mixing process and then atomized and sprayed. In step S3, the process of in-situ mixing of the main drug solution and various adjuvants according to the instantaneous mixing ratio inside the microfluidic mixing device includes: Convert the instantaneous mixing ratio into a driving electrical signal; The microchannel valve opening of the piezoelectric actuator inside the microfluidic mixing device is controlled by the driving electrical signal, so that the main drug liquid, anti-drift agent and penetration agent are ultrasonically mixed in the nozzle end chamber; In step S3, the step of periodically injecting the fluorescent tracer during the mixing process followed by atomization spraying includes: The fluorescent tracer channel inside the microfluidic mixing device is activated at preset time intervals; A fixed amount of fluorescent tracer is injected into the current mixing chamber and mixed with the main drug solution and various additives, so that a specific batch of droplets released by the nozzle has fluorescent luminescence properties.

[0037] Specifically, the system's main control unit sends an instruction file containing the instantaneous mixing ratio control data to the control board of the microfluidic mixing device installed at the end of the agricultural drone's nozzle via the CAN bus. Upon receiving the instruction file, the control board converts the mass percentage values ​​into corresponding PWM drive signals. Different duty cycles of the PWM drive signals correspond to the required flow rates of the main pesticide, anti-drift adjuvant, and penetration adjuvant, respectively.

[0038] The microfluidic mixing device contains independent inlet microchannels connected to the main drug tank, the anti-drift agent tank, and the penetration agent tank. Each inlet microchannel is equipped with a microchannel valve made of piezoelectric material. The control board applies generated PWM drive signals to the piezoelectric actuators of the corresponding microchannel valves. The piezoelectric actuators generate corresponding mechanical deformations based on the duty cycle of the received PWM drive signals, thereby controlling the precise opening of the microchannel valves.

[0039] The system uses a predefined fluid dynamics formula to calculate the mapping relationship between the microchannel valve opening and the fluid flow rate: In the formula, Indicates the first The volumetric flow rate of a liquid component; This represents the inherent flow coefficient of the microchannel valve; Indicates the deformation angle of the piezoelectric actuator The effective flow cross-sectional area of ​​the controlled microchannel valve; This indicates a constant pressure difference between the inlet channel and the mixing chamber; Indicates the first The density of a liquid component.

[0040] Through precise control The main drug solution, anti-drift agent, and penetration agent are injected simultaneously into the mixing chamber located at the end of the nozzle according to the calculated instantaneous mixing ratio.

[0041] Simultaneously with the main drug solution and various adjuvants entering the mixing chamber at the nozzle tip, the system activates the ultrasonic transducer located on the outer wall of the mixing chamber. The ultrasonic transducer receives high-frequency alternating current signals, generating high-frequency mechanical vibrations, and transmitting mechanical energy in the form of ultrasonic waves to the fluid medium within the mixing chamber. The ultrasonic waves generate cavitation and acoustic flow effects in the fluid, forcing the main drug solution, anti-drift adjuvant, and penetration adjuvant to complete in-situ mixing at the molecular level within microseconds, forming a homogeneous mixed drug solution.

[0042] During the in-situ mixing process described above, the system triggers a fluorescence tracer injection procedure at preset time intervals. The microfluidic mixing device has an independent fluorescence tracer channel, which is connected to a micro reservoir containing a fluorescence tracer agent.

[0043] When the preset time interval is reached, the control board sends a fixed-length pulse signal to the miniature electromagnetic pump in the fluorescence tracer channel. Based on this fixed-length pulse signal, the miniature electromagnetic pump activates a preset-length activation cycle, injecting a fixed, minute amount of fluorescence tracer additive into the currently ultrasonically vibrating mixing chamber in a single pass.

[0044] Under the influence of ultrasound, the injected fluorescent tracer rapidly and uniformly fuses with the mixed drug solution in the current chamber. Subsequently, this batch of mixed drug solution, containing the main drug solution, proportioning agent, and fluorescent tracer, is mechanically broken into fine droplets and sprayed out through a centrifugal atomizing disc or high-pressure nozzle at the end of the nozzle. Due to the addition of the fluorescent tracer, specific batches of droplets released from the nozzle exhibit fluorescent luminescence properties under illumination of a specific wavelength of light, while batches without the fluorescent tracer retain their normal physical appearance. This process is periodically repeated at time intervals, enabling the periodic release of specific batches of optically marked droplets during continuous atomization spraying operations.

