Agricultural unmanned aerial vehicle path planning method based on rasterized map and related equipment

Abstract

The invention discloses an agricultural unmanned aerial vehicle path planning method based on a rasterized map and related equipment. The method comprises the following steps: acquiring boundary information of a farmland area to be operated, and performing digital rasterization processing on the farmland area to generate a raster map; performing area division on the farmland area based on the grid map to form a main operation area and turning-around areas located at the two ends of the main operation area; and receiving a starting point and a starting advancing direction specified by a user, generating a bow-shaped reciprocating main operation path covering the main operation area in the main operation area by taking the starting point and the starting advancing direction as path generation constraints, and generating a bow-shaped reciprocating operation path covering the space range of the U-turn area in the space range of the U-turn area. The problems that an existing full-coverage path planning algorithm mainly pursues that a path does not repeatedly sweep the whole area, but the general design cannot meet the special working condition of an agricultural unmanned aerial vehicle in the ploughing operation can be solved.

Description

Technical Field

[0001] This application relates to the field of computers, and more specifically, to a method and related equipment for path planning of agricultural drones based on gridded maps. Background Technology

[0002] Existing full-coverage path planning algorithms primarily aim to ensure that the path sweeps across the entire area without repetition. However, their general design cannot meet the special working conditions of agricultural drones in plowing operations. First, the starting point and direction of the operation need to be flexibly specified by the farmer based on the actual conditions of the field, rather than being fixed by the algorithm. Second, the agricultural machine must reserve turning areas at both ends of the field and use a special fishtail-style turning maneuver to lift the plow and turn, during which plowing is impossible. Finally, due to the existence of the turning area, uncultivated blank areas will be left at the edge of the field. Summary of the Invention

[0003] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0004] To address the issue that existing full-coverage path planning algorithms primarily aim to scan the entire area without repetition, but their general design cannot meet the special working conditions of agricultural drones in plowing operations, this invention proposes, firstly, a path planning method for agricultural drones based on gridded maps, for plowing operations of agricultural drones or autonomous agricultural machinery. The method includes: Obtain the boundary information of the farmland area to be operated, and digitize the farmland area into a raster to generate a raster map, where each raster corresponds to a small arable unit; The farmland area is divided into regions based on the grid map to form a main work area and turning areas located at both ends of the main work area, and the spatial range of the main work area and the turning areas is marked in the grid map. The system receives the user-specified starting point and starting direction of travel, and uses the starting point and starting direction of travel as path generation constraints to generate a bow-shaped reciprocating main work path covering the main work area, and generates a bow-shaped reciprocating work path covering the space of the U-turn area.

[0005] Optional, also includes: For the end of each work row in the bow-shaped reciprocating main work path, a fishtail-shaped turning path is generated in the corresponding turning area, and the fishtail-shaped turning path is automatically connected to the end of the work row to guide the execution of non-tillage turning actions such as lifting the plow, turning around, aligning with the next work row, and lowering the plow. Based on the grid map, the system automatically identifies uncultivated blank areas formed near the boundary of the main working area due to the existence of the turning area, and generates a horizontally reciprocating edge-closing path for the blank areas to achieve coverage of the blank areas. The bow-shaped reciprocating main operation path, each of the fishtail-shaped turning paths, and the edge-closing path are integrated to generate a seamless final operation path, which is then output to the agricultural drone or the autonomous agricultural machine, so that the agricultural drone or the autonomous agricultural machine can complete the automatic plowing operation according to the final operation path.

[0006] Optionally, the step of dividing the farmland area based on the raster map includes: Based on preset turning width parameters, plow operation width parameters, and field safety margin parameters, turning zone boundaries are generated at opposite ends of the main working area along the forward and reverse directions of the starting travel direction, respectively. The grids within the turning zone boundaries are marked as turning zone grids, and the remaining arable grids are marked as main working area grids, so as to restrict tillage and covering to be carried out only within the main working area grids.

[0007] Optionally, generating a bow-shaped reciprocating main work path covering the main work area includes: The normal division direction of the main working row is determined by the starting direction of travel, and the main working area is divided into parallel rows at equal intervals according to the working width of the plow. The first work row and its travel endpoint are determined based on the projection position of the starting point in the row, and a reciprocating sequence is generated according to the rule of alternating reverse of adjacent work rows, so that the bow-shaped reciprocating main work path covers all main work grids in the main work area while satisfying the constraints of the starting point and the starting travel direction.

[0008] Optionally, generating a fishtail-shaped U-turn path within the corresponding U-turn area includes: When the plowing trigger point is reached at the end of the work row, the path geometric constraints of the first deflection segment, the swing segment and the alignment segment are generated in sequence in the turning area, so that the heading angle smoothly transitions from the current work row direction to the next work row direction. A plowing trigger point is generated at the end of the alignment section, so that the agricultural drone or the autonomous agricultural machine enters the tillage state of the next work row at the plowing trigger point, thereby reducing the non-tillage time during the turning process and reducing the trajectory deviation caused by the sudden change of heading.

[0009] Optionally, generating a horizontally reciprocating edge-finishing path for the blank area includes: In the grid map, compare the tillage coverage record of the final operation path with the grid set of the main operation area, extract the uncovered blank area grid and perform connected component segmentation on the blank area grid; For each connected domain, the tackling direction is determined along its long side, and a horizontally reciprocating parallel tackling line sequence is generated within the blank area, so that the spacing between adjacent tackling lines matches the working width of the plow. The tackling path is then added to the end of the final working path to complete the coverage of the blank area.

[0010] Secondly, the present invention also proposes a path planning device for agricultural drones based on gridded maps, used for plowing operations of agricultural drones or autonomous agricultural machinery, the device comprising: The acquisition unit is used to acquire the boundary information of the farmland area to be operated, and to digitize and rasterize the farmland area to generate a raster map, wherein each raster corresponds to a small arable unit. A division unit is used to divide the farmland area based on the grid map to form a main working area and a turning area located at both ends of the main working area, and to identify the spatial range of the main working area and the turning area in the grid map; The planning unit is used to receive the starting point and starting direction of travel specified by the user, and use the starting point and starting direction of travel as path generation constraints to generate a bow-shaped reciprocating main operation path covering the main operation area, and generate a bow-shaped reciprocating operation path covering the space range of the U-turn area.

[0011] Thirdly, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the agricultural unmanned aerial vehicle path planning method based on a raster map as described in any of the first aspects above.

[0012] Fourthly, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the agricultural unmanned aerial vehicle path planning method based on a raster map according to any one of the preceding claims of the first aspect.

[0013] In summary, the agricultural drone path planning method based on a raster map proposed in this application obtains the boundary information of the farmland area to be operated, and digitizes the farmland area into a raster map, where each raster corresponds to a small arable unit. Based on the raster map, the farmland area is divided into a main operating area and turning areas located at both ends of the main operating area, and the spatial range of the main operating area and the turning areas is marked in the raster map. The method receives the user-specified starting point and starting direction of travel, and uses these as path generation constraints to generate a bow-shaped reciprocating main operating path covering the main operating area, and a bow-shaped reciprocating operating path covering the spatial range of the turning areas. Thus, the continuous geometric space of farmland is transformed into discrete raster units, and the coverage planning is transformed from a geometrically continuous problem into a problem of traversing and serializing a set of raster cells. Building upon the discrete representation, the system further models regions with different operational constraints separately through region partitioning. The main operational area emphasizes high efficiency and row coverage, while the turning area emphasizes accessibility, turning space, and the feasibility of non-deep tillage actions. Subsequently, user-specified starting points and directions are introduced as hard constraints to ensure that the order and direction of the generated first operational row and subsequent reciprocating rows closely match the actual entrance location and plot boundary constraints. Finally, a bow-shaped reciprocating walking pattern is adopted, essentially equivalent to alternating the direction of parallel operational rows, allowing adjacent rows to transition with only short lateral movements. This achieves high time efficiency and low trajectory curvature abrupt changes in most tillage conditions. The first row selection and row segmentation mechanism allows for the direct generation of the first executable path even in common situations such as the entrance not being in a corner, the entrance direction being restricted by roads, or the starting point being in the middle of the boundary, reducing manual guidance and temporary adjustments. The system not only generates paths but also verifies which grids are not covered after execution, providing direct data support for subsequent edge finishing, re-tilling, or continuation of operations after abnormal interruptions. By explicitly defining the turning area and generating regular reciprocating channels, the spatial requirements at the field edge are pre-defined, reducing boundary crossings, plow dragging, or multiple reversing corrections caused by insufficient turning space. The turning area path segment can be switched via operation mode commands to achieve non-deep tillage passage or shallow operation coverage, making this solution suitable for both typical plowing and certain complex operation scenarios such as drone field inspection, field edge spraying, and compaction. Attached Figure Description

[0014] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1A schematic flowchart of a path planning method for agricultural drones based on a raster map is provided for an embodiment of this application; Figure 2 A schematic diagram of a path planning device for agricultural drones based on a gridded map, provided in an embodiment of this application; Figure 3 This is a schematic diagram of an electronic device for path planning of agricultural drones based on a gridded map, provided in an embodiment of this application. Detailed Implementation

[0015] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will now be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0016] To address the issue that existing full-coverage path planning algorithms primarily aim to ensure the path traverses the entire area without repetition, but their general design cannot meet the specific working conditions of agricultural drones in plowing operations, please refer to [the relevant documentation / reference]. Figure 1 This is a schematic flowchart of a path planning method for agricultural drones based on a gridded map provided in an embodiment of this application, which may specifically include steps S110 to S130.

