Cotton layered fertilization row control system and method based on navigation

Abstract

The invention discloses a navigation-based cotton layered fertilization row control system and method, and the system comprises a sensing module which is used for collecting pose data in real time, and fusing the pose data into a fused pose according to a self-adaptive multi-mode filtering algorithm; the decision-making module is connected with the sensing module and used for receiving the fusion pose, generating a query area, obtaining grid matrix data of the query area and calculating fertilization data according to the grid matrix data; and an execution module connected with the decision module and used for receiving the fertilization data, obtaining multi-source depth calibration data, obtaining a physical driving signal according to the multi-source depth calibration data, and controlling an execution mechanism to fertilize according to the physical driving signal and the fertilization data. According to the invention, accurate control of fertilizer feeding is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for agricultural machinery, and in particular to a navigation-based control system and method for stratified fertilization of cotton rows. Background Technology

[0002] Traditional cotton basal fertilizer application relies on manual or mechanical uniform application, which has significant drawbacks. Fertilizer application locations are difficult to match with root distribution, easily damaging cotton seedlings and reducing fertilizer utilization; it cannot respond to the spatial heterogeneity of soil moisture and fertility, leading to localized over- or under-fertilization; and the fixed-depth fertilization pattern is difficult to adapt to the evolving nutrient requirements of cotton roots at different growth stages (seedling stage and flowering / bolling stage). Insufficient fertilization precision, lack of stratified control, and weak variable execution severely restrict yield improvement, cost optimization, and farmland protection. Therefore, there is an urgent need for an intelligent fertilization system that achieves centimeter-level row navigation through high-precision pose (position and heading angle) control, integrating centimeter-level precise row alignment, dynamic configuration of stratified depth, and real-time control of fertilizer flow rate to overcome traditional limitations.

[0003] Current improvement methods, such as satellite navigation-assisted or variable-rate fertilization machinery, are limited by the lack of a closed-loop coordination of the entire chain of positioning, decision-making, and execution, which makes it impossible to accurately complete the stratified fertilization control, thus affecting fertilizer utilization.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a navigation-based cotton stratified fertilization row control system and method, which aims to solve the problem that existing technologies such as satellite navigation-assisted or variable displacement fertilization machinery are limited by the lack of a closed-loop coordination of the entire positioning-decision-execution chain, resulting in the inability to accurately complete stratified fertilization control and thus affecting fertilizer utilization.

[0006] To achieve the above objectives, the present invention provides a navigation-based cotton stratified fertilization row control system, the navigation-based cotton stratified fertilization row control system comprising:

[0007] The perception module is used to collect pose data in real time and fuse the pose data into a fused pose according to an adaptive multimodal filtering algorithm.

[0008] The decision module connected to the perception module is used to receive the fused pose, generate a query region, obtain the raster matrix data of the query region, and calculate fertilization data based on the raster matrix data.

[0009] The execution module connected to the decision module is used to receive the fertilization data, acquire multi-source depth calibration data, obtain a physical drive signal based on the multi-source depth calibration data, and control the execution mechanism to perform fertilization based on the physical drive signal and the fertilization data.

[0010] Optionally, the sensing module includes:

[0011] The multi-source redundant positioning and sensing unit is used to acquire real-time dynamic differential signals based on satellite receivers, acquire triaxial acceleration and triaxial angular velocity data based on inertial measurement units, and acquire wheel rotation speed based on wheel speed encoders.

[0012] An adaptive fusion processing unit connected to a multi-source redundant positioning and sensing unit is used to calculate satellite positioning pose based on the real-time dynamic differential signal, calculate inertial measurement pose based on the three-axis acceleration and the three-axis angular velocity data, calculate wheel speed displacement based on the wheel rotation speed, and generate corresponding weight coefficients. Based on an adaptive multimodal filtering algorithm, the unit calculates the fused pose based on the satellite positioning pose, the inertial measurement pose, the wheel speed displacement, and the weight coefficients, and sends it to the decision module.

[0013] Optionally, the adaptive fusion processing unit includes:

[0014] The data calculation subunit is used to calculate the satellite positioning attitude, inertial measurement attitude, and wheel speed displacement based on the real-time dynamic differential signal, the triaxial acceleration, the triaxial angular velocity data, and the wheel speed.

[0015] The weight calculation subunit connected to the data calculation subunit is used to obtain the satellite positioning accuracy attenuation factor and wheel speed confidence, and calculate the satellite positioning pose weight coefficient, obtain the acceleration variance corresponding to the inertial measurement pose and calculate the inertial measurement pose weight coefficient, obtain the signal confidence of the wheel speed displacement and calculate the wheel speed displacement weight coefficient.

[0016] The fusion subunit connected to the weight calculation subunit is used to calculate the fused pose based on the adaptive multimodal filtering algorithm, according to the satellite positioning pose, the inertial measurement pose, the wheel speed displacement, the satellite positioning pose weight coefficient, the inertial measurement pose weight coefficient, and the wheel speed displacement weight coefficient.

[0017] Optionally, the adaptive fusion processing unit further includes:

[0018] The failure degradation monitoring subunit is used to detect the satellite positioning accuracy attenuation factor in real time. When the satellite positioning accuracy attenuation factor exceeds a set threshold, the corresponding failure status is sent to the compensation subunit.

[0019] The compensation subunit connected to the failure degradation monitoring subunit is used to, upon receiving a failure situation, compensate for pose drift based on the stored historical path trajectory and dynamically fuse high-frequency wheel speed integral and inertial measurement unit data, perform short-term positioning prediction, obtain the corresponding fused pose, and send it to the decision module.

[0020] Optionally, the decision module:

[0021] An edge computing point unit is used to receive the fused pose and generate a query region, obtain the raster matrix data of the query region based on the cloud prescription database, wherein the raster matrix data includes a field soil fertility raster matrix and a soil moisture content raster matrix, and parses it;

[0022] The parameter calculation unit connected to the edge computing point unit is used to calculate the spatial mean of soil fertility and the spatial mean of soil moisture content corresponding to the query area based on the parsed field soil fertility raster matrix and soil moisture content raster matrix. Based on the spatial mean of soil fertility and the spatial mean of soil moisture content, it calculates the shallow target fertilization depth and the deep target fertilization depth. Based on the parsed field soil fertility raster matrix and soil moisture content raster matrix, it calculates the shallow real-time fertilizer discharge and the deep real-time fertilizer discharge. It then encapsulates the shallow target fertilization depth, the deep target fertilization depth, the shallow real-time fertilizer discharge, and the deep real-time fertilizer discharge into fertilization data.

