A multi-terminal seedling transplanting path optimization method based on improved simulated annealing algorithm

Abstract

The present invention relates to a transplanting method for seedlings in pots in agricultural machinery. The purpose is to provide a method for optimizing the transplanting path of multi-terminal seedlings in pots based on an improved simulated annealing algorithm, which can reduce the walking distance of a manipulator with multiple terminal actuators of a greenhouse seedling transplanting machine, thereby improving the work efficiency. The technical solution is a method for optimizing the transplanting path of multi-terminal seedlings in pots based on an improved simulated annealing algorithm, including the planning of the transplanting path of seedlings in a target plug tray and the planning of the path for taking seedlings in a seedling supply plug tray, splicing the transplanting path of the seedlings in the target plug tray and the path for taking seedlings in the seedling supply plug tray end to end, and generating a global transplanting path for seedlings.

Description

Technical Field

[0001] The present invention relates to a transplanting method for pot seedlings in agricultural machinery, and particularly to an optimization method for the transplanting path of multi-end effector pot seedling replanting based on an improved simulated annealing algorithm. Background Art

[0002] During the greenhouse plug seedling cultivation process, there are problems such as missed seeding, non-germination of seeds, and abnormal growth of seedling leaves. Normal pot seedlings account for about 80%-95% of the total number. For the purposes of preventing pests and diseases, improving the utilization rate of plug trays, and avoiding empty planting or missed planting during subsequent mechanized batch transplanting in greenhouse farms, within a certain period after the plug seedlings are cultivated, the pot seedling clumps containing inferior pot seedlings must be removed, and then the healthy pot seedling clumps developed from the alternative seedling trays are replanted to obtain a healthy seedling tray without inferior pot seedlings. Traditionally, manual identification and transplanting operations are carried out, which is considered a labor-intensive industry with low operation efficiency, high cost, and high labor intensity. The greenhouse pot seedling removal and replanting machine detects the health status and position information of pot seedlings through machine vision, and uses a manipulator with a multi-end effector to clamp healthy pot seedlings from the seedling supply plug tray and replant them into the target plug tray, which can solve the above problems.

[0003] The positions of the holes to be replanted in the target plug tray and the healthy pot seedlings in the seedling supply plug tray are highly random. The greenhouse pot seedling removal and replanting machine controls the manipulator with a multi-end effector to travel back and forth between the target plug tray and the seedling supply plug tray. Different sequences of replanting the holes to be replanted in the target plug tray, different sequences of clamping healthy pot seedlings in the seedling supply plug tray, and different sequences of selecting the end effector of the manipulator result in diverse transplanting paths, leading to variable lengths of the replanting paths. Selecting a suitable replanting path can significantly reduce the walking distance of the end effector. Due to the large amount of hole data, the method of the controller traversing and calculating to screen out the shortest and optimal path cannot meet the real-time requirements. Currently, there are replanting path planning methods for single-end effectors, but the replanting path planning method for multi-end effectors needs to be developed. Summary of the Invention

[0004] The object of the present invention is to overcome the deficiencies of the above background art and provide an optimization method for the transplanting path of multi-end pot seedling replanting based on an improved simulated annealing algorithm, which can reduce the walking distance of the manipulator with a multi-end effector of the greenhouse pot seedling removal and replanting machine, thereby improving the operation efficiency.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] An optimization method for the transplanting path of multi-end pot seedling replanting based on an improved simulated annealing algorithm includes replanting path planning in the target plug tray and seedling taking path planning in the seedling supply plug tray, splicing the replanting path in the target plug tray and the seedling taking path in the seedling supply plug tray end to end to generate a global replanting operation transplanting path;

[0007] The path planning process for supplementary seeding in the target seedling tray is as follows: Mark and encode the positions to be supplemented in the target seedling tray; Group the holes to be supplemented according to the number of end effectors, which is called the group to be supplemented; Obtain the planting sequence of the centroids of the manipulators corresponding to the holes to be supplemented within the group to be supplemented and the selection of the end effector number corresponding to a certain hole to be supplemented through an improved simulated annealing algorithm; Generate the shortest planting paths for different groups to be supplemented in the target seedling tray.

[0008] The path planning process for seedling picking in the seedling supply tray is as follows: Define the search area based on the start and end points of the shortest planting paths of different groups to be supplemented in the target seedling tray, and mark and encode the healthy pot seedlings within the search area; Select the best positions of the healthy pot seedlings within the search area through an improved simulated annealing algorithm. The number of selected healthy pot seedlings is the same as the number of holes to be supplemented within the corresponding group to be supplemented, and this group of healthy pot seedlings is called the seedling supply group; Obtain the picking sequence of the centroids of the manipulators corresponding to the selected healthy pot seedlings within the seedling supply group and the selection of the end effector number corresponding to a certain healthy pot seedling through an improved simulated annealing algorithm; Generate the shortest picking paths for different seedling supply groups in the seedling supply tray.

