A high-priority replenishment order sharing bottleneck collaborative rescheduling method and system for an intelligent manufacturing workshop

Abstract

The application discloses a high-priority replenishment order insertion shared bottleneck collaborative rescheduling method and system for an intelligent manufacturing workshop, which comprises the following steps: collecting equipment and resource information in a manufacturing scene and forming a risk coding state; uniformly constructing collected data into a dynamic random graph; identifying an order insertion trigger condition of a high-priority replenishment task and screening a candidate task set based on the risk coding state and the dynamic random graph; establishing an upper-layer discrete optimization model based on the candidate task set to obtain a disturbance-controlled collaborative rearrangement scheme; establishing a middle-layer resource and timing coordination model based on the disturbance-controlled collaborative rearrangement scheme to obtain a resource occupation timing scheme; establishing a lower-layer safety motion control model based on the resource occupation timing scheme to obtain an executable reference trajectory and a control amount; and updating a dynamic graph edge weight, a risk cost and a coordination parameter based on data obtained from the model, and performing online learning enhancement correction on a subsequent rolling window to complete collaborative rescheduling.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and industrial logistics collaborative scheduling technology, specifically involving a high-priority replenishment order sharing bottleneck collaborative rescheduling method and system for intelligent manufacturing workshops. Background Technology

[0002] In-manufacturing logistics systems are widely found in lithium battery workshops, semiconductor factories, automobile assembly workshops, and large-scale warehousing and logistics systems. These systems typically involve mobile robots, unmanned forklifts, tractors, robotic arms, and inspection equipment working collaboratively around tasks such as material replenishment, inter-area transfer, buffer switching, and anomaly verification. In real production environments, the most disruptive engineering scenario is often not the general task allocation, but rather the sudden insertion of high-priority material replenishment tasks into the existing execution sequence during continuous production, leading to material shortages at critical workstations, main channel congestion, missed energy replenishment, and cascading disruptions to subsequent tasks.

[0003] Existing technologies typically address localized problems through various approaches, such as multi-robot task allocation, real-time scheduling in smart factories, graph model optimization, path planning, energy-aware dispatching, or non-preemptive scheduling. For example, some solutions focus on Dynamic Multi-Robot Task Allocation (MRTA) and online dispatching, while others introduce graph models, graph learning, or hypergraph search into real-time task allocation in smart factories. Still others focus on energy-aware scheduling, non-preemptive task optimization, or flexible workshop transportation collaboration. Although these solutions cover technical directions such as equipment-task matching, path planning, energy optimization, and general production scheduling, most remain at the level of general MRTA, general smart factory task scheduling, or general multi-robot non-preemptive optimization.

[0004] However, existing technologies generally lack a technical solution directly addressing the complex scenario of "high-priority replenishment order insertion + shared bottleneck resource avoidance + energy replenishment constraints + controlled rescheduling of disturbances." Some existing solutions only perform task assignment, failing to integrate redistribution, sorting, energy replenishment insertion, and shared bottleneck flow limiting. Other solutions rely on relatively complete global information, making direct integration into the rolling scheduling processes of Manufacturing Execution Systems (MES), Warehouse Management Systems (WMS), or Warehouse Control Systems (WCS) difficult. Furthermore, some existing solutions do not explicitly model this problem as a multi-stage, stochastic, mixed-integer, strongly coupled high-dimensional combinatorial optimization and optimal control problem, thus failing to simultaneously address combinatorial complexity representation and online solution requirements. Therefore, it is necessary to provide a more suitable collaborative rescheduling method for shared bottlenecks within the manufacturing logistics field. Summary of the Invention

[0005] This invention aims to address the shortcomings of existing technologies and provides the following solutions: A high-priority replenishment order sharing bottleneck collaborative rescheduling method for smart manufacturing workshops includes the following steps: Collect data on material shortage risks at workstations, task pool status, equipment status, shared bottleneck resource occupancy status, and energy replenishment resource availability status, and generate risk codes accordingly. The workstation material shortage risk, the task pool status, the equipment status, the shared bottleneck resource occupancy status, and the energy replenishment resource availability status are all uniformly constructed into a dynamic random graph; Based on the risk coding status and the dynamic random graph, identify the order insertion triggering conditions for high-priority replenishment tasks and filter the set of candidate tasks that can enter the current scrolling window. Based on the candidate task set, an upper-level discrete optimization model is established around task acceptance, task assignment, task reassignment and service sorting, and the perturbation-controlled collaborative rearrangement scheme is obtained using the upper-level discrete optimization model. Based on the disturbance-controlled collaborative rearrangement scheme, a mid-level resource and timing coordination model is established for shared bottleneck resources, intersections, buffer positions, and power replenishment positions, and the resource occupancy timing scheme is obtained using the mid-level resource and timing coordination model. Based on the resource occupancy timing scheme, a lower-level safe motion control model is established for equipment dynamics, collision avoidance constraints and continuous control input, and the executable reference trajectory and control quantity are obtained using the lower-level safe motion control model; Based on the perturbation-controlled collaborative rescheduling scheme, the resource occupancy timing scheme, the reference trajectory, and the control quantity, as well as the feedback data after execution, the dynamic graph edge weights, risk costs, and coordination parameters are updated, and the subsequent rolling window is enhanced and corrected through online learning to complete the collaborative rescheduling.

