Pesticide spraying flow control system based on multi-sensor information fusion

Abstract

This invention provides a pesticide spraying flow control system based on multi-sensor information fusion, comprising a multi-sensor information acquisition unit, a central control unit, and an execution drive unit. The multi-sensor information acquisition unit is used to collect data in real time on the travel speed, terrain slope, pesticide viscosity, and pipeline flow rate during pesticide spraying operations. The central control unit is configured to: perform parallel advance prediction; generate feedforward control quantities based on the advance prediction results and a dynamic compensation model; the dynamic compensation model is configured to perform coupled compensation based on the terrain slope angle and its rate of change, and compensation based on the temperature-calibrated pesticide viscosity; adaptively tune control parameters online and generate feedback control quantities based on the deviation between the real-time pipeline flow rate and the target flow rate; and weightedly fuse the feedforward control quantities and feedback control quantities to generate a total control command. The execution drive unit is used to adjust the opening of the flow regulating valve according to the total control command to achieve closed-loop control of the spraying flow rate.

Description

Technical Field

[0001] This invention belongs to the field of agricultural intelligent equipment and control technology, specifically relating to a pesticide spraying flow control system based on multi-sensor information fusion. Background Technology

[0002] Precision agriculture is an important direction for the development of modern agriculture. Its core lies in achieving efficient resource utilization and ecological environmental protection through refined and intelligent management of the agricultural production process. As a crucial link in plant protection operations, the precision of pesticide spraying flow control directly affects the application effect, pesticide utilization rate, and control costs. Traditional pesticide spraying equipment often uses mechanical adjustment or simple open-loop control methods. The flow rate is significantly affected by factors such as pipeline pressure fluctuations, changes in travel speed, the physical properties of the pesticide solution, and terrain undulations, often resulting in uneven application, over-spraying, or missed spraying, making it difficult to meet the technical requirements of modern precision agriculture for variable-rate pesticide application.

[0003] To improve control accuracy, closed-loop control strategies have been widely adopted in existing technologies. Currently, the mainstream solution for commercially available self-propelled sprayers is to use a metering pump (diaphragm pump or piston pump) as the power source, coupled with a pressure regulating and reflux valve to maintain the pump outlet pressure within a set range, and then adjust the flow rate through a proportional control valve, using a flow meter feedback signal to form a PID closed-loop control. This solution is structurally mature, cost-controllable, and can maintain good flow stability under flat terrain and uniform operating conditions.

[0004] However, in the complex field operation environment, the above solution reveals obvious technical bottlenecks, mainly in the handling of the following three types of disturbances:

[0005] First, there is the rapid change in travel speed. The amount of pesticide applied per unit area is determined by the ratio of flow rate to travel speed, and a rapid, synchronous response in flow rate is required when speed changes. However, PID control is essentially a passive correction of existing deviations. When speed changes abruptly, the flow rate deviation has already occurred, and the controller needs a full adjustment cycle to bring the flow rate back to the target value, resulting in significant dynamic tracking lag.

[0006] Second, the impact of terrain slope disturbances. Some researchers believe that pressure regulating return valves can maintain constant system pressure, and slope disturbances can be ignored. However, the actual situation is more complex: on the one hand, pressure regulating return valves have inherent response lag. When the sprayer quickly crosses field ridges or enters / exits slopes, the pressure before the valve fluctuates transiently within hundreds of milliseconds. The pressure regulating valve cannot respond in time, causing the actual pressure difference across the proportional control valve to deviate from the design operating point, resulting in transient disturbances in the flow rate. On the other hand, modern self-propelled sprayers can have a boom width of over 12 meters. When operating on slopes, there is a significant height difference between the two ends of the boom and the pump source. The gravitational component of the pesticide column generates additional hydrostatic pressure in the pipeline, causing uneven actual flow distribution at each nozzle. These two types of slope disturbances are particularly prominent in complex terrains such as hills and mountains, and are important factors affecting the uniformity of pesticide application across all terrains.

[0007] Third, the dynamic changes in pesticide solution viscosity. The viscosity of pesticide solutions varies significantly depending on the type, ratio, and temperature. Furthermore, the viscosity-temperature characteristics of the pesticide solution slowly drift during continuous operation due to factors such as evaporation and concentration. Viscosity changes directly affect the actual flow rate at the same valve opening degree, making it difficult for a fixed-parameter PID controller to adapt to the dynamic drift in viscosity.

[0008] More importantly, the three types of disturbances mentioned above often coexist and are coupled in actual field operations, further weakening the performance of traditional PID control. Increasing the controller gain in pursuit of fast response can easily lead to flow overshoot and continuous oscillation under the coupling of multiple disturbances; while conservative parameter settings will make the system response sluggish, making it difficult to balance dynamic performance and stability.

[0009] Some studies have attempted to introduce feedforward compensation, using the detected travel speed to calculate the theoretical flow rate as a feedforward quantity to improve tracking performance when speed changes. However, such methods typically only consider speed as a single disturbance factor, without systematically integrating multi-source information such as slope and viscosity. Furthermore, the feedforward model fails to predict the arrival of disturbances, thus still suffering from the inherent limitation of compensation lag.

[0010] Existing pesticide spraying flow control systems still suffer from technical bottlenecks when dealing with complex and ever-changing field conditions, such as slow dynamic response, insufficient compensation for slope disturbances, and poor robustness against multi-disturbance coupling. Developing a flow control system capable of proactively predicting multi-dimensional disturbances and achieving coordinated proactive compensation and precise closed-loop regulation is of significant practical necessity and application value for improving the technical level of precision spraying equipment and promoting pesticide reduction and efficiency. Summary of the Invention

[0011] To address the shortcomings and deficiencies of existing technologies, this invention provides a pesticide spraying flow control system and method based on multi-sensor information fusion. It aims to solve the core technical problems of traditional flow control schemes for self-propelled sprayers operating in hilly terrain, such as dynamic response lag, insufficient slope disturbance compensation, and poor robustness under multi-source disturbance coupling. This invention employs a predictive-feedforward-feedback collaborative control architecture. A multi-sensor information acquisition unit acquires real-time multi-dimensional operating condition information during spraying operations, including travel speed, terrain slope, pesticide viscosity, and pipeline flow rate. A multi-dimensional time-series prediction module performs short-term advance predictions of travel speed, terrain slope, and pesticide viscosity to match the response delay of the actuators, outputting the advance prediction values. A feedforward controller, combined with a dynamic compensation model, actively compensates for speed changes, slope disturbances, and pesticide viscosity drift. The dynamic compensation model integrates the static influence of the slope angle with the dynamic influence of the slope change rate to achieve coupled compensation, and completes adaptation compensation based on the temperature-calibrated pesticide viscosity. An adaptive feedback controller with online parameter self-tuning accurately corrects the remaining flow error after feedforward compensation. Finally, a control quantity fusion module weights and fuses the feedforward and feedback control quantities according to the dynamically adjusted weights of the operating conditions to generate a total control command that drives the actuator to adjust the opening of the flow regulating valve, achieving precise closed-loop control of the spraying flow.

[0012] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0013] A pesticide spraying flow control system based on multi-sensor information fusion includes a multi-sensor information acquisition unit, a central control unit, and an execution drive unit;

[0014] The multi-sensor information acquisition unit is used to collect data in real time on the travel speed, terrain slope, pesticide viscosity, and pipeline flow rate during pesticide spraying operations.

