Control method for controlling follow-up speed change of drip irrigation disc of seeder based on Beidou speed measurement

Abstract

The invention discloses a control method for controlling the follow-up speed change of a drip irrigation disc of a seeder based on Beidou speed measurement. The control method comprises the following steps: S1, collecting and preprocessing position information and speed information of a tractor; s2, inputting the preprocessed data into a fusion calculation unit, and executing space-time weight calibration; s3, inputting the fusion speed sequence into a speed prediction network, extracting multi-scale features, and generating a speed guide feature map; s4, generating a target rotating speed instruction according to the mapping function and operation parameter modulation; s5, inputting a target rotating speed instruction to a motor control unit, and generating a current signal through dual-channel torque regulation and control; s6, a motor is driven to rotate the drip irrigation disc, the actual rotating speed is collected, and an error sequence is constructed; and S7, inputting the error sequence into a calibration module, generating a correction parameter to update a current signal, and realizing synchronous control. According to the invention, high-precision synchronous control of the drip irrigation disc on the speed of the tractor is realized, and the drip irrigation consistency, the response speed and the operation stability are improved.

Description

Technical Field

[0001] This invention relates to the field of motor drive and speed control technology, and in particular to a method for controlling the speed of a seeder drip irrigation disc based on Beidou speed measurement. Background Technology

[0002] In modern agricultural production, the synchronization of sowing and water and fertilizer management directly affects crop emergence rate, growth uniformity, and agricultural input utilization efficiency. To improve operational quality, some seeders have integrated drip irrigation trays to achieve coordinated sowing and drip irrigation operations. However, traditional drip irrigation tray drive methods are mostly based on mechanical transmission or constant speed control, lacking the ability to dynamically respond to changes in operating speed. This often leads to drip irrigation position lag or overlap in actual operations, seriously affecting sowing quality and the uniformity of water and fertilizer distribution.

[0003] Current tractor speed acquisition methods primarily rely on wheel speed sensors, which are susceptible to slippage, mud, and other adverse conditions, leading to unstable speed measurement and insufficient control accuracy. Furthermore, existing motor control systems generally employ a single-channel speed feedback structure, failing to balance feedforward prediction and error compensation under changing actual operating conditions. This results in slow response, delayed adjustment, and an inability to achieve real-time synchronization between the drip irrigation tray and the tractor's motion.

[0004] In addition, existing methods for controlling the linkage between sowing and drip irrigation generally lack high-precision spatiotemporal fusion algorithms. Speed ​​prediction models are mostly linear fitting or empirical rule settings, which are difficult to adapt to the complex and ever-changing rhythm of field operations. In scenarios with drastic speed changes, tortuous paths, or short-term stops and restarts, the adjustment of the drip irrigation disc rotation speed has a large lag, which cannot meet the needs of precision agriculture for dynamic synchronous control.

[0005] Furthermore, current mainstream methods do not incorporate high-precision spatiotemporal sensing techniques based on BeiDou, nor do they combine deep attention mechanisms to mine velocity change characteristics at different time scales, resulting in inaccurate judgments of future motion trends, which in turn affects the matching and real-time performance of motor torque control strategies.

[0006] Therefore, how to provide a method for controlling the drip irrigation disc and speed change of a seeder based on Beidou speed measurement is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of this invention is to propose a method for controlling the speed of a seeder drip irrigation disc based on Beidou speed measurement. This invention integrates dual-source speed measurement of Beidou navigation and encoder, a speed prediction network based on attention mechanism, a dual-channel torque control structure, and a dynamic error calibration mechanism. It describes in detail how to achieve precise tracking control of the drip irrigation disc to the tractor speed in complex field environments, and has the advantages of high speed perception accuracy, strong drip irrigation synchronization, and fast torque response.

[0008] A method for controlling the speed of a seeder drip irrigation tray and its rotation speed based on BeiDou speed measurement according to an embodiment of the present invention includes the following steps:

[0009] S1. Acquire the position and speed information of the tractor during operation and perform preprocessing;

[0010] S2. Input the preprocessed position and velocity information into the fusion computing unit, and use an improved spatiotemporal weight calibration algorithm to synchronously correct the multi-source data and output the fused velocity sequence.

