GPS positioning-based unmanned agricultural machine operation path planning method
By using a GPS-based unmanned agricultural machinery operation path planning method, which combines real-time positioning and historical data to dynamically adjust the path, the static problem of path planning in agricultural machinery operations is solved, achieving higher operational adaptability and accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WEIFANG HUABO AGRI EQUIP CO LTD
- Filing Date
- 2026-03-06
- Publication Date
- 2026-05-15
AI Technical Summary
Existing unmanned agricultural machinery operation path planning methods fail to fully consider the differences in the real-time location of agricultural machinery and the internal conditions of the field during actual operation, lack dynamic adjustment mechanisms, and cannot adapt to changes in the physical characteristics of the field caused by historical operations.
By comparing the real-time geographic coordinate point sequence based on GPS positioning with the boundary of the work plot, an operation path density distribution map is generated. An initial guidance path network is generated using the path potential energy gradient algorithm. The path wear coefficient is dynamically corrected by combining the historical data of agricultural machinery operation to form an adaptive guidance path.
It improves the response accuracy and flexibility of path planning, can proactively avoid areas that have been over-operated or over-compacted in the past, optimizes the non-uniform evolution of agricultural machinery working conditions in the field, and enhances the adaptability and accuracy of agricultural machinery operations.
Smart Images
Figure CN121784802B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of agricultural machinery navigation path planning technology, specifically a method for planning unmanned agricultural machinery operation paths based on GPS positioning. Background Technology
[0002] In existing automated navigation operations for agricultural machinery, a common path planning method involves pre-planning a fixed operational path for the entire field based on pre-measured boundary coordinates. This method pre-determines a static and global path, which remains unchanged throughout the operation. This conventional approach fails to adequately consider the real-time location of the agricultural machinery at the start of actual operations, as well as the potential spatial differences in the field's internal conditions during the operation.
[0003] Even when technologies exist capable of localized path adjustments, they typically lack a crucial triggering mechanism and a data foundation for continuous optimization. Their adjustments are often based on pre-defined rules or single real-time sensor data, failing to consider the precise real-time positioning of the agricultural machinery as the core condition for triggering path calculations. Furthermore, path generation and subsequent optimization generally lack quantitative consideration of the cumulative effects of historical operations within the same plot, making it impossible for the path network to adapt to the non-uniform changes in the physical characteristics of the field caused by past operations.
[0004] Key challenges in improving the adaptability and accuracy of unmanned agricultural machinery operations include ensuring that path planning is precisely aligned with the real-time operating location of the agricultural machinery, and enabling the planned paths to dynamically adapt to the continuously evolving working conditions within the field due to historical operations. This requires path planning technology to dynamically trigger internal calculations based on real-time positioning information and to continuously self-correct by integrating historical operation data. Summary of the Invention
[0005] This invention aims to solve at least one of the technical problems existing in the prior art;
[0006] Therefore, this invention proposes a GPS-based unmanned agricultural machinery operation path planning method, including:
[0007] Obtain the real-time geographic coordinate point sequence of agricultural machinery obtained through the GPS positioning system;
[0008] The real-time geographic coordinate point sequence is compared with the pre-stored set of geographic coordinates of the work plot boundary to determine the real-time positioning status of the agricultural machinery within the work plot.
[0009] Based on the real-time positioning status, the operation path density field calculation process for the internal area of the operation site is initiated to generate an initial density distribution map of the operation path.
[0010] Based on the initial density distribution map of the operation path, the path potential gradient algorithm is used to generate an initial guidance path network for unmanned agricultural machinery covering the internal area of the operation plot.
[0011] After the initial guidance path network for unmanned agricultural machinery is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial guidance path network for unmanned agricultural machinery, thus forming an adaptive guidance path for unmanned agricultural machinery.
[0012] The agricultural machinery control unit controls the agricultural machinery to perform unmanned operations based on the aforementioned unmanned driving adaptive guidance path.
[0013] Furthermore, obtaining the real-time geographic coordinate sequence of agricultural machinery through the GPS positioning system includes:
[0014] Receive continuous positioning signals from the agricultural machinery vehicle's GPS receiver, and decode them to obtain the agricultural machinery's real-time longitude and latitude data;
[0015] The real-time longitude and latitude data are collected at a preset sampling frequency to form an ordered set of real-time geographic coordinate points;
[0016] The set of real-time geographic coordinate points is filtered to remove abnormal coordinate points caused by GPS signal drift, resulting in a smoothed sequence of real-time geographic coordinate points.
[0017] Furthermore, comparing the real-time geographic coordinate point sequence with the pre-stored set of geographic coordinates of the work plot boundary to determine the real-time positioning status of the agricultural machinery within the work plot includes:
[0018] Read the predefined set of geographic coordinates of the work site boundary, which defines the polygonal closed boundary of the work site;
[0019] Calculate the positional relationship between the latest coordinate point in the real-time geographic coordinate point sequence and the closed boundary of the polygon of the work site;
[0020] If the latest coordinate point is located inside the closed boundary of the polygon, then the instantaneous positioning status is a point within the work area;
[0021] If the latest coordinate point is located outside the closed boundary of the polygon, then the instantaneous positioning status is an external point of the work area;
[0022] If the latest coordinate point is located on the closed boundary of the polygon, then the instantaneous positioning status is the boundary point of the work area.