[0015] The central control unit is signal-connected to the multi-sensor information acquisition unit and the execution drive unit, respectively. The central control unit is configured as follows:

[0016] Based on the historical time-series data of travel speed, terrain slope, and drug viscosity, the travel speed, terrain slope, and drug viscosity at a preset time in the future are predicted in parallel. The preset time is not shorter than the response delay of the execution drive unit.

[0017] Based on the advanced prediction results, a feedforward control quantity is generated by combining the dynamic compensation model. The dynamic compensation model is configured to perform coupled compensation based on the terrain slope angle and its rate of change, and to perform compensation based on the temperature-calibrated viscosity of the liquid.

[0018] Based on the deviation between the real-time flow rate and the target flow rate in the pipeline, the control parameters are adaptively tuned online and feedback control quantities are generated.

[0019] The feedforward control quantity and the feedback control quantity are weighted and fused together. The fusion weight can be dynamically adjusted to generate a total control command and output it to the execution drive unit.

[0020] The execution drive unit is used to adjust the opening of the flow regulating valve according to the overall control command to achieve closed-loop control of the spray flow.

[0021] Furthermore, the central control unit uses a speed prediction submodule, a slope prediction submodule, and a viscosity prediction submodule to respectively achieve advanced prediction of travel speed, terrain slope, and drug viscosity;

[0022] The speed prediction submodule adopts an adaptive linear prediction algorithm with a forgetting factor, establishes an autoregressive model based on historical speed sequences, and updates the model parameters online using the recursive least squares method.

[0023] The slope prediction submodule uses the Kalman filter algorithm to establish a state equation with the slope angle and its rate of change as state variables, and integrates attitude observations to achieve optimal estimation and advance prediction of the slope.

[0024] The viscosity prediction submodule is based on the Andrade viscosity-temperature physical model. It uses a recursive least squares method with a forgetting factor to identify viscosity-temperature characteristic parameters online and outputs an advanced prediction value of the drug solution viscosity by superimposing a historical residual trend correction term.

[0025] Furthermore, in the dynamic compensation model, the coupled compensation of the terrain slope angle and its rate of change is achieved through a slope compensation coefficient. This slope compensation coefficient is calculated by integrating the static influence of the slope angle and the dynamic influence of the slope rate of change, and the calculation formula is as follows:

[0026]

[0027] Where α is the slope compensation coefficient. This is the static slope compensation coefficient. θ is the dynamic slope compensation coefficient, θ is the real-time slope angle, θ̇ is the slope change rate, and sgn is the sign function. The sign function takes a value of 1 when the slope change rate is greater than 0 (corresponding to an uphill scenario) and a value of -1 when the slope change rate is less than 0 (corresponding to a downhill scenario). and Identified through calibration experiments.

[0028] Furthermore, in the dynamic compensation model, the viscosity compensation of the liquid medicine is achieved through a viscosity compensation coefficient, which is calculated based on the ratio of the equivalent viscosity of the liquid medicine calibrated to a preset standard temperature to the reference viscosity; the calibration of the equivalent viscosity of the liquid medicine is completed based on the difference between the measured viscosity of the liquid medicine, the real-time temperature of the liquid medicine, and the preset standard temperature.

[0029] Furthermore, the central control unit first calculates the conventional feedforward quantity based on the target application rate per unit area, real-time travel speed, slope compensation coefficient, and viscosity compensation coefficient, and then weights and fuses the conventional feedforward quantity with the predicted feedforward quantity calculated based on the advanced prediction results to generate the final feedforward control quantity.

[0030] The final formula for calculating the feedforward control quantity is:

[0031]

[0032] in, For the final feedforward control quantity, To predict feedforward quantities, η is the conventional feedforward quantity, and η is the prediction confidence coefficient, which can be dynamically adjusted based on historical prediction errors.

[0033] Furthermore, the central control unit uses an improved PID algorithm to generate feedback control quantity, taking the deviation between the real-time flow rate and the target flow rate in the pipeline, as well as the rate of change of the deviation, as input quantities, and online adaptively tuning the proportional coefficient, integral coefficient, and derivative coefficient of the PID algorithm through fuzzy rules;

[0034] The fuzzy rules pre-set two levels of explicit reference benchmarks: the target flow rate is used as the first reference benchmark to classify the deviation into continuous intervals, and the deviation change rate is used as the system's preset nominal threshold for deviation change rate to classify the deviation change rate into continuous intervals. Each classification includes three order of magnitude intervals: high, medium, and low. The corresponding tuning strategies include:

[0035] When the deviation is in the high order of magnitude range, increase the proportional coefficient to speed up the system response, while limiting the integral coefficient to avoid integral saturation;

[0036] When the deviation is in the middle range, decrease the proportional coefficient while gradually increasing the integral coefficient to eliminate the steady-state error of the system.

[0037] When the deviation is in the low order of magnitude range, increase the integral coefficient and the derivative coefficient to improve the steady-state control accuracy of the system and suppress high-frequency disturbances;

[0038] When the rate of change of deviation is in the high order of magnitude range, increase the differential coefficient to suppress flow overshoot.

[0039] Furthermore, the central control unit can dynamically adjust the weighting coefficients of the weighted fusion according to the working conditions. Under conditions where the slope and the viscosity of the liquid change drastically, the weighting coefficients of the feedforward control quantity are increased, while under conditions where the system tends to operate in a steady state, the weighting coefficients of the feedforward control quantity are decreased.

[0040] Furthermore, the central control unit also includes a working condition identification module;

[0041] The working condition identification module is used to analyze the characteristics of the working scene based on the real-time data acquired by the multi-sensor information acquisition unit, and divide the field working conditions into stable working mode, transient change mode, slope transition mode and strong interference mode. Based on the divided working condition modes, the control strategies and parameter constraints of the advance prediction, feedforward compensation, feedback tuning and weighted fusion links are dynamically adjusted.

[0042] Furthermore, the actuation drive unit includes a proportional control valve and a valve position feedback sensor;

[0043] The control signal input terminal of the proportional regulating valve is connected to the output terminal of the central control unit, and is used to receive the overall control command and adjust the valve core opening.

[0044] The valve position feedback sensor is used to detect the actual opening degree of the proportional control valve in real time and feed the detection result back to the central control unit to form a closed-loop control of the valve position, thereby eliminating the influence of actuator hysteresis and dead zone on the flow control accuracy.

[0045] Furthermore, the multi-sensor information acquisition unit includes a speed sensor, an attitude sensor, a viscosity sensor, a flow meter, and a pressure transmitter;

[0046] The speed sensor is used to collect the travel speed of the spraying operation, the attitude sensor is used to collect the terrain slope and slope change rate, the viscosity sensor integrates a temperature acquisition module to collect the viscosity and temperature of the pesticide solution, the flow meter is used to collect the real-time flow of the pipeline, and the pressure transmitter is used to collect the pressure of the spraying pipeline system.

