A target spraying control method and device based on strong tracking Kalman filter

Abstract

This application discloses a target-based drug delivery control method and apparatus based on a strong-tracking Kalman filter. The method determines the target drug delivery flow rate based on continuous image sequences and vehicle motion parameters. The real-time flow rate of the main pipeline is input into the strong-tracking Kalman filter, and the predicted covariance matrix is ​​corrected by calculating the asymptotic fading factor to obtain a flow rate estimate. A feedforward compensation amount is generated based on the difference between the target flow rate and the flow rate estimate, and a feedback compensation amount is generated based on the difference between the pressure setpoint and the real-time pressure data. These are fused into a composite command to adjust the backflow valve pressure. A PWM control signal is generated based on the target flow rate and the adjusted target pressure, and the nozzle is controlled to deliver drug to the target based on the PWM control signal. This method achieves active pre-adjustment and pressure stabilization control for flow rate step disturbances, improving the drug delivery accuracy under random dynamic target flow rates.

Description

Technical Field

[0001] This application relates to the field of precision agriculture, and in particular to a method and apparatus for targeted drug application control based on strong tracking Kalman filtering. Background Technology

[0002] Precise targeted pesticide application is a key technology for achieving reduced pesticide use and increased efficiency in precision agriculture. Large boom sprayers, during operation, need to dynamically calculate the target pesticide flow rate based on the real-time identification of the target weed area and the vehicle's speed by the vision system. However, changes in field lighting and leaf shading cause random disturbances in the visual detection results, resulting in random fluctuations in the target flow rate command. Directly using this as control input can easily cause system oscillations. The superposition of these visual disturbances and the frequent opening and closing of the nozzles further exacerbates the random fluctuations in the main pipeline flow rate, making it difficult for the control system to establish a stable pressure baseline.

[0003] Furthermore, the numerous nozzles on the ultra-wide spray boom need to be frequently opened and closed according to the target distribution. Each opening and closing action will cause an instantaneous flow step in the main pipeline, which will evolve into severe pressure fluctuations through the nonlinear coupling of the hydraulic system. In the existing technology, constant pressure control cannot adapt to large changes in the number of nozzles, PID pressure regulation is lagging in response to fast time-varying disturbances, and single-nozzle PWM control lacks system-level flow closed-loop correction, making it difficult to maintain stable application under the combined disturbances of random changes in target flow and dynamic switching of multiple nozzles. Summary of the Invention

[0004] This application provides a target application control method and device based on strong tracking Kalman filtering, which solves the technical problem in the prior art where frequent opening and closing of nozzles causes step disturbances in the main pipeline flow, resulting in severe pressure fluctuations and unstable application rate control.

[0005] In a first aspect, this application provides a target-based drug delivery control method based on strong tracking Kalman filtering, including:

[0006] Acquire continuous image sequences of the work area and vehicle motion parameters of the spraying vehicle, and determine the target spraying flow rate corresponding to the work area based on the continuous image sequences and vehicle motion parameters;

[0007] Acquire real-time traffic flow data and real-time pressure data of the main pipeline, and input the real-time traffic flow data into a pre-built strong tracking Kalman filter so that the strong tracking Kalman filter can calculate the residual sequence based on the real-time traffic flow data and the predicted output value of the strong tracking Kalman filter.

[0008] Calculate the sequence covariance matrix corresponding to the residual sequence, dynamically solve the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline.

[0009] Based on the first difference between the target application rate and the estimated application rate, a flow feedforward compensation amount is generated; based on the second difference between the preset pressure setpoint and the real-time pressure data, a pressure feedback control amount is generated.

[0010] The flow feedforward compensation and pressure feedback control are combined to generate a composite control command; the composite control command is sent to the reflux pressure regulating valve to adjust the target pressure of the main pipeline.

[0011] Based on the target application flow rate and target pressure, a PWM control signal for the nozzle solenoid valve is generated, and the nozzle solenoid valve is controlled to perform targeted application of pesticide to the work area according to the PWM control signal.

[0012] Secondly, this application provides a target-based drug delivery control device based on strong tracking Kalman filtering, comprising:

[0013] The target application flow rate determination module is configured to acquire a continuous image sequence of the work area and the vehicle motion parameters of the application vehicle, and determine the target application flow rate corresponding to the work area based on the continuous image sequence and the vehicle motion parameters.

[0014] The residual sequence calculation module is configured to acquire real-time flow data and real-time pressure data of the main pipeline, input the real-time flow data into a pre-built strong tracking Kalman filter, so that the strong tracking Kalman filter can calculate the residual sequence based on the real-time flow data and the predicted output value of the strong tracking Kalman filter.

[0015] The traffic flow prediction module is configured to calculate the sequence covariance matrix corresponding to the residual sequence, dynamically solve the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline.

[0016] The composite control module is configured to generate a flow feedforward compensation amount based on a first difference between the target application flow rate and the estimated flow rate; and to generate a pressure feedback control amount based on a second difference between a preset pressure setpoint and real-time pressure data.

[0017] The composite control command generation module is configured to fuse the flow feedforward compensation amount and the pressure feedback control amount to generate a composite control command; and send the composite control command to the return pressure regulating valve to adjust the target pressure of the main pipeline.

[0018] The execution module is configured to generate a PWM control signal for the nozzle solenoid valve based on the target application flow rate and target pressure, and control the nozzle solenoid valve to perform target application to the work area based on the PWM control signal.

[0019] Thirdly, this application provides a readable medium including executable instructions, which, when executed by a processor of an electronic device, cause the electronic device to perform any of the methods described in the first aspect.

[0020] Fourthly, this application provides an electronic device including a processor and a memory storing execution instructions, wherein when the processor executes the execution instructions stored in the memory, the processor performs the method as described in any of the first aspects.

[0021] This application provides a target-based pesticide application control method and apparatus based on a strong-tracking Kalman filter. The method acquires a continuous image sequence of the work area and vehicle motion parameters of the pesticide application vehicle, and determines the target pesticide application flow rate corresponding to the work area based on the continuous image sequence and vehicle motion parameters. It acquires real-time flow and pressure data of the main pipeline, inputs the real-time flow data into a pre-constructed strong-tracking Kalman filter, and enables the strong-tracking Kalman filter to calculate a residual sequence based on the real-time flow data and the predicted output value of the strong-tracking Kalman filter. The method calculates the sequence covariance matrix corresponding to the residual sequence, dynamically solves for the asymptotic decay factor based on the sequence covariance matrix, and utilizes the asymptotic decay factor... The prediction covariance matrix of the strong-tracking Kalman filter is corrected to obtain the flow rate estimate for the main pipeline. A flow feedforward compensation is generated based on the first difference between the target application flow rate and the flow rate estimate. A pressure feedback control is generated based on the second difference between the preset pressure setpoint and the real-time pressure data. The flow feedforward compensation and pressure feedback control are fused to generate a composite control command. This composite control command is sent to the return pressure regulating valve to adjust the target pressure of the main pipeline. A PWM control signal for the nozzle solenoid valve is generated based on the target application flow rate and target pressure, and the nozzle solenoid valve is controlled to perform targeted application to the work area based on the PWM control signal. This achieves proactive pre-adjustment of flow rate step disturbances caused by frequent nozzle opening and closing and stable control of the main pipeline pressure, improving the accuracy of targeted application under random dynamic target flow rates.