[0009] Further, the marking and encoding of the positions to be supplemented in the target seedling tray specifically means that each hole to be supplemented in the target seedling tray is marked with positive real numbers in the order from top to bottom and from right to left. Thus, the marking and encoding actually implicitly contain the hole position and the seedling health information.

[0010] Further, the grouping of the holes to be supplemented according to the number of end effectors specifically means that assuming there are 10 holes to be supplemented in the target seedling tray, the marking and encoding set of the holes to be supplemented is {1, 2, 3, ……, 10}. Assuming the number of end effectors attached to the manipulator is 5, the holes to be supplemented can be divided into the group to be supplemented 1 {1, 2, 3, 4, 5} and the group to be supplemented 2 {6, 7, 8, 9, 10}. Those with less than 5 are grouped separately. The basis for this grouping method depends on the working mode of the manipulator's supplementary seeding and transplanting operation: Step 1, the manipulator starts from the starting point and goes to the seedling supply tray to pick 5 healthy pot seedlings in a certain order; Step 2, the manipulator with 5 healthy pot seedlings goes from the seedling supply tray to the target seedling tray; Step 3, the manipulator plants the 5 healthy pot seedlings into 5 holes to be supplemented in the target seedling tray in a certain order; Step 4, the manipulator returns to the seedling supply tray; Step 5, repeat Steps 1 to 4 until all the holes to be supplemented in the target seedling tray are planted with healthy seedlings. Steps 1 to 4 are called a set of supplementary seeding and transplanting operations.

[0011] Furthermore, the planting sequence of the centroid of the manipulator corresponding to the holes to be filled in the group to be filled and the selection of the end effector number corresponding to a certain hole to be filled are obtained by improving the simulated annealing algorithm. Specifically, assume that the end effectors arranged in a row on the manipulator are labeled as A - B - C - D - E from right to left. Assume that the physical position of end effector C is the centroid of the manipulator. The position of the centroid of the manipulator relative to the hole to be filled changes with the order of releasing the healthy pot seedlings held by the end effectors. Convert the position of the hole to be filled into the corresponding centroid position of the manipulator. By improving the simulated annealing algorithm, obtain the best sequence of the centroid positions of the manipulator corresponding to traversing the holes to be filled in the group to be filled. By improving the simulated annealing algorithm, calculate the best selection of the end effector number corresponding to a certain hole to be filled. Finally, obtain the planting sequence of the centroid of the manipulator corresponding to the holes to be filled in the group to be filled and the selection of the end effector number.

[0012] Furthermore, the position of the centroid of the manipulator relative to the hole to be filled changes with the order of releasing the healthy pot seedlings held by the end effectors. Specifically, assume that there is a row of a total of 5 holes (named a, b, c, d, e from right to left), and hole c is the hole to be filled. When the manipulator plants the hole to be filled using the healthy pot seedling on end effector E, the centroid of the manipulator is at the position of hole a. Therefore, when the manipulator plants the hole to be filled using the healthy pot seedlings on end effectors E, D, C, B, and A respectively, the centroid of the manipulator is at the positions of holes a, b, d, d, and e respectively.

[0013] Furthermore, the conversion of the position of the hole to be filled into the corresponding centroid position of the manipulator is as follows. Assume that the position of the hole to be filled is (5, 5), and the meaning of this coordinate is that the hole to be filled is in the 5th column and 5th row from left to right and from bottom to top. Assume that end effector E is used to plant this hole to be filled, then the corresponding centroid position of the manipulator for this hole to be filled is (5 + 2, 5), and the meaning of this coordinate is that the centroid of the manipulator is in the 7th column and 5th row. Thus, the position conversion of the hole to be filled to the centroid of the manipulator is completed.

[0014] Further, the best order for traversing the centroid positions of the manipulators corresponding to the replanting of the holes to be filled in the group to be filled obtained by improving the simulated annealing algorithm is specifically as follows: Assume that the positions of the holes to be filled in group 1 to be filled are {hole to be filled 1(3, 10), hole to be filled 2(2, 9), hole to be filled 3(4, 8), hole to be filled 4(5, 7), hole to be filled 5(1, 7)}. Assume that the array A = {0, -1, +1, -2, +2} represents that for the holes to be filled 1, 2, 3, 4, and 5, they respectively receive healthy pot seedlings released from the end effectors C, B, D, A, and E. Any healthy pot seedling on an end effector can only be selected once in a group of replanting operations. Then, the centroid coordinates corresponding to the holes to be filled selected by this end effector are {centroid 1(3 + 0, 10), centroid 2(2 - 1, 9), centroid 3(4 + 1, 8), centroid 4(5 - 2, 7), centroid 5(1 + 2, 7)}. The calculation process of the improved simulated annealing algorithm is as follows:

