Rapid evaluation collaborative optimization scheduling method and system for workshop multi-load AGV scheduling and path planning

Abstract

The invention belongs to the related technical field of AGV scheduling, and discloses a rapid evaluation collaborative optimization scheduling method and system for workshop multi-load AGV scheduling and path planning. The method comprises the following steps: S1, determining the total number of tasks of a to-be-processed workshop, the total number of AGVs, the number of trays on each AGV and the bearing capacity of a single tray; s2, initializing individuals in the bee colony and the total number of the individuals; calculating the transportation completion time of each individual and the shortest transportation completion time of the bee colony; s3, updating the bee colony and calculating the transportation completion time of each individual after updating; and S4, repeating S3 until a preset termination condition is reached, thereby obtaining the shortest transportation completion time and the individual corresponding to the shortest transportation completion time, and realizing a scheduling process. According to the invention, the transportation efficiency is obviously improved, and the production cost is reduced, so that the economic benefits of enterprises are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field related to AGV scheduling, and more specifically, relates to a rapid evaluation collaborative optimization scheduling method and system for multi-load AGV scheduling and path planning in a workshop. Background Art

[0002] Workshop logistics is an indispensable and important link in the operation of manufacturing enterprises, involving the entire transfer process of materials from raw materials to finished products. Due to the complexity and diversity of the logistics link, including storage, handling, and loading and unloading, etc., the efficiency of workshop logistics directly affects production costs and overall production efficiency. With the rapid development of the manufacturing industry and the continuous increase in market demand, optimizing the scheduling and management of workshop logistics has become increasingly important. In-depth research on the optimization methods of workshop logistics can not only significantly shorten the production cycle, improve resource utilization rate, but also enhance the competitiveness and economic benefits of enterprises, and thus promote the sustainable development of the entire manufacturing industry. Therefore, the research on workshop logistics has important theoretical value and practical significance. At the same time, it can also effectively improve the economic benefits of enterprises, and thus promote the further development of infrastructure projects.

[0003] Multi-load automated guided vehicles (AGVs) have shown significant characteristics in modern logistics scheduling and have become an important part of manufacturing automation. It can carry multiple goods simultaneously, which significantly improves the transportation efficiency and meets the high demand for material handling in a large-scale production environment. At the same time, the scheduling process needs to comprehensively consider multiple factors such as the loading capacity of AGVs, the priority of transportation tasks, and real-time road conditions changes, making the scheduling strategy not only complex but also highly flexible.

[0004] In the current research on multi-load automated guided vehicle (AGV) logistics scheduling, scheduling and path planning often deeply explore their respective sub-problems separately, which results in only local optimization being achieved and the effective collaboration between the two not being realized. Especially in a complex production environment, the scheduling of multi-load AGVs needs to comprehensively consider multiple factors such as task allocation, loading capacity, and transportation paths. At the same time, many studies still mainly rely on the A* algorithm for path planning. Although this algorithm has good performance in theory, it often faces high computational time consumption in practical applications. Especially as the number of map nodes increases, the running time significantly extends, which affects the overall scheduling efficiency. This problem is particularly prominent in a complex logistics environment because a multi-load multi-AGV (automated guided vehicle) system requires efficient and safe path planning and scheduling. Therefore, for the existing multi-load multi-AGV logistics workshop, researching a collaborative optimization scheduling method considering AGV scheduling and path planning has important engineering value, which will help improve the efficiency and response ability of the system. Summary of the Invention

[0005] In view of the above deficiencies or improvement requirements of the prior art, the present invention provides a rapid evaluation collaborative optimization scheduling method for the scheduling and path planning of multi-load AGVs in a workshop, which solves the scheduling problem of complex multi-load AGVs in a workshop.