[0022] The further effects of the aforementioned non-conventional preferred method will be explained below in conjunction with specific embodiments. Attached Figure Description

[0023] To more clearly illustrate the embodiments of this application or the existing technical solutions, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1A flowchart illustrating a target-based drug delivery control method based on strong tracking Kalman filtering, provided in an embodiment of this application;

[0025] Figure 2 A flowchart illustrating another target-based drug delivery control method based on strong tracking Kalman filtering provided in an embodiment of this application;

[0026] Figure 3 This is a schematic diagram of a target drug delivery control device based on strong tracking Kalman filtering, provided in an embodiment of this application.

[0027] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] Precise targeted pesticide application is a key technology for achieving reduced pesticide use and increased efficiency in precision agriculture. Large boom sprayers, during operation, need to dynamically calculate the target pesticide flow rate based on the real-time identification of the target weed area and the vehicle's speed by the vision system. However, changes in field lighting and leaf shading cause random disturbances in the visual detection results, resulting in random fluctuations in the target flow rate command. Directly using this as control input can easily cause system oscillations. The superposition of these visual disturbances and the frequent opening and closing of the nozzles further exacerbates the random fluctuations in the main pipeline flow rate, making it difficult for the control system to establish a stable pressure baseline.

[0030] Furthermore, the numerous nozzles on the ultra-wide spray boom need to be frequently opened and closed according to the target distribution. Each opening and closing action will cause an instantaneous flow step in the main pipeline, which will evolve into severe pressure fluctuations through the nonlinear coupling of the hydraulic system. In the existing technology, constant pressure control cannot adapt to large changes in the number of nozzles, PID pressure regulation is lagging in response to fast time-varying disturbances, and single-nozzle PWM control lacks system-level flow closed-loop correction, making it difficult to maintain stable application under the combined disturbances of random changes in target flow and dynamic switching of multiple nozzles.

[0031] To address this issue, this application proposes a target-based dosing control method based on strong-tracking Kalman filtering. This method aims to solve the technical problem in existing technologies where frequent nozzle opening and closing causes step disturbances in the main pipeline flow, leading to severe pressure fluctuations and unstable dosing control. In this embodiment, the target-based dosing control method based on strong-tracking Kalman filtering includes:

[0032] Step 101: Obtain a continuous image sequence of the work area and the vehicle motion parameters of the spraying vehicle, and determine the target spraying flow rate corresponding to the work area based on the continuous image sequence and vehicle motion parameters.

[0033] In actual field operations, due to the continuous changes in lighting conditions over time, coupled with the shading effect of crop leaves on weeds, the detection results of a single frame image often exhibit random fluctuations, making it difficult to directly use as a stable control input. Therefore, this embodiment uses a continuous image sequence rather than a single frame image as the visual perception input. The spraying vehicle travels at a constant or variable speed along the work row direction, and the onboard visual perception system continuously acquires images of the work area at a fixed frame rate, forming a continuous image sequence. At the same time, vehicle motion parameters are acquired in real time through a vehicle speed sensor, including the vehicle's current speed and attitude information.

[0034] A pre-trained target detection model processes a continuous image sequence to identify and locate weed targets, outputting target detection results. The target detection results include the bounding box position and area information of each weed target. The target detection results and vehicle motion parameters are input into an embodied intelligent prediction model to predict the position and area changes of each weed target within a preset time window, resulting in a dynamic target set. The dynamic target set includes the predicted position and predicted area of ​​each weed target. Based on the predicted area and the vehicle speed in the vehicle motion parameters, the target pesticide application rate is calculated.

[0035] Specifically, the formula for calculating the target application rate is as follows:

[0036]

[0037] in, For the first The target application rate for each control cycle. For the first The target number of weeds within each control cycle. For the first Preset dose coefficient for each weed target. For the first Predicted area of ​​each weed target. For the first The vehicle speed is controlled within each control cycle. The target application rate is directly proportional to both the weed coverage area and the vehicle speed. The larger the coverage area or the faster the vehicle speed, the higher the instantaneous application rate is required to ensure a constant application rate per unit area.

[0038] After obtaining a continuous image sequence, the system first calls a pre-trained object detection model to process the sequence frame by frame. The object detection model can employ a deep learning-based architecture, specifically trained for the characteristics of field weeds, enabling rapid identification and accurate localization of weed targets in complex backgrounds. The output of the object detection model is the object detection result, which includes the bounding box coordinates and corresponding area information of each weed target in the current frame. The bounding box positions describe the spatial distribution of the weed targets in the image coordinate system, while the area information reflects the degree of weed coverage.

[0039] Relying solely on the detection results of the current frame is insufficient to address the combined effects of visual disturbances and vehicle motion. Due to the continuous movement of the spraying vehicle, there is an inherent response delay between the nozzle and the target area. If the control is directly driven by the detection results of the current frame, the spraying action will always lag behind the target position, resulting in spraying deviation.

[0040] Therefore, this embodiment introduces an embodied intelligent prediction model, which takes the target detection results and vehicle motion parameters as inputs, and makes a forward-looking prediction of the position and area change trends of each weed target within a preset time window.

[0041] The embodied intelligent prediction model comprehensively considers vehicle displacement, image perspective transformation, and target motion patterns, outputting a dynamic target set. This dynamic target set includes the predicted position and area of ​​each weed target at the prediction time. By introducing a prediction mechanism, the system can perceive changes in the spatial distribution of targets in advance, effectively compensating for response delays.

[0042] After obtaining the predicted area of ​​each weed target, the target pesticide application rate for the current work area can be calculated by combining the vehicle speed from the vehicle motion parameters. Specifically, the target pesticide application rate is positively correlated with the weed coverage area; the larger the coverage area, the more pesticide is required. At the same time, the target pesticide application rate is also affected by vehicle speed; the faster the vehicle speed, the larger the area swept by the spray boom per unit time, and correspondingly, a higher instantaneous flow rate is needed to maintain a constant pesticide application rate per unit area.

[0043] Step 102: Obtain real-time flow data and real-time pressure data of the main pipeline, and input the real-time flow data into the pre-built strong tracking Kalman filter so that the strong tracking Kalman filter can calculate the residual sequence based on the real-time flow data and the predicted output value of the strong tracking Kalman filter.