[0047] Compared to existing technologies, this invention and its preferred scheme adopt a predictive-feedforward-feedback collaborative control architecture, breaking through the inherent limitations of passive correction in traditional PID control. By matching the response delay of the actuator with multi-dimensional time-series advance prediction, it achieves proactive advance compensation for multi-source disturbances in field operations, effectively solving the core problem of dynamic response lag in traditional schemes, significantly improving the dynamic tracking performance of flow control, and reducing flow overshoot and fluctuations. This invention achieves comprehensive perception of all-dimensional operating conditions such as travel speed, terrain slope, and pesticide viscosity through multi-sensor information fusion. Addressing the core pain points of hilly terrain operations, it designs a coupled compensation mechanism that integrates the static influence of slope angle and the dynamic influence of slope change rate, while simultaneously employing a pesticide viscosity adaptation compensation strategy based on temperature calibration. This comprehensively covers the compensation needs for multi-source coupled disturbances in field operations, solving the problems of insufficient slope disturbance compensation and poor adaptability to pesticide viscosity drift in existing schemes. It significantly improves the uniformity of pesticide application under complex terrain and variable pesticide conditions, effectively... This invention reduces the problems of overspraying and missed spraying. Through adaptive feedback control with online parameter self-tuning and a feedforward-feedback fusion strategy adapted to operating conditions, the control parameters and fusion weights are dynamically adjusted according to the operating conditions. This balances the system's anti-disturbance capability in strong disturbance scenarios with control accuracy in steady-state operating scenarios, avoiding the shortcomings of traditional fixed-parameter control that cannot balance response speed and system stability. It significantly improves the system's robustness and environmental adaptability under multi-disturbance coupled operating conditions. This invention forms a full-link closed-loop control architecture from operating condition perception, disturbance prediction, active compensation to precise adjustment, which is suitable for the actual operating needs of self-propelled sprayers in hilly areas. It effectively improves the control accuracy of variable-rate spraying, helps to achieve pesticide reduction and efficiency improvement, and is in line with the development direction of precision agriculture. The valve position inner loop closed-loop control in its preferred scheme can effectively eliminate the impact of actuator hysteresis and dead zone on control accuracy. The operating condition identification module can realize intelligent adaptation to different operating scenarios, further enhancing the system's control performance and practical value. Attached Figure Description

[0048] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0049] Figure 1 This is a schematic diagram of the principle structure of the pesticide spraying flow control system based on multi-sensor information fusion in an embodiment of the present invention.

[0050] Figure 2 This is the main control flowchart of the pesticide spraying flow control method in this embodiment of the invention;

[0051] Figure 3 This is a flowchart of the calculation process of the feedforward controller in an embodiment of the present invention;

[0052] Figure 4This is a parameter self-tuning logic diagram of the adaptive feedback controller in an embodiment of the present invention;

[0053] Figure 5 This is a schematic diagram illustrating the working principle of the control quantity fusion module in an embodiment of the present invention;

[0054] Figure 6 The figures are comparison images of the grayscale images of water-sensitive paper in the embodiments of the present invention; in the figures, (a) is the grayscale image of water-sensitive paper under the traditional PID control scheme, (b) is the grayscale image of water-sensitive paper under the speed feedforward + PID control scheme, and (c) is the grayscale image of water-sensitive paper under the collaborative control scheme of the present invention. Detailed Implementation

[0055] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:

[0056] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0057] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0058] To address the technical problems of traditional PID control in the operation of self-propelled sprayers in hilly terrain, such as dynamic response lag, insufficient slope disturbance compensation, and poor robustness of multi-source disturbance coupling, this invention proposes a predictive-feedforward-feedback collaborative control architecture and method based on multi-sensor information fusion. The system collects multi-dimensional operating condition information in real time through speed sensors, attitude sensors, viscosity sensors, flow meters, and pressure transmitters. The multi-dimensional time-series prediction module uses adaptive linear prediction, Kalman filtering, and adaptive Andrade parameter identification methods to make short-term advance predictions of future travel speed, terrain slope, and pesticide viscosity. The feedforward controller combines a dynamic compensation model with the advance prediction values ​​to achieve proactive advance compensation for speed changes, slope disturbances, and viscosity drift. The adaptive feedback controller uses an improved PID algorithm with online parameter self-tuning to accurately correct residual errors. The control quantity fusion module weightedly fuses the feedforward and feedback to generate a total control command, driving a proportional control valve to achieve closed-loop flow control, thereby achieving proactive advance compensation and precise closed-loop regulation of multiple disturbances such as travel speed, terrain slope, and pesticide viscosity.

[0059] The present invention provides a pesticide spraying flow control system and method based on multi-sensor information fusion, the specific technical contents of which mainly include:

[0060] I. Pesticide Spraying Flow Control System

[0061] The pesticide spraying flow control system based on multi-sensor information fusion of the present invention includes a multi-sensor information acquisition unit, a central control unit, an execution drive unit, and a human-machine interaction unit.

[0062] The multi-sensor information acquisition unit, which communicates with the central control unit, includes a speed sensor, an attitude sensor, a viscosity sensor, a flow meter, and a pressure transmitter, which are used to collect data in real time on the sprayer's travel speed, terrain slope, pesticide viscosity, pipeline flow rate, and system pressure during the spraying operation.

[0063] The central control unit is the core of the system's control and is connected to the multi-sensor information acquisition unit, the execution drive unit, and the human-machine interaction unit. It includes a multi-dimensional time-series prediction module, a feedforward controller, an adaptive feedback controller, and a control quantity fusion module.

[0064] The multi-dimensional time-series prediction module is used to make short-term predictions of future travel speed, terrain slope, and liquid viscosity based on historical sensor data sequences, and outputs advanced prediction values. The multi-dimensional time-series prediction module includes a speed prediction submodule, a slope prediction submodule, and a viscosity prediction submodule.

[0065] The speed prediction submodule adopts an adaptive linear prediction algorithm with a forgetting factor, establishes an autoregressive model based on historical speed sequences, and updates the model parameters online through recursive least squares method to achieve multi-step advanced speed prediction.

[0066] The slope prediction submodule uses the Kalman filter algorithm to establish a state equation with the slope angle and its rate of change as state variables, and integrates the observations from the attitude sensor to achieve optimal estimation and advance prediction of the slope.

[0067] The viscosity prediction submodule is based on the Andrade viscosity-temperature physical model. It uses the recursive least squares method with a forgetting factor to identify viscosity-temperature characteristic parameters online and adds a historical residual trend correction term to achieve adaptive tracking of the viscosity-temperature characteristics drift of the drug solution and output the viscosity prediction value in advance.

[0068] The feedforward controller calculates the feedforward control quantity based on the predicted value and the dynamic compensation model, thereby achieving proactive anticipatory compensation for upcoming disturbances. The feedforward controller has a built-in dynamic compensation model, which includes slope compensation coefficients and viscosity compensation coefficients, as detailed below:

[0069] 1. The slope compensation coefficient, taking into account both the static influence of the slope angle and the dynamic influence of the slope change rate, is calculated using the following formula:

[0070]

[0071]

[0072] In the formula, This is the slope compensation coefficient. This is the static slope compensation coefficient, reflecting the influence of the slope angle itself; The dynamic slope compensation coefficient reflects the rate of slope change. The effect of this is used to suppress flow overshoot or undershoot caused by sudden changes in slope. Parameters , The flow rate was identified through a flow rate calibration experiment on a standard ramp.

[0073] 2. Viscosity compensation coefficient, calculated based on the ratio of the temperature-calibrated drug solution viscosity to the reference viscosity, using the following formula:

[0074]

[0075] in, This is the viscosity compensation coefficient. To convert to the equivalent viscosity at a standard temperature of 25°C, As the reference viscosity, =0.3~0.5 is the crop type coefficient.

[0076] The feedforward controller first calculates the conventional feedforward quantity based on the dynamic compensation model. The conventional feedforward quantity is calculated according to the following formula:

[0077]

[0078] in, For conventional feedforward parameters, A represents the target application rate per unit area, and V represents the real-time travel speed. This is the slope compensation coefficient. This is the viscosity compensation coefficient.

