An automatic cart dispatching method suitable for product assembly

Abstract

The application discloses an automatic transfer vehicle scheduling method suitable for product assembly, and relates to the fields of intelligent manufacturing and intelligent logistics scheduling. The method comprises the following steps: based on order information and process route information, determining an updated process route and a scheduling process node set, and fusing site state, map topology information, abnormal event information, visual information and tool information to generate a distribution task set; unifying a basic scheduling state, a process model and the distribution task set into a rule fact set, and performing rule matching and constraint reasoning with a rule base to construct a hard constraint set and a penalty cost function; under the hard constraint set, eliminating distribution tasks in the distribution task set to generate a candidate scheme set; and taking the minimum penalty cost function as a target, screening the candidate scheme set to obtain a scheduling scheme, and completing automatic transfer vehicle scheduling. The application can improve the scheduling quality and efficiency of product assembly.

Description

Technical Field

[0001] This application relates to the field of intelligent manufacturing and intelligent logistics scheduling technology, and in particular to an automated guided vehicle (AGV) scheduling method suitable for product assembly. Background Technology

[0002] In the assembly areas of products such as ejection seats, materials, specialized tools, and peripheral equipment are characterized by multiple categories, multiple batches, multiple workstations operating in parallel, and strong cycle time constraints. To ensure continuous operation and safety compliance at each workstation, Automated Guided Vehicles (AGVs) are typically used on-site to handle tasks such as material delivery, tool delivery, tool return, work-in-process transfer, and empty-load retrieval.

[0003] In many cases, most AGV scheduling methods primarily rely on the shortest distance or shortest travel time for order assignment, neglecting factors such as assembly process routes, dependencies between processes, mutually exclusive station resources, quality access control, tool specification verification, and on-site traffic control. Furthermore, visual inspection results, tool cabinet status, order cycle time, and anomaly handling information are often not uniformly integrated into the scheduling model. This leads to situations where defective materials may still be sent to workstations, tool mismatches are difficult to prevent in a timely manner, abnormal tasks are difficult to divert, and the quality traceability chain during execution is incomplete, resulting in low scheduling efficiency. Summary of the Invention

[0004] The purpose of this application is to provide an automated guided vehicle (AGV) scheduling method suitable for product assembly, which can improve the scheduling quality and efficiency of product assembly.

[0005] To achieve the above objectives, this application provides the following solution.

[0006] This application provides an automated guided vehicle (AGV) scheduling method applicable to product assembly, comprising: acquiring the basic scheduling status of product assembly; the basic scheduling status includes: AGV status, site status, map topology information, order information, process route information, abnormal event information, visual information, and tool information; updating the process route and the set of scheduled process nodes based on the order information and the process route information to determine the updated process route and the set of scheduled process nodes; based on the updated process route and the set of scheduled process nodes, integrating the site status, the map topology information, the abnormal event information, the visual information, and the tool information to generate a delivery task set; uniformly constructing the basic scheduling status, the process model, and the delivery task set into a rule fact set, and performing rule matching and constraint reasoning with a rule base to construct a hard constraint set and a penalty cost function; under the hard constraint set, eliminating delivery tasks in the delivery task set to generate a candidate solution set; and filtering the candidate solution set with the goal of minimizing the penalty cost function to obtain a scheduling scheme and complete the AGV scheduling.

[0007] According to the specific embodiments provided in this application, this application has the following technical effects: This application fundamentally solves the problem of insufficient consideration of factors by acquiring the basic scheduling status of product assembly, specifically including the status of automated guided vehicles (AGVs), site status, map topology information, order information, process route information, abnormal event information, visual information, and tool information. Then, based on the order information and process route information, it determines the updated process route and the set of scheduled process nodes, enabling the confirmation and updating of process routes and scheduled process nodes during the scheduling process, making the next task more specific. Furthermore, based on the updated process route and the set of scheduled process nodes, it integrates site status, map topology information, abnormal event information, visual information, and tool information to generate a delivery task set, making the generated delivery task set more consistent with the various subdivided tasks in the actual production process. Finally, it integrates the basic... The scheduling status, process model, and delivery task set are uniformly constructed into a rule fact set, and rule matching and constraint reasoning are performed with the rule base to build a hard constraint set and a penalty cost function. This enables the generation of prerequisites for subsequent optimization based on various information of product assembly, solving the problem of inaccurate scheduling caused by the inability to update the prerequisites for subsequent optimization in real time. Under the hard constraint set, delivery tasks in the delivery task set are eliminated to generate a candidate solution set. With the goal of minimizing the penalty cost function, the candidate solution set is screened to obtain a scheduling scheme, completing the scheduling of automated guided vehicles. First, the hard constraint set is used to eliminate unexecutable tasks, and then the penalty cost function is used to complete local optimization, finally obtaining an accurate scheduling scheme, improving the scheduling quality and efficiency of product assembly. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A flowchart illustrating the automated guided vehicle (AGV) scheduling method for product assembly provided in this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] This application belongs to the interdisciplinary field of discrete manufacturing process modeling, rule engine reasoning, path planning and traffic control, real-time scheduling of multiple AGVs and event rescheduling. It can be applied to business scenarios such as assembly and distribution, circulation of peripheral tools, tool delivery and return, and anomaly diversion and traceability. For example, in the ejection seat assembly site, this application can be used for the delivery and retrieval of materials, special tools and peripheral tools between multiple workstations, and to realize verification and transfer or isolation transfer when quality interception or tool anomaly occurs. In the automotive final assembly site, this application can be used for line-side delivery of components such as seats, dashboards, and doors, and can dynamically adjust vehicles, routes and arrival times according to workstation cycle time, site occupancy and road congestion. In the aerospace component assembly site, this application can be used for the delivery, return and traceability management of high-value parts and special tooling, thereby improving the scheduling quality and efficiency of product assembly.

[0012] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0013] In one exemplary embodiment, such as Figure 1 As shown, an automated guided vehicle (AGV) scheduling method suitable for product assembly is provided. This method is executed by computer equipment, specifically by a computer device such as a terminal or a server alone, or by both a terminal and a server. In this embodiment, the method is described using a server as an example, and includes the following steps S1 to S6.

[0014] Step S1: Obtain the basic scheduling status of product assembly; the basic scheduling status includes: automated guided vehicle status, site status, map topology information, order information, process route information, abnormal event information, visual information, and tool information.

[0015] As an feasible approach, the status of automated guided vehicles (AGVs) includes: AGV number, location, load, remaining battery power, task status, availability, and area permissions; site status includes: site number, site type, site location, concurrency capacity, operation time window, occupied status, and disabled status; map topology information includes: map nodes, roadside, road direction, speed limit, meeting point, level crossing, and charging point; order information includes: order number, product model or batch, target workstation, planned time, priority, and current order status; process route information includes: process route number, process version, operation number, operation sequence, required workstation, required materials, and required tools; abnormal event information includes: event type, occurrence time, occurrence location, affected objects, affected scope, and handling status; visual information includes: material code, inspection time, inspection location, inspection result, release status, and abnormality type; tool information includes: tool code, compatible operation, location, in-place status, specification status, and planned retrieval or return time.

[0016] Specifically, the task status in the automated guided vehicle (AGV) includes: idle, performing a task, and charging; the availability status includes: available, occupied, low battery, faulty, offline, and maintenance. The site status includes: site type includes: loading point, material warehouse, assembly station, buffer station, tool cabinet access point, quality inspection station, verification station, isolation area, storage point, and charging point; concurrency capacity indicates the number of tasks a site can simultaneously receive and process; the operation time window indicates the time interval during which a site is allowed to perform arrival, loading / unloading, handover, or waiting operations; the occupation status includes: idle, occupied, queued, and fully loaded; the disabled status indicates that the site is temporarily unable to receive tasks due to fault, manual locking, maintenance, or process change. In the map topology information, map nodes represent locations within the AGV's traversable road network. Sites access the map topology by binding one or more map nodes; road directions include: two-way, one-way, and prohibited. Order information comes from the MES (Manufacturing Execution System), where the current order status includes: not started, executing, paused, pending review, completed, and canceled. The process route information comes from the MES (Manufacturing Execution System). The process types include: material loading, component assembly, tool retrieval, functional testing, verification, isolation processing, and transfer to warehouse. The abnormal event information includes event types such as: congestion, AGV malfunction, site unavailable, road unavailable, temporary order insertion, process change, quality interception, verification failure, tool unavailable, and manual intervention; the handling status includes: pending confirmation, confirmed, processing, frozen, diverted, restored, and closed. The visual information includes release status: released, pending verification, and intercepted; the abnormality types include: material mismatch, quantity abnormality, appearance abnormality, labeling abnormality, and batch abnormality. The tool information includes on-site status: in-situ, borrowed, being returned, and lost; the specification status includes: qualified, pending verification, overdue, damaged, and mismatched.

