Intelligent storage multi-mobile-robot task allocation method based on improved genetic algorithm

Abstract

The invention discloses an intelligent storage multi-mobile robot task allocation method based on an improved genetic algorithm, and the method comprises the steps: allocating a plurality of tasks to a plurality of mobile robots in an environment for execution according to the position distribution, weight and other indexes of the plurality of tasks in intelligent storage, and enabling the movement distance of all robots to be shortest. In order to solve the intelligent storage multi-robot task allocation problem, the traditional genetic algorithm has the problems of inappropriate coding mode, low convergence speed, local optimal solution and the like. In order to solve the problems, selection, crossover and mutation operations of a traditional genetic algorithm are improved, and an improved genetic algorithm is provided. The algorithm can overcome the defects of a traditional genetic algorithm in solving the multi-robot task allocation problem, and the feasibility of the algorithm is verified through practical application.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehouse scheduling, and more particularly to a multi-robot task allocation method. Background Technology

[0002] Since the 20th century, with the rapid development of the world economy and technology, more and more countries have realized the importance of manufacturing to their economic development, and the development of high-end manufacturing has attracted attention from all countries. Traditional warehousing and logistics mainly adopts a "person-to-goods" model, i.e., manual handling. This method is not only inefficient and costly, but also prone to errors, making it difficult to complete tasks in a timely manner when the workload is large, and it also poses certain safety hazards. Therefore, a new handling model has emerged: "goods-to-person." This aims to reduce labor costs, improve logistics sorting efficiency, and achieve a high degree of automation in logistics warehousing.

[0003] Task allocation for multiple mobile robots is a crucial and priority issue in intelligent warehousing systems. It involves optimizing the system based on its current operational conditions and processes using appropriate algorithms to achieve the desired results. While the allocation problem is relatively simple for traditional single-robot systems, the demands of modern industrial development have become increasingly sophisticated, leading to the emergence of multi-robot systems. Current research on multi-robot task allocation focuses on achieving optimal combinations of tasks and robots, making it essentially a combinatorial optimization problem.

[0004] To address the shortcomings of traditional genetic algorithms in solving multi-robot task allocation problems, this invention proposes an intelligent warehousing multi-mobile robot task allocation method based on an improved genetic algorithm. Summary of the Invention

[0005] In view of the above background, this invention provides a task allocation method for intelligent warehouse multi-mobile robots based on an improved genetic algorithm. It aims to solve the problems of unsuitable encoding methods when using traditional genetic algorithms to solve task allocation problems, which lead to solutions that do not conform to the working conditions, slow convergence speed, poor output effect, and easy getting trapped in local optima. The method aims to generate a task allocation scheme with faster convergence speed and obtain a better solution with a shorter total movement distance.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for task allocation of multiple mobile robots in intelligent warehousing based on an improved genetic algorithm includes the following steps:

[0008] The mathematical model of the problem has the following four constraints.

[0009] (1) The starting point and target point of the mobile robot are constrained, and the path formed by the starting point and the end point of all robots is a closed loop.

[0010] FR K =ON LR K =ON

[0011] In the above formula, FR k —The first task performed by robot k, LR k —The last task performed by robot k, ON—the position of the starting point and the target point.

[0012] (2) Robot and task matching constraints: each robot can only execute a set of tasks assigned by the system at a time, and each task can only be executed by one robot once.

[0013]

[0014]

[0015]

[0016]

[0017]

[0018] In the above formula, i represents the task point T. i Number, j — Task point T j Number, k — Robot R k serial number.

[0019] (3) Robot load constraint: The total weight of all tasks in the task group assigned to each robot must be less than the robot's maximum load capacity. All robots in the environment are of the same model and have the same rated load of W.

[0020]

[0021] w i —The weight of task i, W —The rated load weight of all robots.

[0022] (4) Robot number constraint: The number of robots working in the environment must be less than or equal to the total number of robots in the warehouse environment.

[0023] n now ≤n

[0024] n now — The number of robots in operation in the environment at any given time. n — The total number of robots.

[0025] (1) Obtain the number of tasks and the number of robots and encode them. Traditional encoding methods are not suitable for the specific problem in this paper because they are prone to producing solutions that do not meet the problem constraints in subsequent genetic operations.

[0026] In the two-segment encoding method, the task portion and the breakpoint portion together form the genotype of a chromosome. The length of the task sequence portion corresponds to the number of tasks to be executed, and its specific genotype number represents the sequential number of the tasks that the robot needs to execute. The breakpoint portion is the splitting point for each task group. The breakpoints divide the chromosome according to the position of the task genotypes on the chromosome corresponding to their numbers. Assuming there are m robots in total, the length of the breakpoint portion is n-1, and the total length of the chromosome is m+n-1. The two-segment encoding concatenates all task numbers and uses the breakpoints as a gene trimming tool to trim the task into multiple task groups, which are then distributed to each robot, thereby achieving effective decoding of all task groups.

