Optimization Method for Task Scheduling of Multi-Capacity Automated Guided Vehicle System

Abstract

This invention relates to a task scheduling optimization method for a multi-capacity Automated Guided Vehicle (AGV) system, belonging to the field of intelligent manufacturing. First, when the AGV system meets the task allocation triggering conditions, the task information in the system is acquired. Second, the set of AGVs to be scheduled is determined, and the allocatable task set is determined according to the method for determining the allocatable task set. Third, the optimal task sequence is determined according to the artificial immune-gray wolf optimization algorithm. Finally, the optimal AGV allocation scheme is determined according to the task allocation rules. This method optimizes the task scheduling method of the AGV system through an intelligent optimization algorithm, solving the problems of excessive control information interaction and low flexibility between the scheduling system and the AGV in the prior art, thus improving the efficiency and flexibility of the scheduling system.

Description

Technical Field

[0001] This invention relates to the field of intelligent control, and specifically to a task scheduling optimization method for a multi-capacity automated guided vehicle (AGV) system. Background Technology

[0002] Automated Guided Vehicles (AGVs) are a prime example of logistics robots. AGV systems, comprised of multiple AGVs, are widely used in various manufacturing systems and warehousing and distribution fields. Based on rated load capacity, they can be categorized into single-load AGVs and multi-load AGVs. Multi-load AGVs can load and unload multiple trailers, allowing for the loading of multiple loads from multiple dispatch points or the unloading of multiple loads from multiple delivery points within a single material handling task. This results in higher single-vehicle transport capacity and greater flexibility in delivery operations. However, their multi-tasking nature also increases the difficulty of system task scheduling.

[0003] To ensure the completeness of materials and the timeliness of delivery, it is necessary to guarantee the rationality of the material delivery task sequence and AGV allocation in task scheduling. However, a review of existing literature on AGV material handling reveals that research mainly focuses on single-capacity AGVs, with limited research on task scheduling problems in multi-capacity AGV systems. The existing research primarily concentrates on the deadlock prevention scheduling problem for multi-capacity AGVs, such as the Chinese invention patent with publication number CN107544513A. There is currently no relevant research literature on the optimization of multi-capacity AGV task scheduling. Summary of the Invention

[0004] To address the technical deficiencies in existing technologies, this invention proposes a task scheduling optimization method for multi-capacity Automated Guided Vehicle (AGV) systems. By optimizing the task scheduling method of the AGV system through intelligent optimization algorithms, this invention solves the problems of excessive control information interaction and low flexibility between the scheduling system and the AGV in existing technologies, thereby improving the efficiency and flexibility of the scheduling system.

[0005] This invention is achieved through the following technical solution:

[0006] A method for optimizing task scheduling in a multi-capacity automated guided vehicle (AGV) system, including the following steps:

[0007] Step 1: When the AGV system meets the task allocation triggering conditions, obtain the task information in the system;

[0008] Step 2: Determine the set of AGVs to be scheduled, and determine the set of allocable tasks according to the method for determining the set of allocable tasks;

[0009] Step 3: Determine the task sequence based on the artificial immunity-gray wolf optimization algorithm;

[0010] Step 4: Determine the AGV allocation scheme according to the task allocation rules;

[0011] Step 5: Update the system status according to the AGV allocation plan, and return to Step 1.

[0012] Furthermore, the task allocation triggering conditions include periodic triggering and event triggering.

[0013] Furthermore, the set of AGVs to be scheduled consists of idle AGVs in the system that have completed the previous task and have not been assigned a new task.

[0014] Furthermore, the method for determining the allocatable task set includes the following steps:

[0015] Step 1.1: Determine the set of tasks to be assigned based on the task information in the system;

[0016] Step 1.2: Determine the pre-allocated task set according to the deadlock avoidance rules;

[0017] Step 1.3: Based on the total idle capacity of the AGV set to be scheduled, determine the set of allocable tasks, and number the tasks in the set of allocable tasks in sequence.

[0018] Furthermore, the set of tasks to be assigned is a collection of all tasks generated in the AGV system that have not yet been assigned to AGVs for transportation.

[0019] Furthermore, the pre-allocated task set is the set of remaining tasks after the tasks in the task set to be allocated are prohibited from dangerous tasks according to the deadlock avoidance rules.