[0006] To achieve the above object, according to one aspect of the present invention, there is provided a rapid evaluation collaborative optimization scheduling method for the scheduling and path planning of multi-load AGVs in a workshop, the method comprising the following steps:

[0007] S1 Determine the total number of tasks n in the workshop to be processed, the total number of AGVs m, the number of pallets on each AGV, and the load capacity of a single pallet;

[0008] S2 Initialize the individuals and the total number of individuals in the bee colony. Each individual corresponds to a sequence of tasks. Each sequence of tasks contains n tasks and m - 1 delimiters. The m - 1 delimiters divide the n tasks into m groups and assign the m groups of tasks to the AGVs in turn;

[0009] For the tasks assigned to each AGV, obtain the loading position and target machine of each task. Take the initial position of each AGV, the loading position and target machine of each task as nodes respectively; partition the tasks on each AGV according to the volume of the tasks and the number and load capacity of the pallets; starting from the initial position of the AGV, traverse all the nodes in each area to form a path. Taking the shortest path as the goal, with the loading position nodes in each area prior to the target machine nodes, calculate the optimal path and the time for the current AGV to complete transportation according to the optimal path. Sum up the times for all AGVs in each individual to complete transportation according to the optimal path to obtain the transportation completion time of each individual;

[0010] S3 Update the bee colony and calculate the transportation completion time of each individual after the update. Determine whether the transportation completion time of the individual after the update is better than that of the individual before the update. If it is better than the individual before the update, retain the individual after the update; otherwise, do not update the individual;

[0011] S4 Repeat S3 until a preset termination condition is reached, thereby obtaining the shortest transportation completion time and the individual corresponding to the shortest transportation completion time, so as to implement the scheduling process.

[0012] Further preferably, in step S2, the A* algorithm is used to calculate the obtained optimal path.

[0013] Further preferably, in step S2, the obtained optimal path is calculated in combination with a rapid evaluation algorithm. The steps of the rapid evaluation algorithm are as follows:

[0014] Calculate the theoretical value of the path length between two adjacent nodes according to the following formula, and take the minimum sum of the theoretical values of the path lengths of all adjacent nodes as the goal to calculate the obtained optimal path;

[0015] d t (A,B) = |x a - x b | + |y a - y b | + k p × Error now + k i × Error sum + k d × Error delta

[0016]

[0017] Error sum + = Error now

[0018] Error delta = Error now - Error last

[0019] Among them, x a and y a are the coordinates of node A, x b and y b are the coordinates of node B, k p is the proportional term coefficient, k i is the integral term coefficient, k d is the differential term coefficient, Error now is the proportional term, Error sum is the integral term, Error delta is the differential term.

[0020] Further preferably, the actual value is the optimal transportation completion time obtained by the previous generation of bee colony using the A* algorithm, and the theoretical value is the optimal transportation completion time obtained by the previous generation of bee colony using the fast evaluation algorithm.

[0021] Further preferably, the actual value is the average value of the transportation completion times of three individuals selected from the top 20% sorted by the shortest transportation completion time among the previous generation of bee colonies and the individual ranked first.

[0022] Further preferably, in step S3, each individual in the bee colony is divided into a leading bee and a following bee. When updating the bee colony, the leading bee adopts a global search algorithm, and the following bee adopts a neighborhood selection strategy for updating.

[0023] Further preferably, the neighborhood search strategy includes seven neighborhoods, and one of the seven neighborhoods is selected to update the follower bee. The seven neighborhoods are as follows:

[0024] Neighborhood 1: Randomly select two AGVs in an individual, and exchange one task number in the codes of the two AGVs to form a new individual.

[0025] Neighborhood 2: In an individual, select the AGV with the longest transportation completion time and exchange the task number on another AGV to form a new individual.

[0026] Neighborhood 3: Insert the task number on one AGV in an individual into the task sequence of another AGV to form a new individual.

[0027] Neighborhood 4: In an individual, insert one task number in the task sequence of the AGV with the longest transportation completion time into the task sequence of another AGV in the individual, that is, select the AGV with the longest transportation completion time in the individual and insert one task number in its task sequence into the task sequence of another AGV in the individual to form a new individual.

