Cell-based breeding seeding path control method and system

Abstract

The application provides a sowing path control method and system based on cell breeding, relates to the technical field of sowing path control methods, and comprises the following steps: matching and calibrating an electronic planting scheme map with actual coordinates of a field plot; combining all paths into a population, and updating the population according to a fitness coefficient; selecting a path with the highest fitness coefficient in the updated population as a predetermined sowing path and starting a seed metering device; when it is judged that a seeding machine enters an allowable error range of the predetermined end point coordinates, stopping sowing; when the seeding machine enters a start point position of a next cell, an automatic confirmation is made by a control system, and the start point coordinates and the end point coordinates of the next cell are set as new comparison targets. The four-parameter similarity transformation model is adopted to complete accurate registration of theoretical coordinates and actual geographical coordinates; and a dynamic cross-entropy mechanism is introduced under the genetic algorithm framework, so that the algorithm maintains population diversity in the iteration process.

Description

Technical Field

[0001] This invention relates to the field of sowing path control methods, specifically to a sowing path control method and system based on plot breeding. Background Technology

[0002] In the process of variety selection, regional trials, and improved seed breeding of crops such as rapeseed, wheat, and soybeans, numerous plot sowing trials are required. Plot sowing requires each variety to be precisely sown on a fixed area, typically with a certain number of rows and row length, to ensure the accuracy and comparability of the trials. Currently, plot sowing relies heavily on manual measurement with measuring tapes and marking with stakes to determine the start and end points of the sowing rows, or on simple GPS navigation for straight-line control. However, this method is inefficient and labor-intensive. For hundreds or thousands of experimental plots, repeated manual measurement and marking are time-consuming and labor-intensive, resulting in slow sowing speed. Furthermore, it suffers from poor accuracy and is prone to errors. Human visual judgment and the tightness of the measuring tape can affect measurement accuracy, and fatigue or external interference can easily lead to marking errors and inconsistent row lengths, directly impacting the reliability of the experimental data. Most existing automated sowing methods can only travel along fixed straight paths and cannot dynamically adjust the sowing path according to actual field conditions, making it difficult to guarantee sowing quality.

[0003] In the prior art, document CN117369452A proposes a method to determine the global driving path of an unmanned agricultural machine based on its attribute information and the area information of the area to be operated. The global driving path of the unmanned agricultural machine is pre-planned, and the unmanned agricultural machine is tracked and controlled throughout the process using real-time location information, and the unmanned agricultural machine is automatically controlled to perform corresponding agricultural machine behaviors. However, this method is mainly aimed at continuous operation in large-area farmland and does not involve the boundary-constrained zone-by-zone sowing control of multiple independent, sequentially arranged zones in the context of small-area breeding. Furthermore, it cannot effectively correct deviations from the predetermined path during sowing. Therefore, there is an urgent need for a sowing path control method and device based on small-area breeding.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a sowing path control method and system based on plot breeding to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The sowing path control method based on plot breeding includes the following steps: S1: Extract the starting point coordinates, ending point coordinates, and standard row direction of the current target cell according to the cell number order in the experimental field. With the boundary of the target cell as a constraint, generate a path node sequence between the starting point coordinates and the ending point coordinates at a fixed sampling interval. Construct several candidate paths around the standard row direction and set the direction angle and speed of each path node of each candidate path to obtain the candidate path population. S2: For each path in the candidate path population, calculate the benefit coefficient of each path node according to the set direction angle and speed. Further calculate the fitness coefficient of each path based on the benefit coefficient of the path node, and iteratively update the candidate path population. The iterative update strategy adopts the method of retaining the current best path and reorganizing and mutating the remaining paths. After reaching the preset iteration termination condition, output the path with the highest fitness coefficient as the predetermined seeding path. S3: Perform sowing operations based on the predetermined sowing path. During the movement, continuously acquire the current position coordinates of the sowing machine, compare the deviation between the current position coordinates and the predetermined sowing path, and perform path correction. Continuously compare the distance between the current position coordinates and the endpoint coordinates of the current target cell. When the current position coordinates enter the allowable error range of the endpoint coordinates, the sowing of the current target cell is stopped. S4: After the current target cell is stopped, select the next target cell in the preset numbering order and repeat S1 to S3 until all cells in the experimental field are sown.

[0007] Furthermore, a path node sequence is generated discretely between the starting point coordinates and the ending point coordinates at a fixed sampling interval, specifically as follows: Using the starting and ending coordinates of the current target cell as endpoints and the boundary of the target cell as constraints, equidistant sampling is performed on the line segment between the starting and ending points at a fixed sampling interval to generate a series of path nodes that constitute a path node sequence.

[0008] Furthermore, several candidate paths are constructed around the standard row direction, specifically: Using the standard row direction of the current target cell as the reference direction, a random direction offset following a normal distribution is independently generated for each node in the path node sequence. The mean of this offset is zero and the standard deviation is the preset direction variation intensity. The obtained random offset is superimposed on the standard row direction to obtain the actual direction angle of each path node. Within the preset minimum and maximum speed range, all nodes on this path are randomly assigned the same speed value as the driving speed of the entire path. The path generation process is repeated continuously, generating a complete candidate path each time, until the number of generated candidate paths reaches the preset population size. All candidate paths are then aggregated to form a candidate path population.

[0009] Furthermore, the revenue coefficient of the path node is calculated in the following way: For each candidate path in the candidate path population, obtain the Euclidean distance from each path node to the nearest experimental field boundary, and calculate the boundary safety weight of the path node based on the Euclidean distance: in, Indicates the first Boundary security weights of each path node; This indicates the preset safe operating distance threshold; Indicates the first Euclidean distance from each path node to the boundary of the nearest experimental plot; The index of the path node; Get the Priority of each path node in the cell Calculate the seeding priority weight: in, Represents path nodes The seeding priority weight; Represents path nodes The priority level of the residential community; Indicates the priority weight coefficient; Based on the boundary safety weight and seeding priority weight of each path node, calculate its corresponding profit coefficient: in, Represents path nodes The profit coefficient.

[0010] Further, the fitness coefficient is calculated as follows: Based on the changes in steering angle and speed of each path node relative to the previous path node, a steering penalty coefficient is constructed, the expression of which is as follows: in, Indicates the first The turning penalty coefficient for each path node, index It is an integer greater than or equal to 2; Indicates the first The direction and angle of each path node; Indicates the first The speed of each path node; Indicates the first The speed of each path node; , These represent the preset first and second penalty weight coefficients, respectively. ; Calculate the fitness coefficient for each path based on the reward coefficient and turning penalty coefficient of each path node: in, Indicates the first The number of path nodes in the path; Indicates the first The fitness coefficient of each path; Indicates the first The first path The turning penalty coefficient for each path node; Indicates the first The first path The revenue coefficient of each path node; For path indexing.

[0011] Furthermore, the iterative update strategy retains the current optimal path and reorganizes the remaining paths, specifically as follows: Sort all candidate paths by fitness coefficient from largest to smallest, retain the top-ranked candidate paths in the next generation population and mark them as the optimal paths, and do not participate in subsequent recombination and mutation operations. Two candidate paths are randomly selected from the remaining candidate paths. The candidate path with the highest fitness coefficient is selected to enter the next generation population. This step is repeated until the size of the next generation population reaches the preset population size. For paths that have reached the preset population size, random pairings are performed. For the two paired paths, since the path node sequence lengths of the two paths are the same, the index of a common path node is randomly selected as the intersection point, and the dynamic cross-entropy of the two paired paths is calculated: in, Represents dynamic cross-entropy; This represents the preset maximum cross-entropy; This represents the preset minimum cross-entropy; Indicates the first path in the parent path The direction and angle of each path node; Indicates the first generation in the parent generation path The direction and angle of each path node; This represents the index of a randomly selected path node; When the dynamic cross-entropy is greater than the preset cross-entropy, the parent and mother paths are swapped from the first... For each path node, from the first path node to the last path node, generate two new child paths; otherwise, copy all path nodes of the parent and mother paths directly as child path nodes.