[0020] Furthermore, the allocatable task set is determined by comparing the number of tasks N1 in the pre-allocated task set with the total idle capacity N2 of the AGV set to be scheduled. When the total idle capacity N2 of the AGVs to be scheduled is greater than the number of tasks N1 in the pre-allocated task set, the pre-allocated task set is the allocatable task set; otherwise, the tasks are sorted in descending order according to their urgency, and the top N2 tasks are selected to form the allocatable task set.

[0021] Furthermore, the artificial immunity-gray wolf optimization algorithm includes the following steps:

[0022] Step 2.1: Encode the tasks in the assignable task set using an encoding method based on the assignable task set;

[0023] Step 2.2: Set the total number N of tasks in the allocatable task set as the population size, initialize the first n populations using the maximum minimum distance clustering algorithm, and initialize the last Nn populations using a random method;

[0024] Step 2.3: Evaluate the fitness of individuals in the population using the target index, and stratify the individuals using the fast non-dominated ordination method;

[0025] Step 2.4: Determine if the algorithm meets the iteration termination condition. If yes, output the result; otherwise, proceed to step 2.5.

[0026] Step 2.5: Based on the hierarchical ranking results, perform social stratification using the Gray Wolf algorithm to determine α wolves, β wolves, and δ wolves;

[0027] Step 2.6: Select N using a variable probability method for the parent population. J A select few individuals, for N J Cloning operations are performed on elite individuals to obtain a scale of N. C Cloned populations;

[0028] Step 2.7: Determine whether the algorithm has reached the predetermined number of iterations. If yes, proceed to step 2.9; otherwise, proceed to step 2.8.

[0029] Step 2.8: Perform a single-level mutation operation on the cloned population to obtain a mutant population, adjust the mutant population to form a progeny population, and proceed to step 2.10;

[0030] Step 2.9: Perform multi-level mutation operations on the cloned population to obtain a mutant population, adjust the mutant population, and form a progeny population;

[0031] Step 2.10: Merge the parent and offspring populations to form a population of size N+N. C A mixed population;

[0032] Step 2.11: Use an elite preservation strategy to select better individuals from the mixed population to form the next generation of parent population, and proceed to step 2.3.

[0033] Furthermore, the encoding method based on the allocatable task set involves arranging the execution order of tasks in the allocatable task set using task numbers as objects, forming a task execution sequence. This sequence serves as the individual encoding when the artificial immune-gray wolf optimization algorithm sorts the tasks in the allocatable task set.

[0034] Furthermore, the target indicators include two objectives: minimizing the system's latency and minimizing the road network imbalance.

[0035] Furthermore, the iteration termination conditions include reaching a set number of iterations and the optimal solution remaining unchanged for multiple consecutive generations.

[0036] Furthermore, the fast non-dominated sorting method divides the population into multiple non-dominated levels based on Pareto dominance relationships. The first non-dominated level represents the set of non-dominated individuals obtained from multi-objective numerical comparisons in the current population; the second non-dominated level represents the set of non-dominated individuals obtained from multi-objective numerical comparisons after removing individuals from the first non-dominated level; this division continues until the last level consists entirely of non-dominated individuals, and no further levels can be created. Here, α wolves originate from individuals in the first non-dominated level, β wolves and δ wolves originate from individuals in the second and third levels, respectively, and the remaining individuals are called ω wolves.

[0037] Furthermore, the variable probability method first determines the total number N of elite individuals. J Among them, elite individuals include α wolves, β wolves, and δ wolves obtained through social stratification, and the remaining elite individuals N J -3 is allocated to the first, second, third, fourth and higher non-dominated levels with four different probabilities: a, b, c, and d. If the population has fewer non-dominated levels, the slots for lower levels belong to the non-dominated level above it. Where a + b + c + d = 1, and a > b > c > d.

[0038] Furthermore, the single-level mutation operation employs an improved Gaussian mutation operator and performs mutation with a probability of z (0 < z < 1). The calculation method for the mutation operation is as follows:

[0039]

[0040] In the formula, t ij It is the j-th gene in the i-th individual of the clone population, where i = 1, 2, ..., N. C j = 1, 2, ..., N, N C Let N be the size of the clone population, and N be the total number of tasks in the allocable task set. It is t ij The genes that produce new individuals after mutation, N G (0, 1) is a Gaussian variable with a mean of 0 and a variance of 1.

