Tractor and implement cooperative operation under the action of precision synchronization control method of inter-plant weeding and seedling avoidance

Abstract

The application relates to the field of agricultural machinery automatic control, and discloses a precision synchronous control method for inter-plant weeding and seedling-avoiding actions under tractor and machine tool cooperative operation. The method uses a visual sensor to identify crops and maps a physical distance to a shift register stack; millimeter wave radar and inertial measurement unit data are collected, an equivalent longitudinal velocity vector of a weeding tool relative to the ground is solved based on a hitching geometric model, and interference of wheel slip and machine tool swing on velocity calculation is eliminated; a displacement pulse is calculated by using a residual accumulation strategy, and shift registers are synchronously driven to suppress quantitative error accumulation; online correction coefficients are updated in combination with real-time hydraulic state parameters and action feedback, and a dynamic advance trigger distance is calculated. The application realizes accurate decoupling of multi-body kinematics and compensation of time-varying characteristics of a hydraulic actuator, and improves inter-plant weeding and seedling-avoiding precision and operation stability in an unstructured farmland environment.

Description

Technical Field

[0001] This invention relates to the field of automated control of agricultural machinery, specifically a method for precise synchronous control of weeding and seedling avoidance actions between plants under the cooperative operation of tractors and implements. Background Technology

[0002] Interplant mechanical weeding is a key step in reducing the use of agricultural chemical herbicides and achieving ecological agriculture. During interplant weeding operations, the weeding components need to quickly extend and retract between crop plants, which places high demands on the position synchronization accuracy and action response consistency of the control system.

[0003] Existing weeding machines primarily operate using tractor-tethered operation. Regarding implement positioning and speed sensing, current technologies typically utilize tractor wheel speed sensor data or chassis GNSS speed as the motion reference for the weeding implement, treating the tractor and implement as a rigidly connected unit for position calculation. However, in actual farmland environments, loose soil can cause significant slippage of the tractor's drive wheels, making the wheel speed data inaccurate in reflecting the vehicle's speed relative to the ground. Furthermore, the weeding implement is usually attached to the rear of the tractor via a three-point suspension system. This connection is not ideally rigid; the implement will yaw and sway relative to the tractor due to ground undulations and changes in traction resistance. Ignoring this relative motion leads to a deviation between the actual velocity vector of the weeding blade tip and the theoretically calculated value, resulting in longitudinal errors in the working position.

[0004] In terms of control algorithm implementation, to adapt to different operating speeds, control systems often employ a shift register model to map physical distances to logical storage bits. This discretization process inevitably involves quantization and rounding of continuous physical displacements. Existing control logic often fails to handle the rounding residuals during the quantization process. Although the error in a single calculation is small, in long-distance continuous operations, these small errors accumulate unidirectionally, causing the crop marker position in the logical space to gradually deviate from its actual position in the physical space, resulting in a continuous drift in weeding accuracy.

[0005] In actuator control, hydraulic drive systems are widely used due to their high power density. Existing control systems typically use fixed action delay time parameters to compensate for the response lag of hydraulic systems. However, the viscosity of hydraulic oil is extremely sensitive to temperature changes, and the hydraulic circuit pressure fluctuates with engine speed and load. During long-term operation, as the hydraulic oil temperature rises, the oil viscosity decreases, causing the actuator's response speed to drift. Fixed control parameters cannot adapt to these time-varying physical characteristics, resulting in inconsistent timing of weeding blade action during periods of large temperature differences (morning and evening) or before and after equipment thermal equilibrium, affecting the success rate of weeding operations. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a precise synchronous control method for weeding and seedling avoidance actions between plants in tractor-implant cooperative weeding operations. This method solves the problem that in existing tractor-implant cooperative weeding operations, the lateral sway of the implement relative to the traction platform leads to deviations in the calculation of the weeding blade position. At the same time, the accumulation of quantization errors generated by traditional discrete control and the response delay drift of the hydraulic system with changes in oil temperature further reduce the synchronization accuracy of the seedling avoidance actions between plants, causing crop damage or missed weeding.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements, comprising the following steps: The system uses a visual sensor to identify crop seedlings and calculates the longitudinal physical distance between the crop seedlings and the weeding blades. The longitudinal physical distance is then mapped to the shift register stack in the control unit to generate crop tags.

[0008] The inertia and velocity data of the traction platform and the weeding implement are collected, and the equivalent longitudinal velocity vector of the weeding blades relative to the ground is calculated based on the coupling geometry model. Multibody kinematics decoupling calculations are used to eliminate the interference of wheel slippage and implement oscillation on the velocity calculation.

[0009] The displacement pulse increment is calculated using the equivalent longitudinal velocity vector, which drives the read and write pointers of the shift register stack to perform shift operations, so that the logical position of the crop marker in the shift register stack is synchronized with the actual position of the crop seedling.

[0010] The pressure and temperature values ​​of the hydraulic circuit are obtained, and the dynamic advance triggering distance of the weeding execution component is calculated by combining the preset response model and online correction coefficient to compensate for the response delay of the hydraulic system caused by pressure fluctuations and changes in oil viscosity.

[0011] The system detects crop markers located at the dynamic advance trigger distance in the shift register stack. When a crop marker is detected, an avoidance action instruction is generated, and feedback data during the action process is collected to update the online correction coefficient.

[0012] Furthermore, the process of mapping the longitudinal physical distance to the shift register stack to generate crop markers includes: constructing a shift register stack with a circular buffer structure and setting the spatial resolution represented by each unit of storage bit; obtaining the reference read / write pointer of the weeding blade in the shift register stack at the current moment; dividing the longitudinal physical distance by the spatial resolution and rounding down to obtain the distance offset; adding the reference read / write pointer to the distance offset and taking the modulo of the total capacity length of the shift register stack to obtain the target write index; and setting the storage unit in the shift register stack corresponding to the target write index to the valid state. To accommodate visual recognition errors and crop swaying, a continuous crop protection window is set within the adjacent range of the target write index.