[0017] Step S2: Update the process route and the set of scheduled process nodes based on the order information and process route information, and determine the updated process route and the set of scheduled process nodes.

[0018] As one feasible approach, step S2 specifically includes steps S21 to S23: Step S21: Construct pre- and post-order constraints, resource usage mutual exclusion constraints, and time window and cycle time constraints.

[0019] Specifically, three types of constraints are established. The first type is the pre- and post-order constraint, which limits a scheduling process node to start only after its preceding scheduling node has been completed. For example, the assembly process of a certain component can only start after the corresponding materials and special tools have been delivered. The second type is the resource occupancy mutual exclusion constraint, which limits the same workstation, buffer position, level crossing, meeting point, charging point, or tool cabinet access point from being occupied by conflicting tasks at the same time. The third type is the time window and cycle time constraint, which limits the tasks to be completed within the allowed time range to avoid workstation supply interruptions or early arrival causing congestion.

[0020] Step S22: Verify the process route information to obtain the updated process route.

[0021] Specifically, the process route is the standard execution sequence of an order from material loading, component assembly, functional testing, verification, isolation processing, to transfer and warehousing. The preferred method is to read the currently effective process version, operation number, sequence relationship, cycle time requirement, required workstations, and required resources from the MES (Manufacturing Execution System) to verify the current process route for the order and obtain the updated process route.

[0022] Step S23: Based on the updated process route, under the constraints of sequential order, mutual exclusion of resource occupancy, and time window and cycle time, the order information is decomposed to obtain the updated set of scheduling process nodes.

[0023] Specifically, after verifying and updating the process route, the order is broken down into a set of scheduling process nodes. Each set of scheduling process nodes contains multiple scheduling process nodes. Each scheduling process node includes a material delivery node, a tool delivery node, a tool return node, a reusable tool transfer node, a waiting node, a quality inspection node, a review node, and an isolated transfer node. Each scheduling process node is bound to its execution location, required resources, earliest start time, latest finish time, and preceding dependent nodes.

[0024] Step S3: Based on the updated process route and the set of scheduled process nodes, integrate site status, map topology information, abnormal event information, visual information and tool information to generate a delivery task set.

[0025] Specifically, the set of scheduling process nodes formed in step S2 is used as the basis for task generation. The scheduling process nodes that require the participation of automated guided vehicles (AGVs) are converted into delivery tasks. For process nodes that do not involve material delivery, tool delivery, tool return, circulation of reusable tools, quality inspection transfer, verification transfer, isolation transfer, or warehousing transfer, no delivery tasks for AGVs are generated.

[0026] Each delivery task in the delivery task set includes at least the task number, task type, associated order number, associated process number, source station, destination station, required resources, priority, earliest departure time, latest arrival time, release status, alternative station set, and execution status.

[0027] The task type, associated order number, associated process number, required resources, priority, earliest departure time, and latest arrival time are determined based on the updated process route and scheduling process node set in step S2. The source and destination stations are determined based on the execution positions bound to the scheduling process nodes in step S2, combined with the station status collected in step S1. The road network access location corresponding to the station is determined based on the map topology information collected in step S1. The release status is determined based on the visual information collected in step S1, including release status, pending review status, and interception status. Release status indicates that entry into the normal delivery process is allowed; pending review status indicates that the delivery target is limited to the review station or temporary buffer; interception status indicates that entry into the assembly station is prohibited, and the current delivery task at the entry station is frozen, transferred to the review station or isolation area. The alternative station set is determined based on the station status, map topology information, and abnormal event information collected in step S1. When the target site is disabled, fully loaded, unreachable, or affected by abnormal events, the system determines the alternative site based on the available sites and their road network access relationships. The execution status is determined based on the abnormal event information, site status, map topology information, visual information, and tool information collected in step S1, which is used to identify whether the task is executable, needs to be frozen, needs to be diverted, needs to be reassigned, or needs to be postponed.

[0028] Therefore, step S3 transforms the process nodes formed in step S2 into delivery tasks that can be executed by automated guided vehicles, and integrates the basic scheduling status collected in step S1 into the task attributes to form the delivery task set required for subsequent rule reasoning, candidate generation and rolling optimization.

[0029] Step S4: Construct a set of rule facts from the basic scheduling state, process model and delivery task set, and perform rule matching and constraint reasoning with the rule base to build a set of hard constraints and penalty cost function.

[0030] Specifically, the rule fact set is not newly collected data, but a structured expression of the output results of the aforementioned steps, used to describe the constraint relationships between vehicles, stations, roads, process nodes, delivery tasks, and abnormal events at the current scheduling moment.

[0031] The facts derived from step S1 include: the location, remaining battery power, current load, task status, area permissions, and availability of the automated guided vehicle (AGV); station concurrency capacity, operation time window, and disabled status; as well as the status of map nodes, roadside, one-way constraints, speed limits, meeting points, level crossings, and charging points. The facts derived from step S2 include: process nodes, process sequence relationships, cycle time requirements, required workstations, required resources, resource capacity constraints, and time window constraints. The facts derived from step S3 include: delivery task number, task type, source station, target station, required resources, priority, earliest departure time, latest arrival time, release status, alternative station set, and execution status. Abnormal event facts include: congestion, AGV malfunction, station unavailability, temporary order insertion, process changes, quality interception, and non-compliance during review.

[0032] As one feasible approach, step S4 specifically includes steps S41 to S44: Step S41: Construct a rule base; the rule base includes: process sequence rules, quality release rules, automated guided vehicle (AGV) capacity and status rules, site resource capacity rules, map traffic control rules, tool cabinet specification rules, time window and cycle time rules, and abnormal rescheduling rules.

[0033] Specifically, the rule selection in the rule base is defined according to the structure of "triggering condition - constraint action - rule priority - hit reason".

[0034] The process pre- and post-process rules are used to limit the start order between delivery tasks and assembly processes based on the process nodes and pre- and post-process relationships in the process model. The hard constraints generated when this rule is applied include: subsequent process nodes can only start after their preceding process nodes are completed; the corresponding assembly process cannot be released until the material delivery tasks and tool delivery tasks associated with it are completed; delivery tasks that do not satisfy their preceding dependencies cannot be scheduled as executable tasks. The soft constraint penalties for this rule include at least one or more of the following: order delay penalties, workstation supply disruption risk penalties, and lockout disturbance penalties, used to evaluate the delay of preceding tasks, workstation supply risks, and the disturbances caused by rescheduling to the existing process plan.

[0035] The quality release rule is used to determine whether materials are allowed to enter the assembly station, verification station, temporary buffer station, or isolation area based on visual inspection results, review results, and task release status. Hard constraints generated after this rule is triggered include: materials in an intercepted state are not allowed to enter the assembly station; materials in a pending review state are not allowed to directly enter the assembly station, but can only enter the verification station or temporary buffer station; normal delivery tasks corresponding to materials that fail review should be frozen or terminated; tasks that have not obtained release status cannot participate in scheduling as normal entry tasks. The soft constraint penalties corresponding to this rule include at least one or more of the following: order delay penalty, station supply interruption risk penalty, empty driving distance penalty, and lockout disturbance penalty, used to evaluate the impact of review waiting, isolation transfer, target site changes, and quality anomalies on existing plans.