[0041] Furthermore, the size of the clone population N C The calculation method is as follows:

[0042]

[0043] In the formula, N J The size of the elite population, the size of the clone population, and N J Proportional; round(.) is the floor function; β is a random number between 0 and 1; E iThe number of clones of each individual in the elite population is related to its fitness level, which is the non-dominant level of the individual.

[0044] Furthermore, in adjusting the mutant population, to ensure that each gene in the mutant population is unique, the genes of the mutant population individuals need to be adjusted sequentially from the first gene to the last gene. The gene to be adjusted is compared with the previously successfully mutated gene. If the gene to be adjusted is not a duplicate of the previously successfully mutated gene, the mutation is successful; otherwise, the result is discarded. Then, based on the principle of similarity, an integer that is not a duplicate of the previous result is selected from the first to the last gene as the result of the gene mutation. After gene adjustment, the mutant population individuals become offspring population individuals.

[0045] Furthermore, the multi-level mutation operation involves using the offspring formed by the single-level mutation as the parent generation of the multi-level mutation, performing the mutation operation again to form new offspring, and selecting the optimal offspring from the offspring of the single-level mutation and the offspring of the multi-level mutation as the offspring of this mutation operation.

[0046] Furthermore, the elite preservation strategy involves first using a fast non-dominated sorting method to stratify the individuals in the mixed population based on their non-dominated level, then sorting the non-dominated individuals within the same stratum based on their crowding distance, where individuals with higher non-dominated levels and larger crowding distances are preferred over individuals with lower non-dominated levels and smaller crowding distances. Finally, the best N individuals from the mixed population are selected as the parent population for the next generation based on the non-dominated level stratification and crowding distance method for update operations.

[0047] Furthermore, the task sequence refers to the order in which tasks in the allocatable task set are executed.

[0048] Furthermore, the task allocation rule adopts an AGV load balancing evaluation method for task allocation. First, the total transport distance of each idle AGV that has completed all tasks is calculated, and the AGVs are arranged in ascending order of total transport distance. The available tasks are then allocated to AGVs with shorter total transport distances based on their load capacity, in order to ensure the load balancing of the system's AGVs.

[0049] Furthermore, the AGV allocation scheme is the correspondence between tasks in the allocable task set and AGVs to be scheduled in the system.

[0050] Compared with the prior art, the present invention has at least the following beneficial effects or advantages:

[0051] This invention proposes a task scheduling optimization method for a multi-capacity automated guided vehicle (AGV) system. By optimizing the task scheduling method of AGV through a manual-immune algorithm, it solves the problems of excessive control information interaction and low flexibility between the scheduling system and AGV in the prior art, thereby improving the efficiency and flexibility of the scheduling system. Attached Figure Description

[0052] The present invention will now be described in further detail with reference to the accompanying drawings:

[0053] Figure 1 This is a flowchart of the task scheduling optimization method for the multi-capacity automated guided vehicle system of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Because the scheduling systems for single-load and multi-load automated guided vehicles are quite complex, the following assumptions are made regarding the conveying system:

[0056] 1. All multi-load AGVs are loaded at a fixed location, and then the multi-load AGVs are transported at a uniform speed according to the planned unloading sequence.

[0057] 2. Once the AGV starts transporting goods, it will not automatically change its path or adjust the unloading order.

[0058] 3. Idle multi-capacity AGVs stay at specific task waiting points to wait for the next material task, without affecting the passage of other multi-capacity AGVs in the path network;

[0059] 4. The system path network is a one-way single lane and does not support overtaking;

[0060] 5. The multi-capacity AGV loads trailers sequentially from front to back and unloads trailers in a first-in-last-out manner.

[0061] 6. The unloading time for multi-capacity AGVs is fixed. If multiple AGVs are at the same loading and unloading point, they need to wait in line.

[0062] 7. When the AGV begins to perform a material handling task, it searches for the shortest one-way path from its current location to the target workstation in the unidirectional guide path network through path planning. Furthermore, path replanning is not used during traffic control to avoid motion conflicts within the AGV system.