[0011] S3. Input the fused velocity sequence into the velocity prediction network based on the attention mechanism, extract velocity change features at different temporal scales, and output a velocity guidance feature map for control.

[0012] S4. Generate a target rotation speed command for the drip irrigation tray based on the speed guidance feature map. The target rotation speed command is generated by the speed control mapping function and jointly modulated with the task parameters.

[0013] S5. Input the target speed command to the brushless motor drive control unit, construct a dual-channel torque regulation structure through parallel error feedback path and feedforward speed path, and output current control signal.

[0014] S6. Drive the brushless motor to perform the drip irrigation tray rotation operation according to the current control signal, and collect the current actual speed of the drip irrigation tray to construct the error sequence between the target speed and the actual speed;

[0015] S7. Input the error sequence into the dynamic calibration module based on sliding window weighting to generate real-time correction parameters for updating the current control signal, thereby achieving continuous synchronization between the drip irrigation tray and the tractor speed.

[0016] Optionally, the location information is collected by the Beidou navigation terminal, and the speed information is collected jointly by the Beidou navigation terminal and the encoder installed on the drive wheel shaft.

[0017] Optionally, the preprocessing includes timestamp alignment, coordinate transformation, noise filtering, packet loss interpolation, and scale normalization.

[0018] Optionally, S2 specifically includes:

[0019] S21. Input the preprocessed position information sequence and velocity information sequence into the fusion calculation unit. The position information is the sequence of horizontal and vertical coordinates of the sampling point, and the velocity information is the instantaneous velocity sequence of the corresponding sampling point.

[0020] S22. In the fusion computing unit, a sliding window process is performed on the velocity information sequence to generate a local velocity reference sequence within each window;

[0021] S23. In the fusion computing unit, the improved spatiotemporal weight calibration algorithm is invoked to generate a corresponding weight sequence based on the current position coordinate changes and velocity differences. The weights are adjusted by both position distance and velocity deviation.

[0022] S24. Perform weighted fusion based on the velocity information sequence and the weight sequence, and output the fused velocity sequence.

[0023] Optionally, the fusion computing unit receives the preprocessed position information sequence and velocity information sequence, performs sliding window segmentation, local reference extraction, weight generation and weighted calculation operations, and the improved spatiotemporal weight calibration algorithm constructs dynamic weighting factors in the fusion computing unit according to the changes in sampling point distance and velocity differences, and performs weighted fusion processing on the original velocity sequence based on the weighting factors to generate a fused velocity sequence.

[0024] Optionally, S3 specifically includes:

[0025] S31. Input the fused velocity sequence into an attention-based velocity prediction network;

[0026] S32. In the velocity prediction network, three sliding windows of different time lengths are set to construct velocity subsequence feature matrices of three scales, representing short-term, medium-term and long-term velocity change sequences respectively.

[0027] S33. At each time scale, the velocity change feature representation of the corresponding scale is calculated using the attention mechanism. Each feature representation is generated by a weighted sum of the query matrix, the key matrix and the value matrix.

[0028] S34. The velocity change features at the three scales are spliced ​​together in chronological order to construct a velocity guidance feature map, which is used to drive the generation process of the target rotational speed.

[0029] Optionally, the velocity prediction network is a multi-scale sequence modeling network based on an attention mechanism, consisting of an input layer, three sets of sliding window extraction layers with different temporal scales, a feature mapping layer, an attention calculation layer, and a feature fusion output layer. Each scale sliding window corresponds to short-term, medium-term, and long-term velocity sequences, respectively. A query matrix, a key matrix, and a value matrix are generated through matrix calculation, and then weighted output features are calculated. The velocity guidance feature map is a temporal feature set composed of attention-weighted output features from each scale, representing the dynamic change trend of the fused velocity at different scales, and is used for the generation and control of the target rotational speed.

[0030] Optionally, S4 specifically includes:

[0031] S41. Input the velocity-guided feature map into the velocity control mapping function to generate an initial target rotational speed vector arranged by time steps;

[0032] S42. Extract the set of task parameters, and jointly modulate the initial target speed vector with the task parameters to generate a modulated target speed sequence. The target speed sequence is achieved by applying a weighting factor of the task parameters to the initial target speed.

