Robot task allocation method based on improved multi-target genetic algorithm

Abstract

The invention discloses a robot task allocation method based on an improved multi-target genetic algorithm, and relates to the technical field of robot task allocation, and the method comprises the following steps: building a multi-target robot task allocation problem model, and initializing model parameters; and optimizing the multi-target robot task allocation problem model by adopting an improved multi-target genetic algorithm to obtain a task allocation result. According to the method, the population is divided into a plurality of sub-populations, improved crossover mutation operation, population knowledge sharing interaction operation and fusion selection operation are introduced, an improved multi-target genetic algorithm is formed and applied to optimization of a multi-target robot task allocation problem model, and the problems that an existing robot task allocation technology is poor in solving quality and high in efficiency can be effectively solved. The solving speed is slow, and the diversity of solution sets is poor.

Description

Technical Field

[0001] The present application relates to the technical field of robot task allocation, and in particular to a robot task allocation method based on an improved multi-objective genetic algorithm. Background Art

[0002] In recent years, robotics has been widely used in various fields. From manufacturing to healthcare, to the military and service industries, the use of robots not only improves work efficiency but also reduces the occurrence of human errors. Compared with a single robot, the use of multiple robots can significantly improve the performance and task completion of the system. Therefore, robotics research has gradually expanded from single robots to multi-robot systems (MRS), attracting more and more attention from researchers. In multi-robot systems, the task allocation problem is a key research topic. Multi-robot task allocation refers to designing an allocation scheme to assign multiple tasks to different robots in the system in order to achieve the goals of shortest task execution time, minimum resource consumption, and highest task completion.

[0003] In the prior art, Chinese patent CN111832725A discloses a multi-robot multi-task allocation method and device based on an improved genetic algorithm, the method comprising: establishing a consumption cost model; constructing an initial population; performing gene crossover processing on chromosomes in the current initial population as parent chromosomes to form a second population; eliminating half of the chromosomes in the second population to form a third population; performing mutation processing on the third population to generate a fourth population as a new initial population; calculating the cost value of each chromosome in the fourth population according to the consumption cost model, obtaining the chromosome with the smallest cost value, repeating the iteration until the convergence condition is met, and outputting the best allocation plan.

[0004] Although the above-mentioned existing technologies have made some progress in the field of multi-robot task allocation, the optimization effect in complex and multi-constrained environments is still not ideal and the solution quality is poor; secondly, when dealing with large-scale task allocation problems, the above-mentioned existing technologies have high computational complexity and slow solution speed; in addition, the above-mentioned existing technologies do not consider the conflicts between objectives in multi-objective optimization, and the diversity of solution sets is poor. Summary of the invention

[0005] The present application provides a robot task allocation method based on an improved multi-objective genetic algorithm, which is used to solve the problems of poor solution quality, slow solution speed and poor solution set diversity in existing robot task allocation technologies.

[0006] On the one hand, the present application provides a robot task allocation method based on an improved multi-objective genetic algorithm, comprising the following steps:

[0007] Step S1, establish a multi-objective robot task allocation problem model and initialize the model parameters.

[0008] Step S2, using an improved multi-objective genetic algorithm to optimize the multi-objective robot task allocation problem model to obtain a task allocation result.

[0009] In step S2, the improved multi-objective genetic algorithm includes:

[0010] Step S21, initialize the population and divide the population into several sub-populations.

[0011] Step S22, judging whether the number of iterations reaches the multi-population knowledge sharing interaction mechanism stage, if not, performing an improved crossover mutation operation, if reached, performing a population knowledge sharing interaction operation.

[0012] Step S23, determine whether the number of iterations reaches the fusion selection stage, if not, proceed to the next step, if reached, perform a fusion selection operation.

[0013] Step S24, determine whether the number of iterations reaches the maximum number of iterations. If not, loop to step S22. If reached, output the current optimal solution as the task allocation result.

[0014] In a possible implementation, in step S22, the improved crossover mutation operation includes:

[0015] Each subpopulation performs an improved crossover operation or an improved mutation operation based on the crossover rate and the mutation rate to obtain a subpopulation offspring, and then the subpopulation and the subpopulation offspring are merged and non-dominated sorted, and a new subpopulation is obtained according to the sorting result and the crowding distance.

[0016] In a possible implementation, the improved crossover operation adopts a task combination crossover operation.

