A potato seeding and fertilizing vehicle path planning method and system

Abstract

The application relates to the technical field of intelligent agricultural equipment, and discloses a potato seeding machine and fertilizer adding vehicle path planning method and system, the method comprising the following steps: acquiring a comprehensive data set to predict a speed fluctuation trend; calculating a unit distance fertilizer consumption rate and an adjusted distance, and integrating to obtain a total consumption; judging a replenishment opportunity according to a difference value of a residual amount, and triggering path planning if the difference value is lower than a preset proportion; combining a fertilizer adding vehicle position and a seeder track to search for a potential intersection point with the minimum driving distance; evaluating downtime and fertilizer uniformity to determine an optimal intersection point; updating navigation instructions to generate a real-time route; monitoring a relative distance, activating a protocol to obtain a replenishment confirmation when entering a docking range; and feeding back an efficiency index to determine a next prediction starting point. Through path planning, the application realizes efficient and accurate collaborative operation of the seeder and the fertilizer adding vehicle in a complex field environment, and significantly improves operation efficiency and fertilizer utilization.

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural equipment technology, and in particular to a path planning method and system for a fertilizer loading vehicle on a potato planter. Background Technology

[0002] Currently, with the rapid development of manufacturing technologies for mechanized agricultural and horticultural machinery such as intelligent agricultural power machinery, agricultural equipment is evolving towards intelligence and collaboration, providing efficient and precise operating methods for large-scale potato planting. This collaborative operation method is directly related to the operational efficiency and fertilizer utilization rate throughout the entire planting process.

[0003] In existing technologies, the coordinated operation of fertilizer trucks and seeders typically relies on pre-set fixed refueling points or simple straight-line path planning. This approach assumes that the seeder travels along a predetermined trajectory at a constant speed, and a data acquisition and control system installed on the seeder monitors the remaining fertilizer level in real time. When the fertilizer level falls below a threshold, the system notifies the fertilizer truck to proceed to a pre-set fixed refueling point to wait or meet the seeder. However, in actual field operations, the seeder's speed fluctuates frequently due to various factors such as soil texture, terrain slope, and real-time load on the equipment. This statically preset approach cannot adapt to dynamic changes in a timely manner, leading to significant deviations in refueling timing and meeting points. When terrain undulations cause the seeder's actual speed to be lower than expected, the fertilizer consumption rate slows down, and the fertilizer truck may wait idly at the predetermined meeting point. Conversely, when the seeder's speed increases due to reduced load or downhill conditions, fertilizer consumption accelerates, and the fertilizer truck may run out before reaching the predetermined refueling point, forcing the operation to be interrupted and awaiting refueling.

[0004] In summary, existing technologies are insufficient to accurately establish prediction and planning mechanisms that reflect the actual needs of operations in dynamic and uncertain field environments. They are unable to achieve efficient dynamic planning of fertilizer truck routes, resulting in low operational efficiency and serious resource waste. Summary of the Invention

[0005] This invention provides a method and system for path planning of a potato planter and fertilizer truck, so as to realize efficient and precise collaborative operation of the planter and fertilizer truck in complex field environments, significantly improving operation efficiency and fertilizer utilization.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a path planning method for a potato planter fertilizer truck, comprising:

[0007] A comprehensive dataset is obtained and processed using a time series forecasting method to obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, its driving speed, the amount of remaining fertilizer, terrain slope data, and equipment load distribution data.

[0008] Based on the speed fluctuation trend, calculate the fertilizer consumption rate per unit distance within the corresponding path segment and the expected distance after path uncertainty adjustment. Integrate the fertilizer consumption rate and the expected distance to obtain the total fertilizer consumption for the subsequent path.

[0009] Based on the difference between the total fertilizer consumption and the current remaining fertilizer, a replenishment timing threshold is determined. If the replenishment timing threshold is lower than the preset remaining ratio, the fertilizer truck path planning process is triggered.

[0010] The current location and speed data of the fertilizer truck are obtained, and a subset of candidate coordinates is generated by combining the future path trajectory of the seeder. The fertilizer truck is then filtered based on the time it takes to reach each candidate coordinate and the actual distance it travels, resulting in a set of potential intersection coordinates with the minimum empty driving distance.

[0011] Calculate the expected downtime for each coordinate in the potential intersection coordinate set, predict the local fertilization uniformity value of the seeder at the start-up stage based on the expected downtime, and select the optimal intersection coordinate from the potential intersection coordinate set by combining the local fertilization uniformity value and the expected downtime.

[0012] The coordinate points are extracted from the optimal intersection coordinates, the navigation instructions for the fertilizer truck are updated, and the real-time adjusted travel route is obtained.

[0013] The relative distance between the seeder and the fertilizer truck is monitored according to the travel route. If the relative distance enters the preset docking range, the docking protocol is activated and a coordinated supply confirmation signal is obtained.

[0014] By using the feedback of the coordinated supply confirmation signal, the operation efficiency index is recorded, and the starting point of the next prediction cycle is determined.

[0015] Secondly, the present invention provides a path planning system for a potato planter fertilizer applicator, comprising:

[0016] The data acquisition and prediction module is used to acquire a comprehensive dataset, process the comprehensive dataset using time series prediction methods, and obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, driving speed, remaining fertilizer amount, terrain slope data, and equipment load distribution data.

[0017] The consumption calculation module is used to calculate the fertilizer consumption rate per unit distance within the corresponding path segment and the expected distance after path uncertainty adjustment based on the speed fluctuation trend, and to perform an integral operation on the fertilizer consumption rate and the expected distance to obtain the total fertilizer consumption of the subsequent path.

[0018] The replenishment judgment and triggering module is used to determine the replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount. If the replenishment timing threshold is lower than the preset remaining ratio, the fertilizer truck path planning process is triggered.

[0019] The intersection point search module is used to obtain the current position and speed data of the fertilizer truck, combine it with the future path trajectory of the seeder to generate a subset of candidate coordinates, and filter them based on the time when the fertilizer truck arrives at each candidate coordinate and the actual driving distance to obtain the set of potential intersection coordinates with the minimum empty driving distance.

[0020] The intersection point evaluation module is used to calculate the expected downtime corresponding to each coordinate in the potential intersection coordinate set, predict the local fertilization uniformity value of the seeder in the start-up stage based on the expected downtime, and select the optimal intersection coordinate from the potential intersection coordinate set by combining the local fertilization uniformity value and the expected downtime.

[0021] The navigation update module is used to extract coordinate points from the optimal intersection coordinates, update the navigation instructions of the fertilizer truck, and obtain the real-time adjusted travel route.

[0022] The docking monitoring module is used to monitor the relative distance between the seeder and the fertilizer truck according to the travel route. If the relative distance enters the preset docking range, the docking protocol is activated and a coordinated supply confirmation signal is obtained.

[0023] The feedback and loop module is used to record the operation efficiency index and determine the starting point of the next prediction loop through the feedback of the coordinated replenishment confirmation signal.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] (1) This invention acquires a comprehensive dataset containing multi-source sensor data, extracts the speed fluctuation trend of future paths using time series prediction, and then integrates the fertilizer consumption rate spatially with the path uncertainty adjustment coefficient, thereby achieving dynamic and accurate prediction of fertilizer demand for future paths. This mechanism breaks the limitations of the traditional fixed replenishment mode and effectively solves the problems of uneven fertilization and misjudgment of replenishment timing caused by undulating terrain and slippage of machinery.

[0026] (2) Based on the accurately calculated coordinates of the work interruption points, this invention uses a path optimization algorithm to dynamically search for a set of potential intersection coordinates around the future trajectory of the seeder, and further introduces a model for predicting downtime and the uniformity of local fertilization at the start. Through rigorous multi-dimensional evaluation and elimination of substandard coordinates, the optimal intersection coordinates with the minimum empty driving distance and the least impact on fertilization quality are finally locked, which significantly reduces the ineffective driving mileage of the fertilizer truck and greatly shortens the downtime waiting time of the seeder.

[0027] (3) The present invention innovatively designs a dynamic docking monitoring mechanism based on high frequency differential positioning. When the relative Euclidean distance and heading deviation between the seeder and the fertilizer truck meet the stringent threshold, the docking protocol is automatically activated and microsecond-level clock synchronization and lateral fine adjustment are performed, ensuring the efficiency, stability and safety of the physical docking of the two machines in complex and bumpy farmland environments. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating an embodiment of a path planning method for a fertilizer loading vehicle of a potato planter provided by the present invention;

[0029] Figure 2 This is a schematic diagram of an embodiment of a path planning system for a potato planter fertilizer truck provided by the present invention. Detailed Implementation

[0030] 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.

