Multi-target multi-robot task allocation method based on collaborative ant colony

Abstract

The invention discloses a multi-target multi-robot task allocation method based on a collaborative ant colony, and the method comprises the steps: constructing a multi-target multi-robot task allocation model, and carrying out the design of a collaborative ant colony algorithm; three ant colonies are set in the ant colony algorithm and have three optimization objectives of minimizing the maximum completion time of all the tasks, minimizing the average movement time of all the robots and minimizing the average waiting time of all the tasks, and a multi-objective multi-robot task allocation model is solved; optimal solutions on the three optimization targets are selected through local search, and pheromones of the three collaborative ant colonies are globally updated; and outputting an optimal solution set of the multi-target multi-robot task allocation model. Compared with a traditional multi-target method, the method is stable in performance on the task allocation problem of multiple robots of different scales, and a series of high-quality feasible solutions can be obtained within a short time.

Description

Technical Field

[0001] The present invention relates to the field of multi-robot task allocation problem and intelligent computing search technology, and in particular to a robot task allocation method based on cooperative ant colonies for a multi-robot system. Background Art

[0002] With the advent of the era of industrial intelligence, multi-robot systems, as a new type of intelligent manufacturing and logistics technology, are increasingly becoming an important driving force for promoting industrial upgrading and development. Compared with traditional single robots, multi-robot systems have many advantages. They can perform tasks at a lower cost and higher system reliability, have greater flexibility and scalability, and can adapt to task requirements of different scales and complexities. According to the type of robot, multi-robot systems can be divided into homogeneous multi-robot systems and heterogeneous multi-robot systems. In many practical applications, heterogeneous multi-robot systems have shown excellent adaptability and performance advantages. By having multiple robots work together, tasks can be completed more efficiently, improving production efficiency and quality. For example, in the field of picking and delivering goods, heterogeneous multi-robot systems can complete various complex tasks such as handling, sorting and packaging through the expertise and capabilities of different robots, thereby improving logistics efficiency and flexibility. Therefore, if tasks are reasonably assigned to robots, scheduling their efficient collaboration is the focus of research and application of multi-robot systems.

[0003] In addition, with the rapid development of modern industry and the improvement of intelligence level, the application scenarios of multi-robot systems have gradually become large-scale and complex, which has brought severe challenges to traditional multi-robot task allocation methods. First, the large-scale characteristics make the search space of multi-robot task allocation problems grow exponentially with the increase in the number of robots and tasks. Traditional task allocation methods based on rules or greedy strategies are difficult to solve efficiently or even effectively, and often cannot meet the needs of large-scale task allocation problems. Secondly, the complexity of the application scenarios of multi-robot systems is reflected in the emergence of a variety of complex constraints, such as pre-order constraints between tasks, and resource and capability constraints between tasks and robot alliances. In real applications, there is often a dependency relationship between different tasks in terms of execution order, that is, some tasks must be completed before other tasks, which increases the complexity of the task allocation problem. At the same time, each robot has its own resource and capability limitations when performing tasks, such as carrying capacity, working time, etc. These constraints make the modern multi-robot task allocation problem more difficult and complex, and need to be effectively considered in the task allocation process. At the same time, it is also more challenging to combine the decision maker's multi-objective decision-making perspective to perform task allocation.

[0004] As an important and effective swarm intelligence optimization algorithm, the ant colony optimization algorithm mainly constructs solutions by simulating the foraging behavior of ants in nature, and has good global search capabilities. In addition, since the ant colony optimization algorithm gradually completes the construction of solutions through the movement of ants, the algorithm has a natural advantage in solving discrete combinatorial optimization problems, and is therefore widely used to solve complex discrete combinatorial optimization problems. At the same time, unlike other swarm intelligence optimization algorithms, the heuristic information of the ant colony optimization algorithm introduces problem-related knowledge. By designing appropriate heuristic information, the ant colony optimization algorithm can quickly find high-quality solutions. The multi-robot task allocation problem is essentially a discrete combinatorial optimization problem, so it is feasible to design a new multi-robot task allocation method based on the ant colony optimization algorithm. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the multi-robot task allocation problem models and methods in the prior art, and to provide a multi-objective multi-robot task allocation method based on collaborative ant colonies. Under the conditions of heterogeneous multi-robot collaboration, the pre-order constraints between tasks and the resource and capacity constraints between tasks and robot alliances are considered simultaneously, and a three-objective form is adopted to simultaneously optimize the maximum completion time of all tasks, the average moving time of all robots and the average waiting time of all tasks; at the same time, it is proposed to embed the ant colony algorithm into a multi-colony and multi-objective framework, and use the multi-objective collaborative ant colony algorithm to solve the proposed multi-robot task allocation problem.

[0006] The purpose of the present invention can be achieved by adopting the following technical solutions:

[0007] A multi-objective multi-robot task allocation method based on collaborative ant colonies, comprising:

[0008] Preprocess the task and heterogeneous robot data to obtain problem input and build a multi-objective multi-robot task allocation model;

[0009] A collaborative ant colony algorithm is designed. Three ant colonies are set in the ant colony algorithm. The three ant colonies are set with three optimization objectives: minimizing the maximum completion time of all tasks, minimizing the average movement of all robots, and minimizing the average waiting time of all tasks. The three optimization objectives are used to solve the multi-objective multi-robot task allocation model; the termination condition of the collaborative ant colony algorithm is set, and the constructed collaborative ant colony algorithm is executed iteratively for multiple times. In each iteration, the ant individuals in the three ant colonies respectively construct paths including task-task execution sequence paths and task-robot allocation paths according to the pheromones and heuristic information of their ant colonies, and search for feasible solutions from different directions with the three optimization objectives to obtain new feasible solutions for the ant individuals; the fitness values ​​of the new feasible solutions of all ant individuals are evaluated, and the high-quality solutions on each optimization objective are selected for local search, and the optimal solutions on the three optimization objectives are selected, and the pheromones of the three collaborative ant colonies are globally updated respectively;

[0010] When the termination condition of the collaborative ant colony algorithm is met, the optimization process ends and the optimal solution set of the multi-objective multi-robot task allocation model is output.

[0011] In some embodiments, the step of constructing a path based on the pheromones and heuristic information of the ant colony further includes: the individual ant first selects the next task to be performed based on the static heuristic information and the pheromones between tasks, and then calculates the dynamic heuristic information between the task and the robot and combines the pheromones between the task and the robot to construct a robot alliance that performs the task; at the same time, a local update of the pheromones is performed on each task-task and task-robot path of the feasible solution constructed by each ant.