[0017] S110, Obtain the boundary information of the farmland area to be operated, and digitize the farmland area into a raster to generate a raster map, wherein each raster corresponds to a small arable unit.

[0018] S120, the farmland area is divided into regions based on the grid map to form a main work area and turning areas located at both ends of the main work area, and the spatial range of the main work area and the turning areas is marked in the grid map.

[0019] S130: Receive the starting point and starting direction of travel specified by the user, and use the starting point and starting direction of travel as path generation constraints to generate a bow-shaped reciprocating main operation path covering the main operation area, and generate a bow-shaped reciprocating operation path covering the space range of the U-turn area.

[0020] For example, the boundary information of the farmland area to be cultivated is first obtained through agricultural machinery terminals or agricultural drones. This boundary information can originate from polygon coordinates in the land parcel archive system, manual delineation before operation, or a sequence of boundary points collected by the positioning module during equipment patrol around the field. To ensure stable, calculable, and easy-to-determine path generation, this solution transforms the boundary geometry of the continuous space into a discrete raster map. The continuous area determination of whether a certain area has been cultivated is converted into a discrete marker determination of which raster units are covered by the trajectory sweep, thereby reducing the sensitivity of boundary irregularities, positioning jitter, and soil resistance disturbances to coverage determination. Specifically, a local planar coordinate system consistent with the land parcel coordinates is first established, the bounding rectangle of the farmland boundary is calculated, and the rectangle is divided into grids according to a preset raster resolution. The raster resolution can be determined based on the plow's working width, positioning accuracy, and control response capability. For example, when the effective working width of the plow is on the order of one to two meters, the raster side length can be selected as the width. Smaller gradations allow for more precise boundary approximation, while reserving margin for coverage deviations caused by positioning errors and tillage depth disturbances. Then, arable determination is performed on each grid cell. For example, the arable criterion is that the grid center point falls within the farmland polygon, and grids outside the boundary are marked as unarable. If there are obstacles such as ditches, wells, or utility poles in the field, the corresponding grids are further marked as disabled grids so that they can be automatically truncated to avoid trajectory crossings when generating subsequent work rows. Through the above rasterization process, complex boundaries and obstacle distributions can be solidified onto a unified discrete map carrier, transforming intersection, truncation, and coverage statistics during path planning into calculations and label updates of the grid set. This improves the determinism and real-time maintainability of coverage determination. For example, in plots with arc-shaped or sawtooth boundaries, the endpoints of the rasterized work rows can stably fall near the last column of arable grid cells, without endpoint jumps caused by boundary point sampling jitter, thus reducing the risk of frequent corrections and boundary crossings during actual execution.

[0021] For example, after generating a raster map, the farmland area is divided into a main work area and turning areas at both ends of the main work area. Plowing operations involve two types of behaviors with different constraints: tillage covering and turning alignment at the field edge. If the turning space is not separated from the main work area in the form of a turning area beforehand, the bow-shaped covering rows will often be pushed to the vicinity of the boundary, forcing vehicles to turn outside the boundary or make sharp turns inside the boundary. This can lead to problems such as crossing the boundary, plow dragging and damaging the field edge, or insufficient turning radius causing repeated reversing corrections. This solution first determines the field direction based on the user's planning intent. For example, the received initial travel direction is taken as the main direction, and turning areas are reserved at the forward and reverse ends corresponding to this main direction. The width of the turning area is not fixed, but can be derived from the kinematics of the equipment and the geometric parameters of the plow. It takes into account factors such as the minimum turning radius, the response distance of plow lifting and lowering actions, the tail sweep range caused by the rear overhang of the plow, the boundary safety margin, and the lateral offset margin caused by soil slippage, thereby ensuring that the turning area is sufficient to accommodate... The turning and alignment actions should not excessively encroach on the main working area. Specifically, the farmland grid can be projected and sorted along the main direction axis in the grid map. The grid strip with the smallest projected coordinates is selected as the first turning area, and the grid strip with the largest projected coordinates is selected as the second turning area. The remaining area is defined as the main working area. Simultaneously, masks for the main working area and the turning area are maintained separately in the grid map to identify their spatial extent. If necessary, the mask boundaries can be smoothed or morphologically processed to reduce jagged edges caused by rasterization. This area is then used to delineate... By explicitly solidifying the space requirements for turning at the edge of the field into a calculable constraint, the endpoints of the subsequent main operation path naturally fall near the boundary between the main operation area and the turning area. This avoids the risk of uncontrollable turning caused by the main operation row directly hitting the boundary. It also provides a basis for generating a separate walking trajectory for the turning area. For example, when the entrance is in the middle of the south side of the field and the starting direction is north, the turning areas at the north and south ends will form regular strip-shaped areas at both ends of that direction. When the agricultural machinery completes the plowing turn at the north end, it will not squeeze the boundary ditch, thereby improving the actual feasibility and safety.

[0022] For example, after completing the area division, the system can receive the user-specified starting point and starting direction of travel, and use these as path generation constraints to generate a path. In real farmland operations, the starting point is often determined by the field entrance, the location of the machine-cultivated road, or the farmer's working habits. Traditional full-coverage algorithms often assume that starting from a corner point of the area will cause initial empty runs and repeated manual corrections. Therefore, this solution uses the starting point and direction as hard constraints to determine the unfolding direction of the first work row and the reciprocating sequence, thereby ensuring that the generated path is consistent with the actual entrance. In specific implementation, the starting direction of travel is first mapped to the main direction vector in the grid coordinate system, and a parallel work row strip consistent with this main direction is constructed within the main work area. The spacing is determined based on the effective width of the plow and a small overlap margin can be set to compensate for positioning errors and tillage depth disturbances. Then, the starting point is projected onto each working row, selecting the row with the smallest projection distance and whose projection point falls within the walkable grid area of ​​the main working area as the first row. The starting point of the first row is determined based on the initial direction of travel, allowing vehicles to enter the first row directly from the starting point without first heading towards a corner. Then, a bow-shaped reciprocating main working path is generated according to the alternating direction of adjacent rows, so that each working row is cut into actually cultivable rows within the mask of the main working area, and transitions between rows are completed through short lateral connecting sections, thus achieving high-efficiency strip coverage within the main working area. Simultaneously, to meet the needs of the turning area... This solution defines a bow-shaped reciprocating work path that covers the space of the turning area. Within the turning area mask, a striped reciprocating trajectory is also generated. This trajectory can be implemented with strips in the same main direction as the main work area, or with strips chosen based on the terrain to facilitate alignment. Work mode control information is added to this turning area trajectory segment to distinguish between deep and shallow tillage actions. For example, a plowing command is triggered when entering the turning area trajectory segment, making that segment a passage and alignment segment. Alternatively, in situations requiring shallow tillage and compaction, this segment can be set to shallow tillage mode to achieve actual coverage of the turning area space. Through this generation strategy, the starting point and direction constraints ensure that the first segment of the path closely matches the entrance position. This reduces empty runs and the probability of human intervention. For example, when the entrance is located in the middle of the long side, the system will start from the adjacent row in the middle instead of detouring to the corner first. The bow-shaped reciprocating coverage of the main work area reduces long-distance ineffective travel by connecting adjacent rows in alternating directions, improving work efficiency and making the coverage distribution more uniform. The striped reciprocating trajectory of the U-turn area disperses the traffic pressure at the field head to multiple channels, reducing the risk of repeated compaction forming deep ruts and causing vehicles to get stuck. At the same time, it can also provide a more stable attitude adjustment channel for aligning with the next row. For example, in the muddy plot at the northern end, vehicles can choose different strip channels to complete the alignment each time, avoiding the rapid deterioration of local mud due to long-term use of the same field head trajectory, thereby improving continuous operation capability and accessibility.