[0044] Flow and pressure sensors installed on the main pipeline collect flow and pressure data in real time at a fixed sampling period and upload the results to the system. Due to the frequent opening and closing of the nozzle solenoid valve, the flow signal in the main pipeline is not a stable signal but is superimposed with instantaneous step disturbances caused by valve action. Therefore, the real-time flow data collected by the sensors contains significant noise components, and directly using it for control feedback can lead to unnecessary oscillating responses in the system. To address this, this embodiment constructs a strong tracking Kalman filter to dynamically estimate and suppress noise in the real-time flow data.

[0045] The state vector of the strong tracking Kalman filter is determined; the state vector consists of the main pipeline flow rate, the rate of change of flow rate, and the sensor measurement bias; the state vector is recursively derived into the state prediction value through the state transition matrix, and the state prediction value is mapped into the measurement matrix to obtain the predicted output value; the third difference between the real-time flow data and the predicted output value is calculated to obtain the residual sequence.

[0046] The construction of the strong tracking Kalman filter is based on modeling the dynamic characteristics of the main pipeline flow. In determining the filter's state vector, this embodiment selects three state components to form the state vector: the first is the main pipeline flow, representing the current core quantity being estimated; the second is the flow rate of change, used to describe the dynamic trend of flow changes, enabling the filter to track rapid changes in flow; and the third is the sensor measurement bias, used to identify and compensate for systematic errors in the sensors online, improving estimation accuracy.

[0047] The state vector of a strong tracking Kalman filter is defined as follows:

[0048]

[0049] in, To control the traffic flow on the main road, For the rate of change of flow, The sensor measures the bias. The sampling period is used. Discretize the system to obtain the state transition matrix. :

[0050]

[0051] The discretized state equation of the system is expressed as:

[0052]

[0053] The measurement equation is expressed as:

[0054] in, The measured values ​​from the flow sensor, i.e., real-time flow data, are used in the measurement matrix. , For process noise, For measuring noise.

[0055] The formula for calculating the residual sequence is:

[0056]

[0057] During the state prediction phase, the system recursively calculates the state vector at the current moment using the state transition matrix to obtain the predicted state value for the next moment. The state transition matrix is ​​constructed based on the physical model of the main pipeline flow and reflects the evolution relationship of each state component between adjacent sampling moments.

[0058] The predicted state values ​​are mapped to the measurement space through a measurement matrix to obtain the predicted output value at the current moment, which is the flow observation based on the model prediction by the filter. The difference between the real-time collected flow data and the predicted output value is used to obtain the residual sequence. The residual sequence is a key indicator for measuring the deviation between the model prediction and the actual measurement, and its statistical characteristics directly reflect the strength and nature of the disturbance currently affecting the system.

[0059] Step 103: Calculate the sequence covariance matrix corresponding to the residual sequence, dynamically calculate the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline.

[0060] The sequence covariance matrix reflects the statistical dispersion of the residual sequence. When the system is subjected to sudden disturbances, the covariance of the residual sequence increases significantly, indicating a weakening of the model's predictive ability and a need to increase the weight of new measurement information. When the system is running smoothly, the covariance of the residual sequence is small, and the strong tracking Kalman filter can maintain a high prediction confidence.

[0061] Based on the sequence covariance matrix, the preset measurement noise covariance matrix, and the preset process noise covariance matrix, the asymptotic decay factor is calculated; the predicted covariance matrix is ​​multiplied by the asymptotic decay factor and then superimposed with the process noise covariance matrix to obtain the corrected covariance matrix; the Kalman gain is calculated based on the corrected covariance matrix, and the state prediction value is corrected using the Kalman gain to obtain the flow rate prediction value corresponding to the main pipeline.

[0062] The asymptotic decay factor is the core mechanism that distinguishes the strong-tracking Kalman filter from the standard Kalman filter. The standard Kalman filter uses a fixed prediction covariance matrix, which cannot respond quickly when the system experiences abrupt changes, resulting in significant lag in the filter estimation. In contrast, the strong-tracking Kalman filter introduces an asymptotic decay factor, amplifying the prediction covariance matrix when an increase in residual abnormality is detected, thus increasing the Kalman gain. This allows the filter to more quickly approach the new measurement, achieving a strong tracking capability for abrupt changes.

[0063] Gradual decay factor The solution depends on the following intermediate matrix:

[0064]

[0065] in, , To solve for the intermediate matrix of the asymptotic decay factor, For the measurement matrix, This is the prediction covariance matrix for the previous period. The sequence covariance matrix, As a weakening factor, To measure the noise covariance matrix, This represents the process noise covariance matrix. The asymptotic decay factor... Solve using the following formula:

[0066]

[0067] in, The trace operation represents the sum of the elements on the main diagonal of a matrix.

[0068] The system multiplies the predicted covariance matrix by the dynamically calculated asymptotic decay factor, and then superimposes it with the process noise covariance matrix to obtain the corrected covariance matrix. The corrected covariance matrix reflects the degree of uncertainty of the system, adaptively expands when the flow rate changes abruptly, and tends to converge when the system is running smoothly.

[0069] use Corrected prediction covariance matrix:

[0070]

[0071] in, To correct the covariance matrix, This is the prediction covariance matrix for the previous period. Here is the state transition matrix. Let be the process noise covariance matrix.

[0072] The Kalman gain is calculated based on the modified covariance matrix. The Kalman gain determines the relative confidence of the filter in the predicted and measured values ​​during state updates. Finally, the state prediction is corrected using the Kalman gain, and the weighted sum of the state prediction and the residual is used as the final output of the state estimate, thus obtaining the estimated flow rate for the main pipeline.

[0073] Solving the Kalman gain The status update is calculated using the following formula:

[0074]

[0075]

[0076] in, For traffic estimation, This is the predicted state value. For real-time traffic data, It is a residual sequence.

[0077] Step 104: Generate a flow feedforward compensation amount based on the first difference between the target application flow rate and the estimated flow rate; generate a pressure feedback control amount based on the second difference between the preset pressure setpoint and the real-time pressure data.

[0078] Unlike conventional feedback control, which relies on a passive response mechanism that adjusts after pressure deviation occurs, the flow feedforward compensation in this embodiment is generated based on the first difference between the target drug application flow rate and the estimated flow rate output by the strong tracking filter. This allows for the application of adjustment commands to the return pressure regulating valve as soon as the opening and closing action of the nozzle solenoid valve causes the main pipeline flow rate to deviate from the target value. This suppresses pressure fluctuations before they fully develop, achieving proactive pre-adjustment of flow step disturbances and significantly reducing the amplitude of main pipeline pressure oscillations.

[0079] The first difference directly reflects the deviation between the current main pipeline flow and the target demand. Using it as a feedforward signal input to the control loop can apply compensation in advance before the pressure deviation is fully reflected in the measured pressure, thereby significantly shortening the system's response time to changes in flow demand.

[0080] The operating state vector is input to a pre-trained reinforcement learning adaptive controller to obtain the first pressure regulation component. The operating state vector includes real-time pressure data, flow rate estimate and corresponding rate of change of the estimate, target application flow rate, number of open nozzle solenoid valves, and vehicle speed. The second difference is input to a PID controller to obtain the second pressure regulation component. The first pressure regulation component and the second pressure regulation component are superimposed to obtain the pressure feedback control quantity.