[0023] Furthermore, based on the real-time positioning status, the calculation process for the work path density field within the work site area is initiated to generate an initial density distribution map of the work paths, including:
[0024] When the real-time positioning status is a point within the work area, the target planning area inside the work area is delineated with the point within the work area as the center.
[0025] The target planning area is rasterized and divided into multiple uniform cell grids;
[0026] A path density value is initialized for each cell grid, and the path density value is assigned based on the distance between the grid center point and the boundary of the work area and the soil type attribute of the grid;
[0027] Traverse all cell grids and perform spatial smooth interpolation calculations on the path density values of each cell grid to generate an initial density distribution map of the operation paths covering the entire target planning area.
[0028] Furthermore, based on the initial density distribution map of the work path, the path potential gradient algorithm is used to generate an initial guidance path network for unmanned agricultural machinery covering the internal area of the work site, including:
[0029] In the initial density distribution map of the work path, the path density values are converted into virtual potential energy values to construct a path potential energy field;
[0030] In the path potential energy field, the starting point of the agricultural machinery is selected as the potential energy point, and the preset end point of the operation within the operation plot is selected as the potential energy sink point.
[0031] Starting from the potential energy point, iteratively search along the negative gradient direction of the potential energy field along the path until the potential energy sink is reached, and record this search trajectory as a candidate guiding path.
[0032] By changing the location of potential energy points or potential energy sinks, the iterative search process is repeated to generate multiple candidate guidance paths connecting different start and end positions of operations, thus forming an initial guidance path network for unmanned agricultural machinery.
[0033] Furthermore, after the initial guidance path network for unmanned agricultural machinery is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial guidance path network, forming an adaptive guidance path for unmanned agricultural machinery, including:
[0034] Extract data on the number of times and intensity of agricultural machinery operations were repeated along a specific path or area on the current or similar work plots from the agricultural machinery operation history database.
[0035] Based on the data on the number of repetitive operations and the intensity of operations, the path wear coefficients for different geographical locations within the work area are calculated. Areas with more repetitive operations and higher intensity of operations have larger path wear coefficients.
[0036] The path wear coefficient is mapped to the geographical location corresponding to the initial guidance path network for unmanned agricultural machinery.
[0037] Based on the mapped path wear coefficient, the virtual potential energy value of the corresponding path segment in the initial guidance path network for unmanned agricultural machinery is adjusted, and the path is raised or corrected to generate an adaptive guidance path for unmanned agricultural machinery that is more adapted to the actual ground conditions.
[0038] Further, calculating the positional relationship between the latest coordinate point in the real-time geographic coordinate point sequence and the closed boundary of the polygon of the work site includes:
[0039] The calculation is performed using the ray method, drawing a ray horizontally to the right from the latest coordinate point;
[0040] Calculate the number of intersections between the ray and each edge of the closed boundary of the polygon of the work site;
[0041] If the number of intersection points is odd, then the latest coordinate point is determined to be inside the closed boundary of the polygon;
[0042] If the number of intersection points is even, then the latest coordinate point is determined to be located outside the closed boundary of the polygon.
[0043] Furthermore, in the path potential energy field, selecting the starting point of the agricultural machinery as the potential energy source and selecting a preset end point within the work area as the potential energy sink includes:
[0044] From the real-time geographic coordinate point sequence of the agricultural machinery, select a coordinate point that meets the preset stability conditions as the current starting point of the agricultural machinery, i.e., the potential energy point;
[0045] The coordinates of the endpoint of this unmanned operation are extracted from the task instructions and used as the potential energy sink.
[0046] Both the potential energy point and the potential energy sink are located within the working area.
[0047] Furthermore, based on the data on the number of repetitive operations and the intensity of operations, the path wear coefficient for different geographical locations within the work area is calculated, including:
[0048] Establish a historical operation record table indexed by grid cells within the operation plot, recording the cumulative number of operations and average operation intensity for each grid cell;
[0049] Input the cumulative number of operations and the average operation intensity into the wear calculation function;
[0050] The wear calculation function outputs a value between zero and one as the path wear coefficient of the grid cell. The higher the value, the more severe the path wear at the corresponding position of the grid cell.
[0051] Furthermore, adjusting the virtual potential energy value of the corresponding path segment in the initial guidance path network for unmanned agricultural machinery driving, based on the mapped path wear coefficient, includes:
[0052] Set a basic potential energy adjustment amount;
[0053] Multiply the path wear coefficient by the basic potential energy adjustment to obtain the actual potential energy adjustment value for the grid cell;
[0054] In the path potential energy field, the virtual potential energy value of the grid cell with path wear is increased by the actual potential energy adjustment value, so that the grid cell region with path wear exhibits higher "potential energy" in path planning, guiding the path to detour.