[0033] S43. Extract the target rotation speed corresponding to the current time step from the modulated target rotation speed sequence, and output the target rotation speed as the control command of the drip irrigation tray.

[0034] Optionally, S5 specifically includes:

[0035] S51. Input the control command of the drip irrigation tray corresponding to the modulated target speed to the brushless motor drive control unit as the expected speed of the current control cycle;

[0036] S52. Construct the feedforward speed path and calculate the feedforward torque control command based on the target speed and speed gain;

[0037] S53. Construct an error feedback path, obtain the actual speed and calculate the error with the target speed, and generate an error feedback torque control command through a proportional-integral (PI) control algorithm. The PPI control algorithm is a linear control method that calculates the torque command based on a weighted sum of the speed error of the current control cycle and the historical cumulative error.

[0038] S54. Through the dual-channel torque control structure, the torque commands of the feedforward channel and the error feedback channel are directly synthesized to output the total torque control command.

[0039] S55. Perform current conversion based on the total torque control command to generate a current control signal for driving the brushless motor.

[0040] Optionally, S6 specifically includes:

[0041] S61. Drive the brushless motor to perform the drip irrigation tray rotation operation according to the current control signal, and enter the execution stage of the current control cycle;

[0042] S62. Collect the actual rotation speed of the drip irrigation tray in real time within the control cycle, and construct the actual rotation speed sequence according to the time step;

[0043] S63. Compare the target speed sequence with the actual speed sequence one by one to construct an error sequence. The error is the normalized absolute error, which is used for error compensation adjustment.

[0044] The beneficial effects of the present invention are:

[0045] First, this invention integrates a dual-source speed measurement method combining a Beidou navigation terminal and a wheel axle encoder, and uses an improved spatiotemporal weight calibration algorithm to synchronously correct multi-source data, significantly improving the stability and anti-interference capability of speed perception. It overcomes the problems of slippage and loss of measurement by traditional single wheel speed sensors under complex working conditions, and realizes high-precision dynamic perception of tractor speed.

[0046] Secondly, this invention introduces a velocity prediction network based on an attention mechanism to construct a multi-scale velocity change feature extraction structure. This structure can model and guide velocity change trends at different time scales, effectively improving the foresight and adaptability of target rotation speed command generation. Simultaneously, by combining a joint modulation mechanism of task parameters, the drip irrigation tray control commands can dynamically match the sowing requirements of different crops, densities, and terrains.

[0047] Finally, this invention constructs a dual-channel torque control structure with feedforward and feedback in parallel. By integrating proportional-integral regulation and feedforward control, the brushless motor is driven to respond efficiently to the target speed. Combined with actual speed acquisition and sliding window weighted error calibration, the drip irrigation disc achieves continuous and precise synchronous control of the tractor speed, improving the spatial consistency between sowing and drip irrigation, and significantly improving the quality of agricultural machinery operations and water and fertilizer utilization efficiency. Attached Figure Description

[0048] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0049] Figure 1 This is a flowchart of a method for controlling the speed of a seeder drip irrigation disc and variable speed based on Beidou speed measurement, as proposed in this invention.

[0050] Figure 2 This invention presents a flowchart of a fusion speed measurement and speed prediction control method for a seeder drip irrigation disc and speed change control based on Beidou speed measurement and control.

[0051] Figure 3 This is a flowchart of the target speed generation and torque control method for a seeder drip irrigation disc and shift speed control based on Beidou speed measurement and control proposed in this invention;

[0052] Figure 4 This is a flowchart of the dynamic error calibration and closed-loop adjustment process for a seeder drip irrigation disc and speed control method based on Beidou speed measurement and control proposed in this invention. Detailed Implementation

[0053] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0054] refer to Figure 1-4 A method for controlling the speed of a seeder's drip irrigation disc and variable speed based on BeiDou speed measurement and control includes the following steps:

[0055] S1. Acquire the position and speed information of the tractor during operation and perform preprocessing;

[0056] S2. Input the preprocessed position and velocity information into the fusion computing unit, and use an improved spatiotemporal weight calibration algorithm to synchronously correct the multi-source data and output the fused velocity sequence.