[0013] Furthermore, in the process of solving the equivalent longitudinal velocity vector, the data preprocessing includes: calculating the main vibration center frequency based on the real-time engine speed and the number of cylinders, updating the coefficients of the notch filter accordingly, performing frequency domain adaptive filtering on the collected angular velocity data to filter out mechanical vibration noise; comparing the absolute longitudinal velocity collected by the millimeter-wave radar with the preset zero-velocity judgment threshold; when the absolute longitudinal velocity is less than the zero-velocity judgment threshold, the judgment device enters a stationary state, and the output angular velocity is forcibly set to zero to suppress zero-biased integral drift; when the absolute longitudinal velocity is greater than or equal to the zero-velocity judgment threshold, the filtered angular velocity data is output for subsequent calculations.

[0014] Furthermore, the process of calculating the equivalent longitudinal velocity vector of the weeding blade relative to the ground based on the attachment geometric model includes: establishing a coordinate system and using a discrete-time integration algorithm to calculate the attachment hinge angle between the traction platform and the weeding implement; projecting the absolute longitudinal velocity of the traction platform onto the longitudinal axis of the implement using the attachment hinge angle to obtain the traction velocity component; calculating the tangential linear velocity generated at the end of the weeding blade lever arm due to the swing of the weeding implement around the attachment point as the rotational compensation component; and vector superimposing the traction velocity component and the rotational compensation component to obtain the equivalent longitudinal velocity vector at the end of the weeding blade.

[0015] Furthermore, the process of calculating the displacement pulse increment using the equivalent longitudinal velocity vector employs a residual accumulation strategy, specifically including: multiplying the equivalent longitudinal velocity vector by the sampling time interval within each control cycle to calculate the physical micro-displacement of the current cycle; accumulating the physical micro-displacement with the displacement residual of the previous cycle to obtain the total accumulated displacement; dividing the total accumulated displacement by the spatial resolution of the shift register stack and rounding down to extract the integer part of the displacement pulse increment; subtracting the physical length corresponding to the displacement pulse increment from the total accumulated displacement, and storing the remaining fractional part as the new displacement residual in the next control cycle; updating the read pointer of the shift register stack according to the displacement pulse increment to achieve cyclic shifting of the circular buffer.

[0016] Furthermore, the process of calculating the dynamic advance trigger distance of the weeding execution component includes: pre-constructing a two-dimensional response time mapping table indexed by hydraulic circuit pressure and hydraulic oil temperature; retrieving the basic response time from the mapping table using an interpolation algorithm based on the real-time collected pressure and temperature values; adding the basic response time to the online correction coefficient to obtain the predicted response delay time; calculating the product of the equivalent longitudinal velocity vector and the predicted response delay time, and superimposing a preset mechanical compensation distance to synthesize the dynamic advance trigger distance used for control triggering.

[0017] Furthermore, the instruction generation process includes: dividing the dynamic advance trigger distance by the spatial resolution and rounding it to convert it into a logical offset address; adding the current read pointer to the logical offset address and taking the modulo to calculate the target detection index; reading the logical state value corresponding to the target detection index in the shift register stack; if the logical state value indicates that there is a valid marker, then sending an avoidance action instruction to the drive circuit and recording the timestamp of the instruction sending time.

[0018] Furthermore, the process of updating the online correction coefficients adopts a closed-loop feedback mechanism: physical motion signals are detected by displacement sensors or switch sensors installed at the end of the hydraulic actuator; the moment when the physical motion signal reverses is captured as the start time of the motion; the time difference between the start time of the motion and the command sending time is calculated to obtain the actual physical lag; if the time difference is within the safe threshold range, the observation residual between the actual physical lag and the predicted response delay time used by the current model is calculated; using the exponentially weighted moving average algorithm, the online correction coefficients of the previous period are added to the product of the observation residuals and the learning rate coefficient to obtain the updated online correction coefficients.

[0019] Furthermore, the updated online correction coefficients are subjected to saturation processing: it is determined whether the updated online correction coefficients exceed the preset correction range; if they exceed the range, they are clamped to the boundary value of the correction range, and the clamped value is used for the prediction calculation of the next control cycle.

[0020] This invention provides a method for precise synchronous control of weeding and seedling avoidance actions between plants in a tractor-implant coordinated operation. It has the following beneficial effects: 1. This invention achieves kinematic decoupling between the tractor and implements through multi-source sensor data fusion. By using millimeter-wave radar and inertial measurement unit, wheel slippage interference is eliminated and the swing component of the implement around the attachment point is accurately calculated, thereby obtaining the true equivalent ground speed of the weeding blade tip. This eliminates the speed calculation deviation caused by neglecting suspension clearance and implement sway in traditional rigid models, ensuring the positioning accuracy between plants under conditions of slippery ground or implement attitude fluctuation, and effectively reducing the seedling damage rate in weeding operations.

[0021] 2. This invention employs a displacement pulse quantization strategy based on residual accumulation. When converting continuous physical displacements into discrete shift register control signals, the system automatically buffers and compensates for tiny displacement residuals smaller than the resolution unit, avoiding the cumulative effect of single quantization errors in long-distance operations. This ensures that the crop marker position in the logical mapping space is always strictly synchronized with the physical space, preventing phase shifts in weeding actions caused by accumulated errors and improving the system stability of long-ridge operations.

[0022] 3. This invention constructs a hydraulic adaptive control model that includes physical state perception and closed-loop feedback. By monitoring the pressure and temperature of the hydraulic circuit in real time and dynamically updating the time correction coefficient in combination with the actual action feedback of the actuator, the system can automatically offset the response time drift caused by changes in oil viscosity, thermal decay of solenoid valves or mechanical wear, ensuring that the weeding blades can maintain consistent action timing and operation effect throughout the day and the entire life cycle of the equipment. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the machinery seedling avoidance synchronization control system architecture of the present invention; Figure 2 This is a flowchart of the machinery seedling avoidance synchronization control method of the present invention; Figure 3 This is a comparative schematic diagram of the estimation of the equivalent speed at the end of the implement in this invention; Figure 4 This is a schematic diagram comparing the synchronization accuracy of the two control methods of the present invention; Figure 5 This is a schematic diagram of the online self-calibration response characteristics based on temperature-varying operating conditions according to the present invention.

[0024] Among them, 100 is the traction platform; 200 is the weeding implement; 210 is the weeding blade; 300 is the control unit; 310 is the vision sensor; and 320 is the millimeter-wave radar. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see the appendix Figure 1This invention provides a method for precise synchronous control of weeding and seedling avoidance actions between plants under the cooperative operation of tractor and implement. The method includes a synchronous control system, which mainly includes a traction platform 100, a weeding implement 200, a suspension mechanism, and a control unit 300.