[0006] Preferably, the risk coding status is: in, Indicates the risk coding status. This indicates the risk of production line stoppage due to material shortage. Indicates items with urgent time windows. Indicates task priority. β 1. β 2 and β 3 represents the weighting coefficient in the risk urgency model. j Indicates a task node. t Indicates the decision-making period.

[0007] Preferably, the dynamic random graph is: in, Represents a dynamic random graph. Indicates the objects on site. This represents reachability, conflict, cooperation, and resource competition relationships. It represents travel time, congestion costs, risk exposure, collaboration benefits, and switching costs.

[0008] Preferably, the candidate task set is as follows: in, Represents the current set of tasks. Indicates the threshold for triggering order insertion. Indicates task j At any moment t The deadline H Indicates the length of the scroll window.

[0009] Preferably, the upper-level discrete optimization model is: in, This represents the upper-level discrete decision vector. This indicates that the task accepts variables. Indicates the decision-making period t Internal mission j Is it by the equipment? i Accept, Indicates task j Has a reassignment occurred relative to the previous rolling cycle? Indicates task j In the equipment i The current execution sequence order, Indicates equipment i Whether to enter the energy replenishment mode during this period, max represents finding the maximum value. Represents the mathematical expectation. T This indicates the upper bound of the time range for optimizing the summation. This indicates the benefits or value of the services received for completing the task. This indicates the amount of work completed on time. This indicates the cost of being late. This represents the cost of reassignment and sorting perturbations. This indicates the risk and cost of supply disruption. Indicates the cost of energy consumption. Indicates the weight of the cost of being late. Indicates the weights of reassignment and sorting perturbation costs. This indicates the weight of the risk and cost of material shortage. Indicates the weight of energy consumption costs. It represents a collection of embodied devices. Indicates equipment i During the period t The upper limit of the number of tasks that can be accepted. Indicates the remaining charge or state of charge. This indicates the power threshold that triggers the power replenishment. M Represents a constant.

[0010] Preferably, the mid-level resource and timing coordination model is as follows: Where min represents finding the minimum value. Indicates the time variable at which entry occurs. Indicates the time of departure. Indicates the order of variables. Indicates the energy allocation variable. This indicates the cost of bottleneck channel congestion. This indicates the waiting for the cost of dissemination. This indicates the cost of competing for a replacement energy position. This indicates the cost of congestion at critical workstations. Indicates the congestion cost weight. This indicates the weight of the cost of waiting for propagation. This indicates the weight of the cost of energy replenishment competition. This indicates the weight of the congestion cost at critical workstations. r Indicates energy replenishment resources. Indicates equipment k At any moment t Resources r The moment of departure Indicates equipment k At any moment t For shared resources or energy replenishment resources r The moment of entry, This represents a set of shared bottleneck resources. Represents the resource usage variable. Indicates the bottleneck resource capacity. Indicates the capacity of the filler potential. This represents a set of energy replenishment resources.

[0011] Preferably, the lower-level safety motion control model is: in, Indicates equipment i State variables, Indicates the reference trajectory. Indicates the cost of security. This indicates the cost of controlling energy consumption. Indicates the weight of trajectory tracking error. Indicates the weight of security costs. This indicates the weight of the control input cost. Indicates the weight of energy or electricity cost. Indicates continuous control input. This indicates a random perturbation. Indicates equipment i The dynamic state transition function, This represents the position vector of the equipment in the workshop coordinate system. Indicates equipment k At any moment t The position vector, This indicates the minimum safe distance between the spare parts. This indicates the minimum safe power threshold at which the device is allowed to perform a task.

