An intelligent control method and system for spraying of an agricultural protection unmanned aerial vehicle without a flowmeter

Abstract

The application provides a plant protection unmanned aerial vehicle flowmeter-free spraying intelligent control method and system, and belongs to the unmanned aerial vehicle spraying control field. The method comprises the following steps: in the static stage of the plant protection unmanned aerial vehicle, calibrating the medicine tank of the plant protection unmanned aerial vehicle according to a pre-constructed static flow calibration model to obtain an initial apparent reading; in the dynamic stage of the plant protection unmanned aerial vehicle, calculating a target spraying flow in a current control period according to a second state vector; inputting a first state vector in the current control period into an onboard intelligent model to obtain a predicted spraying flow; iteratively optimizing the predicted spraying flow and the target spraying flow to obtain an optimal pulse width modulation signal value in the current control period; and controlling the spraying unit to output the real spraying flow in the current control period according to the optimal pulse width modulation signal value. The application inversely solves the optimal control instruction required to reach the target flow, so that the plant protection unmanned aerial vehicle can be stably and accurately operated for a long time in a variable environment.

Description

Technical Field

[0001] This application relates to the field of drone spraying control, and in particular to a method and system for intelligent control of plant protection drone spraying without flow meter. Background Technology

[0002] As a key piece of equipment in modern agriculture, agricultural drones have incorporated physical flow meters in order to achieve closed-loop control. However, flow meters themselves have physical limitations such as wear, aging, and susceptibility to the properties of liquids. As a result, their measurement accuracy decreases over time and becomes unreliable. Ultimately, this leads to a decrease in spraying accuracy due to closed-loop control based on inaccurate feedback. Furthermore, the flow meters add extra weight, cost, and the risk of failure. Summary of the Invention

[0003] To address the aforementioned technical problems, this application provides an intelligent control system for agricultural drone spraying without a flow meter.

[0004] In a first aspect, embodiments of this application provide a method for intelligent control of agricultural drone spraying without a flow meter, the method comprising:

[0005] During the static phase of the agricultural drone, the tank of the agricultural drone is calibrated according to the pre-built static flow calibration model to obtain the initial apparent weight reading.

[0006] In the dynamic phase of the agricultural drone, the first state vector of the agricultural drone in the current control cycle is obtained. The first state vector includes the pulse width modulation signal value in the previous control cycle and the battery voltage, the initial apparent weight reading, the vertical acceleration of the drone, the roll angle and the pitch angle in the current control cycle.

[0007] The second state vector of the agricultural drone under the current control cycle is obtained. The second state vector includes the current ground speed, the user-set amount per acre and the spray width. The target spray flow rate under the current control cycle is calculated based on the second state vector.

[0008] The first state vector under the current control cycle is input into the airborne intelligent model to obtain the estimated spray flow rate;

[0009] Based on the predicted spray flow rate and the target spray flow rate, the optimal pulse width modulation signal value for the current control cycle is obtained through iterative optimization.

[0010] The spraying unit is controlled to output the actual spraying flow rate under the current control cycle based on the optimal pulse width modulation signal value.

[0011] Secondly, embodiments of this application provide a flow meter-free intelligent control system for agricultural drone spraying, the flow meter-free intelligent control system for agricultural drone spraying includes:

[0012] The main control unit is used to execute algorithm programs and control logic;

[0013] Weighing unit, used to acquire the weight signal of the medicine box in real time;

[0014] The spraying unit is used to receive the pulse width modulation signal value sent by the main control unit and control the nozzle to spray liquid through the pulse width modulation signal value;

[0015] The flight status unit is used to acquire the vertical acceleration, roll angle, and pitch angle of the agricultural drone.

[0016] The power management unit is used to monitor the battery voltage in real time during each control cycle.

[0017] Data storage unit, used to store model parameters and model structure of airborne intelligent model and static flow calibration model;

[0018] The human-machine interface unit is used to remotely receive control signals from operators.

[0019] Thirdly, embodiments of this application provide an agricultural drone, including a memory and a processor. The memory is used to store a computer program, and the computer program executes the intelligent control method for water-meter-less spraying of agricultural drones provided in the first aspect when the processor is running.

[0020] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when run on a processor, executes the intelligent control method for water spraying without a flow meter provided in the first aspect for agricultural drones.