[0015] Step 1: Set initial parameters such as the initial temperature T 0 = 100, the current temperature T is equal to T 0 = 100, the lowest temperature T min = 1e - 6, the number of inner - loop times i within the same temperature = 10. Each loop is to solve the optimal sorting once (exchange the centroid once). The specific solution process is the operation in Step 2. When the number of solution times reaches the inner - loop times i = 10 times, the temperature is decreased, that is, the current temperature T is multiplied by the cooling coefficient K = 0.2. At this moment, the current temperature T = 20. At this temperature, loop 10 times again, that is, solve the optimal sorting ten times (exchange the centroid ten times). Then continue to decrease the temperature, and continue to solve 10 times after the temperature reduction until the current temperature is less than the set lowest temperature T min , and this loop - solving process stops;

[0016] Step 2: Randomly exchange the elements in the initial centroid coordinate array one - by - one once. For example, exchange the order of centroid 1 and centroid 5, {centroid 5(1 + 2, 7), centroid 2(2 - 1, 9), centroid 3(4 + 1, 8), centroid 4(5 - 2, 7), centroid 1(3 + 0, 10)}. Calculate the path length in this order. At this time, there will be two results. One is that the path length after the exchange is longer than the path length before the exchange. At this moment, keep the order before the exchange unchanged. The other is that the path length after the exchange is less than the path length before the exchange. At this moment, overwrite the order before the exchange with the order after the exchange to become the current optimal path order;

[0017] Step 3: Execute Step 2 ten times, that is, the number of inner - loop times i within the same temperature = 10;

[0018] Step 4: After Step 3 is executed, multiply the current temperature T by the cooling coefficient K to obtain the current temperature;

[0019] Step 5: Repeat Steps 1 to 4 until the current temperature is lower than the minimum temperature T min .

[0020] The exchanges in the above steps are random. To accelerate the search speed, the following improvements are made: The centroids close to the demarcation line are assigned to both ends of the group to be replenished, such as centroid 1 and centroid 3. The centroids far from the demarcation line are assigned to the middle of the group to be replenished, and the farther a centroid is from the demarcation line, the greater the probability that it is assigned to the middle of the group to be replenished. The demarcation line is located in the middle of the target seedling tray and the seedling supply tray. The shortest path of all centroids in the group to be replenished is obtained, such as {centroid 1(3 + 0, 10), centroid 2(2 - 1, 9), centroid 4(5 - 2, 7), centroid 5(1 + 2, 7), centroid 3(4 + 1, 8)}.

[0021] Furthermore, the optimal serial number selection of the end effector corresponding to a certain hole to be replenished is calculated by improving the simulated annealing algorithm. Specifically, the elements in the exchange array A = {0, -1, +1, -2, +2} are exchanged. The steps of the improved simulated annealing algorithm are the same as those above, and the following improvements are made: The end effector E has a greater probability of replanting the holes to be replenished far from the demarcation line, which can make the centroid of the manipulator close to the demarcation line. The probabilities of selecting the end effector for the holes to be replenished farther from the demarcation line from high to low are E, D, C, B, A respectively. After each exchange, the optimal arrangement order of the centroid positions of the manipulator corresponding to the holes to be replenished needs to be recalculated. The optimal serial number selection of the end effector corresponding to the holes to be replenished 1, 2, 3, 4, 5 in the group to be replenished is obtained, such as the array A = {0, +1, -1, -2, +2}, and the optimal centroid arrangement order in the group to be replenished is {centroid 1(3 + 0, 10), centroid 2(2 + 1, 9), centroid 3(4 - 1, 8), centroid 4(5 - 2, 7), centroid 5(1 + 2, 7)}

[0022] Further, a search area is delimited based on the start and end points of the shortest replanting paths of different groups to be replenished in the target seedling tray, and the healthy pot seedlings within the search area are marked and encoded. Specifically, a 1×5 rectangular area A is selected. The sum of the distances from the centroid position of the manipulator corresponding to the last to-be-replenished hole in the previous group to be replenished and the centroid position of the manipulator corresponding to the first to-be-replenished hole in the next group to be replenished is the shortest for this area. The healthy pot seedlings in area A are encoded with positive real numbers in the order from right to left and from top to bottom. In an ideal state, there are 5 healthy pot seedlings in this area so that the manipulator can grab them at one time, which is relatively rare. If the number of healthy pot seedlings in the current search area does not meet the requirements of the next group to be replenished, the search area B (3×6), search area C (5×7), etc. are expanded outward using the breadth-first method, and encoding is performed in the same way. The purpose of delimiting the search area is that after the manipulator replants the holes in the previous group to be replenished, it goes to the seedling supply tray to pick 5 healthy pot seedlings, and then goes to the next group to be replenished for replanting. To make the seedling picking path the shortest, it is necessary to exclude those areas that are obviously too far away and impossible, delimit a search area with the greatest hope, and reduce the search calculation amount.

