Harvester in-field path planning method based on multi-objective optimization

Abstract

The invention discloses a harvester in-field path planning method based on multi-objective optimization, and belongs to the field of agricultural machinery dispatching. According to the invention, the problems of low efficiency and waste of energy and human resources in the field path planning of the existing harvester are solved. The contour of a farmland to be harvested is obtained by adopting a rolling ball method, random sampling is performed in the contour, and a point set P in the contour is obtained; dividing the point set P by using a Delauni triangulation method to obtain a triangle set, and reserving points in the farmland contour in the triangle set; segmenting points in the farmland contour by adopting a DBSCAN algorithm to obtain a sub-region set; a bidirectional path smoothing algorithm is adopted, smooth planning is carried out on the harvester path in each sub-region, and a multi-objective function is established with the minimization of the total path length L, the maximization of the farmland coverage area A, the maximization of the path smoothness S and the minimization of the operation time as objectives; and solving the multi-objective function by adopting a grey wolf algorithm to obtain the optimal path of the harvester in the field. The method is suitable for in-field path planning of the harvester.

Description

Technical Field

[0001] The present invention belongs to the field of agricultural machinery scheduling. Background Art

[0002] With the rapid development of the agricultural modernization process, intelligent agricultural technologies play an increasingly crucial role in improving agricultural production efficiency and promoting sustainable development. As one of the core equipment in agricultural production, the rationality of the operating path of a harvester is directly related to the efficiency of agricultural production.

[0003] Traditional operating path planning for harvesters often follows a relatively single and fixed pattern. This pattern exposes many problems when faced with a complex and changeable farmland environment. On the one hand, in terms of coverage area, traditional path planning is difficult to flexibly adjust according to the actual shape of the farmland, crop distribution, etc., resulting in insufficient harvesting in some areas, unable to achieve the maximum coverage area, and causing waste of crop resources. On the other hand, from the perspective of the length of the operating path, traditional planning lacks an overall optimization consideration of the path, and there are often unnecessary round trips and detours of the harvester in the field, making the length of the operating path too long, which not only consumes a large amount of time and energy but also reduces the operating efficiency. Summary of the Invention

[0004] The present invention aims to solve the problems of low efficiency, waste of energy and human resources in the existing in-field path planning of harvesters, and provides a method for in-field path planning of harvesters based on multi-objective optimization.

[0005] The method for in-field path planning of harvesters based on multi-objective optimization according to the present invention includes:

[0006] Step 1: Use the rolling ball method to obtain the contour of the farmland to be harvested, randomly sample within the contour, and obtain a point set P within the contour;

[0007] Step 2: Use the Delaunay triangulation method to triangulate the point set P to obtain a triangle set T = {t1, t2,..., t m}, where each triangle t j is composed of three points;

[0008] Step 3: Judge whether the radius of the circumcircle of each triangle in the triangle set T = {t1, t2,..., t m} is less than or equal to a threshold. If so, divide the points corresponding to the corresponding triangle into points within the farmland contour;

[0009] Step 4: Use the DBSCAN algorithm (density clustering algorithm) to segment the points within the farmland contour to obtain multiple sub-regions and obtain a sub-region set;

[0010] Step 5: Use the bidirectional path smoothing algorithm to smoothly plan the path of the harvester in each sub-region, and establish a multi-objective function with the goal of minimizing the total path length L, maximizing the covered area A of the farmland, maximizing the path smoothness S, and minimizing the operation time.

[0011] Step 6: Use the Grey Wolf algorithm to solve the multi-objective function and obtain the optimal path of the harvester in the field.

[0012] Furthermore, in the present invention, in Step 4, the process of using the DBSCAN algorithm (density clustering algorithm) to segment the points within the farmland contour and obtain multiple sub-regions is as follows:

[0013] Step 4-1: Randomly select a point p in the point set P i , and obtain the set N of all points within the ε-neighborhood around this point ε (p i );

[0014] Step 4-2: Judge whether N ε (p i ) ≥ MinPts. If so, the point p i is a core point, and execute Step 4-3. Otherwise, judge whether this point p i is a neighborhood point of other points in the point set P. If so, return to execute Step 4-1. Otherwise, mark the point p i as a noise point; return to execute Step 4-1;

[0015] Step 4-3: Take the set N ε (p i ) as a cluster, mark the cluster, update the point set P, return to execute Step 4-1 until the points in the updated point set P are only cluster points or noise points, complete the clustering, and take each cluster as a sub-region for segmentation to obtain multiple sub-regions.

