A multi-capacity AMR scheduling method based on heuristic algorithm

Abstract

The present invention relates to the technical field of intelligent manufacturing equipment scheduling algorithms, and discloses a multi-load AMR scheduling method based on a heuristic algorithm, which aims to solve the problem of low efficiency in scheduling material handling tasks in intelligent manufacturing scenarios. This method first converts the unit platform and path into a weighted directed graph, with the weight being the AMR running time; secondly, the K-means algorithm is used to cluster the platforms to reduce the amount of calculation, and the cluster center is used as a candidate for the task starting point; then, a local optimal scheduling solution is generated based on a greedy algorithm, giving priority to the task with the shortest time consumption while satisfying the capacity constraint; finally, the total time consumption of different cluster center solutions is compared, and the global optimal solution is selected for output. This method can significantly shorten the task completion time, reduce path duplication and resource waste, improve the handling efficiency and loading rate of AMRs, and reduce energy consumption. It conforms to the development trend of green manufacturing, realizes efficient and reasonable scheduling, and is suitable for production and manufacturing scenarios with multi-machine linkage.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent manufacturing equipment scheduling algorithms, and specifically to a multi-load AMR scheduling method based on a heuristic algorithm. Background Art

[0002] With the rapid development of intelligent manufacturing and automation technologies, autonomous mobile robots (AMRs) are increasingly being used in industrial production. AMRs are highly intelligent mobile robotic systems that use sensors to perceive their environment, dynamically plan paths, and autonomously complete complex material handling tasks. Multi-load AMRs, a specialized form of AMR, possess a large carrying capacity and can simultaneously load and transport multiple pieces of cargo or heavy items. Therefore, they are widely used for material handling tasks between multiple machine stations in manufacturing scenarios.

[0003] In actual production environments, task scheduling for multi-load AMRs faces many challenges. Due to the diversity and dynamic nature of production tasks, the starting point, end point, type of material being handled, and task priority of each handling task may vary. In addition, each workstation may be both the starting point and the end point of a task, and task requirements are frequently adjusted due to changes in production plans. For example, when switching products or multiple machines are operating simultaneously, multi-load AMRs need to quickly respond to centralized material demands and reasonably allocate task sequences and path planning. However, traditional scheduling methods typically rely on simple rules and manual decision-making, and have the following problems:

[0004] 1. Unreasonable task assignment: Due to the lack of scientific optimization algorithms, traditional methods make it difficult to reasonably arrange tasks based on the temporal and spatial distribution of tasks and the load capacity of AMRs, resulting in delays or accumulation of some tasks.

[0005] 2. Path duplication and resource waste: Path planning that has not undergone global optimization can easily cause AMRs to run repeatedly in the same area, increasing idle driving rates and energy consumption.

[0006] 3. Low loading rate: The actual carrying capacity of multi-load AMRs is not fully utilized, reducing handling efficiency.

[0007] 4. Large amount of calculation: If all possible task combinations are calculated through exhaustive method to find the optimal solution, the calculation time is too long and cannot meet the needs of real-time scheduling. Summary of the Invention

[0008] To solve the above problems, the present invention proposes a multi-load AMR scheduling method based on a heuristic algorithm, which can reasonably arrange the starting point, sequence and path of each wave of tasks within a limited computing time, thereby reducing the AMR's travel distance, shortening the task completion time, and improving the handling efficiency and loading rate; it helps to improve production efficiency and reduce energy consumption, which is in line with the development trend of green manufacturing.

[0009] The technical solution adopted in the present invention is:

[0010] A multi-capacity AMR scheduling method based on a heuristic algorithm includes the following steps:

[0011] Step 1: Path modeling: Convert the plant platform and path into a weighted directed graph, with the weight being the AMR operating time, and dynamically calculated based on the acceleration and speed parameters.

[0012] Step 2: Platform clustering: Use the K-means algorithm to group the platforms by location to reduce computational complexity, and use the cluster center as a candidate for the task starting point.

[0013] Step 3: Greedy scheduling: Starting from each cluster center, a greedy algorithm is used to prioritize the task with the shortest time consumption, generating a local optimal scheduling solution that meets the capacity constraint.

[0014] Step 4: Global optimization: Compare the total time consumption of all cluster center solutions, select the optimal solution and output it to guide the AMR to perform the task, so as to optimize the path and efficiency.