[0028] Neighborhood 5: Randomly select an AGV in an individual, and exchange two task numbers in the task sequence of the selected AGV to form a new individual.

[0029] Neighborhood 6: In an individual, exchange any two task numbers in the task sequence of the AGV with the longest transportation completion time to form a new individual.

[0030] Neighborhood 7: In an individual, randomly select the task sequence of an AGV and update the task sequence by using the method of task grouping based on distance to form a new individual.

[0031] Further preferably, one of the seven neighborhoods is selected in the following manner:

[0032] Randomly generate a random number p between (0, 1). If the random number p is less than the preset threshold, then randomly select one neighborhood to update the individual; otherwise, calculate the selection probability of each neighborhood, and select the neighborhood with the largest probability to update the individual. The selection probability is calculated according to the following relational expression:

[0033]

[0034] where count g is the cumulative number of times the neighborhood is successfully selected, and count l is the number of times the l-th neighborhood is successfully selected.

[0035] Further preferably, each individual corresponds to a string of task sequences, and the task sequences use integer encoding, and the encoding ∈ {1, 2,..., n}.

[0036] According to another aspect of the present invention, there is provided a rapid evaluation collaborative optimization scheduling system for multi-load AGV scheduling and path planning in a workshop, which system includes an actuator for executing the rapid evaluation collaborative optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as described above.

[0037] Generally speaking, compared with the prior art by the above technical solution conceived by the present invention, the following beneficial effects are achieved:

[0038] 1. The present invention comprehensively considers that the multi-load characteristics of AGVs increase the complexity of scheduling and path planning. With the goal of minimizing the maximum transportation completion time, the bee colony algorithm is used for the encoding and decoding processes to obtain the shortest transportation completion time, simplifying the solution process, significantly improving the overall production efficiency of the workshop, and ensuring the reasonable utilization of resources and the efficient completion of tasks.

[0039] 2. The present invention proposes a rapid evaluation bee colony algorithm (ABC-RE), which rapidly evaluates the path length, significantly accelerating the decoding speed and solving the problem of too long decoding time when using the A* algorithm.

[0040] 3. The present invention introduces a neighborhood structure based on task grouping and an adaptive neighborhood selection strategy during the bee colony update process. Since a relatively large number of neighborhoods are used in this problem, this method dynamically adjusts its usage probability according to the real-time feedback of the solution quality and the success times of each neighborhood structure, significantly improving the performance of the algorithm, enabling it to adaptively focus on the currently most effective search strategy, while avoiding wasting computing resources on inefficient neighborhood structures, thus achieving a better balance between exploration and exploitation, accelerating convergence and improving the quality of the solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flowchart of the rapid evaluation collaborative optimization scheduling method for multi-load AGV scheduling and path planning constructed according to the preferred embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of the encoding scheme of the scheduling model constructed according to the preferred embodiment of the present invention;

[0043] Figure 3 is constructed according to the preferred embodiment of the present invention Figure 2 time window after decoding the encoding scheme in;

[0044] Figure 4 is the global search and neighborhood search modes involved in the present invention constructed according to the preferred embodiment of the present invention;

[0045] Figure 5It is a neighborhood search method for distance-based task grouping constructed according to the preferred embodiment of the present invention. Detailed implementation manners

[0046] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0047] A rapid evaluation collaborative optimization scheduling method and system for multi-load AGV scheduling and path planning in a workshop, the method comprising the following steps:

[0048] S1 Construct a multi-load multi-AGV logistics scheduling model, a workshop topological map, and a collaborative optimization scheduling model that simultaneously considers AGV scheduling and path planning in the workshop, and determine the total number of tasks n in the workshop to be processed, the total number of AGVs m, the number of pallets on each AGV, and the load capacity of a single pallet.