[0017] In a possible implementation, the improved mutation operation adopts a global flip operation and / or a local flip operation.

[0018] In a possible implementation, in step S22, the population knowledge sharing interaction operation includes:

[0019] Each subpopulation compares its own optimized task point quantity range with the task point quantity range of other subpopulations, selects interception operation or insertion operation to perform population interaction, and obtains a new subpopulation.

[0020] In a possible implementation, when the task point quantity range optimized by a subpopulation itself is smaller than the task point quantity range of other subpopulations, a new subpopulation is generated within the optimized task point quantity range by intercepting the other subpopulations.

[0021] When the number of task points optimized by a subpopulation itself is larger than the number of task points of other subpopulations, a new subpopulation is generated within the optimized number of task points for other subpopulations through insertion operations.

[0022] In a possible implementation, in step S23, the fusion selection operation includes:

[0023] The original subpopulation is merged with the current subpopulation and then non-dominated sorting is performed, and a new subpopulation is obtained according to the sorting result and crowding distance.

[0024] A robot task allocation method based on an improved multi-objective genetic algorithm in this application has the following advantages:

[0025] By dividing the population into several subpopulations, and introducing improved crossover mutation operations, population knowledge sharing interaction operations, and fusion selection operations, an improved multi-objective genetic algorithm is formed and applied to the optimization of the multi-objective robot task allocation problem model, which can effectively solve the problems of poor solution quality, slow solution speed, and poor solution set diversity in the existing robot task allocation technology. Among them, several subpopulations and population knowledge sharing interaction operations improve the solution set diversity and solution speed, and improved crossover mutation operations and fusion selection operations further increase the wide distribution of solution sets and improve the solution quality. At the same time, the overall task allocation method can achieve the minimum time and maximum number of completed tasks under the robot endurance constraints, thereby improving the quality of task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 A flowchart of a robot task allocation method based on an improved multi-objective genetic algorithm provided in an embodiment of the present application;

[0028] Figure 2 A schematic diagram of subpopulation division provided in the embodiments of the present application;

[0029] Figure 3 A schematic diagram of the encoding form of the subpopulation provided in the embodiment of the present application;

[0030] Figure 4 A schematic diagram of a task combination cross-operation provided in an embodiment of the present application;

[0031] Figure 5A schematic diagram of a global flip operation provided in an embodiment of the present application;

[0032] Figure 6 A schematic diagram of a partial flip operation provided in an embodiment of the present application;

[0033] Figure 7 A schematic diagram of the interception operation in the population knowledge sharing interactive operation provided in the embodiment of the present application;

[0034] Figure 8 A schematic diagram of the insertion operation in the population knowledge sharing interactive operation provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0036] like Figure 1 As shown, the embodiment of the present application provides a robot task allocation method based on an improved multi-objective genetic algorithm, comprising the following steps:

[0037] Step S1, establish a multi-objective robot task allocation problem model and initialize the model parameters.

[0038] Step S2, using an improved multi-objective genetic algorithm to optimize the multi-objective robot task allocation problem model to obtain a task allocation result.

[0039] In step S2, the improved multi-objective genetic algorithm includes:

[0040] Step S21, initialize the population and divide the population into several sub-populations.

[0041] Step S22, judging whether the number of iterations reaches the multi-population knowledge sharing interaction mechanism stage, if not, performing an improved crossover mutation operation, if reached, performing a population knowledge sharing interaction operation.

[0042] Step S23, determine whether the number of iterations reaches the fusion selection stage, if not, proceed to the next step, if reached, perform a fusion selection operation.

[0043] Step S24, determine whether the number of iterations reaches the maximum number of iterations. If not, loop to step S22. If reached, output the current optimal solution as the task allocation result.

[0044] Specifically, in this embodiment, in step S1, the multi-objective robot task allocation problem model includes: a group of robots (in this embodiment, drones), represented by R = {r1, r2, ..., r q}, r q The qth robot performs a task in a specific area. A set of key locations in the given area is used as the task points, which is represented by B = {b1, b2, ..., b h}, b h represents the hth task point. When the robot reaches the task point, it can complete the corresponding task. The task allocation process is to divide B into q different subsets B1, B2, .., B q , B q represents the qth subset, each subset is assigned to a robot; the task order of each robot in the corresponding subset is recorded as represents the |StS of the ith subset i | sorted tasks; the model goal is to optimize the search task allocation decision by maximizing the number of search tasks completed, minimizing the average search time and minimizing the longest search time, thus forming the three objective functions f1, f2, and f3 of the multi-objective robot task allocation problem model as follows:

[0045] f1:Min argmax C i,k

[0046]

[0047] Among them, C i,k It represents the completion time of the i-th robot performing the k-th task in its task set.