[0016] Furthermore, in the present invention, in Step 5, the multi-objective function is:

[0017] minf = ω1·L - ω2·A - ω3·S + ω4·T

[0018] where p i is the i-th point on the path, and the path length L is the sum of the Euclidean distances between adjacent points; A is the area of the farmland covered by the harvester path, Ω is the boundary of the farmland, and 1 cov (x,y) is an indicator function representing the area covered by the path; θ i is the turning angle between two adjacent segments of the path; ω1 is the weight coefficient of the path length, ρ(pi ) is the crop density at path point p i , where k is a constant representing the basic operation time at unit density.

[0019] Furthermore, in the present invention, in step six, the multi-objective function is solved using the Grey Wolf Algorithm, and the process of obtaining the optimal path of the harvester in the field is as follows:

[0020] Step six one: Randomly initialize the positions X of the Grey Wolf population i =(x i1 ,x i2 ,...,x id )(i = 1, 2,..., n), where n is the population size and d is the position dimension of each wolf; set the initial values of the maximum number of iterations and the convergence constant;

[0021] Step six two: Use the objective function as the fitness function and calculate the fitness function value of each Grey Wolf.

[0022] Step six three: Select the three Grey Wolves with the largest fitness function values as the leadership layer in sequence according to the fitness values; update the positions of each Grey Wolf in the wolf pack according to the positions of the three Grey Wolves in the leadership layer combined with the convergence constant; obtain the positions of each Grey Wolf in the updated wolf pack. If the current number of iterations reaches the maximum or there is a maximum value of the fitness function value in the Grey Wolf population, then take the path of the Grey Wolf corresponding to the maximum fitness function value as the optimal path of the harvester in the field, otherwise execute step six four;

[0023] Step six four: Update the convergence constant a, return to execute step six two until if the current number of iterations reaches the maximum or there is a maximum value of the fitness function value in the Grey Wolf population, then take the path of the Grey Wolf corresponding to the maximum fitness function value as the optimal path of the harvester in the field.

[0024] Furthermore, in the present invention, in step six two, the formula for calculating the fitness function value of each Grey Wolf is:

[0025] f i =ω1·L i -ω2·A i -ω3·S i +ω4·T i

[0026] where L i is the path length of the i-th wolf, A i is the coverage area of the i-th wolf, S i is the path smoothness of the i-th wolf, and T i is the operation time of the i-th wolf.

[0027] Further, in the present invention, in step six three, the formula for updating the position of each gray wolf in the wolf pack is:

[0028]

[0029] X1 = X α - A1·D α

[0030] X2 = X β - A2·D β

[0031] X3 = X δ - A3·D δ

[0032] Wherein, D α = |C1·X α - X i |, D β = |C2·X β - X i |, D δ = |C3·X δ - X i |, wherein A1 = 2a·r1 - a, A2 = 2a·r2 - a, A3 = 2a·r3 - a, C1, C2, C3 are random coefficients, C1 = 2·r1, C2 = 2·r2, C3 = 2·r3, where a is a convergence constant Linearly decreasing from 2 to 0, r1, r2 and r3 are random vectors between [0, 1].