[0055] Compared with the prior art, the beneficial effects of the present invention are:
[0056] By comparing the real-time geographic coordinates of agricultural machinery with pre-stored boundaries of the work area, and using the machinery's "instantaneous positioning status" as a decision signal, the path density field calculation process for the internal area of the work area is initiated. This mechanism changes the traditional model of pre-calculating and fixing the global path for the entire plot, allowing the origin of the path network generation and the calculated density to be closely related to the actual location of the agricultural machinery and its state upon entering the work area. The resulting initial density distribution map of the work path and its subsequent guiding path network are dynamically bound to the real-time entry point and state of the current operation, improving the accuracy and flexibility of path planning in response to the initial work scenario.
[0057] After generating the initial guidance path network, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial network. This coefficient quantifies the cumulative history of agricultural machinery travel or operations in a specific plot area, reflecting the actual changes in physical properties. Using this coefficient, the path weights are adjusted and optimized, enabling the final adaptive guidance path to proactively avoid areas with excessively frequent historical operations, potentially over-compacted areas, or areas with high resistance, or to optimize the agricultural machinery's passage strategy in already compacted areas. This transforms the output of path planning from a fixed geometric line into a dynamic navigation network that absorbs historical experience and continuously optimizes itself with increasing operation frequency, allowing agricultural machinery operations to adapt to the non-uniform evolution of field conditions. Attached Figure Description
[0058] Figure 1 This is a flowchart illustrating the steps of the GPS-based unmanned agricultural machinery operation path planning method described in this invention.
[0059] Figure 2 A flowchart for obtaining real-time geographic coordinate point sequences for agricultural machinery;
[0060] Figure 3 A flowchart generated for the initial density distribution map of the work path;
[0061] Figure 4 A contour map of the potential energy field of the agricultural machinery operation path after correction for path wear coefficient;
[0062] Figure 5 The correlation between average work intensity and wear coefficient. Detailed Implementation
[0063] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0064] See Figure 1 The system acquires a real-time geographic coordinate sequence obtained by the agricultural machinery via GPS positioning. This sequence is compared with a pre-stored set of geographic coordinates for the boundaries of the work area to determine the machinery's immediate location within the work area. Based on this location, the system initiates a path density field calculation process for the internal area of the work area, generating an initial path density distribution map. Using this initial path density distribution map, a path potential gradient algorithm is applied to generate an initial unmanned guidance path network covering the internal area of the work area. After the initial unmanned guidance path network is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the network, forming an adaptive unmanned guidance path. The agricultural machinery control unit then uses this adaptive unmanned guidance path to control the machinery to perform unmanned operations.
[0065] See Figure 2 In one embodiment of the present invention, the specific process of obtaining the real-time geographic coordinate point sequence of agricultural machinery through a GPS positioning system includes: receiving continuous positioning signals from the on-board GPS receiver of the agricultural machinery, and decoding to obtain the real-time longitude and real-time latitude data of the agricultural machinery. The real-time longitude and real-time latitude data are collected at a preset sampling frequency to form an ordered set of real-time geographic coordinate points. The set of real-time geographic coordinate points is filtered to remove abnormal coordinate points caused by GPS signal drift, resulting in a smoothed sequence of real-time geographic coordinate points.
[0066] Obtaining the real-time geographic coordinate point sequence of agricultural machinery via a GPS positioning system includes the following steps: receiving continuous positioning signals from the agricultural machinery's onboard GPS receiver, decoding to obtain the agricultural machinery's real-time longitude and latitude data. In some embodiments, the continuous positioning signals follow the NMEA-0183 protocol format, and the decoding process parses GPGGA statements to obtain the real-time longitude and latitude data. The real-time longitude and latitude data are collected at a preset sampling frequency to form an ordered set of real-time geographic coordinate points. Optionally, the preset sampling frequency is set to 10 Hz, i.e., ten times the real-time longitude and latitude data are collected per second. In specific implementations, the sampling frequency can be adjusted according to the type of agricultural machinery operation. For example, a higher sampling frequency is used in sowing operations to accurately record the path, while a lower sampling frequency is used in transportation operations to save storage space. The set of real-time geographic coordinate points is filtered to remove abnormal coordinate points caused by GPS signal drift, resulting in a smoothed sequence of real-time geographic coordinate points. It can be understood that GPS signal drift is mainly caused by multipath reflection or atmospheric delay, causing the coordinate points to deviate from their true positions.
[0067] In some embodiments, the filtering process employs a sliding window mean filtering algorithm, the formula of which is expressed as:
[0068]
[0069] in: This represents the longitude and latitude values of the filtered coordinate points. This represents the width of the sliding window, i.e., the number of coordinate points contained within the window. Indicates the first in the window The longitude value of each coordinate point. Indicates the first in the window The latitude value of each coordinate point. In specific implementation, the width of the sliding window... Setting it to 5 means that each filter uses the current coordinate point and the previous four historical coordinate points to calculate the mean. Optional, the width of the sliding window. It can dynamically adapt to signal quality, increasing the width to enhance smoothing when signal fluctuations are large. In terms of data comparison, an unfiltered set of real-time geographic coordinate points may exhibit coordinate jumps or outliers in the graphical display, while the filtered sequence of real-time geographic coordinate points presents a continuous and smooth trajectory, reducing the impact of abrupt changes on subsequent processing. It is understandable that filtering is a crucial step in data preprocessing, directly affecting the reliability of the real-time geographic coordinate point sequence. In practical implementation, Kalman filtering can be used instead of sliding window mean filtering to achieve better noise suppression in dynamic environments.
