Workshop material distribution task dynamic acquisition method based on side cloud collaboration

Abstract

In order to overcome the defects of the existing workshop material task distribution system / method, the invention provides a workshop material distribution task dynamic acquisition method based on side cloud collaboration. According to the method, on the basis of dynamically sensed carrying carrier state information and to-be-distributed task state information, a CTB active obtaining strategy suitable for being used in the whole distribution process is designed, multi-target optimization is carried out through an NSGA-II algorithm in combination with the state information of the current carrying carrier and a dynamic information model of the distribution task, and the distribution task state information of the current carrying carrier is obtained. And obtaining a pareto solution set to determine a distribution task set scheme, and realizing a function of actively obtaining a new task which is on the way and is proper in load while a carrying carrier walks under the conditions of meeting a time window, capacity and electric quantity. According to the method, the problems of relatively high no-load rate, relatively long optimization time, relatively long information response time and the like of an original distribution method are solved, the efficiency and the flexibility of workshop distribution are improved, and the load of a distribution center is relieved.

Description

Technical Field

[0001] The invention relates to a method for acquiring workshop material distribution tasks. Background Art

[0002] In recent years, with the rapid development of technologies such as industrial Internet, big data, sensor networks, cloud computing and micro-electromechanical systems, the R&D model and production process of manufacturing enterprises have undergone major changes, which have put forward higher requirements for real-time allocation and dynamic optimization in the material transportation process. The traditional centralized task allocation model has gradually been unable to meet the rapidly changing needs.

[0003] At present, the research on workshop material task distribution system / method mainly includes:

[0004] Diego Augusto de Jesus Pacheco and others from Aarhus University used RFID technology to monitor and optimize the control system of the production line online, which helps to control and optimize the production process.

[0005] The patent document with publication number CN110018672A discloses a dynamic material distribution system and method for discrete manufacturing workshops based on an intelligent perception network. The dynamic material distribution system for discrete manufacturing workshops includes an intelligent perception platform, a data processing module, an application service module, and a PC terminal. The intelligent perception platform collects the material information of the workshop, the remaining amount of materials in the workstation buffer area, the consumption of materials in the workstation buffer area per unit time, and the location information of material tools, distribution personnel, and distribution vehicles in real time, and transmits the above-collected information to the data processing module; the data processing module converts the received data into a format recognizable by the application service module and transmits it to the application service module; the application service module transmits the received information to the PC terminal of the manager, and uses the improved ant colony algorithm to calculate a set of optimal distribution paths. This solution can reduce the uncertainties in the material distribution process, optimize the responsiveness of material demand, and obtain a better distribution scheduling method than the traditional scheduling method.

[0006] Wu Lu and others from Chongqing University proposed an active distribution model centered on workstation groups in workshop material supermarkets. They analyzed the active distribution methods by dividing material characteristics and matching distribution strategies, in order to achieve active and timely distribution to demand workstations and improve timely control of the material distribution process.

[0007] Dong Yuxuan and others from Shanghai Jiao Tong University proposed a collaborative optimization method for material distribution. By establishing an information model of personalized orders, the materials required for assembling each order are decomposed into multiple material distribution operation processes performed by feeding robots, thereby realizing the allocation of material distribution tasks in the assembly line scenario.

[0008] The above research has a good effect on promoting the timely delivery of material transportation tasks, but there are still the following shortcomings:

[0009] 1) The current allocation mode of material distribution tasks is mostly based on the data of material transportation tasks to be completed, handling staff and handling carriers. The material tasks are allocated to appropriate handling staff or handling carriers with the goal of minimizing the total time or the shortest path of the material distribution system. However, when there are more distribution tasks, handling staff and handling carriers, the centralized or central task allocation method is difficult to handle the resulting NP-hard problems. In addition, there are also problems such as long optimization time, difficulty in finding a satisfactory solution efficiently, and difficulty in dealing with transportation paralysis caused by emergencies.

[0010] 2) During the execution of material distribution tasks, the status information of transport carriers (such as carts, AGVs, forklifts, etc.) is usually only considered at the initial allocation, ignoring their real-time status information during transportation. As a result, the transport carriers may have a long period of excess load during the distribution task, wasting transportation resources.

[0011] 3) As the central group for task allocation, there may be an information bottleneck problem, that is, all task and resource information needs to be transmitted to the central system for processing and decision-making, which may lead to bottlenecks and delays in information transmission and increase the burden of communication and processing.

[0012] 4) The current material distribution models and methods rarely involve the optimization of active task acquisition decisions during the transportation process. However, in actual implementation, we found that matching the task path and current remaining load of the carrier during the transportation task with the information of the tasks in the task pool, and supporting the active acquisition decision of the tasks to be delivered on this basis, has an important impact on improving the efficiency of material distribution and reducing the cost of material handling. Summary of the invention

[0013] Based on the above background technology, in order to overcome the shortcomings of the existing workshop material task distribution system / method, the present invention provides a dynamic acquisition method of workshop material distribution tasks based on edge-cloud collaboration.

