A sorting cooperative control method and device for a multi-agent system, electronic equipment, and storage medium

Abstract

The present application relates to the technical field of industrial automation, and in particular to a sorting cooperative control method and device for a multi-agent system, electronic equipment and a storage medium, unified description and state synchronization of multiple resources are realized by constructing a composite data model; through integrating safety and energy consumption constraints in a rolling time window for joint solving, cooperative optimization and conflict prevention of tasks, paths and time are realized; by labeling commitment levels and setting frozen boundaries for instructions, re-scheduling oscillation is effectively suppressed, and near-end execution stability is ensured; decoupling communication is realized through an event message bus, and local adjustment and real-time feedback are carried out by each agent within the space-time constraints, forming an adaptive closed-loop control, thereby improving the throughput efficiency, running stability and scalability of the system in a strongly coupled and strongly disturbed sorting scene.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to a sorting collaborative control method, device, electronic device, and storage medium for multi-agent systems. Background Technology

[0002] In the fields of intelligent manufacturing and logistics sorting, the collaborative scheduling of multi-agent systems (such as AGVs, industrial robots, and vision inspection units) is key to achieving high-throughput, low-latency production. Currently, mainstream solutions mainly include centralized scheduling, distributed negotiation, time-triggered rescheduling, and digital twin-driven methods. Centralized scheduling relies on a central controller to decompose tasks based on preset rules, resulting in a clear structure but insufficient resource coupling modeling and delayed state updates. Distributed negotiation allocates tasks through a bidding mechanism, which, while scalable, is prone to global consistency issues and deadlock risks. Time-triggered rescheduling performs static or heuristic planning within a fixed time window, exhibiting poor adaptability to order fluctuations and real-time disturbances, easily leading to unstable cycle times. While digital twin-driven methods can be simulated and verified in advance, the latency and execution deviations caused by virtual-real mapping make it difficult to directly support highly dynamic real-time scheduling. These existing technologies generally suffer from problems such as fragmented perception-decision-execution links, coarse characterization of multi-resource temporal constraints, weak conflict resolution and deadlock prevention capabilities, and insufficient system scalability, making it difficult to achieve stable, efficient, and traceable collaborative control in sorting scenarios characterized by "strong coupling, strong constraints, and strong disturbances of multiple resources".

[0003] The main shortcomings of existing technologies are as follows: scheduling decisions rely on lagging and coarse-grained state information, making it difficult to build a unified real-time system view; the occupancy-release relationship between multiple resources (such as AGVs, robots, workstations, and vision windows) is not accurately modeled, leading to implicit conflicts and resource starvation; conventional conflict avoidance strategies lack global liveness guarantees, resulting in a high risk of deadlock; frequent rescheduling leads to system oscillations when dealing with disturbances such as grasping failures and equipment micro-faults; system integration relies heavily on tightly coupled architectures and proprietary protocols, resulting in high expansion and maintenance costs; the optimization objective is singular, lacking a multi-indicator configurable trade-off mechanism; and the end-to-end observability and quality traceability capabilities are insufficient, hindering process diagnosis and continuous improvement.

[0004] Therefore, there is an urgent need for a multi-agent sorting, scheduling, and collaborative control method that can incorporate data models, achieve real-time collaboration, be robust against disturbances, and support full-link traceability. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a sorting collaborative control method, device, electronic device, and storage medium for multi-agent systems.

[0006] In a first aspect, embodiments of the present invention provide a sorting cooperative control method for multi-agent systems, the method comprising: A composite data model is constructed by the central scheduling unit. The composite data model is jointly constructed through task definition, resource definition, and state diagram of the temporal occupancy relationship between records. Within the rolling time window, the central scheduling unit integrates safety constraints and energy consumption constraints for joint solution to generate global scheduling instructions. The global scheduling instructions include task assignment, path segment occupancy time window and equipment execution sequence, and mark the commitment level and set the freeze boundary for the execution fragments in the instructions. Global scheduling instructions are issued via the event message bus; Each intelligent agent receives and executes instruction fragments assigned to it through its local control unit, and adjusts its trajectory or actions within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information. Each intelligent agent feeds back the execution status of instruction segments, resource occupancy status, and abnormal events to the central scheduling unit in real time through the event message bus.

[0007] In conjunction with the first aspect, the task definition includes task identifier, type, associated workstation, material identifier, and priority attribute; the resource definition abstracts mobile robots, industrial robots, vision windows, fixture stations, and buffer areas into occupiable entities, whose attributes include resource identifier, type, capability set, and safety domain label; the state graph records the entry timestamp, release timestamp, and reentrancy flag of resources occupied by specific tasks through directed edges.