[0031] Reference Figure 1 The first embodiment of the present invention provides a path planning method for a fertilizer loading vehicle of a potato planter, comprising the following steps:

[0032] Step S11: Obtain a comprehensive dataset and process the comprehensive dataset using time series forecasting to obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, driving speed, remaining fertilizer amount, terrain slope data, and equipment load distribution data.

[0033] Step S12: Based on the speed fluctuation trend, calculate the fertilizer consumption rate per unit distance within the corresponding path segment and the expected distance after path uncertainty adjustment, and perform an integral operation on the fertilizer consumption rate and the expected distance to obtain the total fertilizer consumption for the subsequent path.

[0034] Step S13: Determine the replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount. If the replenishment timing threshold is lower than the preset remaining ratio, the fertilizer truck path planning process is triggered.

[0035] Step S14: Obtain the current position and speed data of the fertilizer truck, generate a subset of candidate coordinates by combining the future path trajectory of the seeder, and filter them based on the time and actual driving distance of the fertilizer truck to reach each candidate coordinate to obtain a set of potential intersection coordinates with the minimum empty driving distance.

[0036] Step S15: Calculate the expected downtime for each coordinate in the potential intersection coordinate set, predict the local fertilization uniformity value of the seeder at the start-up stage based on the expected downtime, and select the optimal intersection coordinate from the potential intersection coordinate set by combining the local fertilization uniformity value and the expected downtime.

[0037] Step S16: Extract coordinate points from the optimal intersection coordinates, update the fattening vehicle navigation instructions, and obtain the real-time adjusted travel route;

[0038] Step S17: Monitor the relative distance between the seeder and the fertilizer truck according to the travel route. If the relative distance enters the preset docking range, activate the docking protocol and obtain a coordinated supply confirmation signal.

[0039] Step S18: Based on the feedback of the coordinated supply confirmation signal, record the operation efficiency index and determine the starting point of the next prediction cycle.

[0040] In step S11, the comprehensive dataset is acquired and processed using a time series forecasting method to obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, its driving speed, the amount of remaining fertilizer, terrain slope data, and equipment load distribution data, including:

[0041] Obtain an initial multidimensional dataset, which includes real-time location coordinates, current velocity vector, remaining fertilizer volume, slope data, and machinery weight distribution.

[0042] The terrain undulation gradient is calculated based on the location coordinates and slope data in the initial multidimensional dataset, and the comprehensive load adjustment coefficient is determined in combination with the equipment weight distribution.

[0043] If the comprehensive load adjustment coefficient exceeds the preset load threshold, the speed vector and slope sequence in the comprehensive dataset are smoothed using a moving average filtering algorithm to obtain a filtered comprehensive dataset. The filtered comprehensive dataset is then processed using a long short-term memory network to obtain a speed prediction sequence.

[0044] Based on the velocity prediction sequence, the vector differences between adjacent nodes are analyzed, and the frequency components are extracted by Fourier transform to obtain the velocity fluctuation trend on the future path.

[0045] First, an initial multidimensional dataset is acquired, which includes real-time position coordinates, current velocity vector, remaining fertilizer volume, slope data, and implement weight distribution. Specifically, the real-time position coordinates and current velocity vector are acquired through a real-time dynamic differential GPS receiver mounted on top of the seeder, providing centimeter-level three-dimensional coordinates and high-precision velocity vector information. The remaining fertilizer volume is monitored and calculated in real-time by an ultrasonic level gauge or weighing sensor installed inside the fertilizer tank. Slope data is acquired through a high-precision inertial measurement unit integrated on the seeder chassis, which can output the vehicle's pitch and roll angles in real time. The implement weight distribution is calculated in real-time by combining the seeder's unloaded weight, the current remaining fertilizer weight, and the real-time weight of the potato seed tubers in the seed metering device, using a multi-axis force sensor network on the chassis suspension system to determine the center of gravity position and the load ratio of each wheel. These heterogeneous sensor data are aggregated via an onboard industrial Ethernet network, and after timestamp synchronization and alignment, a multidimensional initial dataset is constructed.

[0046] Subsequently, the terrain undulation gradient is calculated based on the position coordinates and slope data in the initial multidimensional dataset, and the comprehensive load adjustment coefficient is determined in conjunction with the implement weight distribution. Specifically, the terrain undulation gradient is calculated by continuously recording the slope angle change every five meters during the seeder's movement. The absolute value of the slope difference between adjacent recording points is calculated and divided by this distance interval to obtain the gradient value reflecting the degree of drastic terrain change. The comprehensive load adjustment coefficient is determined based on the offset of the current center of gravity position relative to the geometric center and the total mass of the vehicle. The system presets a standard load state (i.e., the ideal weight distribution when fully loaded and on flat ground). By acquiring the actual load data of the front and rear axles in real time and calculating their ratio to the standard axle load, the system dynamically calculates the impact of the terrain undulation gradient (slope angle) on traction demand. Specifically, the comprehensive load adjustment coefficient is calculated using a nonlinear function with terrain and load bias weights. .in, This is the offset of the current center of gravity relative to the geometric center. This refers to the vehicle's wheelbase. The terrain slope angle is measured by an inertial measurement unit (IMU). and Using a pre-calibrated empirical constant for traction sensitivity, the actual traction resistance of the seeder is measured in real time by a tension sensor under typical operating scenarios with different slopes and load combinations, and the center of gravity offset is recorded simultaneously. and slope angle Using multiple linear regression analysis, the measured increment of traction force was taken as the dependent variable, and... and The constant reflecting the load sensitivity is calculated by fitting the data as the independent variable. and a constant reflecting slope sensitivity For example, when the remaining volume of the fertilizer tank is 30% of its full capacity and the vehicle is on a 15-degree uphill slope, the shift in the center of gravity to the rear causes a decrease in front wheel traction, while the uphill slope increases the component of gravity along the slope. (If positive), the relative offset and ramp resistance calculated by the system increase significantly, thereby dynamically increasing the comprehensive load adjustment coefficient from 1.0 in the standard flat condition to over 1.25.

[0047] Subsequently, if the comprehensive load adjustment coefficient exceeds a preset load threshold, a filtered comprehensive dataset is generated, and a long short-term memory network is used to process the comprehensive dataset to obtain a speed prediction sequence. Specifically, a preset load threshold is set, which is obtained by constructing an orthogonal field test database covering multiple dimensions of variables such as different soil types (e.g., sandy loam, clay), different soil moisture contents (e.g., 10%-30%), and different operating speeds. The system extracts dynamic axle load and sinking data when the traction force undergoes a nonlinear abrupt change (i.e., the critical point of excessive wheel slippage and vehicle sinking) during the experiment, and uses a polynomial fitting algorithm to divide the boundaries of these feature data, thereby generating a three-dimensional threshold lookup table. In actual operation, the system interpolates and calls this lookup table in real time based on the soil parameters and vehicle speed identified by the current sensors to obtain the preset load threshold under the current working conditions. When the comprehensive load adjustment coefficient exceeds this threshold, it indicates that the seeder is currently under heavy load and complex terrain conditions, and its speed will be significantly affected, requiring filtering of the initial multidimensional data. Specifically, a moving average filtering method is used to remove high-frequency noise from the sensors to obtain a filtered comprehensive dataset.

[0048] Subsequently, a Long Short-Term Memory (LSTM) network is used to process the comprehensive dataset. The sequence data of the seeder's position, speed, slope, and load coefficient from the comprehensive dataset are used as input to the LTM network. The pre-training process of the LTM network involves collecting historical operation data, dividing the data into an input sequence and actual speed values ​​for a future time period as labels, and optimizing the network weights using a backpropagation algorithm to enable it to predict future speeds based on historical states. The network structure typically includes an input layer, one or more hidden layers in the LTM network, and a fully connected output layer. The number of neurons in the output layer corresponds to the predicted time step. Through forward computation of the LTM network, a speed prediction sequence for a future time period (e.g., the next 30 seconds) is obtained, i.e., a list of speed values ​​changing over time.

[0049] Finally, based on the velocity prediction sequence, the vector differences between adjacent nodes are analyzed, and frequency components are extracted using Fourier transform to obtain the velocity fluctuation trend along the future path. Specifically, based on the velocity prediction sequence, the velocity vector differences between adjacent time nodes are calculated to form a velocity change rate sequence. A Fourier transform is performed on this sequence to convert it from a time-domain signal to a frequency-domain signal, extracting the main frequency components. Analyzing these frequency components identifies the periodic characteristics of velocity fluctuations. For example, if the energy proportion of a certain frequency component is significantly higher than other frequencies, it indicates that there is a periodic fluctuation in velocity dominated by that frequency. This is usually related to the periodic undulations of the terrain (such as wavy ground) or the mechanical vibration of the seeder itself. Combining these frequency-domain characteristics with the time-domain velocity prediction sequence ultimately forms a comprehensive description of the velocity fluctuation trend along the future path. This trend not only includes the specific numerical prediction of velocity but also the severity and periodicity of velocity changes.