[0012] Furthermore, the remaining paths are mutated, specifically as follows: For each path in the next generation population obtained after recombination, calculate its path repetition rate and calculate the mutation probability based on the path repetition rate. When the mutation probability is greater than the preset mutation probability threshold, mutate all path nodes of that path. The path repetition rate is calculated as follows: The straight-line distances between all adjacent path nodes are summed to obtain the actual total length of the path. The actual total length is then divided by the straight-line distance from the starting point to the ending point of the path to obtain the tortuosity. The entire circumference is divided into several intervals. The directional change angles between all adjacent path nodes are calculated, and these angles are categorized into their corresponding intervals. The probability of each interval occurring is calculated. The probabilities of all intervals are multiplied by the logarithm of that probability, summed, and the result is the directional entropy. The tortuosity and directional entropy are normalized. The normalized tortuosity and directional entropy are multiplied by the tortuosity balance coefficient and the directional entropy balance coefficient, respectively, and summed to obtain a comprehensive index. This comprehensive index is then subtracted from the total to obtain the path repetition rate. Calculate the probability of variation based on path repetition rate: in, This represents the preset base variation rate; This represents the coefficient of influence of repetition rate; Indicates the first The path repetition rate of each path; Indicates the first The probability of mutation of each path; When the mutation probability is greater than the preset mutation probability, all path nodes of the path are mutated: in, This indicates that the mean is 0 and the standard deviation is 0. Gaussian distributed random numbers; This indicates the preset maximum speed of the seed metering device; This indicates the preset minimum speed of the seed metering device; Indicates the mutated first The direction and angle of each path node; Represents the mutated path node speed; Represents the path node before mutation The speed.

[0013] Furthermore, the current location coordinates are compared with the predetermined sowing path to determine the deviation and path correction is performed, specifically as follows: The preset number of paths and the preset remaining number of paths are merged to form a new generation population. The fitness coefficient of each path in the new generation population is calculated. When the difference between the maximum fitness coefficient in the new generation population and the maximum fitness coefficient of the previous generation population is less than a preset difference, the path corresponding to the maximum fitness coefficient in the current new generation population is selected. Combined with the starting point of the seeder, the path nodes in the path are converted into actual sowing coordinates to obtain the predetermined sowing path; otherwise, the fitness coefficient is recalculated and the population is updated. The seeding operation is performed based on a predetermined seeding path. During the movement, the current position coordinates of the seeder are continuously acquired. The index of the path node closest to the current position is searched on the predetermined seeding path. The lateral and longitudinal deviations are calculated. Based on the magnitude of the lateral deviation, the path nodes are dynamically adjusted in the following manner: When the lateral deviation is less than the first preset threshold, the original predetermined sowing path remains unchanged; when the lateral deviation is greater than or equal to the first preset threshold and less than the second preset threshold, a cubic Bézier curve connecting the start and end points is generated, with the current position as the starting point and the path node at the preset distance ahead on the predetermined sowing path as the end point, and the path node between the start and end points in the original predetermined sowing path is replaced by the path node on the Bézier curve; when the lateral deviation is greater than or equal to the second preset threshold, S2 is re-executed with the current position of the seeder as the new starting point and the end point of the original predetermined sowing path as the end point.

[0014] Further, the next target cell is selected according to the preset numbering order, specifically as follows: After the seeding in the current target cell is stopped, as the seeder moves away from the current cell and towards the next cell, the distance between the current position of the seeder and the starting coordinates of the next cell is continuously calculated. When the distance is less than the preset distance, the control system automatically confirms that the seed metering device has reached the starting point of the next cell, sets the starting coordinates and ending coordinates of the next cell as the new comparison targets, and repeats steps S1-S3 until the seeding of all cells is completed.

[0015] The present invention also provides a seeding path control system based on plot breeding, wherein the seeding path control system based on plot breeding is used to execute the above-described seeding path control method based on plot breeding, comprising: Data acquisition module: It is used to extract the starting point coordinates, ending point coordinates and standard row direction of the current target cell according to the cell number order in the experimental field. With the boundary of the target cell as a constraint, it generates a path node sequence between the starting point coordinates and the ending point coordinates at a fixed sampling interval. It also constructs several candidate paths around the standard row direction and sets the direction angle and speed of each path node of each candidate path to obtain the candidate path population. Update Calculation Module: For each path in the candidate path population, calculate the benefit coefficient of each path node according to the set direction angle and speed. Based on the benefit coefficient of the path node, further calculate the fitness coefficient of each path and iteratively update the candidate path population. The iterative update strategy adopts the following approach: retain the current best path, reorganize and mutate the remaining paths. After reaching the preset iterative update termination condition, output the path with the highest fitness coefficient as the predetermined seeding path. The comparison and sowing module is used to perform sowing operations based on a predetermined sowing path. During the movement, it continuously acquires the current position coordinates of the sowing machine, compares the current position coordinates with the predetermined sowing path to perform deviation correction, and continuously compares the current position coordinates with the destination coordinates of the current target cell. When the current position coordinates enter the allowable error range of the destination coordinates, the sowing of the current target cell is stopped. The sowing completion module is used to select the next target cell in a preset numbered order after the current target cell has been stopped from sowing, and repeat the data acquisition module to the comparison sowing module until all cells in the experimental field have been sown.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This application introduces a path iteration optimization mechanism based on genetic algorithms, which can automatically generate sowing paths adapted to actual field conditions. This mechanism not only considers the safety distance of the field boundary and the priority of the plot experiment, but also constrains the changes in direction angle and speed by constructing a turning penalty coefficient, ensuring that the planned path is within the physical execution capability of the seeder. It realizes intelligent path generation under multi-objective constraints, which greatly improves the rationality and executability of the sowing path. It adopts an iterative update strategy that combines retaining the optimal path, recombination, and mutation. In the recombination operation, dynamic cross-entropy based on the difference in direction angle is introduced, and in the mutation operation, the mutation intensity is adaptively adjusted according to the path repetition rate, which effectively avoids the algorithm getting trapped in local optima, ensures population diversity, and thus can search for globally optimal sowing paths. In the path execution stage, a hierarchical path correction mechanism is adopted. According to the lateral deviation between the current position of the seeder and the predetermined path, three different strategies are adopted: maintaining the original path, local smooth regression of Bézier curve, and global replanning. This realizes pure software-level path correction without relying on the underlying hardware controller. At the same time, automatic stopping of sowing is achieved by continuously comparing the distance between the current position and the coordinates of the plot end point, which ensures the accuracy of row length control. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a graph showing the relationship between the Euclidean distance from the path node to the boundary of the nearest experimental field and the profit coefficient. Figure 3 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example: Please see Figures 1-2 The present invention provides a technical solution: The sowing path control method based on plot breeding includes the following steps: S1: Extract the starting point coordinates, ending point coordinates, and standard row direction of the current target cell according to the cell number order in the experimental field. With the boundary of the target cell as a constraint, generate a path node sequence between the starting point coordinates and the ending point coordinates at a fixed sampling interval. Construct several candidate paths around the standard row direction and set the direction angle and speed of each path node of each candidate path to obtain the candidate path population. The path node sequence is generated discretely between the starting coordinates and the ending coordinates at a fixed sampling interval, specifically as follows: Using the starting and ending coordinates of the current target cell as endpoints and the boundary of the target cell as constraints, equidistant sampling is performed on the line segment between the starting and ending points at a fixed sampling interval to generate a series of path nodes that constitute a path node sequence.