[0008] In conjunction with the first aspect, the steps by which the central scheduling unit, within a rolling time window, integrates safety constraints and energy consumption constraints for joint solution to generate global scheduling instructions include: The state graph is updated based on the received sensing events, and the security domain constraints and energy consumption constraints are integrated into the composite data model to form a computable set of optimization constraints. Within the rolling time window, based on the composite data model and constraint set, the task assignment, path segment occupancy time window, and device execution timing are jointly solved; The global scheduling instructions are generated based on the solution results, and each execution segment in the instructions is marked with its commitment level. At the same time, the freeze boundary is set for the current instruction version based on the system state.

[0009] Combining the first aspect, the steps for jointly solving task assignment, path segment occupancy time windows, and device execution timing based on a composite data model and constraint set include: The objective function for optimization is a multi-index interpretable scoring function; the multi-index includes at least system throughput, work-in-process inventory, mobile robot idle rate, and fixture switching frequency.

[0010] In conjunction with the first aspect, within the rolling time window, the steps of jointly solving for task assignment, path segment occupancy time window, and device execution timing based on the composite data model and constraint set also include: A conflict resolution strategy based on segment reservation and priority inheritance is adopted; The strategy includes: allocating exclusive time windows for the mobile robot's path segments and workspace segments; dynamically allocating passage priorities based on task urgency and blocking cost when resource contention is detected; and passing high priority to low priority along the waiting chain if chained waiting is formed.

[0011] In conjunction with the first aspect, the commitment level is used to identify the variability of execution segments in subsequent optimizations, including high commitment, medium commitment, and low commitment; the freeze boundary is the dividing point on the time axis, and execution segments before the freeze boundary are marked as high commitment to ensure execution stability.

[0012] In conjunction with the first aspect, the steps of each intelligent agent receiving and executing instruction fragments assigned to it through its local control unit, and adjusting its trajectory or action within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information, also include: When an exception event is received via the event message bus, it is determined whether the exception event affects the execution segment marked as high commitment within the freeze boundary; If so, a re-verification process is triggered for the affected execution fragment. The re-verification process includes reachingability and mutual exclusion checks, and if the checks fail, equivalent resource replacement is prioritized. If not, then subsequent scheduling instructions will be rearranged normally outside the frozen boundary.

[0013] Secondly, embodiments of this application also provide a sorting collaborative control device for a multi-agent system, the device comprising: The building module is used by the central scheduling unit to build a composite data model. The composite data model is jointly built through task definition, resource definition, and state diagram of the temporal occupancy relationship between records. The generation module is used by the central scheduling unit to jointly solve for safety constraints and energy consumption constraints within a rolling time window, and generate global scheduling instructions. The global scheduling instructions include task assignment, path segment occupancy time window and equipment execution sequence, and mark the commitment level and set the freeze boundary for the execution fragments in the instructions. The dispatch module is used to dispatch global scheduling instructions via the event message bus; The execution module is used by each intelligent agent to receive and execute instruction fragments assigned to itself through the local control unit, and to adjust the trajectory or action within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information. The feedback module is used by each intelligent agent to report the execution status of instruction fragments, resource occupancy status, and abnormal events to the central scheduling unit in real time through the event message bus.

[0014] Thirdly, this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to cause the electronic device to perform the above-described method.

[0015] Fourthly, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the above-described method.

[0016] The embodiments of this invention bring the following beneficial effects: This application provides a sorting collaborative control method, device, electronic device, and storage medium for multi-agent systems. The method includes: a central scheduling unit constructing a composite data model, which is jointly constructed through task definition, resource definition, and a state diagram of the temporal occupancy relationship between records; the central scheduling unit, within a rolling time window, integrates safety constraints and energy consumption constraints for joint solution to generate a global scheduling instruction; the global scheduling instruction includes task assignment, path segment occupancy time window, and device execution sequence, and marks the commitment level and sets the freeze boundary for the execution fragments in the instruction; the global scheduling instruction is issued through an event message bus; each agent receives and executes the instruction fragments assigned to it through its local control unit, and adjusts its trajectory or action within the spatiotemporal constraints committed by the instruction fragment based on local real-time perception information; each agent feeds back the execution status, resource occupancy status, and abnormal events of the instruction fragments to the central scheduling unit in real time through the event message bus.