[0014] The technical solution of the present invention is:

[0015] A method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration includes the following steps:

[0016] Step 1: Construct a transport carrier with information acquisition capabilities;

[0017] Step 2: Acquire the status information of the transport carrier and the task pool in real time, and establish an information model of the distribution process based on the status information, wherein the information model includes the information matrix of each transport carrier and the information matrix of the task pool; the status information of the transport carrier includes the number of the transport carrier, the maximum load weight, the real-time position and the load weight used in real time; the status information of the task pool includes the number of the task, the starting point coordinates, the time window, the load weight and the status;

[0018] Step 3: For any transport carrier numbered j, based on the task priority and the load weight capacity of the transport carrier, generate a set of N advance delivery tasks for it;

[0019] Step 3.1: According to the information model, determine whether the transport carrier numbered j is currently executing a task. If not, proceed to step 3.2; if yes, proceed to step 5;

[0020] Step 3.2 The transport carrier numbered j senses the priority of the delivery tasks to be executed in the task pool, sorts the delivery tasks to be executed according to the priority, and generates a task sequence list;

[0021] Step 3.3 The transport carrier numbered j establishes a constraint condition for the combination of delivery tasks related to its real-time load weight capacity, and according to the constraint condition, the delivery tasks are cumulatively intercepted from the task sequence list in descending order of priority, and the intercepted delivery tasks constitute the prior delivery task set N, and then proceed to step 4;

[0022] Step 4: Use the A* algorithm to plan the optimal path for the current delivery task set N obtained by the transport carrier numbered j;

[0023] Step 5: Determine whether the transport carrier numbered j currently has idle load weight. If so, the task set currently executed by the transport carrier is used as its prior delivery task set N, and the process goes to step 6. If not, the transport carrier numbered j continues to execute the current task.

[0024] Step 6: Establish a multi-objective function with delivery path overlap, detour distance and overall load rate as the target, screen the task pool based on the information model, use the NSGA-Ⅱ algorithm to optimize the multi-objective function, and generate a Pareto frontier solution set;

[0025] Step 7: Select a distribution task set N' from the Pareto frontier solution set according to the material distribution scenario, priority, degree of convenience, detour distance and weight requirements of load rate;

[0026] Step 8: Update the status information of the corresponding delivery tasks in the task pool according to the delivery task set N'.

[0027] Furthermore, step 1 is specifically as follows:

[0028] By installing sensors, software systems and display devices on the transport carrier, and configuring RFID electronic tags at key intersections, entrances and exits, and material starting points in the workshop / warehouse, the transport carrier can have real-time status perception capabilities, including the current position of the transport carrier, the remaining load weight, the path information of the distribution task set, the attributes of the tasks to be picked up, and the communication status; the attributes of the tasks to be picked up include the task number, material location coordinates, material weight, and whether the attributes have been picked up;

[0029] The software system dynamically interacts based on the wireless network environment to obtain the attributes of the tasks to be accepted in the task pool, and relies on the transport carrier's own edge computing capabilities to match the obtained task dynamic attribute information and the transport carrier's real-time dynamic information, so as to actively obtain the new material distribution tasks that are most suitable for the transport carrier to perform, and update and display the results on the display device.

[0030] Furthermore, in step 2:

[0031] The information matrix of each transport carrier is:

[0032]

[0033] Among them, AID j represents the number of the j-th transport carrier; represents the maximum load weight of the jth transport carrier; CO j Indicates the real-time position of the jth transport carrier; represents the load weight used by the j-th transport carrier in real time;

[0034] The information matrix of the task pool is:

[0035]

[0036] Among them, TID k Indicates the number of the delivery task k; STA k Indicates the starting coordinates of the delivery task k; END k represents the destination coordinates of the delivery task k; Q k represents the load weight of delivery task k; LT k represents the earliest start time of task k; RT k Indicates the latest completion time of task k.

[0037] Furthermore, in step 4:

[0038] When the A* algorithm is used to perform path planning for each task in the prior delivery task set N currently obtained by the transport carrier numbered j, the shortest path is used as the objective function, and the shortest path from the starting point to the delivery destination of each delivery task is found by selecting and expanding nodes and performing path selection according to the heuristic function. Finally, the path planning scheme with the shortest total delivery is selected as the optimal path planning result for the prior delivery task set N.

