Multi-robot task allocation method based on improved large neighborhood search under time constraints

Abstract

The present invention discloses a multi-robot task allocation method based on improved large neighborhood search under time constraints. First, the initial task allocation schemes of multiple robots are generated as the initial solution. Then, the simulated annealing algorithm and the adaptive large neighborhood search algorithm are combined to improve the initial solution. Finally, a new minimum density insertion operator is introduced to repair the solution to output the optimal solution. The present invention has superior performance in solving the multi-robot task allocation problem under time constraints. While ensuring the quality of the solution, it reduces the calculation time and improves the overall calculation efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot task allocation, and in particular relates to the design of a multi-robot task allocation method based on improved large neighborhood search under time constraints. Background Art

[0002] With the rapid development of the Internet of Things and smart devices, multi-robot task assignment (MRTA) is increasingly used in various scenarios such as logistics, smart manufacturing, disaster relief, and military reconnaissance. In these scenarios, multiple robots need to efficiently visit a set of target locations, which is a variant of the NP-hard vehicle routing problem (VRP). This means that when the problem scale is large, it is necessary to design effective heuristic algorithms to calculate suboptimal solutions.

[0003] Regarding the MRTA problem, existing algorithms are divided into two categories: exact algorithms and heuristic algorithms. Exact algorithms, such as the Hungarian algorithm, branch-and-bound algorithm, and dynamic programming algorithm, use mathematical and logical reasoning to ensure that the optimal solution to the problem is found. However, the computational cost of such algorithms rises sharply as the scale of the problem increases, resulting in reduced efficiency when dealing with large-scale problems. In sharp contrast to exact algorithms are heuristic algorithms, which use empirical rules or group intelligence that imitates natural phenomena to guide the search process. The goal of heuristic algorithms is to quickly find a solution that is close to the optimal solution, rather than spending a lot of time seeking the theoretical optimal solution. Typical representatives of heuristic algorithms include genetic algorithms, simulated annealing algorithms, neighborhood search algorithms, ant colony algorithms, taboo search algorithms, and auction algorithms, which have shown good solution efficiency and flexibility in practical applications.

[0004] In some scenarios, such as disaster relief, search and rescue, and logistics, some tasks usually need to be performed in a specific order due to urgency and dependency. For example, in disaster relief scenarios, the efficient conduct of search and rescue work often depends on the smoothness of roads. Therefore, the first task is to ensure the safety and smoothness of roads so that the search and rescue team can quickly enter the disaster area to carry out search and rescue work. In the field of logistics, the timely delivery of packages also depends on the preliminary preparation work. Only by successfully obtaining and processing the packages can the subsequent distribution process be ensured to proceed smoothly and meet customer needs. In intelligent manufacturing, each production step is closely connected. In the assembly process, the placement of screws is a prerequisite for the robot to tighten the screws. Therefore, in the wide application of multi-robot systems, the multi-robot task assignment problem with timing constraints (MRTAPC) has become a hot research topic, in which the robot can only perform a task after all the predecessor tasks of the task have been completed. MRTAPC is a variant of the NP-hard VRP problem.

[0005] The simulated annealing algorithm simulates the process of temperature drop and particle stabilization in physics, avoids falling into the local optimum through time-varying probability jumps, and gradually approaches the global optimal solution. This algorithm has wide applicability and is easy to implement, but the solution time may be long, and there is a risk of losing the currently known optimal solution. The large neighborhood search algorithm generates new candidate solutions by iteratively removing and inserting elements in the current solution, and continuously optimizes the quality of the solution in each iteration to seek a better solution. Although this algorithm performs well in global search, its high computational complexity leads to relatively slow convergence and long computation time. Summary of the invention

[0006] The purpose of the present invention is to propose a multi-robot task allocation method based on improved large neighborhood search under timing constraints. Under the conditions of timing constraints and the upper limit of the number of tasks that the robots can complete, tasks are allocated to multiple scattered task points in the environment while minimizing the total time cost of the robots visiting and serving all task points.