[0021] The above-described method, provided in this application, involves calibrating the pesticide tank of an agricultural drone during its static phase using a pre-built static flow calibration model to obtain an initial apparent weight reading. During the dynamic phase, a first state vector of the agricultural drone in the current control cycle is acquired. This first state vector includes the pulse width modulation signal value from the previous control cycle, the battery voltage, the initial apparent weight reading, the vertical acceleration, roll angle, and pitch angle of the current control cycle. A second state vector of the agricultural drone in the current control cycle is also acquired. This second state vector includes the current ground velocity, the user-set dosage per acre, and the spray width. The target spray flow rate for the current control cycle is calculated based on the second state vector. The first state vector of the current control cycle is input into an onboard intelligent model to obtain an estimated spray flow rate. Iterative optimization is performed based on the predicted spray flow rate and the target spray flow rate to obtain the optimal pulse width modulation signal value for the current control cycle. The spraying unit is then controlled to output the actual spray flow rate for the current control cycle based on the optimal pulse width modulation signal value. This application directly solves the optimal control command required to achieve the target flow rate by real-time fusion of multi-dimensional dynamic information, including body acceleration and attitude angle, and rapid iteration of pulse width modulation signal values, enabling agricultural drones to operate stably and accurately in changing environments for a long time. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be considered as a limitation on the scope of protection of this application. In the various drawings, similar components are numbered similarly.

[0023] Figure 1 This paper shows a flowchart of an intelligent control method for water spraying without a flow meter for agricultural drones provided in an embodiment of this application.

[0024] Figure 2 This paper shows another schematic flowchart of the intelligent control method for water spraying without a flow meter provided in an embodiment of this application for agricultural drones.

[0025] Figure 3 This paper illustrates another flowchart of the intelligent control method for water spraying without a flow meter provided in an embodiment of this application for agricultural drones.

[0026] Figure 4 This paper illustrates another flowchart of the intelligent control method for water spraying without a flow meter provided in an embodiment of this application for agricultural drones.

[0027] Figure 5 This paper shows a schematic diagram of a flow meter-free intelligent control system for agricultural drone spraying provided in an embodiment of this application.

[0028] Icons: 500 - Intelligent Control System for Plant Protection Drone Spraying without Flow Meter, 501 - Main Control Unit, 502 - Weighing Unit, 503 - Spraying Unit, 504 - Flight Status Unit, 505 - Power Management Unit, 506 - Data Storage Unit, 507 - Human-Machine Interaction Unit. Detailed Implementation

[0029] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0030] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0031] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0032] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0033] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0034] Example 1

[0035] This application provides an intelligent control method for spraying agricultural drones without a flow meter.

[0036] See Figure 1 The intelligent control method for spraying agricultural drones without flow meters includes steps S101-S106.

[0037] S101: During the static phase of the agricultural drone, the tank of the agricultural drone is calibrated according to the pre-built static flow calibration model to obtain the initial apparent weight reading.

[0038] In this embodiment, the intelligent control system 500 for agricultural drone spraying without a flow meter includes a main control unit 501, a weighing unit 502, a spraying unit 503, a flight status unit 504, a power management unit 505, a data storage unit 506, and a human-machine interaction unit 507. After receiving instructions from the ground station, the main control unit 501 automatically executes a static flow calibration model to calibrate the weight of the pesticide tank under static conditions, obtaining an initial apparent weight reading, which serves as the basis for the airborne intelligent model during the subsequent dynamic phase in the air. Provide reliable baseline data.

[0039] See Figure 2 In one embodiment, constructing a static flow calibration model includes steps S1011-S1015:

[0040] S1011: Under static conditions, after prompting the user to add standard liquid, obtain the initial stable weight of the medicine tank.

[0041] In this embodiment, the intelligent control system 500 for spraying agricultural drones without flowmeters is in a static preparation phase. The main control unit 501 issues a command to the user to add standard liquid. After the user adds the standard liquid to the tank as prompted, the weighing unit 502 initiates a short waiting period. Once it determines that the liquid has reached a stable state, it samples the readings of the weighing unit 502 multiple times and processes these sampled data using an extremum filtering algorithm to obtain the initial stable weight of the tank. .

[0042] S1012: Take out pulse width modulation signal values ​​from the preset pulse width modulation signal value list in sequence to test the water spray volume of the medicine tank, and obtain the final stable weight of the medicine tank corresponding to each pulse width modulation signal value.