[0027] The towing platform 100 is the chassis of the agricultural machinery that provides power output, typically a wheeled tractor. The weeding implement 200 is the working component that performs inter-row weeding operations. The weeding implement 200 is mechanically connected to the rear of the towing platform 100 via a suspension mechanism. The suspension mechanism is a non-rigid connection component, specifically a three-point suspension device, which allows the weeding implement 200 to perform relative movements in the yaw, roll, and pitch directions relative to the towing platform 100.

[0028] The weeding implement 200 is equipped with a weeding execution assembly, which includes a weeding blade 210, a hydraulic cylinder, and a proportional control valve. The piston rod end of the hydraulic cylinder is mechanically connected to the weeding blade 210, driving the blade to perform lateral extension and retraction movements between crop rows to avoid crop seedlings. The proportional control valve is connected in series in the hydraulic circuit, controlling the movement of the hydraulic cylinder by adjusting the flow and direction of the hydraulic oil.

[0029] The control unit 300 is the core processing center of the synchronous control system, establishing electrical or signal connections with the sensing and actuating elements in the synchronous control system. The control unit 300 is equipped with a first-in-first-out shift register stack for simulating physical space and is used to perform kinematic decoupling and hydraulic compensation calculations.

[0030] To achieve accurate perception of the kinematic state of multiple bodies, the synchronous control system is equipped with a multi-source sensor network. The multi-source sensor network includes a vision sensor 310, a millimeter-wave radar 320, a first inertial measurement unit, and a second inertial measurement unit.

[0031] The vision sensor 310 is mounted on the front of the frame of the weeding implement 200 and is located upstream of the weeding blade 210 in the direction of travel. The field of view of the vision sensor 310 covers the crop row and is used to acquire image data containing crop seedlings and to resolve the longitudinal physical distance of the crop seedlings relative to the weeding blade 210.

[0032] The millimeter-wave radar 320 is rigidly mounted under the chassis of the traction platform 100, with its detection beam pointing towards the ground. The millimeter-wave radar 320 is used to measure the absolute longitudinal speed of the traction platform 100 relative to the ground in real time. The absolute longitudinal speed value does not depend on the wheel speed, thus eliminating interference from wheel slip rate.

[0033] The first inertial measurement unit is installed at the geometric center or rear axle center of the traction platform 100, and is used to measure the triaxial angular velocity and acceleration data of the traction platform 100 in the global coordinate system. The second inertial measurement unit is installed at the center of the crossbeam of the weeding implement 200, and is used to measure the triaxial angular velocity and acceleration data of the weeding implement 200.

[0034] The first inertial measurement unit and the second inertial measurement unit together constitute a dual-body attitude sensing system, which is used to calculate the relative hinge angle and instantaneous lateral sway between the traction platform 100 and the weeding implement 200.

[0035] To monitor the physical condition of the hydraulic power circuit, a pressure sensor and a temperature sensor are installed in the main oil inlet line of the hydraulic circuit. The pressure sensor is used to collect the main pressure value of the hydraulic power circuit in real time. The temperature sensor is used to collect the temperature value of the hydraulic oil in real time.

[0036] The control unit 300 is connected to the vision sensor 310, millimeter-wave radar 320, first inertial measurement unit, second inertial measurement unit, pressure sensor and temperature sensor respectively via signal harness to acquire real-time data collected by the sensors.

[0037] The output of the control unit 300 is electrically connected to the proportional control valve and is used to send control command signals to the proportional control valve. The control command signals are generated based on the calculation results of sensor data and are used to precisely control the timing of the weeding blades 210.

[0038] See appendix Figure 2 This invention provides a method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements, comprising the following steps: S100: The vision sensor 310 is used to identify crop seedlings and calculate the longitudinal physical distance between the crop seedlings and the weeding blade 210. The longitudinal physical distance is mapped to the shift register stack in the control unit 300 to generate crop markers. S200 uses the angular velocity data collected by the first inertial measurement unit and the second inertial measurement unit, combined with the engine speed, to perform frequency domain notch filtering, and uses the absolute longitudinal velocity collected by the millimeter-wave radar 320 to perform zero-speed clamping, and outputs purified angular velocity data. S300, based on the purified angular velocity data and the absolute longitudinal velocity collected by the millimeter-wave radar 320, and based on the geometric model of the connection between the traction platform 100 and the weeding implement 200, calculates the equivalent longitudinal velocity vector of the weeding blade 210 relative to the ground. S400 uses the equivalent longitudinal velocity vector to calculate the displacement pulse increment, drives the shift register stack to perform shift operations, and keeps the logical position of the crop marker in the shift register stack synchronized with the actual position of the crop seedling in physical space. S500 acquires the main pipe pressure value collected by the pressure sensor and the hydraulic oil temperature value collected by the temperature sensor. Combined with the preset response model and online correction coefficient, it calculates the predicted response delay time and the corresponding dynamic advance trigger distance of the weeding execution component. The S600 detects crop markers within the dynamic advance trigger distance range in the shift register stack. When a crop marker is detected, it sends an action command to the proportional control valve and collects pressure characteristic data during the action process to update the online correction coefficient.

[0039] S111, construct and initialize the virtual mapping space used to characterize the job path. Allocate a contiguous linear address space in the volatile memory of the control unit 300 as a shift register stack. The stack length is... It needs to cover the total physical travel of the weeding machine 200 from the sensing point to the work point and then to the maximum predicted distance. In this embodiment, the stack length... Determined based on the following constraints: ; in, This represents the maximum effective detection range of the vision sensor 310. The maximum permissible operating speed for the traction platform 100. This represents the maximum possible response delay of the hydraulic system. This is the set spatial resolution. Presetting this length prevents stack address overflow caused by excessively fast processing speeds or significant system lag. (Set spatial resolution) This refers to a tiny physical length element represented by a single bit in the register, designed to meet the centimeter-level control precision requirements of weeding operations. The value is usually set to 1 / 2 to 1 / 5 of the allowable positioning error limit of the weeding tool 210, for example, a value of 1 mm / bit.

[0040] S112, acquire an image data stream containing crop row features. The vision sensor 310 continuously exposes and acquires data on the crop area in the direction of travel of the weeding machine 200 at a preset frame rate. To address unstructured environmental interference such as changes in field lighting and crop shading, the control unit 300 can perform histogram equalization or Gaussian filtering preprocessing on the original image to enhance the salience of features at the crop root and stem locations. Such image preprocessing methods are conventional techniques in the field of digital image processing and will not be elaborated upon here.