[0023] Further, the best position of the healthy pot seedlings in the search area is selected by improving the simulated annealing algorithm. Specifically, assume that area B is a 3×6 rectangular area, and there are a total of 14 healthy pot seedlings in it. The positive real numbers are marked in the order from right to left and from top to bottom for area B = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14}. Arbitrarily select five of the pot seedlings to form a seedling supply group 1 {1, 3, 5, 7, 9}. By improving the simulated annealing algorithm, exchange seedling supply group 1 with the remaining healthy pot seedlings in area B, such as seedling supply group 1 {1, 3, 5, 7, 14}. For this corresponding seedling supply group, calculate the best seedling picking order of the centroid corresponding to the manipulator clamping this healthy pot seedling and the optimal selection of the serial number of the end effector clamped by the manipulator. The calculation method is the same as the method used in the target seedling tray and will not be elaborated here.

[0024] The shortest replanting paths of different groups to be replenished in the target seedling tray and the shortest seedling picking paths of different seedling supply groups in the seedling supply tray are merged, passing through the starting point and the ending point, and the overall seedling replenishment and transplanting path is generated.

[0025] Advantages of the present invention

[0026] Through machine vision detection, the present invention obtains the health status and position information of the pot seedlings in the target seedling tray and the seedling supply tray, and uses the improved simulated annealing algorithm as a technical means to complete the rapid optimization of the seedling replenishment and transplanting path from the seedling supply tray to the target seedling tray, reduce the walking distance of the manipulator with multiple end effectors, and thus improve the operation efficiency of the greenhouse pot seedling weeding and replenishment transplanting machine. Description of the drawings

[0027] Figure 1It is the flow chart of the present invention.

[0028] Figure 2 It is the target plug tray marking coding diagram of the present invention.

[0029] Figure 3 It is the schematic diagram of the change of the center of mass of the manipulator corresponding to different holes to be filled in the present invention.

[0030] Figure 4 It is the visual path diagram of the seedling replenishment path of the target plug tray under any arrangement of the present invention.

[0031] Figure 5 It is the visual path diagram of the seedling replenishment path of the target plug tray with the optimal center of mass sorting of the manipulator in the present invention.

[0032] Figure 6 It is the visual path diagram of the seedling replenishment path of the target plug tray with the optimal selection of the end effector in the present invention.

[0033] Figure 7 It is the schematic diagram of the search area division and healthy plug seedling selection of any seedling supply group in the present invention.

[0034] Figure 8 It is the visual path diagram of the global path optimization of the present invention. Detailed implementation manners

[0035] The present invention will be further described below in conjunction with the method flow chart and embodiments.

[0036] The present invention has obtained the health information of the seedlings in the target plug tray and the seedling supply plug tray of the greenhouse plug seedling removal, replenishment and transplanting machine through machine vision. The inferior plug seedlings in the target plug tray and the seedling supply plug tray have been removed. The holes to be filled in the target plug tray need to be replanted with healthy plug seedlings. The target plug tray is the seedling tray that needs healthy plug seedlings to be transplanted into its holes to be filled. The seedling supply plug tray is the seedling tray that provides healthy plug seedlings for the target plug tray. The target plug tray and the seedling supply plug tray are arranged horizontally and parallel to each other. The manipulator repeatedly plants the healthy plug seedlings in the seedling supply plug tray into the holes to be filled in the target plug tray. The moving path of the manipulator in this process is called the seedling replenishment operation transplanting path. The seedling replenishment operation transplanting path can be divided into the seedling replenishment path planning in the target plug tray and the seedling picking path planning in the seedling supply plug tray.

[0037] The process of supplementary seedling transplanting operation is introduced as follows: Five end effectors arranged side by side in a line (named A, B, C, D, and E from right to left) are fixedly installed on a three-coordinate manipulator. The manipulator starts from the starting point and goes to the seedling supply tray to grab five healthy pot seedlings. The selection of the position of the healthy pot seedlings, which end effector to use to grab the pot seedling, and the grabbing order of the healthy pot seedlings will all result in different path lengths. Here, there is a problem of finding the optimal path. The manipulator moves from the seedling supply tray with five healthy pot seedlings to the target tray. The order of transplanting seedlings into the holes to be filled and which healthy pot seedling on which end effector to use to fill the hole to be filled will also result in different path lengths. Here, there is also a problem of finding the optimal path. After all five healthy pot seedlings on the manipulator are transplanted into the five holes to be filled, the first group of supplementary seedling transplanting operations is completed. Repeat the above steps to start the next group of supplementary seedling transplanting operations until all the holes to be filled in the target tray are transplanted with healthy pot seedlings. The present invention mainly provides a method for planning the optimal path of global supplementary seedling transplanting based on multiple end effectors.