[0033] Based on the operation time of the plant density, the present invention aims to minimize the operation path length, maximize the operation coverage area and path smoothness of the harvester for the path planning of the harvester in the field, optimize the path planning of the harvester in the field, achieve the goal of efficient operation, and avoid the waste of personnel resources. Through multi-objective optimization and the application of a new path planning method, the operation efficiency of the harvester is effectively improved and resource waste is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is the flow chart of the method described in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0036] Specific Embodiment 1: Refer to Figure 1 This specific embodiment will be specifically described. The method for path planning of a harvester in the field based on multi-objective optimization described in this embodiment includes:

[0037] Step 1: Use the rolling ball method to obtain the contour of the farmland to be harvested, randomly sample within the contour, and obtain the point set P within the contour;

[0038] Step 2: Use the Delaunay triangulation method to triangulate the point set P to obtain a triangle set T = {t1, t2,..., t m}, where each triangle t j is composed of three points;

[0039] Step 3: Determine whether the radius of the circumcircle of each triangle in the triangle set T = {t1, t2,..., t m} is less than or equal to the threshold. If so, divide the points corresponding to the corresponding triangle into points within the farmland contour;

[0040] Step 4: Use the DBSCAN algorithm (density clustering algorithm) to segment the points within the farmland contour to obtain multiple sub-regions and obtain a sub-region set;

[0041] Step 5: Use the bidirectional path smoothing algorithm to smoothly plan the path of the harvester in each sub-region, establish a multi-objective function with the goal of minimizing the total path length L, maximizing the covered area A of the farmland, maximizing the path smoothness S, and minimizing the operation time;

[0042] Step 6: Use the grey wolf algorithm to solve the multi-objective function to obtain the optimal path of the harvester in the field.

[0043] Furthermore, in the present invention, in Step 4, the process of using the DBSCAN algorithm (density clustering algorithm) to segment the points within the farmland contour to obtain multiple sub-regions is as follows:

[0044] Step 4-1: Randomly select a point p i in the point set P, and obtain the set N ε (p i ) of all points within the ε-neighborhood around this point;

[0045] Step Four Two: Judge N ε (p i ) ≥ MinPts. If so, point p i is a core point, and execute Step Four Three. Otherwise, judge whether this point p i is a neighborhood point of other points in point set P. If so, return to execute Step Four One. Otherwise, mark the point p i as a noise point; return to execute Step Four One;

[0046] Step Four Three: Take the set N ε (p i ) as a cluster, mark the cluster, update point set P, return to execute Step Four One until the points in the updated point set P are only cluster points or noise points, complete clustering, and divide each cluster as a sub-region to obtain multiple sub-regions.

[0047] Furthermore, in the present invention, in Step Five, the multi-objective function is:

[0048] minf = ω1·L - ω2·A - ω3·S + ω4·T

[0049] Wherein,

[0050] p i is the i-th point on the path, and the path length L is the sum of the Euclidean distances between adjacent points;

[0051] A is the area of the farmland covered by the harvester path, Ω is the boundary of the farmland, and 1 cov (x, y) is an indicator function representing the area covered by the path;

[0052] θ i is the turning angle between two adjacent segments of the path;

[0053] ω1 is the weight coefficient of the path length, ρ(p i ) is the crop density at path point p i and k is a constant representing the basic operation time at unit density.

[0054] Furthermore, in the present invention, in Step Six, the grey wolf algorithm is used to solve the multi-objective function, and the process of obtaining the optimal path of the harvester in the field is:

[0055] Step Six One: Randomly initialize the positions X i = (x i1 , x i2 ,..., xid )(i = 1, 2, ..., n), where n is the population size and d is the position dimension of each wolf; set the initial values of the maximum number of iterations and the convergence constant;

[0056] Step 6-2: Use the objective function as the fitness function and calculate the fitness function value of each gray wolf.

[0057] Step 6-3: Select the three gray wolves with the largest fitness function values as the leadership layer in sequence according to the fitness values; update the positions of each gray wolf in the wolf pack according to the positions of the three gray wolves in the leadership layer and the convergence constant; obtain the positions of each gray wolf in the wolf pack after the update. If the current number of iterations reaches the maximum or there is a maximum value of the fitness function value in the gray wolf population, then take the path of the gray wolf corresponding to the maximum fitness function value as the optimal path of the harvester in the field, otherwise execute Step 6-4;

[0058] Step 6-4: Update the convergence constant a, and return to execute Step 6-2 until if the current number of iterations reaches the maximum or there is a maximum value of the fitness function value in the gray wolf population, then take the path of the gray wolf corresponding to the maximum fitness function value as the optimal path of the harvester in the field.