[0043] In this embodiment, the data storage unit 506 of the plant protection drone spraying intelligent control system 500 stores a preset pulse width modulation signal value list in advance, such as [1300, 1450, 1600, 1800, 1950] (unit: microseconds). Different PWM values ​​correspond to different operating speeds of the water pump, thereby controlling different spraying flow rates of the spraying unit 503.

[0044] Optionally, retrieve from the list one by one. The system is tested to a specific value, for example, by starting the precise 12-second spray timer. During the spray timer, the water pump continues to operate according to the current value. The system operates by spraying liquid from the medicine tank. Once the timer expires, the pump immediately stops, and the system waits for a short period (e.g., 2 seconds) for the liquid to completely stop flowing. The final stable weight of the medicine tank is then obtained using the method described in S1031. .

[0045] S1013: Calculate the actual spraying flow rate under static conditions based on the final stable weight of the medicine tank and the initial stable weight of the medicine tank.

[0046] In this embodiment, after obtaining the initial stable weight of the medicine box... The final stable weight of the medicine box corresponding to each PWM value Subsequently, the intelligent control system 500 for spraying agricultural drones without flow meters can calculate the weight difference. The weight difference This reflects the reduction in liquid weight in the tank after T seconds of spraying at a specific PWM value.

[0047] Furthermore, to convert the weight difference into the actual spray flow rate (unit: liters / minute, L / min), the following calculation is performed. Since the unit of weight is usually grams (g), while the unit of flow rate is required to be liters / minute, the weight difference needs to be converted from grams to kilograms (kg), i.e. kilograms; then, considering that the spraying time is T seconds, it needs to be converted to minutes, i.e. Minutes. Finally, actual traffic. The calculation formula is: For example, if If T = 12 seconds, then Through this calculation, the system can accurately obtain the actual spray flow rate of each PWM value under static conditions, providing key output label data for the subsequent training of the static flow calibration model.

[0048] S1014: Correlate the pulse width modulation signal value, average battery voltage, average tank weight with the corresponding actual spraying flow rate to construct a static operating condition training dataset.

[0049] In this embodiment, after completing the spray test for each PWM value and calculating the actual spray flow rate... The system will then record the average battery voltage during the test. and average medicine box weight The average battery voltage is recorded because fluctuations in battery voltage can affect the pump's operating efficiency, thus impacting the spray flow rate. During each PWM value test, the system samples the battery voltage multiple times using a voltage sensor and calculates the average value for each PWM value. Average medicine box weight The recording takes into account that the remaining liquid in the tank may affect the spraying pressure and flow rate. During the spraying process, the system will sample the reading of the weighing unit 502 in a timely manner and calculate the average value under each PWM value. .

[0050] Furthermore, As input features, As output labels, a static working condition training dataset containing multiple training samples is constructed by sequentially processing all PWM values ​​in the preset pulse width modulation signal value list.

[0051] S1015: Train the static flow calibration model based on the static operating condition training dataset.

[0052] In this embodiment, after constructing the static operating condition training dataset, a static flow calibration model is trained. Common model choices include neural network models and piecewise linear interpolation models. Regardless of the model chosen for training, the trained model can accurately predict the actual spraying flow rate based on the input PWM value, average battery voltage, and average tank weight.

[0053] Finally, the system will write the parameters of the trained static flow calibration model into the data storage unit 506. This model will serve as the airborne intelligent model in subsequent in-flight dynamic intelligent control. The initial state or reference baseline provides an important basis for flow calibration for precise spraying by agricultural drones.

[0054] S102: During the dynamic phase of the agricultural drone, obtain the first state vector of the agricultural drone in the current control cycle. The first state vector includes the pulse width modulation signal value in the previous control cycle and the battery voltage, the initial apparent weight reading, the vertical acceleration of the drone, the roll angle, and the pitch angle in the current control cycle.

[0055] In this embodiment, during the dynamic operating condition phase, at the beginning of each control cycle, the main control unit 501 first runs a data acquisition program to obtain the pulse width modulation signal value from the previous control cycle. During the first control cycle, the initial pulse width modulation signal value of the agricultural drone is set. Then, the battery voltage under the current control cycle is obtained through the power management unit 505. The system monitors the battery's operating status; then, the weight of the pesticide tank on the agricultural drone is measured in real time using the weighing unit 502 to obtain an initial apparent weight reading. The Inertial Measurement Unit (IMU) sensor in Flight Status Unit 504 measures the vertical acceleration of the aircraft. Roll angle Pitch angle These three data points can reflect the changes in the UAV's flight attitude and acceleration, and the measured data are integrated into a state vector for the current control cycle. This provides input for subsequent airborne intelligent models to predict spray flow rates.