[0039] Furthermore, step 6 is specifically as follows:

[0040] Step 6.1 Establish multi-objective function:

[0041]

[0042]

[0043]

[0044] Step 6.2 Establish the constraints of the multi-objective function:

[0045]

[0046] LT i ≤s i ≤RT i -d i (5)

[0047] s i +d i ≤c i ≤RT i (6)

[0048] LT j ≤s j ≤RT j -d j (7)

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] Among them, N is the first delivery task set; q i Represents the material weight of the i-th delivery task; represents the maximum load weight of the transport carrier numbered j; W is the dynamically updated task pool, including all current delivery tasks; TID i and TID j Represents task i and task j respectively; LT i and LT j Respectively represent the earliest start time of task i and task j; RT i and RT j Respectively represent the latest end time of task i and task j; s i and j They represent the calculated actual start time of task i and task j respectively; c i and c j They represent the actual end time of task i and task j respectively; d i and d j Represent the duration of task i and task j respectively; d k represents the duration of task k; RT k represents the latest completion time of task k; v represents the speed of the transport carrier; S represents the first end time of the transport carrier when performing the distribution concentration task; Path(TID i ) indicates the task TID i Location; Dist(Path(TID i ),Path(TID j )) indicates task TID i and TID j Dist(Path) represents the total distance of the path; ADD represents the allowed detour distance; Tat represents the total allowed time to complete all tasks; t l represents the material loading and unloading time of the distribution task; h represents the power consumption rate when the carrier is empty; δ represents the power consumption increment per unit time when each unit weight of material is added to the carrier; E N represents the total power consumption of the delivery task set N of the transport carrier; E j Indicates the current remaining power of the transport carrier; E b Indicates the actual remaining power of the transport carrier; D N Indicates the time required to reach the charging station after completing all tasks in the current task set; Z is an integer; Q i,i+1 represents the load weight of the carrier from coordinate i to coordinate i+1; T i,i+1 Indicates the time taken to transport the carrier from coordinate i to coordinate i+1;

[0055] Step 6.3 uses the NSGA-Ⅱ algorithm to optimize the multi-objective function and obtain the Pareto frontier solution set.

[0056] Furthermore, step 6.3 is specifically as follows:

[0057] Step 6.3.1 encodes the delivery task and generates the initial population;

[0058] Step 6.3.2: judging and screening the individuals according to the constraints of step 6.2, retaining the individuals that meet all the constraints and bringing them into the multi-objective function to calculate the function value;

[0059] Step 6.3.3 For the individuals in the initial population generated in step 6.3.1, perform non-dominated sorting according to the objective function value calculated in step 6.3.2 to determine the non-dominated level of each individual;

[0060] Step 6.3.4 For individuals of the same level, use the crowding operator to calculate their crowding values, and determine the comprehensive crowding of individuals by calculating the proximity distance of individuals in each objective function dimension;

[0061] Step 6.3.5: select the next generation of individuals according to the non-dominated ranking and the comprehensive crowding degree, and then operate them by crossover and mutation, merge the parent generation and the offspring generation, select the top N individuals based on the non-dominated ranking and crowding degree distance, and form the next generation of new population;

[0062] Step 6.3.6 returns the new population generated in step 6.3.5 to step 6.3.2 for multiple iterations until the maximum number of iterations is reached and the iteration is stopped. The final population is the Pareto frontier solution set.

[0063] The beneficial effects of the present invention are:

[0064] 1. The present invention designs a method for dynamically acquiring workshop material distribution tasks in an edge-cloud collaborative manner. Based on the status information of dynamically perceived transport carriers (such as carts, AGVs, forklifts, etc.) and the status information of the tasks to be distributed, a CTB (Concurrent Task-Based) active acquisition strategy suitable for the entire distribution process is designed. Combined with the status information of the current transport carrier and the dynamic information model of the distribution task, the NSGA-Ⅱ algorithm is used for multi-objective optimization to obtain the Pareto solution set to determine the distribution task set scheme, so that the transport carrier can actively acquire tasks along the way during the distribution process to implement the CTB strategy, so that the transport carrier can actively acquire new tasks along the way and with suitable load while walking under the conditions of meeting the time window, capacity and power during the material distribution process. Compared with the traditional static task allocation method, the present invention solves the problems of high empty rate, long optimization time and long information response time of the original allocation method, improves the efficiency and flexibility of workshop distribution, alleviates the load of the distribution center, and can better realize digital precision distribution and intelligent material distribution management. Since the present invention is optimized based on the real-time status of transport carriers and tasks, the centralized or central static global scheduling and optimization is transformed into a dynamic small-scale optimization. Therefore, the present invention solves the NP-hard problem caused by the difficulty of centralized or central task allocation when there are many delivery tasks, transport workers and transport carriers.

[0065] 2. The present invention proposes a strategy for actively acquiring tasks for transport carriers during the distribution process. This strategy can flexibly respond to dynamic distribution needs, and can help transport carriers actively select suitable tasks when the transport carriers have spare loads, thereby reducing the empty load rate of the transport carriers, increasing the flexibility of material distribution, and adapting to dynamic production environment changes in the enterprise.

[0066] 3. The present invention takes into account the status information of the transport carrier during the delivery process and the parameter information of the tasks to be accepted into the existing model, providing important information for the transport carrier to actively select appropriate tasks during the delivery process, and assisting the transport carrier to actively select a more convenient set of delivery tasks, thereby improving the reliability of the delivery process.