[0041] S113, calculate the instantaneous physical distance between the crop seedling and the actuator. Use a feature extraction algorithm to identify the pixel coordinates of the rootstock base point of the crop seedling in the image. Based on pre-calibrated camera intrinsic and extrinsic parameter matrices, the planar coordinates in the pixel coordinate system are transformed into three-dimensional spatial coordinates in the machine coordinate system. In this embodiment, the origin of the implement coordinate system is defined as the rotation center or reciprocating motion centerline of the weeding blade 210. The component values ​​of this coordinate system along the implement's travel direction (i.e., the X-axis of the implement coordinate system) are extracted to obtain the longitudinal physical distance of the crop seedlings relative to the weeding blade 210. It should be noted that if the calculated... A negative value indicates that the crop has passed the point of action of the cutting tool, and the system will automatically discard the invalid data to save computing resources.

[0042] S114, performs a quantization mapping from physical distance to logical index. Based on the calculated vertical physical distance. and preset spatial resolution The relative mapping position of the crop seedling in the shift register stack is calculated. To eliminate the uncertainty introduced by floating-point operations and to achieve address alignment, the mapping process uses the following floor function: ; in, This is the mapped crop target index address. This is the stack read / write pointer index corresponding to the current position of weeding tool 210. This is the floor operator. This represents a modulo operation on the total stack length to achieve circular addressing of the circular buffer. The physical meaning of this formula is that it represents the continuously varying vertical physical distance... By spatial resolution Perform grid-based partitioning to determine which logical storage unit it falls into.

[0043] S115, Generate and write the crop existence marker. Control unit 300 accesses the calculated crop target index address. The storage bit corresponding to that address is set to a logic high level "1". As a preferred implementation, to address visual recognition errors or crop swaying caused by wind, the system can simultaneously set the crop target index address... Adjacent storage bits (e.g.) Both bits are set to "1", thus constructing a crop protection window with a certain tolerance in the logical space. The physical width of this window is... This effectively prevents accidental damage to crops due to system discretization errors. After the write operation is completed, the shift register stack is transformed from its original empty state into a spatial mapping band carrying crop position information, providing a data foundation for subsequent synchronous shift control.

[0044] S211 performs time synchronization and acquisition of multi-source heterogeneous data. The control unit 300 acquires the raw angular velocity of the traction platform from the first inertial measurement unit 430 in parallel via the CANFD bus or the vehicle Ethernet interface. The initial angular velocity of the second inertial measurement unit 440 and the real-time engine speed broadcast by the engine electronic control unit (ECU). Considering the differences in sampling frequency and transmission delay among different sensors (for example, inertial sensors are typically 100Hz-200Hz, while engine speed data is typically 10Hz-50Hz), the control unit 300 uses nearest neighbor interpolation or linear interpolation algorithm to map low-frequency engine speed data onto the high-frequency sampling time axis of the inertial measurement unit, thereby achieving strict alignment of multi-source data in the time dimension.

[0045] S212, construct a parameter-adaptive notch filter based on engine order. Given that the reciprocating piston motion of the tractor engine is the main excitation source of high-frequency chassis vibration, and that this vibration frequency changes linearly with throttle opening, a fixed-frequency filter cannot simultaneously handle both idle and full-load conditions. The control unit 300, based on the current real-time speed... (rpm) Calculate the fundamental center frequency of mechanical vibration. : ; in, The number of engine cylinders. Convert the rotational speed to revolutions per second. This term corresponds to the ignition order characteristics of a four-stroke internal combustion engine. Based on this real-time calculation... The system dynamically updates the coefficients of the second-order infinite impulse response (IIR) notch filter. In this embodiment, in order to filter out vibration noise while preserving the low-frequency oscillation characteristics (typically less than 5Hz) during normal machine operation, the quality factor of the notch filter needs to be strictly limited. . The value range is set to 5 to 10, and the calculation relationship is as follows: ,in It has a -3dB stopband bandwidth. (Higher) The value setting ensures that the filter has an extremely narrow stopband width, thereby avoiding phase distortion or amplitude attenuation of the effective motion signal. (Original angular velocity signal of the traction platform) With the original angular velocity signal of the machine After passing through the filter, the vibration component coupled with the engine speed is attenuated in a specific direction.

[0046] S213, Implement Zero-Voltage Clamping (ZUPT) Correction Based on Earth Observation. Although frequency domain filtering removes high-frequency noise, the microelectromechanical system (MEMS) gyroscope still exhibits temperature- and time-varying zero-bias drift in the low-frequency band. This drift, when integrated, leads to severe divergence in the attitude angle. The control unit 300 incorporates the longitudinal ground velocity measured by the millimeter-wave radar 320. As an unbiased reference standard, radar speed, rather than wheel speed, was chosen as the criterion because on soft farmland soil, even when the vehicle is stationary, wheels may generate false speed signals due to slippage or creep, while radar speed measurement is unaffected by wheel-soil interaction. A zero-speed threshold was set. This threshold is determined based on the background noise level of the radar sensor, and is typically set to between 0.03 m / s and 0.08 m / s. The signal correction logic is as follows: ; in, The purified angular velocity is the final component used in the kinematics solution. This is the intermediate variable after filtering in step S212. When detected... When the value falls below a threshold, the system determines that the machine is physically stationary and forcibly pulls the output angular velocity back to absolute zero. This hard threshold clamping strategy can periodically cut off the accumulation link of integral error, especially in low-speed conditions such as turning at the edge or temporary stopping, which can improve the long-term operational stability of the system.

[0047] S311, Construct a multi-body system coordinate system and calibrate key geometric parameters. Establish a vehicle coordinate system with the projection point of the rear axle center of the traction platform 100 onto the horizontal plane as the origin. ,definition The axle points in the direction of travel along the longitudinal centerline of the vehicle body; a coordinate system for the implement is established with the center of the suspension mechanism's engagement hinge point as the origin. The axis points from the longitudinal axis of the machine's main beam to the tail of the machine. To achieve high-precision kinematic calculations, the following key structural parameters need to be pre-determined and fixed: lever arm length. The longitudinal straight-line distance from the machine's rotation center (i.e., the hinge point) to the working point of the weeding blade 210; radar installation offset vector. This characterizes the spatial positional relationship between the measurement center of the millimeter-wave radar 320 and the origin of the vehicle coordinate system. As a preferred implementation, these parameters can be obtained through high-precision RTK-GNSS point measurement or by consulting mechanical design drawings, and stored in the non-volatile storage area of ​​the control unit 300 after the system's initial calibration.