[0056] S103: Obtain the second state vector of the agricultural drone in the current control cycle. The second state vector includes the current ground speed, the user-set application rate per acre, and the spray width. Calculate the target spray flow rate in the current control cycle based on the second state vector.

[0057] In this embodiment, the main control unit 501 obtains the current ground speed of the UAV through the GPS module in the flight status unit 504. At the same time, the user-defined usage per acre is obtained from the human-computer interaction unit 507. and spray width The target spray flow rate under the current control cycle is calculated using a preset formula. .

[0058] S104: Input the first state vector under the current control cycle into the airborne intelligent model to obtain the estimated spray flow rate.

[0059] In this embodiment, during UAV flight operations, the main control unit 501 collects the current first state vector in real time during each control cycle. and the first state vector under the current control cycle will be obtained. As a trained airborne intelligent model The input vector is then used to call the model function library stored in the data storage unit 506. The model function library performs an efficient forward propagation calculation inside the main control unit 501 (MCU) and outputs a high-precision predicted flow rate that has been compensated for the flight dynamics in real time. .

[0060] It should be noted that: This predicted value is directly used as the real-time feedback signal for the subsequent adaptive spraying control algorithm. The process already includes filtering and decoupling of inertial forces, which is technically far more advanced and accurate than any traditional PID or post-filter.

[0061] See Figure 3 In one embodiment, prior to step S104, the method further includes:

[0062] S1041: Construct a multidimensional nonlinear function model based on the first state vector of each control cycle.

[0063] In this embodiment, in order to accurately decouple the actual spray flow rate from the sensor signals affected by flight dynamics, a multidimensional nonlinear function is first constructed based on the first state vector of each control cycle. ,in The internal parameter set of the constructed multidimensional nonlinear function model is used; by capturing the complex nonlinear relationship between the input vector and the actual spray flow rate, an inverse dynamic model is constructed from "contaminated sensor readings and flight attitude" to "clean actual flow rate", thereby decoupling the accurate value of the spray flow rate.

[0064] S1042: Calculate the actual flow rate based on the average weight of the agricultural drone under each control cycle, obtain the actual flow rate label, and construct a dynamic working condition training dataset based on the first state vector of each control cycle and the actual flow rate label.

[0065] In this embodiment, in order to enable the airborne intelligent model It possesses dynamic decoupling capabilities. During the offline phase, it collects training data through a special "dynamic calibration flight maneuver" (e.g., including standardized acceleration, deceleration, and tilting / swaying). During this process, the system synchronously records the input vector at high frequency. The system calculates the actual flow rate of the agricultural drone under each control cycle based on weight changes and time intervals (control cycles), and uses this as a label for the actual flow rate. Together, we will construct a dynamic working condition training dataset. .

[0066] S1043: Train the multidimensional nonlinear function model based on the dynamic operating condition training set to obtain the airborne intelligent model.

[0067] In this embodiment, a feedforward neural network is used as the airborne intelligent model. The carrier, whose input layer is extended to receive the complete state vector. The training dataset for dynamic operating conditions was used. The supervised learning training mode utilizes backpropagation and gradient descent optimizers to iteratively update the model's internal parameters during training. The goal is to minimize the output of the airborne intelligent model, i.e., the predicted spray flow rate. Similar to real traffic labels loss function between By reaching the minimum value, and through multiple iterations of training, the performance of the model is continuously optimized, so that the error between the model's output, the predicted spray flow rate and the actual flow rate label gradually decreases until it reaches the threshold range.

[0068] In one embodiment, the airborne intelligent model is one of a feedforward neural network, support vector regression, gradient boosting decision tree, and radial basis function network.