[0079] The feedforward controller weighted and fused the advanced prediction value with the regular feedforward quantity to generate the final feedforward control quantity, which is calculated according to the following formula:

[0080]

[0081] in, For the final feedforward control quantity, This is an advanced forecast value. This is a standard feedforward quantity. The confidence coefficient is dynamically adjusted based on historical prediction errors, with a value range of 0.5 to 0.9.

[0082] An adaptive feedback controller is used to adaptively adjust control parameters online based on the deviation between real-time flow and target flow, outputting a feedback control quantity to accurately correct the residual error after feedforward compensation and unmodeled disturbances. The adaptive feedback controller employs an improved PID algorithm, and its control quantity output formula is as follows:

[0083]

[0084] in, Let be the feedback control quantity at time t. The deviation between real-time traffic and target traffic. , , These are the proportional, integral, and differential coefficients, which are updated in real time.

[0085] The adaptive feedback controller has a built-in parameter self-tuning module, which adjusts the parameters based on the deviation. and its rate of change The real-time value is adjusted online based on the following fuzzy rules. , , :

[0086] when When it is large, increase To expedite the response while limiting To prevent the points from becoming saturated;

[0087] when When it is in the medium range, reduce appropriately. and gradually introduce To eliminate steady-state error;

[0088] when When it is small, increase and To improve steady-state accuracy and suppress high-frequency disturbances;

[0089] when When it is large, increase In order to suppress overshoot.

[0090] The control quantity fusion module is used to weightedly fuse the feedforward control quantity and the feedback control quantity to generate the total control command. The control quantity fusion module weightedly adds the feedforward control quantity and the feedback control quantity according to the following formula to generate the total control command:

[0091]

[0092] in, This is the feedforward weighting coefficient, ranging from 0.6 to 0.8, dynamically adjusted according to system operating conditions; it is increased when there are drastic changes in slope and viscosity. Value, enhance feedforward effect; reduce when system tends to stabilize. This value enhances the accuracy of feedback adjustment.

[0093] As a preferred embodiment of the present invention, the central control unit further includes a working condition identification module, which is used to analyze the characteristics of the working scene in real time based on multi-sensor information, divide the field working conditions into stable working mode, transient change mode, slope transition mode and strong interference mode, and dynamically adjust the control strategies and parameter constraints of each module accordingly.

[0094] The actuator drive unit, connected to the central control unit, is used to adjust the opening of the proportional control valve according to the overall control command, so as to achieve precise closed-loop control of the spray flow. The actuator drive unit includes a proportional control valve and a valve position feedback sensor: the control signal input terminal of the proportional control valve is connected to the output terminal of the central control unit; the valve position feedback sensor is used to detect the actual opening of the proportional control valve in real time and feed it back to the central control unit to form a valve position closed-loop control, so as to eliminate the influence of actuator hysteresis and dead zone on the flow control accuracy.

[0095] The human-machine interface unit (HMI) communicates with the central control unit and is used to set target parameters, display real-time operating conditions and system status, and support historical data storage and retrieval. The HMI includes a touch screen and a data storage module: the touch screen is used to set the target application rate, travel speed limit, and crop type parameters, and displays flow rate, pressure, speed, slope, viscosity, valve position, and system alarm information in real time; the data storage module is used to record key operating condition data and control command sequences during the operation.

[0096] As a preferred embodiment of the present invention, the system further includes a wireless communication module and a fault diagnosis and alarm module:

[0097] The wireless communication module is used to upload real-time operating data and control status to the remote monitoring platform, and to receive control commands or parameter updates issued by the platform.

[0098] The fault diagnosis and alarm module is used to monitor the working status of each sensor and actuator in real time. When an abnormal signal, communication interruption or control error is detected, an audible and visual alarm is activated, and the proportional control valve is closed and the shutdown protection action is executed.

[0099] II. Methods for Controlling Pesticide Spraying Flow Rate

[0100] The present invention also provides a method for controlling pesticide spraying flow rate based on the above system, comprising the following steps:

[0101] Step S1: The sprayer's travel speed, terrain slope, pesticide viscosity, pipeline flow rate, and system pressure are acquired in real time through the multi-sensor information acquisition unit;

[0102] Step S2: The multidimensional time series prediction module makes short-term predictions of future travel speed, terrain slope and drug viscosity based on historical sensor data sequences, and outputs advanced prediction values.

[0103] Step S3: The feedforward controller calculates the conventional feedforward quantity based on the dynamic compensation model, and then weights and fuses the conventional feedforward quantity with the advanced prediction value to obtain the feedforward control quantity;

[0104] Step S4: The adaptive feedback controller adjusts the PID parameters online based on the deviation between the real-time flow and the target flow and its rate of change, and outputs the feedback control quantity.

[0105] Step S5: The control quantity fusion module combines the operating condition identification results to weightedly fuse the feedforward control quantity and the feedback control quantity to generate the total control command;

[0106] Step S6: The drive unit adjusts the proportional valve opening according to the overall control command, and achieves precise closed-loop control of the spray flow rate through valve position closed-loop auxiliary control;

[0107] Step S7: The human-machine interaction unit displays the system status in real time and supports setting target parameters and querying historical operation data.

[0108] The above-mentioned solution of this invention uses a multi-dimensional time-series prediction module to predict speed, slope, and viscosity in advance, and combines it with a dynamic compensation model to achieve proactive compensation for multi-source disturbances, overcoming the lag defect of traditional PID control's "post-correction". Experimental results show that, compared with traditional constant-parameter PID control, this invention reduces flow overshoot by about 57% and shortens the settling time to 1.3s under speed step conditions; reduces flow fluctuation amplitude during the transition period by about 45% under slope transition conditions; and achieves a flow error of less than 5% for 91.6% of the time under the combined conditions of speed changes, terrain undulations, and viscosity fluctuations, significantly improving system performance and making it suitable for self-propelled sprayer platforms in various hilly areas.

[0109] The specific implementation examples of the technical solution of the present invention will be described in detail below with reference to the accompanying drawings of the specification of the present invention. It should be understood that the specific examples described herein are only for explaining the present invention and are not intended to limit the present invention.

[0110] This implementation example provides a complete solution for a pesticide spraying flow control system based on multi-sensor information fusion. The application scenario is rice paddy plant protection in hilly areas. It is mounted on a self-propelled sprayer platform, with a spray width of 12 m and a tank capacity of 200 L. The system composition and signal connections are as follows: Figure 1 As shown. The core control flow of the system follows... Figure 2 The main control flowchart shown is combined with... Figure 3 , Figure 4 , Figure 5 The detailed logic of its operation.

[0111] exist Figure 1 In this system, the output terminals of each sensor in the multi-sensor information acquisition unit are connected to the signal input terminal of the central control unit: speed sensor, attitude sensor, viscosity sensor, flow meter, and pressure transmitter transmit the collected signals of travel speed, terrain slope, liquid viscosity, pipeline flow, and system pressure to the signal preprocessing module of the central control unit, respectively. The control signal output terminal of the central control unit is connected to the proportional regulating valve control terminal of the execution drive unit, and the valve position feedback sensor output terminal of the execution drive unit is connected to the feedback signal input terminal of the central control unit, forming a closed-loop control of the valve position. The central control unit has a bidirectional communication connection with the human-machine interaction unit, and also has a bidirectional connection with the wireless communication module and the fault diagnosis and alarm module to realize parameter distribution, operating status upload, and abnormal shutdown protection.