[0048] S312, calculate the real-time relative pose between the traction platform and the weeding implement. Given the time-varying drift characteristics of direct integration by the inertial measurement unit (IMU), this embodiment employs a strategy of fusing differential angular velocity integration with absolute observations to obtain a highly reliable articulation angle. The control unit 300 performs discrete-time integration, and its state update equation is as follows: ; in, Indicates the sequence number of the current discrete control cycle. This represents the hinge angle state value stored by the system at the end of the previous control cycle, which serves as the integration reference for the current cycle. This represents the difference in relative angular velocity measured within the current control cycle, i.e. This characterizes the instantaneous rotational rate of the implement relative to the traction platform during the cycle. The sampling time interval of the system is defined as follows. To eliminate the cumulative integration error, the system introduces an observation correction mechanism: when the millimeter-wave radar 320 detects that the vehicle is traveling in a straight line, it utilizes the GNSS dual-antenna heading difference or visual observation line characteristics at this time to correct the error. Perform zeroing or absolute value calibration to ensure the convergence of hinge angle data over long-term operation.

[0049] S313, derive and synthesize the true equivalent longitudinal velocity of the tip of the weeding blade. The absolute velocity of the weeding blade 210 relative to the ground. It includes not only the traction speed component provided by the traction platform 100, but also the tangential linear velocity component generated by the oscillation of the implement around the attachment point. According to the velocity composition theorem, the weeding blade 210 along the longitudinal axis of the implement... Effective feed rate in direction This can be expressed as: ; in, The longitudinal ground velocity of the traction platform as measured by millimeter-wave radar 320; The sideslip angle at the center of gravity of the traction platform's rear axle is approximately estimated based on the Ackermann steering geometry model or can be ignored at small angles when a dedicated sideslip angle sensor is unavailable. (The formula contains...) The term represents the geometric projection coefficients used to project the velocity vector of the traction platform from the vehicle coordinate system. Projected onto the machine coordinate system In the vertical direction, this term reflects the attenuation of the effective component of traction speed due to machine deflection; in the formula... The term is the rotational linear velocity compensation term, used to calculate the effect of the machine's rotation around the engagement point on the lever arm. The tangential velocity generated at the end point reflects the direct additive effect of the tool oscillation on the end point velocity. When the tool oscillates towards the working side, the actual ground velocity of the cutter increases; conversely, it decreases. The system utilizes this synthesized velocity... As a reference velocity for subsequent spatial synchronization control, it effectively eliminates position control errors caused by the simple rigid body assumption.

[0050] S411 calculates the physical displacement increment within a single control cycle. The control unit 300 uses a high-precision hardware timer based on a real-time operating system (RTOS) with a fixed sampling period. (For example, 10ms) Trigger the displacement calculation task. Read the equivalent longitudinal velocity of the weeding tool tip output in step S300. The discrete-time integral algorithm is used to calculate the instantaneous physical displacement increment of the weeding blade 210 along the crop row direction during the current cycle. : ; Among them, the instantaneous physical displacement increment This characterizes the actual feed rate of the tool within the current time slice, taking into account the coupling effect of vehicle traction and tool sway. To prevent malfunctions in non-operating states, the control unit 300 verifies the system state machine before calculation. If a malfunction is detected... If the value is negative (i.e., the machine is reversing), the system will force the instantaneous physical displacement increment to be negative. Alternatively, subsequent shift operations can be paused to avoid control chaos caused by logical address rollback.

[0051] S412 performs pulse quantization conversion based on accumulated residual compensation. Given that the shift register stack is a digital space constructed based on discrete bits, and the instantaneous physical displacement increment... For continuously changing floating-point values, direct rounding or integer manipulation can introduce cumulative position errors during long-term operations. Therefore, this embodiment constructs an accumulator model with error memory functionality, incorporating the small displacement residuals from the previous cycle that were not quantized into the current calculation cycle. (Corrected pulse count) and state residuals The calculation is as follows: ; ; ; in, This represents the current total cumulative displacement, including historical residuals. This is a zero-rounding operator used to extract the complete displacement steps; The spatial resolution of the shift register stack is a parameter stored as a preset constant in the system configuration area. The remaining sub-pixel physical distance after quantization in the current cycle will be cached for computation in the next cycle. The physical significance of this algorithm lies in using distances smaller than a spatial resolution... The tiny displacements are temporarily stored in the residual variables. In the meantime, until it accumulates to exceed the spatial resolution It is released as a shift pulse at one time, thus realizing lossless integral tracking of continuous physical motion in a digital discrete system.

[0052] S413 drives the shift register stack to perform a phase-locked shift. The control unit 300 then performs the shift based on the calculated number of pulses. The system performs logical shift operations on the shift register stack. As a preferred implementation, to improve processing efficiency and reduce memory overhead, the shift register stack is implemented using a circular buffer data structure. The control unit 300 does not directly move data bits in memory, but instead implements logical shifts by updating read / write pointers. ; in, This is the read pointer index representing the current position of the weeding blade. This is the total length of the stack. (As the pointer...) The update process logically shifts the data in the buffer relative to the pointer, synchronizing the crop marker position with the physical implement position. For newly entered areas within the detection range, the system writes new visual recognition results to the corresponding address at the end of the buffer (as described in step S100), while old data exceeding the maximum predicted distance is automatically overwritten and discarded. This mechanism ensures that the logical space always maps to agricultural information within a specific range in front of the weeding blade.

[0053] S511 acquires and preprocesses the state characteristic parameters of the hydraulic power system. The control unit 300 synchronously reads the system pressure value output by the pressure sensor 450 installed at the main oil inlet of the hydraulic valve group via an analog signal acquisition interface. And the oil temperature value output by the temperature sensor 460 installed on the return side of the hydraulic oil tank. To eliminate pressure noise caused by the periodic pulsation of the plunger pump and the high-frequency switching action of the solenoid valve, the control unit 300 applies a window length of [length missing] to the original pressure signal. (For example The algorithm employs a moving average filtering process. Simultaneously, to ensure system robustness, signal validity verification logic is added to the algorithm input: if a signal is detected... or Exceeding the preset physical reasonable range (e.g.) or This indicates that the sensor may have an open circuit or short circuit fault, and the system will automatically switch to the pre-stored nominal operating parameters (such as...). It participates in subsequent calculations and reports fault codes.