[0069] In this embodiment, in the airborne intelligent traffic model In application scenarios, actual spray flow rates often exhibit complex nonlinear characteristics and are influenced by multiple variables. Among these, when the airborne intelligent flow model... When used as a feedforward neural network (MLP), it can fit arbitrarily complex nonlinear functions through the connection of multiple layers of neurons and the nonlinear transformation of activation functions, and can handle the complex mapping relationship between multivariable inputs and flow outputs very well; when used in airborne intelligent flow models... When using Support Vector Regression (SVR), it maps the input data to a high-dimensional feature space based on a kernel function, searching for an optimal hyperplane in the high-dimensional space to fit the data. It has excellent handling capabilities for nonlinear problems, and is particularly suitable for traffic prediction with small sample sizes; when used in airborne intelligent traffic models... Gradient Boosting Decision Tree (GBDT) improves model accuracy by iteratively constructing multiple decision trees and combining their predictions. It automatically handles interactions between features, effectively captures non-linear relationships, and its lightweight version reduces computational resource requirements while maintaining a certain level of accuracy, making it suitable for airborne environments. (For example, in airborne intelligent traffic models...) It is a radial basis function (RBF) network, which uses radial basis functions as activation functions for hidden layer neurons. It can map the input space to a high-dimensional hidden space, and perform linear partitioning of data in the hidden space, thereby realizing the modeling of nonlinear relationships. It is suitable for complex problems such as traffic forecasting.

[0070] Optionally, different models have their own characteristics in terms of complexity and interpretability. In some airborne application scenarios with high requirements for model interpretability, it may be necessary to choose a relatively simple and easy-to-understand model, such as GBDT related to decision trees. On the other hand, in scenarios with extremely high requirements for prediction accuracy and relatively low requirements for interpretability, a more complex but more predictive model, such as RBF network, can be chosen. Choosing a suitable model according to specific needs can achieve a balance between model complexity and interpretability.

[0071] In one embodiment, after obtaining the airborne intelligent model, the method further includes: converting the model structure and model parameters of the airborne intelligent model into a function library; and embedding the function library into the data storage unit of the plant protection drone's flight control firmware.

[0072] In this embodiment, after completing offline training and obtaining the optimal model parameters... Subsequently, to ensure the airborne intelligent model To ensure efficient and stable operation of agricultural drones in real-world environments, embedded model conversion and solidification are necessary. Specifically, this involves converting and solidifying the optimal model parameters obtained through offline training. The analysis, quantification, and transformation are performed using an embedded AI development toolchain (such as STM32Cube.AI). This step involves analyzing, quantifying, and transforming complex model structures and parameters. It is converted into a C code function library that can run efficiently on the main control unit 501 (MCU) and does not depend on the operating system, and then embedded into the data storage unit 506 of the flight control firmware.

[0073] S105: Iterative optimization is performed based on the predicted spray flow rate and the target spray flow rate to obtain the optimal pulse width modulation signal value for the current control cycle.

[0074] In this embodiment, based on the pulse width modulation signal value triggered by the agricultural drone spraying, the initial pulse width modulation signal value of the agricultural drone is first set. , will initialize Substitute the complete state vector at the current moment into the input of the airborne intelligent model. In the prediction, we obtain ,exist Add a PWM increment on top of the existing one The superimposed pulse width modulation signal value and the remaining state vectors are input into the airborne intelligent model. This makes its output approach the target spray flow rate. This allows us to obtain the optimal pulse width modulation signal value for the current control cycle.

[0075] See Figure 4 In one embodiment, step S105 includes steps S1051-S1055:

[0076] S1051: Input the first batch of data of the first state vector of the current control cycle into the airborne intelligent model to obtain the first predicted spray flow rate.

[0077] In this embodiment, the operating state of the agricultural drone typically does not change abruptly; the PWM value of the previous cycle... This approach provides a relatively close starting point for the current cycle, allowing the iteration process to converge to the optimal solution more quickly. Furthermore, this setting avoids the blind guessing involved from scratch, reduces the number of iterations, and improves solution efficiency. For example, in a stable flight and spraying operation scenario, if the PWM value of the previous cycle brought the spray flow rate close to the target, using it as the initial value for iteration may lead to the optimal solution after only a few adjustments. It should be noted that in the first batch of data, the initial pulse width modulation signal value... The first batch of data as the first state vector, i.e. ,Will Input to airborne intelligent model Perform a forward inference and obtain .

[0078] S1052: Calculate the error spray flow rate between the first predicted spray flow rate and the target spray flow rate.

[0079] In this embodiment, the error spray flow rate for the current control cycle is calculated. Error spray flow rate This reflects the difference between the predicted spray flow rate and the target spray flow rate. A positive value indicates that the predicted spray flow rate is less than the target spray flow rate, and the PWM value needs to be increased to improve the spray flow rate; conversely, if... If the value is negative, the PWM value needs to be reduced.

[0080] S1053: Increase the pulse width modulation signal value of the current cycle by a fixed increment to obtain the second batch of data of the first state vector of the current control cycle, and input the second batch of data into the airborne intelligent model to obtain the second predicted spray flow rate.