[0067] 4. The present invention relies on the edge computing capabilities of the transport carrier itself to match the acquired task dynamic attribute information and the real-time dynamic information of the transport carrier to actively acquire new material distribution tasks. This method combines the concept of edge cloud and pushes computing and decision-making capabilities to the edge of network nodes, making computing resources closer to data sources, reducing delays and network congestion in the distribution process, thereby achieving fast and dynamic active task acquisition and route optimization, speeding up response time and reducing dependence on cloud servers, and improving the efficiency and accuracy of material distribution.

[0068] 5. The present invention improves the coding, crossover and inheritance through the NSGA-Ⅱ algorithm to adapt to the current scenario of material handling by the transport carrier. The coding uses a hybrid coding method combining integer coding and real number coding, that is, the coordinates, number, starting point mark, load weight, left and right time windows of the task are mixed with Boolean values ​​for coding. The crossover operation is to directly exchange the order of the coordinates within the individual to indicate the order of completing the tasks. The mutation operation is to exchange other tasks of the non-original task in the new individual with the task to be accepted in the task pool in a random manner to achieve the mutation operation. The modification of the coding operation is to split the starting point coordinates of the task to increase the arrangement of the task order in the case of multiple tasks, and achieve the goal of "receiving new tasks while walking" while performing path planning. The crossover and mutation operations are to generate and select new populations by exchanging the position order under the condition of satisfying the model, so as to achieve path planning, task selection, and improve the efficiency and accuracy of task allocation, and can better handle dynamic and complex material distribution tasks. Through this algorithm, the transport carrier can actively obtain the most suitable task according to various constraints and optimization goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of the method of the present invention.

[0070] Figure 2 It is a flow chart of the A* algorithm used in the present invention.

[0071] Figure 3 is a flow chart of the NSGA-2 algorithm used in the present invention.

[0072] Figure 4 It is a task scheduling flow chart based on the CBT decision method of the present invention.

[0073] Figure 5 It is a distribution schematic diagram based on the present invention considering the CBT strategy in the distribution process. DETAILED DESCRIPTION

[0074] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0075] Reference Figure 1-5 The method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration provided by the present invention includes the following steps:

[0076] Step 1: Construct a transport vehicle with information acquisition capabilities.

[0077] Transport carriers (such as carts, AGVs, and forklifts) are the basic execution units in the material distribution process of workshops. By setting sensors (such as RFID readers, indoor positioning modules), software systems, and display devices on transport carriers, and configuring RFID electronic tags at key intersections, imports and exports, and material starting points (manufacturing equipment) in workshops / warehouses, transport carriers can be equipped with real-time status perception capabilities, including the current position of the transport carrier, the remaining load weight of the transport carrier, the path information of the distribution task set, the attributes of the tasks to be picked up, and the communication status; the attributes of the tasks to be picked up include the task number, material location coordinates, material weight, and whether the attribute information has been received.

[0078] Among them, the software system dynamically interacts based on the wireless network environment to obtain the attributes of the tasks to be accepted in the task pool W, and relies on the edge computing capabilities of the transport carrier to match the acquired task dynamic attribute information and the real-time dynamic information of the transport carrier to actively obtain the new material distribution tasks that are most suitable for the transport carrier to perform, and update and display the results on the display device to guide the operating staff to perform further transport operations.

[0079] Step 2: Obtain the status information of the transport carrier and the task pool in real time, and establish an information model of the distribution process, which includes the information matrix of each transport carrier and the information matrix of the task pool.

[0080] During the execution of the delivery task, the transport carrier constructed in step 1 obtains the status information of the transport carrier itself and the status information of the task pool in real time, and establishes an information model in the delivery process to digitize and standardize the acquired real-time status information. The information model serves as the key initial information for obtaining the delivery task.

[0081] The status information of each transport carrier includes: the number of the transport carrier, the maximum load weight of the transport carrier, the real-time position of the transport carrier, and the real-time load weight used by the transport carrier. The status information of each transport carrier is stored in its corresponding information matrix V. The information matrix V of the jth transport carrier is j for:

[0082]

[0083] Among them, AID j represents the number of the j-th transport carrier; represents the maximum load weight of the jth transport carrier; CO j Indicates the real-time position of the jth transport carrier; Indicates the load weight used by the jth transport carrier in real time. Based on the information matrix V, the working conditions of the transport carrier can also be tracked and traced in real time.

[0084] The status information of the delivery tasks in the task pool includes: the task number, the starting point coordinates of the task, the time window of the task, the load weight of the task and the status of the task.

[0085] The real-time delivery information of m delivery tasks in the task pool is stored in the information matrix W:

[0086]

[0087] Among them, TID k Indicates the number of the delivery task k; STA k Indicates the starting coordinates of the delivery task k; END k represents the destination coordinates of the delivery task k; Q k represents the load weight of delivery task k; LT k represents the earliest start time of task k; RT k represents the latest completion time of task k. Based on the information matrix W, the quantity and status of the delivery tasks can also be updated in real time.

[0088] Step 3: For any transport carrier numbered j, a set of advance delivery tasks N is generated for it based on the priority of the task and the load weight capacity of the transport carrier.