[0017] This invention achieves unified description and state synchronization of multiple resources by constructing a composite data model; it realizes collaborative optimization and conflict prevention of tasks, paths, and time by integrating security and energy consumption constraints for joint solution within a rolling time window; it effectively suppresses rescheduling oscillations and ensures near-end execution stability by marking commitment levels and setting freeze boundaries for instructions; and it forms adaptive closed-loop control by achieving decoupled communication through an event message bus, with each agent making local adjustments and providing real-time feedback within spatiotemporal constraints, thereby improving the system's throughput efficiency, operational stability, and scalability in strongly coupled and highly disturbed sorting scenarios.

[0018] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating a sorting collaborative control method for a multi-agent system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating an application scenario of a sorting collaborative control method for multi-agent systems provided by an embodiment of the present invention. Figure 3 This is a schematic diagram of a sorting collaborative control device for a multi-agent system provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the electronic device structure provided in an embodiment of the present invention.

[0022] Figure label: 10 - Build Module, 20 - Generate Module, 30 - Deploy Module, 40 - Execute Module, 50 - Feedback Module; 130 - Processor, 131 - Memory, 132 - Bus, 133 - Communication interface. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions 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, 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.

[0024] To facilitate understanding of this embodiment, the technical terms used in this application will be briefly introduced below.

[0025] Task: Refers to the smallest executable unit of work defined by the central scheduling unit, such as grabbing, transferring, or placing. Its definition includes attributes such as task identifier (task_id), type (task_type), associated workstation (workcell_id), material identifier (material_id), and priority.

[0026] Resource: Refers to entities defined by the central scheduling unit that can be occupied by tasks, including Automated Guided Vehicles (AGVs), industrial robots, vision acquisition windows (CameraWindow), tool changers, and buffers. Its definition includes attributes such as resource identifier (res_id), type (res_type), capability set (capability_set), and safety zone tag (safety_zone_tag).

[0027] StateGraph: A component of a composite data model built and maintained by a central scheduling unit, used to record the temporal occupancy relationships between tasks and resources in a graph structure. The graph records the occupancy of specific resources by specific tasks through directed edges, which contain fields such as entry timestamp (enter_ts), release timestamp (release_ts), and reentrant flag (reentrant_flag).

[0028] Commitment Level: A stability marker assigned by the central scheduling unit to each execution segment in the global scheduling instruction. This marker controls the variability of the segment in subsequent optimizations, categorized as high commitment, medium commitment, and low commitment. High commitment segments, in principle, remain unchanged in subsequent optimizations.

[0029] FreezeBoundary: A dividing point set on the timeline by the central scheduling unit for the current instruction version. Execution segments before the freeze boundary are marked as high commitment to ensure the stability of near-end execution; segments after the boundary allow for adjustment in subsequent rolling optimizations.

[0030] EventMessageBus: The system communication hub using a publish / subscribe model, used to asynchronously transmit scheduling instructions, status feedback and abnormal events between the central scheduling unit and the local control units of each agent, thereby achieving decoupled communication between the system components.

[0031] After introducing the technical terms used in this application, the application scenarios and design concepts of the embodiments of this application will be briefly described below.

[0032] In the fields of intelligent manufacturing and logistics sorting, the collaborative operation of multiple intelligent agents (such as AGVs and robots) faces challenges such as complex resource coupling, frequent dynamic disturbances, high risk of conflict and deadlock, and fragmented scheduling and execution links. Traditional centralized or distributed scheduling solutions struggle to ensure execution stability while rapidly responding to real-time changes and optimizing multi-dimensional production indicators.

[0033] Based on this, this application provides a sorting collaborative control method, apparatus, electronic device, and storage medium for multi-agent systems.

[0034] Example 1 This application provides a sorting collaborative control method for multi-agent systems, combining... Figure 1 As shown, the method includes: S110 is a composite data model constructed by the central scheduling unit. The composite data model is jointly constructed through task definition, resource definition, and state diagram of the temporal occupancy relationship between records.

[0035] S120, within the rolling time window, the central dispatch unit integrates safety constraints and energy consumption constraints for joint solution to generate global dispatch instructions; the global dispatch instructions include task assignment, path segment occupancy time window and equipment execution sequence, and mark the commitment level and set the freeze boundary for the execution fragments in the instructions.

[0036] S130 sends out global scheduling instructions through the event message bus.

[0037] S140, each intelligent agent receives and executes the instruction fragments assigned to it through its local control unit, and adjusts its trajectory or actions within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information.

[0038] S150, each intelligent agent feeds back the execution status of instruction segments, resource occupancy status and abnormal events to the central scheduling unit in real time through the event message bus.