[0012] This invention also provides a high-priority replenishment order sharing bottleneck collaborative rescheduling system for intelligent manufacturing workshops. The system applies the above-mentioned method and includes: a risk coding module, a dynamic random graph construction module, an order identification and candidate screening module, an upper-level optimization module, a middle-level coordination module, a lower-level control module, and an online learning module. The risk coding module is used to collect data on workstation material shortage risk, task pool status, equipment status, shared bottleneck resource occupancy status, and energy replenishment resource availability status, and to form a risk coding status. The dynamic random graph construction module is used to uniformly construct the workstation material shortage risk, the task pool status, the equipment status, the shared bottleneck resource occupancy status, and the energy replenishment resource availability status into a dynamic random graph; The order insertion identification and candidate screening module identifies the order insertion triggering conditions of high-priority replenishment tasks based on the risk coding status and the dynamic random graph, and filters the set of candidate tasks that can enter the current scrolling window. The upper-level optimization module establishes an upper-level discrete optimization model based on the candidate task set, focusing on task acceptance, task assignment, task reassignment, and service sorting, and uses the upper-level discrete optimization model to obtain a perturbation-controlled collaborative rearrangement scheme. The mid-level coordination module establishes a mid-level resource and timing coordination model for shared bottleneck resources, intersections, buffer positions, and energy replenishment positions based on the disturbance-controlled collaborative rearrangement scheme, and uses the mid-level resource and timing coordination model to obtain a resource occupancy timing scheme. Based on the resource occupancy timing scheme, the lower-level control module establishes a lower-level safe motion control model for equipment dynamics, collision avoidance constraints, and continuous control input, and uses the lower-level safe motion control model to obtain an executable reference trajectory and control quantity. The online learning module updates the dynamic graph edge weights, risk costs, and coordination parameters based on the perturbation-controlled collaborative rearrangement scheme, the resource occupancy timing scheme, the reference trajectory, the control quantity, and the feedback data after execution. It also performs online learning enhancement correction on subsequent scrolling windows to complete the collaborative rescheduling.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention unifies the high-priority replenishment and order insertion problem and the shared bottleneck resource coordination problem in manufacturing internal logistics, forming a shared bottleneck collaborative rescheduling method comprising seven technical steps. This method uses dynamic random graphs as a unified modeling carrier, multi-stage random mixed integer strongly coupled optimization as its mathematical foundation, and adopts a technical route of "upper-layer acceptance / assignment / reassignment + mid-layer resource timing coordination + lower-layer safe motion control + mechanism design and payment embedding + online learning enhancement" to form a closed loop for shared bottleneck collaborative rescheduling in the manufacturing site. It is applicable to manufacturing sites with cycle time constraints, replenishment constraints, and shared channel resource constraints. Attached Figure Description

[0014] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is an overall flowchart of an embodiment of the present invention; Figure 2 This is a schematic diagram of the system architecture layering according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the timing coordination of shared bottleneck resources and supplementary energy resources in an embodiment of the present invention. Figure 4This is a schematic diagram of a scenario according to an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] Example 1 In this embodiment, as Figure 1 As shown, a high-priority replenishment order sharing bottleneck collaborative rescheduling method for smart manufacturing workshops includes the following steps: S1. Collect data on the risk of material shortage at workstations, the status of the task pool, the status of equipment, the status of shared bottleneck resources, and the status of available replenishment resources, and generate risk codes.

[0019] In this embodiment, the system collects data on workstation material shortage risks, task pool status, equipment location, equipment load, power status, shared bottleneck resource occupancy status, and replenishment resource availability status at the manufacturing logistics site within each rolling scheduling cycle. It also assigns risk codes to unexpected replenishment tasks. Risk coding is not limited to whether a task has arrived; rather, it focuses on characterizing the potential material shortage and production line stoppage risks caused by high-priority replenishment tasks, the remaining tolerable waiting time, and the task priority. This allows subsequent scheduling processes to prioritize identifying replenishment events that truly impact production line continuity.

[0020] Specifically, at the beginning of each rolling scheduling cycle, the system first collects data on the risk of material shortage at workstations, task pool status, equipment location, power consumption, load, energy replenishment availability, and shared bottleneck resource occupancy. Unlike general task allocation schemes, this step does not only collect data on "whether a task exists," but focuses on collecting downtime risk signals and shared bottleneck resource status signals caused by high-priority replenishment orders.

[0021] The collection of embodied equipment is denoted as: in, N This indicates the total number of embodied intelligent devices or mobile replenishment devices.

[0022] The current task set is denoted as: in, Mt It changes over time to reflect the dynamic arrival of high-priority replenishment, inter-regional transfer, and anomaly verification tasks.

[0023] For any device Define its state variables as: in, This represents the position vector of the equipment in the workshop coordinate system. Indicates speed status. Indicates the remaining charge or state of charge. Indicates the current load status. This indicates that the current task sequence has been accepted. Indicates the operating mode.

[0024] For any task Define the task parameter vector as follows: in, This indicates the pickup / delivery location or target workstation location corresponding to the task. Indicates the task deadline. This indicates the benefits or value of the services received for completing the task. Indicates the risk level. Indicates urgency. This describes the task's requirements for resources such as bottleneck channels, cache bits, or power replenishment bits.

[0025] In order to explicitly code high-priority replenishment orders, the risk coding status is: in, Indicates the risk coding status. This indicates the risk of production line stoppage due to material shortage. Indicates items with urgent time windows. Indicates task priority. , and This represents the weighting coefficients in the risk urgency model. j Indicates a task node. t Indicates the decision-making period.

[0026] In one implementation, this step uses the risk of material shortage, bottleneck congestion, and energy replenishment margin as the core dimensions of status acquisition and risk coding to improve the pertinence of subsequent order insertion identification and rescheduling decisions.

[0027] S2. The risk of material shortage at workstations, task pool status, equipment status, shared bottleneck resource occupancy status, and energy replenishment resource availability status are uniformly constructed into a dynamic random graph.

[0028] In this embodiment, workstations, buffer positions, power replenishment points, bottleneck channels, task nodes, and embodied device nodes are uniformly represented as a dynamic random graph: in, Represents a dynamic random graph. Indicates the objects on site. This represents reachability, conflict, cooperation, and resource competition relationships. It represents travel time, congestion costs, risk exposure, collaboration benefits, and switching costs.