[0059] Furthermore, in the present invention, in Step 6-2, the formula for calculating the fitness function value of each gray wolf is:

[0060] f i = ω1·L i - ω2·A i - ω3·S i + ω4·T i

[0061] where L i is the path length of the i-th wolf, A i is the coverage area of the i-th wolf, S i is the path smoothness of the i-th wolf, and T i is the operation time of the i-th wolf.

[0062] Furthermore, in the present invention, in Step 6-3, the formula for updating the positions of each gray wolf in the wolf pack is:

[0063]

[0064] X1 = X α - A1·D α

[0065] X2 = X β - A2·D β

[0066] X3 = X δ - A3·D δ

[0067] Among them, D α = |C1·X α - X i |, D β = |C2·X β - X i |, D δ = |C3·X δ - X i |, where A1 = 2a·r1 - a, A2 = 2a·r2 - a, A3 = 2a·r3 - a, C1, C2, C3 are random coefficients, C1 = 2·r1, C2 = 2·r2, C3 = 2·r3, and a is a convergence constant linearly decreases from 2 to 0, and r1, r2, and r3 are random vectors between [0, 1].

[0068] Specific application process:

[0069] 1. Apply the Alpha Shape (rolling ball method) to obtain basic information such as the farmland contour. Randomly sample within the contour to obtain a point set P, and perform Delaunay triangulation on the point set P to obtain a triangle set T = {t1, t2,..., t m}, where each triangle t j is composed of three points. The Delaunay triangulation ensures that there are no other points inside the circumcircle of all triangles. The radius r j of the circumcircle, if r j ≤α, then the triangle t j belongs to a part of the Alpha shape.

[0070] 2. Use the DBSCAN (density clustering algorithm) to segment the area within the farmland contour and divide the farmland into multiple regular sub - regions.

[0071] Steps of DBSCAN:

[0072] (1) Select an unvisited point p i ∈P, P = {p1, p2,..., p n}; Find the neighborhood of this point: Calculate all the point sets N ε (p i ) within the ε - neighborhood around this point, including p i itself; For any p j ∈N ε (p i ), calculate ||p i - p j ||≤ε, where ||p i - p j|| represents the point p i and p j the Euclidean distance between them; said ε is set according to the actual situation.

[0073] (2) Determine the core points: If N ε (p i ) ≥ MinPts, then the point p i is a core point. For each core point p i , add all the density-reachable points in N ε (p i ) (density-reachable means: starting from the core point p, continuously expanding along the ε-neighborhood, all reachable points belong to the same cluster; if the point q is within the ε-neighborhood of the core point p and p is a core point, then q is a density-reachable point; if q itself is a core point, then continue to expand its neighborhood to form a larger cluster) to the cluster, and continue to expand its neighborhood until no more points can be added. If the neighborhood of a point is less than MinPts and the point is not within the neighborhood of any core point, then the point is marked as noise. Return to step (1), traverse all points in the point set P, obtain multiple clusters of points in the point set P, divide each cluster into a sub-region, and obtain the sub-region set S = {S1, S2,..., S k}, and each S i is a regular farmland sub-region.

[0074] 3. Two-way path smoothing algorithm. Considering the difference in turning radius when the harvester moves forward and backward, design a two-way path smoothing algorithm. Path point smoothing: In path planning, use the least squares method to smooth the path where p i ′ is the optimized path point, and the goal is to minimize the total distance between path points. Difference in turning radius between forward and backward: When moving forward, allow a smaller turning radius R f , where L is the path segment length and θ min is the minimum turning angle when moving forward. When moving backward, the turning radius is larger R b , where θ max is the maximum turning angle when moving backward.

[0075] 4. Multi-objective modeling;

[0076] Minimize the total path length L: where p i is the i-th point on the path, and the path length L is the sum of the Euclidean distances between adjacent points.

[0077] Maximize the covered area A of the farmland: Where A is the farmland area covered by the harvester's path, Ω is the boundary of the farmland, and 1 cov (x, y) is an indicator function representing the area covered by the path.

[0078] Maximize the path smoothness S: Where θ i is the turning angle between two adjacent segments of the path.