[0081] This embodiment also employs a dual-component superposition structure combining a reinforcement learning adaptive controller and a PID controller to address the strong nonlinear characteristics exhibited by the dosing system in scenarios with dynamic switching of multiple nozzles. The system constructs an operating state vector and inputs it into a pre-trained reinforcement learning adaptive controller, outputting a first pressure regulation component.

[0082] The operating status vector contains real-time pressure data reflecting the current pressure status of the system; the flow rate estimate and its corresponding rate of change jointly describe the dynamic evolution trend of the flow rate; the target application flow rate represents the flow demand of the current control cycle; the number of open nozzle solenoid valves reflects the current configuration of the hydraulic load; and the vehicle speed reflects the dynamic changes in the operating conditions.

[0083] By comprehensively sensing the above multi-dimensional states, reinforcement learning adaptive controllers can learn and output optimal adjustment strategies that adapt to complex nonlinear operating conditions, thus overcoming the limitation of traditional PID control in adapting parameters under varying operating conditions.

[0084] The system inputs the second difference between the preset pressure setpoint and the real-time pressure data into the PID controller to obtain the second pressure regulation component, providing stable basic feedback regulation capability. The first and second pressure regulation components are superimposed to obtain the final pressure feedback control quantity, realizing the complementarity of adaptive regulation and feedback control.

[0085] Step 105: Combine the flow feedforward compensation amount with the pressure feedback control amount to generate a composite control command; send the composite control command to the return pressure regulating valve to adjust the target pressure of the main pipeline.

[0086] The flow feedforward compensation and pressure feedback control are superimposed to generate a composite control command that acts on the reflux pressure regulating valve. This composite control command is sent to the reflux pressure regulating valve actuator via a communication interface. By adjusting the reflux flow rate, the hydraulic resistance in the main pipeline is changed, thereby controlling the main pipeline pressure to the target pressure. The target pressure refers to the actual pressure value reached in the main pipeline after adjustment by the reflux pressure regulating valve.

[0087] The fusion of feedforward and feedback control mechanisms creates a complementary relationship between the two control methods on a time scale. The flow feedforward compensation has a fast response speed, intervening before pressure fluctuations accumulate to a significant level, while the pressure feedback control continuously corrects system errors during the steady-state phase, ensuring long-term pressure stability. Together, they form a multi-level closed-loop control of the main pipeline pressure, enabling the system to maintain a stable pressure baseline even under the combined disturbances of dynamic switching between multiple nozzles and random changes in target flow.

[0088] Step 106: Generate a PWM control signal for the nozzle solenoid valve based on the target application flow rate and target pressure, and control the nozzle solenoid valve to perform target application to the work area based on the PWM control signal.

[0089] Once the main pipeline pressure stabilizes at the target pressure, the system extracts the predicted location and area of ​​each weed target based on the dynamic target set output by the embodied intelligent prediction model. Each weed target is mapped to the spray boom coordinate system to determine which nozzle's spray coverage area each weed target falls within. The predicted areas of all weed targets are then summarized to obtain the total weed area within the current working area and the coverage area corresponding to each nozzle's solenoid valve. Based on the proportion of each nozzle's coverage area to the total weed area, the target application flow rate is proportionally allocated to each nozzle's solenoid valve, resulting in the target allocation flow rate for each nozzle's solenoid valve.

[0090] To establish a global consistency correction between the target total demand and the actual system outflow, a global flow correction coefficient is introduced. The global flow correction coefficient is the ratio of the sum of the target allocated flow rates of each nozzle solenoid valve to the real-time flow rate data of the main pipeline, reflecting the degree of deviation of the current actual total outflow rate of the system from the target application flow rate.

[0091] By using a global flow correction coefficient to proportionally scale the target distribution flow of each nozzle solenoid valve, the final distribution flow is obtained. This automatically compensates for the cumulative deviation caused by sensor drift and individual nozzle differences while keeping the relative distribution ratio of each nozzle unchanged, ensuring the consistency between the actual total output flow and the target application flow.

[0092] After obtaining the final allocated flow rate of each nozzle solenoid valve, based on the preset nozzle flow rate model, with the final allocated flow rate as the target and the target pressure as the current operating condition parameter, the nozzle flow rate model is inversely solved to obtain the PWM duty cycle required to make the nozzle flow rate equal to the final allocated flow rate under the current pressure condition, which is the initial PWM duty cycle of each nozzle solenoid valve.

[0093] Transient suppression is applied to the initial PWM duty cycle to eliminate the impact of drastic duty cycle jumps on the main pipeline pressure stability during control cycle switching. At the instant the nozzle solenoid valve opens, an exponential smoothing factor is introduced to perform a soft-start process on the initial PWM duty cycle, allowing the duty cycle to smoothly climb from zero to the target value, avoiding current surges and flow rate jumps at the moment of opening. (Exponential smoothing factor) The calculation formula is:

[0094]

[0095] in, For the current moment, This refers to the opening time of the solenoid valve for the nozzle. This is a time constant set based on the response frequency of the nozzle solenoid valve. The smoothed PWM duty cycle is:

[0096]

[0097] in, For the first The initial PWM duty cycle of each nozzle solenoid valve This is the PWM duty cycle after smoothing.

[0098] After completing the soft-start smoothing process, a rate constraint is further applied to the duty cycle change between adjacent control cycles to prevent water hammer effects in the pipeline caused by sudden changes in the distribution of weeds or abrupt changes in the number of nozzles opened, thus protecting the mechanical structure of the solenoid valve and maintaining stable pipeline pressure.

[0099]

[0100] in, For the first The change in duty cycle of each nozzle solenoid valve between adjacent control cycles For the first The PWM duty cycle of the nozzle solenoid valve after smoothing at the previous moment. This represents the maximum permissible change in duty cycle per step.

[0101] After the above transient suppression processing, the target PWM duty cycle of each nozzle solenoid valve is obtained. Based on this, a PWM control signal is generated and sent to the drive circuit of each nozzle solenoid valve through the communication interface, driving each nozzle to perform precise target application of pesticide to the weeds in the working area according to the precise switching sequence.