[0027] The total distance traveled by all robots is taken as the objective optimization function. The smaller the value of the objective optimization function, the higher the fitness of that individual should be. Therefore, this paper uses the reciprocal of the objective optimization function as the fitness function calculation formula for the genetic algorithm. The fitness formula is as follows:

[0028]

[0029] Where, d ij Let x be the distance cost from task i to task j. ijk Define whether robot k moves from task i to j, and Fitness(i) represents the fitness of the i-th individual.

[0030] (2) The linear sorting roulette wheel selection method is combined with the elite retention strategy to screen the initial population and the non-elite population is used as a subpopulation to continue to participate in the evolution.

[0031] The traditional roulette wheel selection method, also known as proportional selection, proceeds as follows: ① Calculate the population fitness. Assuming there are n individuals in the population, and the fitness of the i-th individual is Fitness(i), then the probability of the i-th individual being selected is: ② Fitness value normalization. Normalize the fitness values ​​of all individuals so that their sum equals 1, satisfying the formula... ③ Construct a roulette wheel. Based on the fitness of individuals, map all individuals to a continuous interval of 0 to 1, similar to a roulette wheel. Individuals with higher fitness occupy a larger sector area in the roulette wheel. ④ Select a parent. Generate a random number from the interval [0,1], and select an individual from the roulette wheel as a parent based on this random number. ⑤ Repeat the selection. Repeat the above process until the number of selected individuals meets the requirement, usually the same number as the number of parents.

[0032] Elite retention is an optimization strategy in genetic algorithms that aims to preserve individuals with the highest fitness in the population, preventing them from being excluded by selection or crossover. Setting the elite retention rate to 0.1 means that the top 10% of individuals in terms of fitness are directly retained as elite individuals and added to the offspring population.

[0033] Suppose there are N individuals in the parent population. First, calculate the fitness of all individuals according to the fitness function and sort them. Then, retain the top 10% of individuals based on the elite retention rate and let them directly enter the offspring. The remaining individuals are selected according to the roulette wheel selection strategy to obtain M individuals. Finally, the individuals retained by the elite and the individuals selected by the roulette wheel are recombined to form the offspring.

[0034] (3) Crossover operation: A partial exchange of gene segments between two chromosomes produces a new individual. An adaptive method is used to adjust the crossover probability. The crossover probability formula is:

[0035]

[0036] p c — Crossover probability, p top / p down —The upper / lower bounds of the crossover probability, f max —Maximum fitness of the parent individual, f sub —Fitness of individuals that have not yet participated in the crossover, f w —The fitness threshold for elite retention.

[0037] When the fitness of the offspring chromosome is greater than the fitness threshold under the elite preservation strategy, the crossover probability is reduced to preserve the offspring chromosome as much as possible. When the fitness of the offspring chromosome is less than the threshold, the maximum crossover probability is used to eliminate the offspring as quickly as possible.

[0038] During the crossover process, due to the special nature of the multi-robot task allocation problem, two identical genotypes cannot appear on a single chromosome. When duplicate genotypes appear, it can lead to a situation where a single task is completed multiple times, but some tasks do not appear in the offspring chromosome. This is obviously not acceptable, so it is necessary to perform gene correction on the offspring chromosomes obtained through mutation.

[0039] To obtain effective offspring, the genotypes of the offspring chromosomes need to be corrected using the paternal genotypes. Taking the first offspring chromosome as an example, priority is given to duplicate genotypes in the offspring that did not participate in the crossing over. Genotypes 8, 6, 2, and 5 are corrected sequentially according to the order 1-4-7-3-9-10 in the paternal chromosome genotypes. If a genotype from the paternal crossing over segment still exists in the offspring crossing over segment, that genotype is not used for repair. The paternal genotype 4 replaces the first duplicated genotype 8 in the offspring; similarly, 3 replaces the second duplicated genotype 6, 9 replaces the duplicated genotype 2, and 10 replaces the duplicated genotype 5, thus obtaining offspring chromosomes that meet the actual requirements.

[0040] (4) Mutation operation: For each individual in the non-elite population, if the crossover condition is met, then the individual is mutated using dynamic mutation probability. The mutation formula is as follows:

[0041]

[0042] p m —Probability of mutation, p mtop / p mdown —The upper / lower bounds of the mutation probability, f max —Maximum fitness of the parent individual, f sub —The fitness of individuals that have not yet participated in the mutation, f w —The fitness threshold for elite retention.

[0043] For offspring chromosomes with a fitness level greater than the elite retention threshold, the mutation probability is reduced to ensure that the offspring is retained as much as possible; for offspring chromosomes with a fitness level less than the elite retention threshold, the maximum mutation probability is used for elimination.

[0044] After crossover, individuals undergo mutation by randomly generating mutation sites and then swapping the first and last positions.