[0049] There are m multi-load AGVs in the workshop performing transportation tasks and n materials to be transported. The storage location and corresponding target machine of each material to be transported are known in advance. Each AGV has multiple pallets. During a single task, multiple materials can be transported, but there is an upper limit on the volume of materials that each pallet can load. If the material volume exceeds the upper limit, multiple pallets are required, and there is no situation where two materials share the same pallet.

[0050] Further, the maximum transportation completion time is the maximum value of the transportation completion times of all tasks on the last AGV. The calculation formula for the maximum completion time in the scheduling model is:

[0051] C max = max(T i ), i = 1, 2, …, n

[0052] In an embodiment of the present invention, an integer coding method is used for coding operations. For this problem, the coding length is n + m - 1, where the task numbers are represented as {1, 2, …, n}. Each task consists of a loading task and a transportation task, and the loading task needs to be executed before the transportation task. n + 1, …, n + m - 1 are used as delimiters to divide the tasks among m AGVs.

[0053] S2 Use a fast evaluation bee colony algorithm (ABC-RE) to optimize and solve the collaborative scheduling model to obtain the best order of task transportation for the AGVs.

[0054] Step 1: Represent a feasible solution of the scheduling model using integer encoding, and sequentially determine the transportation completion time of each AGV to finally obtain the optimization objective.

[0055] Step 2: Set up the bee colony and initialize the individuals.

[0056] Step 3: Initialize the quick evaluation parameter settings and the neighborhood selection probability.

[0057] Step 4: Use OBX for global search and execute the employed bee stage.

[0058] Step 5: Use an adaptive neighborhood selection strategy and execute the onlooker bee stage.

[0059] Step 6: Use variable neighborhood search to execute the scout bee stage.

[0060] Step 7: Check the termination condition. If it is satisfied, output the optimal individual of the bee colony; otherwise, return to Step 4.

[0061] Specifically, initialize the individuals and the total number of individuals in the bee colony. Each individual corresponds to a task sequence. Each task sequence contains n tasks and m - 1 separators. The m - 1 separators divide the n tasks into m groups, and the m groups of tasks are sequentially assigned to the AGVs.

[0062] For the tasks assigned to each AGV, obtain the loading position and the target machine of each task. Take the initial position of each AGV, the loading position and the target machine of each task as nodes respectively; partition the tasks on each AGV according to the volume of the tasks and the quantity and loading capacity of the pallets; starting from the initial position of the AGV, traverse all the nodes in each area to form a path, with the shortest path as the goal. In each area, the loading position node takes precedence over the target machine node, calculate the optimal path and the time for the current AGV to complete transportation according to the optimal path, and sum up the times for all AGVs in each individual to complete transportation according to the optimal path to obtain the transportation completion time of each individual.

[0063] During decoding, for the grouping of tasks, in the coding order, sequentially group the tasks according to the task volume and the loading capacity of the AGV, and then execute the transportation tasks in the coding order. For the loading order within the group, since the loading points are concentrated in the warehouse and most of the paths are one-way, the loading order within the package group is determined using a nearest neighbor-based heuristic algorithm. Then a series of points of the AGV can be obtained, use the A* algorithm to calculate the path length, obtain the time for each AGV to complete handling, and finally obtain the optimization objective C. max。A method for quickly evaluating the path length is proposed. This method is based on the idea of PID, takes the actual value decoded by the A* algorithm as the target, compares the theoretical value calculated by the quick evaluation with the actual value, and realizes the approximation of the actual value. After each iteration, three individuals are randomly selected from the top 20% of the individuals in each generation of the population, and together with the optimal solution of each generation, the difference C between their theoretical value and the actual value is calculated. a max (i) - C t max (i), analogized to PID control, the proportional term Error can be obtained in sequence. now = C a max (i) - C t max (i), the integral term Error sum += Error now The differential term Error delta = Error now - Error last 。The formula for the theoretical value of the path length is:

[0064] d t (A,B) = |x a - x b | + |y a - y b | + k p ×Error now + k i ×Error sum + k d ×Error delta

[0065] In an embodiment of the present invention, the number of bee swarms is PopSize, the initialization method of individuals is randomly generated, and each individual in the bee swarm has an integer coding structure; the running time of the algorithm is set to 0.5×n×m; count g is N g the number of successful neighborhood searches. An adaptive neighborhood selection strategy is adopted, and P g = count g / ∑ l＝1 7 count l represents the probability of selecting an operator for the g-th operation; at initialization, the probabilities of all operators being selected are the same.