[0029] In one implementation, a dynamic random graph is used to explicitly express the main channel congestion status, intersection conflict status, workstation material shortage risk, system risk changes after equipment accepts an order, and the saturation level of the replenishment point, thereby providing a unified modeling basis for collaborative rescheduling of shared bottlenecks after high-priority replenishment orders.

[0030] For device nodes i With task nodes j Define equivalent edge weights: in, Indicates equipment i Response task j The estimated arrival time, This indicates the estimated energy consumption for execution. This represents the combined cost of exposure to bottleneck congestion and material shortage risks. This indicates the benefits of matching equipment capabilities with task requirements; , , and This represents the corresponding weighting coefficient.

[0031] For sharing bottleneck resources Define resource usage variables , indicating equipment i At any moment t Does it consume resources? r This variable will further participate in the timing conflict resolution process during the intermediate coordination steps.

[0032] In one implementation, this step directly uses dynamic graphs to express the coupling relationships between shared bottleneck resources, material shortage risks, energy replenishment competition, and task switching disturbances, in order to support subsequent hierarchical solutions.

[0033] S3. Based on the risk coding status and dynamic random graph, identify the order insertion trigger conditions for high-priority replenishment tasks and filter the set of candidate tasks that can enter the current scrolling window.

[0034] In this embodiment, after obtaining the dynamic random graph, the system identifies whether there are any high-priority replenishment tasks among the newly arriving tasks that meet the insertion conditions. The insertion conditions include at least one or more of the following: the risk of material shortage at the workstation exceeds a threshold, the remaining tolerable waiting time is less than a threshold, the task is strongly correlated with the cycle time of the critical process, or its risk level is significantly higher than that of existing tasks.

[0035] The candidate task set is as follows: in, Represents the current set of tasks. Indicates the threshold for triggering order insertion. Indicates task j At any moment t The deadline H Indicates the length of the scroll window.

[0036] The aforementioned steps, while screening tasks, provide a decision-making entry point for the upper-level discrete optimization: "whether to accept the task." Correspondingly, a task acceptance variable is defined. Its meaning is: if Then the task j It will be included in the current rolling window; otherwise, it will be postponed or rejected in this round of scheduling.

[0037] In one implementation, this step sets "task acceptance / rejection" as an explicit optimization decision to adapt to manufacturing environments with high loads and limited shared bottleneck resources.

[0038] S4. Based on the candidate task set, establish an upper-level discrete optimization model around task acceptance, task assignment, task reassignment and service ranking, and use the upper-level discrete optimization model to obtain a perturbation-controlled collaborative rearrangement scheme.

[0039] In this embodiment, after a high-priority replenishment order is triggered, the upper layer is responsible for task acceptance, task assignment, task reassignment, and service prioritization. The task assignment variables, reassignment variables, prioritization variables, and replenishment variables are defined as follows: , , , .in, Indicates the decision-making period t Internal mission j Is it by the equipment? i Accept, Indicates task j Has a reassignment occurred relative to the previous rolling cycle? Indicates task j In the equipment i The current execution sequence order, Indicates equipment iWhether to switch to energy replenishment mode during this period. For ease of subsequent description, let the upper-level discrete decision vector be: The upper-level discrete optimization model is: in, This represents the upper-level discrete decision vector, and max represents finding the maximum value. Represents the mathematical expectation. T This indicates the upper bound of the time range for optimizing the summation. This indicates the amount of work completed on time. This indicates the cost of being late. This represents the cost of reassignment and sorting perturbations. This indicates the risk and cost of supply disruption. Indicates the cost of energy consumption. Indicates the weight of the cost of being late. Indicates the weights of reassignment and sorting perturbation costs. This indicates the weight of the risk and cost of material shortage. Indicates the weight of energy consumption costs. Indicates equipment i During the period t The upper limit of the number of tasks that can be accepted. This indicates the power threshold that triggers the power replenishment. M This represents a sufficiently large constant. The above two points regarding... The constraints are used to characterize the assignment changes between the current cycle and the previous cycle, thus giving "reassignment" a clear mathematical meaning.

[0040] In one implementation, this step integrates acceptance, assignment, reassignment, sorting, and energy replenishment into the upper-level discrete optimization problem to form a mixed integer control subproblem.

[0041] S5. Based on the disturbance-controlled collaborative rearrangement scheme, a mid-level resource and timing coordination model is established for shared bottleneck resources, intersections, buffer positions and energy replenishment positions, and the resource occupancy timing scheme is obtained using the mid-level resource and timing coordination model.

[0042] In this embodiment, the middle layer establishes a resource and timing coordination model around shared bottleneck resources, intersections, cache bits, and power replenishment bits, such as... Figure 3 As shown. For any device i At any momentt Entering the bottleneck resources r The event can introduce an entry time variable. and departure time variable For any two devices competing for the same resource i and k Introducing the first-to-last variable Further define the energy replenishment position allocation variable. ,in , indicating equipment i During the period t Whether it is allocated to energy replenishment resources r .