[0023] Optionally, the execution module includes:

[0024] The multi-source depth sensing unit is used to acquire the three-dimensional geometric information of the trenching profile, the real-time distance offset of the trencher reference surface relative to the ground surface, and the real-time soil resistance information based on the heterogeneous sensing fusion architecture, and to parse the three-dimensional geometric information into the corresponding geometric depth value. Based on the real-time distance offset of the trencher reference surface relative to the ground surface, the real-time soil resistance information, and the geometric depth value, the corresponding multi-source depth calibration data is generated.

[0025] A multi-source depth fusion closed-loop control unit connected to the multi-source depth sensing unit is used to receive multi-source depth calibration data from the multi-source depth sensing unit and generate a physical drive signal based on the multi-source depth calibration data.

[0026] An actuator drive unit connected to the multi-source depth sensing unit and the decision module is used to receive the fertilization data and the physical drive signal, and control the actuator to perform fertilization according to the fertilization data and the physical drive signal. The fertilization data includes shallow target fertilization depth, deep target fertilization depth, shallow real-time fertilizer discharge rate, and deep real-time fertilizer discharge rate.

[0027] Optionally, the multi-source deep fusion closed-loop control unit:

[0028] The weighted fusion subunit is used to receive multi-source depth calibration data from the multi-source depth sensing unit and calculate the actual operational depth feedback value based on the weighted fusion model.

[0029] The adaptive control law subunit connected to the weighted fusion subunit is used to calculate the instantaneous depth deviation between the target depth and the actual working depth feedback value in real time, and to solve it into a control output quantity based on the target control algorithm.

[0030] An execution drive subunit connected to the adaptive control law subunit is used to convert the control output into the physical drive signal;

[0031] The loop judgment subunit connected to the execution drive subunit is used to determine whether the physical drive signal is less than a preset deviation. If it is less, the physical drive signal is output to the execution mechanism drive unit.

[0032] Optionally, the cyclic judgment subunit is further configured to adjust the weight coefficients of the weighted fusion model and the target control algorithm when the physical driving signal is not less than the preset deviation;

[0033] After adjusting the weight coefficients of the weighted fusion model and the target control algorithm, the process of the multi-source deep fusion closed-loop control unit is repeated until the physical drive signal is less than the preset deviation.

[0034] Optionally, the actuator drive unit includes:

[0035] A hydraulic depth adjustment execution subunit is used to receive the physical drive signal and control the trenching depth according to the physical drive signal.

[0036] The layered trenching execution subunit is used to receive the shallow target fertilization depth and the deep target fertilization depth from the fertilization data, and to perform layered trenching control based on the shallow target fertilization depth and the deep target fertilization depth.

[0037] The dual-channel precision fertilizer dispensing execution subunit, connected to the hydraulic depth adjustment execution subunit and the stratified trenching execution subunit, is used to receive the shallow real-time fertilizer dispensing amount and the deep real-time fertilizer dispensing amount in the fertilizer application data, and to perform stratified fertilizer dispensing control based on the shallow real-time fertilizer dispensing amount and the deep real-time fertilizer dispensing amount.

[0038] Furthermore, to achieve the above objectives, the present invention also provides a navigation-based method for controlling row-to-row stratified fertilization of cotton, wherein the navigation-based method for controlling row-to-row stratified fertilization of cotton specifically includes:

[0039] Real-time acquisition of pose data, and fusion of the pose data into a fused pose using an adaptive multimodal filtering algorithm;

[0040] Based on the fused pose, a query region is generated, the raster matrix data of the query region is obtained, the fertilization data is calculated based on the raster matrix data, and the fertilization data is sent to the execution module;

[0041] Acquire multi-source depth calibration data, obtain physical drive signals based on the multi-source depth calibration data, and control the actuator to perform fertilization based on the physical drive signals and the fertilization data.

[0042] In this invention, the navigation-based cotton stratified fertilization row control system includes: a sensing module for real-time acquisition of pose data, which fuses the pose data into a fused pose using an adaptive multimodal filtering algorithm; a decision module connected to the sensing module for receiving the fused pose, generating a query region, acquiring grid matrix data of the query region, and calculating fertilization data based on the grid matrix data; and an execution module connected to the decision module for receiving the fertilization data, acquiring multi-source depth calibration data, obtaining a physical drive signal based on the multi-source depth calibration data, and controlling the execution mechanism to perform fertilization based on the physical drive signal and the fertilization data. This invention ensures stable, centimeter-level precise row alignment of agricultural machinery through high-precision dynamic positioning and navigation; it utilizes cloud-edge collaborative prescription decision-making to calculate soil spatial variation data in real time, dynamically generating stratified fertilization target depths and fertilizer application rates that match the nutrient requirements of cotton shallow and deep roots; combined with stratified execution mechanisms and depth closed-loop control, it provides real-time feedback on actual depth through multi-source sensing fusion and dynamically adjusts the execution mechanisms based on control algorithms to achieve centimeter-level precision dynamic compensation for trenching depth. This significantly improves the matching degree between fertilization location and root distribution, reduces cotton seedling damage, accurately responds to differences in soil moisture and fertility, and optimizes fertilizer utilization. Attached Figure Description

[0043] Figure 1 This is a structural diagram of a preferred embodiment of the navigation-based cotton stratified fertilization row control system of the present invention;

[0044] Figure 2 This is a schematic diagram of the sensing module of the row control system for cotton stratified fertilization based on navigation according to the present invention.