[0038] The flow chart of the method of the present invention is as Figure 1 shown: The present invention obtains the health information of the pot seedlings in the tray through machine vision, marks and encodes the positions of the holes to be filled in the target tray, and groups the holes to be filled according to the number of end effectors (as Figure 2 shown). The position of the centroid of the manipulator relative to the hole to be filled changes with the order of releasing the healthy pot seedlings on the end effector. This relationship is as Figure 3 shown. The centroid of the manipulator corresponding to the holes to be filled in a single group to be filled is optimally sorted through an improved greedy genetic simulated annealing algorithm (as Figure 4 ,5) and the optimal selection of the end effector number (as Figure 6 ). The shortest supplementary seedling path of the target tray is generated. According to the start and end points of the shortest supplementary seedling paths of different groups to be filled in the target tray, the search area is delimited (as Figure 7 ), and the best healthy pot seedling in the search area is selected through an improved greedy genetic simulated annealing algorithm (as Figure 7 ), and the centroid of the manipulator for taking the best healthy pot seedling in the seedling supply group and the selection of the end effector number are selected in the same method as Figure 5 ,6. The shortest seedling taking path of the seedling supply tray is generated. The paths are spliced, including the starting point and the end point, to obtain the overall path planning diagram.

[0039] P1: The positions of the holes in the target tray and the seedling supply tray in the transplanting machine system are fixed, and the health information of the pot seedlings is also obtained through machine vision. The holes to be filled in the target tray are marked with positive real numbers in the order from top to bottom and from right to left, as Figure 2As shown, assume that the target seedling tray is a 50-hole seedling tray with a 5×10 specification. Then the marking code set of the holes to be filled is {1, 2, 3, ……, 10}. Group the holes to be filled according to the number of end effectors. Figure 3 The number of end effectors in Figure 3 is 5. Group the holes to be filled with 5 holes in a group, and the remaining holes less than 5 are also grouped into one group. Then the holes to be filled can be divided into filling group 1 {1, 2, 3, 4, 5} and filling group 2 {6, 7, 8, 9, 10}.

[0040] End effector C is the position of the overall centroid of the manipulator. The position of the centroid of the manipulator relative to the hole to be filled changes according to the selection of grasping or releasing the end effector. For example, Figure 3 As shown. Assume that there is a hole c to be filled in a row of 5 holes (named a, b, c, d, e from right to left) in a 50-hole seedling tray. When the order of releasing healthy seedlings by the end effector is E, D, C, B, and A, the centroid of the manipulator is located at the positions of holes a, b, c, d, and e respectively. The healthy seedlings on any one end effector can only be selected once in the filling operation of a filling group. Therefore, assume an array A = {+2, +1, 0, -1, -2} to reflect this selection. The numbers in the array represent the positions of the centroid coordinates of the manipulator relative to the hole to be filled under different end effector selections. For example, the number 2 means that when using end effector E to fill a hole to be filled, the centroid coordinate of the manipulator relative to the hole to be filled moves two units to the right. This reflects the selection of the end effector. Specifically, assume the position of the hole to be filled is (5, 5), and the meaning of this coordinate is that the hole to be filled is in the 5th column and 5th row from left to right and from bottom to top. Assume using end effector E to plant the hole to be filled, then the corresponding centroid position of the manipulator for this hole to be filled is (5 + 2, 5), and the meaning of this coordinate is that the centroid of the manipulator is in the 7th column and 5th row.

[0041] P2: Take filling group 1 as an example. Since it is not clear which healthy seedlings on which end effector should be used for the holes to be filled in the filling group, a random array A = {0, -1, +1, -2, +2} applied to filling group 1 is selected. This means that for the holes to be filled 1, 2, 3, 4, and 5, they respectively receive the healthy seedlings released from end effectors C, B, D, A, and E. Therefore, an initial solution, that is, an initial path, can be generated. This initial path is composed of the centroid coordinates of the manipulator corresponding to the holes to be filled. The length of this path is surely not the optimal one. For example, Figure 4As shown. Similarly, it is not clear whether the moving distance of the manipulator corresponding to the initial seedling replenishment order {centroid 1(3+0, 10), centroid 2(2-1, 9), centroid 3(4+1, 8), centroid 4(5-2, 7), centroid 5(1+2, 7)} is the shortest. Therefore, the problem is transformed into finding the shortest path for the manipulator to start from the demarcation line, traverse the centroid positions of the manipulator corresponding to the five holes to be replenished in the group to be replenished 1, and finally return to the demarcation line. The method for finding is to use the improved simulated annealing algorithm.