[0043] The mid-level resource and timing coordination model is as follows: Where min represents finding the minimum value. This indicates the cost of bottleneck channel congestion. This indicates the waiting for the cost of dissemination. This indicates the cost of competing for a replacement energy position. Indicates the cost of congestion at critical workstations Indicates the congestion cost weight. This indicates the weight of the cost of waiting for propagation. This indicates the weight of the cost of energy replenishment competition. This indicates the weight of the congestion cost at critical workstations. Indicates equipment k At any moment t Resources r The moment of departure Indicates equipment k At any moment t For shared resources or energy replenishment resources r The moment of entry, This represents a set of shared bottleneck resources. Indicates the bottleneck resource capacity. Indicates the capacity of the filler potential. This represents the set of replenishment resources. Therefore, the middle layer will use the replenishment trigger variable. With specific energy replenishment resource allocation variables Explicit connections were made to eliminate the symbolic disconnect between upper-level decision-making and middle-level resource constraints.

[0044] In one implementation, this step explicitly addresses the resource-timing coordination problem between shared bottleneck resources and supplementary power positions to reduce the conflict resolution burden of the underlying path planning.

[0045] S6. Based on the resource occupancy timing scheme, a lower-level safe motion control model is established for equipment dynamics, collision avoidance constraints and continuous control input, and the executable reference trajectory and control quantity are obtained by using the lower-level safe motion control model.

[0046] In this embodiment, a safe motion control model is established at the lower level based on equipment dynamics, collision avoidance constraints, and continuous control input, such as... Figure 2 As shown. For any device i Its state update can be written as in, Indicates continuous control input. This indicates a random perturbation. Indicates equipment i The dynamic state transition function.

[0047] The lower-level safety motion control model is as follows: in, Indicates the reference trajectory. Indicates the cost of security. This indicates the cost of controlling energy consumption. Indicates the weight of trajectory tracking error. Indicates the weight of security costs. This indicates the weight of the control input cost. Indicates the weight of energy or electricity cost. Indicates equipment k At any moment t The position vector, This indicates the minimum safe distance between the spare parts. This indicates the minimum safe power threshold at which the device is allowed to perform a task.

[0048] The reference trajectory and executable control variables output in this step are further fed back to steps S4 and S5 to correct the upper-layer service sequence and the middle-layer resource occupation time, thereby ensuring that the layered solution results can be implemented and executed at the physical level.

[0049] In one implementation, this step explicitly incorporates continuous motion control of the embodied intelligent device into the scheduling method so that the scheduling result has an executable motion control interface.

[0050] S7. Based on the disturbance-controlled collaborative rescheduling scheme, resource occupancy timing scheme, reference trajectory and control quantity, as well as the feedback data after execution, update the dynamic graph edge weights, risk costs and coordination parameters, and perform online learning enhancement correction on subsequent rolling windows to complete collaborative rescheduling.

[0051] In this embodiment, in open multi-device platforms or cross-brand collaborative scenarios, the device side may only report partial status, and there may even be deviations in capability, cost, or response time. Therefore, in an optional implementation, the present invention can embed mechanism design and payment layer outside the hierarchical solution framework, supplemented by an online learning enhancement module.

[0052] For any device i, a report type can be defined. With real type and set payment variables The utility of a device can be written as: in, Indicates the device is in the true type The actual cost of implementing upper-level allocation decisions. To ensure the constraints of the mechanism design, it can be further required to satisfy incentive compatibility, individual rationality, and budget balance: in, B Indicates the upper limit of the budget.

[0053] Meanwhile, due to the significant randomness and environmental non-stationarity of this problem, relying solely on a fixed-cost model is insufficient to maintain long-term performance. Therefore, an online learning enhancement module is introduced to continuously adjust the graph weights, risk costs, and coordination parameters. Let the learning parameter vector be denoted as... Then updates can be made based on the execution feedback: in, η Indicates the learning rate. Represents the loss function. This represents feedback data comprised of task completion rate, late arrivals, congestion levels, and power replenishment behavior. To avoid misinterpreting the actual device type... In a preferred embodiment, the online learning parameter vector can also be equivalently denoted as... For the sake of brevity, this embodiment will still be written as... However, it is related to the device type parameters. They differ from each other in their domain and physical meaning.

[0054] The updated parameters are used to adjust edge weights and risk costs, for example: The corrected results are then returned to steps S2, S4, and S5 for the next rolling cycle solution.

[0055] In one implementation, this step addresses the information distortion problem under open collaborative conditions through mechanism design and payment layer embedding, and continuously corrects the key cost parameters in the high-dimensional random mixed integer optimization model through an online learning enhancement module, thereby forming a closed-loop optimization mechanism of "hierarchical decomposition + rolling optimization + learning enhancement".