[0045] Step S4: Use the multispectral sensor on the agricultural drone to obtain the actual settling trajectory of the droplets with fluorescent tracer after atomization spraying in the real environment. In step S4, the step of using the multispectral sensor onboard the agricultural drone to obtain the actual settling trajectory of the droplets carrying the fluorescent tracer after atomized spraying in the real environment includes: Enable the capture mode in the multispectral sensor that matches the wavelength of the fluorescent tracer; By continuously extracting the centroid positions of bright droplets with fluorescent luminescence properties in three-dimensional space through image recognition, the centroid positions are connected in chronological order to generate the actual trajectory centroid coordinate sequence as the actual sedimentation trajectory.

[0046] Specifically, the system sends control commands to a multispectral sensor mounted on the underside of the agricultural drone. The multispectral sensor contains multiple independent optical channels, each equipped with a narrowband filter with a specific center wavelength. Based on the physicoluminescence properties of the fluorescent tracer injected in step S3, the system selectively activates a specific optical channel in the multispectral sensor that matches the emission spectrum of the fluorescent tracer, entering a dedicated capture mode. For example, when Rhodamine B is used as the fluorescent tracer, the system activates the receiving channel with a center wavelength in the 580nm to 600nm range, while simultaneously deactivating other visible and near-infrared channels used for conventional crop canopy reflectance detection to filter out ambient light interference.

[0047] In capture mode, the multispectral sensor continuously captures images of the atomized spraying area below the nozzle of the agricultural drone at a set frame rate. The sensor converts the captured continuous optical signals into a digital image sequence and inputs it to the onboard image processing unit. The onboard image processing unit receives the digital images frame by frame and performs image recognition and feature extraction operations.

[0048] The airborne image processing unit performs grayscale and binarization on each frame of the digital image, sets a fixed grayscale threshold, filters out background pixels below the threshold, and retains bright pixel connected components with pixel values ​​greater than or equal to the threshold. The system identifies these bright pixel connected components as specific batches of fog droplets with fluorescent emission characteristics.

[0049] For each identified bright fog droplet cluster pixel connected region, the onboard image processing unit uses the image moment algorithm to calculate its two-dimensional centroid coordinates. Two-dimensional centroid coordinates ( , The formula for calculating ) is: In the formula, Indicates the coordinates in the image as ( , The grayscale value of the pixel; Represents the zeroth moment of a connected region; and Let represent the first moments of the connected domain.

[0050] After obtaining the two-dimensional centroid coordinates, the onboard image processing unit, in conjunction with the depth information matrix provided by the binocular visual ranging module or lidar onboard the agricultural drone, processes the two-dimensional centroid coordinates ( , The back projection is then applied to a three-dimensional coordinate system with the center of the agricultural drone as the origin. The back projection calculation formula is as follows: In the formula, Represents the three-dimensional coordinates of the centroid in a three-dimensional spatial coordinate system; This represents the depth value of the corresponding pixel extracted from the depth information matrix; and These represent the equivalent focal lengths of the multispectral sensor camera in the horizontal and vertical directions, respectively. and This represents the pixel coordinates of the camera principal point in the image coordinate system.

[0051] The airborne image processing unit extracts and calculates the three-dimensional coordinates of the centroid from each consecutive frame of the image and adds a corresponding timestamp. The system then connects these three-dimensional centroid coordinates in chronological order of the timestamps to generate a time-series coordinate set. This coordinate set is the actual trajectory centroid coordinate sequence, which the system uses as the actual sedimentation trajectory of a specific batch of droplets carrying fluorescent tracers in a real environment.

[0052] Step S5: Compare the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss. When the trajectory deviation loss exceeds the preset calibration threshold, use the trajectory deviation loss to update the closed-loop parameters of the real-time aerodynamic flow field model.