[0007] The technical solution of the present invention is: a multi-robot task allocation method based on improved large neighborhood search under time constraints, comprising the following steps:

[0008] S1. Initialize and set the relevant parameters of multi-robot task allocation and define the set of task points to be inserted.

[0009] S2. Update the set of task points to be inserted based on time cost and timing constraints.

[0010] S3. Determine whether the set of task points to be inserted is an empty set. If so, proceed to step S4; otherwise, return to step S2.

[0011] S4. The initial solution is saved as the solution before being destroyed and the current optimal solution, and the selection weight of the destruction operator and the initial temperature of the simulated annealing are initialized.

[0012] S5. Select a destruction operator according to the selection weight of the destruction operator to destroy the current solution, and add the deleted task points with an in-degree of 0 into the set of task points to be inserted.

[0013] S6. Determine whether the set of task points to be inserted is an empty set. If so, proceed to step S12. Otherwise, proceed to step S7 to repair the set of task points to be inserted using the minimum density insertion operator.

[0014] S7. Determine whether there is only one task point to be inserted in the set of task points to be inserted. If so, find the position of the task point to be inserted that adds the least time cost in the current robot path and insert it, and return to step S6; otherwise, go to step S8.

[0015] S8. Calculate the average distance matrix from multiple task points to be inserted in the set of task points to be inserted to all task points existing in the robot path, and calculate the number matrix of task points in the robot path that have been allocated within the circle near each task point to be inserted, taking each task point to be inserted as the center of the circle and the elements in the average distance matrix as the radius.

[0016] S9. Divide the task point quantity matrix by the average distance matrix to obtain a density matrix for each task point to be inserted.

[0017] S10, interfering with the density matrix, multiplying the point density in the density matrix by a random number of 0.8 to 1.2 in sequence, to prevent the minimum density insertion operator from repeatedly and preferentially inserting certain task points each time.

[0018] S11. Find the task point to be inserted with the minimum density in the density matrix after interference, and insert it into the position with the minimum increase in time cost in the current robot path, and return to step S6.

[0019] S12. Update the current solution and the current optimal solution.

[0020] S13, let the number of iterations and the optimal solution remain unchanged plus 1, and update the selection weight of the destruction operator and the temperature of simulated annealing.

[0021] S14, determine whether the number of iterations and the number of iterations with the optimal solution unchanged have reached the end requirement, if so, output the current optimal solution as the multi-robot task allocation result, otherwise return to step S5.

[0022] Furthermore, the relevant parameters of the multi-robot task allocation in step S1 include the set of m robots , robot initial position set , need service Mission point location set , The maximum number of service tasks allowed by the robot , the timing constraint matrix between tasks , the time cost required to serve each task point , the robot's constant driving speed , the optimal solution remains unchanged and the upper limit of the number of iterations , the maximum number of iterations , Iterations The optimal solution remains unchanged in the number of iterations ,in Represents the ceiling function.

[0023] Further, step S2 includes the following sub-steps:

[0024] S2-1. Set the task points to be inserted All task points in the are inserted into the end of each robot path in turn, the increased time cost is calculated, and then the task point is removed from the end of the robot path.

[0025] S2-2. Insert the task points with the smallest added time cost and satisfying the timing constraints into the robot path, and update the set of task points to be inserted .

[0026] Furthermore, the calculation method of the time cost added in step S2-1 is: continuously traverse the start time and completion time of each task point, and the timing constraint matrix between tasks In response to the start time of a task point being earlier than the completion time of its predecessor task point, the start time of this task point is updated until no task point timing constraints conflict in a round of calculation. The sum of the task completion time of all robots is taken as the time cost. The update formula of the start time of the task point is:

[0027] ;

[0028] in Represents the robot service task point The start time of Indicates the robot completes the task point The completion time, Represents the robot service task point The start time of Represents the robot service task point Service time, Represents the timing constraint matrix between tasks Middle Line The value of the column.

[0029] Further, step S4 includes the following sub-steps:

[0030] S4-1. Save the initial solution as the solution before it is destroyed , and calculate the time cost of the initial solution .