[0070] In one embodiment of the present invention, the specific process of comparing the real-time geographic coordinate point sequence with a pre-stored set of geographic coordinates of the work plot boundary to determine the instantaneous positioning status of the agricultural machinery within the work plot includes: reading the predefined set of geographic coordinates of the work plot boundary, which defines the polygonal closed boundary of the work plot; calculating the positional relationship between the latest coordinate point in the real-time geographic coordinate point sequence and the polygonal closed boundary of the work plot; calculating this positional relationship using the ray method, drawing a ray horizontally to the right from the latest coordinate point, and calculating the number of intersections between this ray and each side of the polygonal closed boundary of the work plot; if the number of intersections is odd, the latest coordinate point is determined to be inside the polygonal closed boundary, and the instantaneous positioning status is an inside point of the work plot; if the number of intersections is even, the latest coordinate point is determined to be outside the polygonal closed boundary, and the instantaneous positioning status is an outside point of the work plot; if the latest coordinate point is on the polygonal closed boundary, the instantaneous positioning status is a boundary point of the work plot.
[0071] The process of comparing the real-time geographic coordinate point sequence with the pre-stored set of geographic coordinates of the work plot boundary to determine the real-time positioning status of the agricultural machinery within the work plot includes the following steps: First, reading the predefined set of geographic coordinates of the work plot boundary. This set defines the polygonal closed boundary of the work plot. In some embodiments, the predefined set of geographic coordinates of the work plot boundary is pre-entered through the agricultural machinery management system, and its data format is a set of latitude and longitude coordinate points arranged in clockwise or counterclockwise order. Second, calculating the positional relationship between the latest coordinate point in the real-time geographic coordinate point sequence and the polygonal closed boundary of the work plot. The positional relationship is calculated using the ray method, drawing a ray horizontally to the right from the latest coordinate point and calculating the number of intersections between the ray and each edge of the polygonal closed boundary of the work plot. When calculating the number of intersections, each edge of the polygonal closed boundary of the work plot is traversed, and it is determined whether the ray intersects the current edge. In some embodiments, the geometric determination formula for whether a ray intersects a line segment is:
[0072]
[0073] in: Indicates the longitude and latitude of the latest coordinate point. and These represent the longitude and latitude of the two endpoints of the current side within the closed boundary of the polygonal boundary of the work area, respectively, with symbols... The symbol represents the logical "XOR" operation. This represents a logical AND operation. When `IntersectionExists` is true, it counts one intersection point. For data comparison, assuming the closed polygon boundary of the work area has five vertices, if the latest coordinate point is inside the polygon, the number of times its ray intersects the boundary edge is odd (e.g., three). If the latest coordinate point is outside the polygon, the number of times its ray intersects the boundary edge is even (e.g., zero or two). It's understandable that the computational efficiency of the ray casting method is directly related to the number of polygon edges. In practice, for complex work areas, the number of polygon vertices may be large. If the latest coordinate point is located on the closed polygon boundary, its latitude or longitude precisely matches a point on an endpoint or line segment of a boundary edge. In this case, the instantaneous positioning status is directly determined as the work area boundary point. Optionally, to avoid misjudgments caused by floating-point precision errors, a tolerance threshold can be set. When the distance between the latest coordinate point and any side of the boundary is less than this threshold, it is determined to be the work area boundary point. Finally, the coordinate point position is determined based on the parity of the number of intersection points. If the number of intersection points is odd, the latest coordinate point is determined to be inside the closed boundary of the polygon, and the immediate positioning status is an internal point of the work area. If the number of intersection points is even, the latest coordinate point is determined to be outside the closed boundary of the polygon, and the immediate positioning status is an external point of the work area.
[0074] See Figure 3 In one embodiment of the present invention, the specific process of initiating the operation path density field calculation process for the internal area of the operation plot based on the real-time positioning status to generate an initial operation path density distribution map includes: when the real-time positioning status is an internal point of the operation plot, delineating the target planning area within the operation plot with the internal point of the operation plot as the center; performing rasterization processing on the target planning area, dividing it into multiple uniform cell grids; initializing a path density value for each cell grid, the path density value being assigned based on the distance between the grid center point and the boundary of the operation plot, and the soil type attribute of the grid; traversing all cell grids, performing spatial smooth interpolation calculation on the path density value of each cell grid to generate an initial operation path density distribution map covering the entire target planning area.