[0023] Understandably, a computationally computable, labelable, and obstacle-avoidable spatial foundation is provided through a rasterized discrete representation. The spatial requirements for tillage coverage and turning alignment are separated and solidified into constraints by dividing the main work area and the turning area. Then, a bow-shaped reciprocating generation under the constraints of the starting point and direction achieves high-efficiency strip coverage starting from any entrance. An executable strip reciprocating channel is generated in the turning area to support field access or shallow field operations. This results in stronger starting point adaptability, higher coverage stability, lower risk of crossing boundaries and getting stuck, and a more engineering-feasible path output form. Especially in farmland with irregular boundaries, limited entrances, and significant differences in soil quality at the field edges, traditional full-coverage algorithms that start from a fixed corner are prone to initial empty runs and repeated road compaction at the field edges. This solution can encode entrance constraints and field edge spatial constraints into the path structure during the planning stage, making the generated path more in line with the actual plowing operation requirements in terms of executability, efficiency, and coverage consistency.

[0024] In some examples, it also includes: For the end of each work row in the bow-shaped reciprocating main work path, a fishtail-shaped turning path is generated in the corresponding turning area, and the fishtail-shaped turning path is automatically connected to the end of the work row to guide the execution of non-tillage turning actions such as lifting the plow, turning around, aligning with the next work row, and lowering the plow. Based on the grid map, the system automatically identifies uncultivated blank areas formed near the boundary of the main working area due to the existence of the turning area, and generates a horizontally reciprocating edge-closing path for the blank areas to achieve coverage of the blank areas. The bow-shaped reciprocating main operation path, each of the fishtail-shaped turning paths, and the edge-closing path are integrated to generate a seamless final operation path, which is then output to the agricultural drone or the autonomous agricultural machine, so that the agricultural drone or the autonomous agricultural machine can complete the automatic plowing operation according to the final operation path.

[0025] For example, in response to the turning and alignment requirements of each work row after reaching the end of the bow-shaped reciprocating main work path, the system generates a fishtail-shaped turning path in the corresponding turning area and automatically connects it with the end of the work row. The turning at the end of the plowing operation is not a simple turn in place or a single arc turn, but is constrained by factors such as the minimum turning radius of the vehicle, the sweeping of the plow rear suspension, the response distance of the plow lifting and lowering actions, and the yaw error caused by uneven soil resistance at the end of the field. If only a simple geometric turn is used to connect adjacent rows, insufficient alignment often occurs, requiring multiple corrections, or there is still a lateral deviation at the moment of plowing, resulting in missed plowing and repeated plowing at the beginning of the new row. Therefore, this solution abstracts the turning action into a fishtail-shaped geometric trajectory of deflection into a straight line alignment within the turning area, so that the heading angle smoothly transitions from the direction of the current work row to the direction of the next work row, and provides a sufficiently long alignment segment before entering the next work row to absorb vehicle slippage and positioning noise. The system sets a plowing trigger point at the end of each work row. This trigger point is usually located near the boundary between the main work area and the turning area, and meets the condition that there is still usable sliding distance after plowing. After the vehicle reaches the plowing trigger point, it enters the turning area and performs the first deflection segment. The first deflection segment guides the vehicle to a preset offset distance inside the turning area with a steering trajectory of limited curvature. Then it enters the swing segment. The function of the swing segment is to controllably recover the lateral offset from the lateral displacement generated by the first deflection segment, so that the vehicle's heading gradually approaches the heading of the next work row and the vehicle's position falls into the entrance capture zone of the next work row. Finally, it enters the alignment segment. The alignment segment is usually an approximately straight trajectory in the same direction as the next work row. The system sets a plowing trigger point at the end of the alignment segment and aligns the plowing trigger point with the starting point of the next work row, so that the vehicle can directly enter the stable tillage state of the next work row after plowing. When generating a fishtail-shaped turning path, the minimum turning radius of the vehicle can be used as the upper limit constraint of curvature, and the sweep range of the plow rear overhang can be limited. The safety distance from the boundary serves as the walkable boundary constraint for the turning area, while the alignment section length and allowable lateral error are used as alignment stability constraints. This ensures that the turning trajectory is executable within the spatial range of the turning area and does not intrude into the boundary danger zone. Through the above design, the number of repeated corrections for turning at the edge of the field can be significantly reduced, the probability of alignment failure caused by changes in soil resistance can be reduced, and the impact of non-cultivation time on the overall operational efficiency can be reduced. For example, in long and narrow fields, if a simple circular arc turning is used, the vehicle often still veers when entering the next row, requiring correction while tilling after plowing, resulting in obvious serpentine re-tillage in the starting area of ​​the new row. The alignment section provided by the fishtail turning can pull the heading and position back into the capture zone before plowing, making the tillage depth at the starting point of the new row more consistent and the coverage more neat. For another example, in fields with softer edges, the tires are prone to slipping when turning, resulting in an actual increase in the turning radius. The swing section and alignment section can provide buffer and correction space for this deviation, preventing the vehicle from running out of the turning area boundary or encroaching on the edge of the main working area, causing missed tillage.

[0026] For example, after completing the bow-shaped reciprocating main operation path and its fishtail-shaped turn-off connection, the system automatically identifies uncultivated blank areas near the boundary of the main operation area due to the reserved turn-off area based on the grid map, and generates a lateral reciprocating edge-closing path for the blank areas. The main operation row is usually restricted to the main operation area to avoid cultivation encroaching on the turn-off area, so that the endpoint of the main operation row stops in advance near the field boundary. On the other hand, since the strip direction of the bow-shaped main operation row is usually laid along the main direction of the plot, when the plot boundary is inclined, bent or uneven, residual grids will be formed near the boundary that cannot be completely covered by the parallel main operation row. These residual grids are in the air Fields often exhibit a strip-like or block-like distribution. Without specific edge-closing, this results in noticeable blank spaces at the ends or sides of the plots. Therefore, this solution leverages the advantages of rasterized discrete representation. By comparing the set of arable raster cells with the set of raster cells already covered by the main operation path, blank space identification is transformed into a set difference operation. The difference set raster cells are then aggregated into several blank connected regions based on connectivity, allowing for the planning of edge-closing walking directions that better suit the shape of each connected region. In specific implementation, during the execution of the main operation or simulation coverage, the system continuously updates the covered markers on the raster map based on the operation trajectory and plow width. Once all main operation rows are planned or executed, the system removes the mask from the main operation area. Unmarked and uncovered graticules are selected as candidate blank graticules, and connected component segmentation is performed on these candidate blank graticules to obtain multiple blank areas. For each blank area, the system calculates its circumscribed rectangle or main direction feature. For example, the long side direction of the blank area is taken as the more economical reciprocating direction, thereby generating a transverse reciprocating edge-collecting row sequence that is different from the main operation row direction. The edge-collecting row spacing matches the effective width of the plow and edge overlap margin can be added to ensure boundary fit. The edge-collecting row segments are also truncated according to the arable mask to avoid crossing the boundary or intruding into the obstacle graticule. Through this method of first identifying and then collecting the edges, the system can collect the edges near the field edge and boundary without compromising the efficiency of the main operation. Targeted covering of remaining areas improves overall coverage integrity and reduces land use loss. For example, in a field with a trapezoidal boundary, after the main working rows are laid out along the long side, a wedge-shaped blank area will appear at the short side end. If this blank area is still cultivated along the main working direction, it requires very short fragmented rows and frequent turning. However, by using lateral back-and-forth edge covering, a small number of longer rows can be used to complete the covering, reducing fragmented operations. Another example is when a wide turning area is reserved. If the endpoint of the main working row stops early, a continuous strip of blank space will be formed near the boundary. Lateral back-and-forth edge covering can complete the re-cultivation along this strip of space in one or a small number of back-and-forth turns, making the blank space elimination more thorough and the coverage distribution more uniform.