[0056] Example 2 To facilitate understanding of the deployment methods and applicable boundaries of this invention in intelligent manufacturing workshops, the following describes the possible scenario construction methods. These scenario construction methods can be used for solution verification in discrete event simulation platforms, digital twin workshops, and pilot demonstration lines, and can also serve as a modeling and implementation reference when this invention is integrated into actual manufacturing sites. Applicable scenarios preferably include manufacturing logistics scenarios characterized by frequent material delivery, significant shared channels, high material shortage costs, and significant energy replenishment constraints, such as lithium battery manufacturing, semiconductor packaging and testing, automobile assembly, 3C assembly, and large-scale warehousing and logistics. Figure 4 As shown.

[0057] First, at the scene modeling level, the workshop floor plan can be abstracted as a dynamic random graph composed of workstation nodes, buffer nodes, replenishment station nodes, energy replenishment nodes, and aisle edges. Multiple embodied intelligent devices or mobile replenishment devices are then configured to perform tasks such as delivery, cross-regional transfer, and anomaly verification on shared main aisles, branch aisles, and bottleneck intersections. On the basic task flow, high-priority replenishment order insertion events can be randomly injected, and operational constraints such as remaining safety stock at workstations, task time limits, main aisle capacity, equipment charge status, speed upper limits, and collision avoidance safety distances can be configured simultaneously to form scene input conditions consistent with the actual manufacturing cycle time. Preferably, the order insertion events can be generated according to a Poisson process, batch switching trigger mechanism, or abnormal material shortage trigger mechanism.

[0058] Second, at the method configuration level, corresponding input and output interfaces can be set around the seven steps of this invention: the status acquisition layer inputs the risk of material shortage at workstations, equipment status, shared bottleneck occupancy status, and energy replenishment resource status; the dynamic graph modeling layer outputs task reachability relationships, resource competition relationships, and risk exposure relationships; the upper optimization layer outputs task acceptance, task assignment, task reassignment, priority adjustment, and energy replenishment trigger results; the middle coordination layer outputs the shared bottleneck resource occupancy sequence and energy replenishment position allocation scheme; the lower control layer outputs reference trajectories and continuous control quantities; and the online learning layer performs rolling corrections on graph weights, risk costs, and coordination parameters based on execution feedback. In this way, this invention can be constructed as a closed-loop operation scenario that connects with MES, WMS, WCS, or on-site scheduling control systems.

[0059] Third, at the configuration comparison level, this invention can be compared with several typical baseline solutions under the same scenario conditions for a differentiated examination. These baseline solutions may include: a direct order insertion method that selects the nearest device to perform the replenishment task solely based on spatial distance or static travel time; a greedy reordering method that allows task reassignment but does not address shared bottleneck conflicts and critical time window coordination; a simplified hierarchical method that only has upper-level task acceptance and allocation but lacks mid-level resource timing coordination and lower-level safety motion control linkage; and a fixed-parameter rolling optimization method that does not include an online learning enhancement module. Through comparison, the differentiated mechanism of this invention under complex constraint scenarios can be more clearly demonstrated.

[0060] IV. At the implementation evaluation level, indicators such as the on-time completion rate of high-priority material replenishment tasks, waiting time at key workstations, number of material shortage events, average task lateness, shared main channel congestion rate, average queue length in bottleneck areas, number of equipment reassignments, average path length, average energy consumption, number of energy replenishment triggers, solution time, and throughput per unit time can be set according to the scenario. Among them, the on-time completion rate of high-priority material replenishment tasks and the number of material shortage events are used to reflect the ability to suppress material shortage risks; the shared main channel congestion rate and the average queue length in bottleneck areas are used to reflect the ability to coordinate shared bottlenecks; the number of equipment reassignments and the average task lateness are used to reflect the ability to coordinate and rearrange disturbances under control; and the average energy consumption, number of energy replenishment triggers, and solution time are used to reflect the feasibility of project deployment.

[0061] V. At the operational level, a baseline task flow can be generated first under conditions of no order insertion disturbance. Then, under the same initial task set and the same random seed, the arrival frequency of high-priority orders, the intensity of shared bottleneck occupancy, and the equipment load rate can be gradually increased to observe the performance of the invention under strong disturbance conditions. Furthermore, key parameters such as the safety stock threshold, rolling optimization window length, shared channel capacity, number of devices, and number of refueling stations can be changed to construct deployment scenarios under different manufacturing configurations. If necessary, a layered ablation scenario can be formed by removing the middle-layer resource timing coordination interface, the lower-layer security motion control interface, the mechanism design and payment layer, or the online learning enhancement module module module by module, to identify the boundary of each technical module's effect on the overall method.

[0062] VI. At the on-site implementation level, typical workstation clusters in actual manufacturing sites can be selected for gray-scale deployment. This invention can be integrated into existing manufacturing execution systems, warehouse management systems, or scheduling control systems. By reading real-time tasks, equipment locations, workstation material availability, and traffic occupancy status, a rolling solution is performed. The generated task rescheduling results, resource coordination results, and control results are then distributed to mobile replenishment equipment, collaborative robots, or unmanned forklifts for execution. Through the above scenario construction method, this invention can cover three application forms within a unified framework: simulation analysis, digital twin verification, and industrial site deployment. This more fully demonstrates its engineering applicability for high-priority replenishment, order insertion, sharing, bottleneck collaborative rescheduling in intelligent manufacturing workshops.