[0112] like Figure 2 As shown, after the system powers on and completes hardware initialization, preset parameter loading, and sensor self-test, it enters a periodic interrupt main control loop with a control cycle of Ts=10ms. Within each control cycle, the following processes are executed sequentially:

[0113] Step 1: Data acquisition and preprocessing. Analog signals from the flow meter, pressure transmitter, and viscosity sensor are acquired through the ADC interface. Travel speed data is acquired through the CAN bus. Slope data from the IMU attitude sensor is read through the SPI interface. The raw acquired data is processed by digital low-pass filtering to suppress high-frequency sensor noise.

[0114] Step 2: Multidimensional time series advance prediction, parallel operation of speed prediction submodule, slope prediction submodule, and viscosity prediction submodule, respectively output the advance prediction values ​​of speed, slope, and viscosity at the next τ=30ms time, and synchronously update the online identification parameters of each submodule;

[0115] Step 3: Calculate the feedforward control quantity. Calculate the conventional feedforward quantity using real-time sensor data and the predicted feedforward quantity using advanced prediction values. After weighted fusion using prediction confidence coefficients, the final feedforward control quantity is output.

[0116] Step 4: Calculate the adaptive feedback control quantity. Calculate the deviation e(t) between the real-time flow and the target flow and the rate of change of the deviation ė(t). After updating the PID coefficients through the fuzzy parameter self-tuning module, calculate the output feedback control quantity.

[0117] Step 5: Operating condition identification and control quantity fusion. Based on multi-sensor data, the current operating condition mode is identified, the feedforward-feedback fusion weight γ is dynamically adjusted, and the feedforward control quantity and feedback control quantity are weighted and superimposed to generate the final total control quantity.

[0118] Step 6: Execute output and data storage. The total control quantity is converted into a PWM duty cycle signal after amplitude limiting and output to the proportional valve drive circuit. At the same time, the valve position feedback signal is collected to realize local closed-loop control. The working condition data and control commands are refreshed synchronously in the human-machine interaction unit and stored in the local SD card at a frequency of 10Hz.

[0119] The feedforward control quantity fusion logic in this embodiment is as follows: Figure 3 As shown, a dual-path fusion architecture combining real-time feedforward and advanced prediction feedforward is adopted: the first path is the real-time feedforward path, which inputs the real-time travel speed, real-time slope compensation coefficient, real-time viscosity compensation coefficient and target application rate per unit area to calculate the conventional feedforward quantity; the second path is the advanced prediction feedforward path, which inputs the advanced prediction values ​​of speed, slope, and viscosity to calculate the predicted feedforward quantity; the two input quantities are weighted and fused by the prediction confidence coefficient η, and the final feedforward control quantity is output to the control quantity fusion module.

[0120] The fuzzy PID parameter self-tuning logic of the adaptive feedback controller in this embodiment is as follows: Figure 4 As shown, using the flow deviation |e(t)| and the deviation change rate |ė(t)| as input quantities, after fuzzification processing, fuzzy inference is performed based on a preset fuzzy rule base. After defuzzification processing, the online correction quantities of the three PID parameters are output, and the proportional coefficient K is updated in real time. p Integral coefficient K i Differential coefficient K d The final output is a feedback control quantity.

[0121] The logic for fusing operating condition identification and control quantities in this embodiment is as follows: Figure 5 As shown, firstly, four characteristic quantities—speed fluctuation rate, slope change rate, viscosity change rate, and flow deviation—are input. The operating condition identification module determines the current operating mode: stable operation mode, transient change mode, ramp transition mode, and strong interference mode. Then, a corresponding feedforward weighting coefficient γ is matched according to the operating condition mode. Specifically, γ = 0.75~0.8 is used for ramp transition mode and transient change mode, γ = 0.7~0.75 is used for strong interference mode, and γ = 0.6~0.65 is used for stable operation mode. Finally, the feedforward control quantity and feedback control quantity are weighted and superimposed according to the weighting coefficient to generate a total control command and output it to the execution drive unit.

[0122] The following is a more detailed explanation:

[0123] I. System Hardware Components

[0124] (1) Multi-sensor information acquisition unit

[0125] The speed sensor is a non-contact radar speed sensor, installed at the bottom of the front end of the sprayer, with a range of 0~30km / h and an accuracy of ±0.1 km / h. The output signal is transmitted to the central control unit via the CAN bus.

[0126] The attitude sensor employs an inertial measurement unit (IMU) integrating a three-axis accelerometer and a three-axis gyroscope, which is fixedly installed near the center of gravity of the spraying vehicle frame to calculate the terrain slope angle in real time. The slope change rate dθ / dt is measured, the angle measurement accuracy is ±0.1°, and the data is output at a frequency of 100 Hz via the SPI interface.

[0127] The viscosity sensor uses an online vibratory viscometer with an integrated PT100 platinum resistance temperature probe, installed between the outlet of the main pipeline pump and the proportional control valve to measure the dynamic viscosity of the drug solution in real time. With temperature T, viscosity range 0.520 mPa·s, temperature range 0~60℃, output 4~20 mA analog signal.

[0128] The flow meter is an electromagnetic flow meter, installed on the spray line downstream of the proportional control valve, with a range of 0~20 L / min, an accuracy of ±0.5%, and an output of 4~20 mA analog signal.

[0129] The pressure transmitter uses a diffused silicon pressure sensor, installed on the pump outlet pipeline, with a range of 0~16 MPa, an accuracy of ±0.25%, and an output of 4~20 mA analog signal.

[0130] (2) Central control unit

[0131] The core processor uses the STM32H743 series high-performance embedded microcontroller with a main frequency of 480 MHz. It is equipped with a multi-channel ADC module for acquiring 4~20 mA analog signals, a CAN communication interface for receiving speed sensor data, and an SPI communication interface for receiving IMU data. The central control unit implements a multi-dimensional timing prediction module, a feedforward controller, an adaptive feedback controller, a working condition identification module, and a control quantity fusion module at the software level.

[0132] (3) Execution drive unit

[0133] The proportional control valve is a direct-acting electromagnetic proportional control valve, where the valve core stroke is linearly proportional to the input control current, with a response time of less than 30 ms. An LVDT valve position sensor is built into the valve body, providing real-time feedback on the actual valve core position, forming a closed-loop valve position control and eliminating the effects of electromagnetic hysteresis and dead zone. The drive circuit uses an H-bridge power drive, receiving the PWM signal (frequency 20 kHz) output from the central control unit and converting it into a current signal to drive the proportional valve coil.

[0134] (4) Human-computer interaction unit

[0135] It adopts a 7-inch industrial touch screen and communicates with the central control unit via an RS485 interface. This allows for setting parameters such as target application rate, crop type, and maximum application speed, and displays real-time data on flow rate, pressure, speed, slope, viscosity, valve position, and system alarms. A built-in SD card data storage module records all operational data at a 10 Hz frequency.

[0136] II. Algorithm Implementation of Each Software Module

[0137] (1) Velocity prediction submodule

[0138] Vehicle motion is continuous, and speed changes are limited by acceleration, so there will be no abrupt changes. Therefore, it is suitable to use time series models to fit trends and extrapolate predictions.

[0139] Let the current time be t, and the historical velocity sequence be: V(t), V(t-1), V(t-2), ..., V(t-n+1), for a total of n historical data.

[0140] Prediction is performed using a p-order autoregressive model AR(p):

[0141]

[0142] in, These are the autoregressive coefficients. This represents the prediction error.

[0143] To adapt to changes in motion characteristics under different working conditions, the recursive least squares method with a forgetting factor is used to update the autoregressive coefficients online.