[0053] In step S5, the step of comparing the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss includes: Extract the predicted trajectory coordinate sequence of a specific batch of droplets containing fluorescent tracers; Align the predicted trajectory coordinate sequence with the actual trajectory centroid coordinate sequence on the time axis; The trajectory deviation loss is obtained by calculating the square integral of the spatial distance difference between the predicted centroid position and the actual centroid position at the same time node. In step S5, the steps of updating the closed-loop parameters of the real-time aerodynamic flow field model using trajectory deviation loss include: When the trajectory deviation loss exceeds the preset calibration threshold, the gradient descent algorithm is used to calculate the residual of the trajectory deviation loss relative to the environmental wind field parameters in the real-time aerodynamic flow field model. The residual is superimposed as a feedback compensation value into the flow field calculation weight of the previous cycle to complete the correction and calibration of the real-time aerodynamic flow field model.

[0054] Specifically, the system extracts the predicted trajectory coordinate sequence generated in step S2, which corresponds to a specific batch of droplets injected with fluorescent tracer in step S3. Simultaneously, the system extracts the actual trajectory centroid coordinate sequence generated in step S4. Because the sampling frequencies of the predicted calculations and the actual image captures differ, the system first aligns the two sets of coordinate sequences along the time axis.

[0055] The system uses the timestamps of the actual trajectory centroid coordinate sequence. Using a baseline time point, a linear interpolation algorithm is employed to resample the predicted trajectory coordinate sequence. For each baseline time point... The system finds two adjacent original timestamps in the predicted trajectory coordinate sequence. and ,in ≤ < Calculate the coordinates of the predicted centroid after alignment. : Through this resampling process, the system obtains a time-aligned predicted trajectory coordinate sequence that perfectly corresponds to the actual time points.

[0056] The system iterates through the aligned time nodes and calculates the number of identical time nodes. Below, the predicted centroid position in the time-aligned predicted trajectory coordinate sequence. The actual centroid position in the actual trajectory centroid coordinate sequence The system calculates the spatial Euclidean distance difference between the time points by performing a square integral on the spatial Euclidean distance difference at all time points, resulting in the trajectory deviation loss used to quantify the degree of trajectory deviation. : In the formula, Indicates the total number of time points; Indicates the time interval between adjacent time points; This represents the Euclidean distance norm in three-dimensional space.

[0057] The system has a fixed preset calibration threshold. This threshold is calculated from the maximum permissible spraying error range for agricultural drones. The system will then calculate the trajectory deviation loss. Compared with the preset calibration threshold Compare and judge.

[0058] When trajectory deviation loss Less than or equal to the preset calibration threshold If the system determines that the prediction accuracy of the current real-time aerodynamic flow field model meets the drug application requirements, it will not perform a model update operation.

[0059] When trajectory deviation loss Greater than the preset calibration threshold At this time, the system triggers the closed-loop parameter update module. The closed-loop parameter update module aims to minimize trajectory deviation loss. Let the objective function be , and calculate using the gradient descent algorithm. The torque of the external environment source term relative to the real-time aerodynamic flow field model in step S1 Environmental wind field parameters, including environmental wind speed components , , The residual gradient.

[0060] by Axial wind speed component For example, the formula for calculating the residual gradient is: After calculating the residual gradients in the three directions, the system uses a preset learning rate. The residual gradient is superimposed as a feedback compensation value onto the current set values ​​of the environmental wind field parameters in the real-time aerodynamic flow field model to calculate new corrected values ​​for the environmental wind field parameters: The system inputs the updated environmental wind field parameter correction values ​​into the momentum control equation in step S1, as the flow field boundary condition weights for the next calculation cycle. Through this process, the parameter correction and physical calibration of the real-time aerodynamic flow field model are completed.