[0015] Furthermore, in step 1, when converting the factory platforms and paths into a weighted directed graph, it is necessary to collect the coordinate data and AMR operation path information of all platforms in the multi-load AMR material handling work scene in the factory, and define the attributes of each platform; the attributes include the task starting point, end point, and material type.

[0016] Furthermore, in step 1, the plant platform and path are converted into a weighted directed graph, including the following steps:

[0017] Step 1.1: abstract the platforms as vertices of the graph and the AMR operation paths as edges;

[0018] In step 1.2, the edge weight is the time T that the AMR runs between two points, and the calculation formula is as follows:

[0019] When the AMR travel distance is greater than or equal to twice the distance required for the AMR to accelerate to its maximum speed:

[0020]

[0021] When the AMR travel distance is less than twice the distance required for the AMR to accelerate to its maximum speed:

[0022]

[0023] Where, The straight distance of the AMR handling task; is the operating speed of the AMR; is the running acceleration of AMR; is the number of AMR rotations; It is the single rotation time of AMR.

[0024] Furthermore, in step 2, the K-means algorithm is used to group the stations by location, including the following steps:

[0025] Step 2.1, initialize clustering parameters, input the coordinate set of all stations, the coordinate set of all stations is expressed as:

[0026]

[0027] Where, Represents the coordinate set of all stations; 、 、 is the horizontal coordinate of the platform; 、 、 is the vertical coordinate of the platform;

[0028] Collect the coordinates of all stations Divided into K categories, that is, containing K cluster centers, K Indicates the number of clusters divided, which is also the number of cluster centers; the cluster center is the starting point of the system operation;

[0029] Step 2.2: For the clustering task, the following model is established. The overall goal is to minimize the sum of squared Euclidean distances from each station to its cluster center.

[0030]

[0031] Where, is the objective function, i.e. the sum of squared errors within the cluster; is the number of clusters; is the total number of data points, i.e., the total number of handling task stations in the plant; The jth data point represents the coordinates of the jth station in the factory ; is the center of the ith cluster, which is the coordinate of the center point calculated by the mean of the coordinates of all stations in the cluster.

[0032] Furthermore, in step 2.1, the coordinates of all stations are collected Divided into Class, including K cluster centers, recorded as:

[0033]

[0034] Where, is the horizontal and vertical coordinates of the i-th initial cluster center; Indicates the preset number of clusters;

[0035] In step 2.2, calculate the data points for each station The distance to each cluster center is based on the principle that the distance from the cluster center to the points in the cluster is less than the distance from the cluster centers in other clusters, and all data in the data set are assigned to the corresponding clusters.

[0036] Each station data point Coordinates The distance to each cluster center is calculated as:

[0037]

[0038] Where, For AMR slave platform To the cluster center The Euclidean distance of is the horizontal coordinate of the j-th station; is the ordinate of the j-th station; is the horizontal coordinate of the i-th cluster center; is the ordinate of the i-th cluster center;

[0039] Then, calculate all stations in the same cluster The mean of the coordinate data is used as the new cluster center. Repeat the above steps until the value of the cluster center no longer changes. The clustering algorithm ends and all stations are divided into K clusters.

[0040] The K cluster centers of the K clusters serve as the initial candidate starting points for subsequent task assignment.

[0041] Furthermore, in step 3, the specific steps of the greedy algorithm are as follows:

[0042] Step 3.1 Initialize AMR status: Set the initial status of the multi-load AMR, including current position, current task status, and capacity;

[0043] The capacity constraint formula is:

[0044] ;

[0045] Where, is the total weight of the current batch of tasks; is the number of tasks, that is, the total number of tasks that the current AMR needs to perform; is the weight of the jth task, which indicates the weight of the material that the AMR needs to carry when performing the jth handling task; The maximum capacity of the AMR, indicating the maximum weight limit that a multi-capacity AMR can carry, is determined by the equipment specifications;

[0046] Step 3.2 Task time calculation and selection: Calculate the time required to reach the starting point of other tasks and the destination of the tasks it has carried based on the topology;

[0047] The calculation formula for task time is:

[0048]

[0049] Where, is the total task time, which indicates the total time required for the AMR to complete all tasks, including the time from the current position to each task point and the cumulative time of the path during the task execution; is the number of path nodes, which indicates the total number of path nodes that the AMR needs to pass through in the process of completing the task; is the i-th path node, indicating the i-th node position of the AMR during task execution; is the i+1th path node, indicating the next node position of the AMR during the task execution process; The time taken between two points, indicating that the AMR slave node Move to Node The time required is determined by the edge weights in the path topology graph;

[0050] Then, the task with the shortest execution time is selected as the next execution target and the AMR status is updated;

[0051] Step 3.3: Repeat task selection and execution until all tasks are completed or the AMR capacity is exhausted; record the total time corresponding to the cluster center. and task execution sequence.