[0144]

[0145]

[0146]

[0147] in, Let be the gain matrix at time t. These are the autoregressive coefficients. Let φ(t) be the vector of autoregressive coefficients to be estimated, and let φ(t) be the vector of historical data. The forgetting factor (usually set to 0.95~0.99) controls the decay rate of historical data.

[0148] To compensate for system latency, multi-step advance prediction is required. An iterative method is used for multi-step prediction, calculating the next 1 step, 2 steps, and so on sequentially. Predicted velocity value at time In this embodiment, p=3 and λ=0.97, with the number of lead steps corresponding to the lead time. =30ms.

[0149] (2) Slope prediction submodule

[0150] Slope changes relatively slowly but are random. Kalman filtering can integrate motion models and sensor observations to achieve optimal state estimation and advance prediction.

[0151] Using the slope θ and its rate of change θ̇ as state variables, a discrete-time state equation is established:

[0152]

[0153] in, The sampling period is This is process noise, describing the randomness of slope changes.

[0154] Observation equation:

[0155]

[0156] in, The slope value is the actual value measured by the attitude sensor. To observe noise.

[0157] Kalman filtering involves two steps: prediction and update.

[0158] Prediction Step:

[0159]

[0160]

[0161] Update steps:

[0162]

[0163]

[0164]

[0165] in, Here is the state transition matrix. For the observation matrix, For process noise covariance, To observe the noise covariance.

[0166] Based on the filtered state estimate, multi-step look-ahead prediction is performed using the state transition matrix:

[0167]

[0168] Thus, the predicted slope value at time τ is obtained. .

[0169] (3) Viscosity prediction submodule

[0170] The viscosity of most pesticide solutions exhibits an exponential relationship with temperature, described by the Andrade equation:

[0171]

[0172] in, Viscosity at reference temperature For activation energy, Let be the gas constant. T is the absolute temperature. Traditionally, is used... , As a fixed constant, it is impossible to track the viscosity-temperature characteristic drift caused by evaporation and concentration during different drug solutions or operations. Therefore, this invention introduces a recursive least squares online parameter identification mechanism with a forgetting factor, implemented in the following three steps:

[0173] Step 1: Linearization of the physical model

[0174] Taking the natural logarithm of both sides of Andrade's formula, we get:

[0175]

[0176] make parameter vector Regression vector Then the Andrade model is transformed into a standard linear regression form:

[0177]

[0178] in, This is used to model residuals. This linearization operation ensures that the parameters of the physical model can be identified online subsequently.

[0179] Step 2: Online identification of recursive parameters with forgetting factor

[0180] At each sampling time, the viscosity sensor measures the value. and temperature sensor measurement value The parameter vector is updated online according to the following recursive formula. :

[0181]

[0182]

[0183]

[0184] in, It is the gain vector; Estimate the covariance matrix for the parameters; ∈[0.95, 0.99] is the forgetting factor, which controls the decay rate of historical data. The smaller the value, the faster the model responds to recent data, making it suitable for scenarios where the properties of the drug solution change rapidly; The larger the value, the more stable the identification results, making it suitable for scenarios where properties drift slowly. This mechanism enables parameter identification to continuously track viscosity-temperature characteristic drift caused by drug solution changes, evaporation and concentration, or large temperature variations.

[0185] Step 3: Adaptive Viscosity Prediction

[0186] Based on parameters obtained in real time Based on the temperature trend output by the slope prediction submodule, the advanced predicted viscosity of the physical model is calculated:

[0187]

[0188] Based on the physical prediction, a trend correction term obtained by fitting a low-order polynomial to the historical viscosity residual sequence is superimposed. The final viscosity prediction value is obtained as follows:

[0189]

[0190] in For the future Viscosity prediction at time [time] ∈ (0, 1] is the trend correction coefficient, which is dynamically adjusted based on the mean square error of the recently identified residuals: when the physical model has high fitting accuracy (small residuals), Take the smaller value, focusing on physical prediction; when the model error is large... Take the larger value to enhance the correction weight of trend extrapolation.

[0191] (4) Dynamic compensation model

[0192] The dynamic compensation model includes a slope compensation coefficient based on the slope angle. and viscosity compensation coefficient based on viscosity value This is used to calculate the conventional feedforward quantity;

[0193] In a spraying pipeline, the flow characteristics of a proportional control valve can be approximated as follows:

[0194]

[0195] in This represents the actual traffic volume. The flow coefficient of the valve. The effective pressure difference across the valve. This refers to the density of the liquid medicine. When working on level ground, The operating point is determined by the pump source pressure and is at the design operating point.

[0196] When the sprayer travels on a slope with an angle of θ, the gravitational component of the pesticide liquid column in the pipeline introduces an additional pressure difference along the pipeline axis. Its approximate value is:

[0197]

[0198] in It is the acceleration due to gravity. This is the effective pipeline length along the slope. This additional pressure difference is directly superimposed on the valve's equivalent pressure difference, causing the actual flow rate to deviate from the target value: uphill. > 0, back pressure before valve increases, flow rate is too low; downhill < 0, the flow rate is too high.

[0199] Furthermore, when the slope changes dynamically, The changes cause transient disturbances in the flow rate, the magnitude of which is related to the rate of change of the slope. A positive correlation is particularly pronounced in scenarios such as entering and exiting slopes and crossing field ridges. Based on the above physical analysis, the slope compensation coefficient... It needs to include both a static component reflecting the steady-state slope effect and a dynamic component reflecting the dynamic changes in slope. Therefore, the following compensation model is established:

[0200] Calculate slope compensation according to the formula:

[0201]

[0202]

[0203] In the formula, This is the slope compensation coefficient. This is the static slope compensation coefficient, reflecting the influence of the slope angle itself; The dynamic slope compensation coefficient reflects the rate of slope change. The effect of this is used to suppress flow overshoot or undershoot caused by abrupt changes in slope. In this example, the parameter... =0.12, =0.05.

[0204] The viscosity compensation coefficient is calculated as follows:

[0205]

[0206] in, This is the viscosity compensation coefficient. This is the viscosity value after temperature calibration. For the reference viscosity, in this example =0.45 (emulsifiable concentrate pesticide for paddy fields).

[0207] When performing viscosity value temperature calibration, a viscosity sensor is used to measure the viscosity of the drug solution at the current temperature. The viscosity value at the current temperature is converted to the equivalent viscosity at a standard temperature of 25℃ using the following calibration formula:

[0208]

[0209] in The actual measured viscosity value is T, and the current temperature value measured by the temperature sensor is T.

[0210] The conventional method for calculating the feedforward quantity of a feedforward controller is as follows:

[0211]

[0212] in, For conventional feedforward parameters, A represents the target application rate per unit area, and V represents the real-time travel speed. This is the slope compensation coefficient. This is the viscosity compensation coefficient.

[0213] (5) Fusion of feedforward control quantities

[0214] The predicted feedforward quantity calculated based on the advanced prediction value is weighted and fused with the conventional feedforward quantity to generate the final feedforward control quantity:

[0215]

[0216] in, For the final feedforward control quantity, This is an advanced forecast value. This is a standard feedforward quantity. To predict the confidence coefficient, the value is dynamically adjusted based on historical prediction errors, ranging from 0.5 to 0.9. In this embodiment, the initial value is 0.7. =30ms.

[0217] (6) Adaptive feedback controller

[0218] The formula for the feedback control output is:

[0219]

[0220] in, Let be the feedback control quantity at time t. The deviation between real-time traffic and target traffic. , , These are the proportional, integral, and differential coefficients, which are updated in real time.