[0079] Minimize the operation time. The operation time is not only affected by the path length but also by the crop density. Areas with high crop density will lead to an increase in the harvesting operation time: Among them, the first term represents the impact of crop density on the operation time. When the density is high, the operation time increases; the second term represents the impact of path length on the operation time. The longer the path, the longer the operation time; ω1 is the weight coefficient of the path length, used to balance the impacts of path length and crop density on the operation time. The crop density at each path point p i is ρ(p i ), and the relationship between the operation time t i and the crop density k is a constant representing the basic operation time at unit density, and ρ(p i ) is the crop density at path point p i . Combine multiple objective functions into a comprehensive objective function through the weighted sum method: minf = ω1·L - ω2·A - ω3·S + ω4·T, where ω1, ω2, ω3, and ω4 are weight coefficients used to balance the relationships among path length, covered area, path smoothness, and operation time.

[0080] Use the Grey Wolf Optimization Algorithm to solve the comprehensive objective function;

[0081] 1. Initialization: Randomly initialize the positions X i =(x i1 , x i2 ,..., x id )(i = 1, 2,..., n), where n is the population size and d is the position dimension of each wolf. Set the algorithm parameters: the maximum number of iterations T, the convergence constant a, and the weights ω1, ω2, ω3, and ω4.

[0082] 2. Calculate fitness: Calculate the fitness value of each grey wolf. The fitness function is defined according to the weighted objective function of path length and farmland coverage area f i = ω1·L i - ω2·A i - ω3·S i + ω4·T i , where L i is the path length of the i-th wolf, and Ai is the coverage area of the i-th wolf, S i is the path smoothness of the i-th wolf, T i is the working time of the i-th wolf.

[0083] 3. Determine the leadership: According to the fitness value, select the optimal solution α, the sub-optimal solution β, the third-optimal solution δ, and other solutions w. Update the position: Update the position of each wolf according to the position of the leading wolf D α = |C1·X α - X i |, D β = |C2·X β - X i |, D δ = |C3·X δ - X i |, where C1, C2, and C3 are random coefficients, A = 2a·r1 - a, C1 = 2·r1, C2 = 2·r2, C3 = 2·r3, where a is the convergence constant, linearly decreases from 2 to 0, and r1, r2, and r3 are random vectors between [0, 1]. Where a linearly decreases from 2 to 0, and r1 and r2 are random vectors between [0, 1]. Update the position of each wolf: X1 = X α - A1·D α X2 = X β - A2·D β X3 = X δ - A3·D δ . The final update is:

[0084] 4. Update the convergence constant: where t is the current iteration number and T is the maximum iteration number.

[0085] 5. Termination condition. If the maximum iteration number T is reached, or the fitness meets the accuracy requirement, the algorithm terminates. Otherwise, return to step 2 for iteration.

[0086] 6. Output. The optimal path solution X α , that is, the multi-objective optimal path planning scheme.

[0087] While the present invention has been described herein with reference to particular embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. Accordingly, it should be understood that numerous modifications may be made to the exemplary embodiments, and other arrangements may be devised, without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the different dependent claims and the features described herein may be combined in a manner different from that described in the original claims. It should also be understood that the features described in connection with separate embodiments may be used in other described embodiments.

Claims

1. A harvester field path planning method based on multi-objective optimization, characterized in that: include: Step 1: Use the rolling ball method to obtain the contour of the farmland to be harvested, and randomly sample within the contour to obtain a point set P within the contour; Step 2: Use the Delaunay triangulation method to triangulate the point set P to obtain a triangle set T = {t1, t2, ..., t m }, where each triangle t j It consists of three points; Step 3: Determine the triangle set T = {t1, t2, ..., t m }Whether the radius of the circumscribed circle of each triangle in is less than or equal to the threshold value, if so, the points corresponding to the corresponding triangles are classified as points inside the farmland contour; Step 4: Use the DBSCAN algorithm to segment the points within the farmland contour, obtain multiple sub-regions, and obtain a sub-region set; Step 5: Use a bidirectional path smoothing algorithm to smoothly plan the path of the harvester in each sub-area, and establish a multi-objective function with the goals of minimizing the total path length L, maximizing the coverage area A of the farmland, maximizing the path smoothness S and minimizing the operation time; Step six: Use the Grey Wolf Algorithm to solve the multi-objective function and obtain the optimal path of the harvester in the field.