[0061] Example 2: When applying pesticides in hilly orchards, extreme microclimate challenges are faced. The thick, uneven canopy of the fruit trees, combined with the drone's downwash airflow impacting the irregular ground and canopy, creates strong rebound updrafts and shear vortices. Sudden gusts of wind are also common in mountainous environments. To address these issues, this invention provides a pesticide adjuvant mixing and spraying control system for agricultural drones, the structure of which is as follows: Figure 2 As shown. The specific implementation process of this system is as follows: The flow field modeling module is used to acquire flight status parameters of agricultural drones and three-dimensional topographic data of the crop canopy below, and to construct a real-time aerodynamic flow field model superimposed with the downwash airflow of the agricultural drone rotor and the environmental wind field. The proportioning calculation module is used to predict the sedimentation trajectory of drug droplets based on the real-time aerodynamic flow field model, calculate the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generate the instantaneous mixing ratio of the main drug and various adjuvants. The mixing and spraying module is used to send the instantaneous mixing ratio to the microfluidic mixing device installed at the end of the nozzle of the plant protection drone. Inside the microfluidic mixing device, the main drug solution and various adjuvants are mixed in situ according to the instantaneous mixing ratio. During the mixing process, fluorescent tracer adjuvants are periodically injected and then atomized and sprayed. The trajectory acquisition module is used to acquire the actual settling trajectory of droplets with fluorescent tracer after atomization spraying in the real environment using the multispectral sensor carried by the agricultural drone. The model update module is used to compare the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss. When the trajectory deviation loss exceeds the preset calibration threshold, the closed-loop parameters of the real-time aerodynamic flow field model are updated using the trajectory deviation loss.

[0062] Specifically, the flow field modeling module is physically connected to the airborne positioning unit, inertial measurement unit, electronic speed controller, and lidar under the fuselage to acquire the three-dimensional spatial position, flight attitude angle, rotor speed parameters, and three-dimensional topographic data of the crop canopy below the agricultural drone. This module integrates a fluid dynamics solver to execute fluid dynamics calculation equations. In the calculation equation, It is the air density constant; The velocity vector of the flow field; It is a time variable; It is a static pressure distribution; Aerodynamic viscosity; For rotor momentum source term; This module represents the environmental wind field source term. It outputs a real-time aerodynamic flow field model containing the velocity vectors of the three-dimensional Eulerian grid nodes to the sizing solution module.

[0063] The proportioning calculation module receives a real-time aerodynamic flow field model via a data bus. This module uses a Lagrange particle tracking model to calculate the trajectory of a single drug droplet under stress, outputting a predicted trajectory coordinate sequence. Based on this coordinate sequence, the proportioning calculation module identifies airflow disturbance regions and calculates the maximum relative velocity component within these regions. Combining this with a preset critical Weber number and target Reynolds number, it calculates the surface tension and dynamic viscosity required for the droplet to reach the target canopy. Subsequently, the module reads the physicochemical parameters of the drug and various adjuvants from the storage unit, solves a set of mixing characteristic equations to generate the instantaneous mixing ratio of the main drug, anti-drift adjuvant, and penetration adjuvant, and converts this ratio data into a digital instruction file for distribution.

[0064] The mixing and spraying module includes a microfluidic mixing device installed at the end of the nozzle of an agricultural drone and a corresponding embedded control board. The control board receives a digital instruction file of the instantaneous mixing ratio and parses it into multi-channel PWM drive signals. The microfluidic mixing device contains a piezoelectric actuator and a microchannel valve. The piezoelectric actuator adjusts the valve opening according to the PWM signal, proportionally controlling the inflow rate of the main pesticide and various adjuvants. An ultrasonic transducer is fixed to the outer wall of the mixing chamber inside the microfluidic mixing device for ultrasonic oscillation mixing of the injected liquid medium. Simultaneously, the microfluidic mixing device integrates a fluorescent tracer channel and a micro-electromagnetic pump. The control board outputs control levels at preset time intervals, driving the micro-electromagnetic pump to inject a fixed trace amount of fluorescent tracer adjuvant into the mixing chamber, so that specific batches of droplets sprayed by the atomizing device after mixing possess corresponding fluorescent luminescence characteristics.

[0065] The trajectory acquisition module is connected to the multispectral sensor and image processing unit mounted on the agricultural drone. The multispectral sensor switches to a narrow-band filter channel that matches the emission spectrum of the fluorescent tracer, continuously capturing image signals of the sprayed area below at a preset frame rate. The image processing unit receives the digital image sequence, extracts the connected components of high-brightness pixels through binarization, and calculates their two-dimensional centroid coordinates using an image moment algorithm. , Combined with depth information provided by the airborne ranging module. This module obtains three-dimensional spatial coordinates through back projection calculation: , In the formula , For the camera's equivalent focal length, , These are the pixel coordinates of the camera's principal point. This module outputs the actual trajectory centroid coordinate sequence as the actual settlement trajectory, ordered by timestamps.

