Unmanned field sugarcane concentrated transportation vehicle operation path planning method and system

Abstract

The invention provides an unmanned field sugarcane concentrated transportation vehicle operation path planning method and system, and belongs to the technical field of farmland operation path planning, and the method comprises the steps: obtaining a marked aerial photography field image; constructing a grid map boundary environment model by marking edge detection and boundary extraction of an aerial field image, graph completion and rasterization processing; on the basis of an expansion algorithm, obstacle pixels in the grid map boundary environment model are traversed through structural elements, and a regularized obstacle contour aligned with the grid boundary is generated; constructing a grid map based on the grid map boundary environment model and the regularized obstacle contour; and on the basis of a semi-spring artificial potential field method and an A * algorithm, according to obstacle distribution, topographic features and operation targets in the grid map, generating an optimal path satisfying the operation of the sugarcane concentrated transportation vehicle. According to obstacle distribution, topographic features and operation targets, mechanical limitation of the minimum turning radius is taken into consideration in path planning, meanwhile, path optimality is guaranteed, errors are reduced, and the transportation efficiency and operation safety of the sugarcane concentrated transportation vehicle are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of farmland operation path planning, and particularly relates to an unmanned field sugarcane collection and transportation vehicle operation path planning method and system. BACKGROUND

[0002] With the development of modern agriculture, various unmanned agricultural implements have become a research hotspot. The application of the unmanned agricultural implements can not only solve the problem of insufficient rural labor force, but also greatly improve the efficiency of rural operation. The rationality and efficiency of path planning of agricultural implements will affect the precision and quality of field operation, and scientific and reasonable path planning is indispensable. In view of the complex and changeable field environment, the unmanned field sugarcane collection and transportation vehicle can bring very high efficiency to agricultural sugarcane harvesting. The path planning of the unmanned field sugarcane collection and transportation vehicle can be divided into global path planning and local path planning. The global path planning refers to finding the optimal path from the starting point to the ending point in the known field working environment in an offline state. The local path planning refers to obtaining the obstacle information on the path in real time by a perception sensor when the field working environment is unknown, and then avoiding the obstacles to obtain the optimal path through real-time calculation, focusing on safety and real-time performance.

[0003] At present, there are two path trajectory planning methods, Hamilton path search and geometric reasoning, for global path planning. There are relatively common and simple shortest tangent method, circular arc method, polynomial function and other algorithms for local path planning. However, the above algorithms do not consider the working width and minimum turning radius of various agricultural implements or sugarcane collection and transportation vehicles in the actual field environment, and the obstacle avoidance strategy is too simple, without considering the complex field working environment and systematically analyzing and researching different obstacles, resulting in certain errors in path planning and affecting the transportation efficiency. SUMMARY

[0004] In order to solve the problem that the existing path planning method is not suitable for sugarcane field operation, the application provides an unmanned field sugarcane collection and transportation vehicle operation path planning method.

[0005] In order to achieve the above purpose, the application provides the following technical scheme: An unmanned field sugarcane collection and transportation vehicle operation path planning method comprises the following steps: Obtaining a sugarcane field block aerial image and a field block parameter, marking a representative sugarcane field working environment block in the aerial image to obtain a marked aerial field block image; Edge detection and boundary extraction are performed on the marked aerial images of fields. The extracted edges are used as the boundary conditions of the environment modeling map. Mathematical analysis is performed on the edges to complete the marked aerial images of fields into regular shapes. The grid unit size is determined according to the plot parameters. The completed regular shapes are rasterized according to the grid unit size. The completed boundaries are subjected to binary dilation to obtain the raster map boundary environment model. Based on the dilation algorithm, obstacle pixels in the grid map boundary environment model are traversed through the structuring element. If there is an obstacle within the coverage area of ​​the structuring element, the center pixel is marked as an obstacle. The neighboring grid is traversed through the structuring element to generate a regularized obstacle outline aligned with the grid boundary. A raster map is constructed based on the raster map boundary environment model and regularized obstacle contours. Based on the semi-spring artificial potential field method and the A* algorithm, the optimal path for sugarcane collection and transportation vehicles is generated according to the obstacle distribution, terrain features and operation objectives in the grid map.

[0006] Preferably, the marked aerial field images include irregular plots with sloping sides, plots with obstacles and missing parts, and irregular plots with curves; the grid map boundary environment model includes a grid map model with regularly distributed obstacles, a grid map model with randomly distributed obstacles, a grid map model with obstacles forming narrow passages, and a grid map model with non-rectangular irregular environments.

[0007] Preferably, the plot parameters of the sugarcane field include the total area of ​​the farmland. The total area of ​​the farmland is calculated by measuring the length and width of the farmland based on the actual scope of farmland operations.

[0008] Preferably, the Canny algorithm is used to perform edge detection and boundary extraction on the marked aerial field images, and the extracted edges are used as boundary conditions for the environmental modeling map. Specifically, this includes the following steps: Import the marked aerial images of the fields into MATLAB, convert the marked aerial images of the fields into grayscale images, and use Gaussian convolution to smooth the converted images; Use the imread function to read the filtered image, and then use the rgb2ray function to convert the image to grayscale. Calculate the magnitude and directional gradient of the grayscale image in the x and y directions; compare the calculated magnitude and directional gradient in four directions, and perform non-maximum suppression-Canny1 edge detection on the image; If the detected pixel is higher than the high threshold, then the position is marked as a strong edge in the output matrix canny2. If the current pixel value is between the low threshold and the high threshold, and there are still strong edges in the neighborhood, then the current pixel is also marked as a strong edge, and the edge detection result is obtained.

[0009] Preferably, the modeling process for the boundary environment model of a raster map containing irregular plots with sloping edges specifically includes the following steps: The obtained area is used to complete the irregular plots with sloping sides. The completed map is then rasterized. Each grid cell in the raster map is set to 0.5 meters. The completed regular plots are then rasterized into a 30x20 raster map. Add the diagonal edge to the raster map. Set all non-existent plots to 1 according to the definition of the raster map, i.e., impassable black raster storing obstacles and other terrain environment information. Set all existing plots other than the plots to 0, i.e., passable and traversable raster. Set all plots traversed by the diagonal edge to impassable black raster storing obstacles and other terrain environment information according to the binary dilation method of the raster map, and obtain the raster map boundary environment model.

[0010] Preferably, it also includes updating the coverage value of sugarcane collection vehicles, specifically including the following steps: Using the grid environment, current coverage matrix, detection radius, and other parameters as input, the coverage function identifies grid points of the sugarcane transport vehicle within the detection range and gradually increases the coverage value of the grid points in the coverage matrix; for obstacle handling, the obstacle coordinates are traversed and the value of the corresponding point in the matrix is ​​fixed to a constant. The coverage function calculates the control parameters of the sugarcane collection vehicle based on the coverage matrix, identifies the grid point with the minimum coverage value within the coverage area, and calculates the angle between the current position and that point. If there are points where the coverage value is lower than the expected threshold, the function adjusts the control parameters to make the sugarcane transport vehicle move toward the area where the coverage is less than the preset value. The coverage function dynamically adjusts the control parameters Vxcoverage and Vycoverage based on the coverage value and the difference between the current coverage and the expected coverage, using coefficients g1 and g4 for adjustment. The resulting control parameters are used to guide the sugarcane transport vehicle to move toward the area where the coverage is lower than the preset value.