[0075] The calculation process for the density field of the work path within the work area is initiated based on the real-time positioning status, generating an initial density distribution map of the work path. This includes the following operations: When the real-time positioning status is within the work area, a target planning area is delineated within the work area, centered on the point within the work area. In some embodiments, the target planning area is set as a square area with a side length of 200 meters, centered on the point within the work area, ensuring coverage of the possible range of a single agricultural machinery operation. The target planning area is rasterized, dividing it into multiple uniform unit grids. In specific implementations, the size of the unit grid is set to 1 meter × 1 meter, thereby achieving a balance between computational accuracy and computational burden. A path density value is initialized for each unit grid. The path density value is assigned based on the distance between the grid center point and the boundary of the work area, and the soil type attribute of the grid. It can be understood that the closer the grid is to the boundary of the work area, the higher its importance in path planning, and different soil types directly affect the ease with which the agricultural machinery passes. The initial calculation of the path density value follows the following formula:
[0076]
[0077] in: This represents the path density value after initialization. This represents the shortest Euclidean distance from the center point of the cell grid to the boundary of the work area. This represents the maximum value of the shortest distance from the center point of all grids within the target planning area to the boundary of the work plot. This represents the weighting coefficients mapped based on soil type attributes. and These are the harmonic coefficients for the distance factor and the soil type factor, respectively. For data comparison, assuming one grid cell's center point is 5 meters from the boundary and another's center point is 50 meters from the boundary, under the same soil type, the former's... The calculated result of the first item is greater than the latter, resulting in its initial path density value. Higher; for soil type attributes, clay region mapping The weighting coefficient may be higher in sandy soil areas, resulting in higher initial path density values in clay areas at the same distance. All cell grids are traversed, and the path density value of each cell is spatially smoothed using interpolation. Optionally, the spatial smoothing interpolation uses an inverse distance weighting method, considering a weighted average of the path density values of each cell and its eight neighboring cells to eliminate local abrupt changes. An initial path density distribution map covering the entire target planning area is generated. In practice, the initial path density distribution map is stored in memory as a two-dimensional matrix data structure, where the row and column indices of the matrix correspond to geographic coordinates, and the values of the matrix elements correspond to path density values.
[0078] In one embodiment of the present invention, the specific process of generating an initial guidance path network for unmanned agricultural machinery covering the internal area of the work plot using a path potential energy gradient algorithm based on the initial density distribution map of the work path includes: converting path density values into virtual potential energy values in the initial density distribution map of the work path to construct a path potential energy field. In the path potential energy field, the starting point of the agricultural machinery is selected as the potential energy point, and a preset work endpoint within the work plot is selected as the potential energy sink. Specifically, from the real-time geographic coordinate point sequence of the agricultural machinery, a coordinate point that meets the preset stability conditions is selected as the current starting point of the agricultural machinery, i.e., the potential energy point; the coordinates of the endpoint position to be completed in this unmanned operation are parsed from the work task instruction and used as the potential energy sink; both the potential energy point and the potential energy sink are located within the internal area of the work plot. Starting from the potential energy point, an iterative search is performed along the negative gradient direction of the path potential energy field until the potential energy sink is reached, and this search trajectory is recorded as a candidate guidance path. By changing the location of potential energy points or potential energy sinks and repeating the iterative search process, multiple candidate guidance paths connecting different start and end positions of operations are generated, forming an initial guidance path network for unmanned agricultural machinery.
[0079] Generating an initial guidance path network for unmanned agricultural machinery covering the internal area of the work site using a path potential energy gradient algorithm based on the initial density distribution map of the work path includes the following operations: In the initial density distribution map of the work path, path density values are converted into virtual potential energy values to construct a path potential energy field. The conversion relationship between path density values and virtual potential energy values follows a monotonically decreasing function, because a higher path density means that the area is more suitable for passage, and therefore a lower virtual potential energy value should be assigned to attract the path to pass. In some embodiments, the conversion formula used is:
[0080]
[0081] in: Indicates the location in grid coordinates The virtual potential energy value at that location, This represents the path density value of the corresponding grid in the initial density distribution map of the job path. It is a fundamental potential energy scaling constant. It is a positive parameter that controls the sensitivity of the transition. It is a natural constant. Data comparison can be presented in a simplified table showing the virtual potential energy values calculated using the formula for different path density values, assuming... , See Table 1.