[0042] The calculation process of the improved simulated annealing algorithm is as follows:

[0043] Step 1: Set initial parameters such as the initial temperature T 0 = 100, the current temperature T is equal to T 0 = 100, the lowest temperature T min = 1e-6, the number of inner-loop times i within the same temperature is 10. Each loop is to solve the optimal sorting once (swap the centroids once). The specific solution process is the operation in Step 2. When the number of solution times reaches the inner-loop times i = 10 times, the temperature is decreased, that is, the current temperature T is multiplied by the cooling coefficient K = 0.2. At this moment, the current temperature T = 20. At this temperature, loop 10 times again, that is, solve the optimal sorting ten times (swap the centroids ten times). Then continue to decrease the temperature, and continue to solve 10 times after the temperature decrease until the current temperature is less than the set lowest temperature T min , and this loop solution process stops;

[0044] Step 2: Randomly swap the elements in the initial centroid coordinate array one by one. For example, swap the order of centroid 1 and centroid 5, {centroid 5(1+2, 7), centroid 2(2-1, 9), centroid 3(4+1, 8), centroid 4(5-2, 7), centroid 1(3+0, 10)}, and calculate the path length in this order. At this time, there will be two results. One is that the path length after the swap is longer than the path length before the swap. At this moment, keep the order before the swap unchanged. The other is that the path length after the swap is less than the path length before the swap. At this moment, overwrite the order before the swap with the order after the swap to become the current optimal path order;

[0045] Step 3: Execute Step 2 ten times, that is, the number of inner-loop times i within the same temperature is 10; Step 4: After Step 3 is executed, multiply the current temperature T by the cooling coefficient K to obtain the current temperature; Step 5: Repeat Steps 1 to 4 until the current temperature is less than the lowest temperature T minTo accelerate the convergence rate, the following improvements are made: The algorithm tends to allocate the centroids near the demarcation line to both ends of the group to be supplemented, such as centroid 1 and centroid 3, and allocate the centroids far from the demarcation line to the middle of the group to be supplemented. Moreover, the farther the centroid is from the demarcation line, the greater the probability that it is allocated to the middle of the group to be supplemented. After the above steps, the centroid seeding order of the manipulator corresponding to the shortest seeding path of group 1 to be supplemented is {centroid 1(3 + 0, 10), centroid 2(2 - 1, 9), centroid 4(5 - 2, 7), centroid 5(1 + 2, 7), centroid 3(4 + 1, 8)}, as Figure 5 shown.

[0046] P3: In P2, only the shortest path of the manipulator centroid arrangement under a specific end-effector selection order is solved, where the to-be-supplemented holes 1, 2, 3, 4, and 5 respectively receive healthy pot seedlings released from end-effectors C, B, D, A, and E, that is, the array A = {0, -1, +1, -2, +2}. The problem of selecting the end-effector, that is, the optimal release order of the end-effector (the optimal choice of which end-effector to use for a specific to-be-supplemented hole), has not been solved. Taking group 1 as an example, the improved simulated annealing algorithm is used to pairwise exchange the elements in array A. For example, before the exchange, the original array is A = {0, -1, +1, -2, +2}, and after one exchange, it becomes A = {0, 1, -1, -2, +2}, that is, the to-be-supplemented holes 1, 2, 3, 4, and 5 respectively receive healthy pot seedlings released from end-effectors C, D, B, A, and E. Since the selection order of the end-effectors changes, it is necessary to calculate the shortest path length of the manipulator centroid under this selection. The calculation method is the same as that in P2, that is, repeat P2 once; at this time, two results will occur. One is that the path length after the exchange is longer than the path length before the exchange, and in this case, the selection order before the exchange remains unchanged. The other is that the path length after the exchange is shorter than the path length before the exchange, and in this case, the selection order after the exchange overwrites the selection order before the exchange and becomes the current optimal end-effector selection order. The exchange step continues until the current temperature is lower than the lowest temperature T min , and the operation steps of the improved simulated annealing algorithm are the same as those in P2. The above exchanges are random and disordered. To accelerate the convergence rate, the following improvements are made: We tend to let end-effector E have a higher probability of supplementing the to-be-supplemented holes far from the demarcation line, which can make the centroid of the manipulator closer to the demarcation line. The probabilities of selecting end-effectors for the to-be-supplemented holes farther from the demarcation line from high to low are E, D, C, B, A. After the above steps, the best element sorting of array A is finally obtained as array A = {0, 1, -1, +2, -2}, that is, the selection order of the end-effectors is C, B, D, E, A corresponding to the to-be-supplemented holes 1, 2, 3, 4, and 5, as Figure 6 shown.

[0047] P4: Through the above steps, the optimal serial numbers of the manipulators corresponding to the to-be-supplemented hole positions in different to-be-supplemented groups in the target plug tray and the order of the highest-quality centroid paths are obtained. As Figure 6 shown. The difference between obtaining the shortest seedling-supplementing path in the target plug tray and the shortest seedling-taking path in the seedling-supplying plug tray is that the positions of the to-be-supplemented hole positions in the target plug tray are known and fixed, while the healthy plug seedlings used for seedling supplementation in the seedling-supplying plug tray are unknown. If the positions of the healthy plug seedlings used for seedling supplementation in the seedling-supplying plug tray can be obtained, then the optimal serial numbers of the manipulators for taking seedlings corresponding to the healthy plug seedlings in different seedling-supplying groups and the order of the highest-quality centroid paths for taking seedlings can be obtained in the same way as in steps P2 and P3.