[0011] Preferably, the step of generating the optimal path for sugarcane collection and transport vehicles based on the semi-spring-type artificial potential field method and A* algorithm, according to the obstacle distribution, terrain features, and operational objectives in the grid map, includes the following steps: Initialize the grid map parameters and determine the initial position of the sugarcane collection vehicle, and set the parameters required by the path planning algorithm; Identify the grids around the sugarcane transport vehicle that it can travel on, calculate the coverage value of these grids, and select targets with coverage values ​​less than a set value. For targets with coverage values ​​less than a set value, the Manhattan calculation method of the A* algorithm is used to calculate the movement cost, and the grid movement with the lowest cost is selected. The A* algorithm is used to plan the sugarcane transport vehicle's journey from the current grid position to the target grid position, and the path is optimized by combining a semi-spring-type APF, a coverage value update strategy, and a new movement rule.

[0012] This invention also provides an unmanned sugarcane field transport vehicle operation path planning system, comprising: The image acquisition module is used to acquire aerial images of sugarcane fields and the plot parameters of sugarcane fields, mark representative plots of sugarcane field operation environment in the aerial images, and obtain marked aerial images of the fields. The raster model construction module is used to perform edge detection and boundary extraction on the marked aerial field images. The extracted edges are used as the boundary conditions of the environment modeling map. The edges are mathematically analyzed and processed to complete the marked aerial field images into regular shapes. The raster unit size is determined according to the plot parameters. The completed regular shapes are rasterized according to the raster unit size. The completed boundaries are binary dilated to obtain the raster map boundary environment model. The obstacle contour generation module is used to generate a regularized obstacle contour that is aligned with the grid boundary by traversing obstacle pixels in the grid map boundary environment model based on the dilation algorithm and using structuring elements. If there is an obstacle within the coverage area of ​​the structuring element, the center pixel is marked as an obstacle. The module then traverses the neighboring grid by structuring elements to generate a regularized obstacle contour that is aligned with the grid boundary. The raster map construction module is used to construct raster maps based on the raster map boundary environment model and regularized obstacle contours. The path planning module is used to generate the optimal path for sugarcane collection and transportation vehicles based on the semi-spring artificial potential field method and A* algorithm, according to the obstacle distribution, terrain features and operation objectives in the grid map.

[0013] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in the unmanned field sugarcane collection and transportation vehicle operation path planning method.

[0014] The present invention also provides a computer-readable storage medium storing a computer program, which, when loaded by a processor, is capable of executing any of the steps in the unmanned field sugarcane transport vehicle operation path planning method.

[0015] The unmanned sugarcane collection and transportation vehicle operation path planning method provided by this invention has the following beneficial effects: This invention addresses the problems of existing algorithms that fail to consider working width, minimum turning radius, and simplistic obstacle avoidance strategies by integrating environmental modeling optimization with algorithmic innovation, thereby achieving path planning that is more in line with the actual conditions in sugarcane fields.

[0016] Firstly, it breaks through at the environmental modeling level, addressing the problem of insufficient consideration of actual field parameters in existing algorithms. It first acquires aerial images of sugarcane fields and plot parameters, marking representative operational environment plots. Then, through edge detection, boundary extraction, and mathematical analysis, it completes the images into regular shapes, laying a precise foundation for subsequent modeling. Next, based on the plot parameters, it determines the raster unit size, rasterizes the regular shapes, and performs binary dilation on the boundaries to construct a raster map boundary environment model. This step fully incorporates the actual size of the plots, allowing the model to implicitly include spatial requirements related to operational width, avoiding the problem of path planning ignoring the operational range.

[0017] Subsequently, an expansion algorithm was used to optimize obstacle handling, addressing the shortcomings of the original obstacle avoidance strategy. By traversing obstacle pixels in the grid model using structuring elements, obstacles were marked and regularized obstacle contours were generated. This enabled a systematic analysis of the spatial distribution of different obstacles in the field, moving beyond simple and vague obstacle avoidance processing and providing data support for subsequent precise obstacle avoidance.

[0018] Finally, the semi-spring-type artificial potential field method and the A algorithm are combined to generate the optimal path, solving the problems of minimum turning radius and path efficiency. The A algorithm excels at global optimization, while the semi-spring-type artificial potential field method can better handle motion constraints. The combination of the two can take into account the mechanical constraints of minimum turning radius in path planning based on obstacle distribution, terrain features, and operational objectives, while ensuring path optimality, reducing errors, and improving the transportation efficiency and operational safety of sugarcane collection vehicles. Attached Figure Description

[0019] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the unmanned sugarcane collection and transportation vehicle operation path planning method according to Embodiment 1 of the present invention; Figure 2 These are actual aerial images of sugarcane fields; Figure 3 To mark aerial images of field plots; Figure 4 This is a grayscale image after high-pass filtering; Figure 5 Non-maximum suppression: Canny1 edge detection results; Figure 6 Dual threshold monitoring: Canny2 edge detection results; Figure 7 This is the edge detection result from Canny; Figure 8 This is the result of Prewitt edge detection; Figure 9 This is the result of edge extraction for a single land parcel; Figure 10 It is an irregular plot of land with a sloping side; Figure 11 To complete the regular rectangular plot; Figure 12 For raster maps; Figure 13 This is a raster map after binary dilation; Figure 14 It is the unexpanded, original form; Figure 15 This is the result after expansion processing; Figure 16 For the final modeling result; Figure 17 Force diagram of an artificial potential field; Figure 18 A semi-spring type APF diagram; Figure 19 The parent node graph; Figure 20 Graph showing the calculation of movement costs; Figure 21 To continue searching the graph; Figure 22 A flowchart for path planning; Figure 23 Comparison of path lengths for different algorithms under different environments; Figure 24 Comparison of turning times for different algorithms under different environments Figure 25 Comparison of path repetition rates for different algorithms under different environments. Detailed Implementation

[0021] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0022] During the harvesting, transportation, and storage of sugarcane, the unique characteristics of sugarcane, such as the tendency of cut surfaces to oxidize and turn red, leading to poor sugar extraction, uneven and irregular growth, tilted growth angles, and discontinuous sugarcane planting plots by small and medium-sized farmers, place high demands on the efficiency of unmanned sugarcane collection vehicles. Therefore, this invention aims to propose a route planning method suitable for field sugarcane collection vehicles, designing the most efficient driving routes to solve efficiency problems and safety issues caused by human factors during sugarcane harvesting and transportation. This will improve the efficiency and quality of sugarcane transportation, reduce labor costs and energy consumption, serve the development of the sugarcane industry, reduce labor force, and lower labor costs.

[0023] The main contents of this invention are as follows: (1) Construction of Field Operation Environment Model: Based on the actual operation range of the farmland, its length and width are measured, and the total area of ​​the farmland is calculated. For irregular plots, edge detection algorithms can be used to extract the farmland boundaries, and then the plots are cropped into regular rectangular areas. After obtaining farmland environmental information in this way, the farmland environment is divided into two-dimensional grid cells of equal size. Each grid cell stores environmental information (such as obstacle information, terrain attributes, etc.). If the grid cell is less than one grid cell, it is expanded into one grid cell. This is used to construct a simulation map of the field operation environment. In this invention, not only are operation models constructed for three different types of field environments, namely "irregular plots with sloping sides", "plots with obstacles and missing parts", and "irregular plots with curves", but also four known grid map environments, namely "regularly distributed obstacles", "randomly distributed obstacles", "obstacles forming narrow passages", and "non-rectangular irregular environment", are constructed.