[0066] S3 Update the bee swarm and calculate the transportation completion time of each individual after the update. Determine whether the transportation completion time of the updated individual is better than that of the individual before the update. If it is better than the individual before the update, retain the updated individual; otherwise, do not update the individual.

[0067] Further, the update of the bee colony individuals: In the population update process of the bee colony algorithm, first record the objective value of each individual before the update operation as f1, which is a theoretical value obtained through a fast evaluation calculation method. Perform a global search on the leading bees in the population, and select a neighborhood for the follower bees to perform the update operation. The update process depends on the mutual information between individuals and the existing theoretical evaluation values. At the end of each generation, calculate the theoretical fitness values of all individuals, randomly select three individuals from the top 20% of the individuals in each generation's population, and add the optimal solution of each generation to calculate its actual value, which guides the population update of the next generation to improve the overall fitness. Each individual in the bee colony is divided into a leading bee and a follower bee. When updating the bee colony, the leading bee adopts a global search algorithm, and the follower bee adopts a neighborhood selection strategy for update.

[0068] Randomly generate a random number p between (0, 1). If the random number p < α, then randomly select a neighborhood, otherwise adopt an adaptive neighborhood selection strategy according to the probability P g = count g / ∑ l＝1 7 count l Select a neighborhood N g , perform a fast evaluation on the obtained new encoding. If it is better than the original encoding scheme, then replace the original encoding scheme to become an individual in the next generation. If the two values are equal, then compare the sum of the total AGV transportation times. If it is less than the original solution, then replace it, otherwise maintain the original encoding unchanged, otherwise do not replace; meanwhile, the success count count g of this neighborhood N g is incremented by 1, and the probability P g of this neighborhood being selected will also increase. After all individuals have been processed, calculate the actual values of the top three individuals and update the parameters of the fast evaluation.

[0069] Further, the design of the operator mainly operates on integer encoding. For global search, an order-based crossover (OBX) is used. For local search, it includes the following 7 types:

[0070] Neighborhood structure 1: Operate on two different AGVs in the encoding. Randomly select two AGVs, and two tasks in the random AGV sequence, and exchange the task numbers at the two positions. Assume that the encoding of the current solution is [6, 5, 8, 12, 10, 3, 2, 11, 1, 4, 9, 7]. If the two randomly selected positions are position 3 and position 10, then the process encoding layer after exchange is [6, 5, 4, 12, 10, 3, 2, 11, 1, 8, 9, 7].

[0071] Neighborhood Structure 2: Operate on the AGV with the longest completion time in the encoding and another randomly selected AGV. Similar to Neighborhood Structure 1, select the AGV with the longest completion time and another randomly selected AGV, randomly select two positions in the random AGV sequence, and swap the task numbers at these two positions.

[0072] Neighborhood Structure 3: Operate on two different AGVs in the encoding. Randomly select two AGVs, randomly select two positions in the random AGV sequence, and insert one task of one AGV into the other AGV. Suppose the encoding of the current solution is [6, 5, 8, 12, 10, 3, 2, 11, 1, 4, 9, 7]. If the two randomly selected positions are position 3 and position 10, then the task encoding layer after swapping is [6, 5, 12, 10, 3, 2, 11, 1, 4, 8, 9, 7].

[0073] Neighborhood Structure 4: Operate on the AGV with the longest completion time in the encoding and another randomly selected AGV. Similar to Neighborhood Structure 3. Select the AGV with the longest completion time and another randomly selected AGV, randomly select two positions in the random AGV sequence, and insert one task of the AGV with the longest completion time into the other AGV.