[0048] P5: Through steps P1 - P4, the shortest seedling-supplementing path of the manipulator in the target plug tray is obtained. There are a large number of optional healthy plug seedlings in the seedling-supplying plug tray. The purpose of delimiting the search area is that after the manipulator completes the supplementation of the to-be-supplemented hole positions in the previous group, it goes to the seedling-supplying plug tray to pick 5 healthy plug seedlings, and then goes to the next to-be-supplemented group for supplementation. To make the seedling-taking path the shortest, it is necessary to exclude those areas that are obviously too far away and impossible, and delimit a search area with the greatest hope. The search area is delimited according to the start and end points of the shortest seedling-supplementing paths of different to-be-supplemented groups in the target plug tray. For the sake of clear expression, a 128-hole plug tray (16×8) is selected here. As Figure 7 shown, the healthy plug seedlings in the search area are marked and encoded. Specifically, a 1×5 rectangular area A is selected. The position of the hole in the third column of this area has the shortest sum of distances from the centroid position of the manipulator corresponding to the last to-be-supplemented hole position in the previous group and the centroid position of the manipulator corresponding to the first to-be-supplemented hole position in the next group. The healthy plug seedlings in area A are encoded with positive real numbers in the order from right to left and from top to bottom. In an ideal state, there are 5 healthy plug seedlings in this area so that the manipulator can grab them at one time. This situation is relatively rare. If this situation exists, it is determined that the 5 healthy plug seedlings in area A are the 5 healthy plug seedlings needed by the next to-be-supplemented group. If the number of healthy plug seedlings in area A does not meet the needs of the next to-be-supplemented group, the breadth-first method is used to expand outward to search areas B (3×6), C (5×7), etc., and encoding is carried out in the same way. The purpose of delimiting the search area is to reduce the optional range of the manipulator for taking seedlings, only search the most likely area corresponding to the shortest path, and reduce the overall calculation amount.

[0049] P6: As Figure 7As shown, the number of healthy pot seedlings in area A does not meet the needs of the next set of seedlings to be supplemented. Therefore, the search area is extended to area B, which is a 3×6 rectangular area. There are a total of 14 healthy pot seedlings in it. The area B is marked with positive real numbers in the order from right to left and from top to bottom as B={1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14}. Currently, it is not clear which 5 healthy pot seedlings in the area are needed. Therefore, any five healthy pot seedlings are selected to form the seedling supply group 1 {1, 3, 5, 7, 9}. Steps P7 and P8 are executed to obtain the optimal serial number selection of the manipulator for picking seedlings and the optimal centroid path order of picking seedlings corresponding to this seedling supply group 1. The shortest path under this selection is recorded as the current optimal selection of healthy pot seedlings. Through the improved simulated annealing algorithm, the seedlings in the seedling supply group 1 are exchanged one by one with the remaining healthy pot seedlings in area B. For example, the seedling supply group 1 {1, 3, 5, 7, 14}. The shortest path under this selection is calculated again. At this time, two results will appear. One is that the path length after the exchange is longer than the path length before the exchange. At this moment, the order before the exchange remains unchanged. The other is that the path length after the exchange is shorter than the path length before the exchange. At this moment, the selection after the exchange covers the selection before the exchange and becomes the current optimal selection of healthy pot seedlings. The above exchange and calculation are repeated until the current temperature is lower than the lowest temperature T min , and the best seedling supply group and the corresponding optimal serial number selection of picking seedlings and the optimal centroid path order of picking seedlings are calculated.

[0050] P7: The best sorting of the centroids of the healthy pot seedlings in the seedling supply group for the manipulator to pick seedlings is obtained through the improved simulated annealing algorithm. This step is similar to the calculation of the centroid seedling supplement order of the manipulator in the target seedling tray and will not be elaborated here.

[0051] P8: The optimal serial number selection of the end effector for picking seedlings is calculated through the improved simulated annealing algorithm (which end effector should be used to pick a certain healthy pot seedling). This step is similar to the selection of the optimal serial number of the end effector for seedling supplement in the target seedling tray and will not be elaborated here.

[0052] P9: The seedling supplement path in the target seedling tray and the seedling picking path in the seedling supply tray are spliced and integrated to obtain the optimal global seedling supplement and transplant path of the manipulator based on multiple end effectors. See Figure 8 .