[0045] Figure 3 This is a schematic diagram of the decision-making module of the cotton stratified fertilization row control system based on navigation according to the present invention;

[0046] Figure 4 This is a schematic diagram of the execution module of the cotton stratified fertilization row control system based on navigation according to the present invention;

[0047] Figure 5 This is a schematic diagram of the multi-source deep fusion closed-loop control unit of the cotton stratified fertilization row control system based on navigation in this invention;

[0048] Figure 6 This is a flowchart of a preferred embodiment of the navigation-based cotton stratified fertilization row control method of the present invention;

[0049] Figure 7 This is a general workflow of a preferred embodiment of the navigation-based cotton stratified fertilization row control method of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0051] Traditional cotton basal fertilizer application relies on manual or mechanical uniform application, which has significant drawbacks. Fertilizer application locations are difficult to match with root distribution, easily damaging cotton seedlings and reducing fertilizer utilization; it cannot respond to the spatial heterogeneity of soil moisture and fertility, leading to localized over- or under-fertilization; and the fixed-depth fertilization model is difficult to adapt to the nutrient requirements of cotton roots at different growth stages (seedling stage and flowering / boll-forming stage). Insufficient fertilization precision, lack of stratified control, and weak variable-rate execution severely restrict yield improvement, cost optimization, and farmland protection. There is an urgent need for a high-precision pose (position and heading angle) control system to achieve centimeter-level row navigation, integrating centimeter-level precise row alignment, dynamic configuration of stratified depth, and real-time control of fertilizer flow rate to overcome traditional limitations. Existing improvement solutions (such as satellite navigation-assisted or variable-rate fertilization machinery) are limited by the lack of a closed-loop coordination across the entire positioning-decision-execution chain. Navigation systems struggle to maintain centimeter-level accuracy under complex farmland conditions; prescription decisions are detached from the real-time environment, making it impossible to calculate layered parameters online; and the lack of multi-source depth perception and closed-loop control in the actuators leads to significant deviations in operating depth from set values. Therefore, existing technologies require further improvement.

[0052] To address one or more of the aforementioned problems, the navigation-based cotton stratified fertilization row control system of the present invention includes a sensing module, a decision module, and an execution module. The sensing module is connected to the decision module, and the decision module is connected to the execution module. The sensing module is used to collect pose data in real time, fuse the pose data into a fused pose using an adaptive multimodal filtering algorithm, and send it to the decision module. The decision module is used to receive the fused pose, generate a query region, obtain the raster matrix data of the query region, calculate fertilization data based on the raster matrix data, and send the fertilization data to the execution module. The execution module is used to receive the fertilization data, obtain multi-source depth calibration data, obtain a physical drive signal based on the multi-source depth calibration data, and control the execution mechanism to perform fertilization based on the physical drive signal and the fertilization data.

[0053] The preferred embodiment of the present invention describes a navigation-based cotton stratified fertilization row control system, such as... Figure 1 As shown, the navigation-based cotton stratified fertilization row control system includes:

[0054] The perception module 11 is used to collect pose data in real time and fuse the pose data into a fused pose according to an adaptive multimodal filtering algorithm.

[0055] The decision module 12, connected to the perception module, is used to receive the fused pose, generate a query region, obtain the grid matrix data of the query region, and calculate fertilization data based on the grid matrix data.

[0056] The execution module 13, which is connected to the decision module, is used to receive the fertilization data and acquire multi-source depth calibration data, obtain a physical drive signal based on the multi-source depth calibration data, and control the execution mechanism to perform fertilization based on the physical drive signal and the fertilization data.

[0057] Specifically, the overall architecture of the system presents a three-layer topology: a perception module, a decision-making module, and an execution module. The perception module outputs real-time pose information, i.e., fused pose, which is input to the decision-making module to determine the query area. The resulting fertilization data is encapsulated into CAN command frames and sent to the execution module via a high-speed CAN bus for control execution. The perception module, decision-making module, and execution module are tightly coupled through data flow and control flow. Each layer adopts a modular design to jointly achieve the system functions of centimeter-level precise row navigation, dynamic configuration of layer depth, and real-time control of fertilizer flow.

[0058] Furthermore, the sensing module includes a multi-source redundant positioning sensing unit and an adaptive fusion processing unit;

[0059] The multi-source redundant positioning and sensing unit is used to acquire real-time dynamic differential signals based on satellite receivers, acquire triaxial acceleration and triaxial angular velocity data based on inertial measurement units, and acquire wheel rotation speed based on wheel speed encoders.

[0060] An adaptive fusion processing unit connected to a multi-source redundant positioning and sensing unit is used to calculate satellite positioning pose based on the real-time dynamic differential signal, calculate inertial measurement pose based on the three-axis acceleration and the three-axis angular velocity data, calculate wheel speed displacement based on the wheel rotation speed, and generate corresponding weight coefficients. Based on an adaptive multimodal filtering algorithm, the unit calculates the fused pose based on the satellite positioning pose, the inertial measurement pose, the wheel speed displacement, and the weight coefficients, and sends it to the decision module.

[0061] Specifically, the multi-source redundant positioning and sensing unit collects multi-source heterogeneous data in real time through the hardware sensing layer. Specifically, it receives real-time dynamic (RTK) differential signals through a satellite receiver; collects the three-axis acceleration and three-axis angular velocity data of the carrier through an inertial measurement unit (IMU) module; and further, obtains the wheel speed through a wheel speed encoder.

[0062] Furthermore, the adaptive fusion processing unit includes a data calculation subunit, a weight calculation subunit, and a fusion subunit:

[0063] The data calculation subunit is used to calculate the satellite positioning attitude, inertial measurement attitude, and wheel speed displacement based on the real-time dynamic differential signal, the triaxial acceleration, the triaxial angular velocity data, and the wheel speed.

[0064] The weight calculation subunit connected to the data calculation subunit is used to obtain the satellite positioning accuracy attenuation factor and wheel speed confidence, and calculate the satellite positioning pose weight coefficient, obtain the acceleration variance corresponding to the inertial measurement pose and calculate the inertial measurement pose weight coefficient, obtain the signal confidence of the wheel speed displacement and calculate the wheel speed displacement weight coefficient.

[0065] The fusion subunit connected to the weight calculation subunit is used to calculate the fused pose based on the adaptive multimodal filtering algorithm, according to the satellite positioning pose, the inertial measurement pose, the wheel speed displacement, the satellite positioning pose weight coefficient, the inertial measurement pose weight coefficient, and the wheel speed displacement weight coefficient.