[0089] Step 3.1: Based on the information model established in step 2, determine whether the transport carrier numbered j is currently executing a task. If not, proceed to step 3.2; if yes, proceed to step 5;

[0090] Step 3.2 The transport carrier numbered j senses the priority of the delivery tasks to be executed in the task pool, that is, obtains the end time of each delivery task to be executed in the time window, and sorts the delivery tasks to be executed according to the priority of the delivery tasks to generate a task sequence list; among which, the priority of the delivery task is determined according to the latest end time of the delivery task. The earlier the latest end time of the delivery task, the higher its priority;

[0091] Step 3.3 The transport carrier numbered j establishes a constraint condition for the combination of delivery tasks related to its real-time load weight capacity, and according to the constraint condition, it intercepts delivery tasks from the task sequence list generated in step 3.2 in descending order of priority until no new delivery tasks can be intercepted; at this time, the delivery tasks intercepted by the transport carrier constitute the prior delivery task set N, and the process goes to step 4;

[0092] The constraints are:

[0093]

[0094] Where N represents the set of prior delivery tasks, N = {1, 2…, n}; Qi represents the material weight of the i-th delivery task; Indicates the maximum load weight of the transport carrier numbered j;

[0095] The information of the first delivery task set N is stored in the matrix H:

[0096]

[0097] Step 3.4 determines whether the transport carrier numbered j currently has idle load weight. If so, the task set currently executed by the transport carrier is used as its prior delivery task set N, and the process goes to step 6; if not, the transport carrier numbered j executes the current task.

[0098] Step 4: Perform optimal path planning for the advance delivery task set N currently obtained by the transport carrier numbered j. The planning goal is to make the total delivery path of the advance delivery set N as the shortest.

[0099] Reference Figure 2 , the A* algorithm is used to plan the path for each task in the advance delivery task set N currently obtained by the transport carrier numbered j; when planning the path, the shortest path is used as the objective function, and the nodes are selected and expanded, and the path is selected according to the heuristic function to find the shortest path.

[0100] In the process of searching the shortest path by the A* algorithm, we select the best node to expand based on the distance and heuristic estimate of the current node, and calculate the distance and heuristic estimate of the neighboring nodes. By continuously updating the distance and path of the node, we gradually find the shortest path from the starting point to the delivery destination of each delivery task. Finally, we select the path planning scheme with the shortest total delivery as the optimal path planning result for the entire set of N delivery tasks.

[0101] The specific process is as follows:

[0102] Step 4.1 Create a directed weighted graph: Create a directed weighted graph based on the layout of the workshop and the coordinates of the task starting point. The nodes of the directed weighted graph represent the locations in the workshop, the edges represent the movement paths between two locations, and the weight of the edges represents the distance of the movement paths.

[0103] Step 4.2 Initialize the start node and target node: select the real-time location of the transport carrier as the start node, and the delivery destination of all delivery tasks as the target node.

[0104] Step 4.3 Initialize the distance and calculate the heuristic estimate: For each node in the directed weighted graph, initialize the distance to infinity (INF), set the distance of the starting node to 0, and use Manhattan distance as the heuristic function to calculate the heuristic estimate of each node to the target node.

[0105] Step 4.4 Create a priority queue: Create a priority queue to store the nodes to be expanded, and create a closed set to store the nodes that have been explored.

[0106] Step 4.5 performs an A* search: add the starting node to the priority queue, maintain an accumulated value of the current shortest path (for example, using the total path length), and repeat the following steps until all target nodes are found or the priority queue is empty:

[0107] a. Select the node with the smallest heuristic evaluation value from the priority queue as the current node;

[0108] b. If the current node is one of the target nodes, check whether all paths to the target node have been found. If so, update the cumulative value of the current shortest path;

[0109] c. For each neighbor node of the current node, calculate the distance from the start node to the neighbor node. If the neighbor node is already in the closed set, ignore it. If the neighbor node is not in the closed set, further determine whether it is in the priority queue. If the neighbor node is not in the priority queue, add it to the priority queue and update its distance and path information. If the neighbor node is already in the priority queue, check whether the current path is better. If so, update the distance and path information.

[0110] d. Add the current node to the closed set;

[0111] Step 4.6 Reconstruct the shortest path: For each target node found, starting from the starting node, find the nodes on the shortest path in turn according to the path information of the nodes, and combine the shortest paths of each task to form the total path.

[0112] Among them, an additional condition is introduced in step c of step 4.5 to ensure that the task reaches the starting point before continuing the path. That is, an attribute is added to each node to indicate whether it has reached the starting point. When calculating the distance and path of the neighbor node, if the neighbor node has not reached the starting point, its distance can be set to infinity to exclude it as a candidate for the next expansion node.

[0113] Step 5: Determine whether the transport carrier numbered j currently has idle load weight. If so, use the task set currently executed by the transport carrier as its prior delivery task set N and proceed to step 6. If not, the transport carrier numbered j continues to execute the current task.

[0114] Step 6: Establish a multi-objective function, screen the task pool W based on the information model of the task distribution process, use the NSGA-Ⅱ algorithm to optimize the multi-objective function, and generate the Pareto frontier solution set (refer to Figure 3-4 ).