Claims

1. A multi-terminal seedling transplanting path optimization method based on an improved simulated annealing algorithm, including the planning of the seedling replenishment path in the target plug tray and the planning of the seedling removal path in the seedling supply plug tray, the seedling replenishment path in the target plug tray and the seedling removal path in the seedling supply plug tray are spliced ​​end to end to generate a global seedling replenishment operation transplanting path; The planning process of the seedling filling path in the target hole tray is as follows: mark and encode the holes to be filled in the target hole tray; mark and group the holes to be filled according to the number of end effectors; obtain the planting order of the manipulator mass center corresponding to the holes to be filled in the group and the selection of the end effector sequence corresponding to a hole to be filled by using the improved greedy simulated annealing algorithm; generate the shortest planting path for different groups to be filled in the target hole tray; The planning process of seedling retrieval path in seedling supply plug tray is as follows: the search area is delineated according to the starting and ending points of the shortest replanting path of different groups to be replanted in the target plug tray, and the healthy seedlings in the search area are marked and encoded; the best healthy seedling position in the search area is selected by the improved greedy simulated annealing algorithm, and the number of selected healthy seedlings is the same as the number of holes to be replanted in the corresponding group to be replanted, and this group of healthy seedlings is called the seedling supply group; the seedling retrieval order of the manipulator's centroid corresponding to the selected healthy seedlings in the seedling supply group and the end effector serial number selection corresponding to a certain healthy seedling are obtained by the improved greedy simulated annealing algorithm; the shortest seedling retrieval path for different seedling groups in the seedling supply plug tray is generated.

2. The method for optimizing the multi-terminal seedling transplanting path based on the improved simulated annealing algorithm according to claim 1, characterized in that: The marking and coding of the acupuncture points to be supplemented in the target acupuncture tray is specifically to mark each acupuncture point to be supplemented in the target acupuncture tray with a positive real number in the order from top to bottom and from right to left.

3. The method for optimizing the multi-terminal seedling transplanting path based on the improved simulated annealing algorithm according to claim 2, characterized in that: The holes to be filled are marked and grouped according to the number of end effectors, and the number of holes to be filled in each group is determined according to the number of end effectors; if the number of holes to be filled is greater than the number of end effectors, the holes are divided into several groups, and if the number of holes to be filled is less than the number of end effectors, the holes are grouped separately.

4. The method for optimizing the multi-terminal seedling transplanting path based on the improved simulated annealing algorithm according to claim 3, characterized in that: The steps for transplanting seedlings for each group of holes to be filled are: step 1, the robot starts from the starting point and goes to the seedling supply tray to pick up a group of healthy seedlings in a certain order; step 2, the robot with a group of healthy seedlings goes from the seedling supply tray to the target tray; step 3, the robot transplants a group of healthy seedlings into a group of holes to be filled in the target tray in a certain order; step 4, the robot returns to the seedling supply tray; step 5, repeat steps 1 to 4 until all the holes to be filled in the target tray are planted with healthy seedlings.

5. The method for optimizing the multi-terminal seedling transplanting path based on the improved simulated annealing algorithm according to claim 4, characterized in that: The method of obtaining the order of planting the center of mass of the manipulator corresponding to the holes to be filled in the group to be filled and the selection of the end effector serial number corresponding to a certain hole to be filled by using the improved greedy simulated annealing algorithm is to first convert the position of the hole to be filled into the center of mass position of the corresponding manipulator, then obtain the best order of traversing the center of mass positions of the manipulator corresponding to the planting of the holes to be filled in the group to be filled by using the improved greedy simulated annealing algorithm, and then calculate the best serial number selection of the end effector corresponding to a certain hole to be filled by using the improved greedy simulated annealing algorithm.

6. The multi-terminal seedling transplanting path optimization method based on improved simulated annealing algorithm according to claim 1 is characterized in that: The position of the center of mass of the manipulator relative to the hole to be filled changes with the order of releasing the healthy seedlings clamped by the end effector.

7. The method for optimizing the multi-terminal seedling transplanting path based on the improved simulated annealing algorithm according to claim 1, characterized in that: The improved greedy simulated annealing algorithm is used to obtain the optimal order of the center of mass positions of the manipulators corresponding to the holes to be filled in the group to be filled, and the steps are as follows: (1) The centroids close to the dividing line are allocated to the two ends of the group to be supplemented, and the centroids far from the dividing line are allocated to the middle of the group to be supplemented, and the centroids farther from the dividing line are allocated to the middle of the group to be supplemented with a greater probability; the dividing line is located between the target plug tray and the seedling supply plug tray, so as to obtain the shortest path of all centroids of the group to be supplemented; (2) A search area is defined according to the starting and ending points of the shortest replanting paths of different groups to be replanted in the target plug tray, and the healthy seedlings in the search area are marked and coded; if the number of healthy seedlings in the current search area does not meet the needs of the next group to be replanted, the breadth-first method is used to expand the search area B (3×6), search area C (5×7), etc., and encode them in the same way; (3) The improved greedy simulated annealing algorithm is used to select the best healthy seedling position in the search area, and the area is marked with positive real numbers in order from right to left and from top to bottom; the seedling group is exchanged with the other healthy seedlings in the area through the improved greedy simulated annealing algorithm, and the optimal seedling picking sequence corresponding to the centroid of the healthy seedlings in the seedling group is calculated by the robot and the optimal selection of the end effector sequence for the robot to grip the seedling group.

Images