[0052] Furthermore, in step 4, the total time consumption of all cluster center solutions is compared, that is, the total time consumption corresponding to all cluster centers is compared. , ,…, ; Select the optimal solution output, that is, select the minimum value:

[0053]

[0054] Where, is the optimal total time, which means the total time of the solution with the shortest time among all scheduling solutions, that is, the optimal solution finally selected; is the total time of the k-th plan, which means the total time required for the scheduling plan generated based on the k-th cluster center;

[0055] Output the total time of the optimal solution And the corresponding task execution sequence, AMR performs the handling task according to this sequence.

[0056] The beneficial effects of the present invention are:

[0057] This paper proposes a multi-capacity AMR scheduling method based on a heuristic algorithm. Through path modeling, platform clustering, greedy scheduling, and global optimization, it addresses the problems of irrational task assignment, path duplication, low loading rate, and high computational complexity in traditional scheduling methods. This method has the following significant benefits:

[0058] 1. Improved Task Scheduling Rationality: K-means clustering is used to group stations and generate locally optimal scheduling solutions based on the cluster center, avoiding the irrational task assignments associated with traditional manual decision-making. Combined with heuristic algorithms, this approach can quickly adapt to changes in production plans, such as product switching and task additions, achieving efficient task allocation.

[0059] 2. Reduce path duplication and resource waste: Using weighted directed graph modeling and a greedy algorithm, AMR routes are rationally planned, reducing empty trips and path duplication, thereby lowering energy consumption. The scheduling plans of different cluster centers are compared, and the one with the shortest total time is selected to ensure global optimality of path planning.

[0060] 3. Improve handling efficiency and loading rate: A capacity constraint formula ensures that the AMR's actual carrying capacity is fully utilized, avoiding inefficiencies caused by overloading or underloading. A greedy algorithm prioritizes the shortest tasks, significantly reducing AMR task execution time and improving overall handling efficiency.

[0061] 4. Reduce computational complexity to meet real-time requirements: K-means clustering is used to group stations, reducing the search space for task allocation and path planning, significantly reducing computational complexity. Using a greedy algorithm instead of an exhaustive search method significantly reduces computational time while ensuring scheduling quality, meeting real-time scheduling requirements.

[0062] 5. Complying with the development trend of green manufacturing: By optimizing path planning and reducing idle driving rates, AMR energy consumption is effectively reduced, complying with the requirements of green manufacturing and sustainable development. Efficient scheduling reduces the time cost of material handling, indirectly improving the overall efficiency of the production line.

[0063] 6. Scalability and versatility: This method is not only applicable to multi-machine material handling in intelligent manufacturing but can also be extended to other fields such as logistics warehousing and e-commerce sorting, demonstrating its broad applicability. The number of clusters and capacity constraints can be adjusted to meet specific needs, adapting to different production environments and equipment specifications.

[0064] In summary, the heuristic algorithm-based multi-capacity AMR scheduling method proposed in this paper, through scientific model construction and efficient algorithm design, has achieved remarkable results in rationalizing task allocation, optimizing routes, improving handling efficiency, and reducing computational complexity. Furthermore, its energy-saving and consumption-reducing features align with the development trend of green manufacturing, and possess significant application value and promotional significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0066] Figure 1 Flowchart of the multi-capacity AMR scheduling method based on the heuristic algorithm of the present invention;

[0067] Figure 2 For the present invention, all points and paths are converted into a weighted directed topological graph;

[0068] Figure 3 A flowchart of the workstation clustering method of the present invention;

[0069] Figure 4 This is a flowchart of the greedy algorithm of the present invention. DETAILED DESCRIPTION

[0070] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0071] In a manufacturing plant, there are z stations, with ≤ z stations operating during each shift each day. Raw materials, semi-finished products, and finished products need to be moved between stations, meaning each station either ships or receives materials. A multi-capacity AMR is deployed to accomplish this task. The multi-capacity AMR has a built-in robotic arm for grabbing and loading materials. Due to production task switching and machine switching, the starting point, end point, materials, task sequence, and route of each wave of handling tasks performed by the multi-capacity AMR may vary. Traditional scheduling methods execute tasks in the order they are generated. This can lead to circuitous routes and empty AMRs during operation, reducing handling efficiency.