[0102] As can be seen from the above technical solutions, the beneficial effects of this embodiment are:

[0103] This application provides a target-based pesticide application control method based on a strong-tracking Kalman filter. The method acquires continuous image sequences of the work area and vehicle motion parameters of the pesticide application vehicle, and determines the target pesticide application flow rate corresponding to the work area based on the continuous image sequences and vehicle motion parameters. It acquires real-time flow data and real-time pressure data of the main pipeline, inputs the real-time flow data into a pre-constructed strong-tracking Kalman filter, and enables the strong-tracking Kalman filter to calculate a residual sequence based on the real-time flow data and the predicted output value of the strong-tracking Kalman filter. The method calculates the sequence covariance matrix corresponding to the residual sequence, dynamically solves for the asymptotic decay factor based on the sequence covariance matrix, and utilizes the asymptotic decay factor... The prediction covariance matrix of the strong-tracking Kalman filter is corrected to obtain the flow rate estimate for the main pipeline. A flow feedforward compensation is generated based on the first difference between the target application flow rate and the flow rate estimate. A pressure feedback control is generated based on the second difference between the preset pressure setpoint and the real-time pressure data. The flow feedforward compensation and pressure feedback control are fused to generate a composite control command. This composite control command is sent to the return pressure regulating valve to adjust the target pressure of the main pipeline. A PWM control signal for the nozzle solenoid valve is generated based on the target application flow rate and target pressure, and the nozzle solenoid valve is controlled to perform targeted application to the work area based on the PWM control signal. This achieves proactive pre-adjustment of flow rate step disturbances caused by frequent nozzle opening and closing and stable control of the main pipeline pressure, improving the accuracy of targeted application under random dynamic target flow rates.

[0104] Figure 1 The example shown is only a basic embodiment of a target-based drug delivery control method based on strong tracking Kalman filtering according to this application. With certain optimizations and extensions, other preferred embodiments of the target-based drug delivery control method based on strong tracking Kalman filtering can be obtained.

[0105] like Figure 2 The image shows another specific embodiment of a target-based drug delivery control method based on strong tracking Kalman filtering according to this application.

[0106] In this embodiment, a target-based drug delivery control method based on strong tracking Kalman filtering includes the following steps:

[0107] Step 201: Obtain a continuous image sequence of the work area and the vehicle motion parameters of the spraying vehicle, and determine the target spraying flow rate corresponding to the work area based on the continuous image sequence and vehicle motion parameters.

[0108] Step 202: Obtain real-time flow data and real-time pressure data of the main pipeline, and input the real-time flow data into the pre-built strong tracking Kalman filter so that the strong tracking Kalman filter can calculate the residual sequence based on the real-time flow data and the predicted output value of the strong tracking Kalman filter.

[0109] Step 203: Calculate the sequence covariance matrix corresponding to the residual sequence, dynamically solve the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline.

[0110] Step 204: Generate a flow feedforward compensation amount based on the first difference between the target application flow rate and the estimated flow rate; generate a pressure feedback control amount based on the second difference between the preset pressure setpoint and the real-time pressure data.

[0111] Step 205: Combine the flow feedforward compensation amount with the pressure feedback control amount to generate a composite control command; send the composite control command to the return pressure regulating valve to adjust the target pressure of the main pipeline.

[0112] Step 206: Generate a PWM control signal for the nozzle solenoid valve based on the target application flow rate and target pressure, and control the nozzle solenoid valve to perform target application to the work area based on the PWM control signal.

[0113] Step 207: Based on the predicted location and predicted area, determine the total area of ​​weeds corresponding to the weed target and the coverage area corresponding to each nozzle solenoid valve.

[0114] The embodied intelligent prediction model outputs a dynamic target set containing the predicted location and area of ​​each weed target within a preset time window. Using the predicted location as a reference, the system maps each weed target to the boom coordinate system, determining which nozzle(s) or nozzles each weed target falls within its spray coverage area.

[0115] The system summarizes the predicted areas of all weed targets in the dynamic target set to obtain the total weed area within the current working area. Furthermore, when determining the coverage area of ​​each nozzle solenoid valve, the system performs a mapping calculation based on the predicted location of each weed target and the spatial layout of the nozzle solenoid valves.

[0116] Specifically, the spray boom of the spraying vehicle has multiple nozzle solenoid valves evenly distributed laterally, each corresponding to a spray coverage strip of a certain width on the ground. Based on the predicted location of each weed target, the system determines which nozzle(s) its spatial coordinates fall within, and accumulates the predicted weed areas falling within the corresponding coverage strip to obtain the coverage area corresponding to that nozzle solenoid valve. For weed targets located at the boundary of adjacent nozzle coverage strips, the system assigns them to the corresponding nozzles according to a preset area allocation rule, ensuring that each weed area has exactly one assigned nozzle, avoiding duplicate spraying or missed spraying.

[0117] Step 208: Based on the proportion of the coverage area to the total area of ​​weeds in the work area, allocate the target application flow rate to each nozzle solenoid valve to obtain the target allocation flow rate corresponding to each nozzle solenoid valve.

[0118] Based on the coverage area of ​​each nozzle and the total area of ​​weeds, the target application flow rate is proportionally allocated to the solenoid valves of each nozzle according to the ratio of the coverage area of ​​each nozzle to the total area of ​​weeds, thus obtaining the target allocated flow rate for each nozzle.

[0119] The larger the target area of ​​weeds, the more pesticide the corresponding nozzle needs to handle, and therefore a higher flow rate should be allocated. By allocating according to area ratio, the system can achieve differentiated and precise application of pesticides to each nozzle while ensuring that the total global pesticide application rate is consistent with the target flow rate, so that the amount of pesticide solution received per unit area of ​​weeds tends to be more even.

[0120] If there are no weeds within the coverage area of ​​a nozzle solenoid valve, the corresponding coverage area is zero, the target flow rate is also zero, and the nozzle will remain closed during the control cycle, thereby achieving precise exemption for non-weed areas and effectively reducing unnecessary pesticide application.

[0121] Step 209: Calculate the global traffic correction coefficient based on real-time traffic data; the global traffic correction coefficient is the ratio of the sum of traffic allocated to each target to the real-time traffic data.

[0122] After obtaining the target flow rate allocation for each nozzle solenoid valve, the system calculates the global flow correction coefficient based on the real-time flow data of the main pipeline. The real-time flow data is collected in real-time by flow sensors installed on the main pipeline, reflecting the actual total outflow of the main pipeline within the current control cycle. The global flow correction coefficient is obtained by comparing the sum of the target flow rates allocation for each nozzle solenoid valve with the real-time flow data.

[0123] Global traffic correction factor The calculation formula is:

[0124]

[0125] in, For the real-time traffic data of the main road, This represents the total number of nozzle solenoid valves. Allocate the sum of traffic to the target. It is a preset positive number used to prevent the denominator from being zero.

[0126] The global flow correction factor directly reflects the deviation between the current actual total outflow rate and the target application rate. When the actual outflow rate of the main pipeline is lower than the target application rate, the global flow correction factor is greater than 1, and the target allocation flow rate for each nozzle will be adjusted upwards as a whole. When the actual outflow rate of the main pipeline is higher than the target application rate, the global flow correction factor is less than 1, and the target allocation flow rate for each nozzle will be adjusted downwards as a whole. When the two are equal, the global flow correction factor is equal to 1, and the target allocation flow rate for each nozzle does not need to be adjusted as a whole.