[0054] S512, based on multi-dimensional response characteristic map retrieval of fundamental physical delays. The response speed of a hydraulic system is affected by the dual coupling of oil viscosity (determined by temperature) and driving force (determined by pressure): high viscosity at low temperatures leads to increased resistance of valve core movement, while low pressure results in insufficient cylinder starting acceleration. Based on this physical characteristic, this embodiment pre-constructs a two-dimensional response characteristic mapping table. This mapping table was obtained through offline bench testing. Specifically, at pressure nodes (e.g., every 2 MPa) and temperature nodes (e.g., every 5°C) covering the operating range, step commands were sent to the solenoid valve, and the physical time from energization to completing 90% of the stroke was recorded. The control unit 300 then uses real-time data... and Using bilinear interpolation algorithm from the mapping table The basic response time under the current operating conditions is analyzed in the middle. : ; in, It is a bilinear interpolation operator; This is an amplitude limiting function used to force the input parameters to be constrained to boundary values ​​when they exceed the calibration range of the mapping table, thus preventing unreasonable extreme values ​​from being generated by extrapolation calculations. and These are the minimum and maximum physically permissible response times of the hydraulic system (e.g., ...). This step ensures that the model output always conforms to the physical limits of the hydraulic system.

[0055] S513, synthesizes the dynamic lead-triggered distance including an online correction term. Basic response model. This only reflects the standard characteristics of the system at the time of manufacture and cannot characterize the slow time-varying characteristics caused by seal wear, oil contamination, or thermal decay of the electromagnetic coil. Therefore, a closed-loop correction mechanism is introduced into the system. This is combined with the equivalent longitudinal velocity of the weeding blade tip output in step S300. Calculate the dynamic lead distance for the issuance of weeding action commands. : ; ; in, The total predicted response delay time is the current estimate of the system. This variable represents the complete time span from the moment the control unit issues an electrical signal command to the moment the hydraulic actuator overcomes inertia and resistance and actually moves to the effective working position. This is the adaptive correction coefficient calculated through closed-loop feedback in the previous work cycle (its specific update algorithm will be detailed in subsequent step S600), initially set to 0. In the formula, It consists of two parts: an open-loop reference value derived from a table based on the current hydraulic physical state. And the closed-loop compensation value derived from historical error learning. The combination of these two factors constitutes the best estimate of the system's current true response capability. This is the inherent compensation distance determined by the mechanical structure. Physically, it represents the geometrical distance required for the cutting edge of the weeding blade 210 to travel from contacting the boundary of the crop protection zone to completely covering the area. It is typically taken as the blade's radius of action or the safety margin set for the protection zone. The calculated... This refers to the equivalent spatial distance required to pre-deliver action commands within the logical mapping space, prior to their physical arrival time. This is achieved through real-time adjustment. The system can offset response time fluctuations caused by changes in operating conditions (such as the difference between cold oil in the morning and hot oil at noon), ensuring that the weeding blades always cut into the soil at the optimal time.

[0056] S611, perform predictive triggering determination based on spatial mapping. In each control cycle, the control unit 300 determines the dynamic advance triggering distance output in step S513. The corresponding trigger index position is retrieved in the shift register stack. Due to the use of a circular buffer structure, the simple physical distance needs to be converted relative to the current read pointer. The logical offset address. Control unit 300 calculates the target detection index. : ; in, This is to convert the physical lead distance into an offset of discrete storage bits; This is the current weeding tool position pointer updated in step S413; the modulo operation ensures the loop validity of the index address within the circular buffer. The control unit 300 reads the index from the shift register stack... Logical state value at the location .like This indicates the distance in front of the current tool position. There are crops to be avoided at the location. At this moment, the control unit 300 immediately outputs a high-level start signal to the hydraulic solenoid valve drive circuit and records the timestamp of the command issuance. The action is then marked as an event to be verified and stored in a circular queue. This triggering mechanism based on dynamic indexing allows the trigger position to slide back and forth along the virtual space axis as the response capability of the hydraulic system changes, thus ensuring that the physical action point always coincides with the crop position.

[0057] S612, Capture motion execution feedback and physical response delay. To construct a closed-loop error signal, the system needs to accurately capture the physical moment when the hydraulic actuator completes an effective action. In this embodiment, a high-response Hall switch or magnetostrictive displacement sensor is deployed at the end of the hydraulic cylinder stroke. When the weeding blade 210 moves to a preset effective avoidance position (typically defined as 85% to 95% of the total stroke to ensure the blade has completely left the crop protection zone), the sensor output signal undergoes a level flip. The control unit 300 captures this flip moment via a hardware interrupt. This is combined with a jitter filter to eliminate mechanical vibration interference. Then, the true physical response time of this action is calculated. : ; During this process, the system initiates timeout monitoring logic: if and The time difference exceeds the preset safety threshold If the time is 300ms (e.g., 300ms), it is determined that the hydraulic valve core is stuck or the oil circuit is blocked. The system will automatically trigger the fault shutdown protection to prevent invalid calculation data from polluting the model parameters.

[0058] S613, an online adaptive update of the prediction model based on observation residuals. Addressing the time-varying nonlinear characteristics of hydraulic systems, this update aims to improve the prediction response time. Approximates real response time throughout the entire lifecycle. The system uses an exponentially weighted moving average algorithm to adjust the correction coefficient. Iterative correction is performed. This algorithm is essentially a discrete-time integral controller, capable of suppressing random noise from a single measurement while tracking the long-term drift trend of system parameters. The update rule is as follows: ; ; in, For the first Delay prediction residuals for each action; This is the learning rate coefficient, with a value ranging from 0.05 to 0.15. The determination of this coefficient is based on the dynamic stability analysis of the system. An excessively large value will cause the correction amount to be overly sensitive to random noise, resulting in parameter oscillations; If the value is too small, it will lead to a lag in tracking sudden operating conditions (such as drastic changes in oil temperature). The updated correction coefficients are constrained by the saturation function. Within the range (e.g., ±50±50ms), the feedback is directly sent to step S513 to participate in the next cycle. Calculation. Through this real-time learning mechanism, the system can automatically compensate for changes in the time constant caused by hydraulic oil aging, filter blockage, or electromagnetic coil temperature rise, achieving adaptive aging compensation without manual intervention.