[0057] S3. Input the fused velocity sequence into the velocity prediction network based on the attention mechanism, extract velocity change features at different temporal scales, and output a velocity guidance feature map for control.

[0058] S4. Generate a target rotation speed command for the drip irrigation tray based on the speed guidance feature map. The target rotation speed command is generated by the speed control mapping function and jointly modulated with the task parameters.

[0059] S5. Input the target speed command to the brushless motor drive control unit, construct a dual-channel torque regulation structure through parallel error feedback path and feedforward speed path, and output current control signal.

[0060] S6. Drive the brushless motor to perform the drip irrigation tray rotation operation according to the current control signal, and collect the current actual speed of the drip irrigation tray to construct the error sequence between the target speed and the actual speed;

[0061] S7. Input the error sequence into the dynamic calibration module based on sliding window weighting to generate real-time correction parameters for updating the current control signal, thereby achieving continuous synchronization between the drip irrigation tray and the tractor speed.

[0062] This invention achieves real-time tracking control of the tractor speed by the drip irrigation disc of the seeder through the construction of a complete process that integrates speed perception, predictive analysis, torque regulation and dynamic calibration, thereby improving the accuracy of drip irrigation and the synchronization of operations.

[0063] In this embodiment, the location information is collected at a fixed frequency by a Beidou navigation terminal installed on the top of the tractor, which collects the latitude and longitude coordinates and altitude data of the work path. The speed information includes the instantaneous travel speed obtained by the Beidou navigation terminal and the wheel axle angular velocity data collected by an incremental rotary encoder installed on the drive wheel axle of the tractor. The Beidou speed information and the encoder angular velocity information are linked by a timestamp to form an original speed data pair.

[0064] This invention employs a dual-source speed measurement method using a Beidou navigation terminal and an encoder, which improves the stability and accuracy of speed information and effectively reduces speed detection errors caused by slippage, signal interruption, etc.

[0065] In this embodiment, the preprocessing includes: performing a unified timestamp alignment operation on the collected location information and velocity information to construct a synchronous time-series data stream; converting the latitude and longitude coordinates in the location information into path coordinates in a two-dimensional plane coordinate system using a Gaussian projection algorithm; removing high-frequency noise components in the velocity information using a first-order low-pass filter; completing missing encoder data using cubic spline interpolation; and unifying velocity information from different sources to the same numerical scale range using a maximum-minimum normalization function.

[0066] This invention improves the spatiotemporal consistency and computability of input data by performing preprocessing operations such as timestamp alignment, coordinate transformation and noise filtering on the raw data, laying the foundation for subsequent speed fusion.

[0067] In this embodiment, S2 specifically includes:

[0068] S21. The preprocessed location information sequence and speed information sequence Input to the fusion computing unit, x i Let y represent the horizontal coordinate of the i-th sampling point. i v represents the vertical coordinate of the i-th sampling point. i Let N represent the instantaneous velocity of the i-th sampling point, and N represent the total number of sampling points.

[0069] S22. In the fusion computing unit, a sliding window of length w is used for velocity smoothing to construct a local velocity reference sequence. The calculation method is as follows:

[0070]

[0071] in, This represents the average velocity within the j-th window, where j = 1, 2, ..., N-w+1;

[0072] S23. In the fusion computing unit, the improved spatiotemporal weight calibration algorithm is invoked to construct the weight sequence. The calculation method is as follows:

[0073]

[0074] Among them, w i d represents the weighting coefficient for the i-th sampling point. i This represents the Euclidean distance between adjacent sampling points. | represents the speed difference, α, β, γ are weighting coefficients, satisfying 0≤α,β,γ≤1, and α+γ=1;

[0075] S24. Based on the velocity information sequence V, the weight sequence W, and the local reference velocity sequence... Perform the fusion operation and output the fused velocity sequence. The calculation method is as follows:

[0076]

[0077] in, The fusion rate of the i-th sampling point is represented by 1-w. i This represents the weighting coefficient for the local reference velocity.