[0050] In step S12, the calculation of the fertilizer consumption rate per unit distance within the corresponding path segment and the estimated distance after path uncertainty adjustment based on the speed fluctuation trend, and the integration of the fertilizer consumption rate and the estimated distance to obtain the total fertilizer consumption for the subsequent path, includes:

[0051] Based on the speed fluctuation trend, a pre-acquired flow calibration curve is retrieved, and combined with the pre-acquired fertilizer particle flowability characteristic value, a dynamic speed control sequence for the fertilizer dispenser is generated.

[0052] The dynamic speed control sequence is input into a pre-established discrete flow calculation model to obtain an instantaneous mass flow sequence. Combined with the speed fluctuation trend, the fertilizer consumption rate curve per unit distance is calculated.

[0053] Calculate the ground slip rate, generate path uncertainty adjustment coefficients using the ground slip rate and terrain gradient data in the comprehensive dataset, and obtain the predicted distance after path uncertainty adjustment by weighting the planned path segments according to the path uncertainty adjustment coefficients;

[0054] The fertilizer consumption rate curve is integrated over the predicted distance to obtain the total fertilizer consumption for the subsequent path.

[0055] First, based on the speed fluctuation trend, a pre-acquired flow calibration curve is retrieved, and combined with the pre-acquired fertilizer particle flowability characteristic value, a dynamic speed control sequence for the fertilizer discharger is generated. Specifically, the flow calibration curve is a nonlinear mapping relationship obtained by fitting multiple sets of tests on a laboratory bench with different discharge motor speeds and actual discharge mass for different types and ratios of compound fertilizers. This nonlinear mapping relationship is usually fitted with a cubic polynomial using the least squares method to fully compensate for the nonlinear error caused by the motor torque dead zone at low speeds and the saturation cutoff effect of the discharge trough at high speeds.

[0056] The fertilizer granule flowability characteristic value is a parameter that comprehensively reflects the physical properties of the fertilizer, calculated using an empirical formula. The empirical formula is obtained by pre-selecting hundreds of batches of potato-specific compound fertilizers of different varieties and storage conditions in a laboratory environment, measuring their actual average particle size, angle of repose, and actual moisture content calibrated by the drying method, and simultaneously testing their actual discharge deviation at standard rotation speed on a fertilizer discharge platform. Through multiple linear regression analysis, the contribution ratio of these three physical parameters to the degree of fertilizer flowability obstruction is determined, thus solidifying them into the weighting coefficients of the empirical formula. In this embodiment, moisture content has the most significant impact on fertilizer agglomeration and decreased flowability; therefore, its corresponding weighting coefficient is set to the highest, specifically 0.5. The angle of repose directly reflects the internal friction characteristics of the granule group under gravity, and its corresponding weighting coefficient is set to 0.3. The average particle size mainly affects the physical filling rate of the fertilizer discharge channel wheel, and its corresponding weighting coefficient is set to 0.2.

[0057] In actual operation, the system multiplies the normalized values ​​of these three physical parameters collected in real time with their corresponding weighting coefficients, and then sums the three products to obtain the current particle flowability characteristic value of the batch of fertilizer. Subsequently, it calculates the theoretical discharge required to maintain uniform fertilization. This calculation process uses basic arithmetic derivation. First, it obtains the target fertilizer application amount per hectare pre-issued by the agronomic system. Considering that one hectare equals ten thousand square meters, the system divides this total amount by ten thousand, thus accurately converting the macroscopic fertilizer application amount per hectare into the theoretical target fertilizer application mass per square meter. Next, it reads the actual operating width of the seeder, that is, the total lateral width covered by all fertilizer discharge pipes. Multiplying the calculated theoretical target fertilizer application mass per square meter by the actual operating width yields the total fertilizer mass that the seeder must discharge for each step forward (i.e., traveling one meter). This value is defined as the target fertilizer application amount per unit distance.

[0058] Finally, the system extracts the speed fluctuation trend value for each future second from the time series forecasting method, i.e., the current instantaneous forward speed. The target fertilizer application rate per unit distance is directly multiplied by this instantaneous forward speed to obtain the theoretical discharge rate required to maintain uniform fertilization. Then, the flow rate calibration curve under laboratory standard conditions is corrected in real time using the previously calculated fertilizer particle flowability characteristic value. By looking up the corrected calibration curve, the target rotational speed of the fertilizer discharge motor in the corresponding future time series is calculated, thereby generating a dynamic speed control sequence for the fertilizer discharger.

[0059] Subsequently, the dynamic speed control sequence is input into a pre-established discrete flow calculation model to obtain an instantaneous mass flow sequence. Combined with the speed fluctuation trend, the fertilizer consumption rate curve per unit distance is calculated. It should be noted that the discrete flow calculation model is a mathematical derivation model based on the physical geometry and volumetric efficiency of the fertilizer discharge channel wheel. This model calculates the effective volume of a single-ring channel wheel, multiplies it by the bulk density and volumetric filling coefficient of the fertilizer, and then multiplies it by the instantaneous speed in the control sequence to obtain the instantaneous mass flow rate falling into the fertilizer delivery pipeline per second, forming the instantaneous mass flow sequence. The volumetric filling coefficient refers to the ratio of the actual volume occupied by fertilizer particles during the actual rotation and discharge process of the fertilizer discharge channel (or between the teeth of the fertilizer discharge wheel) to the theoretical maximum volume of the fertilizer discharge channel. Because fertilizer particles have pores and an angle of repose, this coefficient is usually less than 1. A photoelectric distance sensor installed at the fertilizer inlet of the fertilizer dispenser monitors the height of the upper fertilizer layer (i.e., lateral pressure) in real time. Combined with pre-stored physical bulk density and angle of repose parameters of this type of fertilizer, the volumetric filling coefficient at the current rotational speed and material level is obtained using a lookup table. Dividing the flow rate value at each time point in the instantaneous mass flow rate sequence by the expected vehicle speed in the corresponding time-point speed fluctuation trend converts the time-dimensional flow rate into a spatial-dimensional consumption rate, thus obtaining a unit-distance fertilizer consumption rate curve characterizing how many kilograms of fertilizer are consumed per meter traveled.

[0060] Subsequently, the ground slip ratio is calculated. A path uncertainty adjustment coefficient is generated using the ground slip ratio and the terrain gradient data in the comprehensive dataset. The planned path segments are then weighted according to this adjustment coefficient to obtain the estimated distance after path uncertainty adjustment. It should be noted that due to factors such as uneven field surfaces, variations in soil texture leading to slip ratio changes, and terrain slope, the actual travel path length of the seeder may deviate from the geometric length of the planned path, necessitating the introduction of a path uncertainty adjustment coefficient. This coefficient integrates the ground slip ratio and terrain gradient data. The ground slip ratio is calculated by comparing the drive wheel speed with the ground speed measured via differential GPS. Specifically, the drive wheel speed is collected in real-time by a speed sensor mounted on the drive wheel half-shaft. The theoretical forward speed of the wheel is calculated by combining this with the static rolling radius of the tires. Simultaneously, the actual ground speed of the seeder is calculated using positioning data output from an RTK-GPS receiver. After normalizing both to the same time reference, the ground slip ratio is obtained by subtracting the actual ground speed from the theoretical forward speed and then dividing by the theoretical forward speed. The terrain gradient data is used to estimate the additional travel distance caused by inclines and declines. System-defined path uncertainty adjustment factor .in, For ground slip ratio, This is the terrain gradient angle. When slippage occurs ( When the vehicle's actual forward distance is less than the theoretical wheel rotation distance, the fertilizer discharge per unit area is lower, the denominator decreases, and the path uncertainty adjustment coefficient is greater than 1; when operating on uphill or downhill slopes ( When the actual length of the diagonal path that the vehicle travels close to the ground is greater than the horizontal projection length planned on the two-dimensional map, A value less than 1 also makes the path uncertainty adjustment coefficient greater than 1. Therefore, the path uncertainty adjustment coefficient is a coefficient greater than 1. By weighting the planned path segments, the expected distance after path uncertainty adjustment is obtained, which is closer to the actual distance the seeder will travel.

[0061] Finally, the fertilizer consumption rate curve is integrated over the predicted distance to obtain the total fertilizer consumption for the subsequent path. Specifically, the fertilizer consumption rate curve per unit distance is integrated over the predicted distance after path uncertainty adjustment, and the consumption corresponding to each small distance on the curve is summed up to accurately obtain the total amount of fertilizer that the seeder will theoretically consume after completing the predicted path segment.