[0063] Figure 1 This is a flowchart of the task scheduling optimization method for a multi-capacity automated guided vehicle (AGV) system according to the present invention. The task scheduling optimization method for a multi-capacity AGV system specifically includes the following steps:

[0064] Step 1: When the AGV system meets the task allocation triggering conditions, obtain the task information in the system;

[0065] When the AGV system calls the task scheduling optimization method of the multi-capacity automated guided vehicle system, it will first check whether the task allocation trigger condition is met. If the task allocation trigger condition is met, it will obtain the task information in the system, including the type of tasks assigned to AGVs but not yet completed, the starting point of the transport, the ending point of the transport, the task start loading time, the task deadline, and the running path status, as well as the type of tasks not assigned to AGVs, the starting point of the transport, the ending point of the transport, the task creation time, and the task deadline.

[0066] In this embodiment of the invention, the task allocation triggering conditions include periodic triggering and event triggering. Periodic triggering is triggered periodically at fixed time intervals, while event triggering is triggered when a set event occurs. In this embodiment of the invention, the set event includes the number of tasks in the task set to be allocated reaching a certain value.

[0067] It should be noted that the task allocation triggering condition can also be either periodic triggering or event triggering. The set event in the event triggering does not have to be the set event in the embodiment of the present invention. For example, the set event can also be the occurrence of multiple AGV tasks with large loads in the task set to be allocated.

[0068] Step 2: Determine the set of AGVs to be scheduled, and determine the set of allocable tasks according to the method for determining the set of allocable tasks;

[0069] When the AGV system meets the task allocation trigger condition, the set of AGVs to be scheduled is determined, and the allocable task set is determined according to the method for determining the allocable task set. The set of AGVs to be scheduled and the method for determining the allocable task set in the AGV system are then evaluated. The set of AGVs to be scheduled consists of idle AGVs in the system that have completed their previous task and have not been assigned a new task. If the set of AGVs to be scheduled or the allocable task set is empty, there are no idle AGVs to perform transport tasks or no transport tasks requiring AGV allocation; therefore, no task scheduling is needed, and the process proceeds to step 1, waiting for the system to meet the task allocation trigger condition again. If the set of AGVs to be scheduled and the allocable task set are not empty, the process proceeds to step 3.

[0070] Step 3: Determine the task sequence based on the artificial immunity-gray wolf optimization algorithm;

[0071] If the set of AGVs to be scheduled and the set of allocable tasks are not empty, the tasks in the set of allocable tasks are sorted according to the Artificial Immune-Grey Wolf Optimization Algorithm to determine the optimal task sequence. The task sequence is the order in which the tasks in the set of allocable tasks are executed.

[0072] Step 4: Determine the AGV allocation scheme according to the task allocation rules;

[0073] After the tasks in the allocatable task set are sorted using the artificial immunity-gray wolf optimization algorithm, they are assigned to idle AGVs for transport according to the task allocation rules, thus determining the AGV allocation scheme. The AGV allocation scheme is the correspondence between tasks in the allocatable task set and AGVs to be scheduled in the system, that is, determining the tasks assigned to the AGVs to be scheduled after task scheduling.

[0074] In this embodiment of the invention, the task allocation rule adopts an AGV load balancing evaluation method for task allocation. First, the total transport distance of each idle AGV that has completed all tasks is calculated, and the AGVs are arranged in ascending order of total transport distance. The available tasks are then allocated to AGVs with shorter total transport distances based on the load capacity of the AGVs with larger load capacities, in order to ensure the load balancing of the AGVs in the system.

[0075] It should be noted that the task allocation rules are not limited to the methods mentioned above, and can also be based on other methods, such as the shortest unloaded distance or the remaining power of the AGV.

[0076] Step 5: Update the system status according to the AGV allocation plan, and return to Step 1.

[0077] Update the system status according to the AGV allocation scheme, including the system path occupancy status, AGV status, task status, etc. Then return to step 1 and wait for the next task allocation trigger condition to be met before performing the next task scheduling.

[0078] The method for determining the allocatable task set includes the following steps:

[0079] Step 1.1: Determine the set of tasks to be assigned based on the task information in the system;

[0080] Based on the task information obtained from the system, a set of tasks that have been generated in the AGV system but have not yet been assigned to AGVs for transportation is determined, namely the task set to be assigned.