[0027] For example, to enable agricultural drones or autonomous agricultural machinery to continuously complete the main operation, field turn, and final edge finishing from the starting point without manual intervention, the system integrates the bow-shaped reciprocating main operation path, each fishtail-shaped turn path, and the edge finishing path to generate a seamless final operation path and output it to the execution end. If the path planning result is output as multiple isolated segmented tasks, the execution end needs to manually confirm or switch strategies between segments. However, in the field environment, positioning errors, communication fluctuations, and soil disturbances will amplify the risks of such manual switching, leading to problems such as missed execution, repeated execution, or plowing at the wrong position. Therefore, this solution organizes the path segments into a single task sequence according to the time sequence and spatial continuity at the planning end, and embeds action instructions such as plow status and speed limits in the sequence, so that the execution end can complete the entire process according to a unified state machine. The system uses the sequence of work rows in the main work path as its framework. A corresponding fishtail-shaped turn-off path is added to the end of each work row, ensuring continuity between the start of the fishtail-shaped turn-off path and the end of the work row in terms of position and heading. Simultaneously, the end of the fishtail-shaped turn-off path and the start of the next work row are both within the capture zone in terms of position and heading. After all work rows are completed, the edge-closing path is inserted as an appendix to the end of the final sequence. Alternatively, when better scheduling is achieved due to the distribution of blank areas, the edge-closing path is inserted into the window after certain field ends to reduce empty runs. During integration, the system can perform smoothing and sparsification processing on path nodes to balance control accuracy and communication bandwidth. Action events are added to each critical node, such as adding a plowing event at the start of the fishtail-shaped turn-off segment, a plowing event at the end of the alignment segment before entering the stabilization segment of the next work row, and an edge-closing mode event at the start of the edge-closing segment to adjust row spacing tolerance and speed limit. At output, the final work path can be encoded as follows: A sequence of waypoints or curve parameters with timestamps and event markers is sent to the execution controller via the vehicle bus or wireless link. The execution controller then generates tracking control commands and drives the steering, power, and plow lifting actuators to complete the operation. By integrating the output of this path, the need for task switching and manual intervention can be significantly reduced, making the operation from start to finish more continuous, reducing missed tillage and re-tillage caused by inconsistent switching, and improving the level of automation and overall efficiency of the entire process. For example, in field environments with unstable communication, if the main operation and edge finishing are two independent tasks, the second task may not be able to start in time due to link interruption, resulting in unprocessed gaps. However, the unified output final operation path can be completed autonomously by the execution end according to the event sequence after a single issuance. Furthermore, in operation scenarios at night or with low visibility, manual intervention to switch tasks is more prone to errors. The integrated event-based path allows the execution end to automatically raise and lower the plow when it reaches the designated spatial position, reducing the probability of misoperation and improving operation consistency.

[0028] In some examples, the process of dividing the farmland area based on the raster map includes: Based on preset turning width parameters, plow operation width parameters, and field safety margin parameters, turning zone boundaries are generated at opposite ends of the main working area along the forward and reverse directions of the starting travel direction, respectively. The grids within the turning zone boundaries are marked as turning zone grids, and the remaining arable grids are marked as main working area grids, so as to restrict tillage and covering to be carried out only within the main working area grids.

[0029] For example, the system first receives the user-specified starting direction of travel and converts it into a direction reference in the grid map coordinate system. The spatial positions at both ends of the main work area are not fixed to two corner points of the farmland, but should be defined around the actual work direction of travel. Otherwise, when the entrance is on the side or the starting direction specified by the farmer is inconsistent with the long side of the plot, the turning area may be incorrectly generated at the boundary where turning is not required, resulting in a lack of reserved space at the end where plowing and turning are actually needed. Therefore, in the implementation of this scheme, it is preferable to establish the grid map in a unified planar coordinate system and represent the starting direction of travel as a unit direction vector. Further, its orthogonal direction vector is calculated as the lateral division direction of the row strip, so that the forward and reverse ends along the starting direction of travel have calculable meaning. In engineering implementation, the grid coordinate points can be projected onto this direction vector to obtain the scalar coordinates of each grid in the direction of travel, so that all arable grids in the farmland can be sorted according to this scalar coordinates and the starting and ending positions of the fields can be determined. This makes the process of generating the turning area highly sensitive to the starting direction and adaptive to the entrance constraint. It avoids the errors in the location of the turning area caused by traditional algorithms that only judge the end of the field based on the up, down, left, and right of the geometric boundary. For example, when the farmland is rectangular but the entrance is in the middle of the west side and the farmer requests to work from west to east, the system will identify the east end and the west end as the two ends of the field and generate the turning area there, instead of mistakenly generating the turning area at the north and south boundaries. This ensures that the subsequent fishtail turning action has enough space at the true end of the field.

[0030] For example, after determining the directional reference, the system calculates the actual offset distance of the turning zone boundary based on preset turning width parameters, plow working width parameters, and field safety margin parameters. The turning zone width must simultaneously meet the dual constraints of the space requirements for the vehicle to complete turning and alignment while lifting the plow and the main working row's tillage coverage not encroaching on the field safety zone. The turning width parameter expresses the longitudinal reserved space required for the vehicle's turning and fishtail turning trajectory; the plow working width parameter expresses the lateral step distance and boundary contact capability of the main working row's strip coverage; and the field safety margin parameter expresses the buffer required for boundary ditches, field ridges, work entrances, and the safe distance for personnel and equipment. In implementation, the offset distance of the turning zone boundary can be set as a combination of the turning width parameter and the safety margin parameter, and a correction term related to the plow working width can be introduced when necessary, for example, when the plow working width is large and boundary contact errors may accumulate. In practice, the safety margin can be appropriately increased to prevent the last row of tillage from approaching the boundary excessively. In engineering, the system can store these parameters as a configurable parameter set, allowing for rapid adjustments under different vehicle models, different plows, and different plot safety requirements. When generating the turning zone boundary, the offset distance is converted into the number of grids. The number of grid layers corresponding to the turning zone bandwidth is obtained by dividing the offset distance by the grid side length and rounding it up or down. Through this step, the turning zone boundary no longer relies on empirical manual delineation, but can be directly derived from the parameters and can adapt to changes in equipment configuration. This reduces the risk of turning failure due to an overly narrow turning zone or excessive compression of the main working area due to an overly wide turning zone. For example, in the same plot, after replacing with a farm machine with a longer rear overhang and a larger turning radius, simply increasing the turning width parameter can automatically expand the turning zone boundary, avoiding the fishtail turning trajectory sweeping to the outside of the field ridge or multiple corrections due to insufficient space in the turning zone.

[0031] For example, after obtaining the offset distance of the turning zone boundary, the system generates the turning zone boundary at the opposite ends of the main working area along the forward and reverse directions of the starting direction of travel. The turning zone boundary should correspond one-to-one with the end position of the starting direction of travel, thereby cutting out a strip area from the front end and the rear end of the farmland along the direction of travel as the turning zone, ensuring that the bow-shaped working row can naturally enter the turning zone and perform the plow lifting and fishtail turning actions when it reaches the end. The system calculates the projected scalar value of all arable grids along the initial direction of travel. The minimum and maximum values ​​of the projected scalar value are taken as the extreme values ​​of the field ends along the reverse and forward directions, respectively. Then, the system moves back a preset offset distance from these two extreme values ​​to form two boundary sections orthogonal to the initial direction of travel. The intersection of these sections with the gridded arable area constitutes the discrete representation of the turning zone boundary in the grid space. If the farmland boundary is irregular, causing the intersection line of the section to be sawtooth-shaped, the system can perform boundary smoothing at the grid level to ensure the connectivity of the boundary strip area, thereby avoiding the situation where the boundary strip breaks and the turning zone grid is discontinuous. The boundary of the turning area can stably cover the actual field area and be aligned with the endpoint of the main work line. This ensures that the transition distance from the end of each work line to the turning area is controllable and that there is sufficient space for turning actions. This also applies to irregular plots. For example, in a trapezoidal plot, the intersection length of the cross section at the forward end and the boundary line is shorter, while the intersection line at the reverse end is longer. The system can still generate turning area grids with consistent bandwidth at both ends, thus ensuring that there is sufficient space for turning actions at both ends and that the turning area at one end is not insufficient due to the geometric asymmetry of the plot.