[0078] This invention achieves dynamic adaptation to differences in location and speed by using a sliding window segmentation of sampling points and a weighted fusion algorithm, which significantly improves the continuity and representativeness of the fusion speed.

[0079] In this embodiment, the fusion computing unit receives the preprocessed position information sequence and velocity information sequence, performs sliding window segmentation, local reference extraction, weight generation and weighted calculation operations, and the improved spatiotemporal weight calibration algorithm constructs dynamic weighting factors in the fusion computing unit according to the changes in sampling point distance and velocity differences, and performs weighted fusion processing on the original velocity sequence based on the weighting factors to generate a fused velocity sequence.

[0080] This invention embeds a spatiotemporal weight calibration mechanism in the fusion computing unit, which can generate dynamic weights based on displacement distance and velocity fluctuations, thereby achieving more accurate velocity fusion results and improving the accuracy of pre-control.

[0081] In this embodiment, S3 specifically includes:

[0082] S31, Fuse the velocity sequence Input to the velocity prediction network, where The fusion rate of the i-th sampling point is represented by N, and the sequence length is represented by N.

[0083] S32. In the velocity prediction network, a subsequence feature matrix at different time scales is constructed using three-scale time windows τ1, τ2, τ3, defined as:

[0084]

[0085] Among them, F k The scale is τ k The subsequence feature matrix, where t represents the current time, τ kThis represents the length of the k-th time-scale window, where k∈{1,2,3};

[0086] S33. Apply an attention mechanism to the feature matrix of each subsequence at each scale to calculate the attention-weighted features. The calculation formula is as follows:

[0087]

[0088] Among them, Q k =F k ·W Q , where K is the query matrix. k =F k ·W K , where V is the bond matrix. k =F k ·W V , is a value matrix, W Q W K W V These are the trainable weight matrices, where softmax(·) represents the standard normalization function, and A is the weight matrix. k The scale is τ k Attention-weighted features;

[0089] S34. Concatenate the attention-weighted features A1, A2, and A3 at each scale along the time dimension to generate a velocity-guided feature map M, defined as:

[0090] M = concat(A1, A2, A3);

[0091] Where M represents the velocity-guided feature map, and concat(·) represents the matrix concatenation operation.

[0092] This invention enhances the ability to model different speed change trends and improves the system's response capability in variable speed operation scenarios by constructing a speed prediction network that combines a multi-scale sliding window with an attention mechanism.

[0093] In this embodiment, the velocity prediction network is a multi-scale sequence modeling network based on an attention mechanism. It consists of an input layer, three sets of sliding window extraction layers with different temporal scales, a feature mapping layer, an attention calculation layer, and a feature fusion output layer. Each scale sliding window corresponds to short-term, medium-term, and long-term velocity sequences, respectively. A query matrix, a key matrix, and a value matrix are generated through matrix calculation, and then weighted output features are calculated. The velocity guidance feature map is a temporal feature set composed of attention-weighted output features from each scale, representing the dynamic change trend of the fused velocity at different scales, and is used for the subsequent generation and control of the target rotational speed.

[0094] The velocity prediction network of this invention extracts velocity features through an attention mechanism and constructs a velocity-guided feature map, providing a structured input for the accurate generation of the target rotational speed and enhancing prediction stability and context sensitivity.

[0095] In this embodiment, S4 specifically includes:

[0096] S41. Input the velocity guidance feature map M into the velocity control mapping function f. m (·), generate the initial target rotational speed vector The calculation formula is:

[0097] r t =f m (M t ) = W m ·M t +b m ;

[0098] Where M represents the velocity-guided feature map unfolded in time sequence, M t ∈M represents the feature vector at the t-th time step, W m The weight matrix, b, represents the speed control mapping function. m R represents the bias vector. t Let T represent the initial target rotational speed vector generated at time t, where T represents the number of time steps in the feature map.

[0099] S42. Extract the task parameter set P task The initial target speed vector is modulated together with the task parameters to generate a modulated target speed sequence. The calculation formula is:

[0100]

[0101] Where, p k ∈P task Let λ represent the parameters of the k-th job task. k This represents the weight coefficient of the parameter for the k-th task. This represents the modulated target rotational speed vector at time t.