[0115] Step 6.1 Establish a multi-objective function;

[0116] Based on the idea of ​​active acquisition of delivery tasks and parallel delivery, the traditional global task static allocation strategy of the carrier is improved into a dynamic active acquisition strategy in the process, and the objective function is designed as the distribution path overlap, distribution detour distance and overall load rate for the CTB (Concurrent Task-Based) active acquisition strategy model for the entire distribution process. Therefore, through step 2, the attribute information of the tasks to be picked up in the task pool is obtained, and according to the previous delivery task set N and the optimal path planning obtained in steps 3, 4 and 5, the multi-objective function can be expressed as follows:

[0117]

[0118]

[0119]

[0120] Where n represents the number of coordinate sets consisting of n task starting points in the previous delivery task set N; p represents the number of coordinate sets consisting of p task starting points in the current delivery task set after obtaining a new delivery task; Indicates the maximum load weight of the transport carrier numbered j; Q k,k+1 Indicates the load weight of the carrier between adjacent coordinates; Q k,k+1 Indicates the load weight of the carrier between adjacent coordinates; T k,k+1 Indicates that the load weight of the transport carrier is Q k,k+1 Load time at 10 ... i 、x i+1 、x k 、x k+1 ,y i ,y i+1 ,y k ,y k+1 They represent the starting point coordinates in the above different coordinate sets respectively; m represents the number of all coordinate sets in the generated delivery task set.

[0121] Step 6.2 Establish the constraints of the multi-objective function:

[0122]

[0123] LT i ≤si ≤RT i -d i (5)

[0124] s i +d i ≤c i ≤RT i (6)

[0125] LT j ≤s j ≤RT j -d j (7)

[0126]

[0127]

[0128]

[0129]

[0130]

[0131] Among them, N is the first delivery task set; q i Represents the material weight of the i-th delivery task; represents the maximum load weight of the transport carrier numbered j; W is the dynamically updated task pool, including all current delivery tasks; TID i and TID j Represents task i and task j respectively; LT i and LT j Respectively represent the earliest start time of task i and task j; RT i and RT j Respectively represent the latest end time of task i and task j; s i and j They represent the calculated actual start time of task i and task j respectively; c i and c j They represent the actual end time of task i and task j respectively; d i and d j Represent the duration of task i and task j respectively; d k represents the duration of task k; RT k represents the latest completion time of task k; v represents the speed of the transport carrier; S represents the first end time of the transport carrier when performing the distribution concentration task; Path(TID i ) indicates the task TID i Location; Dist(Path(TID i),Path(TID j )) indicates task TID i and TID j Dist(Path) represents the total distance of the path; ADD represents the allowed detour distance; Tat represents the total allowed time to complete all tasks; t l represents the material loading and unloading time of the distribution task; h represents the power consumption rate when the carrier is empty; δ represents the power consumption increment per unit time when each unit weight of material is added to the carrier; E N represents the total power consumption of the delivery task set N of the transport carrier; E j Indicates the current remaining power of the transport carrier; E b Indicates the actual remaining power of the transport carrier; D N Indicates the time required to reach the charging station after completing all tasks in the current task set; Z is an integer; Q i,i+1 represents the load weight of the carrier from coordinate i to coordinate i+1; T i,i+1 Indicates the time taken to transport the carrier from coordinate i to coordinate i+1.

[0132] Formula (4) indicates that the total load weight of the first delivery task set N must be less than the maximum load weight of the transport carrier numbered j. Formulas (5) to (11) indicate that the delivery process must be within the time window required by the delivery task. Formula (12) indicates that the transport carrier numbered j must consider whether the power is sufficient when obtaining the delivery task set.

[0133] Step 6.3: Optimize the multi-objective function using NSGA-Ⅱ algorithm;

[0134] Step 6.3.1 encodes the delivery task and generates the initial population;

[0135] The coordinates, numbers, starting point marks, load weights, and left and right time windows of all tasks (including the preceding task set N and all tasks in the task pool W) are mixed with integer coding and real number coding, that is, they are mixed with Boolean values ​​for coding; the number of random individuals to be generated is obtained according to the population size and the given number of coordinate individuals; tasks in the task pool are randomly selected to generate the remaining individuals, and the given individuals (i.e., the coordinates of the tasks in the preceding delivery task set) and the randomly generated remaining individuals are merged into the initial population.

[0136] When generating the initial population, predefined operations are performed on specific individuals. For those delivery tasks that have been accepted, they are encoded as immutable genotypes and only crossover operations are performed to ensure that they remain unchanged in subsequent mutation operations.

[0137] Step 6.3.2 Individual judgment and function value calculation;

[0138] According to the constraints of step 6.2, the individuals are judged and screened, the population that does not meet the constraints is corrected, and the individuals that meet all the constraints are retained. These retained individuals are used to generate the solution set; according to the objective functions f1, f2, and f3 of the three goals of delivery path overlap, delivery detour distance, and overall load rate, the coordinates of the retained individuals are substituted into the objective functions of these three goals respectively, and the objective function values ​​of the retained individuals are calculated.