[0081] Step 1.2: Determine the pre-allocated task set according to the deadlock avoidance rules;

[0082] When it is determined that the set of tasks to be assigned in the system is not empty, deadlock avoidance rules are adopted to prohibit some dangerous tasks in the set of tasks to be assigned, so as to prevent deadlock from occurring in the system. Then, the remaining tasks in the set of tasks to be assigned are combined into a pre-assigned task set.

[0083] It should be noted that deadlock avoidance rules are not the focus of this invention, so they will not be elaborated on further. If necessary, please refer to Chinese invention patent with publication number CN107544513A.

[0084] Step 1.3: Determine the set of tasks that can be assigned based on the total idle capacity of the AGV set to be scheduled;

[0085] When the pre-allocated task set and the AGV set to be scheduled are not empty, the first step is to compare the number of tasks N1 in the pre-allocated task set with the total idle capacity N2 of the AGV set to be scheduled. If the total idle capacity N2 of the AGV set to be scheduled is greater than the number of tasks N1 in the pre-allocated task set, then there are enough idle AGVs to transport all the tasks in the pre-allocated task set, and no task selection is required. The pre-allocated task set is then the allocatable task set. Conversely, if the total idle capacity N2 of the AGV set to be scheduled is less than the number of tasks N1 in the pre-allocated task set, there is a situation where there are more transport tasks than idle AGVs. In this case, the tasks are sorted in descending order according to their urgency, and the top N2 tasks are selected to form the allocatable task set.

[0086] It should be noted that if any one of the task set to be assigned, the pre-assigned task set, the allocable task set, or the AGV set to be scheduled in the system is empty, no task scheduling is required until the task assignment trigger condition is met again and all the above sets are not empty, in which case task scheduling will be performed.

[0087] The artificial immunity-gray wolf optimization algorithm includes the following steps:

[0088] Step 2.1: Encode the tasks in the assignable task set using an encoding method based on the assignable task set;

[0089] When the system's set of assignable tasks is not empty, the artificial immune-gray wolf optimization algorithm is needed to sort the tasks. First, an encoding method based on the set of assignable tasks is used to encode the tasks in the set.

[0090] In this embodiment of the invention, the encoding method based on the allocatable task set is to arrange the task execution order in the allocatable task set using task numbers as objects, forming a task execution sequence. This sequence represents the individual encoding of the task sorting in the allocatable task set by the artificial immune-gray wolf optimization algorithm. For example: B = [b1, b2, ..., b j , ..., b N For j, task b is the corresponding task. j The position of [b1, b2, ..., b] in the sorted sequence of all available tasks, with encoding length N equal to the total number of tasks in the available task set. j , ..., b N From first to last, the load capacity A of the multi-capacity AGV. v Divided into Clusters facilitate the allocation of tasks from a cluster to AGVs.

[0091] Step 2.2: Set the total number N of tasks in the allocatable task set as the population size, initialize the first n populations using the maximum minimum distance clustering algorithm, and initialize the last Nn populations using a random method;

[0092] First, set the total number N of tasks in the allocatable task set as the population size. Then, initialize the first n populations using the maximum minimum distance clustering algorithm and the last Nn populations using a random method. This can ensure the system's on-time performance and road network balance, and is more likely to assign tasks with similar unloading points to the same AGV, while assigning tasks with distant unloading points to different AGVs.

[0093] Step 2.3: Evaluate the fitness of individuals in the population using the target index, and stratify and rank the individuals according to the fast non-dominated ordination method;

[0094] Based on the obtained new population, the fitness of individuals in the population is evaluated using target indicators, and individuals are stratified and ranked according to the fast non-dominated ranking method to facilitate the selection of the optimal individuals.

[0095] In this embodiment of the invention, the target indicators include two objectives: minimizing the system's latency rate and minimizing the road network imbalance. Then, a fitness function is determined based on the objectives, and the fitness values ​​of tasks in the allocable task set are calculated by clustering them with idle AGVs under the system state.

[0096] It should be noted that the target indicators may adopt one or more of the above rules, or other rules.

[0097] Step 2.4: Determine if the algorithm meets the iteration termination condition. If yes, output the result; otherwise, proceed to step 2.5.