[0221] The adaptive feedback controller has a built-in parameter self-tuning module, which adjusts the parameters based on the deviation. and its rate of change The real-time value is adjusted online based on the following fuzzy rules. , , :

[0222] when When it is large, increase To expedite the response while limiting To prevent the points from becoming saturated;

[0223] when When it is in the medium range, reduce appropriately. and gradually introduce To eliminate steady-state error;

[0224] when When it is small, increase and To improve steady-state accuracy and suppress high-frequency disturbances;

[0225] when When it is large, increase In order to suppress overshoot.

[0226] (7) Control quantity fusion

[0227] The feedforward control input and the feedback control input are weighted and fused to generate the total control command:

[0228]

[0229] in, The feedforward weighting coefficient is dynamically adjusted by the operating condition identification module, and is improved in ramp transition mode and transient change mode. Enhanced feedforward function reduces [damage] in stable operation mode To enhance the accuracy of feedback adjustment.

[0230] III. System Control Flow

[0231] After the system is powered on and initialized, it enters a periodic main control loop with a control period of Ts = 10 ms. The steps are as follows:

[0232] Step S1: Data Acquisition and Preprocessing. During each control cycle interrupt, the microcontroller reads analog signals such as flow rate, pressure, viscosity, and temperature via ADC, speed data via CAN, and IMU attitude data via SPI. The raw data is then subjected to digital low-pass filtering to suppress sensor noise.

[0233] Step S2: Multidimensional Time Series Prediction. The velocity prediction submodule, slope prediction submodule, and viscosity prediction submodule run in parallel, each outputting future... Predicted velocity value at time 30ms Slope prediction value Viscosity prediction value Simultaneously update online identification parameters.

[0234] Step S3: Calculation of feedforward control input. The conventional feedforward path calculates this using real-time sensor data. The advanced prediction path is calculated based on the predicted value. The two paths are fused to obtain the final feedforward control quantity. The process is as follows Figure 3 As shown.

[0235] Step S4: Adaptive feedback control quantity calculation. Calculate the deviation between the real-time flow rate and the target flow rate. and its rate of change The feedback control quantity is calculated after the fuzzy parameter self-tuning module updates the PID coefficients. The parameter self-tuning logic is as follows: Figure 4 As shown.

[0236] Step S5: Working Condition Identification and Control Quantity Fusion. The working condition identification module determines the current working scenario and dynamically adjusts the fusion weights. feedforward control quantity With feedback control quantity The final total control quantity is generated by weighted summation. , fusion logic such as Figure 5 As shown.

[0237] Step S6: Output Execution. The fused total control quantity is limited and then converted into a PWM duty cycle output to the proportional valve drive circuit. The LVDT valve position feedback forms a local closed loop, ensuring that the valve core accurately tracks the command. All operating data is refreshed in real time on the touch screen and stored on the SD card at a frequency of 10Hz.

[0238] Experimental verification:

[0239] To verify the practical effectiveness of this invention, a comparative test was conducted in a standard experimental field, using traditional fixed-parameter PID closed-loop control as the control group and the complete control system of this invention as the experimental group. The target application rate was A = 300 mL / mu. The flow rate error was defined as the difference between the measured flow rate and the target flow rate divided by the target flow rate, and the data was statistically analyzed using the control cycle as the sampling unit.

[0240] Operating Condition 1: Uniform speed operation on level ground (baseline condition). Travel speed V = 6 km / h, slope θ = 0°, spray solution is clean water (μ = 1.0 mPa·s, T = 25℃), continuous spraying for 120 s. The steady-state flow error was ±2.8% for the control group and ±1.2% for the experimental group. This invention's feedforward compensation provides accurate initial control values, requiring only minor corrections in the feedback.

[0241] Condition 2: Step speed change. Accelerate from 4 km / h to 8 km / h on flat ground (acceleration time is about 2 s), maintain a constant speed for 10 s, and then decelerate to 4 km / h. The liquid is water.

[0242]

[0243] The speed prediction submodule of this invention predicts the speed trend in advance and adjusts the valve opening ahead of time, reducing the overshoot by about 57% and shortening the adjustment time to 1.3s.

[0244] Condition 3: Slope transition. Drive at 6 km / h from flat ground (0°) onto an uphill slope (+10°), continue for 15 seconds, then turn onto a downhill slope (-10°), and finally return to flat ground. The chemical solution is water.

[0245]

[0246] This invention uses a slope compensation model and Kalman filter prediction to initiate compensation before disturbances occur, reducing the fluctuation amplitude during the transition period by about 45%.

[0247] Operating Condition 4: High-viscosity pesticide application. Speed ​​V = 6 km / h, slope θ = 0°, high-viscosity suspension pesticide (μ = 4.2 mPa·s, T = 20℃), continuous spraying for 120 s. Control group steady-state error ±6.5%; experimental group ±4.8%. The viscosity compensation coefficient is automatically corrected, and the online identification mechanism continuously updates Andrade parameters A and B.

[0248] Operating Condition 5: Multi-disturbance coupled comprehensive operating condition. Speed ​​variation of 4~8 km / h, terrain including alternating flat ground, uphill (+8°), and downhill (-5°), emulsifiable concentrate pesticide (μ≈2.5 mPa·s, T=28℃), total duration 300 s.

[0249]

[0250] In real-world field conditions with multiple coupled disturbances, the performance of traditional PID control deteriorates significantly. This invention achieves a synergistic effect of multi-sensor fusion sensing, multi-dimensional time-series predictive compensation, and adaptive online parameter tuning, resulting in a flow rate error of <5% for 91.6% of all operating conditions, thus meeting the requirements for precision pesticide application.

[0251] The ultimate goal of flow control is to achieve uniform deposition of the pesticide solution on the target. To this end, a field validation experiment based on water-sensitive paper was conducted to verify the uniformity of spray deposition, extending the evaluation of flow control performance from the pipeline level to the field deposition level. The experiment was conducted in a corn seedling field. A sampling section was set every 5 m along the sprayer's direction of travel. Three sheets of water-sensitive paper (26 mm × 76 mm) were evenly arranged along the spray boom direction on each section and fixed to the top of the crop canopy. Six sampling sections were set for each spray stroke, for a total of 18 sampling points. The pesticide solution used in the experiment was clean water, the target application rate was set at 200 L / ha, and the travel speed was 6 km / h. Three control schemes were used: traditional PID, velocity feedforward + PID, and coordinated control. Each scheme was repeated three times, with new water-sensitive paper used each time.

[0252] Water-sensitive paper samples were digitized using a 600 dpi flatbed scanner, converted to 8-bit grayscale images, and then processed for image analysis. The Otsu thresholding method was used to separate the droplet deposition area (dark color) from the uncolored background (light color). The coverage rate of each sheet of water-sensitive paper, i.e., the percentage of droplet area to the total area, and the number of droplets per unit area, were calculated. A comparison of typical water-sensitive paper grayscale images for three control schemes is provided. Figure 6 As shown.

[0253] It can be intuitively observed from grayscale images that: Figure 6 As shown in (a), the droplet distribution on the water-sensitive paper in the traditional PID scheme is uneven, with some areas showing obvious dense patches while adjacent areas show sparse coverage, reflecting the spatial non-uniformity of flow fluctuations; Figure 6 As shown in (b), the droplet distribution of the velocity feedforward + PID scheme is improved compared with the traditional PID scheme, but local differences in density are still visible; Figure 6 As shown in (c), the droplet distribution of the collaborative control scheme provided in the above embodiments of the present invention is the most uniform, and the coverage density of each region is relatively consistent.