[0062] In step S13, the step of determining a replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount, and triggering a fertilizer truck path planning process if the replenishment timing threshold is lower than a preset remaining ratio, includes:

[0063] The current remaining amount of fertilizer is collected by a gravity sensor. The total amount of fertilizer consumed is subtracted from the current remaining amount of fertilizer to obtain the difference. If the difference is positive, the percentage of the current remaining amount of fertilizer to the total full load is calculated, and the percentage is determined as the replenishment timing threshold.

[0064] If the replenishment timing threshold is lower than the preset remaining ratio, the coordinates of the expected operation interruption point where the fertilizer is exhausted are deduced based on the fertilizer consumption rate curve.

[0065] Obtain the real-time location information of the fertilizer truck and associate it with the coordinates of the operation interruption point to generate a collaborative scheduling request;

[0066] In response to the collaborative scheduling request, the fertilizer truck path planning engine is activated, and the optimal driving path from the real-time location of the fertilizer truck to the coordinates of the operation interruption point is calculated based on road network data, thereby triggering the fertilizer truck path planning process.

[0067] First, the current remaining fertilizer amount is collected using a gravity sensor. The total fertilizer consumption is subtracted from the current remaining fertilizer amount to obtain the difference. If the difference is positive, the percentage of the current remaining fertilizer amount to the total load is calculated, and this percentage is determined as the replenishment timing threshold. Specifically, a strain gauge gravity sensor installed at the bottom of the seeder's feed hopper collects micro-voltage signals in real time. After analog-to-digital conversion and temperature compensation, these signals are mapped to the current actual weight of fertilizer (i.e., the current remaining fertilizer amount). The previously calculated total fertilizer consumption required to complete subsequent operations is used as a subtraction factor and subtracted from the current remaining fertilizer amount. If the result is negative, it indicates that the current fertilizer is insufficient to support the completion of the planned path; if it is positive, the percentage of the current remaining fertilizer amount to the total load is calculated, and this percentage is defined as the replenishment timing threshold.

[0068] Subsequently, if the replenishment timing threshold is lower than the preset remaining ratio, the coordinates of the expected work interruption point where fertilizer is exhausted are calculated based on the fertilizer consumption rate curve. Specifically, the preset remaining ratio is a safety baseline set to prevent the fertilizer applicator from sucking up empty cells, leading to missed application. It is usually determined by calculating the dead zone volume at the bottom of the feed hopper and adding a 10% redundancy (e.g., set to 20% of full load). When the calculated replenishment timing threshold is lower than this percentage, the system initiates the reverse calculation mechanism. Starting from the current position, the system performs a positive spatial accumulation and integration of the aforementioned fertilizer consumption rate curve per unit distance along the planned work path. When the accumulated consumption is exactly equal to the current remaining fertilizer, the corresponding mileage on the path is recorded. Combining the geometric equation of the planned path, the absolute latitude, longitude, and elevation of this mileage point in the field global coordinate system are calculated. This point is the coordinate of the work interruption point where the seeder will be forced to stop without external replenishment.

[0069] Next, the real-time location information of the fertilizer delivery vehicle is obtained and associated with the coordinates of the work interruption point to generate a collaborative scheduling request. Specifically, the system sends a location polling command to the fertilizer delivery vehicle in standby or cruising state via an IoT communication module (such as a 4G / 5G private network or LoRa radio) to obtain its RTK-GPS real-time coordinates and current driving direction. The real-time location and speed status of the fertilizer delivery vehicle, along with the calculated coordinates of the seeder's work interruption point, are packaged into a structured data message. This message contains information such as task priority, required fertilizer type, required quantity, and estimated deadline, thereby generating a standard collaborative scheduling request.

[0070] Finally, in response to the collaborative scheduling request, the fertilizer truck's path planning engine is activated. Based on road network data, the optimal driving path from the fertilizer truck's real-time location to the coordinates of the work interruption point is calculated, triggering the fertilizer truck's path planning process. Specifically, after receiving the scheduling request, the onboard edge computing terminal or the cloud-based scheduling center immediately wakes up the fertilizer truck's path planning algorithm engine. The engine retrieves a pre-stored or UAV-updated high-precision road network topology map of the field. This map exists in the form of a graph structure, where nodes represent intersections, edges represent roads, and the edge weights are calculated comprehensively based on road length, road surface material (mudiness), and passage width. Using Dijkstra's algorithm, starting from the fertilizer truck's current location and ending at the work interruption point, a sequence of nodes with the minimum cumulative weight (representing travel time or energy consumption) is searched in the graph structure; this is the optimal driving path, thus formally triggering and entering the fertilizer truck's active scheduling and optimization planning process.

[0071] In step S14, the current position and speed data of the fertilizer truck are acquired, and a subset of candidate coordinates is generated by combining it with the future path trajectory of the seeder. Based on the time the fertilizer truck arrives at each candidate coordinate and the actual distance traveled, a set of potential intersection coordinates with the minimum empty travel distance is obtained, including:

[0072] The current location and speed of the fertilizer truck and the future path data of the seeder are obtained. A dynamic geofence is set with the future trajectory line of the seeder as the center. Field road nodes that fall within the dynamic geofence are used as the preliminary range of potential intersection coordinates.

[0073] Within the initial range of the potential intersection coordinates, spatial topology analysis is performed in conjunction with farmland operation constraints to eliminate nodes that do not meet the accessibility and operation safety constraints, and retain coordinates that meet the constraints to form a subset of candidate coordinates.

[0074] The speed of the fertilizer truck is obtained, and the candidate coordinate subset is adjusted by integrating the fertilizer truck speed to determine the optimized coordinate set;

[0075] For each coordinate point in the optimized coordinate set, a path planning algorithm is called to calculate the actual driving distance from the current position of the fattening vehicle to that coordinate point along the passable road network. Based on the actual driving distance, all coordinate points are sorted, and several coordinate points with the smallest distance are extracted to form a potential intersection coordinate set with the minimum empty driving distance.

[0076] First, the system acquires the current location and speed of the fertilizer truck and the future path data of the seeder. A dynamic geofence is then established centered on the seeder's future trajectory line. Field road nodes falling within this geofence are considered as the initial range of potential meeting points. Specifically, the system continuously reads the fertilizer truck's dynamic parameters (current coordinates and speed value output from the speedometer) and extracts the coordinates of a series of discrete trajectory points planned for its future journey and the expected timestamps of each point from the seeder's navigation controller. A dynamic geofence is then established centered on the seeder's future trajectory line. The radius of this geofence is determined by multiplying the fertilizer truck's current speed by a preset maneuvering time tolerance window (e.g., fifteen minutes). Only field road nodes falling within this bounding box or geofence are considered to have a physical probability of meeting, thus defining the initial geographical range of potential meeting points.

[0077] Subsequently, within the initial range of the potential intersection coordinates, spatial topology analysis is performed in conjunction with farmland operation constraints. Nodes that do not meet the accessibility and operational safety constraints are eliminated, and the coordinates that meet the constraints are retained to form a subset of candidate coordinates. Specifically, within the road network area determined in the initial range, farmland operation constraints are further superimposed. Spatial topology analysis is performed using a Geographic Information System (GIS) to exclude road sections that are too narrow to accommodate two vehicles parking side by side, slope points with a gradient exceeding the parking brake safety limit of fertilizer trucks, and farmland points located in unsown areas that are prone to soil compaction. After accessibility and operational safety screening, the remaining road intersections, wide U-turn areas at the edge of fields, or widened sections of dedicated farm roads constitute a highly feasible subset of candidate coordinates.

[0078] Subsequently, the speed of the fertilizer delivery vehicle is acquired, and the candidate coordinate subset is adjusted based on this speed to determine the optimized coordinate set. Specifically, for each coordinate point in the candidate subset, the arrival time of the seeder (based on the integral of speed fluctuation trend) and the arrival time of the fertilizer delivery vehicle (based on the current speed of the fertilizer delivery vehicle and the road network distance) are calculated. If the estimated arrival time of the fertilizer delivery vehicle at full speed is still later than the estimated arrival time of the seeder by more than a maximum tolerable waiting threshold (e.g., five minutes), it indicates that the intersection point will cause an unacceptable long downtime for the seeder, and therefore the coordinate point is removed from the candidate subset. Through this dynamic elimination mechanism based on strict spatiotemporal synchronization, the remaining coordinates that are perfectly matched in time logic are retained to form the optimized coordinate set.