[0081] In this embodiment, the pulse width modulation signal value of the current period is used. Apply a very small known increment The second batch of data used to construct the first state vector The second batch of data collected Input into the airborne intelligent model A second model inference is performed to obtain... .

[0082] S1054: Calculate the local gradient of the airborne intelligent model with respect to the pulse width modulation signal value at the current operating point using the difference method, based on the first predicted spray flow rate and the second predicted spray flow rate;

[0083] In this embodiment, the airborne intelligent model is calculated using differential computation. Local gradient with respect to the PWM input at the current operating point: The differential gradient value of PWM represents the change in predicted spray flow rate for each unit change in PWM value around the current PWM value, and also reflects the characteristics of the airborne intelligent model. The sensitivity to PWM input provides an important basis for subsequent PWM correction.

[0084] S1055: Calculate the optimal pulse width modulation signal value for the current control cycle based on the local gradient and the error spray flow rate.

[0085] In this embodiment, based on the error calculated above... and gradient It can be directly calculated to eliminate The required PWM correction amount, i.e. This correction amount It can accurately determine the direction and magnitude of the PWM value that need to be adjusted based on the magnitude of the error and the model's sensitivity to PWM.

[0086] Furthermore, based on the pulse width modulation signal value of the previous control cycle... Obtain the final optimal PWM solution ,Right now ,this Under the current state vector, it enables the airborne intelligent model The output predicted spray flow rate is closest to the target spray flow rate. The above-described PWM iterative solution process can quickly and accurately find the optimal PWM value, providing a reliable control signal for the precise spraying of agricultural drones.

[0087] In one embodiment, prior to step S106, the method further includes: performing a safety limit on the optimal pulse width modulation signal value of the current control cycle based on the standard pulse width modulation signal value width of the agricultural drone.

[0088] In this embodiment, the main control unit obtains the value calculated in step S1055. Afterwards, it cannot be directly output to the spraying unit because the spraying unit 503 has a specific operating range. If the PWM signal value exceeds this range, the spraying unit will malfunction or even damage the equipment. Therefore, the main control unit 501 needs to... Safety limiting is implemented. Specifically, the PWM signal value is limited to the range of 1050µs to 1950µs. This range is determined based on the physical characteristics of the spraying unit and actual working requirements, ensuring that the spraying unit operates in a safe and stable state.

[0089] Furthermore, after completing the amplitude limiting process, the main control unit 501 outputs the processed PWM signal as a control signal to the spraying unit 503 through a timer, ensuring that the spraying unit 503 can work according to the predetermined pattern and achieve precise spraying control.

[0090] S106: Control the spraying unit to output the actual spraying flow rate under the current control cycle according to the optimal pulse width modulation signal value.

[0091] In this embodiment, after obtaining the optimal pulse width modulation signal value through step S105, the optimal PWM value controls the spraying unit 503 to output the actual spraying flow rate under the current control cycle, which is close to the target spraying flow rate.

[0092] Furthermore, after the current control cycle ends, the intelligent control system 500 for the agricultural drone's flow meter-less spraying returns to step S101 and repeats all the above steps (S101 to S106), forming a continuous, high-frequency adaptive closed-loop control that can adapt to changes in flight attitude, speed, voltage, and pesticide dosage in real time. In this closed-loop control system, each control cycle is a complete adjustment process. The system continuously recalculates the optimal PWM control value based on the current actual state and target requirements, and outputs it to the spraying unit 503. Through this high-frequency adaptive adjustment, the system can promptly respond to various changing factors, such as sudden changes in the drone's flight attitude, adjustments in flight speed, fluctuations in power supply voltage, and reductions in pesticide dosage, always maintaining a stable and precise spraying flow rate to ensure optimal operational results for the agricultural drone.

[0093] The intelligent control method for meterless spraying of agricultural drones provided in this embodiment calibrates the drone's tank according to a pre-built static flow calibration model during the static phase of the drone to obtain an initial apparent weight reading. During the dynamic phase, the method acquires a first state vector of the drone in the current control cycle, which includes the pulse width modulation signal value from the previous control cycle, the battery voltage, the initial apparent weight reading, the vertical acceleration, roll angle, and pitch angle. It also acquires a second state vector of the drone in the current control cycle, which includes the current ground velocity, the user-set dosage per acre, and the spray width. The target spraying flow rate for the current control cycle is calculated based on the second state vector. The first state vector is input into an onboard intelligent model to obtain an estimated spraying flow rate. Iterative optimization is performed based on the predicted spraying flow rate and the target spraying flow rate to obtain the optimal pulse width modulation signal value for the current control cycle. Finally, the spraying unit is controlled to output the actual spraying flow rate for the current control cycle based on the optimal pulse width modulation signal value. This application directly solves the optimal control command required to achieve the target flow rate by real-time fusion of multi-dimensional dynamic information, including body acceleration and attitude angle, and rapid iteration of pulse width modulation signal values, enabling agricultural drones to operate stably and accurately in changing environments for a long time.