[0036] The Automated Guided Vehicle (AGV) capability and status rules are used to filter candidate vehicles based on vehicle location, remaining battery power, current load, task status, area access permissions, and availability. Hard constraints generated upon rule fulfillment include: unavailable or faulty AGVs cannot execute new tasks; AGVs whose load capacity does not meet task requirements cannot be assigned a task; AGVs without access to the target area cannot enter the corresponding area; AGVs whose estimated remaining battery power does not meet battery redundancy requirements cannot execute corresponding delivery tasks; and AGVs performing uninterrupted tasks cannot be reassigned. The corresponding soft constraint penalties include at least one or more of the following: empty driving distance penalty, energy consumption and charging penalty, order delay penalty, and lockout disturbance penalty. These penalties are used to evaluate the impact of candidate vehicle distance, vehicle energy consumption, battery safety margin, charging insertion, and vehicle reassignment on the scheduling plan.

[0037] Site resource capacity rules are used to restrict tasks from entering corresponding sites based on site concurrency capacity, job time windows, disabled status, cache capacity, and resource occupation periods. Hard constraints generated after a rule is triggered include: disabled sites cannot be target sites; tasks exceeding the site's concurrency capacity or cache capacity limit cannot enter that site; resources such as workstations, cache slots, tool cabinet access points, and charging points cannot be repeatedly occupied by conflicting tasks within the same time period; candidate arrival times that do not meet the site's job time window cannot be used. The corresponding soft constraint penalties include at least one or more of the following: order delay penalties, workstation supply disruption risk penalties, congestion penalties, conflict risk penalties, and lockout disturbance penalties. These are used to evaluate site waiting times, resource conflicts, workstation supply risks, localized site congestion, and the impact of site changes on existing plans.

[0038] Map traffic control rules are used to filter candidate routes based on map nodes, roadside, one-way constraints, speed limits, restricted areas, meeting points, intersections, charging points, and safety distance requirements. Hard constraints generated after a rule is applied include: candidate routes must not violate one-way traffic; they must not pass through restricted areas or unusable roads; they must not exceed road speed limits; conflicting vehicles must not simultaneously occupy meeting points, narrow passages, or intersections; minimum safety distances must be maintained between vehicles; and intersections subject to time-limited restrictions must not be open during prohibited hours. The corresponding soft constraint penalties include at least one or more of the following: empty driving distance penalty, congestion penalty, conflict risk penalty, energy consumption and charging penalty, and order delay penalty. These penalties are used to evaluate detour distance, congestion level, vehicle conflict risk, route energy consumption, and task arrival delays caused by traffic control.

[0039] The tool cabinet specification rules are used to determine the feasibility of tool delivery, tool return, and spare tool scheduling based on tool codes, compatible processes, specification status, in-situ status, borrowed status, planned return time, and next scheduled retrieval time. Hard constraints generated when this rule is triggered include: no normal tool delivery task should be generated if a tool is not in place, its status is unqualified, or it does not match the current process; tools that have been borrowed and are not yet due for return should not be reassigned; conflicting retrieval tasks should not be scheduled when tool cabinet access points or tool buffer slots are occupied; and tools that do not meet specification verification should not be delivered to assembly stations. The corresponding soft constraint penalties include at least one or more of the following: order delay penalty, workstation supply interruption risk penalty, empty travel distance penalty, conflict risk penalty, and lockout disturbance penalty. These penalties are used to evaluate tool waiting, workstation supply interruptions due to tool mismatch, additional travel distance caused by tool return or spare tool scheduling, tool cabinet retrieval conflicts, and disturbances to existing plans caused by tool task adjustments.

[0040] The time window and takt time rules are used to adjust task feasibility, arrival time, and task priority based on the earliest departure time, latest arrival time, station operation time window, process takt time requirements, and workstation supply interruption threshold. The hard constraints generated after this rule is applied include: tasks cannot be executed earlier than the earliest departure time; task arrival times cannot fall within the station's inoperable time period; candidate solutions that cannot meet the latest arrival time under strict takt time requirements should be eliminated; and solutions that conflict with process takt time and cannot be corrected through waiting, reassignment, or route changes should not be adopted. The corresponding soft constraint penalties include at least one or more of the following: order delay penalty, workstation supply interruption risk penalty, empty driving distance penalty, congestion penalty, and lockout disturbance penalty. These penalties are used to evaluate additional travel and waiting caused by task delays, workstation supply interruptions, early or late scheduling, route changes due to avoiding congestion, and disturbances to the existing plan caused by time adjustments.

[0041] The abnormal rescheduling rule is used to determine the affected tasks, locked-in scope, and rescheduling backtracking steps when events such as congestion, Automated Guided Vehicle (AGV) malfunction, site unavailability, temporary order insertion, process change, quality interception, or non-compliance occur. The hard constraints generated after this rule is triggered include: malfunctioning AGVs cannot continue to undertake unfinished tasks; unavailable sites cannot be used as new target sites; congestion exceeding thresholds or closed road sections cannot be used as candidate paths; quality interception tasks cannot continue to enter assembly stations; locked tasks directly affected by safety, quality, equipment malfunction, or site unavailability events should be unlocked and included in the mandatory rescheduling scope. The corresponding soft constraint penalties include at least one or more of the following: order delay penalty, station supply interruption risk penalty, empty driving distance penalty, congestion penalty, conflict risk penalty, energy consumption and charging penalty, and lock-in disturbance penalty. These penalties are used to evaluate the delays, supply interruptions, detours, congestion, conflicts, energy consumption, and plan disturbances caused by abnormal events.

[0042] Step S42: Extract facts from the basic scheduling status, process model, and delivery task set to form a set of rule facts.

[0043] Step S43: Match the rule fact set with the rules in the rule base to construct a hard constraint set and a soft constraint set; the soft constraint set includes: order delay penalty, workstation supply interruption risk penalty, empty driving distance penalty, congestion penalty, conflict risk penalty, energy consumption and charging penalty, and lockout disturbance penalty.

[0044] Specifically, the rule engine matches the set of rule facts with the rules in the rule base. During matching and reasoning, the set of rule facts is first matched with the triggering conditions of each rule in the rule base. When a fact or combination of facts meets the rule triggering conditions, the corresponding rule is determined to be hit, and a set of hard constraints and a set of soft constraints are generated based on the constraint actions of the rule.

[0045] Step S44: Construct a penalty cost function based on the soft constraint set.

[0046] As an feasible approach, the expression for the penalty cost function is: .

[0047] in, Let the penalty cost function be used. , , , , , and These are all corresponding weighting coefficients, used to characterize the relative importance of different optimization objectives in the overall objective; For order delay penalties; This is a penalty item for the risk of supply disruption to workstations; This is a penalty for empty driving distance. This is a penalty item for congestion levels; This is a penalty item for conflict risk; Energy consumption and charging penalties are included. To lock the perturbation penalty item. Among them, , , , , , and Generally, the non-negativity condition is met.

[0048] As an implementable method, the order delay penalty is used to characterize the scheduling loss caused by a delivery task failing to arrive at the target workstation or resource point before the specified latest time. The expression for the order delay penalty is: .

[0049] in, The set of tasks within the scrolling window; For the task The estimated arrival time under the current scheduling scheme; For the task The latest allowed arrival time under the current scheduling scheme; This indicates that a penalty is only applied when a task times out; otherwise, the penalty value is zero. This means that the later the task arrives compared to the latest allowed arrival time, the greater the penalty for order delay.

[0050] The workstation supply interruption risk penalty item characterizes the risk of a supply interruption caused by materials or tools failing to arrive at the workstation before the risk point is reached. The expression for the workstation supply interruption risk penalty item is: .

[0051] in, This is a penalty item for the risk of supply disruption to workstations; For the task The critical moment of supply disruption risk for the corresponding workstation; the critical moment can be determined jointly by the moment when the safety stock is depleted and the moment of cycle time gap, and its expression is: .

[0052] in, This indicates the moment when the workstation's safety stock is depleted; This indicates the moment when a workstation experiences a shortage of materials or tools due to the pace of production. This indicates that the earlier of the two times is taken as the risk threshold. This means that if the delivery arrival time is later than the risk threshold, the larger the difference, the greater the penalty for supply disruption.