[0139] Step 6.3.3 Non-dominated sorting;

[0140] For the individuals in the initial population generated in step 6.3.1, non-dominated sorting is performed according to the objective function value calculated in step 6.3.2, and different individuals are divided into rank1, rank2, rank3, etc. to determine the non-dominated rank of each individual.

[0141] Step 6.3.4 calculates the crowding degree;

[0142] For individuals of the same level, the crowding operator is used to calculate their crowding values, and the comprehensive crowding of individuals is determined by calculating the neighboring distances of individuals in each objective function dimension. Individuals with a larger comprehensive crowding are more densely distributed in the target space, while individuals with a smaller comprehensive crowding are more isolated to maintain a uniform distribution on the Pareto frontier.

[0143] Step 6.3.5 generates a new population;

[0144] According to the non-dominated level obtained in step 6.3.3 and the comprehensive crowding value obtained in step 6.3.4, the next generation of individuals are selected; then the genetic operations such as crossover (exchanging the order of coordinate values ​​within the individual) and mutation (randomly exchanging a task other than the original task in the new individual with the task to be intercepted in the task pool) are used to operate on the selected next generation of individuals, and the parent and offspring are merged, and the top N individuals are selected based on the non-dominated sorting level and crowding distance to form the next generation of new population.

[0145] Step 6.3.6 Iterate multiple times and output the results;

[0146] Return the new population generated in step 6.3.5 to step 6.3.2 for multiple iterations until the maximum number of iterations is reached. The final population is the Pareto frontier solution set. Each solution in the Pareto frontier solution set represents the advantages and disadvantages of the three objective functions of different individuals (distribution task sets). For example, objective function 1 of individual A may be better than that of individual B, but objective function 2 of individual B may be better than that of individual A.

[0147] Step 7: Select the distribution task set N' from the Pareto frontier solution set.

[0148] Because different material handling scenarios have different specific goals and requirements. After determining the goals and demand preferences of the current scenario, according to the weight requirements of material distribution scenario, priority, degree of convenience, detour distance and load rate, a trade-off is made in the Pareto frontier solution set, and the distribution task set N' is selected from the Pareto frontier solution set.

[0149] Step 8: Delivery task update.

[0150] According to the distribution task set N' selected in step 7, the information status of the corresponding distribution task in the task pool is modified to have been accepted to realize the real-time update of the distribution demand. After the transport carrier numbered j dynamically generates the advance distribution task set N, continue to repeat the operations of steps 5 to 8 during the task distribution process to realize the efficient and timely distribution of workshop materials.

Claims

1. A method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration, characterized in that: The following steps are involved: Step 1: Construct a transport carrier with information acquisition capabilities; Step 2: Acquire the status information of the transport carrier and the task pool in real time, and establish an information model of the distribution process based on the status information, wherein the information model includes the information matrix of each transport carrier and the information matrix of the task pool; the status information of the transport carrier includes the number of the transport carrier, the maximum load weight, the real-time position and the load weight used in real time; the status information of the task pool includes the number of the task, the starting point coordinates, the time window, the load weight and the status; Step 3: For any transport carrier numbered j, based on the task priority and the load weight capacity of the transport carrier, generate a set of N advance delivery tasks for it; Step 3.1: According to the information model, determine whether the transport carrier numbered j is currently executing a task. If not, proceed to step 3.2; if yes, proceed to step 5; Step 3.2 The transport carrier numbered j senses the priority of the delivery tasks to be executed in the task pool, sorts the delivery tasks to be executed according to the priority, and generates a task sequence list; Step 3.3 The transport carrier numbered j establishes a constraint condition for the combination of delivery tasks related to its real-time load weight capacity, and according to the constraint condition, the delivery tasks are cumulatively intercepted from the task sequence list in descending order of priority, and the intercepted delivery tasks constitute the prior delivery task set N, and then proceed to step 4; Step 4: Use the A* algorithm to plan the optimal path for the current delivery task set N obtained by the transport carrier numbered j; Step 5: Determine whether the transport carrier numbered j currently has idle load weight. If so, the task set currently executed by the transport carrier is used as its prior delivery task set N, and the process goes to step 6. If not, the transport carrier numbered j continues to execute the current task. Step 6: Establish a multi-objective function with delivery path overlap, detour distance and overall load rate as the target, screen the task pool based on the information model, use the NSGA-Ⅱ algorithm to optimize the multi-objective function, and generate a Pareto frontier solution set; Step 7: Select a distribution task set N' from the Pareto frontier solution set according to the material distribution scenario, priority, degree of convenience, detour distance and weight requirements of load rate; Step 8: Update the status information of the corresponding delivery tasks in the task pool according to the delivery task set N'.