[0094] Example 2

[0095] In addition, this application provides an intelligent control system for agricultural drone spraying without a flow meter, which is applied to electronic devices.

[0096] like Figure 5 As shown, the intelligent control system 500 for agricultural drone spraying without a flow meter includes:

[0097] The main control unit 501 is used to execute algorithm programs and control logic;

[0098] Weighing unit 502 is used to acquire the weight signal of the medicine box in real time;

[0099] The spraying unit 503 is used to receive the pulse width modulation signal value sent by the main control unit and control the nozzle to spray liquid through the pulse width modulation signal value;

[0100] Flight status unit 504 is used to acquire the vertical acceleration, roll angle and pitch angle of the agricultural drone.

[0101] The power management unit 505 is used to monitor the battery voltage in real time during each control cycle.

[0102] Data storage unit 506 is used to store the model parameters and model structure of the airborne intelligent model and the static flow calibration model;

[0103] The human-machine interface unit 507 is used to remotely receive control signals from operators.

[0104] The intelligent control system 500 for spraying agricultural drones without flow meters provided in this embodiment can realize the intelligent control method for spraying agricultural drones without flow meters provided in Embodiment 1. To avoid repetition, it will not be described again here.

[0105] The intelligent control system for meterless spraying of agricultural drones provided in this embodiment calibrates the drone's tank according to a pre-built static flow calibration model during the static phase of the drone, obtaining an initial apparent weight reading. During the dynamic phase, it acquires a first state vector for the drone in the current control cycle, including the pulse width modulation signal value from the previous control cycle, the battery voltage, the initial apparent weight reading, the vertical acceleration, roll angle, and pitch angle. It also acquires a second state vector for the drone in the current control cycle, including the current ground velocity, the user-set dosage per acre, and the spray width, and calculates the target spraying flow rate for the current control cycle based on the second state vector. The first state vector is input into the onboard intelligent model to obtain an estimated spraying flow rate. Iterative optimization is performed based on the predicted spraying flow rate and the target spraying flow rate to obtain the optimal pulse width modulation signal value for the current control cycle. Finally, the spraying unit is controlled to output the actual spraying flow rate for the current control cycle based on the optimal pulse width modulation signal value. This application directly solves the optimal control command required to achieve the target flow rate by real-time fusion of multi-dimensional dynamic information, including body acceleration and attitude angle, and rapid iteration of pulse width modulation signal values, enabling agricultural drones to operate stably and accurately in changing environments for a long time.

[0106] Example 3

[0107] Furthermore, this application provides an agricultural drone, including a memory and a processor. The memory stores a computer program, and the computer program executes the flow meter-free intelligent control method for agricultural drone spraying provided in Embodiment 1 when it runs on the processor.

[0108] The agricultural drone provided in this embodiment of the invention can execute the steps of the intelligent control method for spraying agricultural drones without flow meters provided in the above-described method embodiment 1. To avoid repetition, these steps will not be repeated here.

[0109] Example 4

[0110] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the intelligent control method for water spraying without a flow meter for agricultural drones provided in Embodiment 1.

[0111] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0112] The computer-readable storage medium provided in this embodiment can implement the intelligent control method for water spraying without flow meter provided in Embodiment 1. To avoid repetition, it will not be described again here.