[0082] Table 1: Example of path density value to virtual potential energy value conversion table
[0083] ;
[0084] It is understandable that after the virtual potential energy field is constructed, each geographical location within the work area corresponds to a potential energy value, forming a potential energy contour map. The starting point of the agricultural machinery is selected as the potential energy source point within the path potential energy field, and a preset work endpoint within the work area is selected as the potential energy sink point. In specific implementation, a coordinate point that meets preset stability conditions is selected from the real-time geographic coordinate sequence of the agricultural machinery as the current starting point of the agricultural machinery, i.e., the potential energy source point. The preset stability conditions include requiring the standard deviation of the positional changes of multiple consecutive coordinate points to be less than a threshold and the speed to be lower than a set value, to ensure that the agricultural machinery is in a static or low-speed stable state. The coordinates of the endpoint position to be completed in this unmanned operation are parsed from the work task instruction and used as the potential energy sink point. Both the potential energy source point and the potential energy sink point are verified to be located within the work area. Starting from the potential energy point, an iterative search is performed along the negative gradient direction of the path's potential energy field until the potential energy sink is reached. This search trajectory is recorded as a candidate guiding path. In each search step, the gradient direction of the current position in the path's potential energy field is calculated. The gradient direction indicates the direction in which the virtual potential energy value decreases the fastest. Optionally, a numerical difference method, such as the central difference method, is used to calculate the gradient. Moving one step along the negative gradient direction, i.e., the direction of virtual potential energy decrease, to reach a new position, this process is repeated until the distance between the new position and the potential energy sink is less than the convergence tolerance. The position of the potential energy point or the potential energy sink is changed, and the iterative search process is repeated to generate multiple candidate guiding paths connecting different start and end positions of the operation, forming an initial guiding path network for unmanned agricultural machinery. In some embodiments, the change of the potential energy point is based on the real-time position update of the agricultural machinery, and the change of the potential energy sink is based on the sequential switching of multiple target points in the operation task. It can be understood that the initial guiding path network for unmanned agricultural machinery consists of multiple smooth curved paths that tend to traverse high path density areas, providing the agricultural machinery with basic global path options.
[0085] See Figure 4In the path potential energy field correction stage of unmanned agricultural machinery operation path planning, the introduction of the path wear coefficient enables dynamic adaptive optimization of the initial guiding path network. Specifically, the initial path potential energy field within the work area is derived from the path density distribution, and its virtual potential energy value shows a monotonically decreasing relationship with path suitability. Based on historical agricultural machinery operation data, the cumulative number of operations and average operation intensity of each grid unit within the plot are statistically analyzed. A path wear coefficient in the range of 0 to 1 is generated through a wear calculation function, and this coefficient is positively correlated with the degree of path wear. In the path potential energy field, the path wear coefficient is multiplied by the basic potential energy adjustment amount to obtain the actual potential energy adjustment value, and the virtual potential energy value of the corresponding grid unit is incrementally corrected, making high-wear areas exhibit a higher "potential energy barrier," thereby guiding the path to automatically avoid obstacles. The corrected potential energy contour map shown in the figure clearly presents this effect: the virtual potential energy value inside the work area (10m≤X≤40m, 10m≤Y≤40m) is generally raised, and the contour lines are more densely distributed. In particular, a significant high potential energy area is formed in the core area where historical operations are frequent (around X=25m, Y=25m). This effectively avoids repeated compaction of the same area by agricultural machinery and improves the adaptability and sustainability of the work path.
[0086] In one embodiment of the present invention, after the initial guidance path network for unmanned agricultural machinery is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial guidance path network and form an adaptive guidance path for unmanned agricultural machinery. The specific process includes: extracting data on the number of times and intensity of repeated operations along a specific path or area on the current or similar work plot from the agricultural machinery operation history database; calculating the path wear coefficient for different geographical locations within the work plot based on the number of repeated operations and intensity data; establishing an operation history table indexed by grid cells within the work plot, recording the cumulative number of operations and average intensity for each grid cell; inputting the cumulative number of operations and average intensity into a wear calculation function; the wear calculation function outputs a value between zero and one as the path wear coefficient for the grid cell, where a higher value indicates more severe path wear at the corresponding location of the grid cell; mapping the path wear coefficient to the geographical location corresponding to the initial guidance path network for unmanned agricultural machinery; adjusting the virtual potential energy value of the corresponding path segment in the initial guidance path network for unmanned agricultural machinery based on the mapped path wear coefficient; and setting a basic potential energy adjustment amount. Multiplying the path wear coefficient by the base potential energy adjustment yields the actual potential energy adjustment value for each grid cell. In the path potential energy field, the virtual potential energy value of grid cells with path wear is increased by the actual potential energy adjustment value, causing these grid cell regions to exhibit higher "potential energy" in path planning and guiding the path to detour. By adjusting the path potential energy value, an adaptive guidance path for unmanned agricultural machinery is generated that is more suited to actual ground conditions.