[0072] Therefore, determining the order in which tasks are assigned is a technical problem that needs to be solved urgently. In addition, if all possible task combinations are calculated and then compared, with current technology, it will take a long time and cannot meet the timeliness requirements of task scheduling.

[0073] In order to solve the above problems, the present invention proposes a multi-load AMR scheduling method based on a heuristic algorithm to reasonably arrange the starting point, sequence, and route of each wave of tasks to improve the multi-load AMR handling efficiency. Figure 1 As shown, the multi-capacity AMR scheduling method based on the heuristic algorithm includes the following steps:

[0074] Step 1 Path Modeling:

[0075] The plant platforms and paths are converted into a weighted directed graph, with the weights representing the AMR runtime. Dynamic calculations are performed based on acceleration and velocity. This step aims to abstract the physical platform locations and AMR paths into a weighted directed graph, providing a foundational model for subsequent optimization.

[0076] Specifically, first establish the path topology:

[0077] Collect the coordinate data and AMR operation path information of all stations in the multi-load AMR material handling work scene in the production workshop, and define the attributes of each station; the attributes include the task start point, end point and material type.

[0078] Then, the workshop platform and path are transformed into a weighted directed graph:

[0079] Step 1.1, as Figure 2 As shown in the figure, the platform is abstracted as the vertex of the graph, and the number in the circle represents the number of the unit platform; the AMR operation path is abstracted as the edge, and the number on the circle line represents the distance between two nodes; after all points and paths are converted into a weighted directed topological graph, it is Figure 2 As shown;

[0080] Step 1.2, where the weight of the edge is the time T that the AMR runs between two points.

[0081] There are two cases for calculating the weight T of the edge of the topological graph. The specific methods are as follows:

[0082] When the AMR travel distance is greater than or equal to twice the distance required for the AMR to accelerate to its maximum speed:

[0083]

[0084] When the AMR travel distance is less than twice the distance required for the AMR to accelerate to its maximum speed:

[0085]

[0086] Where, The straight distance of the AMR handling task; is the operating speed of the AMR; is the running acceleration of AMR; is the number of AMR rotations; It is the single rotation time of AMR.

[0087] Assumption: There are 5 stations in the workshop, numbered 、 ,…, , whose coordinates are: , , , , ; AMR can run between any stations.

[0088] AMR performance parameters are: maximum operating speed , running acceleration , single rotation time ; Each path may contain several rotations. Assuming that each path rotates at most once, that is .

[0089] The straight distance between platforms Calculated by the Euclidean distance formula:

[0090]

[0091] For example, arrive The distance is: ;

[0092] Each station may be a starting point or a destination. Material handling tasks include raw materials, semi-finished products, finished products, etc. The specific weight is not considered here.

[0093] According to the assumed parameters, the coordinate set of the known stations is:

[0094]

[0095] These coordinates represent the physical locations of the stations in the factory; we assume that there are paths connecting all stations, i.e. a fully connected graph. The directionality of the paths is determined by the starting and ending points of the transport tasks. Each station may be both a starting point and an ending point. For example: It may be the starting point of raw materials or the end point of finished products; It may be a transfer station for semi-finished products. The task attributes of each station include the type of materials being transported, such as raw materials, semi-finished products, and finished products.

[0096] The platform is abstracted as the vertex of the graph, which is recorded as: ;Abstract the paths between platforms as edges of the graph, recorded as: ; The direction of the edge is determined by the starting and ending points of the task. For example: if the task is from arrive , then the edge is ( , ).

[0097] The edge weight T represents the time required for an AMR to travel from one station to another, and is calculated based on two scenarios:

[0098] calculate arrive Time: Distance traveled ; The required running distance at the highest speed is: Because of the walking distance , so use the first formula to calculate:

[0099]

[0100] Constructing a weighted directed graph ;

[0101] in, , , T is the weight of each edge.

[0102] Running distance required at maximum speed:

[0103] formula Indicates the minimum distance required for the AMR to decelerate after reaching its maximum speed.

[0104] when ≥ At a, the AMR can accelerate to the highest speed and then decelerate.

[0105] when < When the AMR cannot reach the maximum speed, it runs directly according to the acceleration and deceleration curve.

[0106] Each spin adds a fixed amount of time , need to be adjusted according to the actual path .