[0021] In the above process, by using the starting and ending coordinates of the current target cell as endpoints and the boundary of the target cell as constraints, equidistant sampling is performed on the line segment between the starting and ending points at a fixed sampling interval. This discretizes the continuous seeding path into a series of path nodes evenly distributed at fixed intervals. This provides a standardized and structured data foundation for subsequent candidate path construction, path node-level direction angle and speed settings, as well as node index-based payoff coefficient calculation, fitness evaluation, crossover and mutation operations. At the same time, the constraint of the target cell boundary ensures that all generated path nodes are within the legal range of the current target cell, avoiding seeding paths going out of bounds. The fixed sampling interval ensures the consistency of the number of nodes among different candidate paths.

[0022] Several candidate paths are constructed around the standard row direction, specifically: Using the standard row direction of the current target cell as the reference direction, a random direction offset following a normal distribution is independently generated for each node in the path node sequence. The mean of this offset is zero and the standard deviation is the preset direction variation intensity. The obtained random offset is superimposed on the standard row direction to obtain the actual direction angle of each path node. Within the preset minimum and maximum speed range, all nodes on this path are randomly assigned the same speed value as the driving speed of the entire path. The path generation process is repeated continuously, generating a complete candidate path each time, until the number of generated candidate paths reaches the preset population size. All candidate paths are then aggregated to form a candidate path population.

[0023] In the above process, the preset minimum speed is set to 0.5 m / s to 0.8 m / s, while the preset maximum speed is limited by the seeder's terrain adaptability, turning safety, and the maximum working frequency of the seed metering device. Too high a speed will lead to increased deviation in the seed landing position, unstable sowing depth, and the risk of tipping over during emergency turns. The preset maximum speed is usually set to 1.5 m / s to 2.5 m / s. In small-plot breeding and seeding scenarios, the number of path nodes is typically between 10 and 50, resulting in a relatively limited search space. Therefore, the population size is generally set to 20 to 100 paths. For plots with larger areas and more path nodes, the population size can be appropriately increased to 100 to 200 paths to ensure sufficient search capability. For plots with smaller areas or limited computing resources, a size of 20 to 50 paths can be set to balance computational efficiency and path quality. By using the standard row direction of the current target cell as the reference direction, a random directional offset is independently generated for each node in the path node sequence. This offset follows a normal distribution with a mean of zero and a standard deviation equal to the preset directional variation intensity. This offset is then superimposed on the standard row direction, thereby assigning each path node a different actual directional angle. This achieves ordered random exploration of the seeding direction near the standard row direction. Simultaneously, within a preset minimum and maximum speed range, all nodes along the entire path are randomly assigned the same speed value, ensuring a constant speed during the path's movement and simplifying the parameter complexity of the initial population. By continuously repeating the above path generation process, a complete candidate path is generated each time until the number of candidate paths reaches the preset population size. All candidate paths are then aggregated to form a candidate path population. The value of the preset directional variation intensity is usually based on the minimum turning radius of the seeder. Sampling interval with adjacent path nodes The upper limit is determined by the physical turning capability of the seeder: the maximum allowable change in direction angle between adjacent nodes is approximately... The preset direction variation intensity should be significantly less than this value to ensure that the generated candidate path is within the actual steerable range. Typically, it is set to... to , Indicates the preset direction variation intensity; for example, when the sampling interval Meters, minimum turning radius of the seeder When the distance is 1 meter, the maximum permissible directional change is approximately 0.167 radians, or about 9.6°. At this point, the preset directional variation intensity can be set between 0.03 and 0.08 radians. The value of the preset directional variation intensity directly affects the diversity of the initial path population: a value that is too small will cause all candidate paths to almost coincide with the standard row direction, resulting in insufficient population diversity and making it difficult for the algorithm to find better non-linear paths; a value that is too large will cause the candidate path directions to fluctuate drastically, exceeding the actual turning capability of the seeder or producing unreasonable zigzag paths. In practical applications, the preset directional variation intensity is stored as a system preset parameter in the control system, and a value of 0.05 to 0.12 radians is generally recommended.

[0024] S2: For each path in the candidate path population, calculate the benefit coefficient of each path node according to the set direction angle and speed. Further calculate the fitness coefficient of each path based on the benefit coefficient of the path node, and iteratively update the candidate path population. The iterative update strategy adopts the method of retaining the current best path and reorganizing and mutating the remaining paths. After reaching the preset iteration termination condition, output the path with the highest fitness coefficient as the predetermined seeding path. The revenue coefficient of a path node is calculated in the following way: For each candidate path in the candidate path population, obtain the Euclidean distance from each path node to the nearest experimental field boundary, and calculate the boundary safety weight of the path node based on the Euclidean distance: in, Indicates the first Boundary security weights of each path node; This indicates the preset safe operating distance threshold; Indicates the first Euclidean distance from each path node to the boundary of the nearest experimental plot; The index of the path node; In the above process, The distance should be at least half the width of the seeder body plus a safety margin to ensure that the actual outer edge of the seeder does not touch the field boundary when the path node is within this distance; typically, 0.5 to 1.0 times the working width of the seeder is used as the safety margin. The basic reference value, for example, when the working width of the seeder is 2 meters, It can be set to 1 meter to 2 meters; in practical applications, The value is stored in the control system as a preset parameter, and it is generally recommended to take a value of 1.0 meter to 3.0 meter. Using hyperbolic tangent function Calculate boundary safety weights The core reason is that this function can calculate the Euclidean distance from a path node to the boundary of the nearest experimental plot. Relative to the safe operating distance threshold The ratio is smoothly mapped to Within the interval. When the path node is far from the field boundary, that is... Much larger hour, A value approaching 1 indicates that the node is in a safe zone, and the boundary safety weight is close to its maximum value; when a path node is close to the field boundary, it means... When it approaches 0, A value close to 0 indicates that the node is near the boundary, posing a higher security risk, and the boundary security weight is close to its minimum. The advantages of this design are twofold: firstly, the hyperbolic tangent function is continuous and monotonically increasing throughout its domain, reflecting the smooth impact of distance changes on security and avoiding abrupt weight changes due to threshold switching; secondly, the function value is always positive and does not exceed 1, ensuring that the boundary security weight term... The yield coefficient has a controllable range of magnitude, which facilitates weighted combination with other items such as sowing priority weight.

[0025] Get the Priority of each path node in the cell Calculate the seeding priority weight: in, Represents path nodes The seeding priority weight; Represents path nodes The priority level of the residential community; Indicates the priority weight coefficient; Based on the boundary safety weight and seeding priority weight of each path node, calculate its corresponding profit coefficient: in, Represents path nodes The profit coefficient.