[0053] The empty-run distance penalty is used to reduce the proportion of vehicles operating without load, thereby improving the transportation utilization rate of AGVs. The expression for the empty-run distance penalty is: .

[0054] in, Number the vehicle; The number of vehicles participating in the dispatch; For vehicles The set of empty driving sections in the current plan; Number the empty driving section; For vehicles In the The unloaded travel distance on each empty road segment. This item means that the system aims to minimize the unloaded travel distance of the vehicle while meeting task requirements. If a normalized form is needed, it can also be written as: .

[0055] in, Indicates vehicle Total empty driving distance; Indicates vehicle Total driving distance; This represents the smallest positive number that prevents the denominator from being zero.

[0056] The congestion penalty term characterizes the congestion level of each key road segment in the road network during the statistical period, thereby minimizing the allocation of a large number of tasks to highly congested areas during optimization. The overall expression is: .

[0057] in, This represents the set of road segments involved in the calculation; Indicates the road segment number; Indicates road segment The congestion index is a cumulative assessment of the congestion level across different road sections; the more severe the congestion, the greater the total penalty.

[0058] In the congestion penalty item, the basic congestion coefficient of a road segment can be determined based on the ratio of average travel time to free travel time, and the formula is as follows: .

[0059] in, Indicates road segment The basic congestion coefficient; Indicates road segment Average travel time within the statistical period; Indicates road segment The theoretical travel time under free-flow conditions. This coefficient reflects the degree to which the traffic efficiency of a road segment decreases relative to the ideal state; the larger the ratio, the more severe the congestion.

[0060] To further describe the congestion situation, a queue length normalization index can be introduced, the formula of which is: .

[0061] in, Indicates road segment The normalized value of the queue length; This indicates the average queue length of the road segment within the statistical period; This indicates the maximum allowable queue length for that road segment or an empirically set queue normalization baseline. The higher this indicator, the more severe the traffic congestion on that road segment.

[0062] A waiting time normalization index can also be introduced to characterize the degree of vehicle stagnation on this road segment. The formula is as follows: .

[0063] in, Indicates road segment The normalized value of the waiting time; This indicates the average waiting time for this road segment within the statistical period; This represents the baseline value used for waiting time normalization. The larger this value is, the more severe the vehicle congestion on that road segment due to yielding, queuing, or conflicts.

[0064] Congestion calculations can also consider the number of vehicles passing through per unit time to characterize the traffic flow pressure of a road segment. The normalization formula can be written as: .

[0065] in, Indicates road segment The normalized value of the number of vehicles per unit time; Indicates the road segments passed through during the statistical period The number of vehicles; Indicates the length of the statistical period; This represents the normalized baseline value of the traffic capacity of this road segment per unit time. The larger this value is, the greater the traffic pressure the road segment experiences per unit time.

[0066] Taking into account the basic congestion coefficient, queue length, waiting time, and traffic flow pressure, the road segment congestion index can be expressed as: .

[0067] in, This represents the combination coefficient of the various components within congestion level, used to adjust the relative roles of average travel time, queue length, waiting time, and traffic intensity in congestion assessment. The congestion penalty term can be expanded; its expression is as follows: .

[0068] in, Number the road segment; The number of road segments involved in the calculation; , , and All are combination coefficients; For road section The basic congestion coefficient; For road section The normalized value of the queue length; For road section The normalized value of the waiting time; For road section The normalized value of the number of vehicles per unit time is given by the formula. This formula shows that the system can generate a congestion penalty value that guides route selection and task allocation by weighting and combining various congestion indicators.

[0069] If the internal combination coefficients of congestion degree are normalized, then the following conditions can still be met: . .

[0070] The above settings mean that each weight is used to represent the importance of the corresponding indicator and should not be negative; and when the sum of the weights within the congestion level is 1, it is convenient to unify congestion indicators from different sources into the same comprehensive evaluation framework.

[0071] The conflict backoff penalty term characterizes the waiting and backoff costs incurred by vehicles in situations such as meeting oncoming traffic, cutting in front of other vehicles, occupying intersections, and avoiding narrow passages. The backoff distance needs to be considered; the expression for the conflict risk penalty term is: .

[0072] in, , and All are conflict penalty coefficients; For vehicles The number of times waiting occurs due to conflicts during execution; For vehicles Total rollback time due to conflict rollback; For vehicles Total backtracking distance. This item represents the goal of optimizing scheduling to reduce delays and detours caused by vehicle conflicts.

[0073] The energy consumption and charging penalty terms are used to balance energy consumption, power safety margin, and charging task insertion frequency during task execution. The expressions for the energy consumption and charging penalty terms are: .

[0074] in, , and All are energy penalty coefficients; For vehicles The estimated energy consumption for performing the task within the current scrolling window; For vehicles The degree of insufficient safety margin of power supply; For vehicles The number of charging attempts inserted within the scrolling window. This item is used to prevent the solution from focusing solely on task completion while ignoring battery constraints, which could lead to the vehicle experiencing low battery risk or excessively frequent charging interruptions during execution.

[0075] The degree of insufficient safety margin of battery power can be calculated using the following formula: .

[0076] in, This indicates a pre-set safe power threshold; Indicates vehicle The system estimates the remaining battery power after completing the currently assigned tasks. If the estimated remaining battery power is higher than the safety threshold, this value is set to zero; if it is lower than the safety threshold, the shortfall is penalized, thus guiding the system to prioritize more reasonable charging and task allocation strategies.

[0077] The lock-and-disturbance penalty term is used to limit the extent of changes to the existing plan after an event-triggered rescheduling, thereby ensuring the continuity and stability of system execution. The expression for the lock-and-disturbance penalty term is: .

[0078] in, The set of tasks that allow for the evaluation of disturbances near the locked interval; , and All are disturbance penalty coefficients; For the task The amount of change in arrival or start time between the old and new schemes; For the task The change in the number of vehicles being executed; For the task The amount of change in the execution path. This item means that, while allowing for local adjustments, we should try to keep the established plan from changing drastically.

[0079] Among them, the change in the number of vehicles executing the task can be defined as a piecewise variable, used to determine whether the task has been reassigned to other vehicles, and its expression can be written as: .

[0080] This formula indicates that when the task If the executing vehicle differs before and after rescheduling, it is recorded as 1; if the executing vehicle remains unchanged, it is recorded as 0. It can be calculated based on the degree of difference between the old and new paths in terms of length, node sequence, or path segment composition; for Then it can be determined by the absolute value of the difference between the corresponding times of the old and new schemes.

[0081] Specifically, the hard constraint set represents the feasibility conditions that the scheduling scheme must meet, and eliminates infeasible candidate automated guided vehicles, candidate paths, candidate stations, or candidate start times. The penalty cost function constructed based on the soft constraint set represents the optimization preference for evaluating candidate schemes under the premise of satisfying hard constraints, and includes at least one or more of the following: order delay penalty, workstation supply interruption risk penalty, empty running distance penalty, congestion penalty, conflict risk penalty, energy consumption and charging penalty, and lockout disturbance penalty.

[0082] When multiple rules are triggered simultaneously, they are processed according to their priority. Rules related to safety, quality, equipment failure, and traffic control have higher priority than rules related to efficiency optimization. Hard constraint sets are used primarily for feasibility screening. Penalty cost functions are used to evaluate feasible candidate solutions. Each rule trigger is recorded, including the triggering fact, the triggered rule, the constraint action, the rule priority, the reason for the trigger, and the generated hard constraint set and penalty cost function, for subsequent execution display, anomaly tracing, and scheduling strategy optimization.

[0083] Step S5: Under the hard constraint set, eliminate delivery tasks in the delivery task set to generate a candidate solution set.

[0084] The candidate solution set includes multiple candidate solutions. Each candidate solution must include at least a delivery task, a candidate automated guided vehicle (AGV), a candidate route, and a candidate start time. Candidate AGVs should meet requirements for load capacity, interface compatibility, area access control, power redundancy, and availability. Candidate routes should meet requirements for one-way traffic, no-entry zones, speed limits, mutual exclusion of meeting points, time-limited crossing controls, and safe distances. Candidate start times should meet requirements for the completion time of preceding nodes, resource availability time, site operation time window, and task time window.