[0059] Specific application example: Weeding between corn seedlings Implementation scenarios and hardware configurations: This embodiment was applied to a contiguous corn planting base in the North China Plain, where mechanical weeding was carried out between corn plants at the "V5" leaf age (plant height approximately 25-30cm).

[0060] Traction Platform 100: Selected Dongfanghong LX904 wheeled tractor, equipped with a high-precision steering angle sensor.

[0061] Weeding implement 200: A custom-designed rear-mounted inter-row weeder equipped with two sets of hydraulically driven finger-disc weeding blades 210.

[0062] Perception system: Vision sensor 310: Baslerace2Pro industrial camera, 120fps, mounted at the center of the machine beam, 80cm above the ground.

[0063] Millimeter-wave radar 320: TIAWR1843 millimeter-wave radar, installed at the bottom of the rear axle of the tractor, is used to measure ground speed.

[0064] Inertial Measurement Unit (IMU): Two XsensMTi-630s were selected, one installed under the tractor seat and the other at the center of the weeder beam.

[0065] Control unit 300: Based on the NVIDIA Jetson Xavier NX embedded computing platform, it runs the synchronization control algorithm of this invention.

[0066] Work process description: On the day of the operation, the surface soil moisture content was 22% (relatively moist), and the ambient temperature gradually increased from 18°C ​​in the morning to 28°C at noon.

[0067] Phase 1: Kinematic Decoupling. When the tractor performs S-shaped obstacle avoidance maneuvers at the edge of the field, the wheels slip due to the wet soil (slip rate of about 12%), and the implements experience severe lateral movement due to suspension sway.

[0068] Traditional methods rely solely on wheel speed sensors to calculate displacement. (See attached image.) Figure 3As shown by the dotted line (direct estimation of traditional wheel speed), its curve is relatively smooth, but the value deviates from the actual situation and fails to reflect the dynamic changes caused by the swing of the implement end. Because the system misjudged the blade speed as constant and lost the swing component, the weeding blade cut too slowly when the implement swung to the left, damaging 3 corn seedlings.

[0069] The method of this invention demonstrates that the millimeter-wave radar measures the actual ground speed as 1.2 m / s (lower than the wheel speed of 1.4 m / s). Simultaneously, the IMU detects an angular velocity oscillation of 15° / s relative to the tractor. The control unit calculates the equivalent longitudinal velocity. As attached Figure 3 As shown by the solid line (the actual speed after decoupling in this invention), the curve has a high peak and exhibits fluctuating characteristics, vividly demonstrating the physical fact that the combined speed increases when the machine's swing is superimposed on the traction speed. The system successfully captured this fluctuation and accurately calculated the pulse accordingly, dynamically adjusting the shift frequency of the shift register, successfully ensuring that the cutter avoided all seedlings.

[0070] Phase Two: Position Synchronization Accuracy In long-distance straight-line operations (200 meters in length), the focus is on examining the system's ability to control cumulative errors.

[0071] Traditional method results in approximately 5mm of cumulative error per 10 meters of travel due to unprocessed quantization residuals from the shift register. (See attached image.) Figure 4 As shown by the dotted line (error of traditional control method), the error curve rises steadily in a sloping manner, exceeding the physical boundary of the crop protection window (e.g., ±3cm) defined by the dotted line (allowable range for safe operation) in just a few seconds. When the operation reaches 100 meters, the cumulative error exceeds 5cm, which means that the timing of the weeding blade's action is seriously off, directly uprooting a large area of ​​normal crops.

[0072] The method of this invention is characterized by the introduction of a residual accumulation strategy to temporarily store micro-displacements of less than 1 mm. After a 200-meter operation, as shown in the attached figure... Figure 4 As shown by the solid line (error of the control method of this invention), the error curve fluctuates slightly around 0 throughout, and never exceeds the safe range represented by the upper and lower dotted lines. This fully demonstrates the effectiveness of the "residual accumulation" and "shift register" strategies, keeping the synchronization error within ±5mm.

[0073] Phase 3: Hydraulic Thermal Drift Adaptation As the operation progresses, the hydraulic oil temperature rises from 20°C to 75°C, the oil viscosity decreases, and the physical response characteristics of the hydraulic system change.

[0074] Traditional method: Set a fixed solenoid valve opening advance of 50ms. (See attached image) Figure 5As shown by the dashed line (traditional fixed parameter setting), this is a horizontal straight line, illustrating that the traditional method cannot adapt to environmental changes. Meanwhile, the attached... Figure 5 The solid line (actual physical response delay) shows a clear downward trend, indicating that the higher the oil temperature, the faster the hydraulic action. In the high-temperature zone (right side of the chart), the dashed line is much higher than the solid line, meaning that the system mistakenly perceives the action as slow when it is actually very fast (shortened to 25ms), causing the cutter to move prematurely and retract before reaching the seedling, leaving behind weeds that have not been cleared.

[0075] The method of this invention is characterized by: the system detecting an increase in oil temperature and using a Hall sensor to detect the actual response time. The length has been shortened. The system utilizes an "online correction coefficient" and an "EWMA algorithm" for real-time learning, and the results are shown in the attached figure. Figure 5 The black dots (online predicted values ​​of this invention) are shown in the diagram. These black dots are closely distributed around the solid line representing the actual physical values, proving that the prediction model can track real-time changes in physical delay. Based on this, the system automatically reduces the lead-ahead trigger distance. This ensured that the weed control rate remained above 95%.

[0076] Experimental data conclusions: After a comparative test covering 20 acres of land over 4 consecutive hours: Seedling damage rate: 5.8% with traditional methods, reduced to 0.6% with the method of this invention.

[0077] Weed control rate: The traditional method achieves 82%, while the method of this invention improves it to 96.5%.

[0078] System stability: Under conditions of tractor slippage and sharp turns, the present invention did not experience synchronization loss.