[0032] For example, after generating the boundaries of the two turning zones, the system marks the grids within the turning zone boundaries as turning zone grids and the remaining arable grids as main operation zone grids. During path planning and cover statistics, the system restricts tillage cover to only within the main operation zone grids. The main function of the turning zone is to provide space for plow turning and alignment. If the turning zone and the main operation zone are combined, the main operation row cover will extend to the edge of the field, forcing the plow to continuously till deeply in the field edge area, causing damage to the field ridge edges and loosening of the field edge structure. It will also make it difficult for the fishtail turning path to complete alignment according to the expected trajectory under the resistance of deep tillage. Therefore, in implementation, the system maintains a region label attribute for each grid, and the region label includes at least the main operation row. The system includes area labels and turnaround area labels. When generating the bow-shaped main operation path, only the cultivated state segment of the path is allowed to sweep the main operation area grid, while the non-cultivated passage segment of the path is allowed to enter the turnaround area grid to complete the fishtail turn or attitude adjustment. This restriction can be implemented in two ways: First, when generating operation row segments at the planning level, the main operation area mask is used as a walkable and cultivable constraint, so that the endpoint of the operation row is automatically truncated after intersecting with the main operation area mask; Second, at the execution level, a plow status event is attached to the trajectory segment. Before entering the turnaround area grid, a plow lifting event is triggered and the coverage statistics of this segment are excluded from the cultivated coverage. When aligned with the next operation row and about to return to the main operation area grid, a plow dropping event is triggered and the cultivated coverage statistics are restored. On the one hand, separating the turning area from the cultivated cover range avoids soil structure damage and boundary crossing risks caused by repeated deep plowing in the field head area. On the other hand, it ensures that the main operation area coverage statistics have a clear target set, so that subsequent blank identification and edge planning only target the remaining grid near the boundary of the main operation area for supplementary plowing, thereby improving the integrity of the cover and the interpretability of the plan. For example, in plots with ditches or narrow field ridges at the field head, restricting cultivated cover to only within the grid of the main operation area can ensure that the last row of deep plowing does not approach the edge of the ditch, and the fishtail turning section will not cause the vehicle to deviate due to the resistance of deep plowing when lifting the plow in the turning area, thereby improving both safety and operational consistency.

[0033] In some examples, generating a bow-shaped reciprocating main job path covering the main job area includes: The normal division direction of the main working row is determined by the starting direction of travel, and the main working area is divided into parallel rows at equal intervals according to the working width of the plow. The first work row and its travel endpoint are determined based on the projection position of the starting point in the row, and a reciprocating sequence is generated according to the rule of alternating reverse of adjacent work rows, so that the bow-shaped reciprocating main work path covers all main work grids in the main work area while satisfying the constraints of the starting point and the starting travel direction.

[0034] In some examples, generating a fishtail-shaped U-turn path within the corresponding U-turn area includes: When the plowing trigger point is reached at the end of the work row, the path geometric constraints of the first deflection segment, the swing segment and the alignment segment are generated in sequence in the turning area, so that the heading angle smoothly transitions from the current work row direction to the next work row direction. A plowing trigger point is generated at the end of the alignment section, so that the agricultural drone or the autonomous agricultural machine enters the tillage state of the next work row at the plowing trigger point, thereby reducing the non-tillage time during the turning process and reducing the trajectory deviation caused by the sudden change of heading.

[0035] For example, the system generates a plowing trigger point when an agricultural drone or autonomous agricultural machine is tilling along the current work row and approaching the end of the work row. During plowing, the plow's entry into the soil generates significant longitudinal resistance and lateral yaw torque. If the vehicle begins a large-curvature turn while the plow is still in the soil, on the one hand, the plow's drag will increase the vehicle's turning radius and cause sluggish heading response; on the other hand, it will create undesirable arc-shaped trenches and soil accumulation at the edge of the field, affecting the tillage depth and flatness of the next work row's entry area. Therefore, at the planning level, the plowing action must be treated as the starting event of the turning trajectory, and its trigger position must be geometrically constrained. Specifically, the system can locate the intersection of the current work row and the end boundary of the main work area in a grid map, and retreat a certain distance along the current work row direction towards the inside of the main work area as the plowing trigger point. This retreat distance can be determined by the response time of the plow lifting actuator, the vehicle's current speed, the risk of soil slippage at the edge of the field, and the positioning error margin, ensuring the vehicle... Before reaching the boundary between the main working area and the turning area, the plow is lifted and the vehicle enters a low-resistance turning state. At the same time, the plow lifting trigger point must meet the requirements of not encroaching on the deep tillage prohibition zone of the turning area and avoid premature plowing, which may cause missed tillage at the end of the main working area. On the execution side, when the vehicle reaches the plow lifting trigger point, the controller triggers the plow lifting command and enters the non-tillage passage mode. At the same time, the trajectory segment after plowing can be marked as a non-tillage segment in the coverage statistics to avoid miscalculation of coverage. Through the above settings, the high-resistance tillage segment and the high-curvature turning segment are spatially decoupled, so that the vehicle has a more stable heading response and more predictable trajectory tracking performance when turning around. It also reduces the distortion of ditches and soil accumulation caused by plow dragging at the edge of the field. For example, if the soil at the edge of the field is wet and the tires are prone to slipping, if the plow is not lifted in advance, the vehicle will experience obvious sideslip when it starts to turn, resulting in the actual turning radius being larger than the planned radius, which may lead to approaching the field ridge or crossing the boundary. Setting the plow lifting trigger point in advance can significantly reduce lateral resistance, thereby improving the efficiency of turning space utilization and safety margin.

[0036] For example, after generating the plow trigger point and entering the U-turn area, the system sequentially generates the path geometric constraints of the first deflection segment, the swing segment, and the alignment segment within the U-turn area to form a fishtail-shaped U-turn path. Switching from one work row to an adjacent work row requires not only changing the heading angle but also adjusting the vehicle's position to within the entry capture zone of the next work row. Simple single arcs or sharp turns often cannot simultaneously ensure both the executability and alignment stability under curvature constraints, especially in field scenarios with positioning noise, tire slippage, and limited U-turn area space. If the heading changes are too concentrated, it will amplify the trajectory tracking error and cause lateral deviation when entering the next row. Therefore, this solution incorporates U-turn dynamics... The system is decomposed into three geometric constraints with different functions: the first deflection segment guides the vehicle from the current working direction into the turning area and generates the necessary lateral offset, while not exceeding the vehicle's minimum turning radius constraint; the return segment recovers and corrects the lateral offset while maintaining continuous curvature or limited variation, gradually bringing the vehicle's heading closer to the next working direction and its position closer to the entrance centerline of the next working line; the alignment segment provides an approximately straight or low-curvature segment aligned with the next working direction, allowing the control system sufficient distance to achieve heading stability and lateral error convergence before the plowing operation. In practice, the system can use the entrance centerline of the next working line... As the alignment target, based on the current end position of the work line, the walkable area at the boundary of the U-turn zone, the minimum turning radius of the vehicle, and the safety margin, the target point and upper limit of curvature of the first deflection segment are calculated to ensure that the end of the first deflection segment falls within the safe area inside the U-turn zone. Then, based on the relative pose relationship between the end of the first deflection segment and the starting point of the alignment segment, a swing segment is constructed, ensuring that the difference between the heading at the end of the swing segment and the heading of the next work line does not exceed a preset angle threshold and the lateral deviation does not exceed a portion of the width of the entrance capture zone. Subsequently, the alignment segment is generated and its length is set to meet the distance requirements for the controller to achieve attitude stabilization. Simultaneously, to achieve a smooth transition, the system can apply continuity constraints to the connection points of the three segments, for example... For example, the heading at the connection point must be continuous and the curvature change must not exceed the preset upper limit to avoid control overshoot caused by sudden changes in heading. This step can significantly reduce sudden changes in heading and curvature during the turn-around stage, making the trajectory easier for actual vehicles to track, and reducing the lateral deviation and serpentine correction requirements when entering the next work row. This shortens the non-cultivation turn-around time and improves the consistency of cultivation in the starting area of ​​the work row. For example, when the plot is narrow and the width of the turn-around area is limited, a single arc turn-around often requires additional alignment before entering the next row, while the backswing and alignment sections of the fishtail path can be aligned and captured in a limited space, allowing the vehicle to enter the next row for stable cultivation with a single turn-around.