[0085] Step S6: With the goal of minimizing the penalty cost function, the candidate solution set is screened to obtain the scheduling scheme and complete the scheduling of the automated guided vehicle.

[0086] As an implementable approach, step S6 specifically includes: calculating the penalty cost function value of each candidate scheme in the candidate scheme set based on the penalty cost function; and determining the candidate scheme with the smallest penalty cost function value as the scheduling scheme.

[0087] Specifically, the scheduling scheme is a joint scheduling scheme of task, automated guided vehicle (AGV), path, and execution time, which includes at least the task number, the executing AAV, the pickup station, the target station, the planned path, the planned departure time, the planned arrival time, the resource occupation period, and the task execution order.

[0088] After obtaining the scheduling plan in step S6, closed-loop and event-driven rescheduling is executed. The scheduling plan generated in step S6 is sent out for execution, and feedback information such as arrival, pickup, release, waiting, charging, anomalies, detection results, and manual intervention is received. Feedback is used to update the basic scheduling status and generate task execution records, resource usage records, rule hit records, and scheduling decision records. When events such as congestion, automated guided vehicle (AGV) malfunction, site unavailability, temporary order insertion, process change, quality interception, or review failure occur, the system updates the basic scheduling status according to the event type and scope of impact, and re-executes steps S1 to S6 to obtain the rescheduled scheduling plan; among them, quality-related events are also used to update the task release status and generate review transfer or isolation transfer tasks.

[0089] During rescheduling, locked tasks not directly affected by security, quality, equipment failure, or site unavailability events remain unchanged; tasks directly affected by the aforementioned events are unlocked and included in the mandatory rescheduling scope. For affected tasks that have not yet started after the locked interval, reassignment, route change, time change, target site change, or cache site change can be performed; for tasks that have been picked up but not delivered, execution can be resumed using alternative sites, relay transfer, temporary caching, verification transfer, or isolation transfer methods.

[0090] To support audit traceability and continuous optimization, this application generates structured logs for each rule hit, candidate selection, optimization decision, and execution result. These logs include at least the following fields: Task ID, Process Node ID, Automated Guided Vehicle ID, Path ID, Planned / Actual Time, Hit Rule and Reason, Resource Consumption Conflict, Rescheduling Trigger Event, and Handling Result. Logs can be cached at the edge when the network is unavailable and re-uploaded after recovery. Key rule versions and handling results can be tamper-proofed and stored using signatures or hash chains.

[0091] Example 1 An assembly workshop needs to complete order WO-ES-001, the product of which is an ES-A type ejection seat assembly. The assembly process route for this order is as follows: .

[0092] in, Indicates the material warehouse. Indicates the material loading point. This indicates the skeleton assembly station. This indicates the guide rail assembly station. This indicates the visual quality inspection station. This indicates the latch assembly station. This indicates the point where finished products are received into the warehouse.

[0093] Guide rail assembly station A special torque tool, TL-17, is required and is currently stored in tool cabinet TG-01. The workshop is equipped with three AGVs: AGV-01, AGV-02, and AGV-03.

[0094] . This means the current scrolling window covers delivery tasks to be executed within the next 110 minutes.

[0095] Step 1: Acquire the basic scheduling state at time t. In the first step, the system acquires the basic scheduling state at the current scheduling time t.

[0096] 1.1 AGV Status Acquisition: At the current moment, the system has acquired the status of 3 AGVs. The system's preset safe power threshold is: .

[0097] Table 1 AGV Status Acquisition Results

[0098] As shown in Table 1, the remaining power of AGV-03 meets the following requirements.

[0099] .

[0100] Therefore, the AGV-03 is currently not suitable for long-distance delivery tasks, but only for short-distance movement tasks or priority charging tasks.

[0101] 1.2 Site Status Data Collection: The main sites involved in the current workshop include the material warehouse. Feeding point Frame assembly station Guide rail assembly station Visual quality inspection station Lock assembly station Finished product receiving point Tool cabinet access port TG-01, temporary buffer slot BS-1, and charging point .

[0102] Table 2. Site Status Collection Results

[0103] As shown in Table 2, the concurrency capacity of P3 is 1, therefore only one task is allowed to occupy P3 at the same time for arrival, loading / unloading or handover operations.

[0104] 1.3 Map Topology Information Collection: The system collects map topology information of the workshop road network, including data such as map nodes, roadsides, road directions, speed limits, meeting points, level crossings, and charging points.

[0105] Table 3 Map Topology and Road Segment Status

[0106] The basic congestion coefficient for R3 is as follows.

[0107] .

[0108] This indicates that the travel time on this main road section is significantly longer compared to the free-traffic state. Although the distance is long, the current level of congestion is low, so it can be considered as an alternative route.

[0109] 1.4 Order, Visual, and Tool Information Collection: Table 4 Order, Visual and Tool Information

[0110] 1.5 First step output: The first step output is the basic scheduling state set S(t).

[0111] Step 2: Verification of process route and breakdown of scheduling process nodes.

[0112] In the second step, based on the order information and process route information collected in the first step, the system reads the currently effective process version from the MES system and verifies the assembly process route corresponding to order WO-ES-001.

[0113] .

[0114] The process route includes steps such as skeleton loading, skeleton positioning and assembly, guide rail assembly, visual quality inspection, buckle assembly, finished product warehousing, and tool return.

[0115] 2.1 Process Route Verification: The process route for order WO-ES-001 can be represented as follows: .

[0116] Table 5. Verification Results of the Process Route

[0117] Due to the guide rail assembly process OP 30 The TL-17 tool is required, therefore the system generates additional nodes related to tool delivery and tool return.

[0118] 2.2 Decomposition of Scheduling Process Nodes: The system decomposes the process route into a set of scheduling process nodes: .

[0119] Table 6 Set of Scheduling Process Nodes

[0120] 2.3 Precedence and Postcedence Constraints: Based on the process route, the system establishes precedence and postcedence constraints: .

[0121] Because the guide rail assembly process requires tool TL-17, the tool delivery node... Semi-finished products must be delivered to Completed previously, i.e.: .

[0122] Tool return node It should be at the guide rail assembly node Execute upon completion, i.e.: .

[0123] 2.4 Resource Mutual Exclusion Constraints: Due to Its concurrency capacity is 1, meaning it cannot receive both tool delivery tasks and semi-finished product delivery tasks simultaneously. Therefore, for all arriving... Task and The resource usage should meet the constraint of non-overlapping: .in, and Representing tasks The start and end times of occupying the target site resources. Similarly, the concurrency capacity of the tool cabinet retrieval port TG-01 is 1, and tool retrieval tasks and tool return tasks must not occupy this retrieval port at the same time.

[0124] 2.5 Time Window and Cycle Constraints: Each scheduling process node has an earliest start time and a latest finish time. For any node, its start time... and completion time It should meet the following requirements: and .

[0125] For example, semi-finished product delivery nodes The earliest start time is 08:30, and the latest completion time is 08:50. If 08:00 is taken as the zero point, then: and .

[0126] 2.6 Second Step Output: The second step outputs the updated process route, the set of scheduled process nodes, pre- and post-order constraints, resource mutual exclusion constraints, and time window and cycle time constraints. This output serves as the basis for generating the delivery task set in the third step.

[0127] Step 3: Integrate multi-source information and generate a delivery task set In the third step, based on the set of scheduling process nodes formed in the second step, and integrating the site status, map topology information, abnormal event information, visual information, and tool information collected in the first step, the system converts the scheduling process nodes that require the participation of automated guided vehicles (AGVs) into delivery tasks. Workstation operation nodes that do not involve AGVs do not generate AGV delivery tasks.

[0128] 3.1 Delivery Task Set Generation: In this embodiment, the system generates a delivery task set: .