[0012] In some implementations, the specific expressions of the three optimization objectives are as follows:

[0013]

[0014]

[0015]

[0016] FT i =WT i +ET i

[0017]

[0018]

[0019]

[0020] Where m is the number of tasks, n is the number of robots, i, k are task indexes, j is the robot index, f1(X) is the maximum completion time of all tasks, f2(X) is the average moving time of all robots, f3(X) is the average waiting time of all tasks, and FT i Represents task t i Completion time, WT i Represents task t i Waiting time, ET i Represents task t i The actual execution time (when i = 0, FT i =WT i =ET i =0), TT j Represents robot r j The total moving time, T represents a set of m tasks, R represents a set of n heterogeneous robots, AR i Indicates the execution of task t i The Robot League contains a collection of robots, Represents robot r j The completion time of the previous task, dist(TP i ,TP j ) represents task t i and task t j The Euclidean distance between the task points, Represents robot r j The task point of the previous task, TP i Represents task t i The mission point location, V j Represents robot r j The moving speed, L represents the number of different capabilities required by the task and that the robot can provide, Indicates the execution of task t i The required first ability value, Represents robot r j The first capability value, RP0, represents the common starting point of all robots. Represents robot r j The task point location of the hth task to be executed, |RT j | indicates robot r j The total number of tasks executed.

[0021] In some embodiments, constructing a path based on the pheromone and heuristic information of the ant colony further includes: the ant individual first constructs a path based on the static heuristic information n 1 and the pheromone τ between task-task paths 1,cSelect the next task to be performed and then calculate the dynamic heuristic information η between the task and the robot path 2 And combine the task with the pheromone τ between robots 2,c A robot alliance is built to perform the task, and at the same time, local updates are performed on each task-task and task-robot path of the feasible solution constructed by each individual ant to construct a new feasible solution for the individual ant.

[0022] In some embodiments, the pheromone of the ant colony further includes a pheromone matrix τ accumulated between tasks by each ant colony. 1,c and the pheromone matrix τ accumulated between the task and the robot 2,c , the two pheromone matrices are in the following form:

[0023]

[0024]

[0025] Among them, m represents the number of tasks, n represents the number of robots, and c represents the index of the three ant colonies. (0≤i,k≤m) represents the ant colony c in task t i With the task k The pheromone value between Indicates that ant colony c is in task t i With robot j pheromone values ​​between .

[0026] In some embodiments, the feasible solution is further obtained by each ant individual using the static heuristic information n between tasks. 1 and the dynamic heuristic information η between the task and the robot 2,c Build;

[0027] Represents task t i and task t k The static heuristic information between is calculated as follows:

[0028]

[0029] Among them, dist(TP i ,TP k ) represents task t i and task t k The Euclidean distance between the task points;

[0030] Task i and robot r j Dynamic heuristic information between The calculation formula is as follows:

[0031]

[0032]

[0033] Where Δf c If you select robot r j Execute Task i , the target value increment of the solution obtained on the target c, FT i Represents task t i The completion time, T f Indicates the set of tasks that have been assigned, AR i Indicates that the allocated space is used to execute task t i A collection of robots, Indicates the last assigned to robot r g The mission point location, WT i Represents task t i waiting time.

[0034] In some embodiments, the process of constructing a feasible solution further comprises:

[0035] (i) The feasible solution X c,e RT with each robot's task execution sequence j Initialized to empty;

[0036] (ii) Select currently executable tasks from all unassigned tasks and add them to the set T p ;

[0037] (iii) Based on the pheromone matrix τ 1,c and static heuristic information η 1 From the set T p Select the next task to be assigned;

[0038] (iv) Construct execution task t i The Robot Alliance, which includes the robots in the AR i Initialized to empty;

[0039] (v) All the tasks not assigned to task t i The robot is added to the set R'; the dynamic heuristic information η of each robot in the set R' is calculated 2,c ; Binding pheromone τ 2,c Select execution task t from the set R' i Robot j ;

[0040] (vi) Repeat step (v) until the constructed robot alliance can meet the task ti capacity and resource constraints;

[0041] (vii) According to the robot alliance AR i Sort each robot in descending order according to its number as the robot to perform task t i The Robot Alliance;

[0042] (viii) According to the selected task t i and Robot Alliance z =(AID z ,AR i ), update the feasible solution X c,e RT with each robot's task execution sequence j ;

[0043] (viiii) Repeat steps (ii)-(viii) until all tasks are assigned and output a feasible solution X c,e .

[0044] In some implementations, a crossover operation is used to perform local searches between different feasible solutions and merge them into a new feasible solution. The crossover operation further includes pruning, splicing, removing duplicate tasks, filling in missing tasks, and repairing violated precedence constraints.

[0045] In some embodiments, the local search further includes: determining whether the new robot alliance can meet the resource and capability requirements of the task; if so, updating the corresponding new robot alliance in the feasible solution; otherwise, re-performing the mutation operation until the above mutation operation is performed on all robot alliances, and finally obtaining a new feasible solution.

[0046] In some embodiments, the step of performing a global update of the pheromone of the three cooperative ant colonies further includes: 1,c and τ 2,c The global update expression is as follows:

[0047]

[0048] Among them, s∈{1,2}, c∈{1,2,3} represent the indexes of three ant colonies and also represent the indexes of three optimization objectives, ρ g = 0.1 indicates the evaporation rate of the global update of pheromone, represents the optimal feasible solution on target c, Δτ c represents the pheromone increment on ant colony c, and the calculation formula is as follows:

[0049]

[0050] in, represents the optimal feasible solution on target c The target value on target c.

[0051] Compared with the prior art, the beneficial effects and advantages of the present invention are:

[0052] 1) A multi-objective multi-robot task allocation model is constructed under heterogeneous multi-robot collaboration conditions, which takes into account multiple complex constraints such as the order constraints between tasks. An efficient multi-objective collaborative ant colony algorithm designed based on the ant colony algorithm and the multi-swarm multi-objective framework is used to solve the model. Dynamic heuristic information and dual pheromones are used to construct feasible solutions, which can effectively narrow the search space and utilize historical experience, thereby improving the efficiency of multi-objective multi-robot task allocation.

[0053] 2) Three ant colonies are set with three optimization goals: minimizing the maximum completion time of all tasks, minimizing the average moving time of all robots, and minimizing the average waiting time of all tasks. The three optimization goals are used to solve the multi-objective multi-robot task allocation model; the multi-objective multi-robot task allocation model is solved from the three goals to obtain the optimal solution set, which provides an optimization solution for multi-objective multi-robot task allocation;

[0054] 3) A new fusion-based local search method is implemented to fully integrate the high-quality solution information of different target populations and further comprehensively explore the Pareto frontier. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is an overall flow chart of a multi-objective multi-robot task allocation method based on collaborative ant colonies of the present invention;

[0056] Figure 2 This is a specific algorithm flow chart of a multi-objective multi-robot task allocation method based on collaborative ant colonies according to an embodiment of the present invention;

[0057] Figure 3 is a schematic diagram of a crossover operation of a local search in an embodiment of the present invention;

[0058] Figure 4 It is a schematic diagram of a feasible solution in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] like Figure 1 As shown, a multi-objective multi-robot task allocation method based on collaborative ant colonies of the present invention comprises the following steps:

[0061] Preprocess the task and heterogeneous robot data to obtain problem input to build a multi-objective multi-robot task allocation model;

[0062] A collaborative ant colony algorithm is designed. Three ant colonies are set in the ant colony algorithm. The three ant colonies are set to have three optimization objectives: minimizing the maximum completion time of all tasks (i.e., minimizing the completion time of the last completed task), minimizing the average movement time of all robots (i.e., minimizing the average total movement time of all robots between task points), and minimizing the average waiting time of all tasks (i.e., minimizing the average start time of all tasks). The three optimization objectives are used to solve the multi-objective multi-robot task allocation model. The termination condition of the collaborative ant colony algorithm is set. The constructed collaborative ant colony algorithm is executed iteratively for multiple times. In each iteration, the ant individuals in the three ant colonies respectively construct paths including task-task execution sequence paths and task-robot allocation paths according to the pheromones and heuristic information of the ant colony they are in. The feasible solutions are searched from different directions with the three optimization objectives to obtain new feasible solutions for the ant individuals.