[0024] (2) Path planning algorithm design: For the four ideal grid map environments mentioned above, namely "obstacles are regularly distributed", "obstacles are randomly distributed", "obstacles form narrow passages", and "non-rectangular irregular environment", the fusion improved A* algorithm and the half-spring APF algorithm were used for path planning simulation. The innovation of full coverage value update strategy and movement rules is included. Coverage-based control parameters and obstacle avoidance are integrated together to guide the movement of sugarcane collection vehicles in grid-based environments. This ensures effective coverage of the field operation environment while avoiding obstacles. At the same time, it minimizes the path repetition rate and number of turns while achieving full coverage, improves the safety and efficiency of the path, and consumes less energy. The four known grid map environments were verified by four different algorithms. The simulation data proved the superiority of the fusion algorithm over the other three unimproved algorithms. At the same time, the basic field operation environment was simulated to ensure that the path planning algorithm has practical significance.

[0025] Example 1 This embodiment takes Guangxi Zhuang Autonomous Region as an example. As a region with a large sugarcane planting area, Guangxi Zhuang Autonomous Region's farmland is characterized by "predominantly dry land, severe fragmentation, and climate sensitivity." The farmland environment is characterized by scattered and discontinuous land, irregular shapes, irregular field ridges, and complex obstacle types, which brings great difficulties to path planning. The boundaries of the fields are mostly irregular polygons due to terrain constraints. This fragmented land pattern requires frequent turning adjustments during mechanical operations. However, when agricultural machinery operates in the field, its operating path is a full-area coverage path, which needs to meet the requirements of minimizing the number of turns, minimizing the path repetition rate, and maximizing the effective operating path ratio. Therefore, the global operating path of the unmanned sugarcane transport vehicle is mostly a parallel straight line. However, there are special situations where multiple turns are required for correction. The actual operating path generally appears in a mixed form of straight lines and various curves. Except for special cases such as U-turns, the operating path will not have intersections of straight lines or curves.

[0026] This invention uses irregularly shaped sugarcane fields as the research scenario. Addressing the complex terrain and varied obstacles in sugarcane fields in Southwest China, and considering the actual operational characteristics of the field environment and the specific needs of sugarcane harvesting, it acquires field images through drone aerial photography, extracts boundaries using Canny edge detection, and constructs a raster map model using binary morphological dilation processing. A path planning algorithm based on an improved A* algorithm and the elastic APF artificial potential field method is implemented to achieve a full-coverage assignment and update strategy for field path planning, solving the problem of full-coverage harvesting in irregular fields with various field obstacles under a known map.

[0027] Specifically, such as Figure 1 As shown, the unmanned sugarcane collection and transportation vehicle operation path planning method provided by the present invention includes the following steps: Step 1: Modeling the field operation environment in sugarcane fields Step 11: Obtaining the site for field operations in sugarcane fields Aerial images of sugarcane fields were obtained through drone aerial photography. Figure 2 After acquiring aerial images, representative sugarcane field operation environments within the aerial images are marked, resulting in marked aerial field images. Figure 3 The purple, red, and cyan plots represent "irregular plots with sloping sides", "plots with obstacles and missing parts", and "irregular plots with curves", respectively.

[0028] Step 12: Extraction of plot boundaries in the sugarcane field work environment The Canny algorithm is used to perform edge detection and boundary extraction on the marked aerial images of fields, and the extracted edges will be used as the boundary conditions of the environment modeling map.

[0029] Specifically, edge detection includes the following steps: The first step is to import the marked aerial images of the fields into MATLAB, convert them to grayscale images, and then use Gaussian convolution to smooth the converted images, such as... Figure 4 .

[0030] The image is read using the `imread` function, then converted to grayscale using the `rgb2ray` function. `Img = double(I)` converts the pixel values ​​from `uint8` (unsigned integer type) to `double` (double-precision floating-point number type), and `sz = size(I)` returns the image. `aif` and `n` set the standard deviation and size of the Gaussian kernel in the Gaussian filter. `n0 = floor((n+1) / 2)` calculates the coordinates of the Gaussian kernel center point. `h = zeros(n, n)` initializes the Gaussian kernel matrix. Calculate the Gaussian convolution kernel. `Img_n=uint8(conv2(Img, h, 'same'))` converts the image pixels back to uint8 type after the final Gaussian filtering.

[0031] The second step is to calculate the magnitude and directional gradients in the x and y directions.

[0032] The for function is used to loop through the ranges i and j from 2 to sz(1)-1 and sz(2)-1 to avoid accessing pixels outside the image boundary. Then, the Sobel operator is used to calculate the gradient in the x and y directions. The gradient magnitude and directional gradient are calculated by sqrt(Sx^2 + Sy^2) and atan2, respectively.

[0033] The third step is to compare in four directions and perform non-maximum suppression.

[0034] The gradient direction is divided into four main directions: horizontal (0° or 180°), 45°, vertical (90°), and 135°. Each main direction covers a range of ±22.5°. The 0° direction corresponds to [-22.5°, 22.5°) or [157.5°, 180°) and [0°, 22.5°), while the 45° direction corresponds to [22.5°, 67.5°). The direction is determined by the comparison pixels. M(i, j)>=M(i, j + 1)&&M(i, j)>= M(i, j - 1) corresponds to the horizontal direction of 0° or 180°, and M(i, j)>= M(i + 1, j - 1)&&M(i, j)>= M(i - 1, j + 1) corresponds to the 45° direction. The non-maximum suppression-Canny1 edge detection results are as follows: Figure 5 As shown.

[0035] The fourth step is dual-threshold monitoring and edge connectivity. The results of dual-threshold monitoring are as follows: Figure 6 As shown.

[0036] Calculate the low threshold lowTh and the high threshold higtTh. If a pixel is higher than the high threshold higtTh, mark that position as a strong edge in the output matrix canny2 (assign a value of 1). If the current pixel value is between the low threshold lowTh and the high threshold, and there are strong edges in the neighborhood, mark the current pixel as a strong edge as well (assign a value of 1).

[0037] Step 5: Obtain the edge detection results ( Figure 7 ): After obtaining the image from the Canny edge detection, compare it with the image obtained by the Prewitt operator ( Figure 8 In comparison, it is clear that the image obtained using Canny edge detection has clearer edges, stronger noise resistance, and higher positioning accuracy. Therefore, the Canny algorithm is used for image edge extraction.

[0038] Step 6: Edge detection and extraction for individual plots: Using an image coordinate picking tool, pixels were extracted from the image for "irregular plots with sloping sides," "plots with obstacles and missing parts," and "irregular plots with curves." After obtaining the pixels, the target region image was obtained using the `imageReduced = imcrop(originalImage)` function. Then, Canny edge detection was used to obtain the images separately. The results are as follows. Figure 9 .

[0039] Step 13: Grid-based mapping of the sugarcane field operation environment.

[0040] After obtaining the edge of the sugarcane field operation environment, it is also necessary to know the specific size and area of ​​the sugarcane plot in the aerial photograph. This requires carrying an RTK module and a high-resolution drone (such as DJI Phantom 4 RTK, Mavic 3, etc.), and combining it with ground station GCP and GIS tools to calculate the result.

[0041] Taking the purple plot, an "irregular plot with a sloping side," as an example, ideally, it would be a rectangle with a length of 15 meters, a width of 10 meters, and a right side length of 11.5 meters. Figure 10 It's easy to calculate that its area is 150 square meters. Use this area to complete the "irregular plot with a hypotenuse," as shown below. Figure 11 The completed map is then rasterized, with each grid cell set to 0.5 meters. The completed regular plots, after rasterization, will form a 30x20 grid map. Figure 12 As shown.

[0042] Then, the "hypotenuse" is added to the raster map. All previously uncompleted plots are set to 1 according to the raster map definition, representing impassable black raster cells storing obstacle and other terrain information. Existing plots outside the completed plots are set to 0, representing passable and traversable raster cells. Plots traversed by the "hypotenuse" are also set to impassable black raster cells storing obstacle and other terrain information using the raster map's binary dilation method. The final result is as follows: Figure 13 .