[0037] For example, after the fishtail-shaped turning path is generated, the system generates a plowing trigger point at the end of the alignment section, so that the agricultural drone or autonomous agricultural machine can enter the tillage state of the next work row at this plowing trigger point. If the plowing occurs when the heading is not yet stable or the lateral deviation is still large, the plow will convert the residual yaw error into obvious furrow deviation and lateral drag the moment it enters the soil. This will cause over-tillage, missed tillage, or tillage depth fluctuations at the beginning of the next work row. In addition, the sudden increase in resistance after the plow enters the soil will weaken the vehicle's correction ability, causing the trajectory deviation to persist for a certain distance in the beginning. Therefore, this solution sets the plowing trigger point at the end of the alignment section and aligns it with the beginning of the next work row. The system ensures that the vehicle completes the plowing into the soil in a low-curvature, low-deviation posture, minimizing the disturbance at the moment of entry. In specific implementation, the system can determine the longitudinal position of the plowing trigger point relative to the starting point of the next working row based on the width of the entry capture zone, the allowable lateral error threshold, the lateral convergence characteristics of the vehicle controller, and the positioning error margin. This ensures that the target heading at the trigger point is consistent with the heading of the next working row and that the target position falls into the central area of ​​the capture zone. At the same time, speed and acceleration constraints can be set before and after the plowing trigger point to reduce the impact of entering the soil. An operation state switching event is written at the plowing trigger point so that the execution end automatically triggers the plow to descend and switches to tillage and cover statistics when it reaches the point. By aligning the plowing timing with the alignment completion timing, the equipment can maintain a smaller lateral error and form straighter, more uniform furrows from the very first section of the new row. This reduces trajectory deviations caused by sudden changes in heading and the need for repeated corrections to the field head area, thereby reducing non-cultivation time during the turning process and improving operational efficiency. For example, when working at night or in poor visibility, manual plowing often tends to be done earlier to catch up, but this can lead to a noticeable sloping entry at the start of the new row and the formation of a re-cultivated zone. An automated plowing trigger point strategy can ensure that plowing occurs at the end of the alignment section without relying on manual judgment, thereby improving the quality of the initial section and the overall coverage consistency.

[0038] In some examples, generating a horizontally reciprocating edge-closing path for the blank area includes: In the grid map, compare the tillage coverage record of the final operation path with the grid set of the main operation area, extract the uncovered blank area grid and perform connected component segmentation on the blank area grid; For each connected domain, the tackling direction is determined along its long side, and a horizontally reciprocating parallel tackling line sequence is generated within the blank area, so that the spacing between adjacent tackling lines matches the working width of the plow. The tackling path is then added to the end of the final working path to complete the coverage of the blank area.

[0039] In some cases, considering that conventional bow-shaped reciprocating paths typically cover the main work area with parallel rows in a fixed direction, when the boundary of the main work area is locally inclined, bent, or curved, the fixed-direction rows are not aligned with the boundary normal, resulting in wedge-shaped, triangular, or elongated blank areas near the boundary. While a single row may appear close to the boundary, after accumulating multiple rows, continuous uncovered grid strips still appear near the boundary. Furthermore, traditional lateral edge finishing often requires a large number of short, fragmented rows, frequent turning around, low efficiency, and the risk of repeated compaction of the field edge. By allowing the row direction to gradually rotate at a small angle relative to the initial direction of travel only within the boundary neighborhood buffer layer, and this rotation angle decreasing with increasing distance from the boundary, the row direction is made to better conform to the boundary direction near the boundary, while remaining consistent with the main work direction in the inner layer. This reduces the area and shape complexity of the wedge-shaped blank areas without significantly compromising the efficiency of the main work. Based on this, some examples also include: A buffer layer area of ​​preset width is constructed within the boundary neighborhood of the main work area. Based on the distance from each grid in the buffer layer area to the boundary of the main work area, a rotation angle is assigned to the work belt corresponding to the bow-shaped reciprocating main work path, so that the work belt undergoes a gradual small rotation relative to the starting direction of travel within the buffer layer area, and the rotation angle gradually decreases in the direction away from the boundary of the main work area. Based on the micro-rotated work row strip, work row segments are generated within the main work area grid to cover the buffer layer area, and the work row segments are incorporated into the bow-shaped reciprocating main work path to reduce the wedge-shaped uncultivated blank area caused by the angle between the main work row strip and the boundary of the main work area.

[0040] For example, in a grid map, the system first extracts the boundary grid set of the main working area and generates a boundary neighborhood buffer layer region with a preset buffer layer width. The buffer layer width can be set according to the plow width, allowable edge overlap margin, and boundary curvature. For example, a narrower buffer layer is selected for plots with relatively straight boundaries, and a wider buffer layer is selected for plots with significantly curved or sloping boundaries. Subsequently, the system calculates the shortest distance from each grid to the boundary of the main working area within the buffer layer, forming a discrete distance field. This distance field is used for subsequent allocation of rotation angles, so that row bands within the same distance range from the boundary have similar micro-rotation trends. The system already has a set of parallel row band centerlines under the original main working direction. For row bands falling into the influence range of the buffer layer, the system calculates the rotation angle based on the distance field value of the corresponding grid of the row band. The rotation angle is usually a small angle and satisfies a monotonically decaying relationship of "the closer to the boundary, the larger the rotation; the farther away from the boundary, the smaller the rotation." To maintain executability, rotation angle changes should be limited to a range that the vehicle can track, and an upper limit can be set on the rotation angle difference between adjacent strips to avoid abrupt changes in local direction. The system then locally adjusts the centerline of the strip within the buffer layer according to the assigned rotation angle, forming a strip centerline with a slightly rotated outer layer and an inner layer maintaining its original direction. For each rotated strip centerline, the system intersects with the main work area grid mask to obtain the set of actually cultivable row segments, and inserts these segments into the main work path in a bow-shaped reciprocating sequence. To avoid sharp angles at the junctions of rotated and non-rotated areas, the system introduces continuity constraints at the row segment connections, such as ensuring continuous heading at the connection points and limiting curvature changes. At the grid level, connection trajectories can be generated through local smoothing, making the path easier for the controller to track in the boundary neighborhood. After path generation, the system evaluates the coverage within the buffer layer, statistically analyzing the area and shape characteristics of uncovered grid cells within the buffer layer. If the wedge-shaped blank area remains significant, the upper limit of the rotation angle of the row strips near the boundary can be increased, or the width of the buffer layer can be expanded; if the coverage already meets the threshold, the current parameters are maintained. This evaluation process is mainly used for parameter adaptation during the planning phase and can be achieved without introducing complex online control. Through the buffer layer and distance field, the micro-rotation of the row strips has a clear spatial range of action, preventing the rotation of all row strips in the field, thus maintaining high-efficiency strip coverage within the main working area. Through progressive rotation angle allocation, the row strip direction near the boundary is more closely aligned with the boundary, the wedge-shaped blank area is reduced, the blank shape is more regular, and the number of short, fragmented row segments required for subsequent edge finishing is reduced. Through continuous processing of connecting segments, micro-rotation does not introduce angles that are difficult to track with control, resulting in smaller actual trajectory deviations. Overall, this reduces the workload in the edge finishing stage, shortens the total operation time, and reduces the risk of soil structure damage caused by repeated compaction near the boundary.

[0041] In some cases, while the geometry of a fishtail turn is typically designed based on the vehicle's minimum turning radius and the width of the turning area, real-world terrain often presents variations in longitudinal slope, localized wet and soft areas, and uneven compaction. These factors can cause tire sideslip, understeer, or oversteer, resulting in different actual effects of the turning trajectory under the same geometric parameters. By establishing a longitudinal slope and load-bearing attribute layer in the turning area and mapping it to a turning risk level, and then using this risk level to adjust the segment length and curvature constraints of the fishtail turn, the turning becomes more conservative and controllable in high-risk terrain and more compact and efficient in low-risk terrain. This reduces time wasted due to alignment failures and sideslip, as well as quality issues in the initial segment. Based on this, some examples also include: A grid attribute layer reflecting the longitudinal slope of the field and the soil bearing capacity is established in the grid of the U-turn area, and the U-turn risk level is determined based on the grid attribute layer when generating the fishtail-shaped U-turn path; The length parameters and curvature constraint parameters of at least one of the first deflection segment, swing segment, and alignment segment of the fishtail-shaped turning path are adaptively adjusted according to the turning risk level. When the turning risk level increases, the upper limit of curvature of the swing segment is reduced and the length of the alignment segment is extended to reduce the probability of sideslip during the turning process and improve the alignment stability when entering the next work line.