[0031] S4-2, the current optimal solution Assigned to , and the optimal time cost Assigned to .

[0032] S4-3. Initialize the selection weight of the destruction operator .

[0033] S4-4. Initialize the initial temperature of simulated annealing .

[0034] Furthermore, the destruction operator in step S5 includes a random deletion operator, a path deletion operator and a region deletion operator.

[0035] The random delete operator randomly selects and deletes Mission points to destroy the current solution ,in Indicates between and A random number between .

[0036] The path deletion operator randomly selects a robot and determines Whether it exceeds the number of tasks currently executed by the robot, if so, directly remove the entire task path of the robot, otherwise randomly delete it from the current task of the robot A mission point.

[0037] The region deletion operator selects a task point as the center and then identifies and selects the closest task point and remove it from the robot's path.

[0038] Further, step S12 includes the following sub-steps:

[0039] S12-1. Get the latest solution after repair , and calculate the time cost of the latest solution .

[0040] S12-2. Determine whether If so, update the current optimal solution , optimal time cost , let the selection weight of the destruction operator Add parameters , the optimal solution does not change the number of iterations , and go to step S12-4, otherwise go to step S12-3.

[0041] S12-3. Determine whether If yes, go to step S12-4, otherwise go to step S12-5.

[0042] S12-4. Update the current solution , the time cost of the current solution , let the selection weight of the destruction operator Add parameters , proceed to step S13.

[0043] S12-5. Use simulated annealing to decide whether to accept the latest solution , in response to the random number , then accept the latest solution , update the current solution , the time cost of the current solution , let the selection weight of the destruction operator Add parameters , enter step S13, where the parameter .

[0044] Furthermore, step S13 includes the following sub-steps:

[0045] S13-1, set the number of iterations The optimal solution remains unchanged in the number of iterations Add 1.

[0046] S13-2. Update the selection weight of the destruction operator :

[0047] ;

[0048] in represents the selection weight of the destruction operator after the update, Representation Operator The score obtained in this iteration, Represents the utility parameter.

[0049] S13-3. Update the temperature of simulated annealing ,in represents the updated simulated annealing temperature, in response to the optimal solution unchanged number of iterations The upper limit of the number of iterations for the optimal solution When the temperature of the simulated annealing is 0.3, 0.6 or 0.9 times of .

[0050] Furthermore, the end requirement in step S14 is:

[0051] Iterations Less than the maximum number of iterations Half of the optimal solution, and the number of iterations remains unchanged The upper limit of the number of iterations to reach the optimal solution remains unchanged ;

[0052] or number of iterations Greater than the maximum number of iterations Half of the maximum number of iterations and less than the upper limit 3 / 4 of the optimal solution, and the number of iterations remains unchanged The upper limit of the number of iterations to reach the optimal solution remains unchanged 3 / 4;

[0053] or number of iterations Greater than the maximum number of iterations 3 / 4 and less than the maximum number of iterations , and the optimal solution does not change the number of iterations The upper limit of the number of iterations to reach the optimal solution remains unchanged half of;

[0054] or number of iterations Reached the maximum number of iterations .

[0055] The beneficial effects of the present invention are as follows: the present invention combines the global exploration of the simulated annealing algorithm and the local optimization capability of the adaptive large neighborhood search algorithm, and is specifically optimized for the MRTAPC problem, which can not only find potential high-quality solutions in a wide search space, but also conduct refined searches in a local range. The present invention has superior performance in solving the multi-robot task allocation problem with timing constraints, while ensuring the quality of the solution, reducing the calculation time and improving the overall calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Shown is a flow chart of a multi-robot task allocation method based on improved large neighborhood search under timing constraints provided by an embodiment of the present invention.

[0057] Figure 2 Shown is a schematic diagram of a destruction operator provided by an embodiment of the present invention.

[0058] Figure 3 FIG. 4 is a schematic diagram of a minimum density insertion operator provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] Now, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the accompanying drawings are only exemplary and are intended to explain the principles and spirit of the present invention, rather than to limit the scope of the present invention.