[0043] This allows us to obtain the raster map environment boundary model for the purple plot, which is an irregular plot with a sloping edge. The same method can be used to model the environment boundaries of the other two terrain types: red plots (plots with obstacles and missing parts), cyan plots (irregular plots with curves), and other field environment plots. Specifically, we first use a drone to obtain aerial images and plot parameters. Then, we use Canny edge detection to extract plot edges, perform mathematical analysis on these edges to complete them into a rectangular image, determine the raster unit size based on the plot parameters, and rasterize the image. Areas that have been completed are defined as impassable black grids, while previously existing areas are defined as passable white grids (except where obstacles were present). Finally, we perform binary dilation on the completed boundaries, thus completing the raster map boundary environment model.

[0044] Step 14: Handling obstacles in the sugarcane field work environment.

[0045] This invention employs a rasterized environment modeling method, dividing the harvester's operating space into raster cells of equal size. Obstacle areas are marked with black grids, while passable areas are marked with white grids. To address the issue of blurred boundaries for irregular obstacles, a dilation operator from binary morphology is introduced. Binary dilation is a fundamental operation in mathematical morphology, essentially traversing obstacle pixels in the map using structuring elements (such as 3×3 or 5×5 kernels). If an obstacle exists within the coverage area of ​​the structuring element, the central pixel is marked as the obstacle. The mathematical expression is: ; Where B is the original set of obstacles, A is the structuring element, Â is the mapping of A about the origin, and S is the expansion result. Each grid cell represents a unit length, and the size of the entire grid map is the unit length represented by the XY axis. The expansion radius determines the distance the obstacles expand. The actual lengths corresponding to the unit length and expansion radius need to be calculated based on the actual map and the size of the agricultural machinery.

[0046] By traversing the neighboring grid using structuring elements, a regularized obstacle profile is generated that is strictly aligned with the grid boundary. Figure 14 In its original form, Figure 15 (The result is shown in the original text). This operation, based on the safe movement radius of the sugarcane transport vehicle, expands the area occupied by obstacles to avoid the path planning getting stuck in local optima.

[0047] Similarly, for irregular sugarcane farmland plots in the environment, this invention can also use the binary dilation method to set the boundary of the sugarcane farmland with black grids, constraining the sugarcane collection and transportation vehicle to operate in irregular environmental plots, so as to achieve the effect of full coverage operation of sugarcane collection and transportation vehicle in less than ideal field environment.

[0048] Step 15: Generate the field operation environment for sugarcane fields.

[0049] In MATLAB raster map modeling, the first step is to determine the map parameters, including the map extent and resolution. `map_width` determines the number of rows in the map raster, `map_height` determines the number of columns, and `resolution` determines the map resolution (i.e., the actual length of each raster). For ease of drawing, this invention draws the raster map as a 30x30 raster map with a resolution of 1.5 (i.e., 1.5 units per raster). The actual parameters should be determined according to the size of the plot.

[0050] Next, the impassable areas (obstacles) in the grid are set. Here, the present invention uses the obstacles matrix to set the obstacle map. The two rows on the left, from top to bottom, determine the x and y coordinates of the obstacle, and the two rows on the right, from top to bottom, determine the width and height of the obstacle. Position passes the values ​​processed by the obstacles matrix to rectangle to draw the obstacle graphic. FaceColor determines the color of the drawn obstacle, and the last three parameters are the RGB three-color ratio.

[0051] Finally, use `figure;imagesc(occupancyGrid)` to create a new window to display the raster map and draw obstacles. `colormap` determines the color of the passable and obstacle areas of the raster map. `axis equal` forces the x-axis and y-axis to scale consistently, ensuring that the raster map is not stretched. `xlabel` and `title` set the axis labels and the raster map title, respectively.

[0052] The following section describes the environmental raster modeling of the red and cyan plots mentioned earlier, representing "plots with obstacles and missing parts" and "irregular plots with curves," respectively. A 3x3 obstacle is added to the raster map of the cyan plot ("irregular plot with curves") at (20,6) for testing. The final result is shown below. Figure 16.

[0053] This invention employs methods such as plot parameter acquisition (using RTK drones), edge detection, rasterization (0.5-meter resolution), and obstacle regularization (expanding obstacle areas using a dilatation algorithm). Ultimately, raster maps of three typical plot types (including hypotenuses, curves, and locally missing plots) are constructed, and the modeling effectiveness is verified using MATLAB.

[0054] Step 2: Improve the path planning algorithm.

[0055] After completing the grid environment modeling of farmland plots, path planning algorithms are the core technology for achieving autonomous navigation and efficient operation of agricultural machinery. This invention selects the fusion of the A* algorithm and the Artificial Potential Field (APF) method as the path planning algorithm.

[0056] (1) Artificial potential field method The basic idea of ​​the artificial potential field method is to construct a repulsive force field around the obstacle and a gravitational potential field around the target point, similar to an electromagnetic field in physics. The controlled object is subjected to repulsive and gravitational forces in the composite field composed of these two potential fields. The resultant force of the repulsive and gravitational forces guides the movement of the controlled object and searches for a collision-free obstacle avoidance path. More intuitively, the potential field method compares the obstacle to a mountain peak with a high potential energy value on a plain, while the target point is a valley with a low potential energy value.

[0057] exist Figure 17 In the middle, obstacle 2 gave the sugarcane transport vehicle F req2 The repulsive force of obstacle 1 caused the sugarcane transport vehicle F to give way to the obstacle 1. req1 The repulsive force, at the same time the target point is given to the sugarcane transport vehicle F att The resultant force F obtained by the sugarcane transport vehicle in the entire potential field is obtained by analyzing the force of the attraction, the repulsive force of the two obstacles, and the attraction of the target point. The direction of this resultant force is the direction of the robot's movement. However, the APF algorithm is prone to getting stuck in local optima when the repulsive and attractive forces are balanced, causing the sugarcane transport vehicle to stop moving, i.e., a local extremum problem. In areas with dense obstacles, the interaction of the repulsive force field may also cause the robot to be unstable, affecting the smoothness of the path and the reachability of the target.

[0058] To address the aforementioned issue—that the APF algorithm is prone to getting stuck in local extrema and thus stagnates—the A* algorithm is integrated with the APF algorithm to solve the problem, thus preventing the sugarcane transport vehicle from getting stuck in a dead zone loop.

[0059] 1. Calculation of gravitational and repulsive potential fields: The gravitational potential field is mainly related to the distance between the sugarcane transport vehicle and the target point. The greater the distance, the greater the potential energy experienced by the sugarcane transport vehicle; the smaller the distance, the smaller the potential energy experienced by the sugarcane transport vehicle. Therefore, the function of the gravitational potential field is: ; Where η is the proportional gain coefficient, p(q,q) g Let q be a vector representing the position q of the sugarcane transport vehicle and the position q of the target point. g Euclidean distance between them | qq g | The vector direction is from the position of the sugarcane transport vehicle to the target point.

[0060] The corresponding gravitational force F att (X) represents the negative gradient of the gravitational field: ; The factor determining the repulsive potential field of the obstacle is the distance between the sugarcane transport vehicle and the obstacle. When the sugarcane transport vehicle is not within the obstacle's influence range, its potential energy is zero. After the sugarcane transport vehicle enters the obstacle's influence range, the greater the distance between the two, the smaller the potential energy of the sugarcane transport vehicle; the smaller the distance, the greater the potential energy of the sugarcane transport vehicle. The potential field function of the repulsive potential field is: ; Where k is a proportionality coefficient. ( Let be a vector, with a direction from the obstacle to the sugarcane collection vehicle, and a magnitude equal to the distance between the sugarcane collection vehicle and the obstacle. |, is a constant representing the maximum distance at which the obstacle affects the sugarcane transport vehicle.