[0063] Evaluate the fitness of the new feasible solutions of all ant individuals, select the high-quality solutions on each optimization objective for local search, select the optimal solutions on the three optimization objectives, and perform global updates of the pheromones of the three collaborative ant colonies respectively; when the termination condition of the collaborative ant colony algorithm is met, the optimization process ends and outputs the optimal solution set of the multi-objective multi-robot task allocation model.

[0064] like Figure 2 As shown, the specific algorithm flow of the embodiment of the present invention includes the following steps:

[0065] S1. Preprocess the m tasks and n heterogeneous robot data as the problem input to obtain task t i Attributes (TID i ,TP i ,TD i, TC i ,TR i ,PC i ) and robot r j Attributes (RID j ,V j ,RC j ,RR j ), where TID i =i, i∈[1,m] and RID j =j, j∈[1,n] represent tasks ti and robot r j Number, TP i =(tx i ,ty i ) represents task t i The geographical coordinates of the mission point, TD i Represents task t i The maximum execution time of Indicates the execution of task t i The different abilities required, Indicates the execution of task t i H different resources are needed, PC i Indicates that when executing task t i The task that needs to be completed before (i.e. task t i The set of precedent tasks) is V j Represents robot r j The moving speed, Represents robot r j L different abilities, Represents robot r j H different resources;

[0066] S2. Design collaborative ant colony algorithm:

[0067] S2-1. Initialize the collaborative ant colony algorithm, including initializing pheromone and static heuristic information, and initializing the Archive Pareto set: set three ant colonies in the ant colony algorithm, and set the three ant colonies to minimize the maximum completion time of all tasks (i.e., minimize the completion time of the last completed task), minimize the average moving time of all robots (i.e., minimize the average total moving time of all robots between task points), and minimize the average waiting time of all tasks (i.e., minimize the average start time of all tasks) as optimization goals;

[0068] S2-1. Construct the solution space. Based on the random starting point, each ant colony uses two pheromone matrices τ accumulated between task-task and task-robot. 1,c and τ 2,c Construct a feasible solution; first, a random strategy is used to construct a feasible solution, and then the pheromone matrix τ is optimized according to the three optimization target values ​​of the feasible solution. 1,c and τ 2,c Initialize, c∈{1,2,3} represents three ant colonies; at the same time, the static heuristic information η between the common tasks of the three ant colonies in the ant colony algorithm 1Initialization: Initialize the set Archive that stores all Pareto solutions to an empty set, and use three ant colonies to simultaneously solve the multi-objective multi-robot task allocation model from the three optimization goals of minimizing the maximum completion time of all tasks (i.e., minimizing the completion time of the last completed task), minimizing the average movement time of all robots (i.e., minimizing the average total movement time of all robots between task points), and minimizing the average waiting time of all tasks (i.e., minimizing the average start time of all tasks). The waiting time of a task is the time when the task starts to execute;

[0069] S3, set the ant colony index to start accumulating from c=1 until all ant colonies update pheromones, and set the termination condition of the collaborative ant colony algorithm (such as the maximum number of iterations): execute the constructed collaborative ant colony algorithm iteratively for multiple times, and in each iteration, the ant individuals in the three ant colonies respectively construct the task-robot path according to the pheromones and heuristic information of their ant colonies, and locally search for feasible solutions from three optimization target directions (i.e., minimizing the maximum completion time of all tasks, minimizing the average moving time of all robots, and minimizing the average waiting time of all tasks); the step of constructing the task-robot path further includes: the ant individuals first construct the task-robot path according to the static heuristic information η 1 and the pheromone τ between task-task paths 1,c Select the next task to be performed and then calculate the dynamic heuristic information η between the task and the robot path 2 And combine the task with the pheromone τ between robots 2,c Build a robot alliance to perform the task, and at the same time perform local updates of pheromones on each task-task and task-robot path of the feasible solution constructed by each individual ant to construct a new feasible solution for the individual ant;

[0070] S4, evaluate the fitness values ​​of the new feasible solutions constructed by all ant individuals (i.e., the objective function values ​​of the feasible solutions on the three optimization objectives), add all these feasible solutions to the set Archive, and eliminate all dominated solutions in the set Archive; then, select the high-quality solutions on each objective from the set Archive for local search;

[0071] S5, select the optimal solutions for the three objectives from the set Archive, and globally update the pheromones of the three ant colonies respectively;

[0072] S6. Determine whether the termination condition of the algorithm is met. If the termination condition is met, the optimization process ends and the set Archive is output; otherwise, return to step S3 to continue iteration.

[0073] Furthermore, the multi-objective multi-robot task allocation model in step S1 allows multiple robots to form an alliance to collaboratively perform a task, and only when the task t i All predecessor tasks of have been completed, and task t i When the required capabilities and resources are available, the task t i can be executed.

[0074] Furthermore, in the steps S2-S3, each ant colony has a pheromone matrix τ accumulated between tasks. 1,c and the pheromone matrix τ accumulated between the task and the robot 2,c , the two pheromone matrices are in the following form:

[0075]

[0076] Among them, m represents the number of tasks, n represents the number of robots, and c∈{1,2,3} represents the indexes of the three ant colonies. (0≤i,k≤m) represents the ant colony c in task t i With the task k When i = 0 or k = 0, task t i or task t k represents the starting point of all robots; when i = k, task t i With the task k is the same task; when i≠k, task t i With the task k For two different tasks, Indicates that ant colony c is in task t i With robot j The pheromone value between them, i and j represent the index of the task and robot respectively. When i=0 or j=0, 0≤i≤m,0≤j≤n;

[0077] Pheromone Matrix τ 1,c and τ 2,c The initial value τ 0,c The calculation formula is as follows:

[0078]

[0079] Among them, X r represents the initial feasible solution constructed using a random strategy, f c (X r ) indicates X r The target value on target c.