[0107] Through the above steps, a weighted directed topology graph within the factory was successfully established. This graph, with platforms as vertices, paths as edges, and edge weights representing AMR runtimes, accurately reflects the operating costs of AMRs between different platforms. This topology graph provides a foundational model for subsequent cluster analysis and task scheduling, ensuring the scientific and operational nature of the scheduling plan.

[0108] Step 2: Platform clustering:

[0109] Use the K-means algorithm to group stations by location to reduce computational complexity, and use the cluster centers as candidate starting points for tasks. The purpose of this step is to reduce computational complexity through clustering and improve scheduling efficiency.

[0110] like Figure 3 As shown in the figure, the K-means algorithm is used to group the stations by location, which includes the following steps:

[0111] Step 2.1, initialize clustering parameters:

[0112] Input the coordinate set of all stations. The coordinate set of all stations is expressed as:

[0113]

[0114] Where, Represents the coordinate set of all stations; 、 、 is the horizontal coordinate of the platform; 、 、 is the vertical coordinate of the platform.

[0115] Collect the coordinates of all stations Divided into Class, that is, contains K cluster centers, K It indicates the number of clusters divided, and also the number of cluster centers; the cluster center is the starting point of the system operation.

[0116] Collect the coordinates of all stations Divided into Class, including K cluster centers, recorded as:

[0117]

[0118] Where, is the horizontal and vertical coordinates of the i-th initial cluster center; Indicates the preset number of clusters.

[0119] Step 2.2, for clustering task:

[0120] The following model is established, and the overall goal is to minimize the sum of squared Euclidean distances from each station to its cluster center;

[0121]

[0122] Where, is the objective function, i.e. the sum of squared errors within the cluster; is the number of clusters; is the total number of data points, i.e., the total number of handling task stations in the plant; The jth data point represents the coordinates of the jth station in the factory ; is the center of the ith cluster, which is the coordinate of the center point calculated by the mean of the coordinates of all stations in the cluster.

[0123] Iterative optimization clustering results:

[0124] Calculate the data points for each station The distance to each cluster center is based on the principle that the distance from the cluster center to the points in the cluster is less than the distance from the cluster centers in other clusters, and all data in the data set are assigned to the corresponding clusters.

[0125] Each station data point Coordinates The distance to each cluster center is calculated as:

[0126]

[0127] Where, For AMR slave platform To the cluster center The Euclidean distance of is the horizontal coordinate of the j-th station; is the ordinate of the j-th station; is the horizontal coordinate of the i-th cluster center; is the vertical coordinate of the i-th cluster center.

[0128] Then, calculate all stations in the same cluster The mean of the coordinate data is used as the new cluster center; the above steps are repeated until the value of the cluster center no longer changes, the clustering algorithm ends, and all stations are divided into K clusters.

[0129] Output clustering results:

[0130] The K cluster centers of the K clusters serve as the initial candidate starting points for subsequent task assignment.

[0131] Assumption: There are 5 stations in the factory, numbered 、 ,…, , whose coordinates are: , , , , ; then set the number of clusters , that is, dividing the platform into 2 clusters.

[0132] The K-means clustering operation steps are as follows:

[0133] Random selection Initial cluster centers: Assume that the two randomly selected initial cluster centers are: , corresponding to the platform ; , corresponding to the platform .

[0134] The initial cluster center set is: ;

[0135] Calculate the distance from each station to the two cluster centers:

[0136] ; ;

[0137] Therefore, the platform Distance from initial cluster center If the station is closer, it is assigned to cluster 1; similarly, the distances of other stations are calculated and assigned to the nearest cluster.

[0138] The allocation results are as follows:

[0139] First round allocation results: Cluster 1: , , Cluster 2: , .

[0140] Update cluster centers:

[0141] For cluster 1 contains , , :Then the mean is calculated as:

[0142]

[0143] For cluster 2, , :Then the mean is calculated as:

[0144]

[0145] The new set of cluster centers is: ;

[0146] Calculate the distance from each station to the new cluster center again and redistribute the stations;

[0147] for :

[0148] ; ;

[0149] Therefore, the station is still assigned to cluster 1; similarly, the other stations are reallocated.

[0150] If the new cluster center is the same as the previous round, the algorithm converges; otherwise, continue to iterate. After several rounds of iteration, the final clustering result is:

[0151] Cluster 1: , , , cluster center Cluster 2: , , cluster center .