[0066] Specifically, such as Figure 2 As shown, the data computing subunit receives real-time kinematic (RTK) differential signals through a satellite receiver to output the satellite positioning pose P. sat The inertial measurement unit (IMU) module acquires the triaxial acceleration and triaxial angular velocity data of the carrier to output the inertial measurement pose P. imuFurthermore, the wheel speed is obtained through a wheel speed encoder, and the displacement P is calculated based on the integral of the wheel speed. whl Wherein, the displacement P whl The calculation formula is:

[0067]

[0068] In the formula, υ(τ) represents the instantaneous linear velocity measured by the wheel speed encoder.

[0069] The weight calculation subunit performs real-time quality assessment on the collected multi-source heterogeneous data and dynamically allocates fusion weight coefficients. Specifically, it monitors in real time the precision attenuation factor (DOP) of the satellite receiver output signal, the variance of the acceleration output by the inertial measurement unit (IMU), and the confidence level of the wheel speed encoder output signal. Based on the monitored data quality parameters, it dynamically allocates fusion weight coefficients for each sensor data, including satellite positioning pose weight coefficient w1, inertial measurement pose weight coefficient w2, and wheel speed displacement weight coefficient w3, wherein w1+w2+w3=1.

[0070] Wherein, the satellite positioning pose P sat The weighting coefficient w1 is determined based on its precision decay factor (DOP), and Confwhl is the wheel speed confidence level (fixed value 0.2 or 0.05), specifically:

[0071]

[0072] Wherein, the inertial measurement pose P imu The weighting coefficient w2 is determined based on its acceleration variance, specifically as follows:

[0073]

[0074] The wheel speed displacement weighting coefficient w3 is determined based on the signal confidence level. Specifically, it is 0.2 when the confidence level of the wheel speed encoder is greater than 90%, and 0.05 otherwise.

[0075] The fusion subunit uses an adaptive multimodal filtering algorithm (AMF) to calculate the final fused pose. Specifically, dynamically allocated weight coefficients w1, w2, and w3 are applied to the corresponding sensor pose data, where the fused pose P... out The following weighted fusion formula was used for calculation:

[0076] P out =w1·P sat +w2·P imu +w3·P whl ;

[0077] And it satisfies w1+w2+w3=1; where, P sat Pimu represents the satellite positioning pose, and P represents the inertial measurement pose. whl This indicates the wheel speed and displacement.

[0078] Furthermore, the adaptive fusion processing unit also includes a failure degradation monitoring subunit and a compensation subunit:

[0079] The failure degradation monitoring subunit is used to detect the satellite positioning accuracy attenuation factor in real time. When the satellite positioning accuracy attenuation factor exceeds a set threshold, the corresponding failure status is sent to the compensation subunit.

[0080] The compensation subunit connected to the failure degradation monitoring subunit is used to, upon receiving a failure situation, compensate for pose drift based on the stored historical path trajectory and dynamically fuse high-frequency wheel speed integral and inertial measurement unit data, perform short-term positioning prediction, obtain the corresponding fused pose, and send it to the decision module.

[0081] This invention further compensates for satellite signal failure or degradation. The failure / degradation monitoring subunit monitors the satellite positioning accuracy attenuation factor (DOP) in real time. When the DOP exceeds a set threshold, indicating satellite signal failure, the failure status is sent to the compensation subunit. The compensation subunit compensates for pose drift by dynamically fusing high-frequency wheel speed integrals with IMU data, and performs short-term positioning prediction based on stored historical path trajectories via kinematic extrapolation. Specifically, the displacement acquired by the wheel speed encoder is the high-frequency wheel speed integral, and the attitude data output by the inertial measurement unit (IMU) is the IMU data. Dynamic fusion compensation for pose drift and short-term positioning prediction involves dynamically fusing high-frequency wheel speed integrals with IMU data to compensate for pose drift, and performing short-term positioning prediction based on stored historical path trajectories via kinematic extrapolation. Finally, the fused pose P is output. out As the final positioning result, and the standard deviation σ of the positioning accuracy under flat terrain conditions is less than or equal to 2 cm, the failure degradation monitoring subunit and compensation subunit adaptively switch to a compensation mode based on wheel speed integration, IMU, and historical trajectory when the satellite signal fails, ultimately outputting pose information with centimeter-level accuracy, providing a reliable benchmark for fertilization row control.

[0082] In the decision-making module, a cloud-edge collaborative prescription decision-making module that integrates real-time pose input to the decision-making module is used to trigger prescription data requests based on geographic coordinates; at the same time, it drives the path tracking controller to calculate the deviation between the agricultural machinery's travel direction and the preset path, and generates steering control commands to achieve precise line tracking.

[0083] Furthermore, the decision-making module includes an edge computing point unit and a parameter calculation unit:

[0084] An edge computing point unit is used to receive the fused pose and generate a query region, obtain the raster matrix data of the query region based on the cloud prescription database, wherein the raster matrix data includes a field soil fertility raster matrix and a soil moisture content raster matrix, and parses it;

[0085] The parameter calculation unit connected to the edge computing point unit is used to calculate the spatial mean of soil fertility and the spatial mean of soil moisture content corresponding to the query area based on the parsed field soil fertility raster matrix and soil moisture content raster matrix. Based on the spatial mean of soil fertility and the spatial mean of soil moisture content, it calculates the shallow target fertilization depth and the deep target fertilization depth. Based on the parsed field soil fertility raster matrix and soil moisture content raster matrix, it calculates the shallow real-time fertilizer discharge and the deep real-time fertilizer discharge. It encapsulates the shallow target fertilization depth, the deep target fertilization depth, the shallow real-time fertilizer discharge, and the deep real-time fertilizer discharge into fertilization data and sends it to the execution module.

[0086] Specifically, such as Figure 3 As shown, the decision-making module relies on edge computing nodes and cloud prescription database to build a real-time data interaction link, and has a built-in spatiotemporal coupling engine for hierarchical fertilization parameters. The execution logic of the spatiotemporal coupling engine is specifically manifested in a process-oriented working mode.