[0039] This application provides a sorting collaborative control method for multi-agent systems. This method is executed collaboratively by a central scheduling unit and distributed local control units of each agent, communicating via an event message bus. Combined with... Figure 2 The system operation principle shown in the (control flow) is based on the construction of a closed-loop information flow and control flow that runs through the perception layer, scheduling layer and operation terminal. Figure 2 This is not just a flowchart, but also a visual representation of the logical architecture of this method. The central scheduling unit acts as... Figure 2 The core of the scheduling layer relies on a dynamic, unified composite data model (i.e., a "task-resource-state diagram"). This model is built upon abstract mappings of the physical world: for example, robots on-site are abstracted as "Robot" resources, with their `capability_set` defining the specifications of parts they can grasp; AGVs are abstracted as mobile resources; conveyor line stations are abstracted as "Buffer" resources; and the field of view of industrial cameras is abstracted as a "CameraWindow" resource. These definitions form the basis of all scheduling.

[0040] This method is executed collaboratively by a central scheduling unit and distributed local control units of each intelligent agent, communicating via an event message bus. Taking a typical application scenario such as a laser-cut parts sorting system as an example, the central scheduling unit acts as the core decision-making node, its hardware typically deployed within a control cabinet or a connected server. Each intelligent agent corresponds to a physical entity performing the sorting task, such as a robot, automated guided vehicle (AGV), or conveyor line. Industrial cameras serve as key sensing units, providing visual feedback to the system. When the central scheduling unit starts, it constructs a unified composite data model based on work orders and resource information obtained from upper-level systems (such as Manufacturing Execution System (MES) or Warehouse Management System (WMS), and referring to the actual equipment layout and type. Specifically, robots are abstracted as "Robot" resources with grasping and placing capabilities, AGVs as mobile resources with transfer capabilities, conveyor lines and their stations as "Buffer" or "Station" resources, and the field of view of industrial cameras as "CameraWindow" resources. This allows for a precise mapping of the physical sorting system in the information space, laying the foundation for subsequent collaborative scheduling and control.

[0041] In conjunction with the first aspect, the task definition in step S110 includes task identifier, type, associated workstation, material identifier, and priority attribute; the resource definition abstracts the mobile robot, industrial robot, vision window, fixture station, and buffer area into occupiable entities, whose attributes include resource identifier, type, capability set, and safety domain label; the state graph records the entry timestamp, release timestamp, and reentrancy flag of the resource being occupied by a specific task through directed edges.

[0042] In this application, each specific sorting operation is defined as a task. For example, an instruction to "transfer part A from buffer B to processing station C" is instantiated as a task object, whose attributes include: a unique task_id (such as T001), a task_type of "transfer", a workcell_id of the code for station C, a material_id of the code for part A, and a priority value set according to the order delivery date.

[0043] All physical or logical entities in the application scenario are abstracted as resources. Referring to the example above, a KUKA robot is defined as a resource with `res_type` set to `Robot`, `res_id` set to `R001`, `capability_set` describing the size and weight range of parts it can grasp, and `safety_zone_tag` associated with the safety geofence of its work area. AGVs, vision windows (corresponding to industrial cameras), gripper quick-change stations, and buffer zones are defined in a similar manner.

[0044] The central scheduling unit constructs and maintains a dynamic state graph. When task T001 plans to use robot R001, a directed edge from T001 to R001 is added to the state graph. This edge records the planned enter_ts (e.g., start grasping at 10:00:00), the planned release_ts (e.g., release at 10:00:05), and sets the reentrant_flag according to the characteristics of the workstation (e.g., for high-temperature workstations, the same task is not allowed to re-enter during the cooling period, so the flag is set to no). This state graph is the core basis for all subsequent scheduling decisions.

[0045] In conjunction with the first aspect, step S120 includes: S121, update the state graph based on the received sensing events, and integrate the security domain constraints and energy consumption constraints into the composite data model to form a computable set of optimization constraints.

[0046] Step S121 is the core data preparation stage before dynamic optimization. The central scheduling unit continuously subscribes to various sensing events (scene_state) on the event message bus. For example, when an industrial camera detects a new workpiece in place, or an AGV reports its precise location, the corresponding event triggers a real-time update of the state graph. These updates include: inserting new task nodes into the state graph (such as a grabbing task for a new workpiece), or modifying the state attributes of existing resource nodes (such as marking a buffer as "occupied").

[0047] More importantly, step S121 explicitly integrates two types of key production constraints into a unified composite data model: Safety zone constraint integration: Dynamic safety zone information (such as temporarily defined restricted areas) from safety control systems (e.g., laser scanners) is translated into updates to the safety_zone_tag attribute of specific resources. In subsequent optimizations, any task assignment or path planning must ensure that the agent's trajectory always remains within the reach_set_ref of its authorized safety zone.