[0079] Finally, the distance from the current position of the heavy-duty vehicle to the optimized coordinate set is calculated, and a potential set of intersection coordinates with the minimum empty driving distance is searched based on this distance. Specifically, for each point in the optimized set, the path planning algorithm is invoked to calculate not only the straight-line distance (Euclidean distance), but more importantly, the Manhattan distance or the actual driving trajectory length along the actual passable road network. The empty driving distance is defined here as the extra invalid driving mileage incurred by the heavy-duty vehicle when it deviates from its optimal cruising route to reach the intersection point. The system sorts the actual driving distances of all points and extracts the top 3 points in terms of distance. These points ensure a perfect encounter between the two vehicles in time while maximizing fuel and driving time savings for the heavy-duty vehicle, thus forming the final set of potential intersection coordinates with the minimum empty driving distance used for detailed evaluation.

[0080] In step S15, the calculation of the expected downtime corresponding to each coordinate in the potential intersection coordinate set, the prediction of the local fertilization uniformity value at the start-up stage of the seeder based on the expected downtime, and the selection of the optimal intersection coordinate from the potential intersection coordinate set by combining the local fertilization uniformity value and the expected downtime, includes:

[0081] Calculate the estimated downtime for each coordinate in the potential intersection coordinate set. The estimated downtime is obtained by adding the fertilizer replenishment time to the time difference between the estimated arrival time of the seeder and the estimated arrival time of the fertilizer truck.

[0082] The instantaneous flow deviation value during the startup phase is determined based on the estimated downtime and the preset path curvature of the seeder after it starts at each coordinate point. The instantaneous flow deviation value is obtained based on a preset mapping table of downtime and flow recovery delay.

[0083] The instantaneous flow rate deviation value is input into a pre-trained particle distribution density prediction model to obtain the local fertilization uniformity value;

[0084] The coordinates with local fertilization uniformity values ​​lower than the preset uniformity threshold are removed, and the coordinate with the minimum expected downtime is selected from the remaining coordinates as the optimal intersection coordinate.

[0085] First, the estimated downtime for each coordinate in the potential intersection coordinate set is calculated. This estimated downtime is obtained by adding the fertilizer resupply time to the estimated arrival time of the seeder and the fertilizer truck. Specifically, for candidate points in the set, the system extracts the expected arrival timestamps from the seeder's navigation controller and the fertilizer truck's path planner. The difference between these two timestamps yields the waiting time difference. If the fertilizer truck arrives first, the seeder does not need to wait for it, and the difference is zero; if the seeder arrives first, the difference is positive. A constant term is added to this waiting time difference. This constant term represents the average standard operating time (e.g., fixed at fifteen minutes) for the robotic arm docking, valve opening, material gravity descent, and interface separation. The sum of these two terms is the overall estimated downtime the seeder must endure if it chooses to resupply at that point.

[0086] Subsequently, based on the estimated downtime and the preset path curvature of the seeder after starting at each coordinate point, the instantaneous flow deviation value of the startup phase is determined. This instantaneous flow deviation value is obtained based on a preset mapping table of downtime and flow recovery delay. Specifically, the system reads the geometric curvature of the next path the seeder will enter after completing refueling at the candidate coordinate point. Simultaneously, it retrieves a mapping table established in advance through numerous field bench tests. This mapping table records the time delay required for the hydraulic motor or stepper motor of the fertilizer discharge system to restart, reach the target speed, and achieve stable discharge under different downtimes due to physical factors such as cooling, pressure unloading, and compaction and settling of fertilizer particles in the pipeline. The longer the downtime, the more pronounced the system response lag, resulting in an actual fertilizer discharge flow rate significantly lower than the theoretical commanded flow rate at the moment of restart. The system calculates a negative percentage value, i.e., the instantaneous flow deviation value of the startup phase, by looking up the table and using two-dimensional interpolation, combined with the path curvature at the starting point.

[0087] Subsequently, the instantaneous flow deviation value is input into a pre-trained particle distribution density prediction model to obtain a local fertilization uniformity value. Specifically, the local fertilization uniformity value is used to characterize the uniformity of fertilizer distribution within the working area after the seeder restarts its operation. It should be noted that the establishment and pre-training process of the particle distribution density prediction model involves collecting sample data including various instantaneous flow deviation values, seeder starting acceleration, wind speed, and soil roughness as input features, as well as the actual fertilization uniformity coefficient of variation (CV value) collected by manually weighing fertilizer collection trays laid in the field, serving as the label. The model structure employs a multilayer perceptron (MLP) or a one-dimensional convolutional neural network (1D-CNN), continuously adjusting neuron weights through backpropagation to minimize the mean square error between the predicted uniformity and the actual CV value. After pre-training, the particle distribution density prediction model can accurately capture the complex nonlinear relationship between flow fluctuations and the final ground fertilizer distribution. Input the calculated instantaneous flow deviation value and the corresponding environmental parameters into the pre-trained model, and the model will output a percentage prediction value, representing the local fertilization uniformity value of the initial stage after the refueling at the junction.

[0088] Finally, coordinates with local fertilization uniformity values ​​below a preset uniformity threshold are removed, and the coordinate with the shortest expected downtime is selected from the remaining coordinates as the optimal intersection coordinate. Specifically, the preset uniformity threshold is a baseline standard set based on agronomic requirements and the sensitivity of specific potato varieties to nutrient distribution, typically set at 85%. The system iterates through all candidate intersection coordinates and evaluates the uniformity values ​​output by the model. If a coordinate experiences a large deviation in initial flow rate due to excessive downtime, resulting in the model predicting a fertilization uniformity of only 70%, below the threshold, the coordinate is directly removed to prevent crop yield reduction during the initial stage of replenishment at that point. In the set of coordinates that survive after uniformity constraint filtering, the system selects the coordinate with the shortest expected downtime using a simple numerical comparison algorithm. This point ensures both agronomical fertilization quality and minimizes machine downtime, and is ultimately confirmed by the system as the optimal intersection coordinate.

[0089] In step S16, extracting coordinate points from the optimal intersection coordinates and updating the fattening vehicle navigation instructions to obtain the real-time adjusted travel route includes:

[0090] The optimal intersection coordinates are analyzed to obtain the navigation endpoint node, and the global optimal path is generated by combining the real-time positioning coordinates of the fertilizer truck.

[0091] The globally optimal path is smoothed based on the pre-obtained vehicle turning radius to generate a feasible driving trajectory;

[0092] Obtain real-time obstacle information, replan the feasible driving trajectory based on the real-time obstacle information to obtain a collision-free geometric path, and generate a fattening vehicle navigation command sequence based on the collision-free geometric path;

[0093] The sequence of navigation instructions for the fattening vehicle is issued to drive the vehicle and form a real-time adjusted route for the fattening vehicle to travel towards the optimal intersection coordinates.

[0094] First, the optimal intersection coordinates are analyzed to obtain the navigation endpoint node. Combined with the real-time positioning coordinates of the fertilizer truck, a globally optimal path is planned and generated. Specifically, the system performs protocol parsing on the optimal intersection coordinate data packet determined in the previous step, separating the high-precision WGS84 latitude and longitude coordinates and the corresponding logical field plot numbers, which serve as the absolute endpoint node for the path planning engine. Simultaneously, the current high-frequency RTK-GPS positioning coordinates of the fertilizer truck are read as the starting point. Based on the existing road network topology, a global path search algorithm, such as the A* algorithm, is run again, considering the latest road condition and traffic information (such as updating edge weights based on the occupancy status of other agricultural machinery), to generate a skeleton path connected by a series of discrete intersection nodes, i.e., the globally optimal path.

[0095] Subsequently, the globally optimal path is smoothed based on the pre-acquired vehicle turning radius to generate a feasible driving trajectory. Specifically, the globally optimal path is usually composed of multiple straight segments, forming broken lines at the nodes, which does not comply with the non-holonomic kinematic constraints of the heavy-duty vehicle chassis (i.e., the vehicle cannot turn around on the spot or make a turn with infinite curvature). The system extracts the wheelbase and maximum front wheel steering angle of the heavy-duty vehicle to calculate its theoretical minimum turning radius. At each inflection point of the global path, the Dubins Curve algorithm is used to generate a circular arc or transition curve with continuous curvature and greater than the minimum turning radius to replace the abrupt angle. The processed path is not only positionally continuous but also has first-order continuous heading angles, thus transforming the theoretical path into a feasible driving trajectory that the vehicle controller can smoothly follow.

[0096] Next, real-time obstacle information is acquired, and the feasible driving trajectory is replanned based on this information to obtain a collision-free geometric path. A navigation command sequence for the truck is then generated based on this collision-free geometric path. Specifically, while the truck travels along a smooth trajectory, its onboard LiDAR or binocular vision sensors scan the environment ahead in real time, detecting sudden obstacles (such as temporarily parked farm tools, fallen rocks, or people). When an obstacle is detected encroaching on the safety envelope of the predetermined trajectory, the system triggers a local obstacle avoidance replanning algorithm. This algorithm, while maintaining the trend towards the global target, samples and generates multiple sets of obstacle avoidance Bezier curves in the space surrounding the vehicle, evaluates the obstacle avoidance cost, and selects a local curve that bypasses the obstacle and smoothly transitions back to the original trajectory, thus piecing together a collision-free geometric path. The system then discretizes this geometric path, calculates the heading angle and expected linear velocity that the vehicle should maintain at each path point, and encapsulates it into a standard machine-readable protocol format according to a time series or distance interval, thereby generating the truck navigation command sequence.