[0045] (5) Merge the evolved subpopulations with elite individuals to form a new population;

[0046] (6) Determine whether the evolution termination condition is met. The algorithm termination condition is set as follows: the algorithm terminates when it reaches the set maximum number of iterations. If the condition is met, generate an allocation scheme; otherwise, return to step (2) until a task allocation scheme is generated.

[0047] The results show that both algorithms found the optimal solution that met the constraints within the maximum number of iterations. In the traditional genetic algorithm, the robot's total travel distance was 715.28m. After about 600 iterations, the algorithm got stuck in a local optimum. Due to the limitations of the traditional selection strategy, the population diversity was reduced. At the same time, the fixed crossover and mutation probabilities of the traditional genetic algorithm meant that many excellent individuals might be eliminated, which significantly slowed down the convergence speed of the algorithm. The iteration result stalled for about 800 iterations before a new result was obtained. Subsequently, after about 1400 iterations, the result tended to be 720m. Finally, after 1782 iterations, the result stabilized. The improved genetic algorithm robot initially exhibits an iterative process similar to the traditional genetic algorithm. However, after approximately 750 iterations, the result approaches 720m. This is because the improved selection and adaptive probability crossover mutation strategy ensures that superior individuals are preserved in the genetic operations, resulting in a more balanced convergence process with better solutions appearing throughout the process. There are no obvious local optima, and the iteration speed is significantly faster. Finally, after 1625 iterations, it reaches stability with a total travel distance of 704.73m, which is shorter than that of the traditional genetic algorithm. Attached Figure Description

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0049] Figure 1 This is a schematic diagram of the improved encoding method in this invention.

[0050] Figure 2 This is the improved selection process of the present invention.

[0051] Figure 3(a) shows the parent chromosomes before crossover in the crossover example of this invention.

[0052] Figure 3(b) shows the offspring chromosomes after crossing over in the crossover example of this invention.

[0053] Figure 4 This is a schematic diagram of the offspring chromosome correction process in this invention.

[0054] Figure 5 This is a schematic diagram of the mutation process in this invention.

[0055] Figure 6 This is a task distribution coordinate diagram in an example of the present invention.

[0056] Figure 7 This is the task information table in the example of the present invention.

[0057] Figure 8 These are two genetic algorithm parameter settings in examples of this invention.

[0058] Figure 9(a) shows the task allocation results before algorithm improvement in the example of this invention.

[0059] Figure 9(b) shows the task allocation result after algorithm improvement in the example of this invention.

[0060] Figure 10(a) is a schematic diagram of the task allocation result before algorithm improvement in the example of the present invention.

[0061] Figure 10(b) is a schematic diagram of the task allocation result after algorithm improvement in the example of the present invention.

[0062] Figure 11(a) is the algorithm iteration curve before algorithm improvement in the example of the present invention.

[0063] Figure 11(b) is an iterative curve of the algorithm after the algorithm improvement in the example of the present invention.

Claims

1. A method for task allocation of multiple mobile robots in intelligent warehousing based on an improved genetic algorithm, the specific steps of which include: Step 1: Input the m task point attributes and n mobile robots participating in task allocation and scheduling into the task allocation system. Step 2: The system calls the improved genetic algorithm to allocate tasks, using the attributes of each task point and the number of mobile robots as inputs to solve the task allocation model. Step 3: Output the task allocation scheme obtained by solving the algorithm through the system.

2. The intelligent warehousing multi-mobile robot task allocation method based on an improved genetic algorithm according to claim 1, characterized in that, Step 1: After the scheduling system receives the task instruction, it determines that there are n tasks to be moved. Each task contains information such as task coordinates, task weight, and task number. At the same time, there are m mobile robots in the warehouse. The quantity relationship between the two must satisfy the condition m < n. The system saves the data in the database and updates it in real time.

3. The intelligent warehousing multi-mobile robot task allocation method based on an improved genetic algorithm according to claim 1, characterized in that, The system call algorithm step described in step two involves packaging the improved algorithm on the MATLAB platform into a DLL file for the C# platform to call.

4. The intelligent warehousing multi-mobile robot task allocation method based on an improved genetic algorithm according to claim 1, characterized in that, In step two, the population coding method of the improved genetic algorithm is to adopt a two-stage coding method that is more suitable for the problem in this paper.

5. The intelligent warehousing multi-mobile robot task allocation method based on an improved genetic algorithm according to claim 1, characterized in that, Step two, which improves the genetic algorithm, requires setting initial parameters, including the maximum number of iterations, population size, elite retention rate, maximum / minimum crossover probability, and maximum / minimum mutation probability.

6. The method for task allocation of multiple mobile robots in intelligent warehousing based on an improved genetic algorithm according to claim 1, characterized in that, Step two involves improving the genetic algorithm, including encoding, initializing the population, selection, crossover, and mutation operations, repeating the evolutionary process until the iteration ends.

Images