[0087] The multi-source deep fusion closed-loop control unit receives the fused pose and generates a query region, i.e., high-precision position coordinates (x, y). The real-time high-precision position coordinates (x, y) of the agricultural machinery trigger a prescription request. The corresponding engine dynamically generates a spatial query region based on these coordinates. This region is preferably a circular area with radius R. The engine then sends a vector rasterized prescription data request to the cloud prescription database. The cloud prescription database stores the field soil fertility raster matrix F. ij With soil moisture content raster matrix W ij The engine receives and parses the raster matrix data returned from the cloud in real time.

[0088] Furthermore, in the parameter calculation unit, corresponding to the engine, adaptive calculation is performed based on a preset agronomic decision model. This adaptive calculation process specifically includes:

[0089] Calculate the spatial mean F of soil fertility within the query area. avg Spatial mean of soil moisture content W avg The results are obtained by calculating using the following formulas:

[0090]

[0091] Based on the spatial mean F avg W avgUsing preset agronomic experience coefficients a, b, c, d, e, and f, the shallow target fertilization depth D1 and the deep target fertilization depth D2 are calculated. D1 and D2 are obtained by the following formulas:

[0092] D1=α·F avg +b·W avg +c;

[0093] D1=α·F avg +b·W avg +c;

[0094] Meanwhile, based on the soil fertility value F of the grid cell corresponding to the real-time location. ij Moisture content value W ij Using preset agronomic experience coefficients k, m, n, and p, the real-time fertilizer application rates Q1 (shallow layer) and Q2 (deep layer) are calculated. Q1 and Q2 are obtained using the following formulas:

[0095] Q1=k·F ij +m·W ij ;

[0096] Q2=n·F ij +p·W ij ;

[0097] The engine further executes the instruction encapsulation and synchronization process. This process calculates the layered fertilization target parameters, namely the shallow target depth D1, the deep target depth D2, the shallow real-time fertilizer discharge Q1, and the deep real-time fertilizer discharge Q2, and combines them with the synchronization timestamp to form a data load. This load is then encapsulated into a standard data frame, i.e., fertilization data, according to the CAN2.0B protocol and transmitted in real time to the execution module via the high-speed CAN bus, ensuring the timing consistency and control accuracy of multi-axis coordinated actions.

[0098] Furthermore, the execution module includes a multi-source depth sensing unit, a multi-source depth fusion closed-loop control unit, and an execution mechanism drive unit:

[0099] The multi-source depth sensing unit is used to acquire the three-dimensional geometric information of the trenching profile, the real-time distance offset of the trencher reference surface relative to the ground surface, and the real-time soil resistance information based on the heterogeneous sensing fusion architecture, and to parse the three-dimensional geometric information into the corresponding geometric depth value. Based on the real-time distance offset of the trencher reference surface relative to the ground surface, the real-time soil resistance information, and the geometric depth value, the corresponding multi-source depth calibration data is generated.

[0100] A multi-source depth fusion closed-loop control unit connected to the multi-source depth sensing unit is used to receive multi-source depth calibration data from the multi-source depth sensing unit and generate a physical drive signal based on the multi-source depth calibration data.

[0101] An actuator drive unit connected to the multi-source depth sensing unit and the decision module is used to receive the fertilization data and the physical drive signal, and control the actuator to perform fertilization according to the fertilization data and the physical drive signal. The fertilization data includes shallow target fertilization depth, deep target fertilization depth, shallow real-time fertilizer discharge rate, and deep real-time fertilizer discharge rate.

[0102] Specifically, such as Figure 4 As shown, the multi-source depth sensing unit acquires the corresponding sensing information, and the corresponding multi-source depth fusion closed-loop control unit generates physical drive signals based on the information from the multi-source depth sensing unit, and finally executes the corresponding control operations through the actuator drive unit.

[0103] The multi-source depth sensing unit adopts a heterogeneous sensing fusion architecture, which further integrates a trench geometry sensing subunit, a surface reference sensing subunit, and a soil physical state sensing subunit. The trench geometry sensing subunit is installed 20cm behind the trencher on a support structure to acquire the three-dimensional geometric information of the trench profile in real time and parse the geometric depth value D. geo The surface reference sensing subunit is installed at the frame crossbeam to continuously monitor the real-time distance offset D of the trenching execution unit's reference surface relative to the ground surface. ref The soil physical state sensing subunit is embedded in the tip of the trencher to dynamically detect real-time soil resistance information R in the working area. soil Furthermore, the output data of each sensing subunit must undergo strict spatiotemporal synchronization calibration.

[0104] Furthermore, the multi-source deep fusion closed-loop control unit includes a weighted fusion subunit, an adaptive control law subunit, an execution drive subunit, and a loop judgment subunit:

[0105] The weighted fusion subunit is used to receive multi-source depth calibration data from the multi-source depth sensing unit and calculate the actual operational depth feedback value based on the weighted fusion model.

[0106] The adaptive control law subunit connected to the weighted fusion subunit is used to calculate the instantaneous depth deviation between the target depth and the actual working depth feedback value in real time, and to solve it into a control output quantity based on the target control algorithm.

[0107] An execution drive subunit connected to the adaptive control law subunit is used to convert the control output into the physical drive signal;

[0108] The loop judgment subunit connected to the execution drive subunit is used to determine whether the physical drive signal is less than a preset deviation. If it is less, the physical drive signal is output to the execution mechanism drive unit.

[0109] Specifically, such as Figure 5 As shown, the multi-source deep fusion closed-loop control unit includes a weighted fusion subunit, an adaptive control law subunit, an execution drive subunit, and a loop judgment subunit. The weighted fusion subunit receives multi-source depth calibration data D from the multi-source depth sensing unit. geo D ref and R soil The actual operation depth feedback value D is calculated using a weighted fusion model. act The weighted fusion model is characterized as follows:

[0110]

[0111] Where R0 = 50 N / cm 2 The reference value for the calibrated resistance is dynamically allocated based on the sensor confidence level, and the normalization constraint α+β+γ=1 is satisfied.