[0048] Energy consumption constraint integration: The real-time energy consumption strategy of the system or equipment (such as the total power consumption limit, the power limit corresponding to the time-of-use electricity price) is quantified as the power_cap_tag attribute of the resource or system. In the optimization model, this is reflected in the constraint on the number of devices running simultaneously or their power-intensive actions (such as high-speed movement, frequent fixture switching).

[0049] Through the above updates and integrations, the central scheduling unit has constructed a computable set of optimization constraints that is synchronized with the physical world and contains complete spatiotemporal relationships and hard constraints.

[0050] S122, within the rolling time window, based on the composite data model and constraint set, jointly solves the task assignment, path segment occupancy time window and device execution timing.

[0051] Step S122 is the core computational process of global optimization. The central scheduling unit sets a forward-sliding rolling time window (e.g., the next 5-10 minutes) in each scheduling cycle (e.g., every second) and performs integrated solution within this window.

[0052] Solution input: The input is the composite data model (containing all tasks to be processed, available resources and their current state graph) and constraint set output from the previous step, which is updated in real time.

[0053] Solution Variables and Objective: The solver needs to simultaneously determine three tightly coupled decision variables: task assignment (which AGV or robot performs which task), path segment occupancy time window (allocating precise entry and exit times for each AGV on critical path segments), and device execution sequence (the precise start and end times of the robot's grasping and placing actions). The optimization objective is a multi-index interpretable scoring function, which normalizes and weights key performance indicators such as system throughput, work-in-process inventory, AGV idle rate, and fixture changeover frequency, seeking the optimal comprehensive score.

[0054] S123 generates a global scheduling instruction based on the solution results, marks the commitment level of each execution segment in the instruction, and sets a freeze boundary for the current instruction version based on the system state.

[0055] Step S123 transforms the mathematical solution into executable instructions rich in control metadata, outputting a series of action sequences with precise timestamps and resource bindings. The central scheduling unit encapsulates these into structured global scheduling instructions, which include specific executable segments such as "At time T1, AGV-001 departs from position A, occupies path segment S1-S2, and arrives at workstation W1 at time T2".

[0056] Through the three sub-steps S121-S123, step S120 completes the entire closed loop from real-time state perception to generating stable, executable global scheduling instructions containing intelligent control strategies.

[0057] In conjunction with the first aspect, step S122 includes: S1221, the objective function for optimization is a multi-index interpretable scoring function; the multi-index includes at least system throughput, work-in-process quantity, mobile robot idle rate and fixture switching frequency.

[0058] Since the setting of the objective function directly determines the quality of the scheduling scheme, unlike the traditional single objective (such as minimizing the completion time) or fixed weight summation optimization method, step S1221 of this application sets a multi-index interpretable scoring function, which aims to balance and optimize multiple key operational indicators, so that the scheduling scheme is more in line with the comprehensive benefit needs of actual production.

[0059] The design and calculation process of the objective function is as follows: Indicator definition and quantification: System throughput: refers to the total number of tasks expected to be completed within a rolling time window. This metric directly reflects system capacity, and its goal is to maximize it.

[0060] Work-in-Process (WIP): This refers to the instantaneous backlog of materials in the system corresponding to all started but unfinished tasks within a rolling time window (usually the average or maximum value within the time window). Controlling WIP helps reduce inventory costs, shorten logistics cycles, and reduce management complexity.

[0061] Mobile robot empty-run rate (AGVIdleRate): refers to the proportion of a mobile robot's mileage or time spent without a task to its total mileage or time spent. Reducing the empty-run rate means higher utilization of transportation resources and lower ineffective energy consumption.

[0062] Tool Change Frequency: This refers to the estimated number of times an industrial robot will need to change its end effector (gripper) to adapt to different workpieces. Frequent changes increase non-productive time, wear and tear, and potential failure risks, and therefore require optimization.

[0063] The objective function is designed as a weighted sum of multiple indicators, where the weight coefficients for each indicator are not fixed but stored as system parameters. Operators or upper management systems can dynamically adjust these weights based on the priorities of different production shifts (e.g., day shifts aim for maximum capacity, night shifts focus on energy conservation), order structure characteristics, or real-time energy strategies, enabling the scheduling strategy to flexibly adapt to diverse production objectives. Each candidate scheduling scheme calculates a comprehensive score and outputs the contribution value of each sub-indicator. This makes scheduling decisions no longer a "black box," allowing managers to understand why scheme A was chosen over scheme B (e.g., scheme A achieves a higher comprehensive score due to significantly reducing empty running rates), thereby enhancing the system's transparency and credibility.