[0097] Finally, the navigation command sequence for the extended-duty vehicle is issued to drive the vehicle, thereby forming a real-time adjusted route for the extended-duty vehicle to reach the optimal intersection coordinates. Specifically, the system sends the generated navigation command sequence to the extended-duty vehicle's underlying drive-by-wire chassis controller (including the steering actuator, electronic throttle, and brake-by-wire system) at a fixed control frequency (e.g., 50Hz) via the vehicle's CAN bus or Ethernet. The chassis controller uses a model predictive control (MPC) algorithm to track these commands and drive the steering motor and drive motor. As commands are continuously issued and executed, the extended-duty vehicle rolls out a real trajectory in the physical world, continuously avoiding obstacles and smoothly approaching the target, thus completing the physical closed loop of the real-time adjusted route from the current position to the optimal intersection coordinates.

[0098] In step S17, the step of monitoring the relative distance between the seeder and the fertilizer truck according to the travel route, and activating the docking protocol and obtaining a coordinated supply confirmation signal if the relative distance enters a preset docking range, includes:

[0099] The real-time positioning coordinates and velocity vectors of the seeder and the fertilizer truck are obtained, and the relative Euclidean distance and heading deviation angle between the seeder and the fertilizer truck are calculated using differential positioning data.

[0100] If the relative Euclidean distance is less than the preset docking range threshold and the heading deviation angle is within the allowable alignment sector, a docking ready status code is generated.

[0101] In response to the docking ready status code, the docking protocol is activated, a data transmission channel is established, and a handshake request frame is sent.

[0102] The handshake request frame is parsed, clock synchronization is performed and the lateral control is fine-tuned. When the alignment error between the supply interface and the receiving port converges, a collaborative supply confirmation signal is output.

[0103] First, the real-time positioning coordinates and velocity vectors of the seeder and the fertilizer truck are acquired. Differential positioning data is then used to calculate the relative Euclidean distance and heading deviation angle between the two vehicles. Specifically, as the two vehicles gradually approach each other, the system exchanges Real-Time Kinematic (RTK) positioning data packets at a high frequency (e.g., 10Hz) via a low-latency vehicle-to-vehicle (V2V) communication link. Using the distance formula between two points in three-dimensional space, the straight-line physical distance between the antenna centers of the two vehicles, i.e., the relative Euclidean distance, is accurately calculated. Simultaneously, the heading angle (the angle relative to true north) in the current velocity vectors of the two vehicles is extracted. By calculating the algebraic difference between the heading angles of the two vehicles and normalizing it to the range of -180 degrees to +180 degrees, the heading deviation angle used to assess the parallel alignment of the two vehicles is calculated.

[0104] Subsequently, if the relative Euclidean distance is less than the preset docking range threshold and the heading deviation angle is within the allowable alignment sector, a docking ready status code is generated. Specifically, the system continuously compares the calculated relative Euclidean distance with the preset docking range threshold (usually set to 50 to 90 meters, depending on the braking performance of the two vehicles and the stable connection distance of the communication module, such as 75 meters). Simultaneously, the system compares the heading deviation angle with the allowable alignment sector (set based on the mechanical swing degrees of freedom of the resupply cantilever, such as between -35 and +35 degrees). Only when both the distance and attitude indicators meet the aforementioned strict range limits is it considered that the two vehicles have entered a safe and controllable close-range interaction zone. The system's underlying logic circuitry and control software then jointly flip the status bit, generating a high-priority docking ready status code flag.

[0105] Subsequently, in response to the docking-ready status code, the docking protocol is activated, a data transmission channel is established, and a handshake request frame is sent. Specifically, when the central controller detects that the status code is set, it immediately triggers the interrupt handler, waking up the dormant docking protocol stack. The system establishes an encrypted and reliable data transmission channel between the two vehicles based on TCP / IP or a dedicated industrial protocol. The fertilizer truck, as the master control unit, constructs a data packet containing its unique device identifier (ID), current timestamp, checksum, and preparation docking request instruction according to the protocol specifications—that is, a handshake request frame—and sends it to the seeder through a dedicated channel. After receiving the frame, the seeder performs a CRC check; if there is no error, it replies with an acknowledgment frame, completing the electronic handshake process. Finally, the handshake request frame is parsed to perform clock synchronization and fine-tune the lateral control quantity. When the alignment error between the supply interface and the receiving port converges, a coordinated supply confirmation signal is output. Specifically, when parsing the handshake frame, the seeder extracts the timestamp of the fertilizer truck, combines it with network transmission delay estimation, and adjusts the local system clock through the network time protocol to ensure that the control systems of the two vehicles are strictly aligned at the millisecond level, avoiding misalignment and oscillation of control instructions due to time differences. For physical alignment, a laser rangefinder or machine vision system mounted on the side of the vehicle body is used to measure the lateral distance between the two vehicles in real time. This distance is compared with the set ideal docking distance (e.g., 1.5 meters), and the difference is input into a proportional-integral-derivative (PID) controller. The controller outputs a lateral fine-tuning command to the fertilizer truck's steering system, controlling the vehicle to move laterally slowly. When the sensor detects that the offset between the fertilizer truck's supply robotic arm interface and the seeder's receiving port center in three-dimensional space remains stable within the tolerance range (e.g., 0.15 meters) for two consecutive seconds, the alignment error is considered to have converged. At this time, the main control system pulls a high-level signal or sends a specific message to formally output a coordinated supply confirmation signal, instructing the robotic arm to extend and the valve to open.

[0106] In step S18, the step of recording the operation efficiency index and determining the starting point of the next prediction cycle through the feedback of the coordinated supply confirmation signal includes:

[0107] The operation completion timestamp and supply valve opening duration contained in the coordinated supply confirmation signal are analyzed, and the instantaneous filling volume and average operation flow rate are calculated by combining them with the real-time material transfer rate.

[0108] A comprehensive efficiency feature vector is constructed based on the instantaneous filling volume and the average operation flow rate. The comprehensive efficiency feature vector is then input into a preset operation resource consumption deduction model to generate a resource demand matrix.

[0109] Based on the resource demand matrix, the next optimal intersection node is retrieved, the expected arrival time corresponding to the optimal intersection node is locked, and the starting point of the next prediction cycle is determined.

[0110] First, the system analyzes the operation completion timestamp and supply valve opening duration contained in the coordinated supply confirmation signal, and calculates the instantaneous filling volume and average operating flow rate by combining this with the real-time material transfer rate. Specifically, when the supply action is completed and the robotic arm retracts, the system generates a confirmation signal packet containing the completion status. The central controller intercepts and analyzes this signal packet, extracting the precise operation completion timestamp recorded at the packet header, and the actual opening duration (accurate to milliseconds) of the supply valve from fully open to fully closed, as fed back by the underlying actuator. During valve opening, a Coriolis mass flow meter installed in the conveying pipeline collects the real-time material transfer rate at high frequency. The system performs time integration on this rate over the entire valve opening duration, accurately accumulating and calculating the total mass of fertilizer actually transported during this supply process, i.e., the instantaneous filling volume. Dividing the instantaneous filling volume by the supply valve opening duration yields the average operating flow rate, reflecting the overall performance of the material conveying system.

[0111] Subsequently, a comprehensive efficiency feature vector is constructed based on the instantaneous filling volume and the average operation flow rate. This comprehensive efficiency feature vector is then input into a preset operation resource consumption projection model to generate a resource demand matrix. Specifically, to comprehensively evaluate the efficiency of this collaborative operation, the system concatenates and normalizes the calculated instantaneous filling volume, average operation flow rate, and characteristic parameters such as the average engine fuel consumption rate and the total downtime of the two vehicles during the replenishment process obtained from the CAN bus, constructing a multi-dimensional comprehensive efficiency feature vector. It should be noted that the establishment and pre-training process of the operation resource consumption projection model includes: collecting similar comprehensive efficiency feature vectors from a large number of historical operation cycles as input; using the total actual fuel consumption, total working time, and subsequent fertilizer depletion time nodes within the corresponding cycle as multi-objective output labels; constructing a network architecture using a Transformer model based on a self-attention mechanism; and training the network parameters using a large amount of historical real data through backpropagation and gradient descent to enable it to uncover the implicit correlation between efficiency indicators and future resource consumption rates. The currently constructed comprehensive efficiency feature vector is input into the trained inference model. After forward computation, the model outputs a multi-dimensional matrix-based prediction result. This resource demand matrix details the predicted demand values ​​and time points for the seeder in terms of oilseeds, time margin, and the next fertilizer replenishment amount over a future period.