[0112] The adaptive control law subunit is further manifested as a PID control law subunit, which calculates the target depth D in real time. tar The actual depth D is output by deep fusion modeling. act The instantaneous depth deviation e(t) between them has the following relationship:

[0113] e(t) = D tar -D act

[0114] Then, the PID control algorithm is applied to calculate the control output u(t):

[0115]

[0116] Wherein, the proportional gain coefficient K p The integral cumulative coefficient Ki is used to suppress deep abrupt changes in real time, and the differential prediction coefficient K is used to eliminate steady-state errors caused by historical biases. d Used to predict trends in soil condition changes.

[0117] The execution drive subunit further converts the control output u(t) into a corresponding physical drive signal, such as an analog voltage signal, to regulate the action of the hydraulic depth adjustment execution unit in real time. Specifically, it adjusts the opening of the electro-hydraulic proportional valve, thereby achieving centimeter-level precise dynamic adjustment of the trenching depth through changes in the stroke of the hydraulic cylinder.

[0118] Furthermore, the cyclic judgment subunit is also used to adjust the weight coefficients of the weighted fusion model and the target control algorithm when the physical driving signal is not less than the preset deviation;

[0119] After adjusting the weight coefficients of the weighted fusion model and the target control algorithm, the process of the multi-source deep fusion closed-loop control unit is repeated until the physical drive signal is less than the preset deviation.

[0120] In this invention, the target value of the physical drive signal is |e(t)|≤2cm, that is, 2cm is the preset deviation. When the physical drive signal is less than the preset deviation, the physical drive signal is output to the actuator drive unit.

[0121] When the physical driving signal deviates from the preset deviation, the parameter adjustment process is triggered. That is, when the soil condition changes drastically, the weight coefficients α, β, and γ of the multi-source fusion model are dynamically allocated first, based on the real-time confidence key parameters of the data from each sensing unit; when the system response is lagging, the PID control parameter K is dynamically adjusted first. p K i K d .

[0122] The actuator drive unit includes a hydraulic depth adjustment actuator subunit, a layered trenching actuator subunit, and a dual-channel precision fertilizer discharge actuator subunit;

[0123] A hydraulic depth adjustment execution subunit is used to receive the physical drive signal and control the trenching depth according to the physical drive signal.

[0124] The layered trenching execution subunit is used to receive the shallow target fertilization depth and the deep target fertilization depth from the fertilization data, and to perform layered trenching control based on the shallow target fertilization depth and the deep target fertilization depth.

[0125] The dual-channel precision fertilizer dispensing execution subunit, connected to the hydraulic depth adjustment execution subunit and the stratified trenching execution subunit, is used to receive the shallow real-time fertilizer dispensing amount and the deep real-time fertilizer dispensing amount in the fertilizer application data, and to perform stratified fertilizer dispensing control based on the shallow real-time fertilizer dispensing amount and the deep real-time fertilizer dispensing amount.

[0126] Specifically, the actuator drive unit of the present invention further includes a layered ditching execution subunit, a dual-channel precision fertilizer dispensing execution subunit, and a hydraulic depth adjustment execution subunit. The layered ditching execution subunit is used to receive the shallow target fertilization depth D1 and the deep target fertilization depth D2 from the fertilization data, and performs layered ditching control according to the shallow target fertilization depth and the deep target fertilization depth, that is, to realize precise layered ditching operation. The dual-channel precision fertilizer dispensing execution subunit is used to receive the shallow real-time fertilizer dispensing amount Q1 and the deep real-time fertilizer dispensing amount Q2 from the fertilization data. The dual-channel precision fertilizer dispensing execution subunit is configured with an independent dual-channel control interface, wherein the first channel is used to respond to the shallow real-time fertilizer dispensing amount Q1 command, and the second channel is used to respond to the deep real-time fertilizer dispensing amount Q2 command, together realizing the precise measurement and application of fertilizer in layers. The hydraulic depth adjustment execution subunit is used to receive the physical drive signal and perform ditching depth control according to the physical drive signal.

[0127] The navigation-based cotton stratified fertilization row control system of the present invention includes: a sensing module for real-time acquisition of pose data, and fusing the pose data into a fused pose according to an adaptive multimodal filtering algorithm; a decision module connected to the sensing module for receiving the fused pose, generating a query region, acquiring grid matrix data of the query region, and calculating fertilization data based on the grid matrix data; and an execution module connected to the decision module for receiving the fertilization data, acquiring multi-source depth calibration data, obtaining a physical drive signal based on the multi-source depth calibration data, and controlling the execution mechanism to perform fertilization based on the physical drive signal and the fertilization data. This invention ensures stable, centimeter-level precise row alignment of agricultural machinery through high-precision dynamic positioning and navigation; it utilizes cloud-edge collaborative prescription decision-making to calculate soil spatial variation data in real time, dynamically generating stratified fertilization target depths and fertilizer application rates that match the nutrient requirements of cotton shallow and deep roots; combined with stratified execution mechanisms and depth closed-loop control, it provides real-time feedback on actual depth through multi-source sensing fusion and dynamically adjusts the execution mechanisms based on control algorithms to achieve centimeter-level precision dynamic compensation for trenching depth. This significantly improves the matching degree between fertilization location and root distribution, reduces cotton seedling damage, accurately responds to differences in soil moisture and fertility, and optimizes fertilizer utilization.

[0128] Furthermore, such as Figure 6 As shown, based on the above-mentioned navigation-based cotton stratified fertilization row control system, the present invention also provides a navigation-based cotton stratified fertilization row control method, wherein the navigation-based cotton stratified fertilization row control method includes:

[0129] Step S61: Collect pose data in real time, and fuse the pose data into a fused pose according to an adaptive multimodal filtering algorithm;

[0130] Step S62: Based on the fused pose, generate a query region, obtain the raster matrix data of the query region, calculate the fertilization data based on the raster matrix data, and send the fertilization data to the execution module;

[0131] Step S63: Obtain multi-source depth calibration data, obtain physical drive signal based on multi-source depth calibration data, and control the actuator to perform fertilization based on the physical drive signal and the fertilization data.