[0064] During the joint solution process in step S122 (e.g., based on heuristic search, mathematical programming, or reinforcement learning algorithms), the solver generates multiple candidate scheduling schemes (covering different combinations of task assignment, path planning, and time scheduling). For each candidate scheme, its comprehensive score is calculated using the aforementioned objective function. Finally, the scheme with the highest comprehensive score is selected as the joint solution result for the current rolling time window. By adopting this multi-index interpretable scoring function as the optimization objective, the scheduling method of this invention can systematically coordinate the performance of multiple dimensions such as production efficiency (throughput), liquidity (work-in-process), resource utilization (empty-running rate), and equipment wear and tear (switching frequency), outputting scheduling instructions that better reflect the optimal overall operational benefits, thereby overcoming the limitations of existing technologies where the objective function is singular and it is difficult to balance multiple objective requirements.

[0065] In conjunction with the first aspect, step S122 also includes: S1222, adopts a conflict resolution strategy based on segment reservation and priority inheritance; The conflict resolution strategy includes: allocating exclusive time windows for the path segments and workspace segments of the mobile robot; dynamically allocating passage priorities based on task urgency and blocking cost when resource contention is detected; and passing high priority to low priority along the waiting chain if chained waiting is formed.

[0066] Step S1222, embedded within the joint solution process of step S122, is a crucial component of the core algorithm logic. It aims to prevent and resolve potential conflicts and deadlocks among multiple agents sharing spatiotemporal resources at their root. This strategy is a preventative rather than reactive conflict management mechanism, and its principle and process are as follows: Strategy 1: Allocate exclusive time windows for the mobile robot's path segments and workspace segments: Path segment reservation: The AGV's travel network (such as channels, intersections, loading / unloading points) is divided into logically critical segments. During joint solution, the scheduling algorithm not only plans paths for the AGVs, but more importantly, it precisely allocates an exclusive occupancy time window ([enter_ts, release_ts]) for each AGV on each critical segment of its path. This means that within this time window, the segment is considered "reserved" by that AGV, and the scheduling schemes of other AGVs must avoid this spatiotemporal region.

[0067] Workspace Segment Reservation: Similarly, for the operating space of industrial robots, especially key areas such as shared workstations, fixture exchange areas, and palletizing positions, these are also abstracted as space segments that can be occupied, and exclusive usage time windows are allocated to robot tasks during the solution process.

[0068] By transforming the risk of conflict in physical space into the mutual exclusion constraint of time windows in the scheduling model, this strategy can ensure that the generated schemes are conflict-free in the spatiotemporal dimension during the planning stage, fundamentally avoiding the risk of collision during the execution process.

[0069] Strategy 2: When resource contention is detected, dynamically allocate passage priority based on task urgency and blocking cost. When the joint solution algorithm deduces a scenario in which multiple agents apply for the same resource (segment) at similar times during the search or optimization process, it is considered to have detected resource contention.

[0070] At this point, the system will not simply adopt a first-come, first-served or fixed-priority rule. Instead, it will comprehensively evaluate the task's urgency (e.g., the task's inherent priority attribute, the urgency of the approaching delivery date) and the cost of blocking (e.g., how many subsequent tasks will be delayed and how many resources will be idle if the agent waits), and dynamically calculate and allocate the real-time passage priority for this competition.

[0071] Dynamic prioritization makes scheduling decisions more intelligent, enabling judgments based on global impact and prioritizing tasks with a greater impact on overall system performance, thereby optimizing the overall workflow.

[0072] Strategy 3: If a chain of waits occurs, the higher priority wait is passed down to the lower priority wait along the chain. Chained waiting: In complex scenarios, a chained or circular waiting relationship may form, where A waits for resources held by B, B waits for resources held by C, and C may indirectly wait for A. This is a typical precursor to deadlock.

[0073] Priority inheritance mechanism: Once the system detects such a chained wait during the solution process, it will immediately trigger the priority inheritance rule. That is, the task with the highest priority in the chain will have its priority passed down along the waiting chain (towards lower priorities). This means that lower priority tasks in the chain will inherit a higher temporary priority.

[0074] After inheriting high priority, tasks with lower priority in the chain will gain priority scheduling rights. The solver will adjust its scheme accordingly, usually forcing lower priority agents at the tail or in the chain to give up competition, choose alternative paths, or release resources early, thereby actively breaking the waiting chain and preventing circular waiting and deadlock.