Claims

1. A method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements, characterized in that, Includes the following steps: The visual sensor (310) is used to identify crop seedlings and calculate the longitudinal physical distance between the crop seedlings and the weeding blade (210). The longitudinal physical distance is then mapped to the shift register stack in the control unit (300) to generate crop tags. The inertia and velocity data of the traction platform (100) and the weeding implement (200) are collected, and the equivalent longitudinal velocity vector of the weeding blade (210) relative to the ground is calculated based on the attachment geometric model. The displacement pulse increment is calculated using the equivalent longitudinal velocity vector, which drives the read / write pointer of the shift register stack to perform a shift operation, so that the logical position of the crop marker is synchronized with the physical position of the crop seedling. The hydraulic circuit pressure and hydraulic oil temperature values ​​are obtained, and the dynamic advance triggering distance of the weeding execution component is calculated by combining the preset response model and online correction coefficient. The system detects crop markers located at the dynamic advance trigger distance in the shift register stack. When a crop marker is detected, an avoidance action command is generated, and feedback data during the action process is collected to update the online correction coefficient.

2. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements as described in claim 1, characterized in that, The step of mapping the longitudinal physical distance to the shift register stack within the control unit (300) to generate crop tags specifically includes: Construct a shift register stack with a circular buffer structure and set the spatial resolution represented by each unit of storage bit; Obtain the reference read / write pointer of the weeding tool (210) in the shift register stack at the current moment; Calculate the ratio of the longitudinal physical distance to the spatial resolution and round down to obtain the distance offset; Add the reference read / write pointer to the distance offset, and take the modulo of the total capacity length of the shift register stack to obtain the target write index; The memory cell in the shift register stack corresponding to the target write index is set to an active state, and a continuous crop protection window is set in the adjacent range of the target write index.

3. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements as described in claim 1, characterized in that, In the step of collecting the inertia and velocity data of the traction platform (100) and the weeding implement (200) and calculating the equivalent longitudinal velocity vector, the data preprocessing includes: The main vibration center frequency is calculated based on the real-time engine speed and the number of cylinders. The coefficients of the notch filter are updated, and the collected angular velocity data is filtered in the frequency domain to remove mechanical vibration noise. The absolute longitudinal velocity collected by the millimeter-wave radar (320) is compared with the preset zero velocity determination threshold. When the absolute longitudinal velocity is less than the zero velocity determination threshold, the determination device enters a stationary state and forcibly sets the output angular velocity to zero. When the absolute longitudinal velocity is greater than or equal to the zero velocity determination threshold, the filtered angular velocity data is output for calculation.

4. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements as described in claim 3, characterized in that, The steps for calculating the equivalent longitudinal velocity vector of the weeding blade (210) relative to the ground based on the attached geometric model specifically include: Establish a coordinate system and use an integral algorithm to calculate the hinge angle between the traction platform (100) and the weeding implement (200); By projecting the absolute longitudinal velocity onto the longitudinal axis of the implement using the hook-and-hook angle, the traction speed component is obtained. Calculate the tangential linear velocity of the weeding implement (200) at the position of the weeding blade (210) due to the swing around the attachment point, and use it as the rotational compensation component; The traction speed component is superimposed with the rotational compensation component to obtain the equivalent longitudinal velocity vector.

5. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements as described in claim 1, characterized in that, The step of calculating the displacement pulse increment using the equivalent longitudinal velocity vector is implemented through a residual accumulation strategy, specifically including: In each control cycle, the equivalent longitudinal velocity vector is multiplied by the sampling time interval to obtain the physical micro-displacement; The total cumulative displacement is obtained by adding the physical micro-displacement to the displacement residual of the previous cycle. Calculate the ratio of the total cumulative displacement to the spatial resolution of the shift register stack and round down to obtain the displacement pulse increment for the current cycle; Subtract the physical length corresponding to the displacement pulse increment from the total cumulative displacement, and update the displacement residual to enter the next control cycle; The read pointer of the shift register stack is updated according to the displacement pulse increment to achieve cyclic shifting of the circular buffer.

6. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements according to claim 1, characterized in that, The step of calculating the dynamic advance triggering distance of the weeding execution component by combining the preset response model and online correction coefficient specifically includes: A two-dimensional response time mapping table indexed by pressure and temperature is pre-constructed; Based on the collected hydraulic circuit pressure and hydraulic oil temperature values, the basic response time is obtained from the mapping table using an interpolation algorithm; The predicted response delay time is obtained by adding the base response time to the online correction coefficient. The dynamic advance triggering distance is synthesized by multiplying the equivalent longitudinal velocity vector and the predicted response delay time, and superimposing a preset mechanical compensation distance.

7. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements as described in claim 6, characterized in that, The step of detecting the crop marker located at the dynamic lead trigger distance in the shift register stack and generating instructions includes: Calculate the ratio of the dynamic advance trigger distance to the spatial resolution and round it to obtain the logical offset address; Add the current read pointer to the logical offset address and take the modulo to calculate the target detection index; Read the logical state value corresponding to the target detection index from the shift register stack; If the logic state value is a valid flag, an avoidance action command is sent to the drive circuit, and the timestamp of the command sending time is recorded.

8. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements according to claim 1, characterized in that, In the step of updating the online correction coefficient with feedback data during the data acquisition process, the acquisition of feedback data includes: Physical motion signals are detected by displacement sensors or switch sensors installed at the end of the hydraulic actuator; The moment when the physical action signal flips is taken as the start time of the physical action; Calculate the time difference between the start time of the physical action and the time of command transmission to obtain the actual physical lag; If the time difference exceeds a preset safety threshold, a fault shutdown protection will be triggered.

9. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the coordinated operation of tractors and implements as described in claim 8, characterized in that, The step of updating the online correction coefficient includes: Calculate the observation residual between the actual physical lag and the predicted response delay time used by the current model; Using the exponentially weighted moving average algorithm, the online correction coefficient of the previous period is added to the product of the observation residual and the learning rate coefficient to obtain the updated online correction coefficient.

10. The method for precise synchronous control of weeding and seedling avoidance actions between plants under the cooperative operation of tractors and implements according to claim 9, characterized in that, The updated online correction coefficients need to undergo saturation processing: Determine whether the updated online correction coefficient exceeds the preset correction range; If the value exceeds the range, the updated online correction coefficient is clamped to the boundary value of the correction range, and the clamped value is used for prediction calculation in the next control cycle.

Images