[0042] For example, the system establishes longitudinal slope and bearing capacity attributes within the U-turn area grid. The longitudinal slope attribute can be obtained from elevation data, using the elevation difference between grid cells to calculate the slope magnitude and aspect. The bearing capacity attribute can be inferred from historical traffic settlement, wheel speed difference, slip ratio estimation, soil moisture sensing, or ground conductivity, and can be discretized into multiple levels. This attribute layer is consistent with the U-turn area mask, enabling spatial location of different combinations of steep and soft or gentle and hard slopes. The system combines longitudinal slope and bearing capacity to form a U-turn risk level; for example, a large longitudinal slope and poor bearing capacity are classified as high risk; a small longitudinal slope and good bearing capacity are classified as low risk; and intermediate situations are classified as medium risk. The risk level can be determined by taking statistical values ​​from the candidate area of ​​each U-turn path during planning, such as taking the maximum risk or a certain proportion of high-risk grid cells, to avoid misjudgments caused by local extreme points. The system then identifies the locations of risk-sensitive sections. For example, the first deflection section and the sway section are considered more sensitive to sideslip, while the alignment section is considered more sensitive to attitude stability, providing a basis for subsequent parameter adjustments. When the risk level increases, the system lowers the upper limit of curvature in the sway section to make the sway smoother and reduce lateral force requirements; simultaneously, it extends the length of the alignment section, allowing the vehicle more sufficient heading and lateral error convergence distance before entering the next row; if necessary, it moves the plow trigger point forward so that the vehicle is already in a low-resistance state and has a longer buffer when entering the turning zone. Conversely, in low-risk areas, the system allows for a more compact sway curvature and a shorter alignment section, thereby reducing non-cultivated travel distance. At the same time, the system can apply lower speed and acceleration limits to high-risk sections to reduce the probability of slippage and understeer. In high-risk turning zones, the plow trigger point can be slightly delayed relative to the end of the alignment section to ensure that heading stability reaches the threshold before entering the soil; in low-risk turning zones, the plow trigger point can be closer to the end of the alignment section to shorten the non-cultivated distance. This strategy prevents the disturbance at the moment of entry into the soil from amplifying trajectory deviations in high-risk areas. Therefore, the attribute layer explicitly models the invisible risks in the turning zone, upgrading the turning parameters from fixed geometric rules to an adaptive strategy that matches the field conditions. Reducing curvature and speed and lengthening the alignment segment in high-risk areas significantly reduces the probability of sideslip and alignment failure, thereby minimizing time loss due to repeated turning corrections. Shortening the alignment segment and using a more compact curvature in low-risk areas reduces uncultivated distances and improves efficiency. Overall, this improves the success rate of turning and the straightness of the new row starting segment, reducing serpentine furrows and areas of missed or overcultivated tillage.

[0043] In some cases, given the highly repetitive nature of traffic in U-turn areas, using the same "shortest path" for every U-turn can create deep ruts at the edge of the field, increasing rolling resistance, causing tires to get stuck, making steering difficult, and even requiring manual rescue. This problem is often not apparent at the beginning of operations but only becomes apparent with increased traffic. Dividing the U-turn area into multiple usable lanes and accumulating a rut risk index for each lane allows for planning U-turns that consider not only length but also the risk index, proactively rotating lanes to distribute traffic pressure across multiple areas. Based on this, some examples also include: Multiple usable U-turn lanes are generated within the U-turn area based on preset lane division rules, and a rut accumulation risk index is established for the U-turn lanes. The rut accumulation risk index is determined based on the number of passages of the grid corresponding to the U-turn lane, vehicle load parameters, and soil moisture or slipperiness level parameters. When generating the fishtail-shaped turning path at the end of the work row of the bow-shaped reciprocating main work path, a target turning channel is selected based on the rut accumulation risk index and the number of times the same turning channel is selected consecutively is limited. This allows the turning process at the end of different work rows to be distributed among the multiple turning channels, thereby reducing the risk of vehicles getting stuck due to the formation of deep ruts at the edge of the field.

[0044] For example, the system divides the U-turn area mask into multiple channels according to preset rules. Each channel can be represented as a set of strip-shaped grids, with the width of each channel at least covering the vehicle's wheelbase and lateral error margin, ensuring sufficient spacing between channels to achieve pressure distribution. The channel orientation can be consistent with or orthogonal to the main working direction, crucially supporting the entry and alignment of the fishtail U-turn section. For each channel, the system establishes a rut accumulation risk index. This index can be determined by the number of passages through the channel grid, vehicle load parameters, soil slipperiness level, or moisture content estimation. The number of passages can be obtained from path execution records; the load can be obtained from equipment parameters or estimations; the slipperiness level can be inferred from sensors, weather input, or wheel speed differences. Index updates can be incrementally accumulated for each passage through the channel grid, with higher increments assigned under wet and soft conditions, making the risk sensitive to environmental changes. When generating a fishtail turnaround path at the end of a work line, the system selects a lower-risk channel from the candidate channel set as the target channel and aligns the deflection entry point and alignment segment of the fishtail turnaround section with the entrance and exit positions of that channel. To avoid concentrated use of a single channel in the short term, the system sets continuous usage restrictions, such as limiting consecutive turns from using the same channel, or forcibly switching to another channel when the risk index of a channel exceeds a threshold. This selection strategy balances feasibility and risk control, not only pursuing the shortest path. When the risk index of all channels is high, the system can trigger conservative strategies, such as lowering the upper limit of the turnaround speed, increasing the length of the alignment segment, prioritizing channels in areas with better load-bearing capacity, or temporarily activating wider channels to reduce pressure per unit area, thereby avoiding further increasing the probability of vehicles getting stuck in extremely muddy conditions. Through channel division, traffic in the turnaround area is no longer fixed on a single shortest trajectory, possessing scalable spatial freedom. By accumulating the risk index, the system can identify the trend of deep ruts forming in advance and divert traffic before vehicles get stuck. Lane rotation significantly reduces repeated compaction at single points, maintains the load-bearing capacity of the turning area, and reduces turning failures and manual intervention caused by vehicles getting stuck or slipping. Overall, it improves the stability of long-term continuous operation, reduces unplanned downtime, and increases available operating time and efficiency.

[0045] In some cases, conventional practices often set the plowing trigger point and the plowing drop trigger point to a fixed distance or a fixed grid position. However, there are three types of dynamic factors that cannot be ignored: First, actuator response differences, as the response delay of the lifting actuator varies under different loads, temperatures, or hydraulic conditions; second, soil resistance differences, as the plow needs a certain time and travel distance to reach a stable penetration depth after dropping, and this is significantly related to soil hardness, moisture content, and adhesion; and third, speed fluctuations, as the speed during the turnaround and alignment sections may be adjusted by the controller or affected by slippage. If a fixed trigger point is still used, plowing may occur too late, resulting in deep tillage and dragging near the boundary of the turnaround area, or plowing may occur too early, resulting in missed tillage at the end of the main operation; or dropping the plow may occur too early, resulting in misalignment and sloping entry into the furrow, or dropping the plow too late, resulting in blanking at the start of a new row. The actuator delay and soil penetration establishment time can be estimated online, and the trigger point can be dynamically adjusted by combining real-time speed and soil resistance level, transforming the trigger point from a static position to the optimal time and position based on the current state. Based on this, some examples also include: During the execution of the fishtail-shaped U-turn path, displacement response data of the plow lifting actuator and plow entry pressure establishment data are collected to estimate the plow lifting response delay and plow entry establishment time online, and to obtain the real-time travel speed of the vehicle in the U-turn area and soil adhesion or resistance level parameters. Based on the plowing response delay, the plowing entry time, the real-time travel speed, and the soil adhesion or resistance level parameters, the positions of the plowing trigger point and the plowing trigger point in the path direction are dynamically corrected so that the plowing can be completed in advance without missing the end of the main work area and the plowing can be completed after alignment and stabilization, thereby reducing the impact of the turning point heading deviation on the starting tillage quality of the next work row.