[0026] In the above process, the priority level is divided into three levels, and the corresponding values ​​are set as follows: Level 1 is the core variety of the community, taking... Level 2, i.e., conventional test area Level 3 refers to border crossing, protection crossing, or backup cell. The numerical design employs an arithmetic progression of 1:2:3 to maintain a fixed ratio in the priority weight difference between different levels. This ensures that the sowing priority weight of the first-level plot is three times that of the third-level plot and twice that of the second-level plot, thus reflecting a clear and quantifiable priority difference in the benefit coefficient. The rationale for this numerical design lies in the following: the arithmetic progression method is simple and intuitive, facilitating rapid understanding and setting by breeding experimenters during the design phase; simultaneously, the baseline value of 1.0 ensures that the third-level plot still has a basic priority contribution and is not completely ignored due to its low level, while the upper limit value of 3.0, combined with the subsequent priority weight coefficient, can produce significant differentiation, allowing the path nodes of important plots to gain a clear advantage in fitness assessment. Priority weight coefficient A single fixed value should not be used. Instead, a tiered system should be established based on the different priority levels and their corresponding sowing quality assurance requirements, reflecting the differentiated weighting adjustments between different levels. Specifically, for Level 1 plots, since the requirements for sowing accuracy and seedling uniformity are the highest, strong priority assurance is required. Therefore, a larger priority weighting coefficient should be adopted, such as... to This makes the seeding priority weight of the first-level cell... Reaching a score of 4.5 to 6.0, it dominates the profitability coefficient, guiding the path planning algorithm to prioritize the optimal route through the primary cell. For secondary cells, a medium-priority weight coefficient is used, such as... to This makes its seeding priority weight The priority weight should be between 1.6 and 2.4, maintaining a moderate balance with the boundary security weight. For Level 3 cells, a smaller priority weight coefficient should be used, such as... to This makes its seeding priority weight The value is between 0.3 and 0.6, far lower than the typical value of the boundary safety weight term, indicating that the seeding priority factor contributes little to the profit coefficient, and path planning focuses more on boundary safety; among which , , These represent the priority weight coefficients for level one, level two, and level three, respectively. Dependent variable Indicates the first The benefit coefficient of each path node directly reflects its overall performance in path planning. Its technical advantage lies in providing a node-level quantitative evaluation basis for subsequent fitness coefficient calculations, enabling the algorithm to distinguish the quality of different path nodes. The independent variables in the formula include boundary safety weights. Euclidean distance from the path node to the nearest experimental plot boundary Safe operating distance threshold and seeding priority weight Among them, the boundary security weight It itself is made of and It is derived from the hyperbolic tangent function, therefore Overall, this reflects the safety cost of path nodes due to their proximity to field boundaries, while This reflects the contribution of the importance of the cell where the node is located to the revenue. Regarding the influence relationship, the boundary security weight term... Distance from path node to field boundary There is a positive correlation, meaning that nodes farther from the field boundary receive a higher contribution to boundary security, guiding planting paths to prioritize areas far from the boundary; planting priority weights Priority level with the community There is a positive correlation, meaning that path nodes in higher priority cells receive higher revenue contributions, ensuring that seeding paths in important cells are prioritized for optimization.

[0027] In the above embodiments, 20 sets of data on the Euclidean distance from the path node to the nearest experimental plot boundary and the corresponding revenue coefficient of the path node are given to reflect the change of the revenue coefficient with the Euclidean distance, as shown in Table 1: Table 1: Relationship between Euclidean distance from path nodes to the nearest experimental plot boundary and the profit coefficient. In Table 1 above, a safe operating distance threshold is given. Rice, sowing priority weight It can be seen that Follow The value increases monotonically as it increases: when When rice is near the edge of the field, rice; when That is, when the safe operating distance threshold is reached, The growth rate of the return coefficient exhibits a characteristic of being rapid at first, then slowing down, and finally stabilizing: From 0.1 meters to 1.0 meters, It increased by approximately 0.23; From 1.0 meter to 2.0 meter range, It increased by approximately 0.55, with faster growth in later stages, due to the boundary security weight term. Follow This is due to the increased and accelerated growth; the data relationship shows that the farther the path node is from the field boundary, the higher the benefit coefficient, and the easier it is to be prioritized by the algorithm in path planning, thereby guiding the sowing path to actively move away from the field edge and improving the safety of operation.

[0028] The fitness coefficient is calculated as follows: Based on the changes in steering angle and speed of each path node relative to the previous path node, a steering penalty coefficient is constructed, the expression of which is as follows: in, Indicates the first The turning penalty coefficient for each path node, index It is an integer greater than or equal to 2; Indicates the first The direction and angle of each path node; Indicates the first The speed of each path node; Indicates the first The speed of each path node; , These represent the preset first and second penalty weight coefficients, respectively. ; In the above process, The value range is typically set from 0.5 to 2.0, used to constrain the change in direction angle between adjacent path nodes. The larger the value, the higher the algorithm's requirement for the straightness of the path, and the smoother the turning of the generated path; The smaller the value, the greater the directional fluctuation of the path is allowed, which is beneficial for exploring curved paths to adapt to complex field shapes. The value range is typically set to 1.0 to 4.0, and it must be ensured that... It is used to constrain the speed variation between adjacent path nodes. The larger the value, the higher the requirement for the algorithm's speed stability, in order to avoid affecting the uniformity of seeding due to drastic speed fluctuations. A smaller value allows for a wider range of speed variations, which is beneficial for flexibly adjusting speed in areas requiring deceleration, such as near field boundaries or at turns; in practical applications... and Stored as system preset parameters in the control system, the recommended typical values ​​are: , ; Dependent variable Indicates the first The turning penalty coefficient for each path node directly reflects the degree of change in turning angle and speed of that node relative to the previous node. The technical effect is to provide a node-level penalty basis for subsequent fitness coefficient calculations, ensuring that path nodes with excessively sharp turns or large speed fluctuations receive higher penalty values, thereby guiding the algorithm to prioritize smooth, stable seeding paths. The independent variables in the formula include the direction angle of the current node. , the direction and angle of the previous node The speed of the current node The speed of the previous node and the preset first penalty weight coefficient Second penalty weight coefficient Regarding the influencing relationship, the steering angle change term... The absolute values ​​of the direction angle difference show a square positive correlation, meaning that the greater the change in direction angle, the faster the penalty value increases, effectively suppressing sharp turning behavior; the speed change term... There is a linear positive correlation between the penalty value and the absolute value of the speed difference; that is, the greater the speed change, the more linearly the penalty value increases, thus suppressing drastic speed fluctuations. The velocity variation term has a higher weight in the penalty coefficient, which is reflected in the fact that the requirement for velocity stability is higher than that for directional smoothness in the breeding and sowing scenario.

[0029] Calculate the fitness coefficient for each path based on the reward coefficient and turning penalty coefficient of each path node: in, Indicates the first The number of path nodes in the path; Indicates the first The fitness coefficient of each path; Indicates the first The first path The turning penalty coefficient for each path node; Indicates the first The first path The revenue coefficient of each path node; For path indexing.

[0030] In the above process, this design unifies the two objectives of "maximizing benefits" and "minimizing costs" in path planning within a comparable numerical framework. On one hand, it considers the path from the first to the second... The revenue coefficient of all path nodes The overall benefit of the path is accumulated so that it reflects the sum of the boundary safety contribution and the planting priority contribution of each node. The larger the accumulated benefit coefficient, the better the path performs in terms of being far from the field boundary and prioritizing important plots. On the other hand, the path from the 2nd to the 3rd node is accumulated. Turning penalty coefficient for all path nodes The stress is accumulated so that the overall path penalty reflects the total cost of changes in direction and speed between adjacent nodes. A larger accumulated value of the steering penalty coefficient indicates poorer performance in terms of driving smoothness and speed stability. The fitness coefficient is calculated by subtracting the accumulated steering penalty coefficient from the accumulated benefit coefficient. It incorporates both positive contributions (benefits) and negative costs (penalties), ensuring that paths with high fitness coefficients possess both good safety and priority coverage, as well as smooth and stable driving characteristics. Furthermore, the benefit coefficient accumulates from the first node to the... The turning penalty coefficient accumulates from the second node to the third node. The difference in the node index range between the two nodes avoids the penalty calculation for the starting node, that is, the starting node has no previous node to compare, thus ensuring the mathematical rigor of the calculation formula. The maximum cumulative limit of the return coefficient is Instead The reason is that the last path node is the first... The first path node is typically the endpoint coordinate of the current target cell. At this node, the seeder has completed the seeding operation for the current cell and executed a stop-seeding operation. Therefore, this path node itself does not involve the actual seeding process, and its revenue coefficient should not be included in the overall revenue accumulation of the path. Similarly, the accumulation range of the turning penalty coefficient is from the second node to the first... The first path node cannot be calculated because the steering penalty coefficient requires comparing the direction angle and velocity changes between the current path node and the previous path node. Since the first node has no previous node for comparison, its steering penalty coefficient cannot be calculated. Meanwhile, the second... The node is used as the endpoint path node, although its relative to the first node can be calculated. The change in steering at each node occurs after sowing has stopped and does not affect the smoothness of driving during sowing; therefore, it should not be included in the penalty accumulation.