[0127] Due to factors such as fitting errors in the nozzle flow model, individual differences in solenoid valve response characteristics, and sensor measurement drift, the actual total outflow from each nozzle can continuously deviate from the target drug delivery flow rate. Without correction, these deviations will accumulate over multiple control cycles, ultimately leading to a significant deviation between the actual drug delivery per unit area and the expected dose, affecting the overall accuracy of target delivery. By introducing a global flow correction coefficient, the system can dynamically correct these accumulated deviations in a closed-loop manner within each control cycle, ensuring consistency between the overall drug delivery rate and the target drug delivery flow rate.

[0128] Step 210: Scale the allocated flow of each target proportionally using the global flow correction coefficient to obtain the final allocated flow of each nozzle solenoid valve.

[0129] After obtaining the global flow correction coefficient, the system uses the global flow correction coefficient to proportionally scale the target distribution flow of each nozzle solenoid valve to obtain the final distribution flow corresponding to each nozzle solenoid valve.

[0130] Final traffic allocation The calculation formula is:

[0131]

[0132] in, This is the global traffic correction factor. For the first The target flow rate of each nozzle solenoid valve. The advantage of proportional scaling is that the relative flow rate distribution ratio between each nozzle remains completely consistent before and after scaling. That is, the differentiated distribution pattern determined by the proportion of weed area in the coverage area of ​​each nozzle is completely preserved, and no new distribution imbalance is introduced due to global correction operation.

[0133] The spray boom of the pesticide application vehicle is divided into multiple collaborative control zones along its length. Each collaborative control zone contains at least one nozzle solenoid valve. For each collaborative control zone, the sum of the target allocated flow rates corresponding to each nozzle solenoid valve within the collaborative control zone is taken as the zone flow rate of the collaborative control zone. The average expected flow rate of all collaborative control zones is calculated. Based on the fourth difference between the zone flow rate and the average expected flow rate, a zone collaborative compensation amount corresponding to the collaborative control zone is generated. According to the preset allocation weight, the zone collaborative compensation amount is superimposed on the corresponding target allocated flow rate to obtain the corrected allocated flow rate, and the corrected allocated flow rate is taken as the target allocated flow rate.

[0134] In this embodiment, the system further introduces a correction mechanism based on the collaborative control region. After performing regional collaborative compensation on the target allocation flow, it performs proportional scaling to eliminate hydraulic coupling interference caused by the extremely uneven flow demand in each region due to the patchy distribution of weeds on the ultra-wide spray boom.

[0135] The system divides the spray boom of the application vehicle into several coordinated control zones along its length. Each coordinated control zone contains at least one nozzle solenoid valve, and adjacent coordinated control zones are hydraulically correlated. For each coordinated control zone, the target allocated flow rates corresponding to all nozzle solenoid valves within it are summed to obtain the regional flow rate of that zone. The regional flow rate reflects the total flow demand intensity of that local area within the current control cycle.

[0136] The system calculates the average flow rate of all zones as the expected average flow rate, representing the ideal flow rate benchmark that each zone should approach under normal conditions within the global scope of the spray boom. For collaborative control zones with flow rates higher than the expected average flow rate, the actual hydraulic load is heavier, requiring an appropriate reduction in the distribution flow rate of their internal nozzles to alleviate pressure shocks. For collaborative control zones with flow rates lower than the expected average flow rate, the distribution flow rate of their internal nozzles can be appropriately increased to fully utilize the system's margin. Therefore, the system generates the corresponding zone collaborative compensation amount for each collaborative control zone based on the fourth difference between the flow rate of each zone and the expected average flow rate.

[0137] Regional collaborative compensation amount The calculation formula is:

[0138]

[0139] in, For the cooperative gain coefficient, The expected value of the average flow rate. For the first Regional flow in a coordinated control area.

[0140] The system distributes the regional collaborative compensation amount to the target distribution flow of each nozzle solenoid valve in the corresponding collaborative control area according to the preset distribution weight, and the result is obtained by superimposing the weights.

[0141] Correct the allocation flow The calculation formula is:

[0142]

[0143] in, The preset allocation weights are used. The corrected allocation flow rate is taken as the target allocation flow rate to obtain the final allocation flow rate for each nozzle solenoid valve. Through the combined effect of regional collaborative correction and global flow correction coefficients, the final allocation flow rate obtained by each nozzle solenoid valve maintains the differentiated allocation characteristics to the target while its distribution along the length of the spray bar tends to be more balanced. Therefore, the local pressure drop in the main pipeline caused by concentrated nozzle opening in local areas and its impact on the flow accuracy of nozzles in other areas are mitigated, and the risk of coupling interference in the hydraulic system under uneven load distribution conditions is correspondingly reduced.

[0144] Step 211: Based on the preset nozzle flow model, calculate the initial PWM duty cycle of each nozzle solenoid valve according to the final allocated flow rate and target pressure.

[0145] After obtaining the final allocated flow rate of each nozzle solenoid valve, the system, based on a preset nozzle flow rate model, uses the final allocated flow rate as the target and the target pressure as the current operating condition parameter to inversely calculate the initial PWM duty cycle of each nozzle solenoid valve. The nozzle flow rate model describes the quantitative mapping relationship between the nozzle flow rate and the PWM duty cycle under a given pipeline pressure condition. It is usually established through offline calibration experiments. The model fits the correspondence between the PWM duty cycle and the measured flow rate under different pressure conditions to form a mapping function that can be looked up or solved online.

[0146] The nozzle flow rate model takes the following form:

[0147]

[0148] in, For the first The flow rate of each nozzle solenoid valve For the first The flow coefficient of a single nozzle solenoid valve

[0149] For the first The basic duty cycle of the nozzle solenoid valve. For target pressure, This is the preset pressure index.

[0150] When calculating the initial PWM duty cycle, the system substitutes the final distributed flow rate of each nozzle solenoid valve into the nozzle flow model for inverse kinematics to obtain the basic duty cycle required to make the nozzle flow rate equal to the final distributed flow rate under the target pressure condition. :

[0151]

[0152] in, This provides real-time pressure data for the main pipeline.

[0153] Considering that the main pipeline pressure fluctuates during actual operation due to frequent nozzle opening and closing, there may be a deviation between the real-time pressure data and the target pressure, thus affecting the degree of consistency between the nozzle flow rate and the target flow rate. Therefore, a compensation model based on pressure sensitivity is further introduced to correct the base duty cycle:

[0154]

[0155] in, This is the duty cycle after pressure compensation. The target pressure is determined by introducing pressure sensitivity compensation. This allows the system to adjust the PWM duty cycle of each nozzle in real time when the main pipeline pressure fluctuates, ensuring that the nozzle flow rate accurately tracks the final allocated flow rate even under pressure fluctuations.

[0156] Duty cycle after pressure compensation Then, a saturation constraint is applied to limit the duty cycle within the effective operating range allowed by the solenoid valve, thus obtaining the initial PWM duty cycle. :

[0157]

[0158] in, As a saturation function, the duty cycle is limited to or between. It is the minimum duty cycle signal that drives the nozzle flow solenoid valve. , It is the duty cycle signal when the solenoid valve for driving the nozzle flow rate is fully open. .