[0087] After the initial guidance path network for unmanned agricultural machinery is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial guidance path network to form an adaptive guidance path for unmanned agricultural machinery. This includes the following steps: Extracting data from the agricultural machinery operation history database on the number of times and intensity of agricultural machinery repeatedly operates along a specific path or area on the current or similar operation plots. The intensity data may include indicators such as the load weight, travel speed, or operation depth during operation. Calculating the path wear coefficient for different geographical locations within the operation plot based on the number of repeated operations and intensity data; establishing an operation history table indexed by grid cells within the operation plot; recording the cumulative number of operations and average intensity for each grid cell; in some embodiments, the operation history table is a two-dimensional data table isomorphic to the rasterized map, with each entry storing the cumulative number of operations and average intensity for the corresponding grid cell. The cumulative number of operations and average operation intensity are input into the wear calculation function. The wear calculation function outputs a value between zero and one as the path wear coefficient of the grid cell. The higher the value, the more severe the path wear at the corresponding location of the grid cell. In other words, the path wear coefficient quantifies the degree of ground compaction or trench deepening caused by historical operations. One expression of the wear calculation function is as follows:
[0088]
[0089] in: This represents the calculated path wear coefficient. This indicates the cumulative number of jobs performed by a grid cell. This represents the average workload of the grid cells. and These are the weighting coefficients for the cumulative number of tasks and the average task intensity, respectively. This represents the natural exponential function. For data comparison, assume the cumulative number of jobs for grid cell A is... For 10 times, the average workload The cumulative number of jobs in grid cell B is 0.5. For 50 times, the average workload is... It is 0.8, with the same weighting coefficient. and Below, the path wear coefficient of mesh cell A The calculated result is approximately 0.37, representing the path wear coefficient of mesh cell B. The calculated result is approximately 0.92, indicating that the path wear condition is more severe in grid cell B. The path wear coefficient is mapped to the geographical location corresponding to the initial guidance path network for unmanned agricultural machinery. In practice, the mapping process involves querying the grid cell containing each path point and reading the path wear coefficient corresponding to that grid cell. This is done by adjusting the virtual potential energy value of the corresponding path segment in the initial guidance path network for unmanned agricultural machinery based on the mapped path wear coefficient, and setting a basic potential energy adjustment amount. The path wear coefficient With the adjustment of the basic potential energy Multiplying yields the actual potential energy adjustment value for the grid cell. In the path potential energy field, the virtual potential energy value of the mesh cells exhibiting path wear is increased by the actual potential energy adjustment value. This allows grid cell regions with path wear to exhibit higher "potential energy" in path planning. In some embodiments, the base potential energy adjustment amount... The value is set to 20% of the original potential field average. The guidance path bypasses severely worn areas, thus generating an adaptive guidance path for unmanned agricultural machinery that is more adapted to actual ground conditions. Optionally, the path wear coefficient... Grid cells exceeding a set threshold (e.g., 0.8) can be marked as restricted areas in the guidance path network, in addition to increasing their virtual potential energy value. It can be understood that by dynamically introducing a path wear coefficient and adjusting the potential energy field, the adaptive guidance path for unmanned agricultural machinery can proactively avoid repeated travel in severely compacted areas, helping to protect soil structure and optimize overall operational efficiency.
[0090] See Figure 5In the dynamic correction process of adaptive guidance paths for unmanned agricultural machinery, the correlation analysis between path wear coefficient and average work intensity is a key parameter verification step. The figure uses average work intensity (0–1) as the horizontal axis and path wear coefficient (0–1) as the vertical axis, visually presenting the strong positive correlation between the two. From the data distribution, as the average work intensity increases, the path wear coefficient shows a significant upward trend: when the average work intensity is below 0.3, the path wear coefficient is mainly concentrated in the 0.1–0.2 range, indicating that low-intensity work has a limited impact on path wear; when the average work intensity increases to the 0.5–0.7 range, the path wear coefficient correspondingly climbs to 0.25–0.35, reflecting the direct driving effect of work intensity on path compaction and trench deepening; and when the average work intensity exceeds 0.8, the path wear coefficient generally reaches above 0.35, with some data points even approaching 0.45, confirming the engineering law that high-intensity work significantly exacerbates path wear. The distribution characteristics of the scatter plot provide empirical evidence for calibrating the average work intensity weighting coefficient in the wear calculation function. During the dynamic correction phase of path planning, this correlation allows the average work intensity to be used as one of the core input variables. The wear calculation function then generates a precise path wear coefficient, which in turn adjusts the path potential energy field, guiding agricultural machinery to actively avoid high-wear areas, thereby protecting soil structure and optimizing operational efficiency.
[0091] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A method for unmanned agricultural machinery operation path planning based on GPS positioning, characterized in that, The method includes: Obtain the real-time geographic coordinate point sequence of agricultural machinery obtained through the GPS positioning system; The real-time geographic coordinate point sequence is compared with the pre-stored set of geographic coordinates of the work plot boundary to determine the real-time positioning status of the agricultural machinery within the work plot. Based on the real-time positioning status, the calculation process for the work path density field within the work area is initiated to generate an initial density distribution map of the work paths, including: When the real-time positioning status is a point within the work area, the target planning area inside the work area is delineated with the point within the work area as the center. The target planning area is rasterized and divided into multiple uniform cell grids; A path density value is initialized for each cell grid, and the path density value is assigned based on the distance between the grid center point and the boundary of the work area and the soil type attribute of the grid; Traverse all cell grids and perform spatial smooth interpolation calculation on the path density value of each cell grid to generate an initial density distribution map of the operation path covering the entire target planning area; Based on the initial density distribution map of the work path, the path potential gradient algorithm is used to generate an initial guidance path network for unmanned agricultural machinery covering the internal area of the work plot, including: In the initial density distribution map of the work path, the path density values are converted into virtual potential energy values to construct a path potential energy field; In the path potential energy field, the starting point of the agricultural machinery is selected as the potential energy point, and the preset end point of the operation within the operation plot is selected as the potential energy sink point. Starting from the potential energy point, iteratively search along the negative gradient direction of the potential energy field along the path until the potential energy sink is reached, and record this search trajectory as a candidate guiding path. By changing the location of potential energy points or potential energy sinks, the iterative search process is repeated to generate multiple candidate guidance paths connecting different start and end positions of operations, thus forming an initial guidance path network for unmanned agricultural machinery. After the initial guidance path network for unmanned agricultural machinery is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial guidance path network for unmanned agricultural machinery, thus forming an adaptive guidance path for unmanned agricultural machinery. The agricultural machinery control unit controls the agricultural machinery to perform unmanned operations based on the aforementioned unmanned driving adaptive guidance path.
2. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 1, characterized in that, The real-time geographic coordinate sequence obtained by the agricultural machinery through the GPS positioning system includes: Receive continuous positioning signals from the agricultural machinery vehicle's GPS receiver, and decode them to obtain the agricultural machinery's real-time longitude and latitude data; The real-time longitude and latitude data are collected at a preset sampling frequency to form an ordered set of real-time geographic coordinate points; The set of real-time geographic coordinate points is filtered to remove abnormal coordinate points caused by GPS signal drift, resulting in a smoothed sequence of real-time geographic coordinate points.
3. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 2, characterized in that, The real-time geographic coordinate point sequence is compared with the pre-stored set of geographic coordinates of the work plot boundary to determine the real-time positioning status of the agricultural machinery within the work plot, including: Read the predefined set of geographic coordinates of the work site boundary, which defines the polygonal closed boundary of the work site; Calculate the positional relationship between the latest coordinate point in the real-time geographic coordinate point sequence and the closed boundary of the polygon of the work site; If the latest coordinate point is located inside the closed boundary of the polygon, then the instantaneous positioning status is a point within the work area; If the latest coordinate point is located outside the closed boundary of the polygon, then the instantaneous positioning status is an external point of the work area; If the latest coordinate point is located on the closed boundary of the polygon, then the instantaneous positioning status is the boundary point of the work area.
4. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 3, characterized in that, After the initial guidance path network for unmanned agricultural machinery is generated, a path wear coefficient based on historical agricultural machinery operation data is introduced to dynamically correct the initial guidance path network, forming an adaptive guidance path for unmanned agricultural machinery, including: Extract data on the number of times and intensity of agricultural machinery operations were repeated along a specific path or area on the current or similar work plots from the agricultural machinery operation history database. Based on the data on the number of repetitive operations and the intensity of operations, the path wear coefficients for different geographical locations within the work area are calculated. Areas with more repetitive operations and higher intensity of operations have larger path wear coefficients. The path wear coefficient is mapped to the geographical location corresponding to the initial guidance path network for unmanned agricultural machinery. Based on the mapped path wear coefficient, the virtual potential energy value of the corresponding path segment in the initial guidance path network for unmanned agricultural machinery is adjusted, and the path is raised or corrected to generate an adaptive guidance path for unmanned agricultural machinery that is more adapted to the actual ground conditions.
5. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 4, characterized in that, Calculating the positional relationship between the latest coordinate point in the real-time geographic coordinate point sequence and the closed boundary of the polygon of the work site includes: The calculation is performed using the ray method, drawing a ray horizontally to the right from the latest coordinate point; Calculate the number of intersections between the ray and each edge of the closed boundary of the polygon of the work site; If the number of intersection points is odd, then the latest coordinate point is determined to be inside the closed boundary of the polygon; If the number of intersection points is even, then the latest coordinate point is determined to be located outside the closed boundary of the polygon.
6. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 5, characterized in that, In the path potential energy field, the starting point of the agricultural machinery is selected as the potential energy source, and the predetermined end point within the work area is selected as the potential energy sink, including: From the real-time geographic coordinate point sequence of the agricultural machinery, select a coordinate point that meets the preset stability conditions as the current starting point of the agricultural machinery, i.e., the potential energy point; The coordinates of the endpoint of this unmanned operation are extracted from the task instructions and used as the potential energy sink. Both the potential energy point and the potential energy sink are located within the working area.
7. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 6, characterized in that, Based on the data on the number of repetitive operations and the intensity of operations, the path wear coefficient for different geographical locations within the work area is calculated, including: Establish a historical operation record table indexed by grid cells within the operation plot, recording the cumulative number of operations and average operation intensity for each grid cell; Input the cumulative number of operations and the average operation intensity into the wear calculation function; The wear calculation function outputs a value between zero and one as the path wear coefficient of the grid cell. The higher the value, the more severe the path wear at the corresponding position of the grid cell.
8. The method for planning unmanned agricultural machinery operation paths based on GPS positioning as described in claim 7, characterized in that, Adjusting the virtual potential energy value of the corresponding path segment in the initial guidance path network for unmanned agricultural machinery based on the mapped path wear coefficient includes: Set a basic potential energy adjustment amount; Multiply the path wear coefficient by the basic potential energy adjustment to obtain the actual potential energy adjustment value for the grid cell; In the path potential energy field, the virtual potential energy value of the grid cell with path wear is increased by the actual potential energy adjustment value, so that the grid cell region with path wear exhibits higher "potential energy" in path planning, guiding the path to detour.