2. The harvester field path planning method based on multi-objective optimization according to claim 1 is characterized in that: In step 4, the DBSCAN algorithm (density clustering algorithm) is used to segment the points within the farmland outline, and the process of obtaining multiple sub-areas is as follows: Step 41: Randomly select a point p in the point set P i , get the set N of all points in the ε-neighborhood around the point ε (p i ); Step 42: Determine N ε (p i )≥MinPts, if so, then point p i is the core point, execute step 4-3, otherwise, determine the point p i Is it a neighboring point of other points in the point set P? If so, return to step 41. Otherwise, the point p i Mark as noise point; Return to step 41; Step 43: Set N ε (p i ) as a cluster, mark the cluster, update the point set P, return to execute step 41, until the points in the updated point set P are only cluster points or noise points, the clustering is completed, and each cluster is divided as a sub-region to obtain multiple sub-regions.

3. The harvester field path planning method based on multi-objective optimization according to claim 1 or 2, characterized in that: In step 5, the multi-objective function is: minf=ω1·L-ω2·A-ω3·S+ω4·T in, p i is the i-th point on the path, and the path length L is the sum of the Euclidean distances between adjacent points; A is the farmland area covered by the harvester path, Ω is the boundary of the farmland, 1 cov (x,y) is the indicator function, which represents the area covered by the path; θ i is the turning angle between two adjacent segments in the path; ω1 is the weight coefficient of path length, ρ(p i ) is the path point p i The crop density at , k is a constant representing the basic operation time at unit density.

4. The harvester field path planning method based on multi-objective optimization according to claim 1 or 2, characterized in that: In step six, the gray wolf algorithm is used to solve the multi-objective function, and the process of obtaining the optimal path of the harvester in the field is: Step 6.

1. Randomly initialize the position X of the gray wolf population i =(x i1 ,x i2 ,...,x id )(i=1,2,...,n), where n is the population size and d is the location dimension of each wolf; Set the maximum number of iterations and the initial value of the convergence constant; Step 62: taking the objective function as the fitness function, and calculating the fitness function value of each gray wolf; Step 63: Select the three gray wolves with the largest fitness function values ​​as the leaders according to the fitness values; update the position of each gray wolf in the wolf pack according to the positions of the three gray wolves in the leadership and the convergence constant; obtain the position of each gray wolf in the wolf pack after the update, if the current number of iterations reaches the maximum or the maximum fitness function value exists in the gray wolf population, then the path of the gray wolf corresponding to the maximum fitness function value is used as the optimal path of the harvester in the field, otherwise execute step 64; Step 64: Update the convergence constant a and return to step 62 until the current number of iterations reaches the maximum or the maximum value of the fitness function exists in the gray wolf population. Then, the path of the gray wolf corresponding to the maximum value of the fitness function is used as the optimal path of the harvester in the field.

5. The method for harvester field path planning based on multi-objective optimization according to claim 4, characterized in that: In step 62, the formula for calculating the fitness function value of each gray wolf is: f i =ω1·L i -ω2·A i -ω3·S i +ω4·T i Among them, L i is the path length of the i-th wolf, A i is the coverage area of ​​the ith wolf, S i is the path smoothness of the ith wolf, T i is the operation time of the i-th wolf.

6. The method for harvester field path planning based on multi-objective optimization according to claim 5, characterized in that: In step 63, the formula for updating the position of each gray wolf in the wolf pack is: X1=X α -A1·D α X2=X β -A2·D β X3=X δ -A3·D δ Among them, D α =|C1·X α -X i |, D β =|C2·X β -X i |, D δ =|C3·X δ -X i |, where A1 = 2a·r1-a, A2 = 2a·r2-a, A3 = 2a·r3-a, C1, C2, C3 are random coefficients, C1 = 2·r1, C2 = 2·r2, C3 = 2·r3 where a is the convergence constant, It decreases linearly from 2 to 0, and r1, r2 and r3 are random vectors between [0,1].

Images