[0129] Table 7 Delivery Task Set

[0130] 3.2 Task Time Conversion: To facilitate subsequent objective function calculations, the system uses 08:00 as the time zero point. Therefore, the latest arrival time for each task can be converted as follows: , .

[0131] The critical moments for each task can be converted into: , .

[0132] because The task of returning tools does not directly correspond to the risk of workstation supply disruption, therefore it is not necessary to set it. Alternatively, the penalty for the risk of supply disruption may be counted as 0.

[0133] 3.3 Release Status Fusion: For tasks with a release status of "Release", the system allows them to enter the normal delivery process; for tasks with a release status of "Pending Inspection and Confirmation", the system retains its candidate branches when generating candidate solutions, but needs to confirm them based on the visual quality inspection results before issuing them for execution.

[0134] when When the visual inspection result is qualified, and The system will switch to a release status; if the visual inspection result is pending review or interception, the system will trigger a rescheduling in step seven.

[0135] 3.4 Third Step Output: The third step output is the delivery task set. Each task in this task set includes at least the task number, task type, associated order number, associated process number, source station, target station, required resources, priority, earliest departure time, latest arrival time, risk threshold time, release status, alternative station set, and execution status.

[0136] Step 4: Rule Base Reasoning and Constraint Generation: In this step, the system integrates the basic scheduling state formed in Step 1, the process model formed in Step 2, and the delivery task set formed in Step 3 into a unified set of rule facts. The data is then input into the rule engine for rule matching and constraint reasoning.

[0137] .in, Represents the basic scheduling state set. Represents the set of scheduling process nodes. Represents a set of delivery tasks. This represents a set of abnormal events. Rules in the rule base are defined according to the structure of "triggering condition—constraint action—rule priority—hit reason". After a rule is hit, the system generates the basis for calculating the penalty costs of hard and soft constraints.

[0138] 4.1 Rule Matching and Constraint Generation Table 8. Rule Matching Results and Constraint Generation

[0139] 4.2 AGV Capability and Status Rules: rule The triggering condition can be expressed as: .in, This represents the remaining battery power of vehicle k at time t. In this embodiment: , , Therefore, the rules If the condition is met, a hard constraint is generated: .in, Indicates task A set of candidate automated guided vehicles.

[0140] 4.3 Site Resource Capacity Rules: For any target site s, its resource occupancy constraints can be expressed as: .in, This indicates the number of times station s is occupied at time t. This represents the concurrency capability of site s. In this embodiment, the concurrency capability of P3 is: Therefore, the time periods occupied by tool delivery task T3 and semi-finished product delivery task T4 must not overlap: .

[0141] 4.4 Map Traffic Control Rules: For any candidate route If it includes restricted road sections If not, then this path is not feasible, that is: In this embodiment, both R3 and R4 meet the road direction and passability requirements, and therefore are not eliminated due to hard constraints. However, R3 is currently significantly congested, so its corresponding candidate solution will incur a high congestion penalty in step S6.

[0142] 4.5 Tool Specification Rules: The triggering conditions for tool specification rules can be expressed as follows: ,or: .

[0143] In this embodiment: , Therefore, the tool delivery task T3 was not removed.

[0144] 4.6 Fourth Step Output: The fourth step outputs the set of hard constraints. .in, Indicates power constraint. This indicates regional access constraints. This indicates site capacity constraints. Indicates traffic control constraints. Representation tool specification constraints, This represents the quality release constraint. The fourth step also outputs the set of soft constraints: This set of soft constraints is used in step six to comprehensively evaluate feasible candidate solutions.

[0145] Step 5: Generate a candidate solution set based on hard constraints: In step 5, the system receives the delivery task set T generated in step 3 and combines it with the set of hard constraints output in step 4. This generates candidate solutions for each delivery task to be performed. Each candidate solution includes at least the delivery task, the candidate automated guided vehicle (AGV), the candidate route, and the candidate start time. For each task... Its candidate solutions can be expressed as: .in, Indicates task The j-th candidate automated guided vehicle, Indicates the corresponding candidate path, Indicates the candidate's departure time. Indicates the arrival time of the candidate. Indicates task The number of candidate solutions.

[0146] 5.1 Candidate Automated Guided Vehicle (AGV) Screening: Candidate AGVs must meet requirements regarding load capacity, battery power, area access permissions, availability, and task status. For tasks... If the vehicle satisfy: , , , Then the vehicle Enter the mission Candidate AGV set In this embodiment, AGV-03 was excluded from multiple long-distance tasks because its remaining battery power was below the safe battery power threshold and it did not have W1 area permissions.

[0147] 5.2 Candidate Route and Start Time Filtering: Candidate routes must meet constraints regarding road direction, traffic restrictions, speed limits, mutually exclusive meeting points, level crossing occupancy, and safe distance. If a candidate route contains restricted or unusable sections, then: The candidate start time should satisfy the earliest departure time of the task, the completion time of the preceding node, the availability of site resources, and the target site's operation time window. For the task Its candidate start time It should meet the following requirements: If the task There exists a set of preceding tasks. Then it should also satisfy: Meanwhile, the estimated arrival time of the mission The target site operation time window should fall within: .

[0148] 5.3 Generation of Typical Candidate Solutions Table 9. Results of Generating Candidate Solutions for Typical Tasks

[0149] As shown in Table 10, the system first eliminates infeasible candidate solutions based on hard constraints; for candidate solutions that do not violate hard constraints but have risks of congestion, detours or waiting, the system retains them and evaluates them in the sixth step by using soft constraint penalty costs.

[0150] 5.4 Forming Candidate Scheme Combinations: The system combines feasible candidate schemes from each task to form several scheduling schemes. A candidate scheduling scheme can be represented as: .in, Let p represent the p-th candidate scheduling scheme. Indicates task The selected candidate solutions. In this embodiment, the system generates three main candidate solutions.

[0151] Table 10 Candidate Scheduling Scheme Combinations

[0152] 5.5 Fifth step output.

[0153] Step 5: Output the set of candidate solutions. Simultaneously, the system transmits the task arrival time, path set, vehicle allocation, estimated energy consumption, empty driving distance, number of conflict waiting times, and resource occupation periods from each candidate solution to step six for calculation: Thus, step five completes the transformation from the delivery task set to the feasible candidate solution set, providing input for step six, rolling time-domain incremental optimization.

[0154] Step 6: Rolling Incremental Optimization in the Temporal Domain: In step 6, the system receives the set of candidate solutions output in step 5 and their corresponding soft constraint penalty costs. Combining the rolling window, locking interval, and historical scheduling schemes, it performs incremental optimization on newly released tasks, tasks that have not yet started, tasks affected by events, and unlocked tasks that are allowed to be adjusted within the current rolling window.

[0155] In this embodiment, Plan-C is selected as the scheme to be calculated from the multiple feasible candidate schemes generated in the fifth step. The task arrangement for Plan-C is as follows: Performed by AGV-01 Performed by AGV-02 Performed by AGV-01 Executed and selected by AGV-01 Alternative channels Performed by AGV-02 Performed by AGV-01 Performed by AGV-02 Performed by AGV-02.

[0156] In this embodiment, the weighting coefficient is taken as: Each weight satisfies the non-negativity condition and also satisfies: This indicates that the above weights meet the normalization requirements.

[0157] (1) Calculation of order delay penalty: In this embodiment, the estimated arrival time and latest allowed arrival time of each task in Plan-C are as follows: . . . . . . . Substituting into the formula, we get: .

[0158] Therefore, none of the tasks in Plan-C exceeded their latest allowed arrival time, so the order delay penalty is 0.

[0159] (2) Calculation of risk penalty for workstation supply interruption: In this embodiment, the critical time of risk for each task has been determined based on the process cycle time and on-site inventory status. The expected arrival time and critical time of risk for each task in Plan-C are as follows: . . . . . . .because The task of returning tools does not directly correspond to the risk of supply disruption at the workstation; therefore, the penalty for this risk is set to 0. Substituting into the formula, we get: Therefore, it can be seen that in Plan-C... , and There is a slight delay beyond the critical risk point, therefore the penalty for the workstation supply interruption risk is 5.