[0152] Through the above steps, K-means clustering was successfully completed, dividing the platform into two clusters and determining the cluster centers for each cluster. These cluster centers can be used as starting points for subsequent task scheduling, significantly reducing computational complexity. This process demonstrates the effectiveness of data dimensionality reduction and improved computational efficiency, laying the foundation for efficient scheduling of multi-capacity AMRs.

[0153] Step 3 Greedy Scheduling:

[0154] Taking each cluster center as the starting point, the greedy algorithm prioritizes the shortest-time tasks to generate a local optimal scheduling solution that meets the capacity constraint. The purpose of this step is to generate a local optimal task scheduling solution for each cluster center as the starting point.

[0155] like Figure 4 As shown, the specific steps of the greedy algorithm are as follows:

[0156] Step 3.1 Initialize AMR status: Set the initial status of the multi-load AMR, including current position, current task status, and capacity;

[0157] The capacity constraint formula is:

[0158]

[0159] Where, is the total weight of the current task; is the number of tasks, that is, the total number of tasks that the current AMR needs to perform; is the weight of the jth task, which indicates the weight of the material that the AMR needs to carry when performing the jth handling task; The maximum capacity of the AMR indicates the maximum weight limit that the multi-capacity AMR can carry, which is determined by the equipment specifications.

[0160] Step 3.2 Task time calculation and selection: Calculate the time required to reach the starting point of other tasks and the destination of the tasks it has carried based on the topology;

[0161] The calculation formula for task time is:

[0162]

[0163] Where, is the total task time, which indicates the total time required for the AMR to complete all tasks, including the time from the current position to each task point and the cumulative time of the path during the task execution; is the number of path nodes, which indicates the total number of path nodes that the AMR needs to pass through in the process of completing the task; is the i-th path node, indicating the i-th node position of the AMR during task execution; is the i+1th path node, indicating the next node position of the AMR during the task execution process; The time taken between two points, indicating that the AMR slave node Move to Node The time required is determined by the edge weights in the path topology graph.

[0164] Then, the task with the shortest execution time is selected as the next execution target and the AMR status is updated.

[0165] Step 3.3: Repeat task selection and execution until all tasks are completed or the AMR capacity is exhausted; record the total time corresponding to the cluster center. and task execution sequence.

[0166] Assumption: AMR maximum capacity The mission weights of each platform are: , , , , ; The calculation of path weight T will not be repeated here.

[0167] First, the cluster center of cluster 1 and the cluster center of cluster 2 As the initial positions of the two scheduling schemes respectively. For each scheme, the initial state of AMR is set as: current position: cluster center position; current task state: idle; current load weight: .

[0168] The initial task list is a collection of tasks for all stations: ; Use the path weight formula to calculate the AMR from the current position To the destination station Time .

[0169] Such as: from arrive The Euclidean distance is: ;

[0170] because Greater than twice the distance the AMR needs to travel to accelerate to its maximum speed: ; Use the first formula to calculate the running time , that is, from arrive The running time is 9.27 seconds. Then, the time from the current cluster center to all stations is calculated and a timetable is formed.

[0171] Initialize the AMR status:

[0172] Set the initial state:

[0173] Current location: Cluster center ; Current mission status: Idle; Current load weight: . Every time you select a task, make sure to add the weight of the task Not exceeding the maximum capacity of the AMR .

[0174] According to the formula Calculate the total time required for the AMR to complete the current task sequence. Select the task with the shortest time from the current task list as the next execution target. For example: Departure, the time to each station is calculated as: , , ;choose As the next task.

[0175] Update the status of the AMR:

[0176] Current Location: ; Current load capacity: .

[0177] Continue to select the shortest tasks from the remaining task list until one of the following conditions is met: all tasks are completed; the AMR reaches its maximum carrying capacity .

[0178] Each time you select a task, record the task number and its corresponding execution order. For example: Task sequence: ; Accumulate the time required for AMR to complete all tasks .

[0179] Repeat the above steps: and As the starting point, calculate the total time of the two solutions and .Compare and , select the solution with the shortest total time; if , , then choose option 2, As the starting point.

[0180] Through the above steps, a greedy algorithm was successfully used to calculate scheduling solutions from different cluster centers. Ultimately, the solution with the shortest total time was selected as the optimal solution, and the task execution sequence was recorded. This process embodies the principle of local optimization and can quickly generate efficient scheduling solutions within a limited time, providing a scientific basis for the practical application of multi-capacity AMRs.