[0112] Finally, based on the resource demand matrix, the next optimal intersection node is retrieved, and the expected arrival time corresponding to the optimal intersection node is locked to determine the starting point of the next prediction cycle. Specifically, the system parses the output resource demand matrix and extracts the time point and consumption amount when fertilizer or fuel will reach the warning line next time. Using the same logic and algorithm engine as S11 to S16 above, combined with the road network topology map of the remaining unplanted paths in the field, the physical intersection node that meets the next supply demand is planned and searched in advance. After confirming that the node meets all safety and efficiency constraints, it is set as the next optimal intersection node. The system records the absolute timestamp corresponding to the arrival of the planter at the node and writes this time node into the global scheduling calendar of the main control program. The determination of this time node marks the successful completion of the current supply scheduling closed loop, and it also serves as the time starting point benchmark for triggering the next spatiotemporal prediction and path planning algorithm (i.e., the next prediction cycle), thereby realizing seamless and continuous intelligent collaboration in potato planting field operations.

[0113] In summary, this invention discloses a path planning method for a fertilizer delivery vehicle on a potato planter. The method includes acquiring a comprehensive dataset; processing the dataset using a time series forecasting method to obtain the speed fluctuation trend of the future path; the comprehensive dataset includes the real-time location coordinates of the fertilizer delivery vehicle, its speed, remaining fertilizer amount, terrain slope data, and implement load distribution data; calculating the fertilizer consumption rate per unit distance within the corresponding path segment and the estimated distance after path uncertainty adjustment based on the speed fluctuation trend; integrating the fertilizer consumption rate and the estimated distance to obtain the total fertilizer consumption for the subsequent path; determining a replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount; and triggering the fertilizer delivery vehicle path planning process if the replenishment timing threshold is lower than a preset remaining proportion; and acquiring the current position and speed data of the fertilizer delivery vehicle and combining it with the future path trajectory of the planter to generate... A subset of candidate coordinates is selected based on the arrival time and actual travel distance of the fertilizer truck at each candidate coordinate, resulting in a set of potential intersection coordinates with the minimum empty travel distance. The estimated downtime for each coordinate in this set is calculated, and the local fertilization uniformity value at the start-up stage of the seeder is predicted based on this estimated downtime. Combining the local fertilization uniformity value and the estimated downtime, the optimal intersection coordinate is selected from the set of potential intersection coordinates. Coordinate points are extracted from the optimal intersection coordinates, and the fertilizer truck's navigation instructions are updated to obtain a real-time adjusted route. The relative distance between the seeder and the fertilizer truck is monitored according to the route. If the relative distance enters a preset docking range, the docking protocol is activated, and a collaborative supply confirmation signal is obtained. Based on the feedback of the collaborative supply confirmation signal, operational efficiency indicators are recorded, and the starting point for the next prediction cycle is determined. This invention, through path planning, achieves efficient and precise collaborative operation between the seeder and the fertilizer truck in complex field environments, significantly improving operational efficiency and fertilizer utilization.

[0114] Reference Figure 2 The second embodiment of the present invention provides a path planning system for a potato planter fertilizer truck, comprising:

[0115] The data acquisition and prediction module is used to acquire a comprehensive dataset, process the comprehensive dataset using time series forecasting methods, and obtain the speed fluctuation trend of the future path.

[0116] The consumption calculation module is used to calculate the fertilizer consumption rate per unit distance within the corresponding path segment and the expected distance after path uncertainty adjustment based on the speed fluctuation trend, and to perform an integral operation on the fertilizer consumption rate and the expected distance to obtain the total fertilizer consumption of the subsequent path.

[0117] The replenishment judgment and triggering module is used to determine the replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount. If the replenishment timing threshold is lower than the preset remaining ratio, the fertilizer truck path planning process is triggered.

[0118] The intersection point search module is used to obtain the current position and speed data of the fertilizer truck, and combine it with the future path trajectory of the seeder to search for the set of potential intersection coordinates with the minimum empty driving distance using a path optimization algorithm;

[0119] The intersection point evaluation module is used to evaluate the downtime corresponding to each coordinate in the set of potential intersection coordinates, and determine the optimal intersection coordinates by comparing the impact of the variable fertilization uniformity.

[0120] The navigation update module is used to extract coordinate points from the optimal intersection coordinates, update the navigation instructions of the fertilizer truck, and obtain the real-time adjusted travel route.

[0121] The docking monitoring module is used to monitor the relative distance between the seeder and the fertilizer truck according to the travel route. If the relative distance enters the preset docking range, the docking protocol is activated and a coordinated supply confirmation signal is obtained.

[0122] The feedback and loop module is used to feed back the collaborative replenishment confirmation signal to the system, record the operation efficiency index, and determine the starting point of the next prediction loop.

[0123] It should be noted that the potato planter fertilizer truck path planning system provided in this embodiment of the invention is used to execute all the process steps of the potato planter fertilizer truck path planning method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0124] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0125] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A path planning method for a fertilizer application vehicle on a potato planter, characterized in that, include: A comprehensive dataset is obtained and processed using a time series forecasting method to obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, its driving speed, the amount of remaining fertilizer, terrain slope data, and equipment load distribution data. Based on the speed fluctuation trend, calculate the fertilizer consumption rate per unit distance within the corresponding path segment and the expected distance after path uncertainty adjustment. Integrate the fertilizer consumption rate and the expected distance to obtain the total fertilizer consumption for the subsequent path. Based on the difference between the total fertilizer consumption and the current remaining fertilizer, a replenishment timing threshold is determined. If the replenishment timing threshold is lower than the preset remaining ratio, the fertilizer truck path planning process is triggered. The current location and speed data of the fertilizer truck are obtained, and a subset of candidate coordinates is generated by combining the future path trajectory of the seeder. The fertilizer truck is then filtered based on the time it takes to reach each candidate coordinate and the actual distance it travels, resulting in a set of potential intersection coordinates with the minimum empty driving distance. Calculate the expected downtime for each coordinate in the potential intersection coordinate set, predict the local fertilization uniformity value of the seeder at the start-up stage based on the expected downtime, and select the optimal intersection coordinate from the potential intersection coordinate set by combining the local fertilization uniformity value and the expected downtime. The coordinate points are extracted from the optimal intersection coordinates, the navigation instructions for the fattening vehicle are updated, and the real-time adjusted travel route is obtained. The relative distance between the seeder and the fertilizer truck is monitored according to the travel route. If the relative distance enters the preset docking range, the docking protocol is activated and a coordinated supply confirmation signal is obtained. By using the feedback of the coordinated supply confirmation signal, the operation efficiency index is recorded, and the starting point of the next prediction cycle is determined.

2. The path planning method for the fertilizer application vehicle of the potato planter according to claim 1, characterized in that, The process involves acquiring a comprehensive dataset, processing it using time series forecasting to obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, its speed, remaining fertilizer quantity, terrain slope data, and equipment load distribution data. Obtain an initial multidimensional dataset, which includes real-time location coordinates, current velocity vector, remaining fertilizer volume, slope data, and machinery weight distribution. The terrain undulation gradient is calculated based on the location coordinates and slope data in the initial multidimensional dataset, and the comprehensive load adjustment coefficient is determined in combination with the equipment weight distribution. If the comprehensive load adjustment coefficient exceeds the preset load threshold, the speed vector and slope sequence in the comprehensive dataset are smoothed using a moving average filtering algorithm to obtain a filtered comprehensive dataset. The filtered comprehensive dataset is then processed using a long short-term memory network to obtain a speed prediction sequence. Based on the velocity prediction sequence, the vector differences between adjacent nodes are analyzed, and the frequency components are extracted by Fourier transform to obtain the velocity fluctuation trend on the future path.

3. The path planning method for the fertilizer application vehicle of the potato planter according to claim 1, characterized in that, The process involves calculating the fertilizer consumption rate per unit distance within the corresponding path segment and the estimated distance after path uncertainty adjustment based on the speed fluctuation trend. Integrating the fertilizer consumption rate and the estimated distance yields the total fertilizer consumption for the subsequent path, including: Based on the speed fluctuation trend, a pre-acquired flow calibration curve is retrieved, and combined with the pre-acquired fertilizer particle flowability characteristic value, a dynamic speed control sequence for the fertilizer dispenser is generated. The dynamic speed control sequence is input into a pre-established discrete flow calculation model to obtain an instantaneous mass flow sequence. Combined with the speed fluctuation trend, the fertilizer consumption rate curve per unit distance is calculated. Calculate the ground slip rate, generate path uncertainty adjustment coefficients using the ground slip rate and terrain gradient data in the comprehensive dataset, and obtain the predicted distance after path uncertainty adjustment by weighting the planned path segments according to the path uncertainty adjustment coefficients; The fertilizer consumption rate curve is integrated over the predicted distance to obtain the total fertilizer consumption for the subsequent path.