[0060] The embodiment of the present invention provides a multi-robot task allocation method based on improved large neighborhood search under time constraints, such as Figure 1 As shown, the following steps S1 to S14 are included:

[0061] S1. Initialize and set the relevant parameters of multi-robot task allocation and define the set of task points to be inserted.

[0062] In the embodiment of the present invention, the relevant parameters of multi-robot task allocation include a set of m robots: , robot initial position set , need service Mission point location set , The maximum number of service tasks allowed by the robot , the timing constraint matrix between tasks , the time cost required to serve each task point , the robot's constant driving speed , the optimal solution remains unchanged and the upper limit of the number of iterations , the maximum number of iterations , Iterations The optimal solution remains unchanged in the number of iterations ,in Represents the ceiling function.

[0063] In the embodiment of the present invention, the set of task points to be inserted is Contains task points that have not yet been assigned to the robot path and have no predecessor tasks or whose predecessor tasks have been added to the robot path.

[0064] S2. Update the set of task points to be inserted based on time cost and timing constraints.

[0065] Step S2 includes the following sub-steps S2-1 to S2-2:

[0066] S2-1. Set the task points to be inserted All task points in the are inserted into the end of each robot path in turn, the increased time cost is calculated, and then the task point is removed from the end of the robot path.

[0067] In the embodiment of the present invention, the calculation method of the increased time cost is: continuously traverse the start time and completion time of each task point, and the timing constraint matrix between tasks In response to the start time of a task point being earlier than the completion time of its predecessor task point, the start time of this task point is updated until no task point timing constraints conflict in a round of calculation. The sum of the task completion time of all robots is taken as the time cost. The update formula of the start time of the task point is:

[0068] ;

[0069] in Represents the robot service task point The start time of Indicates the robot completes the task point The completion time, Represents the robot service task point The start time of Represents the robot service task point Service time, Represents the timing constraint matrix between tasks Middle Line The value of the column. To the task point When there is a time-series relationship requirement, that is, the task point is required The completion time must be earlier than the task point The start time, i.e. ,at this time Value is , otherwise In the process of calculating time, if the task point or mission point If it is not in the robot path, it will not be calculated.

[0070] S2-2. Insert the task points with the smallest added time cost and satisfying the timing constraints into the robot path, and update the set of task points to be inserted .

[0071] S3. Determine whether the set of task points to be inserted is an empty set. If so, proceed to step S4; otherwise, return to step S2.

[0072] S4. The initial solution is saved as the solution before being destroyed and the current optimal solution, and the selection weight of the destruction operator and the initial temperature of the simulated annealing are initialized.

[0073] Step S4 includes the following sub-steps S4-1 to S4-4:

[0074] S4-1. Save the initial solution as the solution before it is destroyed , and calculate the time cost of the initial solution .

[0075] In the embodiment of the present invention, the time cost of the initial solution is Satisfies the following formula:

[0076]

[0077] S4-2, the current optimal solution Assigned to , and the optimal time cost Assigned to .

[0078] S4-3. Initialize the selection weight of the destruction operator .

[0079] S4-4. Initialize the initial temperature of simulated annealing .

[0080] In the embodiment of the present invention, the solution corresponds to a multi-robot task allocation scheme.

[0081] S5. Select a destruction operator according to the selection weight of the destruction operator to destroy the current solution, and add the deleted task points with an in-degree of 0 (there is no predecessor task in the current state or the predecessor task has been added to the robot path) into the set of task points to be inserted.

[0082] like Figure 2 As shown, the destruction operators include random deletion operators, path deletion operators, and region deletion operators.

[0083] Random deletion operator such as Figure 2 As shown in (a), by randomly selecting and deleting Mission points to destroy the current solution ,in Indicates between and Since the relative positions between task points are not considered, the random deletion operator can perform deletion operations at any position of the robot path.