[0031] The iterative update strategy retains the current optimal path and reorganizes the remaining paths, specifically: Sort all candidate paths by fitness coefficient from largest to smallest, retain the top-ranked candidate paths in the next generation population and mark them as the optimal paths, and do not participate in subsequent recombination and mutation operations. In the above process, in practical applications, the initial preset number of candidate paths is often set to 5% to 20% of the preset population size. For example, when the preset population size is 100 paths, the initial preset number can be 5 to 20. For scenarios with a smaller population size, such as 20 to 50 paths, the proportion can be appropriately increased to 20% to ensure sufficient elite retention strength; for scenarios with a larger population size, such as 100 to 200 paths, the proportion can be appropriately reduced to 5% to 10% to maintain population diversity.

[0032] Two candidate paths are randomly selected from the remaining candidate paths. The candidate path with the highest fitness coefficient is selected to enter the next generation population. If the fitness coefficients of the two paths are the same, one of them is randomly selected to enter the next generation population. This step is repeated until the size of the next generation population reaches the preset population size. In the above process, in the scenario of small-plot breeding and sowing, the number of path nodes is usually between 10 and 50, and the search space is relatively limited. Therefore, the preset population size is generally set to 20 to 100 paths. For plots with larger areas and more than 30 path nodes, the population size can be appropriately increased to 100 to 200 paths to ensure sufficient search capability. For scenarios with smaller plots or limited computing resources, the size can be set to 20 to 50 paths to balance computational efficiency and path quality.

[0033] For paths that have reached the preset population size, random pairings are performed. For the two paired paths, since the path node sequence lengths of the two paths are the same, the index of a common path node is randomly selected as the intersection point, and the dynamic cross-entropy of the two paired paths is calculated: in, Represents dynamic cross-entropy; This represents the preset maximum cross-entropy; This represents the preset minimum cross-entropy; Indicates the first path in the parent path The direction and angle of each path node; Indicates the first generation in the parent generation path The direction and angle of each path node; This represents the index of a randomly selected path node; In the above process, the dependent variable Represents dynamic cross-entropy, the magnitude of which determines whether the parent and mother paths exchange subsequent path nodes at the crossover point. The technical effect is that the intensity of the crossover operation can be adaptively adjusted according to the local directional differences between the two paths at the crossover point, avoiding blind exchange or excessive destruction of excellent gene structures caused by using a fixed crossover probability. The independent variable in the formula includes the pre-defined minimum crossover probability. Preset maximum crossover probability Parent path The direction angle of each path node , parental path The direction angle of each path node ,in and The range of values ​​for dynamic cross-entropy is defined, and This reflects the degree of difference in direction and angle between the two paths at their intersection. Regarding the influence relationship, dynamic cross-entropy... Differences in direction and angle A negative correlation exists: when the difference in direction angle between two paths at their intersection is small. The value is relatively large. Approaching When two paths have similar travel directions near the node, exchanging subsequent nodes is more likely to produce beneficial gene combinations. Therefore, a higher crossover probability is used to promote gene exchange. When the directional angles differ significantly, The value is relatively small. Approaching At this point, the two paths have significantly different directions of travel. Directly swapping subsequent nodes may destroy their respective excellent path structures. Therefore, a lower crossover probability is used to protect the existing high-quality genes. generally The value is set between 0.8 and 0.95 to ensure a relatively high probability of intersection when the directional heights are consistent, but not to be fixed at 1 to avoid losing randomness; The default value is 0.4. This value is chosen because even if the two paths reach their maximum directional difference at the intersection point (i.e., their directions are opposite), the angle difference is close. radian, The value is close to -1, but after taking the absolute value in the formula... The value is close to 0, and the crossover probability is still maintained at the baseline level of 0.4, so as to retain a certain opportunity for gene exchange and prevent the population from falling into a local optimum too early.

[0034] When the dynamic cross-entropy is greater than the preset cross-entropy, the parent and mother paths are swapped from the first... For each path node, from the first path node to the last path node, generate two new child paths; otherwise, copy all path nodes of the parent and mother paths directly as child path nodes.

[0035] In the above process, firstly, the path with the highest fitness coefficient among the first predetermined number of paths is directly retained and marked as the optimal path, and it is not involved in subsequent operations. This ensures that the current optimal solution is not lost due to subsequent random operations, guaranteeing the monotonicity of the algorithm's convergence. Secondly, two paths are randomly selected from the remaining paths, and the one with the higher fitness coefficient is selected to enter the next generation. This is a tournament selection strategy, aiming to introduce a certain degree of randomness while favoring individuals with high fitness, maintaining competitive pressure in the population. Finally, for the two paired paths, since their node sequence lengths are the same, a common node index is randomly selected. As the intersection point, and based on the difference in directional angle between the parent and parent generations at this node. Calculate the dynamic crossover probability The reason for using the cosine function is that when the difference in direction angle between two paths at the intersection point is small, When the value is large, the crossover probability approaches 0. This promotes gene exchange to explore better combinations in similar pathways; when the directional angles differ significantly, When the value is small, the crossover probability approaches 0. This design protects the path structure with unique directional characteristics from being excessively destroyed. This dynamic crossover design enables the recombination operation to adaptively adjust the crossover strength based on the local similarity between paths, taking into account the algorithm's exploration and development capabilities. Ultimately, by selectively exchanging all nodes after the crossover point, two new offspring paths are generated. This not only preserves the prefix structure of the parent path but also introduces new gene combinations through suffix exchange, thereby gradually improving the overall fitness level of the population during the iteration process.

[0036] The remaining paths are mutated as follows: For each path in the next generation population obtained after recombination, calculate its path repetition rate and calculate the mutation probability based on the path repetition rate. When the mutation probability is greater than the preset mutation probability threshold, mutate all path nodes of that path. The path repetition rate is calculated as follows: The straight-line distances between all adjacent path nodes are summed to obtain the actual total length of the path. The actual total length is then divided by the straight-line distance from the starting point to the ending point of the path to obtain the tortuosity. The entire circumference is divided into several intervals. The directional change angles between all adjacent path nodes are calculated, and these angles are categorized into their corresponding intervals. The probability of each interval occurring is calculated. The probabilities of all intervals are multiplied by the logarithm of that probability, summed, and the result is the directional entropy. The tortuosity and directional entropy are normalized. The normalized tortuosity and directional entropy are multiplied by the tortuosity balance coefficient and the directional entropy balance coefficient, respectively, and summed to obtain a comprehensive index. This comprehensive index is then subtracted from the total to obtain the path repetition rate. In the above process, the calculation method for the path repetition rate quantifies the geometric uniqueness of a path by integrating two complementary indicators: tortuosity and directional entropy. Tortuosity reflects the overall curvature of the path; a straight path has a tortuosity close to 1, while a curved path has a tortuosity greater than 1. Directional entropy reflects the richness of directional changes along the path; paths with a single directional change have low directional entropy, while paths with a rich variety of directional changes have high directional entropy. After normalizing the tortuosity, the normalized value for a straight path is close to 0, while the normalized value for a curved path is close to 1. After normalizing the directional entropy, the normalized value for a path with a single direction is close to 0, while the normalized value for a path with a rich variety of directions is close to 1. The path repetition rate is obtained by weighted summing the normalized tortuosity and directional entropy and then subtracting this comprehensive index from 1. The core logic is that: a straight path with a single direction has a comprehensive index close to 0 and a path repetition rate close to 1, indicating that the path shape is monotonous and easily repeats with other paths, requiring a higher mutation probability to increase population diversity; a curved path with rich directions has a comprehensive index close to 1 and a path repetition rate close to 0, indicating that the path shape is unique and contributes greatly to diversity, and its excellent structure should be preserved and given a lower mutation probability.