[0181] Step 4 Global Optimization:

[0182] Compare the total time consumption of all cluster center solutions, select the optimal solution and output it to guide the AMR to execute the task, so as to optimize the path and efficiency. The purpose of this step is to select the global optimal solution by comparing the scheduling solutions of different cluster centers.

[0183] Specifically, compare the total time consumption of all cluster center solutions, that is, compare the total time consumption corresponding to all cluster centers , ,…, ; Select the optimal solution output, that is, select the minimum value:

[0184]

[0185] Where, is the optimal total time, which means the total time of the solution with the shortest time among all scheduling solutions, that is, the optimal solution finally selected; is the total time of the k-th plan, which means the total time required for the scheduling plan generated based on the k-th cluster center.

[0186] Output the total time of the optimal solution And the corresponding task execution sequence, AMR performs the handling task according to this sequence to achieve the goals of shortest path, least time and highest loading rate.

[0187] Assumption: The calculation result of the greedy algorithm is:

[0188] by The total time of the plan starting from , the task sequence is ;by The total time of the plan starting from , the task sequence is .

[0189] Select the solution with the shortest total time: ; The optimal solution is solution 2, and the corresponding total time is ;

[0190] Output the optimal transport solution: The task sequence of the optimal solution is Total time: 45s, starting position: Multi-load AMRs follow the optimal task sequence Carry out the transportation task, from the starting point Start and complete and transportation tasks.

[0191] Through the above steps, the optimal scheduling solutions were successfully compared and output. Ultimately, solution 2, with the shortest total time, was selected, and its total time and task execution sequence were output. This process embodies the principle of global optimization, ensuring that the AMR can complete all tasks in the shortest time while meeting path planning and capacity constraints. This method provides a scientific and efficient scheduling basis for the practical application of multi-capacity AMRs.

[0192] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A multi-capacity AMR scheduling method based on a heuristic algorithm, characterized in that: The multi-capacity AMR scheduling method based on a heuristic algorithm includes the following steps: Step 1: Path modeling: Convert the unit platform and path into a weighted directed graph, with the weight being the AMR operating time, and dynamically calculated based on the acceleration and speed parameters; Step 2: Platform clustering: Use the K-means algorithm to group the platforms by location to reduce the amount of computation, and use the cluster center as a candidate for the task starting point. Step 3: Greedy scheduling: Starting from each cluster center, a greedy algorithm is used to prioritize the task with the shortest time consumption, generating a local optimal scheduling solution that meets the capacity constraint. The specific steps of the greedy algorithm are as follows: Step 3.1 Initialize AMR status: Set the initial status of the multi-load AMR, including current position, current task status, and capacity; Step 3.2 Task Time Calculation and Selection: Based on the topology, the AMR calculates the time it takes to reach the starting points of other tasks and the destinations of the tasks it has already carried. Then, the task with the shortest time is selected and executed, updating the AMR status. Step 3.3: Repeat the task selection and execution until all tasks are completed or the AMR capacity is exhausted; record the total time T corresponding to the cluster center. k and task execution sequence; Step 4: Global optimization: Compare the total time consumption of all cluster center solutions, select the optimal solution and output it to guide the AMR to perform the task, so as to optimize the path and efficiency.

2. The multi-capacity AMR scheduling method based on a heuristic algorithm according to claim 1, characterized in that: Step 1: During the process of converting the unit platforms and paths into a weighted directed graph, it is necessary to collect the coordinate data and AMR operation path information of all platforms in the multi-load AMR material handling work scene in the production workshop, and define the attributes of each platform; the attributes include the task start point, end point and material type.

3. The multi-capacity AMR scheduling method based on a heuristic algorithm according to claim 1, characterized in that: In step 1, the unit platforms and paths are converted into a weighted directed graph, which includes the following steps: Step 1.1: abstract the platforms as vertices of the graph and the AMR operation paths as edges; In step 1.2, the edge weight is the time T that the AMR runs between two points, and the calculation formula is as follows: When the AMR travel distance is greater than or equal to twice the distance required for the AMR to accelerate to its maximum speed: When the AMR travel distance is less than twice the distance required for the AMR to accelerate to its maximum speed: Where s is the straight distance of the AMR handling task; v is the running speed of the AMR; a is the running acceleration of the AMR; n rotate is the number of AMR rotations; t ratate It is the single rotation time of AMR.