[0061] The corresponding repulsive force is the negative gradient of the repulsive field: ; Let the position of the sugarcane transport vehicle be (x, y), and the position of the obstacle be (x, y). g， y g Then the function of the gravitational potential field can be transformed into: ; The repulsive field function is: ; 2. Improved artificial potential field algorithm The defects of the artificial potential field method generally include two situations: First, the problem of the target point being unreachable. Because the distance between the obstacle and the target point is too close, when the sugarcane transport vehicle reaches the target point, according to the potential field function, the attraction of the target point drops to zero, while the repulsive force of the obstacle is not zero. At this time, although the sugarcane transport vehicle reaches the target point, it cannot stop under the action of the repulsive force field, thus leading to the problem of the target being unreachable.

[0062] The second problem is getting stuck in a local optimum. When the vehicle is at a certain position, if the combined repulsive force of several obstacles is equal in magnitude and opposite in direction to the gravitational force of the target point, the resultant force is 0. This will cause the vehicle to no longer be "forced" and thus cannot search for a path to avoid obstacles.

[0063] To solve these two problems, considering the actual application of this invention, since every non-boundary, non-obstacle grid needs to be traversed during the full-coverage path search of the sugarcane transport vehicle, the gravitational potential field in the APF can be removed, leaving only the repulsive potential field between obstacles and boundaries. For sugarcane transport vehicles approaching the boundary, the repulsive force of the wall is introduced to prevent collisions. Using a half-spring method, the magnitude of the repulsive force is proportional to the distance from the sugarcane transport vehicle to the nearest boundary, and its direction points towards the boundary. (The repulsive force is proportional to the distance between objects; the closer the distance, the greater the repulsive force. This is similar to the rebound effect of a spring when compressed. One difference of the half-spring method is that when the distance exceeds a certain threshold, the repulsive force does not continue to increase linearly but remains a constant value, effectively preventing the repulsive force from becoming infinitely large.) The horizontal and vertical coordinate components of the repulsive force between the sugarcane transport vehicle and the obstacle, and the repulsive force between the sugarcane transport vehicle and the wall, are added together to obtain the total repulsive force of the sugarcane transport vehicle. The magnitude of the total repulsive force is calculated to adjust the speed of the sugarcane transport vehicle. Based on the magnitude and direction of the repulsive force, calculate the velocity components on the horizontal and vertical axes of the sugarcane transport vehicle. If the current time is 0, it means that the sugarcane transport vehicle is not moving and its velocity is zero. Otherwise, add the repulsive force between the sugarcane transport vehicle and the obstacle and the repulsive force between the sugarcane transport vehicle and the wall to obtain the final obstacle avoidance speed.

[0064] Set boundary repulsion coefficient Safety distance between sugarcane transport vehicles and the boundary Calculation of the boundary repulsive field: ; ; ; The total repulsive force is the resultant force of all the repulsive forces from obstacles and the boundary forces acting on the sugarcane transport vehicle, which is: ; ; The final force on the sugarcane transport vehicle is Figure 18 As shown.

[0065] 3. Specific Applications Calculate the repulsive force of the obstacle: First, we need to define some basic input parameters, including the position coordinates x and y of the sugarcane transport vehicle, the coordinate array matrix Obstacles of the obstacles, the repulsion distance and the range of repulsion force r and R of the wall, the gain coefficients g2 and g3 of the repulsion force, the size of a single grid d, the number of time steps t, and initialize the obstacle repulsion force in the X and Y directions.

[0066] Then, the repulsive force of the obstacle using the half-spring model was calculated. First, a for loop was used to scan the coordinates of the obstacle in the entire grid map, and then sqrt was used to calculate the distance between the sugarcane transport vehicle and the obstacle.

[0067] When the distance between the sugarcane transport vehicle and the obstacle is less than the range of the repulsive force R, the repulsive force is triggered. atan2 calculates the relative angle between the sugarcane transport vehicle and the obstacle to obtain the direction of the repulsive force opposite to the direction from the sugarcane transport vehicle to the obstacle. Then, the magnitude of the repulsive force (half spring) is calculated based on the distance between the sugarcane transport vehicle and the obstacle. g3 is the repulsion coefficient of the obstacle.

[0068] Using mechanical analysis, the repulsive force of the obstacle is decomposed into forces along the X and Y axes: Calculate the boundary repulsion: Initialize the repulsive forces of the boundary walls in the X and Y directions.

[0069] The boundary walls in four directions are detected. If the distance between the sugarcane transport vehicle and the X or Y boundary or the corresponding other boundary is less than the repulsion distance r of the wall, then a thrust in the opposite direction is applied, where g2 is the repulsion coefficient of the boundary.

[0070] Finally, the total repulsive force and velocity of the sugarcane transport vehicle were calculated.

[0071] (2) A* algorithm 1. Preprocessing The A* (A-Star) algorithm is a highly efficient direct search method for finding the shortest path in static road networks, and it is also an effective algorithm for solving many search problems. It is widely used in indoor robot path search, game animation path search, etc. It combines a greedy algorithm (depth-first search) and Dijkstra's algorithm (breadth-first search), and is a heuristic search algorithm.

[0072] The formula for evaluating the quality of a path is: ; f(n) is the cost estimate from the initial state to the target state via state n, g(n) is the actual cost from the initial state to state n in the state space, and h(n) is the estimated cost of the optimal path from state n to the target state. Two state tables are used, called the openList and the closeList. The openList consists of nodes to be examined, and the closeList consists of nodes that have already been examined.

[0073] Initially, node A is defined as the parent node, with a distance of 0 from itself, indicating a completely determined path. It is then moved into the `closeList`. There are 8 nodes surrounding parent node A, defined as child nodes. These child nodes are added to the `openList` to be examined. If a node is neither in the `openList` nor the `closeList`, it means that the node has not yet been found. The path quality is judged based on the movement cost. The cost of a single movement is calculated using the Manhattan method, where the cost of moving a node horizontally or vertically is defined as 10, and the cost of diagonal movement is calculated using the method for calculating the hypotenuse of an isosceles triangle (distance I), which is 14. Figure 19 .

[0074] 2. Start Search The present invention will now start from the following grid map ( Figure 20 Starting from node A in the diagram, the search ends at the target node J, with black grid cells representing obstacle areas. Let's take starting the search from node D as an example: Based on the A* path movement cost evaluation formula Therefore, since moving from grid point A to grid point D is a single-step diagonal movement, the actual cost is... =14; After moving to point D, since h(n) considers the optimal path estimate from the current state to the target state, it does not consider whether there are impassable (obstacle) areas in the grid near point D, and it needs to be calculated completely according to the Manhattan method, only calculating the cost of lateral and vertical movement, that is, from point D to point J, it is necessary to move three steps to the right and one step down, for a total of four steps. Therefore, h(n) = 40. So the path cost of moving from point A to point J is f(n) = 14 + 40 = 54.

[0075] Similarly, calculate the movement path cost of the remaining 7 child nodes around point A, and select the point with the smallest movement path cost to move into the closeList table.

[0076] The path cost calculation results for the eight child nodes surrounding parent node A are shown above. Figure 20 As shown, taking point F as an example, the number in the upper left corner represents the movement path cost f(n), and the two sets of numbers in the lower left corner from left to right represent the actual path cost g(n) and the estimated path cost f(n), respectively. It is easy to see that F is the byte point with the smallest movement cost. Select F from the openList table and put it into the closeList table, and then use it as the new parent node.