[0066] The model update module receives the predicted trajectory coordinate sequence output by the ratio calculation module and the actual settlement trajectory output by the trajectory acquisition module. This module uses a linear interpolation algorithm to align the two sequences along the time axis and calculates the square integral of the spatial distance difference at the same time point to obtain the trajectory deviation loss value. When the trajectory deviation loss value exceeds the system's preset calibration threshold, the model update module initiates a closed-loop correction procedure, using a gradient descent algorithm to calculate the residual gradient of the trajectory deviation loss relative to the environmental wind speed component in the real-time aerodynamic flow field model. This module uses the residual gradient as a compensation weight to superimpose it onto the flow field source term settings, generating new environmental wind field parameters that are fed back to the flow field modeling module. This feedback mechanism eliminates model prediction errors caused by mountain gusts not directly detected by sensors, achieving closed-loop parameter updates for the pesticide application system.

[0067] Finally, it should be noted that the above description is only 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 foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing 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 the formulation and spraying control of pesticide adjuvants by agricultural drones, characterized in that, Includes the following steps: Step S1: Obtain flight status parameters of the agricultural drone and three-dimensional topographic data of the crop canopy below, and construct a real-time aerodynamic flow field model that superimposes the downwash airflow of the agricultural drone rotor and the environmental wind field. Step S2: Predict the sedimentation trajectory of the drug droplets based on the real-time aerodynamic flow field model, calculate the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generate the instantaneous mixing ratio containing the main drug and various adjuvants. Step S3: Send the instantaneous mixing ratio to the microfluidic mixing device installed at the end of the nozzle of the plant protection drone. Inside the microfluidic mixing device, the main drug solution and various adjuvants are mixed in situ according to the instantaneous mixing ratio. During the mixing process, fluorescent tracer adjuvants are periodically injected and then atomized and sprayed. Step S4: Use the multispectral sensor carried by the agricultural drone to obtain the actual settling trajectory of the droplets with the fluorescent tracer after atomization spraying in the real environment. Step S5: Compare the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss. When the trajectory deviation loss exceeds the preset calibration threshold, use the trajectory deviation loss to update the closed-loop parameters of the real-time aerodynamic flow field model.

2. The method for the formulation and spraying control of pesticide adjuvants by a plant protection drone according to claim 1, characterized in that, In step S1, the step of constructing a real-time aerodynamic flow field model that superimposes the rotor downwash airflow of the agricultural drone with the environmental wind field includes: The spatial position, flight attitude angle, and rotor speed of the agricultural drone are obtained as boundary conditions. Based on the three-dimensional topography data of the crop canopy, the superposition effect of rotor torque and air viscosity in the spatial grid is calculated using fluid dynamics calculation methods, and the gridded flow field velocity vector is output as the real-time aerodynamic flow field model.

3. The method for the formulation and spraying control of pesticide adjuvants by a plant protection drone according to claim 1, characterized in that, In step S2, the step of predicting the sedimentation trajectory of the liquid droplets based on the real-time aerodynamic flow field model includes: Extract the flow field velocity vector from the real-time aerodynamic flow field model; By combining the preset initial mass of a single droplet with the aerodynamic drag coefficient, the force state of the liquid droplet in space is calculated using the dynamic force relationship, and a predicted trajectory coordinate sequence in three-dimensional space is generated as the sedimentation trajectory.

4. The method for mixing and controlling the application of pesticide adjuvants by a plant protection drone according to claim 3, characterized in that, In step S2, the step of calculating the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generating the instantaneous mixing ratio of the main drug solution and various adjuvants, includes: Identify airflow disturbance regions present in the predicted trajectory coordinate sequence; The surface tension and dynamic viscosity required to resist airflow deviation are calculated based on the disturbance intensity of the airflow disturbance region. The instantaneous mixing ratio is generated by distributing the required amounts of the main drug solution, anti-drift agent, and penetration agent in reverse order of surface tension and dynamic viscosity.