[0074] Neighborhood Structure 5: Operate on one AGV in the encoding. Randomly select one AGV, randomly select two positions in the sequence of this AGV, and swap the task numbers at these two positions. Suppose the encoding of the current solution is [6, 5, 8, 12, 10, 3, 2, 11, 1, 4, 9, 7]. If the two randomly selected positions are position 10 and position 12, then the task encoding layer after swapping is [6, 5, 8, 12, 10, 3, 2, 11, 1, 7, 9, 4].

[0075] Neighborhood Structure 6: Operate on one AGV in the encoding. Similar to Neighborhood Structure 5. Select the AGV with the longest handling time, randomly select two positions in the sequence of this AGV, and swap the task numbers at these two positions.

[0076] Neighborhood Structure 7: In the encoding, select the task corresponding to one AGV and update this task using the method of distance-based task grouping to update the neighborhood search strategy of the encoding.

[0077] S4 Repeat S3 until a preset termination condition is reached, thereby obtaining the shortest transportation completion time and the individual corresponding to this shortest transportation completion time, so as to implement the scheduling process.

[0078] If the hybrid algorithm satisfies that the current running time reaches 0.5×n×m, then output the optimal individual of the population; otherwise, jump to Step 5 and iterate repeatedly until the termination condition is satisfied.

[0079] The following will further elaborate on the present invention with specific embodiments.

[0080] S1 Please refer to Figure 1 、 Figure 2 、 Figure 3 , Figure 4 and Figure 5 , in a logistics workshop, the number of storage location nodes = 30, the number of machine nodes (the number of target machines) = 30, the number of tasks = 8, the number of AGVs = 2, then the coding length is 8 + 2 - 1 = 10, the task volume (material volume) V i ∈[10, 60], the AGV loading capacity (the number of pallets) is 3, and the volume upper limit of a single pallet = 30.

[0081] S2 Setting and initialization of the population.

[0082] The total number of individuals in the bee colony is PopSize = 60, the number of leading bees and following bees are each PopSize / 2, the initialization method of individuals is random generation, and the objective value of each individual is calculated in turn according to the decoding scheme.

[0083] Figure 3 For Figure 2 the Gantt chart after decoding the coding scheme example in the decoding process example is: in the coding [5, 6, 7, 4, 3, 9, 8, 2, 1], taking 9 as the delimiter, assign the task [5, 6, 7, 4, 3] to AGV1, and the corresponding node sequence it walks through is assign the task [8, 2, 1] to AGV2, and a string of node sequences can also be obtained After decoding, the theoretical fitness value and actual fitness value of AGV1 and AGV2 can be obtained.

[0084] Set the running termination time of the hybrid algorithm to 0.5×n×m, and the number of neighborhood structures is 7. count g is N g the number of successful neighborhood searches, represents the probability of selecting an operation operator for the g-th operation; at initialization, the probabilities of all neighborhood structures being selected are the same, so the probability initial value is 1 / 7.

[0085] S3 Population update

[0086] Update of population individuals: In the leading bee stage, perform global search, using an order-based crossover (OBX). In the following bee stage, perform local search, randomly generate a random number p between (0, 1), if the random number p < α, then randomly select a neighborhood, otherwise adopt an adaptive neighborhood selection strategy, and select a neighborhood N according to the probability select a neighborhood N g, quickly evaluate the obtained new encoding. If it is better than the original encoding scheme, replace the original encoding scheme to become the individual of the next generation. If the values of the two are equal, compare the sum of the total transportation times of the AGVs. If it is less than the original solution, replace it; otherwise, keep the original encoding unchanged. At the same time, the success count g of the neighborhood N g is incremented by 1, and the selection probability P g of the corresponding neighborhood will also increase. After all individuals have been executed, randomly select three individuals from the top 20% of the individuals in each generation's population, and add the optimal solution of each generation to calculate their actual values, and update the parameters of the quick evaluation accordingly.