[0077] 3. Continue searching Then, all child nodes adjacent to node F are checked, ignoring impassable obstacle nodes and nodes already existing in the closeList (i.e., the initial parent node A). Only nodes D, I, H, and C remain. These child nodes are then selected, and the node with the smallest path cost f(n) is chosen from the openList and added to the closeList as the next new parent node. If an adjacent node is already in the open list, it is checked whether this path is better; that is, whether reaching that node via the current node (the node selected in this invention) has a smaller g(n) value. If not, no operation is performed. That is, the g(n) cost from node A to node I is 14, and the g(n) cost from node A to node F and then to node I is 20. If the g(n) cost from node A to node F and then to node I is less than the g(n) cost from node A to node I, then the path from A to node F is not optimal, and I will be chosen as the parent node; otherwise, no changes are made.

[0078] Because the path costs of nodes I and G are the same when F is the parent node, both will be searched. This invention takes searching node I first as an example. Nodes A and F, which have already been moved into the closed list, are ignored, as are impassable obstacles. Nodes K, L, and M, which were not initially searched, are added to the open list. The path movement costs of the child nodes surrounding node I are recalculated, and the one with the lowest cost is selected as the parent node. This process is repeated until the target T is found, completing the algorithm. Figure 21 .

[0079] 4. Specific Applications First, define the A* algorithm function. Input the current point, the set of all points, and obstacles. Initialize the minimum distance to the bisection of the map. Then, set the flag of the current point to 1, i.e., the parent node or the starting point. Then, initialize the target point, the points around the current point, the endpoint, and the path. Then, add the current point to the path (closeLis). Finally, use a loop counter (to prevent infinite loops).

[0080] The algorithm uses a double for loop to iterate through all points (all rows and columns) in the entire raster map to find the nearest uncovered point. Then, it uses an if statement to check whether the point is an obstacle or has been covered. Only if the detected point is neither an obstacle nor covered will the algorithm proceed to the next step of calculating the Euclidean distance between the uncovered point and the current point. The point with the smallest Euclidean distance is then selected as the target point, and the status and coordinates of the target point are updated.

[0081] Use a while loop to probe eight points around the starting point, represented by grid (1, 2, 3, 4, 5, 6, 7, 8). Determine if these points are within the map's boundaries and if there are no obstacles on the movement path.

[0082] Then remove the point at the previous position, that is, the point that has already been traversed and passed through its parent node.

[0083] Start by calculating the movement path cost f(n) of the eight points A* surrounding the current point one by one, and set the movement path of the first adjacent point as f_min. If there is a subsequent adjacent point with a smaller movement cost f(n), then refresh the minimum value and set f_min.

[0084] Move the current point into the traversed points last_position(closeList), move to the point f_min with the minimum move cost, mark the current point as covered by its parent node, pass the coordinates of the point into the Path, and update the path and the current point position.

[0085] At the end of each loop, the around_points set is cleared to ensure that the latest available points around the current point are collected again in the next loop.

[0086] Check if the coordinates of the current point (current_point) are exactly the same as the coordinates of the target point (target_point). If they are the same, set the current point (current) to the target point (target_point), indicating that the end point of the path has been reached.

[0087] (3) Coverage value update strategy This strategy is primarily used to update the coverage value of sugarcane transport vehicles. Taking the grid environment, current coverage matrix, detection radius, and other parameters as input, the `coverage` function identifies grid points within the detection range of the sugarcane transport vehicle and progressively increases the coverage value of these grid points in the coverage matrix (the value is limited to a maximum to prevent exceeding the desired coverage). For obstacle handling, the obstacle coordinates are traversed, and the corresponding point's value in the matrix is ​​fixed to a constant to ensure that obstacles do not interfere with the coverage calculation. The `coverage` function calculates the sugarcane transport vehicle's control parameters based on the coverage matrix, identifies the grid point with the minimum coverage value within the coverage range, and calculates the angle between the current position and that point.

[0088] If there are points where the coverage value is lower than the desired threshold, the function adjusts the control parameters to make the sugarcane transport vehicle move towards the area with lower coverage. The coverage function dynamically adjusts the control parameter V based on the coverage value and the difference between the current coverage and the desired coverage. x coverage and V y The coverage parameter, adjusted using coefficients g1 and g4, guides sugarcane transport vehicles towards areas with low coverage. A minimum coverage value is introduced; in cases of very low coverage, the function sets a minimum threshold to help prevent sugarcane transport vehicles from overly prioritizing areas with very low coverage.

[0089] This strategy integrates coverage-based control parameters and obstacle avoidance, guiding the movement of sugarcane transport vehicles in a grid-based environment while ensuring effective coverage and avoiding obstacles.

[0090] (4) Innovation of movement rules By checking the status of surrounding points, the next movement direction is selected. If the lower left and lower left are not empty, but the left side is empty, the movement is directed to the lower left. If the upper left and upper right are not empty, but the left side is empty, the movement is directed to the upper left. When the vehicle reaches the last row of the map and is at a T-junction, it enters the dead zone with the smallest relative path. The dead zone with the smallest relative path here refers to a situation where the left, lower, and lower left sides are not empty, while the upper and right sides are empty. By determining whether the current position is on the edge of the map and its relative position, the direction to enter the dead zone is selected, thus reducing repeated paths. This movement rule makes the sugarcane transport vehicle's path planning more flexible and adaptable to different environment requirements. Especially when dealing with T-junctions, selecting the dead zone with the smallest relative path allows for more effective obstacle avoidance.

[0091] Step 3: Perform path planning based on multiple algorithms, including the A* algorithm and the artificial potential field method.

[0092] This invention combines a semi-spring-type artificial potential field method with the A* algorithm. This algorithm generates an optimal path for sugarcane transport vehicles based on obstacle distribution, terrain features, and operational objectives in a grid map. The algorithm incorporates a coverage value update strategy and innovates the movement rules, simulating a scenario of achieving full-coverage path planning in irregular and obstacle-filled field environments. This enables efficient obstacle avoidance and full-coverage harvesting in complex farmland environments. The specific process is as follows: Figure 22 As shown, the algorithm steps are as follows: 1) Simulation map environment construction: Initialize the grid map parameters and determine the initial position of the sugarcane collection vehicle, and set the parameters required by the path planning algorithm.

[0093] 2) Calculate the driving grid path: Determine the grids around the sugarcane collection vehicle that can be driven, calculate the coverage value of these grids, and select the targets with lower coverage values.

[0094] 3) Target grid selection: For targets with lower coverage values, calculate the movement cost using the Manhattan calculation method of the A* algorithm, and select the grid with the lowest cost to move.

[0095] 4) Path optimization: The A* algorithm is used to plan the sugarcane transport vehicle from the current grid position to the target grid position, and the path is optimized by combining the semi-spring APF, the coverage value update strategy, and the new movement rules.

[0096] 5) Target completion check: If the grid map covering the entire simulated field environment has been completed, the algorithm ends; if it has not been fully covered, return to step 2) to recalculate and cover.

[0097] This invention focuses on the design of a path planning algorithm for sugarcane transport vehicles in sugarcane fields. For irregular plots and multi-obstacle scenarios, it proposes a solution that integrates an improved A* algorithm and the semi-spring artificial potential field (APF) method. Traditional APF is prone to getting stuck in local optima and the target becomes unreachable. By introducing a semi-spring model to adjust the repulsion force calculation (repulsion force increases non-linearly with distance, and boundary repulsion force is handled independently), it effectively balances obstacle avoidance and path continuity. The A* algorithm pre-calculates the globally optimal path based on Manhattan distance and dynamically adjusts the priority of target points using a coverage value update strategy to ensure efficient coverage. The algorithm's innovations include: coverage value-driven (guiding sugarcane transport vehicles to prioritize traversing low-coverage areas), optimized movement rules (reducing repeated paths at T-junctions), and multi-algorithm collaboration (A* global planning + APF dynamic obstacle avoidance).