[0159] Step 212: Apply transient suppression processing to the initial PWM duty cycle to obtain the target PWM duty cycle of the nozzle solenoid valve, and generate a PWM control signal based on the target PWM duty cycle.

[0160] After obtaining the initial PWM duty cycle of each nozzle solenoid valve, the system applies transient suppression processing to eliminate the impact of drastic duty cycle jumps on the pressure stability of the main pipeline during control cycle switching. Between two adjacent control cycles, if the distribution of weeds changes significantly or the number of nozzles opening abruptly changes, the initial PWM duty cycle of each nozzle may experience a large step change. If such a step change is directly applied to the solenoid valve drive signal, it will not only trigger a new round of pressure shocks in the main pipeline but also accelerate the mechanical wear of the valve components, shortening their service life.

[0161] Transient suppression processing consists of two cascaded stages: soft-start smoothing and rate of change constraint. In the soft-start smoothing stage, the system introduces an exponential smoothing factor based on the switching state of each nozzle solenoid valve. At the moment the nozzle solenoid valve opens, the initial PWM duty cycle is smoothed, so that the duty cycle gradually rises from zero to the target value, avoiding the current surge and flow step at the moment of opening, and obtaining the intermediate PWM duty cycle after soft-start smoothing.

[0162] In the rate-of-change constraint stage, the system further applies a rate limit to the change in the intermediate PWM duty cycle between adjacent control cycles, constraining the change amplitude within the allowable range. This prevents water hammer effects in the pipeline caused by a step change in duty cycle, protecting the mechanical structure of the solenoid valve and maintaining stable pipeline pressure. After the soft-start smoothing and rate-of-change constraint stages are connected in series, the target PWM duty cycle for each nozzle solenoid valve is obtained. Based on the target PWM duty cycle, the system generates a PWM control signal with the corresponding frequency and duty cycle, which is sent to the drive circuit of each nozzle solenoid valve through the communication interface. This drives each nozzle to perform targeted application of pesticides to the weeds in the working area according to a precise switching sequence, achieving accurate target application.

[0163] As can be seen from the above technical solutions, the beneficial effects of this embodiment are as follows: by introducing the predicted area and spatial location information of weeds into the flow distribution calculation, and combining the dynamic closed-loop correction of the collaborative control area correction, the global flow correction coefficient, and the transient suppression processing of the PWM duty cycle, a complete control link from flow distribution to the generation of single nozzle drive signals is constructed. While realizing differentiated target application of each nozzle, it effectively suppresses the hydraulic coupling interference and pressure fluctuations caused by uneven weed distribution and frequent opening and closing of nozzles, and ensures the overall accuracy and system stability of the application rate control under composite disturbance conditions.

[0164] like Figure 3 The image shown is a specific embodiment of a target-based drug delivery control device based on a strong-tracking Kalman filter, as described in this application. This embodiment of a target-based drug delivery control device based on a strong-tracking Kalman filter is used to execute... Figures 1-2 A physical device for a target-based drug delivery control method based on strong-tracking Kalman filtering is provided. Its technical solution is essentially the same as the embodiments described above, and the corresponding descriptions in the embodiments above also apply to this embodiment. This embodiment of a target-based drug delivery control device based on strong-tracking Kalman filtering includes:

[0165] The target application flow rate determination module 301 is configured to acquire a continuous image sequence of the work area and the vehicle motion parameters of the application vehicle, and determine the target application flow rate corresponding to the work area based on the continuous image sequence and the vehicle motion parameters.

[0166] The residual sequence calculation module 302 is configured to acquire real-time flow data and real-time pressure data of the main pipeline, input the real-time flow data into a pre-built strong tracking Kalman filter, so that the strong tracking Kalman filter calculates the residual sequence based on the real-time flow data and the predicted output value of the strong tracking Kalman filter.

[0167] The traffic flow prediction module 303 is configured to calculate the sequence covariance matrix corresponding to the residual sequence, dynamically solve the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline.

[0168] The composite control module 304 is configured to generate a flow feedforward compensation amount based on a first difference between the target application flow rate and the estimated flow rate; and to generate a pressure feedback control amount based on a second difference between a preset pressure setpoint and real-time pressure data.

[0169] The composite control command generation module 305 is configured to fuse the flow feedforward compensation amount and the pressure feedback control amount to generate a composite control command; and send the composite control command to the return pressure regulating valve to adjust the target pressure of the main pipeline.

[0170] The execution module 306 is configured to generate a PWM control signal for the nozzle solenoid valve based on the target application flow rate and target pressure, and control the nozzle solenoid valve to perform target application to the work area based on the PWM control signal.

[0171] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include RAM, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk storage device. Of course, the electronic device may also include other hardware required for other services.

[0172] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0173] Memory is used to store instructions for execution. Specifically, instructions for execution are computer programs that can be executed. Memory can include main memory and non-volatile memory, and it provides the processor with execution instructions and data.

[0174] In one possible implementation, the processor reads the corresponding execution instructions from non-volatile memory into main memory and then executes them. Alternatively, it can obtain the corresponding execution instructions from other devices to form a target-based drug delivery control device based on strong-tracking Kalman filtering at the logical level. The processor executes the execution instructions stored in the memory to implement the target-based drug delivery control method based on strong-tracking Kalman filtering provided in any embodiment of this application.

[0175] The above is as stated in this application. Figure 3The method for implementing a target-based drug delivery control device based on a strong-tracking Kalman filter, as provided in the illustrated embodiment, can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed through integrated logic circuits in the processor's hardware or through software instructions. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.

[0176] The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0177] This application also proposes a readable medium storing execution instructions. When these instructions are executed by a processor of an electronic device, the device can perform a target-based drug delivery control method based on strong tracking Kalman filtering provided in any embodiment of this application, specifically for executing, as... Figure 1 or Figure 2 The method shown.

[0178] The electronic devices in the foregoing embodiments may be computers.

[0179] Those skilled in the art will understand that the embodiments of this application can be provided as methods or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or a combination of software and hardware.

[0180] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0181] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. 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 apparatus that includes that element.

[0182] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A target-based drug delivery control method based on strong tracking Kalman filtering, characterized in that, The method is applied to a pesticide application vehicle, the pesticide application vehicle being equipped with a main pipeline, a backflow pressure regulating valve disposed on the main pipeline, and a nozzle solenoid valve connected to the main pipeline; the method includes: Acquire a continuous image sequence of the work area and the vehicle motion parameters of the spraying vehicle, and determine the target spraying flow rate corresponding to the work area based on the continuous image sequence and the vehicle motion parameters; The real-time flow data and real-time pressure data of the main pipeline are obtained, and the real-time flow data is input into a pre-constructed strong tracking Kalman filter so that the strong tracking Kalman filter calculates the residual sequence based on the real-time flow data and the predicted output value of the strong tracking Kalman filter. Calculate the sequence covariance matrix corresponding to the residual sequence, dynamically solve the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline. A flow feedforward compensation amount is generated based on the first difference between the target application flow rate and the estimated flow rate; a pressure feedback control amount is generated based on the second difference between the preset pressure setpoint and the real-time pressure data. The flow feedforward compensation amount and the pressure feedback control amount are fused to generate a composite control command; the composite control command is sent to the reflux pressure regulating valve to adjust the target pressure of the main pipeline. Based on the target application flow rate and the target pressure, a PWM control signal for the nozzle solenoid valve is generated, and the nozzle solenoid valve is controlled to perform targeted application of the drug to the working area based on the PWM control signal.