[0087] If neighborhood structure 1 is selected: Operate on two different AGVs in the encoding. Randomly select two AGVs, and randomly select two positions in the random AGV sequence, and swap the task numbers at the two positions.

[0088] If neighborhood structure 2 is selected: Operate on the AGV with the maximum completion time and another randomly selected AGV in the encoding. Similar to neighborhood structure 1, select the AGV with the maximum completion time and another randomly selected AGV, and randomly select two positions in the random AGV sequence, and swap the task numbers at the two positions.

[0089] If neighborhood structure 3 is selected: Operate on two different AGVs in the encoding. Randomly select two AGVs, and randomly select two positions in the random AGV sequence, and insert one task of one AGV into the other AGV.

[0090] If neighborhood structure 4 is selected: Operate on the AGV with the maximum completion time and another randomly selected AGV in the encoding. Similar to neighborhood structure 3. Select the AGV with the maximum completion time and another randomly selected AGV, and randomly select two positions in the random AGV sequence, and insert one task of the AGV with the maximum completion time into the other AGV.

[0091] If neighborhood structure 5 is selected: Operate on one AGV in the encoding. Similar to neighborhood structure 5. Select the AGV with the longest handling time, and randomly select two positions in the sequence of this AGV, and swap the task numbers at the two positions.

[0092] If neighborhood structure 7 is selected: Neighborhood search strategy based on distance-based task grouping. Similar to operator 6, randomly select two positions in the transportation layer of the optimal individual, and sequentially replace the transportation equipment allocation in the transportation layer.

[0093] S4 If the bee colony algorithm satisfies that the current running time reaches 0.5×n×m, output the optimal individual; otherwise, jump to step Step3 and iterate repeatedly until the termination condition is met.

[0094] The present invention also provides a rapid evaluation collaborative optimization scheduling system for multi-load AGV scheduling and path planning in a workshop. The system includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it executes the rapid evaluation collaborative optimization scheduling method for multi-load AGV scheduling and path planning as described above.

[0095] Those skilled in the art can easily understand that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A fast evaluation and collaborative optimization scheduling method for multi-load AGV scheduling and path planning in a workshop, characterized in that: The method comprises the following steps: S1 determines the total number of tasks to be processed in the workshop n, the total number of AGVs m, the number of pallets on each AGV, and the carrying capacity of a single pallet; S2 initializes the individuals and the total number of individuals in the swarm. Each individual corresponds to a sequence of tasks. Each sequence of tasks contains n tasks and m-1 separators. The m-1 separators divide the n tasks into m groups and assign the m groups of tasks to the AGVs in sequence. For each task assigned to an AGV, obtain the loading position and target machine of each task, and use the initial position of each AGV, the loading position and target machine of each task as nodes respectively; partition the tasks on each AGV according to the volume of the task and the number and loading capacity of pallets; Starting from the initial position of the AGV, traverse all nodes in each zone to form a path, with the shortest path as the goal. The loading position node in each zone prioritizes the target machine node, calculates the optimal path and the time it takes for the current AGV to complete the transportation according to the optimal path, and sums the time it takes for all AGVs in each individual to complete the transportation according to the optimal path to obtain the transportation completion time of each individual; S3 updates the bee colony and calculates the transportation completion time of each individual after the update, and determines whether the transportation completion time of the updated individual is better than that of the individual before the update. If it is better, the updated individual is retained, otherwise the individual is not updated; S4 repeats S3 until the preset termination condition is reached, thereby obtaining the shortest transportation completion time and the individual corresponding to the shortest transportation completion time, thereby realizing the scheduling process.

2. A fast evaluation and collaborative optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as claimed in claim 1, characterized in that: In step S2, the optimal path is calculated using the A* algorithm.