5. The method for the formulation and spraying control of pesticide adjuvants by a plant protection drone according to claim 1, characterized in that, In step S3, the step of in-situ mixing of the main drug solution and various adjuvants according to the instantaneous mixing ratio inside the microfluidic mixing device includes: The instantaneous mixing ratio is converted into a driving electrical signal; The microchannel valve of the piezoelectric actuator inside the microfluidic mixing device is controlled by the driving electrical signal to mix the main drug solution, anti-drift agent and penetration agent by ultrasonic vibration in the nozzle end chamber.

6. The method for mixing and controlling the application of pesticide adjuvants by a plant protection drone according to claim 1, characterized in that, In step S3, the step of periodically injecting the fluorescent tracer during the mixing process followed by atomized spraying includes: The fluorescent tracer channel inside the microfluidic mixing device is activated at preset time intervals. A fixed tracer amount of the fluorescent tracer is injected into the current mixing chamber and mixed with the main drug solution and various additives, so that a specific batch of droplets released by the nozzle has fluorescent luminescence properties.

7. The method for mixing and controlling the application of pesticide adjuvants by a plant protection drone according to claim 6, characterized in that, In step S4, the step of using the multispectral sensor mounted on the agricultural drone to obtain the actual settling trajectory of the droplets carrying the fluorescent tracer in the real environment after atomized spraying includes: Activate the capture mode in the multispectral sensor that matches the wavelength of the fluorescent tracer; The centroid positions of high-brightness droplets with the fluorescence emission characteristics are continuously extracted in three-dimensional space by image recognition, and the centroid positions are connected in chronological order to generate an actual trajectory centroid coordinate sequence as the actual sedimentation trajectory.

8. The method for mixing and controlling the application of pesticide adjuvants by a plant protection drone according to claim 7, characterized in that, In step S5, the step of comparing the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss includes: Extract the predicted trajectory coordinate sequence corresponding to a specific batch of droplets containing the fluorescent tracer; Align the predicted trajectory coordinate sequence with the actual trajectory centroid coordinate sequence on the time axis; The trajectory deviation loss is obtained by calculating the square integral of the spatial distance difference between the predicted centroid position and the actual centroid position at the same time node.

9. The method for the formulation and spraying control of pesticide adjuvants by a plant protection drone according to claim 1, characterized in that, In step S5, the step of updating the closed-loop parameters of the real-time aerodynamic flow field model using the trajectory deviation loss includes: When the trajectory deviation loss is greater than the preset calibration threshold, the gradient descent algorithm is used to calculate the residual of the trajectory deviation loss relative to the environmental wind field parameters in the real-time aerodynamic flow field model; The residual is added as a feedback compensation value to the flow field calculation weight of the previous cycle to complete the correction and calibration of the real-time aerodynamic flow field model.

10. A control system for the mixing and spraying of pesticide adjuvants by a crop protection drone, characterized in that, The system is used for the method of mixing and spraying chemical adjuvants for agricultural protection drones according to any one of claims 1-9, the system comprising: The flow field modeling module is used to acquire flight status parameters of agricultural drones and three-dimensional topographic data of the crop canopy below, and to construct a real-time aerodynamic flow field model superimposed with the downwash airflow of the agricultural drone rotor and the environmental wind field. The proportioning calculation module is used to predict the sedimentation trajectory of the drug droplets based on the real-time aerodynamic flow field model, calculate the surface tension and dynamic viscosity required for the droplets to reach the target canopy based on the sedimentation trajectory, and generate the instantaneous mixing ratio including the main drug and various adjuvants. The mixing and spraying module is used to send the instantaneous mixing ratio to a microfluidic mixing device installed at the end of the nozzle of the plant protection drone. Inside the microfluidic mixing device, the main drug solution and various adjuvants are mixed in situ according to the instantaneous mixing ratio. During the mixing process, fluorescent tracer adjuvants are periodically injected and then atomized and sprayed. The trajectory acquisition module is used to acquire the actual settling trajectory of droplets carrying the fluorescent tracer after atomized spraying in the real environment using the multispectral sensor carried by the agricultural drone. The model update module is used to compare the settlement trajectory predicted by the real-time aerodynamic flow field model with the actual settlement trajectory to calculate the trajectory deviation loss. When the trajectory deviation loss exceeds a preset calibration threshold, the closed-loop parameters of the real-time aerodynamic flow field model are updated using the trajectory deviation loss.

Images