[0160] (3) Calculation of empty driving distance penalty: In this embodiment, the following is taken: The empty driving distance and total driving distance of each AGV in Plan-C are as follows: .

[0161] Substituting into the formula, we get: Therefore, the empty driving distance penalty for Plan-C is 0.3935.

[0162] (4) Congestion penalty calculation: In this embodiment, the internal combination coefficient of congestion is taken as: And it satisfies: Plan-C chose a route that avoided areas with high levels of congestion. The main corridor, primarily traversing the following sections: .

[0163] for Given: Substituting into the formula, we get: ;for Given: Substituting into the formula, we get: ;for Given: Substituting into the formula, we get: ;for Given: Substituting into the formula, we get: ;for Given: Substituting into the formula, we get: ;for Given: Substituting into the formula, we get: ;for Given: Substituting into the formula, we get: .

[0164] Therefore, the congestion penalty is: Considering rounding, we take: Therefore, Plan-C selects... Alternative lanes reduce congestion penalties.

[0165] (5) Calculation of conflict risk penalty: In this embodiment, the following is taken: In Plan-C, AGV-01 and AGV-02 intersect at Exit Q1. There was one instance of mistimed passage nearby, causing AGV-01 to wait once, but it did not reverse. The conflict data for each vehicle is as follows: , , .

[0166] Substituting into the formula, we get: Therefore, the conflict risk penalty for Plan-C is 2.0.

[0167] (6) Calculation of energy consumption and charging penalty: In this embodiment, the following is taken: In Plan-C, the energy consumption, battery power, and number of charging cycles for each vehicle are as follows: , , , , , .

[0168] Substituting into the formula, we get: Therefore, the energy consumption and charging penalty for Plan-C is 18.0. Among them, AGV-03 was assigned to the charging task due to low battery, so the charging insertion penalty and insufficient battery penalty are included.

[0169] (7) Calculation of lock-in disturbance penalty: This embodiment is the initial scheduling process. There are currently no issued and locked historical plans, therefore: Substituting into the formula, we get: Therefore, the lock disturbance penalty term in the initial scheduling phase is 0.

[0170] (8) Calculation of the overall objective function value: Based on the above calculation results, the penalty values ​​for each item in Plan-C are: .

[0171] Substitute it into the synthesis objective function: .

[0172] Therefore, the comprehensive objective function value of Plan-C is: .

[0173] (9) Comprehensive comparison of candidate schemes: To determine the optimal scheduling scheme within the current rolling window, the system performs the same penalty cost calculation for Plan-A, Plan-B, and Plan-C. In Plan-A, AGV-01 serves as the main vehicle. choose Main passage. Due to The congestion level is high, and there are many instances of vehicle conflicts and waiting times. Calculations show that: Substituting into the comprehensive objective function, we get: .

[0174] In Plan-B, AGV-02 undertakes more tasks and has a lower risk of conflict, but its empty runs and energy consumption are relatively high. Calculations show that: , Substituting into the comprehensive objective function, we get: .

[0175] In Plan-C, AGV-01 and AGV-02 work together to perform tasks. choose The alternative routes, which achieve a relatively balanced result among the risks of supply disruption, congestion, conflict, and energy consumption, are calculated as follows: .therefore: Right now: .

[0176] Therefore, the system selects Plan-C as the optimal scheduling scheme within the current rolling window.

[0177] (10) Step 6 Output: Step 6 outputs the joint scheduling scheme of task-automatic transport vehicle-path-execution time in the current scrolling window. Based on the above calculation results, the system determines Plan-C as the optimal scheduling scheme, and its comprehensive objective function value is: .

[0178] This solution, while meeting hard constraints, takes into account order timeliness, continuous supply to workstations, vehicle empty driving distance, route congestion, vehicle conflict risk, energy consumption and charging arrangements, and plan disturbance control. Since this embodiment is in the initial scheduling phase, the lock disturbance penalty is 0. In the subsequent seventh step, if events such as congestion, station unavailability, automated guided vehicle malfunction, quality interception, or review failure occur, the system will update the status facts and recalculate the corresponding lock disturbance penalty to achieve event-driven rolling time-domain incremental optimization.

[0179] Step 7: Execute closed-loop and event-driven rescheduling: 7.1 Scheduling Scheme Issuance and Status Feedback: In this embodiment, the optimal scheme determined in step six is ​​Plan-C. After the system issues this scheme, it continuously receives information to update the basic scheduling status and task execution status. For example, after AGV-01 completes T1, the system updates the T1 status to "completed" and releases the corresponding station resources.

[0180] 7.2 Abnormal Event Identification and Rescheduling Trigger: When the system detects events such as congestion, automated guided vehicle (AGV) malfunction, site unavailability, road unavailability, temporary order insertion, process change, quality interception, review failure, or tool unavailability, it first determines the scope of the event's impact and updates the basic scheduling status. Then, using the updated basic scheduling status as input, steps one through six are re-executed to obtain the rescheduled scheduling plan. For quality interception or review failure events, the system also updates the task release status and generates a review transfer task or an isolation transfer task.

[0181] In this embodiment, it is assumed that the guide rail assembly station P3 is temporarily unavailable at 08:36. At this time, T1, T2, and T3 have been completed, T4 has picked up the goods but has not yet delivered them to P3, and T5 and T8 have not yet started. Since T4, T5, and T8 are all related to P3, the system identifies them as affected tasks, updates the status of P3 to unavailable, and then re-executes steps one through six with the updated basic scheduling status.

[0182] 7.3 Local rescheduling: During rescheduling, completed tasks remain unchanged; unaffected short-term tasks that have already been issued will, in principle, maintain their original plans; tasks directly affected by abnormal events will be unlocked and included in the scope of mandatory rescheduling.

[0183] In response to the temporary unavailability of P3, the system temporarily relocates the target station of the ongoing T4 task from P3 to the buffer station BS-1. AGV-01 then delivers the semi-finished products to BS-1 for temporary storage. Once P3 is restored, subsequent transfer tasks from BS-1 to P3 are generated or released. The T5 quality inspection transfer task and the T8 tool return task are postponed accordingly. T6 and T7, which are not directly affected, maintain their original vehicles and routes as much as possible, adjusting departure times only based on the completion time of preceding tasks. This approach avoids vehicles carrying materials waiting for extended periods near P3 and reduces the risk of congestion at station entrances and route conflicts.

[0184] 7.4 Recording and Output: The system records the execution process and rescheduling process, including the task number, the executing automated guided vehicle, the original target station, the adjusted target station, the original planned time, the adjusted planned time, the hit rule, the cause of the exception, and the handling method.

[0185] Taking T4 as an example, its record is as follows: At 08:36, the P3 station was detected to be unavailable, and the system hit the station unavailable rule and the abnormal rescheduling rule; the original target station was P3, and the adjusted target station is BS-1; the executing vehicle is still AGV-01; the task status changed from "in execution" to "sent to buffer bit".

[0186] The seventh step ultimately outputs task execution records, resource usage records, abnormal event records, rule hit records, and rescheduling plans. Thus, the system completes closed-loop control from scheduling plan issuance, on-site feedback collection, anomaly identification, local rescheduling to execution traceability.

[0187] The beneficial effects of the automated guided vehicle (AGV) scheduling method for product assembly proposed in this application are mainly reflected in the following aspects: 1. In acquiring the basic scheduling status of product assembly, specific data was obtained including the status of the automated guided vehicle (AGV), site status, map topology information, order information, process route information, abnormal event information, visual information, and tool information. By introducing more comprehensive scheduling status information, the problem of insufficient consideration of factors was fundamentally solved. Specifically, order information and process route information were used to obtain the updated process route and scheduling process node set, enabling the confirmation and updating of the process route and scheduling process nodes during the scheduling process, making the next task more explicit. Since the generation of the delivery task set is based on the updated process route and scheduling process node set, and integrates site status, map topology information, abnormal event information, visual information, and tool information, the generated delivery task set is more consistent with the actual production process. The system breaks down tasks into smaller, more manageable steps. It then unifies the basic scheduling state, process model, and delivery task set into a rule-fact set, and performs rule matching and constraint reasoning with a rule base to construct a hard constraint set and a penalty cost function. This generates prerequisites for subsequent optimization based on various product assembly information, solving the problem of inaccurate scheduling caused by the inability to update prerequisites for subsequent optimization in real time. Under the hard constraint set, delivery tasks in the delivery task set are eliminated to generate a candidate solution set. The candidate solution set is then filtered to minimize the penalty cost function, resulting in a scheduling scheme for automated guided vehicles (AGVs). By first eliminating unexecutable tasks through the hard constraint set and then performing local optimization through the penalty cost function, an accurate scheduling scheme is obtained, improving the scheduling quality and efficiency of product assembly.