[0080] Furthermore, in the steps S2-S3, each ant colony uses the static heuristic information η between tasks.1 and the dynamic heuristic information η between the task and the robot 2,c Construct feasible solutions; (0≤i,k≤m) represents task t i and task t k The static heuristic information between, when i = 0 or k = 0, task t i or task t k represents the starting point of all robots; when i = k, task t i With the task k For the same task, When i≠k, task t i With the task k For two different tasks, The calculation formula is as follows:

[0081]

[0082] Among them, dist(TP i ,TP k ) represents task t i and task t k The Euclidean distance between the task points;

[0083] Task i and robot r j , 1≤j≤n represents the dynamic heuristic information between robot index and The calculation formula is as follows:

[0084]

[0085] Where Δf c If you select robot r j Execute Task i , the target value increment of the solution obtained on the target c, FT i Represents task t i The completion time, T f Indicates the set of tasks that have been assigned, AR i Indicates that the allocated space is used to execute task t i A collection of robots, Indicates the last assigned to robot r g The mission point location, WT i Represents task t i Waiting time

[0086] Furthermore, in step S3, each ant individual e of the ant colony c constructs a feasible solution according to the following steps:

[0087] S301, the feasible solution Xc,e RT with each robot's task execution sequence j Initialized to empty, c∈{1,2,3} represents the index of three ant colonies, and also represents the index of three optimization targets, e represents the index of individual ants, and 1≤j≤n represents the index of robots;

[0088] S302: Select currently executable tasks from all unassigned tasks and add them to the set T p , where the tasks that can be executed are those whose predecessor tasks have been assigned, that is, the predecessor tasks have been assigned to the corresponding robot alliance;

[0089] S303, based on pheromone matrix τ 1,c and static heuristic information η 1 From the set T p Select the next task to be assigned. The specific selection process is as follows:

[0090]

[0091] Where 0≤i,k≤m represents the task index, w represents the index of the last assigned task, parameters α=1 and β=2 are used to balance the importance between pheromone and heuristic information, q is a random number uniformly distributed in the range [0,1], and parameter q0=0.9 represents the probability of the ant colony to develop; if q does not exceed q0, the ant colony with the largest The value of the task t k , otherwise the roulette wheel selection is performed according to the following probabilities:

[0092]

[0093] in, Represents the selection task t k As a task w The probability of the next assigned task after ;

[0094] S304, build execution task t i The Robot Alliance, which includes the robots in the AR i Initialized to empty;

[0095] S305, all the tasks not assigned to i The robot is added to the set R'; the dynamic heuristic information η of each robot in the set R' is calculated according to formulas (5) and (6): 2,c ; Binding pheromone τ 2,c Select execution task t from the set R' i Robot j , similar to step S303, the specific selection process is as follows:

[0096]

[0097] Among them, 0≤i≤m represents the task index, 0≤i≤m, 1≤j,y,g≤n represents the robot index, Indicates that the task t i Assign to robot y The probability of selecting the robot r j Join Collection AR i middle;

[0098] S306, repeat step S305 until the constructed robot alliance can meet the task t i capacity and resource constraints;

[0099] S307, construct a binary sequence of length n, where each bit of the sequence is 0, and then i , sort each robot in descending order according to its number. If robot r j ∈AR i , then set the jth position of the binary sequence to 1; finally, convert the binary sequence into a decimal number as the execution task t i The Robot Alliance ID AID z ;

[0100] S308, according to the selected task t i and Robot Alliance z =(AID z ,AR i ), update the feasible solution X c,e RT with each robot's task execution sequence j ;

[0101] S309, repeat steps S302-S308 until all tasks are assigned and a feasible solution X is output. c,e .

[0102] Furthermore, the pheromone matrix τ in step S3 1,c and τ 2,c The local update process is as follows:

[0103]

[0104] Where s∈{1,2}, when s=1, 0≤i,j≤m both represent the task index (i=0 and j=0 both represent the index of the starting point), Represents task t i With the task jWhen s=2, 0≤i≤m represents the task index, and 0≤j≤n represents the robot index (i=0 and j=0 have no practical meaning at this time, and are only for the convenience of calculation). Represents task t i and robot r j The pheromone between l =0.1 represents the evaporation rate of local renewal of pheromone, X c,e represents a feasible solution constructed by ant e in ant colony c.

[0105] Furthermore, in step S4, the fitness of the new feasible solution constructed by all ant individuals is evaluated. The fitness of the feasible solution X includes three optimization target values: the maximum completion time of all tasks, the average moving time of all robots, and the average waiting time of all tasks. The specific calculation process is as follows:

[0106] S401, the calculation process of the maximum completion time f1(X) of all tasks is as follows:

[0107]

[0108] FT i =WT i +ET i (13)

[0109] Where T represents a set of m tasks, 0≤i≤m represents the task index, and when i=0, t i Indicates the starting point of all robots, FT i Represents task t i Completion time, WT i Represents task t i Waiting time, ET i Represents task t i The actual execution time, when i = 0, FT i =WT i =ET i =0;

[0110] WT i The calculation method is as follows:

[0111]

[0112] Among them, 1≤j≤n represents the robot index, 0≤k≤m represents the task index, which is defined the same as i, AR i Indicates the execution of task t i The Robot League contains a collection of robots, Represents robot r j The completion time of the previous task, Represents robot rj The task point of the previous task; WT i The calculation method means that only when all robots in the alliance reach the task point and the task t i Task t can only be started when all predecessor tasks have been completed. i ET i The calculation method is as follows:

[0113]

[0114] in, Indicates the execution of task t i The required first ability value, Represents robot r j The first ability value, ET i The calculation method indicates that the time it takes for a robot alliance to perform a task is determined by its relatively weakest ability;

[0115] S402, the calculation process of the average moving time f2(X) of all robots is as follows:

[0116]

[0117] Where R represents a set of n heterogeneous robots, 1≤j≤n represents the robot index, and TT j Represents robot r j The total moving time is calculated as follows:

[0118]

[0119] Among them, RP0 represents the common starting position of all robots, Represents robot r j The task point location of the hth task to be executed, |RT j | indicates robot r j Total number of tasks executed;

[0120] S403, the calculation process of the average waiting time f3(X) of all tasks is as follows:

[0121]

[0122] Furthermore, the local search process in step S4 is as follows:

[0123] S411. Select the optimal solution for each target from the Archive collection Add the candidate solution set L;

[0124] S412. Perform 0-1 standardization on all solutions in Archive for each objective respectively; then calculate the average Manhattan distance between each solution in Archive and other solutions for each objective; sort the solutions in Archive in descending order of Manhattan distance for each objective; select the top K solutions for each objective as the sparse solutions (solutions that are far from other solutions in the search space) and add them to the candidate solution set L.

[0125] S413. Randomly select two non-repeating solutions X and Y from the candidate solution set L.

[0126] S414. Generate a random number r1 that follows a uniform distribution within the range [0, 1]. If r1 < R cro , then perform a crossover operation on X and Y to obtain X1 and Y1. Otherwise, (X1, Y1) = (X, Y).