2. The method according to claim 1, characterized in that, Determining the target application rate for the work area based on the continuous image sequence and the vehicle motion parameters includes: The continuous image sequence is processed by a pre-trained target detection model to identify and locate weed targets, and the target detection results are output; the target detection results include the bounding box position and area information of each weed target; The target detection results and the vehicle motion parameters are input into the embodied intelligent prediction model to predict the position and area changes of each weed target within a preset time window, thereby obtaining a dynamic target set; the dynamic target set includes the predicted position and predicted area of ​​each weed target; The target application rate is calculated based on the predicted area and the vehicle speed in the vehicle motion parameters.

3. The method according to claim 1, characterized in that, The step of enabling the strong tracking Kalman filter to calculate the residual sequence based on the real-time traffic data and the predicted output value of the strong tracking Kalman filter includes: Determine the state vector of the strong tracking Kalman filter; the state vector is composed of the main pipeline flow rate, the rate of change of flow rate, and the sensor measurement bias. The state vector is recursively processed by the state transition matrix to obtain the state prediction value, and the state prediction value is mapped by the measurement matrix to obtain the prediction output value. The third difference between the real-time traffic data and the predicted output value is calculated to obtain the residual sequence.

4. The method according to claim 3, characterized in that, The step of dynamically calculating the asymptotic decay factor based on the sequence covariance matrix, and using the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the estimated flow rate corresponding to the main pipeline includes: The gradual decay factor is calculated based on the sequence covariance matrix, the preset measurement noise covariance matrix, and the preset process noise covariance matrix. The predicted covariance matrix is ​​multiplied by the asymptotic decay factor and then superimposed with the process noise covariance matrix to obtain the corrected covariance matrix. The Kalman gain is calculated based on the modified covariance matrix, and the state prediction value is corrected using the Kalman gain to obtain the flow prediction value corresponding to the main pipeline.

5. The method according to claim 2, characterized in that, The step of generating a pressure feedback control quantity based on a second difference between a preset pressure setpoint and the real-time pressure data includes: The operating state vector is input into a pre-trained reinforcement learning adaptive controller to obtain a first pressure regulation component; the operating state vector includes the real-time pressure data, the estimated flow rate and the corresponding rate of change of the estimated flow rate, the target drug application flow rate, the number of opening nozzle solenoid valves, and the vehicle speed; The second difference is input to the PID controller to obtain the second pressure regulation component; The first pressure regulation component and the second pressure regulation component are superimposed to obtain the pressure feedback control quantity.

6. The method according to claim 2, characterized in that, The step of generating the PWM control signal for the nozzle solenoid valve based on the target drug flow rate and the target pressure includes: Based on the predicted location and the predicted area, determine the total area of ​​weeds corresponding to the weed target and the coverage area corresponding to each of the nozzle solenoid valves; Based on the proportion of the coverage area to the total area of ​​weeds in the work area, the target application flow rate is allocated to each of the nozzle solenoid valves to obtain the target allocated flow rate corresponding to each of the nozzle solenoid valves; Based on the real-time traffic data, a global traffic correction coefficient is calculated; the global traffic correction coefficient is the ratio of the sum of the traffic allocated to each target to the real-time traffic data. The global flow correction coefficient is used to scale the target allocated flow rate proportionally to obtain the final allocated flow rate corresponding to each nozzle solenoid valve. Based on the preset nozzle flow model, the initial PWM duty cycle of each nozzle solenoid valve is calculated according to the final allocated flow rate and the target pressure. A transient suppression process is applied to the initial PWM duty cycle to obtain the target PWM duty cycle of the nozzle solenoid valve, and the PWM control signal is generated based on the target PWM duty cycle.

7. The method according to claim 6, characterized in that, Before scaling the allocated traffic to each target using the global traffic correction coefficient, the method further includes: The spray boom of the drug application vehicle is divided into multiple coordinated control zones along its length; each coordinated control zone contains at least one of the nozzle solenoid valves; For each of the cooperative control regions, the sum of the target allocated flow rates corresponding to each of the nozzle solenoid valves within the cooperative control region is taken as the region flow rate of the cooperative control region. Calculate the average expected flow value for all the coordinated control areas, and generate the regional coordinated compensation amount corresponding to the coordinated control area based on the fourth difference between the flow of each area and the average expected flow value. According to the preset allocation weight, the regional collaborative compensation amount is superimposed on the corresponding target allocation flow to obtain the corrected allocation flow, and the corrected allocation flow is used as the target allocation flow.

8. A target-based drug delivery control device based on strong tracking Kalman filtering, characterized in that, include: The target application flow rate determination module is configured to acquire a continuous image sequence of the work area and the vehicle motion parameters of the application vehicle, and determine the target application flow rate corresponding to the work area based on the continuous image sequence and the vehicle motion parameters; The residual sequence calculation module is configured to acquire real-time flow data and real-time pressure data of the main pipeline, and input the real-time flow data into a pre-constructed strong tracking Kalman filter so that the strong tracking Kalman filter calculates the residual sequence based on the real-time flow data and the predicted output value of the strong tracking Kalman filter. The traffic flow prediction module is configured to calculate the sequence covariance matrix corresponding to the residual sequence, dynamically solve the asymptotic decay factor based on the sequence covariance matrix, and use the asymptotic decay factor to correct the prediction covariance matrix of the strong tracking Kalman filter to obtain the traffic flow prediction value corresponding to the main pipeline. The composite control module is configured to generate a flow feedforward compensation amount based on a first difference between the target application flow rate and the estimated flow rate; A pressure feedback control quantity is generated based on the second difference between the preset pressure setting value and the real-time pressure data. The composite control command generation module is configured to fuse the flow feedforward compensation amount and the pressure feedback control amount to generate a composite control command; and send the composite control command to the reflux pressure regulating valve to adjust the target pressure of the main pipeline. The execution module is configured to generate a PWM control signal for the nozzle solenoid valve based on the target application flow rate and the target pressure, and to control the nozzle solenoid valve to perform target application to the working area based on the PWM control signal.

9. A computer-readable storage medium storing a computer program, characterized in that, The computer program is used to execute the target drug delivery control method based on strong tracking Kalman filtering as described in any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the target drug delivery control method based on strong tracking Kalman filtering as described in any one of claims 1-7.

Images