[0188] 2. Multi-AGV collaborative scheduling is achieved through a process of "process modeling—multi-source demand fusion—rule reasoning—candidate generation—rolling optimization—execution closed loop—event rescheduling". Process modeling expresses the sequence of operations, cycle time, and resource usage; multi-source demand fusion unifies MES material / tool ​​requirements, tool cabinet specifications and retrieval plans, and visual inspection release results into delivery tasks; rule reasoning generates hard / soft constraints and interpretable decision suggestions; rolling optimization obtains near-optimal assignment and path timing schemes under real-time constraints; and event rescheduling quickly restores feasible and stable execution plans under dynamic disturbances. To balance configurable on-site strategies with real-time algorithm performance, this invention preferably employs "rule-driven candidate shrinkage + rolling temporal incremental optimization": the rule engine first eliminates infeasible solutions during the candidate generation stage and imposes penalties on soft constraints; the optimizer incrementally solves only unstarted tasks within the rolling window and sets a locking interval for issued short-term plans, thereby reducing recalculation disturbances.

[0189] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0190] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for scheduling automated guided vehicles (AGVs) suitable for product assembly, characterized in that, The automated guided vehicle (AGV) scheduling method applicable to product assembly includes: Obtain the basic scheduling status of product assembly; the basic scheduling status includes: automated guided vehicle status, site status, map topology information, order information, process route information, abnormal event information, visual information, and tool information; Based on the order information and the process route information, the process route and the set of scheduled process nodes are updated to determine the updated process route and the set of scheduled process nodes. Based on the updated process route and the set of scheduled process nodes, the delivery task set is generated by integrating the site status, the map topology information, the abnormal event information, the visual information, and the tool information. The basic scheduling state, the process model, and the delivery task set are uniformly constructed into a rule fact set, and rule matching and constraint reasoning are performed with the rule base to construct a hard constraint set and a penalty cost function. Under the set of hard constraints, delivery tasks in the delivery task set are eliminated to generate a set of candidate solutions; With the goal of minimizing the penalty cost function, the candidate scheme set is filtered to obtain a scheduling scheme, and the automated guided vehicle (AGV) scheduling is completed.

2. The automated guided vehicle (AGV) scheduling method for product assembly according to claim 1, characterized in that, The automated guided vehicle (AGV) status includes: AGV number, location, load, remaining battery power, task status, availability, and area permissions; the station status includes: station number, station type, station location, concurrency capacity, operation time window, occupied status, and disabled status; the map topology information includes: map nodes, roadside, road direction, speed limit, meeting point, level crossing, and charging point; the order information includes: order number, product model or batch, target workstation, planned time, priority, and current order status; the process route information includes: process route number, process version, operation number, operation sequence, required workstation, required materials, and required tools; the abnormal event information includes: event type, occurrence time, occurrence location, affected objects, affected scope, and handling status; the visual information includes: material code, detection time, detection location, detection result, release status, and abnormality type; the tool information includes: tool code, compatible operation, location, in-place status, specification status, and planned retrieval or return time.

3. The automated guided vehicle (AGV) scheduling method for product assembly according to claim 2, characterized in that, Based on the order information and the process route information, the updated process route and the set of scheduled process nodes are determined, specifically including: Construct pre- and post-order constraints, resource usage mutual exclusion constraints, and time window and cycle time constraints; The process route information is verified to obtain an updated process route; Based on the updated process route, under the constraints of the preceding and following sequences, the mutual exclusion constraints of resource occupancy, and the constraints of time window and cycle time, the order information is decomposed to obtain the updated set of scheduling process nodes.

4. The automated guided vehicle (AGV) scheduling method for product assembly according to claim 1, characterized in that, The basic scheduling state, the process model, and the delivery task set are uniformly constructed into a rule-fact set, and then matched and constrained inferences are performed with the rule base to construct a hard constraint set and a penalty cost function, specifically including: Construct a rule base; the rule base includes: process sequence rules, quality release rules, automated guided vehicle (AGV) capacity and status rules, site resource capacity rules, map traffic control rules, tool cabinet specification rules, time window and cycle time rules, and abnormal rescheduling rules; Facts are extracted from the basic scheduling state, the process model, and the delivery task set to form a set of rule facts; The set of rule facts is matched with the rules in the rule base to construct a set of hard constraints and a set of soft constraints. The set of soft constraints includes: order delay penalty, workstation supply interruption risk penalty, empty driving distance penalty, congestion penalty, conflict risk penalty, energy consumption and charging penalty, and lockout disturbance penalty. Based on the set of soft constraints, a penalty cost function is constructed.

5. The automated guided vehicle (AGV) scheduling method for product assembly according to claim 1, characterized in that, With the goal of minimizing the penalty cost function, the candidate solution set is filtered to obtain a scheduling scheme, specifically including: Based on the penalty cost function, calculate the penalty cost function value for each candidate solution in the candidate solution set; The candidate scheme with the smallest penalty cost function value is determined as the scheduling scheme.

6. The automated guided vehicle (AGV) scheduling method for product assembly according to claim 1, characterized in that, The expression for the penalty cost function is: ； in, Let the penalty cost function be used. , , , , , and These are all corresponding weighting coefficients; For order delay penalties; This is a penalty item for the risk of supply disruption to workstations; This is a penalty for empty driving distance. This is a penalty item for congestion levels; This is a penalty item for conflict risk; Energy consumption and charging penalties are included. To lock the perturbation penalty item.

7. The automated guided vehicle (AGV) scheduling method for product assembly according to claim 6, characterized in that, The expression for the order delay penalty is: ； in, For order delay penalties; The set of tasks within the scrolling window; For the task The estimated arrival time under the current scheduling scheme; For the task The latest allowed arrival time under the current scheduling scheme; The expression for the workstation supply interruption risk penalty item is: ； in, This is a penalty item for the risk of supply disruption to workstations; For the task The critical moment of supply disruption risk for the corresponding workstation; The expression for the empty driving distance penalty is: ； in, This is a penalty for empty driving distance. Number the vehicle; The number of vehicles participating in the dispatch; For vehicles The set of empty driving sections in the current plan; Number the empty driving section; For vehicles In the The distance traveled by empty vehicles on each empty driving section; The expression for the congestion penalty term is: ； in, This is a penalty item for congestion levels; Number the road segment; The number of road segments involved in the calculation; , , and All are combination coefficients; For road section The basic congestion coefficient; For road section The normalized value of the queue length; For road section The normalized value of the waiting time; For road section The normalized value of the number of vehicles per unit time; The expression for the conflict risk penalty term is: ； in, This is a penalty item for conflict risk; , and All are conflict penalty coefficients; For vehicles The number of times waiting occurs due to conflicts during execution; For vehicles Total rollback time due to conflict rollback; For vehicles Total rollback distance; The expression for the energy consumption and charging penalty term is: ； in, Energy consumption and charging penalties are included. , and All are energy penalty coefficients; For vehicles The estimated energy consumption for performing the task within the current scrolling window; For vehicles The degree of insufficient safety margin of power supply; For vehicles The number of charging cycles inserted within the scrolling window; The expression for the lock-in perturbation penalty term is: ； in, To lock the perturbation penalty item; The set of tasks that allow for the evaluation of disturbances near the locked interval; , and All are disturbance penalty coefficients; For the task The amount of change in arrival or start time between the old and new schemes; For the task The change in the number of vehicles being executed; For the task The amount of change in the execution path.

Images