[0063] Example 3 In this embodiment, a high-priority replenishment and insertion order sharing bottleneck collaborative rescheduling system for intelligent manufacturing workshops includes: a risk coding module, a dynamic random graph construction module, an insertion order identification and candidate screening module, an upper-level optimization module, a middle-level coordination module, a lower-level control module, and an online learning module.

[0064] The risk coding module collects data on workstation material shortage risks, task pool status, equipment status, shared bottleneck resource occupancy status, and replenishment resource availability status, and generates risk-coded statuses. The dynamic random graph construction module unifies these data into a dynamic random graph. The order insertion identification and candidate filtering module, based on the risk-coded status and the dynamic random graph, identifies the order insertion trigger conditions for high-priority replenishment tasks and filters the set of candidate tasks that can enter the current scrolling window. The upper-level optimization module, based on the candidate task set, establishes an upper-level discrete optimization model around task acceptance, task assignment, task reassignment, and service sorting, and uses this model to obtain a perturbation-controlled collaborative rearrangement scheme. The middle-level coordination module, based on the perturbation-controlled collaborative rearrangement scheme, establishes a middle-level resource and timing coordination model for shared bottleneck resources, intersections, buffer positions, and replenishment positions, and uses this model to obtain a resource occupancy timing scheme. The lower-level control module, based on the resource occupancy timing scheme, establishes a lower-level safe motion control model considering equipment dynamics, collision avoidance constraints, and continuous control inputs. This model is then used to obtain executable reference trajectories and control variables. The online learning module, based on the disturbance-controlled cooperative rearrangement scheme, resource occupancy timing scheme, reference trajectory, control variables, and post-execution feedback data, updates the dynamic graph edge weights, risk costs, and coordination parameters. It also performs online learning enhancement and correction on subsequent rolling windows to complete the cooperative rescheduling.

[0065] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A high-priority replenishment order sharing bottleneck collaborative rescheduling method for a smart manufacturing workshop, characterized in that, Includes the following steps: Collect data on material shortage risks at workstations, task pool status, equipment status, shared bottleneck resource occupancy status, and energy replenishment resource availability status, and generate risk codes accordingly. The workstation material shortage risk, the task pool status, the equipment status, the shared bottleneck resource occupancy status, and the energy replenishment resource availability status are all uniformly constructed into a dynamic random graph; Based on the risk coding status and the dynamic random graph, identify the order insertion triggering conditions for high-priority replenishment tasks and filter the set of candidate tasks that can enter the current scrolling window. Based on the candidate task set, an upper-level discrete optimization model is established around task acceptance, task assignment, task reassignment and service sorting, and the perturbation-controlled collaborative rearrangement scheme is obtained using the upper-level discrete optimization model. Based on the disturbance-controlled collaborative rearrangement scheme, a mid-level resource and timing coordination model is established for shared bottleneck resources, intersections, buffer positions, and power replenishment positions, and the resource occupancy timing scheme is obtained using the mid-level resource and timing coordination model. Based on the resource occupancy timing scheme, a lower-level safe motion control model is established for equipment dynamics, collision avoidance constraints and continuous control input, and the executable reference trajectory and control quantity are obtained using the lower-level safe motion control model; Based on the perturbation-controlled collaborative rescheduling scheme, the resource occupancy timing scheme, the reference trajectory, and the control quantity, as well as the feedback data after execution, the dynamic graph edge weights, risk costs, and coordination parameters are updated, and the subsequent rolling window is enhanced and corrected through online learning to complete the collaborative rescheduling.

2. The high-priority replenishment order sharing bottleneck collaborative rescheduling method for intelligent manufacturing workshops according to claim 1, characterized in that, The risk coding status is: in, Indicates the risk coding status. This indicates the risk of production line stoppage due to material shortage. Indicates items with urgent time windows. Indicates task priority. β 1. β 2 and β 3 represents the weighting coefficient in the risk urgency model. j Indicates a task node. t Indicates the decision-making period.

3. The high-priority replenishment order sharing bottleneck collaborative rescheduling method for intelligent manufacturing workshops according to claim 1, characterized in that, The dynamic random graph is: in, Represents a dynamic random graph. Indicates the objects on site. This represents reachability, conflict, cooperation, and resource competition relationships. It represents travel time, congestion costs, risk exposure, collaboration benefits, and switching costs.

4. The high-priority replenishment order sharing bottleneck collaborative rescheduling method for intelligent manufacturing workshops according to claim 2, characterized in that, The candidate task set is as follows: in, Represents the current set of tasks. Indicates the threshold for triggering order insertion. Indicates task j At any moment t The deadline H Indicates the length of the scroll window.