[0127] S415. Generate a random number r2 that follows a uniform distribution within the range [0, 1]. If r2 < R mut , then perform a mutation operation on X1 and Y1 to obtain X2 and Y2. Otherwise, (X2, Y2) = (X1, Y1).

[0128] S416. Evaluate the fitness values of solutions X2 and Y2, and add X2 and Y2 to the set Archive, and eliminate all dominated solutions in the set Archive.

[0129] S417. Repeat steps S404 - S406 N times, where N is the number of ant individuals in a single ant colony.

[0130] In this method, a feasible solution X is encoded as a (TID, AID) sequence sorted in descending order of task execution priority; thus, as Figure 3 shown, the specific process of performing a crossover operation on feasible solutions X and Y is as follows:

[0131] S421. Clipping: Randomly select two crossover positions pos1 and pos2 (1 < pos1 < pos2 < m); clip solutions X and Y according to pos1 and pos2 to obtain solution segments x 1, x2, x3 and y1, y2, y3.

[0132] S422. Splicing: Splice x1, y2, x3 in sequence to obtain solution X1.

[0133] S423. Remove duplicate tasks: Remove the task number - robot alliance number pairs in solution segments x1 and x3 that are repeated with the tasks in solution segment y2.

[0134] S424, filling in missing tasks: inserting the task number-robot alliance number pairs corresponding to the tasks that appear in solution fragment x2 but do not appear in solution fragment y2 into solution fragment y2, and the order of insertion follows the priority order of the tasks in the feasible solution Y;

[0135] S425, repair the violated pre-order constraints: use the bubble sort method to check whether the pre-order constraints of each task in solution x1 are satisfied. If not, swap the position of the task with the pre-order task. Repeat this step until the pre-order constraints of all tasks in solution x1 are satisfied.

[0136] S426, repeat steps S12-S15 for the solution fragment y1, x2, y3 to obtain the solution Y1;

[0137] S427, output solutions X1 and Y1;

[0138] In addition, the specific process of performing mutation operation on the feasible solution X1 is as follows:

[0139] S431, let X2 = X1, generate a random number r in the range [0,1] that obeys a uniform distribution for each robot alliance that has a feasible solution to X2, and if r<0.5, perform a mutation operation on it;

[0140] S432, the robot alliance a z AID in decimal z Convert to a binary sequence B of length n z , sequence B z If the jth bit is 1, it means that robot r j In this alliance, if it is 0, it means that the robot r j Not in that alliance;

[0141] S433, generate a random index u (1≤u≤n), and convert the binary alliance sequence B z The u-th position of is reversed (i.e. 1 is set to 0, 0 is set to 1), and then converted to a decimal alliance number AID z ', get the new alliance a z ';

[0142] S434, determine the new robot alliance a z 'Whether the task t is satisfied i If the capacity and resource requirements are met, the corresponding alliance in solution X2 is updated, otherwise it returns to step S423 and repeats at most n times;

[0143] S435, output the new feasible solution X2 after mutation;

[0144] It should be noted that the parameters of the local search method in step S4 are set as: K = N, R cro=1.0, R mut =0.1, where N=20 is the number of ants in a single ant colony.

[0145] Furthermore, in step S5, after all ant individuals in all ant colonies have constructed a new feasible solution and the optimal solution on each optimization target has been locally searched, the pheromone matrix of all ant colonies is globally updated. The optimal solutions on three targets are selected from the set Archive, and the pheromones of the three ant colonies are globally updated respectively. The pheromone matrix τ 1,c and τ 2,c The global update expression is as follows:

[0146]

[0147] Among them, c∈{1,2,3} represents the index of three ant colonies and also represents the index of three optimization targets. s∈{1,2}, when s=1, 0≤i,j≤m both represent the task index, i=0 and j=0 both represent the index of the starting point. Represents task t i With the task j When s = 2, 0≤i≤m represents the task index, and 0≤j≤n represents the robot index. At this time, i = 0 and j = 0 have no practical meaning and are only for the convenience of calculation. Represents task t i and robot r j The pheromone between g = 0.1 indicates the evaporation rate of the global update of pheromone, represents the optimal feasible solution on target c, Δτ c represents the pheromone increment on target c (i.e., ant colony c), Δτ c The calculation formula is as follows:

[0148]

[0149] in, represents the optimal feasible solution on target c, represents the optimal feasible solution on target c The target value on target c.

[0150] Compared with the prior art, the present invention has the following advantages and effects:

[0151] (1) This paper proposes a new three-objective multi-robot task allocation problem model. Combined with real application scenarios, under the condition of heterogeneous multi-robot collaboration, it considers the order constraints between tasks and the resource and capacity constraints between tasks and robot alliances, and optimizes the maximum completion time of all tasks, the average moving time of all robots, and the average waiting time of all tasks. The setting of multiple objectives is more practical and helps decision makers balance various objectives and make plans.

[0152] (2) The present invention proposes an efficient multi-objective intelligent computing method, which combines the ant colony algorithm with a multi-swarm multi-objective framework. An efficient multi-objective collaborative ant colony algorithm is designed by combining the ant colony algorithm with a multi-swarm multi-objective framework. Dynamic heuristic information and dual pheromones are used to construct feasible solutions, which can effectively narrow the search space and utilize historical experience.

[0153] (3) This paper proposes a new fusion-based local search method that fully integrates the high-quality solution information of different target populations and further comprehensively explores the Pareto frontier. This method performs stably on multi-robot task allocation problems of different scales and can obtain a series of high-performance Pareto solutions that meet multiple complex constraints in a relatively short time.

[0154] The embodiment is described as follows: In the application scenario of heterogeneous multi-robot system, multiple robots can often collaborate to perform the same task together. The group of multiple robots performing the same task is called a robot alliance. Different robot alliances have different costs (execution time) when performing the same task. There are often certain pre-order constraints between different tasks, such as Figure 3 As shown, there is a precedence constraint between task t6 and task t7. Task t6 needs to be completed before executing task t7, and task t6 is the precedent task of task t7. For example, in a logistics distribution scenario, the robot needs to perform the pickup task before it can perform the delivery task. In addition, there are many different resource and capability constraints between tasks and the robot alliance, that is, the execution of tasks may require a variety of different resources and capabilities, and the robot alliance needs to provide these. Under the condition of satisfying the above constraints, the present invention simultaneously optimizes the three optimization objectives of the maximum completion time of all tasks, the average moving time of all robots, and the average waiting time of all tasks. The specific expressions of the three optimization objectives are as follows:

[0155]

[0156] FT i =WT i +ET i (twenty four)

[0157]

[0158]

[0159] Where m represents the number of tasks, n represents the number of robots, and 0≤i,k≤m represents the task index (when i=0 or k=0, task t i or task t k represents the starting point of all robots), 1≤j≤n represents the robot index, f1(X) represents the maximum completion time of all tasks, f2(X) represents the average moving time of all robots, f3(X) represents the average waiting time of all tasks, FT i Represents task t i Completion time, WT i Represents task t i Waiting time, ET i Represents task t i The actual execution time, when i = 0, FT i =WT i =ET i =0,TT j Represents robot r j The total moving time, T represents a set of m tasks, R represents a set of n heterogeneous robots, AR i Indicates the execution of task t i The Robot League contains a collection of robots, Represents robot r j The completion time of the previous task, dist(TP i ,TP j ) represents task t i and task t j The Euclidean distance between the task points, Represents robot r j The task point of the previous task, TP i Represents task t i The mission point location, V j Represents robot r j The moving speed, L represents the number of different capabilities required by the task and that the robot can provide, Indicates the execution of task t i The required first ability value, Represents robot r j The first capability value, RP0, represents the common starting point of all robots. Represents robot r j The task point location of the hth task to be executed, |RT j | indicates robot r j The total number of tasks executed.