4. The multi-capacity AMR scheduling method based on a heuristic algorithm according to claim 1, characterized in that: In step 2, the K-means algorithm is used to group the stations by location, which includes the following steps: Step 2.1, initialize clustering parameters, input the coordinate set of all stations, the coordinate set of all stations is expressed as: X={(x1,y1),(x2,y2),...,(x m ,y m )} Where X represents the coordinate set of all stations; x1, x2, x m is the horizontal coordinate of the platform; y1, y2, y m is the vertical coordinate of the platform; Divide the coordinate set X of all stations into K categories, that is, it contains K cluster centers, where K represents the number of clusters divided, and also the number of cluster centers; the cluster center is the starting point of the system operation; Step 2.2: For the clustering task, the following model is established. The overall goal is to minimize the sum of squared Euclidean distances from each station to its cluster center. Where J is the objective function, i.e., the sum of squared errors within a cluster; K is the number of clusters; m is the total number of data points, i.e., the total number of handling task stations in the production workshop; x j is the jth data point, representing the coordinate x of the jth station in the plant j =(x j ,y j ); μ i is the center of the ith cluster, which is the coordinate of the center point calculated by the mean of the coordinates of all stations in the cluster.

5. The multi-capacity AMR scheduling method based on heuristic algorithm according to claim 4, characterized in that: In step 2.1, the coordinate set X of all stations is divided into k categories, including k cluster centers, which are recorded as: <h2 style=";text-align:left;direction:ltr">(a1,b1),(a2,b2),...,(a<h2 style=";text-align:left;direction:ltr"> k <h2 style=";text-align:left;direction:ltr"> ,b<h2 style=";text-align:left;direction:ltr"> k <h2 style=";text-align:left;direction:ltr"> ) Where a i ,b i is the horizontal and vertical coordinates of the i-th initial cluster center; k represents the preset number of clusters; In step 2.2, calculate the data point x for each station j The distance to each cluster center is based on the principle that the distance from the cluster center to the points in the cluster is less than the distance from the cluster centers in other clusters, and all data in the data set are assigned to the corresponding cluster; Each station data point x j The coordinates (x j ,y j ) to each cluster center is calculated as: Where, d ij For AMR slave station x j To cluster center k i The Euclidean distance of x j is the horizontal coordinate of the jth station; y j is the vertical coordinate of the jth station; a i is the horizontal coordinate of the i-th cluster center; b i is the ordinate of the i-th cluster center; Then, calculate all stations x in the same cluster j The mean of the coordinate data is used as the new cluster center. Repeat the above steps until the value of the cluster center no longer changes. The clustering algorithm ends and all stations are divided into K clusters. The K cluster centers of the K clusters serve as the initial candidate starting points for subsequent task assignment.

6. The multi-capacity AMR scheduling method based on heuristic algorithm according to claim 1, characterized in that: In step 3, the capacity constraint formula is: Where W is the total weight of the current batch of tasks; m is the number of tasks, that is, the total number of tasks that the current AMR needs to perform; w j is the weight of the j-th task, indicating the weight of the material that the AMR needs to carry when performing the j-th handling task; η is the maximum capacity of the AMR, indicating the maximum weight limit that a multi-load AMR can carry, which is determined by the equipment specifications; The calculation formula for task time is: Where T is the total task time, which indicates the total time required for the AMR to complete all tasks, including the time from the current position to each task point and the cumulative time of the path during the task execution; K is the number of path nodes, which indicates the total number of path nodes that the AMR needs to pass through in the process of completing the task; p i is the i-th path node, indicating the i-th node position of AMR during task execution; p i+1 is the i+1th path node, indicating the next node position of the AMR during the task execution process; t(p i ,p i+1 ) is the time taken between two points, indicating that AMR starts from node p i Move to node p i+1 The time required is determined by the edge weights in the path topology graph.

7. The multi-capacity AMR scheduling method based on heuristic algorithm according to claim 1, characterized in that: In step 4, the total time consumption of all cluster center solutions is compared, that is, the total time consumption T1, T2, ..., T corresponding to all cluster centers is compared. k ; Select the optimal solution output, that is, select the minimum value: Z=min(T1,T2,...,T k ) Where Z is the optimal total time, which means the total time of the shortest scheduling solution among all the scheduling solutions, that is, the optimal solution finally selected; T k is the total time of the k-th plan, which means the total time required for the scheduling plan generated based on the k-th cluster center; Output the total time Z of the optimal solution and the corresponding task execution sequence, and AMR performs the handling task according to this sequence.

Images