[0084] The path deletion operator is as follows: Figure 2 As shown in (b), a robot is randomly selected and judged Whether it exceeds the number of tasks currently executed by the robot, if so, directly remove the entire task path of the robot, otherwise randomly delete it from the current task of the robot A mission point.

[0085] The region deletion operator is Figure 2 As shown in (c), a task point is selected as the center, and then the closest point to the center is identified and selected. The region removal operator considers the spatial relationship between tasks and helps to maintain the structural characteristics of the solution during the removal process.

[0086] S6. Determine whether the set of task points to be inserted is an empty set. If so, proceed to step S12. Otherwise, proceed to step S7 to repair the set of task points to be inserted using the minimum density insertion operator.

[0087] S7. Determine whether there is only one task point to be inserted in the set of task points to be inserted. If so, find the position of the task point to be inserted that adds the least time cost in the current robot path and insert it, and return to step S6; otherwise, go to step S8.

[0088] S8, such as Figure 3 As shown, when the task point set to be inserted When there are multiple task points to be inserted in the robot path, the average distance matrix from multiple task points to be inserted in the task point set to all task points in the robot path is calculated. , and take each task point to be inserted as the center of the circle, and use the average distance matrix The elements in are the radius, and the number matrix of task points in the assigned robot path in the circle near each task point to be inserted is calculated. .

[0089] S9. Divide the task point quantity matrix by the average distance matrix to obtain the density matrix of each task point to be inserted :

[0090] ;

[0091] S10, density matrix Interference is performed, and the density matrix The point density in is multiplied by a random number between 0.8 and 1.2 to prevent the minimum density insertion operator from repeatedly inserting certain task points each time.

[0092] In the embodiment of the present invention, the density matrix It includes two columns, the first column represents a task point to be inserted, and the second column represents the point density corresponding to the task point to be inserted.

[0093] S11. Density matrix after interference Find the task point to be inserted with the minimum density, and insert it into the position where the time cost increases the least in the current robot path, and return to step S6.

[0094] S12. Update the current solution and the current optimal solution.

[0095] Step S12 includes the following sub-steps S12-1 to S12-5:

[0096] S12-1. Get the latest solution after repair , and calculate the time cost of the latest solution .

[0097] S12-2. Determine whether If so, update the current optimal solution , optimal time cost , let the selection weight of the destruction operator Add parameters , the optimal solution does not change the number of iterations , and go to step S12-4, otherwise go to step S12-3.

[0098] S12-3. Determine whether If yes, go to step S12-4, otherwise go to step S12-5.

[0099] S12-4. Update the current solution , the time cost of the current solution , let the selection weight of the destruction operator Add parameters , proceed to step S13.

[0100] S12-5. Use simulated annealing to decide whether to accept the latest solution , in response to the random number , then accept the latest solution , update the current solution , the time cost of the current solution , let the selection weight of the destruction operator Add parameters , enter step S13, where the parameter .

[0101] S13, let the number of iterations and the optimal solution remain unchanged plus 1, and update the selection weight of the destruction operator and the temperature of simulated annealing.

[0102] Step S13 includes the following sub-steps S13-1 to S13-3:

[0103] S13-1, set the number of iterations The optimal solution remains unchanged in the number of iterations Add 1.

[0104] S13-2. Update the selection weight of the destruction operator :

[0105] ;

[0106] in represents the selection weight of the destruction operator after the update, Representation Operator The score obtained in this iteration, Represents the utility parameter.

[0107] S13-3. Update the temperature of simulated annealing ,in represents the updated simulated annealing temperature, in response to the optimal solution unchanged number of iterations The upper limit of the number of iterations for the optimal solution When the temperature of the simulated annealing is 0.3, 0.6 or 0.9 times of .

[0108] S14, determine whether the number of iterations and the number of iterations with the optimal solution unchanged have reached the end requirement, if so, output the current optimal solution as the multi-robot task allocation result, otherwise return to step S5.

[0109] In the embodiment of the present invention, the termination requirement is:

[0110] (1) Number of iterations Less than the maximum number of iterations Half of the optimal solution, and the number of iterations remains unchanged The upper limit of the number of iterations to reach the optimal solution remains unchanged .