[0037] Calculate the probability of variation based on path repetition rate: in, This represents the preset base variation rate; This represents the coefficient of influence of repetition rate; Indicates the first The path repetition rate of each path; Indicates the first The probability of mutation of each path; In the above process, the path repetition rate is... Probability of mutation when converted to this path The core logic is as follows: A path with a higher path repetition rate indicates a monotonous shape and greater similarity to other paths in the population, resulting in a lower contribution to population diversity. Therefore, a higher mutation probability is needed to enhance the random perturbation of this path, thereby exploring new solution spaces and increasing population diversity. Conversely, a path with a lower path repetition rate indicates a unique shape and a greater contribution to diversity, requiring a lower mutation probability to preserve its superior structure. Basic mutation rate This ensures that even paths with a zero repetition rate have a basic chance of mutation, preventing the algorithm from getting trapped in local optima too early; the repetition rate influence coefficient. It determines the moderating power of path repetition rate on mutation probability, and its magnitude reflects the importance attached to maintaining population diversity. The determination method usually adopts empirical calibration. The empirical calibration method involves conducting preliminary experiments under typical field conditions, selecting multiple groups of different... Values ​​such as 0.2, 0.5, 0.8, and 1.0 were used to observe the algorithm's convergence speed and the fitness coefficient of the output path, and the algorithm with the best overall performance was selected. As a default value.

[0038] When the mutation probability is greater than the preset mutation probability, all path nodes of the path are mutated: in, This indicates that the mean is 0 and the standard deviation is 0. Gaussian distributed random numbers; This indicates the preset maximum speed of the seed metering device; This indicates the preset minimum speed of the seed metering device; Indicates the mutated first The direction and angle of each path node; Represents the mutated path node speed; Represents the path node before mutation The speed.

[0039] In the above process, the mutation calculation uses Gaussian perturbation to independently and randomly adjust the direction angle and velocity of the path nodes. The core reason for this is that Gaussian distribution... It can generate a symmetrical and smooth random perturbation centered on the current value within a standard deviation of 0.08. This ensures that the mutated value has approximately a 68% probability of falling within the range of ±0.08 of the original value to retain its original desirable characteristics, while also having approximately a 32% probability of deviating significantly to explore new solution spaces. This achieves a balance between fine-grained local search and global exploration. (Regarding direction and angle...) The mutation is directly superimposed with Gaussian random numbers because the range of change in direction and angle is relatively fixed, requiring no additional boundary constraints. For velocity... The mutation process, after superimposing Gaussian random numbers, adds a segmented pruning process: when the mutated speed is lower than the preset minimum speed... Time to take When the speed exceeds the preset maximum speed Time to take Otherwise, the mutated speed value is maintained; this boundary pruning design ensures that the mutated speed always falls within the controllable physical range of the seeder, avoiding the generation of unexecutable speed values. Furthermore, the algorithm mutates all path nodes simultaneously, enabling a holistic adjustment of the entire path's driving characteristics, resulting in a systematic change in the path's direction and speed rhythm. The settings are based on several factors, including: firstly, the minimum stable operating speed of the seed metering device. Too low a speed may cause intermittent jamming or uneven seed distribution; typically, it is set to 0.3 m / s to 0.5 m / s. The value ranges from 1.2 m / s to 2.0 m / s, and can be appropriately relaxed to 2.0 m / s for routine seed propagation experiments.

[0040] S3: Perform sowing operations based on the predetermined sowing path. During the movement, continuously acquire the current position coordinates of the sowing machine, compare the deviation between the current position coordinates and the predetermined sowing path, and perform path correction. Continuously compare the distance between the current position coordinates and the endpoint coordinates of the current target cell. When the current position coordinates enter the allowable error range of the endpoint coordinates, the sowing of the current target cell is stopped. The current location coordinates are compared with the predetermined sowing path to determine the deviation and path correction is performed. Specifically: The preset number of paths and the preset remaining number of paths are merged to form a new generation population. The fitness coefficient of each path in the new generation population is calculated. When the difference between the maximum fitness coefficient in the new generation population and the maximum fitness coefficient of the previous generation population is less than a preset difference, the path corresponding to the maximum fitness coefficient in the current new generation population is selected. Combined with the starting point of the seeder, the path nodes in the path are converted into actual sowing coordinates to obtain the predetermined sowing path; otherwise, the fitness coefficient is recalculated and the population is updated. In the above process, after each iteration, the path with the highest fitness coefficient in the previous generation is compared with the path with the highest fitness coefficient in the new generation. When the difference between the two is less than a preset difference, it indicates that the population has become unlikely to generate a better new path after multiple generations of iteration, and the algorithm has converged to a stable solution. At this time, the path corresponding to the maximum fitness coefficient in the current new generation is the optimal solution found. This solution is then combined with the starting point of the seeder and converted into actual seeding coordinates as the predetermined seeding path output. Conversely, when the difference in the maximum fitness coefficients is greater than or equal to the preset difference, it indicates that the population is still evolving and there is still a possibility of generating a better path. Therefore, no result is output, and the population continues to be updated and iterated. This approach avoids the waste of computational resources caused by unlimited iteration of the algorithm and also prevents the premature output of suboptimal paths before sufficient convergence, ensuring that the output predetermined seeding path has a high fitness coefficient and quality. In practical applications, the preset difference is usually set using a relative value method, that is, it is set to 0.1% to 1% of the maximum fitness coefficient in the previous generation population. For example, when the maximum fitness coefficient of the previous generation is 100, the preset difference can be set between 0.1 and 1. This relative value setting method can adapt to fitness coefficients of different magnitudes and has good versatility.

[0041] The seeding operation is performed based on a predetermined seeding path. During the movement, the current position coordinates of the seeder are continuously acquired. The index of the path node closest to the current position is searched on the predetermined seeding path. The lateral and longitudinal deviations are calculated. Based on the magnitude of the lateral deviation, the path nodes are dynamically adjusted in the following manner: When the lateral deviation is less than the first preset threshold, the original predetermined sowing path remains unchanged; when the lateral deviation is greater than or equal to the first preset threshold and less than the second preset threshold, a cubic Bézier curve connecting the start and end points is generated, with the current position as the starting point and the path node at the preset distance ahead on the predetermined sowing path as the end point, and the path node between the start and end points in the original predetermined sowing path is replaced by the path node on the Bézier curve; when the lateral deviation is greater than or equal to the second preset threshold, S2 is re-executed with the current position of the seeder as the new starting point and the end point of the original predetermined sowing path as the end point.