3. A fast evaluation and coordinated optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as claimed in claim 2, characterized in that: In step S2, the optimal path is calculated by combining a fast evaluation algorithm, and the steps of the fast evaluation algorithm are as follows: The theoretical value of the path length between two adjacent nodes is calculated according to the following formula, and the optimal path is calculated with the goal of minimizing the sum of the theoretical values ​​of the path lengths of all adjacent nodes; d t (A,B)=|x a -x b |+|y a -y b |+k p ×Error now +k i ×Error sum +k d ×Error delta Error sum +=Error now Error delta =Error now -Error last Among them, x a and a is the coordinate of node A, x b and b is the coordinate of node B, k p is the proportionality coefficient, k i is the integral term coefficient, k d is the coefficient of the differential term, Error now is a proportional term, Error sum is the integral term, Error delta is the differential term.

4. A fast evaluation and coordinated optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as claimed in claim 3, characterized in that: The actual value is the optimal transportation completion time calculated by the previous generation of bee colonies using the A* algorithm. The theoretical value The previous generation of swarms used a fast evaluation algorithm to calculate the optimal transport completion time.

5. A fast evaluation and coordinated optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as claimed in claim 4, characterized in that: The actual value is the average of the transport completion time of three individuals selected from the top 20% of the bee colony in the previous generation, which are sorted according to the shortest transport completion time, and the transport completion time of the first-ranked individual.

6. The method for rapid evaluation and coordinated optimization of workshop multi-load AGV scheduling and path planning as claimed in claim 1, characterized in that: In step S3, the individuals in the bee colony are divided into leading bees and following bees. When updating the bee colony, the leading bees adopt a global search algorithm, and the following bees adopt a neighborhood selection strategy for updating.

7. A fast evaluation and coordinated optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as claimed in claim 6, characterized in that: The neighborhood search strategy includes seven neighborhoods, one of which is selected to update the follower bee. The seven areas are as follows: Neighborhood 1: Randomly select two AGVs from an individual and exchange one task number in the codes of the two AGVs to form a new individual; Neighborhood 2: In an individual, select the AGV with the longest transportation completion time and exchange the task number on another AGV to form a new individual; Neighborhood 3: Insert the task sequence number of an AGV in an individual into the task sequence of another AGV to form a new individual; Neighborhood 4: In one individual, insert a task number in the task sequence of the AGV with the longest transportation completion time into another individual, and select the task sequence of the AGV with the longest transportation completion time in the individual to form a new individual; Neighborhood 5: In an individual, randomly select an AGV and exchange the two task numbers in the task sequence of the selected AGV to form a new individual; Neighborhood 6: In an individual, any two task numbers in the task sequence of the AGV with the longest transportation completion time are exchanged to form a new individual; Neighborhood 7: In an individual, randomly select an AGV's task sequence and use the distance-based task grouping method to update the task sequence to form a new individual.

8. A fast evaluation and coordinated optimization scheduling method for multi-load AGV scheduling and path planning in a workshop as claimed in claim 7, characterized in that: Choose one of the seven neighborhoods. The neighborhood is selected as follows: A random number p between (0, 1) is randomly generated. If the random number p is less than the preset threshold, a neighborhood update individual is randomly selected; otherwise, the selection probability of each neighborhood is calculated, and the neighborhood update individual with the largest probability is selected. The selection probability is calculated according to the following relationship: Among them, count g is the cumulative number of times the neighborhood is successfully selected, count l is the number of times the lth neighborhood is successfully selected.

9. The method for rapid evaluation and coordinated optimization of workshop multi-load AGV scheduling and path planning as claimed in claim 1, characterized in that: Each of the bodies corresponds to a sequence of tasks, and the task sequence is encoded using integers, coded ∈{1,2,…,n}.

10. A fast evaluation and collaborative optimization scheduling system for multi-load AGV scheduling and path planning in a workshop, characterized in that: The system includes an actuator, which is used to execute the rapid evaluation and collaborative optimization scheduling method for workshop multi-load AGV scheduling and path planning as described in any one of claims 1-9.

Images