[0111] (2) Number of iterations Greater than the maximum number of iterations Half of the maximum number of iterations and less than the upper limit 3 / 4 of the optimal solution, and the number of iterations remains unchanged The upper limit of the number of iterations to reach the optimal solution remains unchanged 3 / 4 of.

[0112] (3) Number of iterations Greater than the maximum number of iterations 3 / 4 and less than the maximum number of iterations , and the optimal solution does not change the number of iterations The upper limit of the number of iterations to reach the optimal solution remains unchanged half.

[0113] (4) Number of iterations Reached the maximum number of iterations .

[0114] When any of the above end requirements is met, the current optimal solution is output As the multi-robot task allocation result, otherwise return to step S5.

[0115] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A multi-robot task allocation method based on improved large neighborhood search under time constraints, characterized in that: The following steps are involved: S1. Initialize and set the relevant parameters of multi-robot task allocation and define the set of task points to be inserted; S2, update the set of task points to be inserted based on time cost and timing constraints; S3, determine whether the set of task points to be inserted is an empty set, if so, proceed to step S4, otherwise return to step S2; S4, saving the initial solution as the solution before being destroyed and the current optimal solution, and initializing the selection weight of the destruction operator and the initial temperature of the simulated annealing; S5. Select a destruction operator according to the selection weight of the destruction operator to destroy the current solution, and add the deleted task points with an in-degree of 0 to the set of task points to be inserted; S6, determine whether the set of task points to be inserted is an empty set, if so, proceed to step S12, otherwise proceed to step S7, use the minimum density insertion operator to repair the set of task points to be inserted; S7, determine whether there is only one task point to be inserted in the set of task points to be inserted, if so, find the position of the task point to be inserted that adds the least time cost in the current robot path and insert it, return to step S6, otherwise go to step S8; S8, calculating the average distance matrix from multiple task points to be inserted in the set of task points to be inserted to all task points existing in the robot path, and taking each task point to be inserted as the center of the circle and the elements in the average distance matrix as the radius, calculating the number matrix of task points in the robot path that have been allocated in the circle near each task point to be inserted; S9, dividing the task point quantity matrix by the average distance matrix to obtain a density matrix of each task point to be inserted; S10, interfering with the density matrix, multiplying the point density in the density matrix by a random number of 0.8 to 1.2 in sequence, to prevent the minimum density insertion operator from repeatedly inserting certain task points in priority each time; S11, searching for the task point to be inserted with the minimum density in the density matrix after interference, and inserting it into the position where the time cost is increased the least in the current robot path, and returning to step S6; S12, updating the current solution and the current optimal solution; S13, adding 1 to the number of iterations and the number of iterations with the optimal solution unchanged, and updating the selection weight of the destruction operator and the temperature of simulated annealing; S14, judging whether the number of iterations and the number of iterations in which the optimal solution remains unchanged have reached the end requirement, if so, outputting the current optimal solution as the multi-robot task allocation result, otherwise returning to step S5; The step S12 comprises the following sub-steps: S12-1. Get the latest solution after repair , and calculate the time cost of the latest solution ; S12-2. Determine whether If so, update the current optimal solution , optimal time cost , let the selection weight of the destruction operator Add parameters , the optimal solution does not change the number of iterations , and proceed to step S12-4, otherwise proceed to step S12-3; S12-3. Determine whether If yes, go to step S12-4, otherwise go to step S12-5; S12-4. Update the current solution , the time cost of the current solution , let the selection weight of the destruction operator Add parameters , proceed to step S13; S12-5. Use simulated annealing to decide whether to accept the latest solution , in response to the random number , then accept the latest solution , update the current solution , the time cost of the current solution , let the selection weight of the destruction operator Add parameters , enter step S13, where the parameter ; The step S13 comprises the following sub-steps: S13-1, set the number of iterations The optimal solution remains unchanged in the number of iterations Add 1; S13-2. Update the selection weight of the destruction operator : ； in represents the selection weight of the destruction operator after the update, Representation Operator The score obtained in this iteration, represents the utility parameter; S13-3. Update the temperature of simulated annealing ,in represents the updated simulated annealing temperature, in response to the optimal solution unchanged number of iterations The upper limit of the number of iterations for the optimal solution When the temperature of the simulated annealing is 0.3, 0.6 or 0.9 times of .