[0160] Combination Figure 4 The feasible solution shown here describes the specific implementation of the entire algorithm step by step:

[0161] T1. Preprocess the data of the m tasks and n heterogeneous robots input into the problem to obtain task t i Attributes (TID i ,TP i ,TD i ,TC i ,TR i ,PC i ) and robot r j Attributes (RID j ,V j ,RC j ,RR j ), where TID i =i(i∈[1,m]) and RID j =j(j∈[1,n]) respectively represent tasks t i and robot r j Number, TP i =(tx i ,ty i ) represents task t i The geographical coordinates of the mission point, TD i Represents task t i The maximum execution time of Indicates the execution of task t i The different abilities required, Indicates the execution of task t i H different resources are needed, PC i Indicates that when executing task t i The task that needs to be completed before (i.e. task t i The set of predecessor tasks of task t2 is PC2 = {t7, t6}, and the set of predecessor tasks of task t8 is PC8 = {t3, t4, t 10 ,t5,t9};V j Represents robot r j The moving speed, Represents robot r j L different capabilities can be provided, Represents robot r j H different resources can be provided.

[0162] T2, use three ant colonies to optimize the three objectives simultaneously, and each ant colony uses the two pheromone matrices accumulated between tasks and tasks and between tasks and robots to construct a feasible solution; therefore, first a random strategy is used to construct a feasible solution, and then the pheromone matrix τ is respectively adjusted according to the three objective values ​​of the solution. 1,c and τ 2,c (c∈{1,2,3}, representing 3 ant colonies) is initialized, and the pheromone matrix form and initial value are set as follows:

[0163]

[0164] in, (0≤i,k≤m) represents the ant colony c in task t i With the task k (When i=0 or k=0, task t i or task t k represents the starting point of all robots; when i = k, task t i With the task k is the same task; when i≠k, task t i With the task k for two different tasks), (0≤i≤m,0≤j≤n) represents the ant colony c in task t i With robot j , i and j represent the index of the task and robot respectively. For the convenience of calculation, i=0 and j=0 have no practical meaning. When i=0 or j=0, X r is the initial solution constructed using a random strategy, τ 0,c Represents the pheromone matrix and The initial value of the elements in, f c (X) represents the target value of solution X on target c.

[0165] T3, the three ant colonies in the algorithm use the same static heuristic information η when selecting tasks 1 , for η 1 Initialize, 0≤i,k≤m represents task t i and task t k The static heuristic information between, when i = 0 or k = 0, task t i or task t k represents the starting point of all robots; when i = k, task t i With the task k For the same task, When i≠k, task t i With the task kFor two different tasks, The calculation formula is as follows:

[0166]

[0167] Among them, dist(TP i ,TP k ) represents task t i and task t k The Euclidean distance between the task points; at the same time, the set Archive storing all Pareto solutions is initialized to an empty set.

[0168] T4. Construct the task-robot path through the ant colony algorithm and search for feasible solutions from three target directions. At each step of constructing the task-robot path, the ants first use the static heuristic information η 1 and the pheromone τ between tasks 1,c Select the next task to be performed. The specific selection process is as follows:

[0169]

[0170] Where 0≤i,k≤m represents the task index, w represents the index of the last assigned task, parameters α=1 and β=2 are used to balance the importance between pheromone and heuristic information, q is a random number uniformly distributed in the range [0,1], and parameter q0=0.9 represents the probability of the ant colony to develop; if q does not exceed q0, the ant colony with the largest The value of the task t k , otherwise the roulette wheel selection is performed according to the following probabilities:

[0171]

[0172] in, Represents the selection task t k As a task w The probability of the next assigned task after ;

[0173] Select the next task to be executed i After that, build and execute task t i First, the dynamic heuristic information η between the task and the robot is calculated. 2 , the specific calculation method is as follows:

[0174]

[0175] Where Δf c If you select robot r j Execute Task i , the target value increment of the solution obtained on the target c, FTi Represents task t i The completion time, T f Indicates the set of tasks that have been assigned, AR i Indicates that the allocated space is used to execute task t i A collection of robots, Indicates the last assigned to robot r g The mission point location, WT i Represents task t i The waiting time; then, the dynamic heuristic information η 2 Pheromone τ between tasks and robots 2,c Combined, gradually select available robots to join the current robot alliance until the alliance meets the task t i The execution constraints of , the specific selection process is as follows:

[0176]

[0177] Among them, 0≤i≤m represents the task index, 1≤j,y,g≤n represents the robot index, Indicates that the task t i Assign to robot y The probability of selecting the robot r j Join Collection AR i middle.

[0178] While each ant is building a solution, it locally updates the pheromone matrix τ on each task-task and task-robot path. 1,c and τ 2,c The local update process is as follows:

[0179]

[0180] Where s∈{1,2}, when s=1, 0≤i,j≤m both represent the task index (i=0 and j=0 both represent the index of the starting point), Represents task t i With the task j When s=2, 0≤i≤m represents the task index, and 0≤j≤n represents the robot index (i=0 and j=0 have no practical meaning at this time, and are only for the convenience of calculation). Represents task t i and robot r j The pheromone between l =0.1 represents the evaporation rate of local renewal of pheromone, X c,erepresents a feasible solution constructed by ant e in ant colony c. By locally updating the pheromone and reducing the pheromone on the path, the following ants may choose different tasks or build different robot alliances for the tasks, thereby increasing the diversity of solutions.

[0181] T5. Evaluate the fitness of all new solutions constructed by ants, add all these solutions to the Archive set, and eliminate all dominated solutions in the Archive set.