[0042] In the above process, the core of the graded path correction method lies in adopting differentiated correction strategies based on the magnitude of the lateral deviation to balance the relationship between correction accuracy, computational cost, and driving stability. When the lateral deviation is less than the first preset threshold, it indicates that the deviation between the current position of the seeder and the predetermined sowing path is small and within an acceptable range, requiring no correction operation. The original predetermined sowing path is maintained to preserve the continuity and stability of the journey. When the lateral deviation is greater than or equal to the first preset threshold but less than the second preset threshold, it indicates a moderate degree of deviation, requiring local path correction. In this case, a cubic Bézier curve is generated, connecting the start and end points, with the path node at a preset distance ahead on the predetermined sowing path as the end point. The path node discretized from this Bézier curve replaces the portion between the start and end points of the original predetermined sowing path. The advantage of using cubic Bézier curves is that they can generate a smooth and continuously curvatured transition path, allowing the seeder to smoothly return from the deviation point to the original predetermined sowing path. Simultaneously, the corrected path's direction at the start and end points naturally connects with the original path, avoiding sharp turns. When the lateral deviation is greater than or equal to the second preset threshold, it indicates that the seeder has seriously deviated from the predetermined sowing path. At this point, local correction is indeed difficult to effectively revert, or the reverted path will be too tortuous, affecting sowing quality. Therefore, a global replanning strategy is adopted. Using the seeder's current position as the new starting point and the end point of the original predetermined sowing path as the endpoint, the path planning algorithm in S2 is re-executed to generate a completely new remaining sowing path, replacing the portion of the original predetermined sowing path after the current position. This hierarchical processing approach avoids the waste of computational resources caused by using complex global replanning in any deviation case, and also prevents unnatural paths or reduced seeding quality caused by only making local corrections in the case of large deviations, thus achieving adaptive matching between the deviation correction strategy and the degree of deviation. The first preset threshold is typically set to 2 to 3 times the seeder's positioning accuracy. For example, when using RTK positioning accuracy of ±2.5 cm, the first preset threshold can be set to 5 to 8 cm to ensure that the system does not frequently trigger unnecessary correction operations due to positioning noise. The second preset threshold represents the critical value for severe deviation. When the lateral deviation is greater than or equal to this threshold, it is considered that local correction can no longer effectively regress or the regression path will be too tortuous, requiring a global replanning. The second preset threshold is typically set to 0.3 to 0.5 times the seeding row spacing. For example, when the row spacing is 30 cm, the second preset threshold can be set to 9 to 15 cm, because exceeding this range means that the seeder has significantly deviated from the current row, may intrude into adjacent plots, or pose a serious risk of missed seeding. A buffer zone is formed between the first and second preset thresholds. Within this interval, Bézier curves are used for smooth regression correction, which avoids the computational overhead of frequent replanning while ensuring the ability to replan globally when there is a large deviation.

[0043] S4: After the current target cell is stopped, select the next target cell in the preset numbering order and repeat S1 to S3 until all cells in the experimental field are sown.

[0044] Select the next target cell according to the preset numbering order, specifically: After the seeding in the current target cell is stopped, as the seeder moves away from the current cell and towards the next cell, the distance between the current position of the seeder and the starting coordinates of the next cell is continuously calculated. When the distance is less than the preset distance, the control system automatically confirms that the seed metering device has reached the starting point of the next cell, sets the starting coordinates and ending coordinates of the next cell as the new comparison targets, and repeats steps S1-S3 until the seeding of all cells is completed.

[0045] In the above process, after the current target cell stops sowing, the seeder needs to leave the current cell and head to the starting point of the next cell. At this time, the control system continuously calculates the Euclidean distance between the current position of the seeder and the coordinates of the starting point of the next cell. When this distance is less than the preset distance, the system automatically confirms that the seed metering device has arrived at the starting point of the next cell. Without manual intervention or additional button confirmation, the system sets the starting and ending coordinates of the next target cell as the new comparison targets and repeats steps S1 to S4. The advantages of this design are: on the one hand, through continuous distance monitoring and automatic confirmation mechanisms, the time for manual judgment and operation is saved, improving work efficiency; on the other hand, the preset distance setting ensures that the seed metering device will not start sowing prematurely before reaching the starting point, while also allowing for a certain positioning error and vehicle inertia, avoiding false triggering caused by positioning signal fluctuations; the preset distance is usually set to 0.3 meters to 0.5 meters, which can complete the loading and path planning preparation of the target cell in advance without causing the sowing starting position to shift due to excessive advance.

[0046] Please see Figure 3 The present invention also provides a seeding path control system based on plot breeding, wherein the seeding path control system based on plot breeding is used to execute the above-described seeding path control method based on plot breeding, comprising: Data acquisition module: It is used to extract the starting point coordinates, ending point coordinates and standard row direction of the current target cell according to the cell number order in the experimental field. With the boundary of the target cell as a constraint, it generates a path node sequence between the starting point coordinates and the ending point coordinates at a fixed sampling interval. It also constructs several candidate paths around the standard row direction and sets the direction angle and speed of each path node of each candidate path to obtain the candidate path population. Update Calculation Module: For each path in the candidate path population, calculate the benefit coefficient of each path node according to the set direction angle and speed. Based on the benefit coefficient of the path node, further calculate the fitness coefficient of each path and iteratively update the candidate path population. The iterative update strategy adopts the following approach: retain the current best path, reorganize and mutate the remaining paths. After reaching the preset iterative update termination condition, output the path with the highest fitness coefficient as the predetermined seeding path. The comparison and sowing module is used to perform sowing operations based on a predetermined sowing path. During the movement, it continuously acquires the current position coordinates of the sowing machine, compares the current position coordinates with the predetermined sowing path to perform deviation correction, and continuously compares the current position coordinates with the destination coordinates of the current target cell. When the current position coordinates enter the allowable error range of the destination coordinates, the sowing of the current target cell is stopped. The sowing completion module is used to select the next target cell in a preset numbered order after the current target cell has been stopped from sowing, and repeat the data acquisition module to the comparison sowing module until all cells in the experimental field have been sown.

[0047] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0049] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0050] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for controlling the sowing path based on plot breeding, characterized in that, include: S1: Extract the starting point coordinates, ending point coordinates, and standard row direction of the current target cell according to the cell number order in the experimental field. With the boundary of the target cell as a constraint, generate a path node sequence between the starting point coordinates and the ending point coordinates at a fixed sampling interval. Construct several candidate paths around the standard row direction and set the direction angle and speed of each path node of each candidate path to obtain the candidate path population. S2: For each path in the candidate path population, calculate the benefit coefficient of each path node according to the set direction angle and speed. Further calculate the fitness coefficient of each path based on the benefit coefficient of the path node, and iteratively update the candidate path population. The iterative update strategy adopts the method of retaining the current best path and reorganizing and mutating the remaining paths. After reaching the preset iteration termination condition, output the path with the highest fitness coefficient as the predetermined seeding path. S3: Perform sowing operations based on the predetermined sowing path. During the movement, continuously acquire the current position coordinates of the sowing machine, compare the deviation between the current position coordinates and the predetermined sowing path, and perform path correction. Continuously compare the distance between the current position coordinates and the endpoint coordinates of the current target cell. When the current position coordinates enter the allowable error range of the endpoint coordinates, the sowing of the current target cell is stopped. S4: After the current target cell is stopped, select the next target cell in the preset numbering order and repeat S1 to S3 until all cells in the experimental field are sown.

2. The sowing path control method based on plot breeding according to claim 1, characterized in that, The path node sequence is generated discretely between the starting coordinates and the ending coordinates at a fixed sampling interval, specifically as follows: Using the starting and ending coordinates of the current target cell as endpoints and the boundary of the target cell as constraints, equidistant sampling is performed on the line segment between the starting and ending points at a fixed sampling interval to generate a series of path nodes that constitute a path node sequence.

3. The sowing path control method based on plot breeding according to claim 1, characterized in that, Several candidate paths are constructed around the standard row direction, specifically: Using the standard row direction of the current target cell as the reference direction, a random direction offset following a normal distribution is independently generated for each node in the path node sequence. The mean of this offset is zero and the standard deviation is the preset direction variation intensity. The obtained random offset is superimposed on the standard row direction to obtain the actual direction angle of each path node. Within the preset minimum and maximum speed range, all nodes on this path are randomly assigned the same speed value as the driving speed of the entire path. The path generation process is repeated continuously, generating a complete candidate path each time, until the number of generated candidate paths reaches the preset population size. All candidate paths are then aggregated to form a candidate path population.

4. The sowing path control method based on plot breeding according to claim 1, characterized in that, The revenue coefficient of a path node is calculated in the following way: For each candidate path in the candidate path population, obtain the Euclidean distance from each path node to the nearest experimental field boundary, and calculate the boundary safety weight of the path node based on the Euclidean distance: in, Indicates the first Boundary security weights of each path node; This indicates the preset safe operating distance threshold; Indicates the first Euclidean distance from each path node to the boundary of the nearest experimental plot; The index of the path node; Get the Priority of each path node in the cell Calculate the seeding priority weight: in, Represents path nodes The seeding priority weight; Represents path nodes The priority level of the residential community; Indicates the priority weight coefficient; Based on the boundary safety weight and seeding priority weight of each path node, calculate its corresponding profit coefficient: in, Represents path nodes The profit coefficient.