2. The multi-robot task allocation method based on improved large neighborhood search under time constraints according to claim 1 is characterized in that: The relevant parameters of the multi-robot task allocation in step S1 include a set of m robots , robot initial position set , need service Mission point location set , The maximum number of service tasks allowed by the robot , the timing constraint matrix between tasks , the time cost required to serve each task point , the robot's constant driving speed , the optimal solution remains unchanged and the upper limit of the number of iterations , the maximum number of iterations , Iterations The optimal solution remains unchanged in the number of iterations ,in Represents the ceiling function.

3. The multi-robot task allocation method based on improved large neighborhood search under time constraints according to claim 2 is characterized in that: The step S2 comprises the following sub-steps: S2-1. Set the task points to be inserted All task points in the robot path are inserted into the end of each robot path in turn, the increased time cost is calculated, and then the task point is removed from the end of the robot path; S2-2. Insert the task points with the smallest added time cost and satisfying the timing constraints into the robot path, and update the set of task points to be inserted .

4. The multi-robot task allocation method based on improved large neighborhood search under time constraints according to claim 3 is characterized in that: The calculation method of the time cost added in step S2-1 is: continuously traverse the start time and completion time of each task point, and the timing constraint matrix between tasks In response to the start time of a task point being earlier than the completion time of its predecessor task point, the start time of this task point is updated until no timing constraints of the task point conflict in a round of calculation. The sum of the task completion time of all robots is taken as the time cost. The update formula of the start time of the task point is: ； in Represents the robot service task point The start time of Indicates the robot completes the task point The completion time, Represents the robot service task point The start time of Represents the robot service task point Service time, Represents the timing constraint matrix between tasks Middle Line The value of the column.

5. The multi-robot task allocation method based on improved large neighborhood search under time constraints according to claim 4 is characterized in that: The step S4 comprises the following sub-steps: S4-1. Save the initial solution as the solution before it is destroyed , and calculate the time cost of the initial solution ; S4-2, the current optimal solution Assigned to , and the optimal time cost Assigned to ; S4-3. Initialize the selection weight of the destruction operator ; S4-4. Initialize the initial temperature of simulated annealing .

6. The multi-robot task allocation method based on improved large neighborhood search under time constraints according to claim 5 is characterized in that: The destruction operators in step S5 include a random deletion operator, a path deletion operator and an area deletion operator; The random deletion operator randomly selects and deletes Mission points to destroy the current solution ,in Indicates between and A random number between The path deletion operator randomly selects a robot and determines Whether it exceeds the number of tasks currently executed by the robot, if so, directly remove the entire task path of the robot, otherwise randomly delete it from the current task of the robot Mission points; The region deletion operator selects a task point as the center and then identifies and selects the nearest task point and remove it from the robot's path.

7. The multi-robot task allocation method based on improved large neighborhood search under time constraints according to claim 2 is characterized in that: The end requirement in step S14 is: Iterations Less than the maximum number of iterations Half of the optimal solution, and the number of iterations remains unchanged The upper limit of the number of iterations to reach the optimal solution remains unchanged ; or number of iterations Greater than the maximum number of iterations Half of the maximum number of iterations and less than the upper limit 3 / 4 of the optimal solution, and the number of iterations remains unchanged The upper limit of the number of iterations to reach the optimal solution remains unchanged 3 / 4; or number of iterations Greater than the maximum number of iterations 3 / 4 and less than the maximum number of iterations , and the optimal solution does not change the number of iterations The upper limit of the number of iterations to reach the optimal solution remains unchanged half of; or number of iterations Reached the maximum number of iterations .

Images