[0182] T6. Select high-quality solutions for each target from the set Archive for local search. High-quality solutions for each target include the optimal solution and the first K sparse solutions (i.e., solutions that are far away from other solutions). Therefore, the optimal solution for each target is first added to the candidate solution set L. Then, all solutions in the set Archive are normalized to 0-1 on each target, and the average Manhattan distance between each solution and other solutions is calculated. The solutions are arranged in descending order, and finally, the first K solutions for each target are added to the set L. After that, two non-repeated solutions X and Y are selected from the set L for local search. First, a random number r1 in the range [0,1] that follows a uniform distribution is generated. If r1 <R cro , then perform a crossover operation on X and Y to obtain X1 and Y1, otherwise, (X1, Y1) = (X, Y); then generate a random number r2 that follows a uniform distribution in the range [0, 1]. If r2 <R mut , then perform mutation operation on X1 and Y1 to obtain X2 and Y2, otherwise, (X2, Y2) = (X1, Y1); finally, evaluate the fitness of solutions X2 and Y2, and add X2 and Y2 to the set Archive, eliminating all dominated solutions in the set Archive; repeat the above process N times, where N is the number of ants in a single ant colony.

[0183] In this method, each feasible solution X c,e The code is divided into two parts: the task number sequence ID obtained by sorting the task execution priorities t and the corresponding robot alliance serial number ID a Based on this solution representation, the crossover operation of local search for feasible solutions X and Y is as follows: Figure 3 First, two random crossover positions are generated, 3 and 7 respectively. According to the crossover positions, X and Y are cut into x1, x2, x3 and y1, y2, y3 respectively. Then, x1, y2, x3 are concatenated into a new solution X1, and y1, x2, y3 are concatenated into a new solution Y1. At this time, there may be duplicate tasks or tasks that violate the pre-order constraint in X1 and Y1. Taking solution X1 as an example, the tasks and their alliances that are duplicated with y2 in the solution fragments x1 and x3 are removed, as shown in Figure 3As shown in (3), the repeated tasks in x1 and x3 are t6 and t9, and the corresponding task number-alliance number pairs are (6,6) and (9,5). Then, the remaining missing tasks are filled by inserting the task number-robot alliance number pairs corresponding to the tasks that appear in the solution fragment x2 but not in y2 into the solution fragment y2. The order of insertion follows the priority order of the tasks in the feasible solution Y, as follows: Figure 3 As shown in (4), at this time, tasks t3 and t2 in x2 have not appeared in y2. In the feasible solution Y, the priority of task t3 is higher than that of task t5, and the priority of task t2 is lower than that of task t9. Therefore, the task number-robot alliance number pairs (3,3) and (2,7) in the solution fragment x2 are inserted into the head and tail of the solution fragment y2 respectively to obtain a new solution X1. At this time, the solution X1 already contains all the tasks, but the precedence constraints between tasks may be violated. Therefore, check whether the precedence tasks of each task in X1 have been completed in order from high to low priority. If not, swap the positions of the task number-robot alliance number pairs corresponding to the two. Figure 3 As shown in (5), the predecessor of task t7 is task t6, and the execution priority of task t6 is lower than that of task t7. (7,3) and (6,2) are swapped. Similarly, the predecessor of task t9 is t 10 and t4, and the execution priority of task t4 is lower than that of task t9, so (9,7) and (4,3) are swapped. Repeat the above repair process until the pre-order constraints of all tasks are satisfied. At this time, solution X1 is a new feasible solution that integrates the information of solution X and solution Y, and it is output.

[0184] In the process of local search, the new feasible solution obtained has a certain probability R mut Mutation occurs. Taking the feasible solution X1 as an example, let the new solution X2 = X1. Then, for each robot alliance in X2, generate a random number r that follows a uniform distribution in the range [0,1]. If r < 0.5, perform a mutation operation on it: First, replace the robot alliance a z AID in decimal z Convert to a binary sequence B of length n z , sequence B z If the jth bit is 1, it means that robot r j In this alliance, if it is 0, it means that the robot r j Not in this alliance. Figure 4 As shown, the execution task t 10 The robot alliance is a3, and the binary sequence B3=011 is obtained by binary conversion, and the corresponding robot set is {r2,r1}. Then, a random index u (1≤u≤n) is generated and the binary alliance sequence B zThe u-th position of is reversed (i.e. 1 is set to 0, 0 is set to 1), and then converted to a decimal alliance number AID z ', get the new alliance a z '. Assuming the random index u = 1, the binary sequence B obtained after inversion is z '=010, the corresponding robot set is {r2}, and the corresponding decimal alliance number AID z '=2. Determine whether the new robot alliance a2 can meet the task t 10 If the resource and capability requirements are met, the corresponding alliance in solution X2 will be updated, otherwise the mutation will be repeated up to n times. After performing the above mutation operation on all robot alliances, a new feasible solution X2 is finally obtained;

[0185] T7. After all ant individuals in all ant colonies have constructed a new feasible solution and the optimal solution of each optimization objective has been locally searched, the pheromone matrix of all ant colonies is globally updated. The optimal solutions of the three objectives are selected from the set Archive, and the pheromones of the three ant colonies are globally updated respectively. The pheromone matrix τ 1,c and τ 2,c The global update expression is as follows:

[0186]

[0187] Among them, c∈{1,2,3} represents the index of three ant colonies and also represents the index of three optimization targets. s∈{1,2}, when s=1, 0≤i,j≤m both represent the task index (i=0 and j=0 both represent the index of the starting point). Represents task t i With the task j When s=2, 0≤i≤m represents the task index, and 0≤j≤n represents the robot index (i=0 and j=0 have no practical meaning at this time, and are only for the convenience of calculation). Represents task t i and robot r j The pheromone between g = 0.1 indicates the evaporation rate of the global update of pheromone, represents the optimal feasible solution on target c, Δτ c represents the pheromone increment on target c (i.e., ant colony c), Δτ c The calculation is as follows:

[0188]

[0189] in, represents the optimal feasible solution on target c, represents the optimal feasible solution on target c The target value on target c.

[0190] T8. Determine whether the termination condition of the algorithm is met. If the termination condition is met, the optimization process ends and the set Archive is output; otherwise, return to step T4 to continue iteration.

[0191] In summary, combined with real-world application scenarios, the present invention proposes a multi-objective multi-robot task allocation model, which simultaneously considers multiple complex constraints such as pre-order constraints between tasks under the conditions of heterogeneous multi-robot collaboration. In order to solve the model, the present invention combines the ant colony algorithm with a multi-population multi-objective framework to design an efficient multi-objective collaborative ant colony algorithm, which uses dynamic heuristic information and dual pheromones to construct feasible solutions, and can effectively narrow the search space and utilize historical experience. At the same time, the present invention proposes a new fusion-based local search method, which fully integrates high-quality solution information from different target populations and further comprehensively explores the Pareto frontier. Experiments have shown that compared with traditional multi-objective methods, the method disclosed in the present invention is stable in multi-robot task allocation problems of different scales, and can obtain a series of high-quality feasible solutions in a relatively short time.