[0098] The following path planning simulation analysis verifies the unmanned sugarcane collection vehicle operation path planning method proposed in this invention.

[0099] This invention establishes four different obstacle grid maps and compares them with three control group algorithms and one experimental group algorithm. The four different obstacle grid maps are: a grid map with regularly distributed obstacles, a grid map with randomly distributed obstacles, a grid map with obstacles forming narrow passages, and a grid map of irregular environments. The algorithms compared are: 1. A semi-spring APF algorithm with a coverage update strategy and innovative movement rules (removing A*); 2. A combination of traditional APF and A* with a coverage update strategy and innovative movement rules (removing the improved APF); 3. A semi-spring APF algorithm with A* and innovative movement rules (removing the coverage update strategy); 4. The fusion algorithm proposed in this invention (A* + semi-spring APF + coverage update + innovative movement rules). The experiment uses a 40x40 grid map, with the starting point set to (1,1). Each algorithm is run 20 times in the four different scenarios to ensure the stability and reliability of the results. Furthermore, the algorithm's performance metrics are qualitatively defined as Nice Path (covered path), Path Length (total path length), Turn Number (number of turns), and Repeat Rate (path repetition rate). The final experimental comparison chart is shown below. Figure 23 , 24 As shown in Figure 25.

[0100] The line graphs showing the combined performance of the four algorithms in four different scenarios reveal that the experimental group's fusion algorithm exhibits optimal performance across all four scenarios. It achieves the shortest path length and lowest repetition rate, and significantly improves robustness in complex environments (such as irregular scenarios with slanted sides) through a multi-strategy collaborative mechanism, optimizing the number of turns by 10%-15%. In contrast, the control group's algorithm 3 performs the worst in dynamic and complex scenarios. While algorithms 1 and 2 maintain basic functionality in regular scenarios, they exhibit insufficient flexibility in narrow passages or irregular environments. This fusion algorithm balances path efficiency and stability through a strategy combining a semi-spring-type APF with an A-star. Its low error rate and low repetition rate are particularly suitable for practical scenarios with high operational accuracy requirements, such as agriculture and warehousing, highlighting its strong practical application potential.

[0101] This invention validates the effectiveness of the fusion algorithm through four typical scenarios (regular obstacles, random obstacles, narrow passages, and irregular environments with sloping sides). The experimental group (A* + half-spring APF + coverage value update + movement rules) was compared with three control groups. The results showed that in the regular obstacle scenario, the experimental group reduced path length by 2.15% and the number of turns by 11.7%; in the irregular environment with sloping sides, the path repetition rate decreased by 4.49%, and the number of turns was optimized by 30.21%. The fusion algorithm prioritizes traversing low-coverage areas by dynamically adjusting the coverage value, enhances obstacle avoidance robustness by combining a half-spring repulsion model, and innovates movement rules to reduce repeated paths at T-junctions. Its overall performance is significantly better than traditional methods. Simulation data shows that the algorithm balances path efficiency (shortest length), operational accuracy (lowest repetition rate), and motion stability (fewest turns) in complex farmland environments, providing reliable technical support for practical applications.

[0102] Simulation experiments based on MATLAB in four typical scenarios (regularly distributed obstacles, randomly distributed obstacles, narrow passages, and irregular boundaries) demonstrate that the proposed scheme can achieve collision-free coverage of the entire field, reducing path length by 0.41%-4.49%, decreasing repetition rate by 0.43%-6.85%, and reducing the number of turns by up to 30.21% in complex environments, while significantly improving robustness against sloping terrain. In the four typical scenarios of regularly distributed obstacles, randomly distributed obstacles, narrow passages, and irregular environments, it exhibits the following significant advantages: (1) The path coverage rate and obstacle avoidance success rate both reach 100%; (2) The path length is shortened by up to 6.8% compared with the traditional method, and the repetition rate is reduced to below 0.5; (3) In the irregular scenario with sloping sides, the multi-strategy collaboration mechanism greatly improves the planning robustness and optimizes the number of turns by 17%-30%.

[0103] This invention provides an efficient path planning algorithm for unmanned sugarcane transport vehicles, which effectively improves operational efficiency, reduces labor intensity, and has significant application value for agricultural mechanization in complex terrain.

[0104] The path planning method proposed in this invention has the following advantages: (1) Improve sugarcane harvesting and transportation efficiency: Traditional sugarcane harvesting and transportation methods mainly rely on manual labor, which is labor-intensive and inefficient. By designing a path planning system for an unmanned sugarcane collection and transportation vehicle in the field, the harvesting and transportation routes can be optimized, greatly improving transportation efficiency and saving labor costs.

[0105] (2) Reduce sugarcane harvesting and transportation costs: Unmanned sugarcane collection vehicles in the field can achieve automatic driving functions, reduce human error, reduce losses during transportation, find the optimal route by algorithm, and save fuel and energy. In addition, automated route planning can reduce the need for drivers, further reducing harvesting and transportation costs.

[0106] (3) Providing a reference for the transportation of other crops: Although this invention uses sugarcane as an example, the route planning of unmanned sugarcane collection and transportation vehicles in the field can also be applied to the harvesting and transportation route planning of other crops, such as rice and corn. Therefore, this invention has certain universality and promotional value. It can be seen that the route planning of unmanned sugarcane collection and transportation vehicles in the field is a research topic with important practical significance.

[0107] In summary, the path planning for unmanned sugarcane collection vehicles in the field has significant theoretical and practical value. This path planning method combines the inherent conditions of the sugarcane collection vehicle with external environmental factors, and integrates global and local obstacle avoidance algorithms. It minimizes path length to reduce battery consumption time and maximizes coverage area, ensuring high efficiency in field operations. By selecting appropriate path optimization algorithms, simulation, and real-time data tracking and feedback from the sugarcane collection vehicle, it gains a certain degree of real-time adjustment capability, ultimately planning an optimal route that combines efficiency, safety, and practicality. This route system can not only significantly improve field operation efficiency and reduce costs, but also accelerate the economic development of Guangxi's sugarcane industry, narrow the technological gap between domestic and international practices, and gradually achieve fully automated sugarcane harvesting and transportation. This will improve the relevant technological level of my country's sugarcane industry, ensure product quality and market competitiveness, provide theoretical direction and technical foundation for future sugarcane production research, and also have certain reference value for the fully automated systematic harvesting of other crops.

[0108] Based on the same inventive concept, this invention also provides an unmanned sugarcane field transport vehicle operation path planning system, comprising: The image acquisition module is used to acquire aerial images of sugarcane fields and the plot parameters of sugarcane fields, mark representative plots of sugarcane field operation environment in the aerial images, and obtain marked aerial images of the fields.

[0109] The raster model construction module is used to perform edge detection and boundary extraction on the marked aerial field images. The extracted edges are used as the boundary conditions of the environment modeling map. The edges are mathematically analyzed and processed to complete the marked aerial field images into regular shapes. The raster unit size is determined according to the plot parameters. The completed regular shapes are rasterized according to the raster unit size. The completed boundaries are binary dilated to obtain the raster map boundary environment model.

[0110] The obstacle contour generation module is used to generate regularized obstacle contours aligned with the grid boundaries by traversing obstacle pixels in the grid map boundary environment model based on the dilation algorithm and using structuring elements. If there is an obstacle within the coverage area of ​​the structuring element, the center pixel is marked as an obstacle. The module then traverses the neighboring grids using structuring elements to generate regularized obstacle contours that are aligned with the grid boundaries.