[0075] The conflict resolution strategy described in step S1222 integrates three mechanisms—spatiotemporal reservation, dynamic prioritization, and inheritance resolution—at the global optimization level. This ensures that the final scheduling scheme is not only high-performing but also inherently conflict-free and possesses liveness guarantees (provably free of deadlock). Unlike common existing methods that rely on passive collision avoidance or simple intersection arbitration at the execution end, this strategy fundamentally improves the reliability and efficiency of multi-agent systems in dense, dynamic environments. The execution and effectiveness of this strategy are crucial for step S122 to output high-quality joint solution results.

[0076] In conjunction with the first aspect, the commitment level in step S123 is used to identify the variability of the execution segment in subsequent optimization, including high commitment, medium commitment and low commitment; the freeze boundary is the dividing point on the time axis, and the execution segment before the freeze boundary is marked as high commitment to ensure execution stability.

[0077] Step S123 defines the core metadata mechanism for ensuring the stability of system execution, namely commitment level and freeze boundary. They work together to generate global scheduling instructions to balance the flexibility of scheduling optimization with the determinism of on-site execution.

[0078] The commitment level is a label used to identify the extent of adjustment allowed for each execution segment in subsequent rolling optimization cycles. It is divided into three levels: High Commitment: Identifies segments that are nearing or already in execution (e.g., within the next 30 seconds). This level signifies a strong commitment: in the next optimization cycle, unless an unavoidable anomaly occurs (such as a hardware failure or a security domain mutation), the central scheduling unit will strive to maintain these segments unchanged. This provides stable expectations for the execution end and is the cornerstone of ensuring stable cycle time.

[0079] Medium Commitment: Identifies intermediate segments (e.g., the next 30 seconds to 2 minutes). This level allows for limited, controlled adjustments. Without affecting the execution of any high-commitment segments, the system can modify these segments in subsequent optimizations to absorb disturbances or for further optimization.

[0080] Low Commitment: Identifies long-term or predictive segments (e.g., 2 minutes from now). These segments are primarily used for strategic pre-scheduling and resource demand forecasting, are highly variable, and are continuously replanned and refined through rolling optimization.

[0081] The three-level commitment mechanism essentially establishes a variability gradient on the time axis. It allows the system to maintain high stability in near-term operations while retaining sufficient optimization space in the medium and long term, thereby effectively suppressing the plan oscillation phenomenon commonly found in traditional rescheduling methods, namely, the execution chaos caused by frequent adjustments to near-term instructions.

[0082] A frozen boundary is a clearly defined dividing point on the absolute timeline (e.g., the current time Tnow + 25 seconds). It serves as a benchmark for separating the stable execution zone from the optimization and adjustment zone. The frozen boundary works in conjunction with commitment levels, following a core rule: all execution segments preceding the frozen boundary, regardless of their original planned commitment level, are considered high-commitment segments for processing and protection within the current scheduling period. Thus, the frozen boundary explicitly and solidifies stability guarantees. It clearly informs all agents: instructions before the boundary are highly reliable and should be executed with full force; instructions after the boundary are provisional and may change. This brings two key benefits: Ensuring near-end stability: It provides a clear and unambiguous stable command range for the execution layer (especially the motion controllers of AGVs and robots), enabling it to perform smooth and precise control without worrying about commands being revoked at the next moment.

[0083] Simplified exception handling logic: When a disturbance occurs, the system can quickly determine the scope of impact based on the frozen boundary. As long as the disturbance does not directly affect the segments within the boundary, rearrangement only needs to be performed outside the boundary, greatly simplifying the decision-making logic and protecting critical production cycles from being affected.

[0084] In step S123, after the central scheduling unit completes the solution, it performs the following operations: First, label the commitment level: Based on the difference between the planned start time and the current time of each execution segment, label the segment with the corresponding commitment level according to the preset time threshold (e.g., 0-30 seconds is high commitment, 30-120 seconds is medium commitment, and after 120 seconds is low commitment).

[0085] Next, the freeze boundary is calculated and set: based on the current average command delivery latency, network jitter, the shortest response time of critical equipment, and production cycle requirements, a reliable freeze boundary time point is dynamically calculated. Typically, this boundary is set slightly earlier than the start time of the earliest high-commitment segment to allow for a buffer.

[0086] A hierarchical plan stability management system is constructed by combining commitment levels and frozen boundaries: the frozen boundaries rigidly divide the stable zone and the optimization zone in terms of time; the commitment levels, within the optimization zone, flexibly manage plan segments from a variability perspective. This ensures that the scheduling strategy of this invention can adapt to highly volatile environments while maintaining highly robust production execution.