[0048] The constraints of the multi-objective robot task allocation problem model are as follows:

[0049] S i,k ∈B i

[0050] D(StS i )≤F i

[0051]

[0052] Among them, F i Represents robot r i The maximum endurance time, D(StS i ) represents the completion of the task set StS i The travel time of all tasks in D(StS i ) and C i,k The calculation of is as follows:

[0053]

[0054] Among them, v(r i ) represents the speed of the ith robot, dis(StS i,k-1 ,StS i,k ) represents the task set StS i The Euclidean distance between the positions of the k-1th task point and the kth task point.

[0055] Under the constraints of a fixed number of robots and the maximum endurance of the robots, the multi-objective robot task allocation problem model in this application focuses on completing the search tasks to the maximum extent, minimizing the average search time of the search points, and ensuring that the consumption time of the task points with the longest search time is actually as short as possible. The goal of minimizing the average search time of the search points is to adjust the task planning routes of the entire group of robots to reduce the total time to complete the search tasks. At the same time, by minimizing the consumption time of the task points with the longest search time, the model can avoid the situation where the workload of each robot is extremely unbalanced, that is, one robot may undertake most of the search tasks, while other robots only undertake one or two search tasks. By setting the objective function, it can be ensured that each robot has a relatively fair distribution of tasks.

[0056] Specifically, in this embodiment, the model parameters of the multi-objective robot task allocation problem model include: task point location, number of task points, number of robots, robot driving speed v, robot endurance, maximum number of iterations Tmax, crossover rate, mutation rate, population size Popsize, number of iterations T_Int of the multi-population knowledge sharing interaction mechanism stage, and number of iterations T_mix of the fusion selection stage. In this embodiment, the number of task points is 9, the number of robots is 3, the robot driving speed v is 36 kilometers per hour, the robot endurance is 1.5 hours, the maximum number of iterations is 2500, the crossover rate is 0.5, the mutation rate is 0.5, the population size is 50, the number of iterations of the multi-population knowledge sharing interaction mechanism stage is 100, and the number of iterations of the fusion selection stage is 50.

[0057] Specifically, in this embodiment, in step S21, after the population is initialized, the population is divided into five sub-populations pop1, pop2, pop3, pop4 and pop5, that is, the population is divided into five regions according to the number of task points. In this embodiment, the number of task points is 9, and the five sub-populations pop1, pop2, pop3, pop4 and pop5 are responsible for optimizing the number of tasks in the range of [1,2], [3,4], [5,6], [7,8] and [9], respectively. Figure 2As shown. The initialization of each subpopulation is randomly generated in the area it is responsible for, and the encoding form of the subpopulation is a double-part chromosome structure, as shown Figure 3 shown. Figure 3 In the figure, the first part is the random order of the 9 task points, and the second part is the number of task points that the three robots are responsible for respectively. The corresponding tasks are represented in the order of the first part. Figure 3 In the example, robot 1 is responsible for tasks [7, 3, 6], robot 2 is responsible for tasks [1, 4, 8, 2], and robot 3 is responsible for tasks [9, 5].

[0058] Exemplarily, in step S22, the improved crossover mutation operation includes:

[0059] Each subpopulation performs an improved crossover operation or an improved mutation operation based on the crossover rate and the mutation rate to obtain a subpopulation offspring, and then the subpopulation and the subpopulation offspring are merged and non-dominated sorted, and a new subpopulation is obtained according to the sorting result and the crowding distance.

[0060] Exemplarily, the improved crossover operation adopts a task combination crossover operation.

[0061] like Figure 4 As shown, specifically, in this embodiment, the task combination crossover operation is to randomly combine the number of task points of each robot from the two parent subpopulations Dad and Mom, and the number of task points of the combined subpopulation offspring is the same as the number of task points of the parent subpopulation Mom. Figure 4 In , the left side of the dotted line indicates the task points that have been considered for visit, and the right side of the dotted line and the left side of the dashed line indicate the task points that have not been considered. When performing the task combination crossover operation, some continuous vectors {3}, {6} and {4,8} are randomly selected from each robot of the parent subpopulation Mom, and the remaining task points are arranged according to the sequence of the parent subpopulation Dad, and {5,2,7,9,1} is obtained, as shown in Figure 3 As shown, the generation of offspring involves randomly generating 3 integers that do not exceed the number of remaining tasks. This set of random integers divides {5,2,7,9,1} into {5,2}, {7,9}, and {1}, and the corresponding part of each robot selects the subsequent tasks according to these random integers and combines them with the part selected by the parent subpopulation Mom to create subpopulation offspring.