4. The path planning method for the fertilizer application vehicle of the potato planter according to claim 3, characterized in that, The step of determining a replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount, and triggering a fertilizer truck path planning process if the replenishment timing threshold is lower than a preset remaining ratio, includes: The current remaining amount of fertilizer is collected by a gravity sensor. The total amount of fertilizer consumed is subtracted from the current remaining amount of fertilizer to obtain the difference. If the difference is positive, the percentage of the current remaining amount of fertilizer to the total full load is calculated, and the percentage is determined as the replenishment timing threshold. If the replenishment timing threshold is lower than the preset remaining ratio, the coordinates of the expected operation interruption point where the fertilizer is exhausted are deduced based on the fertilizer consumption rate curve. Obtain the real-time location information of the fertilizer truck and associate it with the coordinates of the operation interruption point to generate a collaborative scheduling request; In response to the collaborative scheduling request, the fertilizer truck path planning engine is activated, and the optimal driving path from the real-time location of the fertilizer truck to the coordinates of the operation interruption point is calculated based on road network data, thereby triggering the fertilizer truck path planning process.

5. The path planning method for the fertilizer application vehicle of the potato planter according to claim 1, characterized in that, The process involves acquiring the current position and speed data of the fertilizer truck, combining it with the future path trajectory of the seeder to generate a subset of candidate coordinates, and then filtering these based on the time the fertilizer truck arrives at each candidate coordinate and the actual distance traveled to obtain a set of potential intersection coordinates with the minimum empty travel distance. This set includes: The current location and speed of the fertilizer truck and the future path data of the seeder are obtained. A dynamic geofence is set with the future trajectory line of the seeder as the center. Field road nodes that fall within the dynamic geofence are used as the preliminary range of potential intersection coordinates. Within the initial range of the potential intersection coordinates, spatial topology analysis is performed in conjunction with farmland operation constraints to eliminate nodes that do not meet the accessibility and operation safety constraints, and retain coordinates that meet the constraints to form a subset of candidate coordinates. The speed of the fertilizer truck is obtained, and the candidate coordinate subset is adjusted by integrating the fertilizer truck speed to determine the optimized coordinate set; For each coordinate point in the optimized coordinate set, a path planning algorithm is called to calculate the actual driving distance from the current position of the fattening vehicle to that coordinate point along the passable road network. Based on the actual driving distance, all coordinate points are sorted, and several coordinate points with the smallest distance are extracted to form a potential intersection coordinate set with the minimum empty driving distance.

6. The path planning method for the fertilizer application vehicle of the potato planter according to claim 1, characterized in that, The process of calculating the estimated downtime for each coordinate in the potential intersection coordinate set, predicting the local fertilization uniformity value during the seeder's start-up phase based on the estimated downtime, and combining the local fertilization uniformity value and the estimated downtime to select the optimal intersection coordinate from the potential intersection coordinate set includes: Calculate the estimated downtime for each coordinate in the potential intersection coordinate set. The estimated downtime is obtained by adding the fertilizer replenishment time to the time difference between the estimated arrival time of the seeder and the estimated arrival time of the fertilizer truck. The instantaneous flow deviation value during the startup phase is determined based on the estimated downtime and the preset path curvature of the seeder after it starts at each coordinate point. The instantaneous flow deviation value is obtained based on a preset mapping table of downtime and flow recovery delay. The instantaneous flow rate deviation value is input into a pre-trained particle distribution density prediction model to obtain the local fertilization uniformity value; The coordinates with local fertilization uniformity values ​​lower than the preset uniformity threshold are removed, and the coordinate with the shortest expected downtime is selected from the remaining coordinates as the optimal intersection coordinate.

7. The method for planning the path of a fertilizer application vehicle for a potato planter according to claim 1, characterized in that, The step of extracting coordinate points from the optimal intersection coordinates, updating the navigation instructions for the fattening vehicle, and obtaining the real-time adjusted travel route includes: The optimal intersection coordinates are analyzed to obtain the navigation endpoint node, and the global optimal path is generated by combining the real-time positioning coordinates of the fertilizer truck. The globally optimal path is smoothed based on the pre-obtained vehicle turning radius to generate a feasible driving trajectory; Obtain real-time obstacle information, replan the feasible driving trajectory based on the real-time obstacle information to obtain a collision-free geometric path, and generate a fattening vehicle navigation command sequence based on the collision-free geometric path; The sequence of navigation instructions for the fattening vehicle is issued to drive the vehicle and form a real-time adjusted route for the fattening vehicle to travel towards the optimal intersection coordinates.

8. The method for path planning of a fertilizer application vehicle for a potato planter according to claim 1, characterized in that, The step of monitoring the relative distance between the seeder and the fertilizer truck according to the travel route, and activating the docking protocol and obtaining a coordinated supply confirmation signal if the relative distance enters the preset docking range, includes: The real-time positioning coordinates and velocity vectors of the seeder and the fertilizer truck are obtained, and the relative Euclidean distance and heading deviation angle between the seeder and the fertilizer truck are calculated using differential positioning data. If the relative Euclidean distance is less than the preset docking range threshold and the heading deviation angle is within the allowable alignment sector, a docking ready status code is generated. In response to the docking ready status code, the docking protocol is activated, a data transmission channel is established, and a handshake request frame is sent. The handshake request frame is parsed, clock synchronization is performed and the lateral control is fine-tuned. When the alignment error between the supply interface and the receiving port converges, a collaborative supply confirmation signal is output.

9. The path planning method for the fertilizer application vehicle of the potato planter according to claim 1, characterized in that, The process of recording operational efficiency indicators and determining the starting point of the next prediction cycle through feedback of the coordinated supply confirmation signal includes: The operation completion timestamp and supply valve opening duration contained in the coordinated supply confirmation signal are analyzed, and the instantaneous filling volume and average operation flow rate are calculated by combining them with the real-time material transfer rate. A comprehensive efficiency feature vector is constructed based on the instantaneous filling volume and the average operation flow rate. The comprehensive efficiency feature vector is then input into a preset operation resource consumption deduction model to generate a resource demand matrix. Based on the resource demand matrix, the next optimal intersection node is retrieved, the expected arrival time corresponding to the optimal intersection node is locked, and the starting point of the next prediction cycle is determined.

10. A path planning system for a potato planter fertilizer applicator, characterized in that, include: The data acquisition and prediction module is used to acquire a comprehensive dataset, process the comprehensive dataset using time series prediction methods, and obtain the speed fluctuation trend of the future path. The comprehensive dataset includes the real-time location coordinates of the fertilizer truck, driving speed, remaining fertilizer amount, terrain slope data, and equipment load distribution data. The consumption calculation module is used to calculate the fertilizer consumption rate per unit distance within the corresponding path segment and the expected distance after path uncertainty adjustment based on the speed fluctuation trend, and to perform an integral operation on the fertilizer consumption rate and the expected distance to obtain the total fertilizer consumption of the subsequent path. The replenishment judgment and triggering module is used to determine the replenishment timing threshold based on the difference between the total fertilizer consumption and the current remaining fertilizer amount. If the replenishment timing threshold is lower than the preset remaining ratio, the fertilizer truck path planning process is triggered. The intersection point search module is used to obtain the current position and speed data of the fertilizer truck, combine it with the future path trajectory of the seeder to generate a subset of candidate coordinates, and filter them based on the time when the fertilizer truck arrives at each candidate coordinate and the actual driving distance to obtain the set of potential intersection coordinates with the minimum empty driving distance. The intersection point evaluation module is used to calculate the expected downtime corresponding to each coordinate in the potential intersection coordinate set, predict the local fertilization uniformity value of the seeder in the start-up stage based on the expected downtime, and select the optimal intersection coordinate from the potential intersection coordinate set by combining the local fertilization uniformity value and the expected downtime. The navigation update module is used to extract coordinate points from the optimal intersection coordinates, update the navigation instructions of the fertilizer truck, and obtain the real-time adjusted travel route. The docking monitoring module is used to monitor the relative distance between the seeder and the fertilizer truck according to the travel route. If the relative distance enters the preset docking range, the docking protocol is activated and a coordinated supply confirmation signal is obtained. The feedback and loop module is used to record the operation efficiency index and determine the starting point of the next prediction loop through the feedback of the coordinated replenishment confirmation signal.

Images