5. The sowing path control method based on plot breeding according to claim 4, characterized in that, The fitness coefficient is calculated as follows: Based on the changes in steering angle and speed of each path node relative to the previous path node, a steering penalty coefficient is constructed, the expression of which is as follows: in, Indicates the first The turning penalty coefficient for each path node, index It is an integer greater than or equal to 2; Indicates the first The direction and angle of each path node; Indicates the first The speed of each path node; Indicates the first The speed of each path node; , These represent the preset first and second penalty weight coefficients, respectively. ; Calculate the fitness coefficient for each path based on the reward coefficient and turning penalty coefficient of each path node: in, Indicates the first The number of path nodes in the path; Indicates the first The fitness coefficient of each path; Indicates the first The first path The turning penalty coefficient for each path node; Indicates the first The first path The revenue coefficient of each path node; For path indexing.

6. The sowing path control method based on plot breeding according to claim 5, characterized in that, The iterative update strategy retains the current optimal path and reorganizes the remaining paths, specifically: Sort all candidate paths by fitness coefficient from largest to smallest, retain the top-ranked candidate paths in the next generation population and mark them as the optimal paths, and do not participate in subsequent recombination and mutation operations. Two candidate paths are randomly selected from the remaining candidate paths. The candidate path with the highest fitness coefficient is selected to enter the next generation population. This step is repeated until the size of the next generation population reaches the preset population size. For paths that have reached the preset population size, random pairings are performed. For the two paired paths, since the path node sequence lengths of the two paths are the same, the index of a common path node is randomly selected as the intersection point, and the dynamic cross-entropy of the two paired paths is calculated: in, Represents dynamic cross-entropy; This represents the preset maximum cross-entropy; This represents the preset minimum cross-entropy; Indicates the first path in the parent path The direction and angle of each path node; Indicates the first generation in the parent generation path The direction and angle of each path node; This represents the index of a randomly selected path node; When the dynamic cross-entropy is greater than the preset cross-entropy, the parent and mother paths are swapped from the first... For each path node, from the first path node to the last path node, generate two new child paths; otherwise, copy all path nodes of the parent and mother paths directly as child path nodes.

7. The sowing path control method based on plot breeding according to claim 6, characterized in that, The remaining paths are mutated as follows: For each path in the next generation population obtained after recombination, calculate its path repetition rate and calculate the mutation probability based on the path repetition rate. When the mutation probability is greater than the preset mutation probability threshold, mutate all path nodes of that path. The path repetition rate is calculated as follows: The straight-line distances between all adjacent path nodes are summed to obtain the actual total length of the path. The actual total length is then divided by the straight-line distance from the starting point to the ending point of the path to obtain the tortuosity. The entire circumference is divided into several intervals. The directional change angles between all adjacent path nodes are calculated, and these angles are categorized into their corresponding intervals. The probability of each interval occurring is calculated. The probabilities of all intervals are multiplied by the logarithm of that probability, summed, and the result is the directional entropy. The tortuosity and directional entropy are normalized. The normalized tortuosity and directional entropy are multiplied by the tortuosity balance coefficient and the directional entropy balance coefficient, respectively, and summed to obtain a comprehensive index. This comprehensive index is then subtracted from the total to obtain the path repetition rate. Calculate the probability of variation based on path repetition rate: in, This represents the preset base variation rate; This represents the coefficient of influence of repetition rate; Indicates the first The path repetition rate of each path; Indicates the first The probability of mutation of each path; When the mutation probability is greater than the preset mutation probability, all path nodes of the path are mutated: in, This indicates that the mean is 0 and the standard deviation is 0. Gaussian distributed random numbers; This indicates the preset maximum speed of the seed metering device; This indicates the preset minimum speed of the seed metering device; Indicates the mutated first The direction and angle of each path node; Represents the mutated path node speed; Represents the path node before mutation The speed.

8. The sowing path control method based on plot breeding according to claim 1, characterized in that, The current location coordinates are compared with the predetermined sowing path to determine the deviation and path correction is performed. Specifically: The preset number of paths and the preset remaining number of paths are merged to form a new generation population. The fitness coefficient of each path in the new generation population is calculated. When the difference between the maximum fitness coefficient in the new generation population and the maximum fitness coefficient of the previous generation population is less than the preset difference, the path corresponding to the maximum fitness coefficient in the current new generation population is selected. Combined with the starting point of the seeder, the path nodes in the path are converted into actual sowing coordinates to obtain the predetermined sowing path. Otherwise, recalculate the fitness coefficient and update the population; The seeding operation is performed based on a predetermined seeding path. During the movement, the current position coordinates of the seeder are continuously acquired. The index of the path node closest to the current position is searched on the predetermined seeding path. The lateral and longitudinal deviations are calculated. Based on the magnitude of the lateral deviation, the path nodes are dynamically adjusted in the following manner: When the lateral deviation is less than the first preset threshold, the original predetermined sowing path remains unchanged; when the lateral deviation is greater than or equal to the first preset threshold and less than the second preset threshold, a cubic Bézier curve connecting the start and end points is generated, with the current position as the starting point and the path node at the preset distance ahead on the predetermined sowing path as the end point, and the path node between the start and end points in the original predetermined sowing path is replaced by the path node on the Bézier curve; when the lateral deviation is greater than or equal to the second preset threshold, S2 is re-executed with the current position of the seeder as the new starting point and the end point of the original predetermined sowing path as the end point.

9. The sowing path control method based on plot breeding according to claim 1, characterized in that, Select the next target cell according to the preset numbering order, specifically: After the seeding in the current target cell is stopped, as the seeder moves away from the current cell and towards the next cell, the distance between the current position of the seeder and the starting coordinates of the next cell is continuously calculated. When the distance is less than the preset distance, the control system automatically confirms that the seed metering device has reached the starting point of the next cell, sets the starting coordinates and ending coordinates of the next cell as the new comparison targets, and repeats steps S1-S3 until the seeding of all cells is completed.

10. A seeding path control system based on plot breeding, wherein the seeding path control system based on plot breeding is used to implement the seeding path control method based on plot breeding according to any one of claims 1-9, characterized in that: Data acquisition module: It is used to extract the starting point coordinates, ending point coordinates and standard row direction of the current target cell according to the cell number order in the experimental field. With the boundary of the target cell as a constraint, it generates a path node sequence between the starting point coordinates and the ending point coordinates at a fixed sampling interval. It also constructs several candidate paths around the standard row direction and sets the direction angle and speed of each path node of each candidate path to obtain the candidate path population. Update Calculation Module: For each path in the candidate path population, calculate the benefit coefficient of each path node according to the set direction angle and speed. Based on the benefit coefficient of the path node, further calculate the fitness coefficient of each path and iteratively update the candidate path population. The iterative update strategy adopts the following approach: retain the current best path, reorganize and mutate the remaining paths. After reaching the preset iterative update termination condition, output the path with the highest fitness coefficient as the predetermined seeding path. The comparison and sowing module is used to perform sowing operations based on a predetermined sowing path. During the movement, it continuously acquires the current position coordinates of the sowing machine, compares the current position coordinates with the predetermined sowing path to perform deviation correction, and continuously compares the current position coordinates with the destination coordinates of the current target cell. When the current position coordinates enter the allowable error range of the destination coordinates, the sowing of the current target cell is stopped. The sowing completion module is used to select the next target cell in a preset numbered order after the current target cell has been sown, and repeat the data acquisition module to the comparison sowing module until all cells in the experimental field have been sown.

Images