[0046] For example, during the execution of a fishtail turn, the system collects data on the displacement of the plow lifting actuator over time, such as the time from issuing the lifting command to the displacement reaching the preset height. Simultaneously, it collects data related to soil penetration after the plow is lowered, such as the time point at which the soil penetration reaches a stable depth obtained through hydraulic pressure, traction resistance estimation, acceleration disturbance, or plow depth sensing. To enhance robustness, the system can accumulate this data during multiple turns, forming a sample set that can be used for estimation. The system uses a recursive estimation method to update the plow lifting response delay and the soil penetration establishment time online, ensuring that the estimated values ​​reflect the current equipment status. At the same time, the system reads real-time travel speed and soil adhesion or resistance level parameters within the turning area. Higher speeds correspond to greater spatial displacement due to actuator delays; greater soil resistance corresponds to longer spatial displacement due to soil penetration establishment time. Based on this, the system converts time quantities into spatial quantities, obtaining dynamic values ​​for plow lifting lead and plow penetration allowance. The system dynamically shifts the plow trigger point along the working row direction: when the plow response delay or speed increases, the trigger point is moved forward towards the inside of the main working area, ensuring the plow is raised before reaching the end of the main working area; when soil resistance is high and there is a risk of dragging, it can also be moved forward appropriately. However, to avoid missing the end of the plow, the system uses the grid bandwidth that still needs to be covered at the end of the main working area as a constraint to ensure that the plow trigger point, after being moved forward, can still complete the final section of coverage. If necessary, this can be compensated by a small overlap at the end of the coverage. The system dynamically corrects the plow trigger point relative to the end of the alignment section: when the plow penetration time becomes longer or soil resistance increases, the plow trigger point is moved forward appropriately within the alignment section, ensuring the plow reaches a stable depth when entering the starting position of the next working row, reducing initial gaps; when the alignment section has poor attitude stability or there is a risk of sideslip, the plow trigger point is moved backward appropriately to ensure that the heading and lateral errors during penetration are within the threshold. To reduce the instantaneous disturbance upon entering the soil, the system can also apply speed and acceleration limits before and after the plowing, making the soil entry process smoother. Through data collection and estimation, the trigger point no longer relies on fixed empirical values ​​and can adapt to actuator aging, oil temperature changes, and load variations. Dynamically moving the plowing trigger point forward can reduce understeering and trajectory deviation caused by the plow's drag during the initial turning phase, and reduce the risk of damage to the field boundary. Dynamically correcting the plowing trigger point can reduce slanted entry and serpentine furrows at the start of new rows, and reduce missed tillage and over-tillage zones in the initial phase. Overall, it reduces ineffective correction actions during the turning process, shortens non-tillage time, and improves the consistency and stability of operation quality.

[0047] Please see Figure 2 One embodiment of the agricultural drone path planning device based on a raster map in this application is used for plowing operations of agricultural drones or autonomous agricultural machinery. The device includes: The acquisition unit 21 is used to acquire the boundary information of the farmland area to be operated, and to digitize and rasterize the farmland area to generate a raster map, wherein each raster corresponds to a small arable unit. The division unit 22 is used to divide the farmland area based on the grid map to form a main working area and a turning area located at both ends of the main working area, and to mark the spatial range of the main working area and the turning area in the grid map. Planning unit 23 is used to receive the starting point and starting direction of travel specified by the user, and use the starting point and starting direction of travel as path generation constraints to generate a bow-shaped reciprocating main operation path covering the main operation area, and generate a bow-shaped reciprocating operation path covering the space range of the U-turn area.

[0048] like Figure 3 As shown, this application embodiment also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements the steps of any of the above-described methods for agricultural drone path planning based on gridded maps.

[0049] Since the electronic device described in this embodiment is the device used to implement the agricultural drone path planning device based on a gridded map in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. Any device used by those skilled in the art to implement the method in this application embodiment is within the scope of protection of this application.

[0050] In practical implementation, when the computer program 311 is executed by the processor, it can achieve the following: Figure 1 Any of the corresponding implementation methods in the embodiments.

[0051] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0052] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A path planning method for agricultural unmanned aerial vehicles (UAVs) based on raster maps, characterized in that, The method for plowing operations using agricultural drones or autonomous agricultural machinery includes: Obtain the boundary information of the farmland area to be operated, and digitize the farmland area into a raster to generate a raster map, where each raster corresponds to a small arable unit; The farmland area is divided into regions based on the grid map to form a main work area and turning areas located at both ends of the main work area, and the spatial range of the main work area and the turning areas is marked in the grid map. The system receives the user-specified starting point and starting direction of travel, and uses the starting point and starting direction of travel as path generation constraints to generate a bow-shaped reciprocating main work path covering the main work area, and generates a bow-shaped reciprocating work path covering the space of the U-turn area.

2. The method as described in claim 1, characterized in that, Also includes: For the end of each work row in the bow-shaped reciprocating main work path, a fishtail-shaped turning path is generated in the corresponding turning area, and the fishtail-shaped turning path is automatically connected to the end of the work row to guide the execution of non-tillage turning actions such as lifting the plow, turning around, aligning with the next work row, and lowering the plow. Based on the grid map, the system automatically identifies uncultivated blank areas formed near the boundary of the main working area due to the existence of the turning area, and generates a horizontally reciprocating edge-closing path for the blank areas to achieve coverage of the blank areas. The bow-shaped reciprocating main operation path, each of the fishtail-shaped turning paths, and the edge-closing path are integrated to generate a seamless final operation path, which is then output to the agricultural drone or the autonomous agricultural machine, so that the agricultural drone or the autonomous agricultural machine can complete the automatic plowing operation according to the final operation path.

3. The method as described in claim 1, characterized in that, The process of dividing the farmland area based on the raster map includes: Based on preset turning width parameters, plow operation width parameters, and field safety margin parameters, turning zone boundaries are generated at opposite ends of the main working area along the forward and reverse directions of the starting travel direction, respectively. The grids within the turning zone boundaries are marked as turning zone grids, and the remaining arable grids are marked as main working area grids, so as to restrict tillage and covering to be carried out only within the main working area grids.

4. The method as described in claim 1, characterized in that, The step of generating a bow-shaped reciprocating main work path covering the main work area includes: The normal division direction of the main working row is determined by the starting direction of travel, and the main working area is divided into parallel rows at equal intervals according to the working width of the plow. The first work row and its travel endpoint are determined based on the projection position of the starting point in the row, and a reciprocating sequence is generated according to the rule of alternating reverse of adjacent work rows, so that the bow-shaped reciprocating main work path covers all main work grids in the main work area while satisfying the constraints of the starting point and the starting travel direction.

5. The method as described in claim 2, characterized in that, The step of generating a fishtail-shaped U-turn path within the corresponding U-turn area includes: When the plowing trigger point is reached at the end of the work row, the path geometric constraints of the first deflection segment, the swing segment and the alignment segment are generated in sequence in the turning area, so that the heading angle smoothly transitions from the current work row direction to the next work row direction. A plowing trigger point is generated at the end of the alignment section, so that the agricultural drone or the autonomous agricultural machine enters the tillage state of the next work row at the plowing trigger point, thereby reducing the non-tillage time during the turning process and reducing the trajectory deviation caused by the sudden change of heading.

6. The method as described in claim 2, characterized in that, The process of generating a horizontally reciprocating edge-finishing path for the blank area includes: In the grid map, compare the tillage coverage record of the final operation path with the grid set of the main operation area, extract the uncovered blank area grid and perform connected component segmentation on the blank area grid; For each connected domain, the tackling direction is determined along its long side, and a horizontally reciprocating parallel tackling line sequence is generated within the blank area, so that the spacing between adjacent tackling lines matches the working width of the plow. The tackling path is then added to the end of the final working path to complete the coverage of the blank area.

7. The method according to any one of claims 1 to 6, characterized in that, Also includes: A buffer layer area of ​​preset width is constructed within the boundary neighborhood of the main work area. Based on the distance from each grid in the buffer layer area to the boundary of the main work area, a rotation angle is assigned to the work belt corresponding to the bow-shaped reciprocating main work path, so that the work belt undergoes a gradual small rotation relative to the starting direction of travel within the buffer layer area, and the rotation angle gradually decreases in the direction away from the boundary of the main work area. Based on the micro-rotated work row strip, work row segments are generated within the main work area grid to cover the buffer layer area, and the work row segments are incorporated into the bow-shaped reciprocating main work path to reduce the wedge-shaped uncultivated blank area caused by the angle between the main work row strip and the boundary of the main work area.

8. A path planning device for agricultural unmanned aerial vehicles based on a raster map, characterized in that, The device is used for plowing operations by agricultural drones or autonomous agricultural machinery, and includes: The acquisition unit is used to acquire the boundary information of the farmland area to be operated, and to digitize and rasterize the farmland area to generate a raster map, wherein each raster corresponds to a small arable unit. A division unit is used to divide the farmland area based on the grid map to form a main working area and a turning area located at both ends of the main working area, and to identify the spatial range of the main working area and the turning area in the grid map; The planning unit is used to receive the starting point and starting direction of travel specified by the user, and use the starting point and starting direction of travel as path generation constraints to generate a bow-shaped reciprocating main operation path covering the main operation area, and generate a bow-shaped reciprocating operation path covering the space range of the U-turn area.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program stored in the memory to implement the steps of the agricultural unmanned aerial vehicle path planning method based on a raster map as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the agricultural drone path planning method based on a raster map as described in any one of claims 1-7.

Images