[0192] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A multi-objective multi-robot task allocation method based on collaborative ant colonies, characterized in that: include: Preprocess the task and heterogeneous robot data to obtain problem input to build a multi-objective multi-robot task allocation model; A collaborative ant colony algorithm is designed. Three ant colonies are set in the ant colony algorithm. The three ant colonies are set with three optimization objectives, namely, minimizing the maximum completion time of all tasks, minimizing the average moving time of all robots, and minimizing the average waiting time of all tasks. The three optimization objectives are used to solve the multi-objective multi-robot task allocation model. The termination condition of the collaborative ant colony algorithm is set. The constructed collaborative ant colony algorithm is executed iteratively for multiple times. In each iteration, the ant individuals in the three ant colonies respectively construct paths including task-task execution sequence paths and task-robot allocation paths according to the pheromones and heuristic information of their ant colonies. The feasible solutions are searched from different directions with the three optimization objectives to obtain new feasible solutions for the ant individuals. The fitness values ​​of the new feasible solutions of all the ant individuals are evaluated. The high-quality solutions on each optimization objective are selected for local search. The optimal solutions on the three optimization objectives are selected, and the pheromones of the three collaborative ant colonies are globally updated. When the termination condition of the collaborative ant colony algorithm is met, the optimization process ends and the optimal solution set of the multi-objective multi-robot task allocation model is output.

2. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1 is characterized in that: The step of constructing a path based on the pheromones and heuristic information of the ant colony further includes: the individual ant first selects the next task to be performed based on the static heuristic information and the pheromones between tasks, then calculates the dynamic heuristic information between the task and the robot and combines the pheromones between the task and the robot to construct a robot alliance that performs the task; at the same time, local updates of pheromones are performed on each task-task and task-robot path of the feasible solution constructed by each ant.

3. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1 is characterized in that: The specific expressions of the three optimization objectives are as follows: FT i =WT i +ET i Where m is the number of tasks, n is the number of robots, i, k are task indexes, j is the robot index, f1(X) is the maximum completion time of all tasks, f2(X) is the average moving time of all robots, f3(X) is the average waiting time of all tasks, and FT i Represents task t i Completion time, WT i Represents task t i Waiting time, ET i Represents task t i The actual execution time, TT j Represents robot r j The total moving time, T represents a set of m tasks, R represents a set of n heterogeneous robots, AR i Indicates the execution of task t i The Robot League contains a collection of robots, Represents robot r j The completion time of the previous task, dist(TP i ,TP j ) represents task t i and task t j The Euclidean distance between the task points, Represents robot r j The task point of the previous task, TP i Represents task t i The mission point location, V j Represents robot r j The moving speed, L represents the number of different capabilities required by the task and that the robot can provide, Indicates the execution of task t i The required first ability value, Represents robot r j The first capability value, RP0, represents the common starting point of all robots. Represents robot r j The task point location of the hth task to be executed, |RT j | indicates robot r j The total number of tasks executed.

4. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1 is characterized in that: The construction of a path based on the pheromone and heuristic information of the ant colony further includes: the ant individual first constructs a path based on the static heuristic information η 1 and the pheromone τ between task-task paths 1,c Select the next task to be performed and then calculate the dynamic heuristic information η between the task and the robot path 2 And combine the task with the pheromone τ between robots 2,c A robot alliance is built to perform the task, and at the same time, local updates are performed on each task-task and task-robot path of the feasible solution constructed by each individual ant to construct a new feasible solution for the individual ant.

5. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1, characterized in that: The pheromone of the ant colony further includes the pheromone matrix τ accumulated between tasks by each ant colony. 1,c and the pheromone matrix τ accumulated between the task and the robot 2,c , the two pheromone matrices are in the following form: Among them, m represents the number of tasks, n represents the number of robots, and c represents the index of the three ant colonies. (0≤i,k≤m) represents the ant colony c in task t i With the task k The pheromone value between Indicates that ant colony c is in task t i With robot j pheromone values ​​between .

6. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1, characterized in that: The feasible solution is further obtained by each ant individual using the static heuristic information η between tasks 1 and the dynamic heuristic information η between the task and the robot 2,c Build; Represents task t i and task t k The static heuristic information between is calculated as follows: Among them, dist(TP i ,TP k ) represents task t i and task t k The Euclidean distance between the task points; Task i and robot r j Dynamic heuristic information between The calculation formula is as follows: Where Δf c If you select robot r j Execute Task i , the target value increment of the solution obtained on the target c, FT i Represents task t i The completion time, T f Indicates the set of tasks that have been assigned, AR i Indicates that the allocated space is used to execute task t i A collection of robots, Indicates the last assigned to robot r g The mission point location, WT i Represents task t i waiting time.

7. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1 or 5, characterized in that: The process of constructing a feasible solution further includes: (i) The feasible solution X c,e RT with each robot's task execution sequence j Initialized to empty; (ii) Select currently executable tasks from all unassigned tasks and add them to the set T p ; (iii) Based on the pheromone matrix τ 1,c and static heuristic information η 1 From the set T p Select the next task to be assigned; (iv) Construct execution task t i The Robot Alliance, which includes the robots in the AR i Initialized to empty; (v) All the tasks not assigned to task t i The robot is added to the set R'; the dynamic heuristic information η of each robot in the set R' is calculated 2,c ; Binding pheromone τ 2,c Select execution task t from the set R' i Robot j ; (vi) Repeat step (v) until the constructed robot alliance can meet the task t i capacity and resource constraints; (vii) According to the robot alliance AR i Sort each robot in descending order according to its number as the robot to perform task t i The Robot Alliance; (viii) According to the selected task t i and Robot Alliance z =(AID z ,AR i ), update the feasible solution X c,e RT with each robot's task execution sequence j ; (viiii) Repeat steps (ii)-(viii) until all tasks are assigned and output a feasible solution X c,e .

8. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1, characterized in that: A crossover operation is used to perform local searches between different feasible solutions and merge them into a new feasible solution. The crossover operation further includes pruning, splicing, removing duplicate tasks, filling in missing tasks, and repairing violated pre-order constraints.

9. The multi-objective multi-robot task allocation method based on collaborative ant colonies according to claim 1, characterized in that: The local search further includes: judging whether the new robot alliance can meet the resource and capability requirements of the task; if so, updating the corresponding new robot alliance in the feasible solution; otherwise, re-performing the mutation operation until all robot alliances are subjected to the above mutation operation, and finally obtaining a new feasible solution.

10. The method for allocating multi-objective and multi-robot tasks based on collaborative ant colonies according to claim 1, characterized in that: The step of globally updating the pheromone of the three cooperative ant colonies further includes the pheromone matrix τ 1,c and τ 2,c The global update expression is as follows: Among them, s∈{1,2}, c∈{1,2,3} represent the indexes of three ant colonies and also represent the indexes of three optimization objectives, ρ g = 0.1 indicates the evaporation rate of the global update of pheromone, represents the optimal feasible solution on target c, Δτ c represents the pheromone increment on ant colony c, and the calculation formula is as follows: in, represents the optimal feasible solution on target c The target value on target c.

Images