[0098] After evaluating the fitness of individual individuals in the population and sorting the population in descending order, determine whether the artificial immunity-gray wolf optimization algorithm meets the iteration termination condition: if yes, the optimal individual in the population is the optimal task sequence; if not, proceed to step 2.5.

[0099] In this embodiment of the invention, the iteration termination conditions include reaching a set number of iterations and the optimal solution remaining unchanged for several consecutive generations.

[0100] It should be noted that the iteration termination condition can be one or more of the rules mentioned above, or it can be any other rule mentioned above.

[0101] Step 2.5: Determine the social stratification of the Grey Wolf algorithm based on the stratified ranking results;

[0102] When the algorithm does not meet the termination condition, it is necessary to determine the α wolf, β wolf, and δ wolf based on the hierarchical sorting results obtained by the fast non-dominated sorting method, and use them to lead other individuals to search for the optimal solution.

[0103] In this embodiment of the invention, the fast non-dominated sorting method divides the population into multiple non-dominated levels based on Pareto dominance relationships. The first non-dominated level represents the non-dominated solution set obtained by comparing multiple objective values ​​in the current population. The second non-dominated level represents the non-dominated individual solution set obtained by comparing multiple objective values ​​after removing the first level in the population. This division continues until no more levels can be created. Specifically, α wolves originate from individuals in the first non-dominated level, β wolves and δ wolves originate from individuals in the second and third levels, respectively, and the remaining individuals are called ω wolves. If the current population has two non-dominated levels, then α wolves originate from individuals in the first level, β wolves and δ wolves originate from individuals in the second level, and the remaining individuals are called ω wolves; otherwise, when the population has three or more non-dominated levels, then α wolves, β wolves, and δ wolves originate from individuals in the first, second, and third levels, respectively, and the remaining individuals are called ω wolves.

[0104] Step 2.6: Select N using a variable probability method for the parent population. J A select few individuals, for N J Cloning is performed on each elite individual to obtain N individuals. C Cloned populations;

[0105] To increase population diversity and prevent premature convergence of the algorithm, a variable probability method is used to select elite individuals from the parent population. J Cloning is performed on each elite individual to obtain N individuals. C Clonal population.

[0106] In this embodiment of the invention, the variable probability method is as follows: First, determine the total number N of elite individuals. J Among them, α wolves, β wolves, and δ wolves, determined through social stratification, are definitely elite individuals, while the remaining elite individuals N J -3 is allocated to the first, second, third, fourth and above non-dominated levels according to four different probabilities: a, b, c, and d. If the population has fewer non-dominated levels, the slots for lower levels belong to the non-dominated level above them. Where a + b + c + d = 1, and a > b > c > d.

[0107] Step 2.7: Determine whether the algorithm has reached the predetermined number of iterations. If yes, proceed to step 2.9; otherwise, proceed to step 2.8.

[0108] Mutation operations achieve local search by employing a gene mutation method. However, general single-level mutation (i.e., performing only one mutation operation) can only perform a small-scale search. When the system gets stuck in a local optimum, it is insufficient to escape the local optimum. Determine if the algorithm has reached the predetermined number of iterations. If yes, proceed to step 2.9; otherwise, proceed to step 2.8.

[0109] Step 2.8: Perform a single-level mutation operation on the cloned population to obtain a mutant population, adjust the mutant population to form a progeny population, and proceed to step 2.10;

[0110] If the number of iterations of the algorithm is within the predetermined number of iterations, a single-level mutation operation is performed on the clone population to obtain a mutant population. Since the individuals in the mutant population have duplicate genes, which does not meet the requirements, the mutant population needs to be adjusted to form a progeny population, and then proceed to step 2.10.

[0111] In this embodiment of the invention, the single-level mutation operation employs an improved Gaussian mutation operator and performs mutation with a probability of 0.5. The mutation operation calculation method is as follows:

[0112]

[0113] In the formula, t ij It is the j-th gene in the i-th individual of the clone population, where i = 1, 2, ..., N. C j = 1, 2, ..., N, N C Let N be the total size of the clone population, and N be the total number of tasks in the allocable task set. It is t ij In the new individuals produced after mutation, the gene N... G (0, 1) is a Gaussian variable with a mean of 0 and a variance of 1.