[0102] S43, Modulate the target rotation speed sequence R * The target rotational speed vector corresponding to the current time step As the output of control commands for the drip irrigation tray.

[0103] This invention generates a target rotation speed sequence by jointly modulating the speed guidance feature map with the task parameters, so that the drip irrigation rotation speed not only responds to the current speed state, but also takes into account the needs of the crop, thus achieving flexible sowing control.

[0104] In this embodiment, S5 specifically includes:

[0105] S51, Target rotation speed vector The corresponding drip irrigation tray control command is input to the brushless motor drive control unit and set as the target speed for the current control cycle;

[0106] S52. Construct the feedforward velocity path and calculate the feedforward torque control command, defined as:

[0107]

[0108] Among them, T f Indicates feedforward torque command, K v Indicates the velocity gain coefficient;

[0109] S53. Construct an error feedback path and obtain the actual rotational speed vector of the drip irrigation tray. Calculate the velocity error ε t Defined as:

[0110]

[0111] The proportional-integral (PI) control algorithm is used. This algorithm is a linear control method that calculates the torque command based on the weighted sum of the speed error in the current control cycle and the historical cumulative error. The calculation formula is as follows:

[0112]

[0113] Among them, T b K represents the error feedback torque command. p K represents the proportional gain. i ε represents the integral gain. t ε represents the speed error of the current control cycle. j represents the speed error of the j-th control cycle, and t represents the current control cycle number;

[0114] S54. Construct a dual-channel torque control structure to directly synthesize the feedforward torque command and the error feedback torque command, and output the total torque control command T. out ;

[0115] S55, according to the total torque control command T out Perform current conversion and generate current control signal I. c Defined as:

[0116]

[0117] Among them, I c K represents the current control signal. t This represents the motor torque constant.

[0118] This invention employs a dual-channel torque control structure, which enhances the dynamic response capability of the brushless motor through a feedforward speed path and an error feedback path, thereby improving the stability and tracking accuracy of the control system and reducing error accumulation.

[0119] In this embodiment, S6 specifically includes:

[0120] S61, based on the current control signal I c The brushless motor is driven to rotate the drip irrigation tray, initiating the rotation process of the current control cycle.

[0121] S62. During the rotational execution process, the actual rotational speed of the drip irrigation tray is collected in each control cycle to construct the actual rotational speed sequence R. act ,in This represents the actual speed vector during the t-th control cycle;

[0122] S63, Based on the target rotational speed sequence R * With the actual speed sequence R act Construct the error sequence E = {ε1, ε2, ..., ε T}, error ε at each time step t Calculated using the following formula:

[0123]

[0124] Where, ε t This represents the normalized velocity error in the t-th control cycle. Represents the target rotational speed vector. δ represents the actual rotational speed vector, and δ represents a non-negative small constant used to avoid the denominator being zero.

[0125] This invention constructs an error sequence based on actual rotation speed acquisition and normalized error calculation, ensuring that the drip irrigation tray's running trajectory closely follows the target command and improving the overall control closed-loop quality.

[0126] Example 1:

[0127] To verify the feasibility of this invention in practice, it was applied to a typical hilly dryland farming scenario. The required equipment was a medium-sized traction seeder with a drip irrigation system, which integrated a Beidou positioning terminal, axle encoder, brushless motor drive components, and an attention-based control algorithm processor. The farming area included multiple undulating roads with frequent speed changes, turns, and intermittent stops, making it a typical complex environment for testing the synchronization capability of the drip irrigation disc speed and the accuracy of the control response.

[0128] In this scenario, traditional drip irrigation control methods, employing mechanical constant-speed modes or simplified linear control algorithms, struggle to effectively adapt to the frequent speed fluctuations of tractors across different terrains. This results in uneven drip tape distribution, repeated watering, and missed irrigation. Particularly on undulating slopes, traditional methods exhibit significant delays in responding to tractor deceleration, severely impacting the precise utilization of water and fertilizer.