[0254] Quantitative analysis was performed on the water-sensitive paper images of all sampling points, and the coefficient of variation (CV) of the coverage was used as a comprehensive evaluation index of deposition uniformity. The results are shown in the table below.

[0255]

[0256] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0257] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0258] This invention is not limited to the preferred embodiment described above. Anyone inspired by this invention can derive other forms of pesticide spraying flow control systems based on multi-sensor information fusion. All equivalent variations and modifications made within the scope of the claims of this invention should be included in the scope of this invention.

Claims

1. A pesticide spraying flow control system based on multi-sensor information fusion, characterized in that: It includes a multi-sensor information acquisition unit, a central control unit, and an execution drive unit; The multi-sensor information acquisition unit is used to collect data in real time on the travel speed, terrain slope, pesticide viscosity, and pipeline flow rate during pesticide spraying operations. The central control unit is signal-connected to the multi-sensor information acquisition unit and the execution drive unit, respectively. The central control unit is configured as follows: Based on the historical time-series data of travel speed, terrain slope, and drug viscosity, the travel speed, terrain slope, and drug viscosity at a preset time in the future are predicted in parallel. The preset time is not shorter than the response delay of the execution drive unit. Based on the advanced prediction results, a feedforward control quantity is generated by combining the dynamic compensation model. The dynamic compensation model is configured to perform coupled compensation based on the terrain slope angle and its rate of change, and to perform compensation based on the temperature-calibrated viscosity of the liquid. Based on the deviation between the real-time flow rate and the target flow rate in the pipeline, the control parameters are adaptively tuned online and feedback control quantities are generated. The feedforward control quantity and the feedback control quantity are weighted and fused together. The fusion weight can be dynamically adjusted to generate a total control command and output it to the execution drive unit. The execution drive unit is used to adjust the opening of the flow regulating valve according to the overall control command to achieve closed-loop control of the spray flow.

2. The pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The central control unit uses a speed prediction submodule, a slope prediction submodule, and a viscosity prediction submodule to achieve advance prediction of travel speed, terrain slope, and drug viscosity, respectively. The speed prediction submodule adopts an adaptive linear prediction algorithm with a forgetting factor, establishes an autoregressive model based on historical speed sequences, and updates the model parameters online using the recursive least squares method. The slope prediction submodule uses the Kalman filter algorithm to establish a state equation with the slope angle and its rate of change as state variables, and integrates attitude observations to achieve optimal estimation and advance prediction of the slope. The viscosity prediction submodule is based on the Andrade viscosity-temperature physical model. It uses a recursive least squares method with a forgetting factor to identify viscosity-temperature characteristic parameters online and outputs an advanced prediction value of the drug solution viscosity by superimposing a historical residual trend correction term.

3. The pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: In the dynamic compensation model, the coupled compensation of the terrain slope angle and its rate of change is achieved through a slope compensation coefficient. This slope compensation coefficient is calculated by combining the static influence of the slope angle and the dynamic influence of the rate of change of slope, and the calculation formula is as follows: Where α is the slope compensation coefficient. This is the static slope compensation coefficient. θ is the dynamic slope compensation coefficient, θ is the real-time slope angle, θ̇ is the slope change rate, and sgn is the sign function. The sign function takes a value of 1 when the slope change rate is greater than 0 (corresponding to an uphill scenario) and a value of -1 when the slope change rate is less than 0 (corresponding to a downhill scenario). and Identified through calibration experiments.

4. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: In the dynamic compensation model, the viscosity compensation of the drug solution is achieved through a viscosity compensation coefficient, which is calculated based on the ratio of the equivalent viscosity of the drug solution calibrated to a preset standard temperature to the reference viscosity. The calibration of the equivalent viscosity of the drug solution is completed based on the difference between the measured viscosity of the drug solution and the real-time temperature of the drug solution and the preset standard temperature.

5. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The central control unit first calculates the conventional feedforward quantity based on the target application rate per unit area, real-time travel speed, slope compensation coefficient, and viscosity compensation coefficient. Then, it weights and fuses the conventional feedforward quantity with the predicted feedforward quantity calculated based on the advanced prediction results to generate the final feedforward control quantity. The final formula for calculating the feedforward control quantity is: in, For the final feedforward control quantity, To predict feedforward quantities, η is the conventional feedforward quantity, and η is the prediction confidence coefficient, which can be dynamically adjusted based on historical prediction errors.

6. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The central control unit uses an improved PID algorithm to generate feedback control quantities. The deviation between the real-time flow rate and the target flow rate, as well as the rate of change of the deviation, are used as input quantities. The proportional coefficient, integral coefficient, and derivative coefficient of the PID algorithm are adaptively tuned online using fuzzy rules. The fuzzy rules pre-set two levels of explicit reference benchmarks: the target flow rate is used as the first reference benchmark to classify the deviation into continuous intervals, and the deviation change rate is used as the system's preset nominal threshold for deviation change rate to classify the deviation change rate into continuous intervals. Each classification includes three order of magnitude intervals: high, medium, and low. The corresponding tuning strategies include: When the deviation is in the high order of magnitude range, increase the proportional coefficient to speed up the system response, while limiting the integral coefficient to avoid integral saturation; When the deviation is in the middle range, decrease the proportional coefficient while gradually increasing the integral coefficient to eliminate the steady-state error of the system. When the deviation is in the low order of magnitude range, increase the integral coefficient and the derivative coefficient to improve the steady-state control accuracy of the system and suppress high-frequency disturbances; When the rate of change of deviation is in the high order of magnitude range, increase the differential coefficient to suppress flow overshoot.

7. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The central control unit can dynamically adjust the weighting coefficients of the weighted fusion according to the working conditions. When the slope and the viscosity of the liquid change drastically, the weighting coefficients of the feedforward control quantity are increased, and when the system tends to operate in a steady state, the weighting coefficients of the feedforward control quantity are decreased.

8. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The central control unit also includes a working condition identification module; The working condition identification module is used to analyze the characteristics of the working scene based on the real-time data acquired by the multi-sensor information acquisition unit, and divide the field working conditions into stable working mode, transient change mode, slope transition mode and strong interference mode. Based on the divided working condition modes, the control strategies and parameter constraints of the advance prediction, feedforward compensation, feedback tuning and weighted fusion links are dynamically adjusted.

9. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The execution drive unit includes a proportional control valve and a valve position feedback sensor; The control signal input terminal of the proportional regulating valve is connected to the output terminal of the central control unit, and is used to receive the overall control command and adjust the valve core opening. The valve position feedback sensor is used to detect the actual opening degree of the proportional control valve in real time and feed the detection result back to the central control unit to form a closed-loop control of the valve position, thereby eliminating the influence of actuator hysteresis and dead zone on the flow control accuracy.

10. A pesticide spraying flow control system based on multi-sensor information fusion according to claim 1, characterized in that: The multi-sensor information acquisition unit includes a speed sensor, an attitude sensor, a viscosity sensor, a flow meter, and a pressure transmitter; The speed sensor is used to collect the travel speed of the spraying operation, the attitude sensor is used to collect the terrain slope and slope change rate, the viscosity sensor integrates a temperature acquisition module to collect the viscosity and temperature of the pesticide solution, the flow meter is used to collect the real-time flow of the pipeline, and the pressure transmitter is used to collect the pressure of the spraying pipeline system.

Images