[0114] In this embodiment of the invention, the total size N of the clone population C The calculation method is as follows:

[0115]

[0116] In the formula, N J The size of the elite population, the size of the clone population, and N J Proportional; round(.) is the floor function; β is a random number between 0 and 1; E i The number of clones of each individual in the elite population is related to its fitness. Equation (2) can ensure that each individual in the elite population has a certain number of clones.

[0117] In this embodiment of the invention, the mutant population is adjusted. To ensure that each gene in the mutant population is unique, the genes of the mutant population individuals need to be adjusted sequentially from the first gene to the last gene. The gene to be adjusted is compared with the previously successfully mutated gene. If the gene to be adjusted is not a duplicate of the previously successfully mutated gene, the mutation is successful; otherwise, the result is discarded. Then, based on the principle of similarity, an integer that is not a duplicate of the previous result is selected from the first to the last gene as the result of the gene mutation. After gene adjustment, the mutant population individuals become offspring population individuals.

[0118] Step 2.9: Perform multi-level mutation operations on the cloned population to obtain a mutant population, adjust the mutant population, and form a progeny population;

[0119] If the algorithm reaches the predetermined number of iterations, it may get stuck in a local optimum. In this case, a multi-level mutation operation is performed on the clone population to obtain a mutant population. Then, the mutant population is adjusted to form the offspring population.

[0120] In this embodiment of the invention, the multi-level mutation operation uses the offspring formed by the single-level mutation as the parent of the multi-level mutation, performs the mutation operation again to form new offspring, and selects the optimal offspring from the offspring of the single-level mutation and the offspring of the multi-level mutation as the offspring of this mutation operation. Since multi-level mutation consumes more resources, it is only used when trapped in a local optimum.

[0121] Step 2.10: Merge the parent population and the offspring population to form a population of size N+N. C A mixed population;

[0122] After mutation, the parent population with a population size of N and the population with a population size of N are... C The offspring populations are merged into a single population of size N+N. C The populations of the two species form a mixed population.

[0123] Step 2.11: Use an elite preservation strategy to select better individuals from the mixed population to form the next generation of parent population, and proceed to step 2.3.

[0124] In order to select better individuals from the mixed population for the next iteration, an elite preservation strategy is adopted to select the better population as the parent population of the next generation, and then proceed to step 2.3.

[0125] In this embodiment of the invention, the elite preservation strategy involves using a fast non-dominated sorting method to stratify the individuals in the mixed population based on their non-dominated level. Then, non-dominated individuals within the same stratum are sorted by crowding distance, where individuals with higher non-dominated levels and larger crowding distances are preferred over individuals with lower non-dominated levels and smaller crowding distances. Finally, the best N individuals from the mixed population are selected as the parent population for the next generation based on the non-dominated level stratification and crowding distance method for update operations.

[0126] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A task scheduling optimization method for a multi-capacity automated guided vehicle (AGV) system, characterized in that, including the steps: Step 1: When the AGV system meets the task allocation trigger condition, obtain the task information in the system; Step 2: Determine the set of AGVs to be scheduled, and determine the set of assignable tasks according to the method for determining the set of assignable tasks; Step 3: Determine the task sequence according to the artificial immune-gray wolf optimization algorithm; Step 4: Determine the AGV allocation plan according to the task allocation rules; Step 5: Update the system state according to the AGV allocation plan, and return to Step 1; Among them, the method for determining the set of assignable tasks includes the following steps: Step 1.1: Determine the set of tasks to be assigned according to the task information in the system; Step 1.2: Determine the set of pre-assigned tasks according to the deadlock avoidance rules; Step 1.3: Determine the set of assignable tasks according to the total idle load of the set of AGVs to be scheduled, and number the tasks in the set of assignable tasks in sequence; The task allocation rule is to use the AGV load balance degree evaluation method for task allocation; first calculate the total handling distance of each idle AGV for all completed tasks, and arrange them in ascending order of the handling total distance. Assign the assignable tasks to the AGVs with less handling total distance according to the load of the multi-load AGV to ensure the AGV load balance degree of the system.

2. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 1, characterized in that, The artificial immune-gray wolf optimization algorithm includes the following steps: Step 2.1: Encoding the tasks in the set of assignable tasks by using the encoding method based on the set of assignable tasks; Step 2.2: Set the total number of tasks in the allocatable task set. Set the population size as the initialization factor, initialize the first n populations using the minimum-maximum distance clustering algorithm, and then initialize the remaining populations using a random method. Individual populations; Step 2.3: Evaluate the fitness of the population individuals with the target index, and use the fast non-dominated sorting method to sort the individuals hierarchically; Step 2.4: Judge whether the algorithm meets the iteration end condition. If so, output the result. If not, enter Step 2.5; Step 2.5: Perform social stratification using the Gray Wolf algorithm based on the stratified ranking results to determine... ; Step 2.6: Select the parent population using a variable probability method. An elite individual, for Cloning operations were performed on elite individuals to obtain a scale of [number]. Cloned populations; Step 2.7: Judge whether the iteration number of the algorithm reaches the predetermined iteration number. If so, enter Step 2.

9. If not, enter Step 2.8; Step 2.8: Perform a single-level mutation operation on the cloned population to obtain a mutated population, adjust the mutated population, form a subpopulation, and enter Step 2.10; Step 2.9: Perform a multi-level mutation operation on the cloned population to obtain a mutated population, adjust the mutated population, form a subpopulation; Step 2.10: Merge the parent and offspring populations to form a population of [size missing]. A mixed population; Step 2.11: Use the elite retention strategy to select better individuals from the mixed population to form the next generation of parent population, and enter Step 2.

3.

3. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 2, characterized in that, The encoding method based on the set of assignable tasks is to arrange the execution order of the tasks in the set of assignable tasks with the task numbers as the objects, form a task execution sequence, and this sequence is used as the individual encoding when the artificial immune-gray wolf optimization algorithm sorts the tasks in the set of assignable tasks.

4. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 2, characterized in that, The fast non-dominated sorting method divides the population into multiple non-dominated levels based on Pareto dominance relationships. The first non-dominated level represents the set of non-dominated individuals obtained from multi-objective numerical comparisons in the current population; the second non-dominated level represents the set of non-dominated individuals obtained from multi-objective numerical comparisons after removing individuals from the first non-dominated level; this division continues until the last level consists entirely of non-dominated individuals, and no further levels can be created. Individuals at the first non-dominant level Individuals originating from the second and third layers respectively; the remaining individuals are called... Wolf.

5. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 4, characterized in that, The variable probability method first determines the total number of elite individuals. Among them, elite individuals include those who have obtained their status through social stratification. Elite individuals, the remaining elite individuals -3 press Four different probabilities are assigned to the first, second, third, fourth and above non-dominated levels; if a population has fewer non-dominated levels, the slots at lower levels belong to the non-dominated level above it. ,and .

6. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 2, characterized in that, The single-level mutation operation is to use an improved Gaussian mutation operator and mutate with a probability of z (0 < z < 1). The calculation method of the mutation operation is: （1） In the formula, It is the j-th gene in the i-th individual of the clone population, where, , Let N be the size of the clone population, and N be the total number of tasks in the allocable task set. yes The genes that produce new individuals after mutation operations It is a Gaussian variable with a mean of 0 and a variance of 1. To The remainder operation; where, The calculation method is as follows: （2） In the formula, The size of the elite population, the size of the clone population, and Proportional; It is a rounding function; A random number between 0 and 1; The number of clones of each individual in the elite population is related to its fitness level, which is the non-dominant level of the individual.

7. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 2, characterized in that, The multi-level mutation operation is to use the offspring formed by the single-level mutation as the parent of the multi-level mutation, perform the mutation operation again to form a new offspring, and select the optimal offspring from the offspring of the single-level mutation and the offspring of the multi-level mutation as the offspring of this mutation operation.

8. The task scheduling optimization method for a multi-capacity automated guided vehicle system according to claim 2, characterized in that, The elite retention strategy involves first using a fast non-dominated sorting method to stratify individuals in the mixed population based on their non-dominated level, then sorting non-dominated individuals in the same stratum based on their crowding distance, where individuals with high non-dominated levels and large crowding distances are preferred over individuals with low non-dominated levels and small crowding distances; finally, selecting the best N individuals from the mixed population based on the non-dominated level stratification and crowding distance method to serve as the parent population for the next generation and perform an update operation.

Images