[0129] This invention inputs Beidou navigation data and wheel axle encoder speed data into a fusion calculation unit, and outputs the fused speed using an improved spatiotemporal weight calibration algorithm; it combines a speed prediction network with an attention mechanism to analyze short-term, medium-term, and long-term speed characteristic trends and generate a speed guidance feature map; it outputs a real-time target speed command through joint modulation of a mapping function and operating parameters; and finally, it controls the rotation of a brushless motor through a dual-channel torque control structure to achieve dynamic synchronization between the drip irrigation tray and the tractor speed.

[0130] In practical applications, the experimental team compared this system with a traditional linear control system. The total test area was approximately 3.8 kilometers long, including multiple slopes, continuous curves, and several deceleration-acceleration sections. Data collected included drip tape length error, drip flow rate deviation per unit area, response delay time, motor current fluctuation rate, and overall operational consistency score. After using this invention, the drip tape length error was significantly reduced to 2.3%, a reduction of approximately 62% compared to traditional control methods; the flow rate deviation per unit area was controlled within ±4.1%, an improvement of approximately 37% compared to traditional systems; the motor response delay under sudden speed changes decreased from 1.2 seconds to 0.35 seconds; and the spatial consistency score of the drip tape increased from 72.6 points to 93.4 points, resulting in more stable, continuous, and balanced operational performance.

[0131] Furthermore, on uphill sections with a slope greater than 8%, the traditional system suffers from a leakage irrigation rate as high as 18.7% due to lag in dynamic response, while the control system of this invention reduces the leakage irrigation rate to 5.2% on the same slope. Statistical analysis of the entire plot shows that water-saving efficiency is improved by approximately 14%, the fluctuation amplitude of motor control current is reduced by nearly 25%, and the overlap rate between sowing and drip irrigation points increases from 82.3% to 96.8%.

[0132] Table 1 Comparison of Control Accuracy and Operational Effectiveness of Drip Irrigation Trays

[0133]

[0134] Table 1 further verifies that the present invention is superior to the traditional solution in terms of drip irrigation control accuracy, response speed and resource utilization. The present invention not only effectively solves the problems of slow response and poor synchronization of the traditional drip irrigation tray control method in complex terrain, but also has higher accuracy, lower energy consumption and better operation consistency, and has significant practical value and promotion significance.

[0135] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for controlling the speed of a seeder's drip irrigation tray and variable speed based on BeiDou speed measurement, characterized in that, The steps include: S1. Acquire the position and speed information of the tractor during operation and perform preprocessing; S2. Input the preprocessed position and velocity information into the fusion computing unit, and use an improved spatiotemporal weight calibration algorithm to synchronously correct the multi-source data and output the fused velocity sequence. S3. Input the fused velocity sequence into the velocity prediction network based on the attention mechanism, extract velocity change features at different temporal scales, and output a velocity guidance feature map for control. S4. Generate a target rotation speed command for the drip irrigation tray based on the speed guidance feature map. The target rotation speed command is generated by the speed control mapping function and jointly modulated with the task parameters. S5. Input the target speed command to the brushless motor drive control unit, construct a dual-channel torque regulation structure through parallel error feedback path and feedforward speed path, and output current control signal. S6. Drive the brushless motor to perform the drip irrigation tray rotation operation according to the current control signal, and collect the current actual speed of the drip irrigation tray to construct the error sequence between the target speed and the actual speed; S7. Input the error sequence into the dynamic calibration module based on sliding window weighting to generate real-time correction parameters for updating the current control signal, thereby achieving continuous synchronization between the drip irrigation tray and the tractor speed.

2. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 1, characterized in that, The location information is collected by the Beidou navigation terminal, and the speed information is collected jointly by the Beidou navigation terminal and the encoder installed on the drive wheel shaft.

3. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 1, characterized in that, The preprocessing includes timestamp alignment, coordinate transformation, noise filtering, packet loss interpolation, and scale normalization.

4. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 1, characterized in that, S2 specifically includes: S21. Input the preprocessed position information sequence and velocity information sequence into the fusion calculation unit. The position information is the sequence of horizontal and vertical coordinates of the sampling point, and the velocity information is the instantaneous velocity sequence of the corresponding sampling point. S22. In the fusion computing unit, a sliding window process is performed on the velocity information sequence to generate a local velocity reference sequence within each window; S23. In the fusion computing unit, the improved spatiotemporal weight calibration algorithm is invoked to generate a corresponding weight sequence based on the current position coordinate changes and velocity differences. The weights are adjusted by both position distance and velocity deviation. S24. Perform weighted fusion based on the velocity information sequence and the weight sequence, and output the fused velocity sequence.

5. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement and control, as described in claim 4, is characterized in that... The fusion computing unit receives the preprocessed position information sequence and velocity information sequence, and performs sliding window segmentation, local reference extraction, weight generation and weighted calculation operations. The improved spatiotemporal weight calibration algorithm constructs dynamic weighting factors in the fusion computing unit according to the changes in sampling point distance and velocity differences, and performs weighted fusion processing on the original velocity sequence based on the weighting factors to generate a fused velocity sequence.

6. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 1, characterized in that, S3 specifically includes: S31. Input the fused velocity sequence into an attention-based velocity prediction network; S32. In the velocity prediction network, three sliding windows of different time lengths are set to construct velocity subsequence feature matrices of three scales, representing short-term, medium-term and long-term velocity change sequences respectively. S33. At each time scale, the velocity change feature representation of the corresponding scale is calculated using the attention mechanism. Each feature representation is generated by a weighted sum of the query matrix, the key matrix and the value matrix. S34. The velocity change features at the three scales are spliced ​​together in chronological order to construct a velocity guidance feature map, which is used to drive the generation process of the target rotational speed.

7. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 6, characterized in that, The velocity prediction network is a multi-scale sequence modeling network based on an attention mechanism. It consists of an input layer, three sets of sliding window extraction layers with different temporal scales, a feature mapping layer, an attention calculation layer, and a feature fusion output layer. Each scale sliding window corresponds to short-term, medium-term, and long-term velocity sequences, respectively. The query matrix, key matrix, and value matrix are generated through matrix calculation, and then the weighted output features are calculated. The velocity-guided feature map is a temporal feature set composed of attention-weighted output features at each scale, representing the dynamic change trend of the fused velocity at different scales, and is used for the generation and control of the target rotational speed.

8. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 6, characterized in that, S4 specifically includes: S41. Input the velocity-guided feature map into the velocity control mapping function to generate an initial target rotational speed vector arranged by time steps; S42. Extract the set of task parameters, and jointly modulate the initial target speed vector with the task parameters to generate a modulated target speed sequence. The target speed sequence is achieved by applying a weighting factor of the task parameters to the initial target speed. S43. Extract the target rotation speed corresponding to the current time step from the modulated target rotation speed sequence, and output the target rotation speed as the control command of the drip irrigation tray.

9. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 1, characterized in that, S5 specifically includes: S51. Input the control command of the drip irrigation tray corresponding to the modulated target speed to the brushless motor drive control unit as the expected speed of the current control cycle; S52. Construct the feedforward speed path and calculate the feedforward torque control command based on the target speed and speed gain; S53. Construct an error feedback path, obtain the actual speed and calculate the error with the target speed, and generate an error feedback torque control command through a proportional-integral (PI) control algorithm. The PPI control algorithm is a linear control method that calculates the torque command based on a weighted sum of the speed error of the current control cycle and the historical cumulative error. S54. Through the dual-channel torque control structure, the torque commands of the feedforward channel and the error feedback channel are directly synthesized to output the total torque control command. S55. Perform current conversion based on the total torque control command to generate a current control signal for driving the brushless motor.

10. The method for controlling the speed of a drip irrigation tray and shifting gears in a seeder based on Beidou speed measurement control according to claim 1, characterized in that, S6 specifically includes: S61. Drive the brushless motor to perform the drip irrigation tray rotation operation according to the current control signal, and enter the execution stage of the current control cycle; S62. Collect the actual rotation speed of the drip irrigation tray in real time within the control cycle, and construct the actual rotation speed sequence according to the time step; S63. Compare the target speed sequence with the actual speed sequence one by one to construct an error sequence. The error is the normalized absolute error, which is used for error compensation adjustment.

Images