2. The method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration according to claim 1 is characterized in that: Step 1 is as follows: By installing sensors, software systems and display devices on the transport carrier, and configuring RFID electronic tags at key intersections, entrances and exits, and material starting points in the workshop / warehouse, the transport carrier can have real-time status perception capabilities, including the current position of the transport carrier, the remaining load weight, the path information of the distribution task set, the attributes of the tasks to be picked up, and the communication status; the attributes of the tasks to be picked up include the task number, material location coordinates, material weight, and whether the attributes have been picked up; The software system dynamically interacts based on the wireless network environment to obtain the attributes of the tasks to be accepted in the task pool, and relies on the transport carrier's own edge computing capabilities to match the obtained task dynamic attribute information and the transport carrier's real-time dynamic information, so as to actively obtain the new material distribution tasks that are most suitable for the transport carrier to perform, and update and display the results on the display device.

3. The method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration according to claim 2 is characterized in that: In step 2: The information matrix of each transport carrier is: Among them, AID j represents the number of the j-th transport carrier; represents the maximum load weight of the jth transport carrier; CO j Indicates the real-time position of the jth transport carrier; represents the load weight used by the j-th transport carrier in real time; The information matrix of the task pool is: Among them, TID k Indicates the number of the delivery task k; STA k Indicates the starting coordinates of the delivery task k; END k represents the destination coordinates of the delivery task k; Q k represents the load weight of delivery task k; LT k represents the earliest start time of task k; RT k represents the latest completion time of task k.

4. The method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration according to claim 3 is characterized in that: In step 4: When the A* algorithm is used to perform path planning for each task in the prior delivery task set N currently obtained by the transport carrier numbered j, the shortest path is used as the objective function, and the shortest path from the starting point to the delivery destination of each delivery task is found by selecting and expanding nodes and performing path selection according to the heuristic function. Finally, the path planning scheme with the shortest total delivery is selected as the optimal path planning result for the prior delivery task set N.

5. The method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration according to claim 4 is characterized in that: Step 6 is as follows: Step 6.1 Establish multi-objective function: Step 6.2 Establish the constraints of the multi-objective function: LT i ≤s i ≤RT i -d i (5) s i +d i ≤c i ≤RT i (6) LT j ≤s j ≤RT j -d j (7) Where N is the set of preceding delivery tasks; qi represents the material weight of the i-th delivery task; represents the maximum load weight of the carrier numbered ; W is the dynamically updated task pool, including all current delivery tasks; TID i and TID j Represents task i and task j respectively; LT i and LT j Respectively represent the earliest start time of task i and task j; RT i and RT j Respectively represent the latest end time of task i and task j; s i and j They represent the calculated actual start time of task i and task j respectively; c i and c j They represent the actual end time of task i and task j respectively; d i and d j Represent the duration of task i and task j respectively; d k represents the duration of task k; RT k represents the latest completion time of task k; v represents the speed of the transport carrier; S represents the first end time of the transport carrier when performing the distribution concentration task; Path(TID i ) indicates the task TID i Location; Dist(Path(TID i ), Path(TID j )) indicates task TID i and TID j Dist(Path) represents the total distance of the path; ADD represents the allowed detour distance; Tat represents the total allowed time to complete all tasks; t represents the material loading and unloading time of the distribution task; h represents the power consumption rate when the carrier is empty; δ represents the power consumption increment per unit time when each unit weight of material is added to the carrier; E N represents the total power consumption of the delivery task set N of the transport carrier; E j Indicates the current remaining power of the transport carrier; E b Indicates the actual remaining power of the transport carrier; D N Indicates the time required to reach the charging station after completing all tasks in the current task set; Z is an integer; Q i,i+1 represents the load weight of the carrier from coordinate i to coordinate i+1; T i,i+1 Indicates the time taken to transport the carrier from coordinate i to coordinate i+1; Step 6.3 uses the NSGA-Ⅱ algorithm to optimize the multi-objective function and obtain the Pareto frontier solution set.

6. The method for dynamically acquiring workshop material distribution tasks based on edge-cloud collaboration according to claim 5 is characterized in that: Step 6.3 is as follows: Step 6.3.1 encodes the delivery task and generates the initial population; Step 6.3.2: judging and screening the individuals according to the constraints of step 6.2, retaining the individuals that meet all the constraints and bringing them into the multi-objective function to calculate the function value; Step 6.3.3 For the individuals in the initial population generated in step 6.3.1, perform non-dominated sorting according to the objective function value calculated in step 6.3.2 to determine the non-dominated level of each individual; Step 6.3.4 For individuals of the same level, use the crowding operator to calculate their crowding values, and determine the comprehensive crowding of individuals by calculating the proximity distance of individuals in each objective function dimension; Step 6.3.5: select the next generation of individuals according to the non-dominated ranking and the comprehensive crowding degree, and then operate them by crossover and mutation, merge the parent generation and the offspring generation, select the top N individuals based on the non-dominated ranking and crowding degree distance, and form the next generation of new population; Step 6.3.6 returns the new population generated in step 6.3.5 to step 6.3.2 for multiple iterations until the maximum number of iterations is reached and the iteration is stopped. The final population is the Pareto frontier solution set.

Images