5. The high-priority replenishment order sharing bottleneck collaborative rescheduling method for intelligent manufacturing workshops according to claim 4, characterized in that, The upper-level discrete optimization model is as follows: in, This represents the upper-level discrete decision vector. This indicates that the task accepts variables. Indicates the decision-making period t Internal mission j Is it by the equipment? i Accept, Indicates task j Compared to the previous rolling cycle, has a reassignment occurred? Indicates task j In the equipment i The current execution sequence order, Indicates equipment i Whether to enter the energy replenishment mode during this period, max represents finding the maximum value. Represents the mathematical expectation. T This indicates the upper bound of the time range for optimizing the summation. This indicates the benefits or value of the services received for completing the task. This indicates the amount of work completed on time. This indicates the cost of being late. This represents the cost of reassignment and sorting perturbations. This indicates the risk and cost of supply disruption. Indicates the cost of energy consumption. Indicates the weight of the cost of being late. Indicates the weights of reassignment and sorting perturbation costs. This indicates the weight of the risk and cost of material shortage. Indicates the weight of energy consumption cost. It represents a collection of embodied devices. Indicates equipment i During the period t The upper limit of the number of tasks that can be accepted. Indicates the remaining charge or state of charge. This indicates the power threshold that triggers the power replenishment. M Represents a constant.

6. The high-priority replenishment order sharing bottleneck collaborative rescheduling method for intelligent manufacturing workshops according to claim 5, characterized in that, The mid-level resource and timing coordination model is as follows: Where min represents finding the minimum value. Indicates the time variable at which entry occurs. Indicates the time of departure. Indicates the order of variables. Indicates the energy allocation variable. This indicates the cost of bottleneck channel congestion. This indicates the need to wait for the cost of dissemination. This indicates the cost of competing for a replacement energy position. This indicates the cost of congestion at critical workstations. Indicates the congestion cost weight. Indicates the weight of the cost of waiting for propagation. This indicates the weight of the cost of energy replenishment competition. This indicates the weight of the congestion cost at critical workstations. r Indicates energy replenishment resources. Indicates equipment k At any moment t Resources r The moment of departure Indicates equipment k At any moment t For shared resources or energy replenishment resources r The moment of entry, This represents a set of shared bottleneck resources. Represents the resource usage variable. Indicates the bottleneck resource capacity. Indicates the capacity of the filler potential. This represents a set of energy replenishment resources.

7. The high-priority replenishment order sharing bottleneck collaborative rescheduling method for intelligent manufacturing workshops according to claim 6, characterized in that, The lower-level safe motion control model is as follows: in, Indicates equipment i State variables, Indicates the reference trajectory. Indicates the cost of security. This indicates the cost of controlling energy consumption. Indicates the weight of trajectory tracking error. Indicates the weight of security costs. This indicates the weight of the control input cost. Indicates the weight of energy or electricity cost. Indicates continuous control input. This indicates a random perturbation. Indicates equipment i The dynamic state transition function, This represents the position vector of the equipment in the workshop coordinate system. Indicates equipment k At any moment t The position vector, This indicates the minimum safe distance between the spare parts. This indicates the minimum safe power threshold at which the device is allowed to perform a task.

8. A high-priority replenishment order sharing bottleneck collaborative rescheduling system for intelligent manufacturing workshops, wherein the system applies the method described in any one of claims 1-7, characterized in that, include: The module includes a risk coding module, a dynamic random graph construction module, an order insertion identification and candidate screening module, an upper-level optimization module, a middle-level coordination module, a lower-level control module, and an online learning module. The risk coding module is used to collect data on workstation material shortage risk, task pool status, equipment status, shared bottleneck resource occupancy status, and energy replenishment resource availability status, and to form a risk coding status. The dynamic random graph construction module is used to uniformly construct the workstation material shortage risk, the task pool status, the equipment status, the shared bottleneck resource occupancy status, and the energy replenishment resource availability status into a dynamic random graph; The order insertion identification and candidate screening module identifies the order insertion triggering conditions of high-priority replenishment tasks based on the risk coding status and the dynamic random graph, and filters the set of candidate tasks that can enter the current scrolling window. The upper-level optimization module establishes an upper-level discrete optimization model based on the candidate task set, focusing on task acceptance, task assignment, task reassignment, and service sorting, and uses the upper-level discrete optimization model to obtain a perturbation-controlled collaborative rearrangement scheme. The mid-level coordination module establishes a mid-level resource and timing coordination model for shared bottleneck resources, intersections, buffer positions, and energy replenishment positions based on the disturbance-controlled collaborative rearrangement scheme, and uses the mid-level resource and timing coordination model to obtain a resource occupancy timing scheme. Based on the resource occupancy timing scheme, the lower-level control module establishes a lower-level safe motion control model for equipment dynamics, collision avoidance constraints, and continuous control input, and uses the lower-level safe motion control model to obtain an executable reference trajectory and control quantity. The online learning module updates the dynamic graph edge weights, risk costs, and coordination parameters based on the perturbation-controlled collaborative rearrangement scheme, the resource occupancy timing scheme, the reference trajectory, the control quantity, and the feedback data after execution. It also performs online learning enhancement correction on subsequent scrolling windows to complete the collaborative rescheduling.

Images