[0062] Exemplarily, the improved mutation operation adopts a global flip operation and / or a local flip operation.

[0063] like Figure 5As shown, specifically, in this embodiment, the global flip operation is to randomly select a segment of the entire task sorting, take the center point {7} as the symmetry point, and symmetrically flip the left and right {5,2,6} and {9,4,8} to obtain {8,4,9} and {6,2,5}.

[0064] like Figure 6 As shown, specifically, in this embodiment, the local flipping operation is performed in a task subset of each robot. Figure 6 In , the segments {3,5}, {7,9}, and {4,8} are randomly selected from the three robot subsets and symmetrically flipped to obtain {5,3}, {9,7}, and {8,4}.

[0065] Specifically, the global flip operation and the local flip operation help to adjust the sequence of the individual selected task points, thereby helping to find the optimal sorting order.

[0066] In this embodiment, a logical execution relationship of the improved crossover and mutation operation is given as follows: if r1 is less than the crossover rate, an improved crossover operation is performed on the subpopulation; if r1 is not less than the crossover rate and r2 is less than the mutation rate, a global flipping operation is performed on the subpopulation; if r1 is not less than the crossover rate and r2 is not less than the mutation rate, for the i-th robot starting from 1, if r3 is less than 0.5, a local flipping operation is performed on the i-th robot of the subpopulation; if r1 is not less than the crossover rate and r2 is not less than the mutation rate, and for the i-th robot starting from 1, r3 is not less than 0.5, a new individual is randomly generated within the range to obtain the subpopulation offspring, wherein r1, r2, and r3 are all random numbers in the range of (0,1).

[0067] In this embodiment, the improved mutation operation adopts a global flip operation or a local flip operation. In other possible embodiments, a local flip operation may be performed on the subpopulation after the global flip operation, so that the improved mutation operation adopts a global flip operation and a local flip operation.

[0068] Specifically, after obtaining the offspring population offspring k Then, the subpopulation pop k Offspring Pop k After merging, perform non-dominated sorting and obtain a new subpopulation pop based on the sorting result and crowding distance. k .

[0069] Exemplarily, in step S22, the population knowledge sharing interaction operation includes:

[0070] Each subpopulation compares its own optimized task point quantity range with the task point quantity range of other subpopulations, selects interception operation or insertion operation to perform population interaction, and obtains a new subpopulation.

[0071] Specifically, when the number of iterations reaches 100 of the iterations in the multi-population knowledge sharing interaction mechanism stage, the population knowledge sharing interaction operation is performed.

[0072] Exemplarily, when the task point quantity range optimized by a subpopulation itself is smaller than the task point quantity range of other subpopulations, a new subpopulation is generated within the optimized task point quantity range by intercepting the other subpopulations.

[0073] When the number of task points optimized by a subpopulation itself is larger than the number of task points of other subpopulations, a new subpopulation is generated within the optimized number of task points for other subpopulations through insertion operations.

[0074] like Figure 7 As shown, specifically, in this embodiment, when the number of task points optimized by a subpopulation itself is smaller than the number of task points of other subpopulations, a new subpopulation is generated within the optimized number of task points by intercepting the first 1 / 4 individuals of other subpopulations. Figure 7 In the , the three robots of the first 1 / 4 individuals of the other subpopulations respectively cut out fragments {3,7}, {1,2} and {4,8}. The total number of tasks is within the optimized number of task points. These three fragments are connected as the number of tasks selected by the new subpopulation individuals. The rest are arranged in the unselected area of ​​the new subpopulation individuals according to the order of the first 1 / 4 individuals of the other subpopulations.

[0075] like Figure 8 As shown, specifically, in this embodiment, when the number of task points optimized by a subpopulation itself is larger than the number of task points of other subpopulations, a new subpopulation is generated within the optimized number of task points by inserting the first 1 / 4 individuals of other subpopulations. Figure 8 In the , a subset of tasks for each robot is selected from individuals with a smaller number of tasks as the starting point of the new subpopulation, and unvisited task points are randomly selected in sequence to ensure that the total number of tasks matches the optimized range of task point numbers.