[0111] The raster map building module is used to construct raster maps based on the raster map boundary environment model and regularized obstacle contours.

[0112] The path planning module is used to generate the optimal path for sugarcane collection and transportation vehicles based on the semi-spring artificial potential field method and A* algorithm, according to the obstacle distribution, terrain features and operation objectives in the grid map.

[0113] The modules in the aforementioned unmanned sugarcane collection and transportation vehicle operation path planning system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0114] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps in the embodiment of the unmanned sugarcane collection and transportation vehicle operation path planning method. Specific implementation methods can be found in the method embodiments, and will not be repeated here.

[0115] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions, on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in the embodiment of the unmanned sugarcane field transport vehicle operation path planning method. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0116] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for planning a work path of an unmanned field sugarcane collection vehicle, characterized in that, The method comprises the following steps: Obtaining a sugarcane field block aerial image and a field block parameter, marking a representative sugarcane field work environment block in the aerial image, and obtaining a marked aerial field block image; Edge detection and boundary extraction are performed on the marked aerial field block image, the extracted edge is taken as a boundary condition of an environment modeling map, mathematical analysis and processing are performed on the edge, the marked aerial field block image is completed into a regular graph, the size of a grid unit is determined according to the field block parameter, the completed regular graph is rasterized according to the size of the grid unit, binary dilation processing is performed on the completed boundary, and a grid map boundary environment model is obtained; Based on the dilation algorithm, the obstacle pixels in the grid map boundary environment model are traversed through a structure element, if there is an obstacle in the coverage range of the structure element, the center pixel is marked as an obstacle, and the neighborhood grid is traversed through the structure element, and a regular obstacle contour aligned with the grid boundary is generated; Based on the grid map boundary environment model and the regular obstacle contour, a grid map is constructed; Based on the semi-spring artificial potential field method and the A* algorithm, an optimal path meeting the sugarcane collection and transportation vehicle operation is generated according to the obstacle distribution, terrain characteristics and operation target in the grid map.

2. The method according to claim 1, wherein, The marked aerial field block image includes an irregular block containing a bevel, a block with obstacles and lacking part of the block, and an irregular block containing a curve; the grid map boundary environment model includes an obstacle regular distribution grid map model, an obstacle random distribution grid map model, an obstacle narrow channel grid map model and a non-rectangular irregular environment grid map model.

3. The method according to claim 1, wherein, The field block parameter of the sugarcane field block includes the total area of the farmland, the length and width of the farmland are measured according to the actual operation range of the farmland, and the total area of the farmland is calculated.

4. The method of claim 1, wherein, Edge detection and boundary extraction are performed on the marked aerial field block image through the Canny algorithm, the extracted edge is taken as a boundary condition of an environment modeling map, and the specific steps include the following steps: The marked aerial field block image is imported into MATLAB, the marked aerial field block image is converted into a gray image, and the converted image is subjected to smoothing filtering through Gaussian convolution; The filtered image is read by using the imread function, and then the image is converted into a gray image by using the rgb2gray function; The amplitude and direction gradient of the gray image in the x and y directions are calculated; the calculated amplitude and direction gradient are compared in four directions, and non-maximum suppression-Canny1 edge detection is performed on the image; If the detected pixel is higher than the high threshold value, the position is marked as a strong edge in the output matrix canny2; if the current pixel value is between the low threshold value and the high threshold value, and there is a strong edge in the neighborhood, the current pixel is also marked as a strong edge, and the edge detection result is obtained.

5. The method according to claim 3, wherein, The modeling process of the grid map boundary environment model of the irregular block containing a bevel includes the following steps: The irregular block containing a bevel is completed by using the obtained area, and the completed map is rasterized, the unit of each grid of the rasterized grid map is set to 0.5 meters, and the rasterized regular block is a 30X20 grid map; The bevel is added to the grid map, and the originally non-existing completed land is all set to 1 according to the definition of the grid map, that is, the non-passable black grid for storing obstacle and other terrain environment information; the originally existing area outside the completed land is set to 0, that is, the passable and traversable grid; for the land through by the bevel, according to the binary dilation method of the grid map, all are set to the non-passable black grid for storing obstacle and other terrain environment information, to obtain the grid map boundary environment model.

6. The method of claim 1, wherein, Further comprising updating the coverage value of the sugarcane gathering vehicle, specifically comprising the following steps: Taking the grid network environment, the current coverage matrix, the detection radius and other parameters as inputs, the coverage function identifies the grid points within the detection range of the sugarcane gathering vehicle and gradually enhances the coverage value of the grid points in the coverage matrix; for the part of obstacle processing, the obstacle coordinates are traversed, and the value of the corresponding point in the matrix is fixed as a constant; The coverage function calculates the control parameters of the sugarcane gathering vehicle according to the coverage matrix, identifies the grid point with the minimum coverage value in the coverage range, and calculates the angle between the current position and the position point; If there is a point with a coverage value lower than the expected threshold, the function adjusts the control parameters to make the sugarcane gathering vehicle move towards the area with coverage less than the preset value; the coverage function dynamically adjusts the control parameters Vxcoverage and Vycoverage according to the coverage value and the difference between the current coverage and the expected coverage, adjusts using coefficients g1 and g4, and uses the obtained control parameters to guide the sugarcane gathering vehicle to move towards the area with coverage less than the preset value.

7. The method according to claim 6, wherein, The semi-spring artificial potential field method and A* algorithm are used to generate an optimal path for the sugarcane gathering vehicle according to the obstacle distribution, terrain characteristics and operation target in the grid map, including the following steps: Initialize the grid map parameters and determine the initial position of the sugarcane gathering vehicle, and set the parameters required by the path planning algorithm; Determine the grid that can be walked around the sugarcane gathering vehicle, calculate the coverage value of these grids, and select the target with coverage value less than the set value; Calculate the moving cost of the target with coverage value less than the set value by the Manhattan calculation method of A* algorithm, and select the grid with the minimum cost to move; Use A* algorithm to plan the path from the current grid position to the target grid position of the sugarcane gathering vehicle, and combine the semi-spring APF and coverage value updating strategy and new moving rules to optimize the path.

8. An unmanned field sugarcane gathering vehicle operation path planning system characterized by, It comprises: An image acquisition module is used to acquire aerial images of sugarcane fields and land parameters of sugarcane fields, mark representative sugarcane field working environment lands in the aerial images, and obtain marked aerial field images; A grid model construction module is used to perform edge detection and boundary extraction on the marked aerial field images, take the extracted edges as boundary conditions of the environment modeling map, perform mathematical analysis and processing on the edges, complete the marked aerial field images into regular patterns, determine the grid unit size according to the land parameters, rasterize the completed regular patterns according to the grid unit size, perform binary dilation processing on the completed boundaries, and obtain a grid map boundary environment model; The obstacle profile generation module is configured to generate a regularized obstacle profile in alignment with the grid boundary by traversing obstacle pixels in the grid map boundary environment model based on a dilation algorithm, and marking a center pixel as an obstacle if there is an obstacle within a coverage range of a structural element, and traversing a neighborhood grid based on the structural element. The grid map construction module is configured to construct a grid map based on the grid map boundary environment model and the regularized obstacle profile. The path planning module is configured to generate an optimal path for the sugarcane gathering and transporting vehicle based on the distribution of obstacles, the terrain features and the work target in the grid map based on a semi-spring artificial potential field method and an A* algorithm.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1-8. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when loaded by the processor, can execute the steps of the method of any one of claims 1 to 7.

Images