[0132] Specifically, such as Figure 7 As shown, the corresponding method of this invention begins with a high-precision line navigation process, specifically by immediately activating the AMF multi-source fusion positioning process after the operation starts. This process receives three heterogeneous data sources in parallel: satellite positioning pose P... sat Its update frequency is no less than 10Hz; the triaxial acceleration and triaxial angular velocity data collected by the inertial measurement unit are used to calculate the attitude and output the pose P. imu The instantaneous linear velocity υ(τ) of the wheel, measured by the wheel speed encoder, is used to calculate the displacement P through numerical integration. whl Furthermore, the system's built-in adaptive fusion processing unit evaluates the quality indicators of each data source in real time, specifically including the precision attenuation factor (DOP) of satellite positioning pose, IMU acceleration variance, and wheel speed signal confidence. Weighting coefficients are assigned based on the dynamic quality evaluation results to ensure that the fused pose consistently meets centimeter-level positioning accuracy (σ≤2cm). Moreover, when satellite signals fail, pose drift is dynamically compensated through high-frequency wheel speed integration and IMU data, and short-term positioning is predicted based on historical path trajectories to ensure that the standard deviation of positioning accuracy σ remains ≤2cm. Based on this high-precision pose information, the path tracking controller uses a PID algorithm to calculate the deviation between the agricultural machinery's travel direction and the preset path, generating steering mechanism control quantities to drive the hydraulic steering system for precise line tracking.

[0133] The corresponding method of this invention further executes the cloud-edge collaborative prescription decision-making process. Specifically, the real-time geographic coordinates (x, y) output by the navigation module trigger the decision-making process, wherein the edge computing node dynamically generates a spatial query area centered on these coordinates, which is preferably a circular coverage area with a radius of R. Further, a vector raster data request is initiated to the cloud prescription database. The requested data includes at least a soil fertility raster matrix and a soil moisture content raster matrix, with a spatial resolution of not less than 0.5m × 0.5m. After receiving the raster data returned from the cloud, the spatiotemporal coupling engine performs a two-layer parameter calculation: the first layer is based on spatial mean calculation, specifically extracting the values ​​of all raster cells within the query area and calculating the spatial mean of soil fertility and soil moisture content; the second layer is based on real-time location calculation, specifically extracting the soil fertility and moisture content values ​​of the raster cells corresponding to the current agricultural machinery location, and further calculating the shallow target depth, deep target depth, shallow real-time fertilizer application rate, and deep real-time fertilizer application rate. The calculated shallow target depth, deep target depth, shallow fertilizer discharge rate, and deep fertilizer discharge rate, along with a high-precision timestamp, are used as the data payload, encapsulated into a data frame conforming to the CAN bus protocol standard, and transmitted in real time via a high-speed CAN bus.

[0134] The method described in this invention further executes a layered operation and depth closed-loop control process. In the layered trenching execution subunit, in response to the target depth command: the shallow trencher sets its initial depth according to D1, and the deep trencher sets its initial depth according to D2. The dual-channel precision fertilizer discharge execution subunit responds to the flow command: the first discharge channel adjusts the discharge shaft speed according to Q1 g / m, and the second channel adjusts it independently according to Q2 g / m. Subsequently, in the deep fusion modeling subunit, a physical drive signal is generated, and the opening of the electro-hydraulic proportional valve is adjusted. Through changes in the hydraulic cylinder stroke, centimeter-level dynamic compensation of the trenching depth is achieved (compensation accuracy ≤ ±1.5 mm) until the stable condition of |e(t)| ≤ 2 cm is met.

[0135] The method described in this invention further performs full-process status monitoring and parameter optimization, specifically by continuously monitoring the operation status and implementing dynamic optimization. When the data response delay in the cloud prescription database exceeds 200ms, historical prescription data in the edge cache is activated. During the execution phase, the persistence of the monitoring depth deviation of the sub-unit is judged by looping: if |e(t)|>3cm continuously for more than 2s, it is determined that the soil condition has changed drastically, and the fusion weight coefficient is adjusted first (e.g., increasing the soil resistance weight γ); if the system response time exceeds 500ms, it is determined that the control is lagging, and the PID parameters are adjusted first (e.g., increasing the derivative coefficient Kd). This optimization process ensures that the system maintains stable operation under complex farmland conditions until the operation completion signal triggers the process termination.

[0136] The method described in this invention further includes operation termination and data archiving. Specifically, when the coverage rate of the operation area reaches 100%, a termination command is automatically sent, and the entire chain data of this operation is encrypted and stored in a cloud database to provide data support for agronomic model iteration.

[0137] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0138] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A navigation-based cotton stratified fertilization row control system, characterized in that, The navigation-based cotton stratified fertilization row control system includes: The perception module is used to collect pose data in real time and fuse the pose data into a fused pose according to an adaptive multimodal filtering algorithm. The decision module connected to the perception module is used to receive the fused pose, generate a query region, obtain the raster matrix data of the query region, and calculate fertilization data based on the raster matrix data. The execution module connected to the decision module is used to receive the fertilization data, acquire multi-source depth calibration data, obtain a physical drive signal based on the multi-source depth calibration data, and control the execution mechanism to perform fertilization based on the physical drive signal and the fertilization data.

2. The navigation-based cotton stratified fertilization row control system according to claim 1, characterized in that, The sensing module includes: The multi-source redundant positioning and sensing unit is used to acquire real-time dynamic differential signals based on satellite receivers, acquire triaxial acceleration and triaxial angular velocity data based on inertial measurement units, and acquire wheel rotation speed based on wheel speed encoders. An adaptive fusion processing unit connected to a multi-source redundant positioning and sensing unit is used to calculate satellite positioning pose based on the real-time dynamic differential signal, calculate inertial measurement pose based on the three-axis acceleration and the three-axis angular velocity data, calculate wheel speed displacement based on the wheel rotation speed, and generate corresponding weight coefficients. Based on an adaptive multimodal filtering algorithm, the unit calculates the fused pose based on the satellite positioning pose, the inertial measurement pose, the wheel speed displacement, and the weight coefficients, and sends it to the decision module.