[0076] Exemplarily, in step S23, the fusion selection operation includes:

[0077] The original subpopulation is merged with the current subpopulation and then non-dominated sorting is performed, and a new subpopulation is obtained according to the sorting result and crowding distance.

[0078] Specifically, in this embodiment, when the number of iterations reaches 50 in the fusion selection stage, the five original subpopulations oldpop k Merge with the current subpopulation and obtain a new subpopulation based on the sorting results and crowding distance.

[0079] Specifically, in this embodiment, when the number of iterations reaches the maximum number of iterations 2500, the Pareto front (pareto), that is, the current optimal solution, is output as the task allocation result.

[0080] The embodiment of the present application forms an improved multi-objective genetic algorithm by dividing the population into several sub-populations, and introduces improved crossover and mutation operations, population knowledge sharing interaction operations, and fusion selection operations, and applies it to the optimization of the multi-objective robot task allocation problem model, which can effectively solve the problems of poor solution quality, slow solution speed, and poor solution set diversity in the existing robot task allocation technology. Among them, several sub-populations and population knowledge sharing interaction operations improve the solution set diversity and solution speed, and the improved crossover and mutation operations and fusion selection operations further increase the wide distribution of solution sets and improve the solution quality. At the same time, the overall task allocation method can achieve the minimum time and maximum number of completed tasks under the robot endurance constraints, thereby improving the quality of task allocation.

[0081] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0082] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A robot task allocation method based on an improved multi-objective genetic algorithm, characterized in that: The following steps are involved: Step S1, establishing a multi-objective robot task allocation problem model and initializing model parameters; Step S2, using an improved multi-objective genetic algorithm to optimize the multi-objective robot task allocation problem model to obtain a task allocation result; In step S2, the improved multi-objective genetic algorithm includes: Step S21, initializing the population and dividing the population into several subpopulations; Step S22, judging whether the number of iterations reaches the stage of multi-population knowledge sharing interaction mechanism, if not, performing an improved crossover mutation operation, if reached, performing a population knowledge sharing interaction operation; Step S23, judging whether the number of iterations reaches the fusion selection stage, if not, proceeding to the next step, if reached, performing the fusion selection operation; Step S24, determine whether the number of iterations reaches the maximum number of iterations. If not, loop to step S22. If reached, output the current optimal solution as the task allocation result.

2. A robot task allocation method based on an improved multi-objective genetic algorithm according to claim 1, characterized in that: In step S22, the improved crossover mutation operation includes: Each subpopulation performs an improved crossover operation or an improved mutation operation based on the crossover rate and the mutation rate to obtain a subpopulation offspring, and then the subpopulation and the subpopulation offspring are merged and non-dominated sorted, and a new subpopulation is obtained according to the sorting result and the crowding distance.

3. A robot task allocation method based on an improved multi-objective genetic algorithm according to claim 2, characterized in that: The improved crossover operation adopts task combination crossover operation.

4. The robot task allocation method based on improved multi-objective genetic algorithm according to claim 2 is characterized in that: The improved mutation operation adopts a global flip operation and / or a local flip operation.

5. The robot task allocation method based on improved multi-objective genetic algorithm according to claim 1 is characterized in that: In step S22, the population knowledge sharing interaction operation includes: Each subpopulation compares its own optimized task point quantity range with the task point quantity range of other subpopulations, selects interception operation or insertion operation to perform population interaction, and obtains a new subpopulation.

6. A robot task allocation method based on an improved multi-objective genetic algorithm according to claim 5, characterized in that: When the task point number range optimized by a subpopulation itself is smaller than the task point number range of other subpopulations, a new subpopulation is generated within the optimized task point number range by intercepting the other subpopulations; When the number of task points optimized by a subpopulation itself is larger than the number of task points of other subpopulations, a new subpopulation is generated within the optimized number of task points for other subpopulations through insertion operations.

7. The robot task allocation method based on improved multi-objective genetic algorithm according to claim 1 is characterized in that: In step S23, the fusion selection operation includes: The original subpopulation is merged with the current subpopulation and then non-dominated sorting is performed, and a new subpopulation is obtained according to the sorting result and crowding distance.

Images