[0113] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0114] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0115] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for intelligent control of agricultural drone spraying without a flow meter, characterized in that, The method includes: During the static phase of the agricultural drone, the tank of the agricultural drone is calibrated according to the pre-built static flow calibration model to obtain the initial apparent weight reading. During the dynamic phase of the agricultural drone, the first state vector of the agricultural drone in the current control cycle is obtained. The first state vector includes the pulse width modulation signal value in the previous control cycle and the battery voltage, the initial apparent weight reading, the vertical acceleration of the drone, the roll angle and the pitch angle in the current control cycle. The second state vector of the agricultural drone under the current control cycle is obtained. The second state vector includes the current ground speed, the user-set amount per acre and the spray width. The target spray flow rate under the current control cycle is calculated based on the second state vector. The first state vector under the current control cycle is input into the airborne intelligent model to obtain the estimated spray flow rate; Based on the estimated spray flow rate and the target spray flow rate, iterative optimization is performed to obtain the optimal pulse width modulation signal value for the current control cycle; The step of iteratively optimizing based on the estimated spray flow rate and the target spray flow rate to obtain the optimal pulse width modulation signal value for the current control cycle includes: The first batch of data from the first state vector of the current control cycle is input into the airborne intelligent model to obtain the first estimated spray flow rate. Calculate the error spray flow rate between the first estimated spray flow rate and the target spray flow rate; The pulse width modulation signal value of the current cycle is increased by a fixed increment to obtain the second batch of data of the first state vector of the current control cycle, and the second batch of data is input into the airborne intelligent model to obtain the second estimated spray flow rate; The first estimated spray flow rate and the second estimated spray flow rate are used to calculate the local gradient of the airborne intelligent model with respect to the pulse width modulation signal value at the current operating point using the difference method; The optimal pulse width modulation signal value for the current control cycle is calculated based on the local gradient and the error spray flow rate. The spraying unit is controlled to output the actual spraying flow rate under the current control cycle based on the optimal pulse width modulation signal value.

2. The method according to claim 1, characterized in that, Before inputting the first state vector under the current control cycle into the airborne intelligent model, the method further includes: A multidimensional nonlinear function model is constructed based on the state vectors of each control cycle; The actual flow rate is calculated based on the average weight of the agricultural drone under each control cycle, and the actual flow rate label is obtained. A dynamic working condition training dataset is constructed based on the state vector of each control cycle and the actual flow rate label. The multidimensional nonlinear function model is trained based on the dynamic operating condition training set to obtain the airborne intelligent model.

3. The method according to claim 2, characterized in that, The airborne intelligent model is one of the following: feedforward neural network, support vector regression, gradient boosting decision tree, and radial basis function network.

4. The method according to claim 2, characterized in that, After obtaining the airborne intelligent model, the method further includes: The model structure and parameters of the airborne intelligent model are converted into a function library; The function library is embedded into the data storage unit of the agricultural drone's flight control firmware.

5. The method according to claim 1, characterized in that, Before controlling the spraying unit to output the actual spraying flow rate under the current control cycle according to the optimal pulse width modulation signal value, the method further includes: Based on the standard pulse width modulation signal width of the agricultural drone, the optimal pulse width modulation signal value for the current control cycle is subjected to a safety limit.

6. The method according to claim 1, characterized in that, Construct a static traffic calibration model, including: After prompting the user to add standard liquid under static conditions, the initial stable weight of the medicine tank is obtained; The pulse width modulation signal values ​​are sequentially selected from the preset pulse width modulation signal value list to test the water spray volume of the medicine tank, and the final stable weight of the medicine tank corresponding to each pulse width modulation signal value is obtained. Calculate the actual spraying flow rate under static conditions based on the final stable weight of the medicine tank and the initial stable weight of the medicine tank; By correlating the pulse width modulation signal value, average battery voltage, and average tank weight with the corresponding actual spraying flow rate, a static operating condition training dataset is constructed. A static flow calibration model is trained based on the static operating condition training dataset.

7. A flow meter-free intelligent control system for agricultural drone spraying, characterized in that, The system includes: The main control unit is used to execute algorithm programs and control logic; Weighing unit, used to acquire the weight signal of the medicine box in real time; The spraying unit is used to receive the pulse width modulation signal value sent by the main control unit and control the nozzle to spray liquid through the pulse width modulation signal value; The flight status unit is used to acquire the vertical acceleration, roll angle, and pitch angle of the agricultural drone. The power management unit is used to monitor the battery voltage in real time during each control cycle. Data storage unit, used to store model parameters and model structure of airborne intelligent model and static flow calibration model; The human-machine interface unit is used to remotely receive control signals from operators.

8. A plant protection drone, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program that executes the intelligent control method for water spraying without a flow meter for agricultural drones as described in any one of claims 1 to 7 when the processor is running.

9. A computer-readable storage medium, characterized in that, It stores a computer program, which, when run on a processor, executes the intelligent control method for water spraying without a flow meter for agricultural drones as described in any one of claims 1 to 7.

Images