3. The navigation-based cotton stratified fertilization row control system according to claim 2, characterized in that, The adaptive fusion processing unit includes: The data calculation subunit is used to calculate the satellite positioning attitude, inertial measurement attitude, and wheel speed displacement based on the real-time dynamic differential signal, the triaxial acceleration, the triaxial angular velocity data, and the wheel speed. The weight calculation subunit connected to the data calculation subunit is used to obtain the satellite positioning accuracy attenuation factor and wheel speed confidence, and calculate the satellite positioning pose weight coefficient, obtain the acceleration variance corresponding to the inertial measurement pose and calculate the inertial measurement pose weight coefficient, obtain the signal confidence of the wheel speed displacement and calculate the wheel speed displacement weight coefficient. The fusion subunit connected to the weight calculation subunit is used to calculate the fused pose based on the adaptive multimodal filtering algorithm, according to the satellite positioning pose, the inertial measurement pose, the wheel speed displacement, the satellite positioning pose weight coefficient, the inertial measurement pose weight coefficient, and the wheel speed displacement weight coefficient.

4. The navigation-based cotton stratified fertilization row control system according to claim 3, characterized in that, The adaptive fusion processing unit further includes: The failure degradation monitoring subunit is used to detect the satellite positioning accuracy attenuation factor in real time. When the satellite positioning accuracy attenuation factor exceeds a set threshold, the corresponding failure status is sent to the compensation subunit. The compensation subunit connected to the failure degradation monitoring subunit is used to, upon receiving a failure situation, compensate for pose drift based on the stored historical path trajectory and dynamically fuse high-frequency wheel speed integral and inertial measurement unit data, perform short-term positioning prediction, obtain the corresponding fused pose, and send it to the decision module.

5. The navigation-based cotton stratified fertilization row control system according to claim 1, characterized in that, The decision-making module: An edge computing point unit is used to receive the fused pose and generate a query region, obtain the raster matrix data of the query region based on the cloud prescription database, wherein the raster matrix data includes a field soil fertility raster matrix and a soil moisture content raster matrix, and parses it; The parameter calculation unit connected to the edge computing point unit is used to calculate the spatial mean of soil fertility and the spatial mean of soil moisture content corresponding to the query area based on the parsed field soil fertility raster matrix and soil moisture content raster matrix. Based on the spatial mean of soil fertility and the spatial mean of soil moisture content, it calculates the shallow target fertilization depth and the deep target fertilization depth. Based on the parsed field soil fertility raster matrix and soil moisture content raster matrix, it calculates the shallow real-time fertilizer discharge and the deep real-time fertilizer discharge. It then encapsulates the shallow target fertilization depth, the deep target fertilization depth, the shallow real-time fertilizer discharge, and the deep real-time fertilizer discharge into fertilization data.

6. The navigation-based cotton stratified fertilization row control system according to claim 1, characterized in that, The execution module includes: The multi-source depth sensing unit is used to acquire the three-dimensional geometric information of the trenching profile, the real-time distance offset of the trencher reference surface relative to the ground surface, and the real-time soil resistance information based on the heterogeneous sensing fusion architecture, and to parse the three-dimensional geometric information into the corresponding geometric depth value. Based on the real-time distance offset of the trencher reference surface relative to the ground surface, the real-time soil resistance information, and the geometric depth value, the corresponding multi-source depth calibration data is generated. A multi-source depth fusion closed-loop control unit connected to the multi-source depth sensing unit is used to receive multi-source depth calibration data from the multi-source depth sensing unit and generate a physical drive signal based on the multi-source depth calibration data. An actuator drive unit connected to the multi-source depth sensing unit and the decision module is used to receive the fertilization data and the physical drive signal, and control the actuator to perform fertilization according to the fertilization data and the physical drive signal. The fertilization data includes shallow target fertilization depth, deep target fertilization depth, shallow real-time fertilizer discharge rate, and deep real-time fertilizer discharge rate.

7. The navigation-based cotton stratified fertilization row control system according to claim 6, characterized in that, The multi-source deep fusion closed-loop control unit: The weighted fusion subunit is used to receive multi-source depth calibration data from the multi-source depth sensing unit and calculate the actual operational depth feedback value based on the weighted fusion model. The adaptive control law subunit connected to the weighted fusion subunit is used to calculate the instantaneous depth deviation between the target depth and the actual working depth feedback value in real time, and to solve it into a control output quantity based on the target control algorithm. An execution drive subunit connected to the adaptive control law subunit is used to convert the control output into the physical drive signal; The loop judgment subunit connected to the execution drive subunit is used to determine whether the physical drive signal is less than a preset deviation. If it is less, the physical drive signal is output to the execution mechanism drive unit.

8. The navigation-based cotton stratified fertilization row control system according to claim 7, characterized in that, The cyclic judgment subunit is also used to adjust the weight coefficients of the weighted fusion model and the target control algorithm when the physical driving signal is not less than the preset deviation; After adjusting the weight coefficients of the weighted fusion model and the target control algorithm, the process of the multi-source deep fusion closed-loop control unit is repeated until the physical drive signal is less than the preset deviation.

9. The navigation-based cotton stratified fertilization row control system according to claim 6, characterized in that, The actuator drive unit includes: A hydraulic depth adjustment execution subunit is used to receive the physical drive signal and control the trenching depth according to the physical drive signal. The layered trenching execution subunit is used to receive the shallow target fertilization depth and the deep target fertilization depth from the fertilization data, and to perform layered trenching control based on the shallow target fertilization depth and the deep target fertilization depth. The dual-channel precision fertilizer dispensing execution subunit, connected to the hydraulic depth adjustment execution subunit and the stratified trenching execution subunit, is used to receive the shallow real-time fertilizer dispensing amount and the deep real-time fertilizer dispensing amount in the fertilizer application data, and to perform stratified fertilizer dispensing control based on the shallow real-time fertilizer dispensing amount and the deep real-time fertilizer dispensing amount.

10. A navigation-based method for row control of stratified fertilization in cotton, characterized in that, The navigation-based cotton stratified fertilization row control method specifically includes: Real-time acquisition of pose data, and fusion of the pose data into a fused pose using an adaptive multimodal filtering algorithm; Based on the fused pose, a query region is generated, the raster matrix data of the query region is obtained, and the fertilization data is calculated based on the raster matrix data. Acquire multi-source depth calibration data, obtain physical drive signals based on the multi-source depth calibration data, and control the actuator to perform fertilization based on the physical drive signals and the fertilization data.

Images