[0087] Finally, instruction encapsulation: global scheduling instructions that specify the commitment level and freeze boundary location are encapsulated into messages and prepared for issuance.

[0088] In conjunction with the first aspect, step S140 also includes: S141, when an abnormal event is received through the event message bus, determine whether the abnormal event affects the execution segment marked as high commitment within the frozen boundary.

[0089] If yes, proceed to step S142; otherwise, proceed to step S143.

[0090] S142 triggers a re-verification process for the affected execution fragment. The re-verification process includes reachingability and mutual exclusion checks, and if the checks fail, priority is given to replacing equivalent resources.

[0091] S143 then performs a regular rearrangement of subsequent scheduling instructions outside the frozen boundary.

[0092] Step S141 is the decision entry point for the system to respond to disturbances and ensure stability. As shown in the event.alarm message stream from the operator, when any agent (such as an AGV or robot) or sensor detects an anomaly (such as grasping failure, device communication timeout, or sudden path blockage), it will immediately format it as a standard anomaly event and report it to the central scheduling unit via the event message bus. Upon receiving the event, the central scheduling unit first performs a critical judgment, assessing whether the anomaly will render any execution segment marked as high commitment within the frozen boundary inoperable. For example, if a robot reports a "gripper jam" fault, it will determine whether it will affect a high commitment grasping segment that is about to begin and will be executed by that robot.

[0093] This judgment directly determines the system's response strategy to balance recovery speed and the impact on the established plan. If it is (a high-commitment segment within the impact boundary): this indicates that the anomaly directly threatens the committed stable production cycle time at the near end, requiring immediate and precise intervention, proceeding to step S142. At this time, the central scheduling unit locks the high-commitment execution segment directly affected by the anomaly and triggers a re-verification process for it in the state diagram. This is the core calculation of re-verification. Based on the latest system state after the anomaly occurred (e.g., the faulty equipment has been marked as unavailable), the central scheduling unit recalculates in the state diagram: Reachability check: Can the affected segment still be completed by the originally planned resources within the original time window? For example, can the faulty robot arrive at the grab point on time before repair?

[0094] Mutual exclusion check: If resources are changed or time is adjusted, will it create new spatiotemporal conflicts with other committed plans in the state diagram?

[0095] If the check fails (i.e., the original segment is not feasible), the equivalent resource replacement strategy is prioritized. This involves searching the resource definition set for equivalent resources with the same or higher capability_set that are available in the current time period (e.g., another robot of the same model and located nearby). The strategy attempts to directly replace the existing faulty resource with this equivalent resource to execute the high-commitment segment, and the time window for that segment is quickly recalculated. Upon successful replacement, only the resource allocation of the affected segment is updated, preserving the original planned timing structure and the integrity of other segments within the frozen boundary to the greatest extent possible, achieving a smooth transition.

[0096] If the judgment result of step S141 is negative (no impact): this indicates that the disturbance mainly affects future plans (after the frozen boundary), and the system has more buffer space to handle it. The process then proceeds to step S143, which involves an absorptive, systematic optimization process for anomalies that do not affect the core stable region. Since the anomaly does not affect high-commitment segments within the frozen boundary, the central scheduling unit strictly limits the processing scope to the area after the frozen boundary, protecting the upcoming stable plan from interference. Under this premise, this anomaly is treated as a new state input. Like handling ordinary state updates, a new joint solution is initiated for medium and low-commitment segments after the frozen boundary within the next or current rolling time window (i.e., step S122). Reordering can adjust task assignment and path planning to avoid anomaly points (such as bypassing fault areas) and reallocate load, thereby systematically absorbing the impact of the disturbance. This approach transforms a local disturbance into an opportunity for global optimization, ensuring near-end cycle stability while optimizing medium- and long-term plans, enabling the system to autonomously recover from the disturbance to a better operating state.

[0097] Steps S141-S143 constitute a hierarchical and intelligent anomaly response system. By using the key criterion of freezing boundaries, it categorizes anomalies into two types: urgent threats and absorbable disturbances. It then employs two strategies to handle these categories: precise re-verification and replacement, and globally optimized boundary rearrangement, respectively. This mechanism ensures that the system can respond quickly to various uncertainties while strictly safeguarding the stability of committed execution plans, serving as the core guarantee for achieving highly robust collaborative control.

[0098] This application provides a sorting collaborative control method for multi-agent systems. The complete operational logic, hierarchical structure, and data interaction relationships of this method are as follows: Figure 2 The control flow diagram illustrates how the system connects the control panel (central scheduling unit carrier), the central scheduling, the sub-scheduling, and the underlying execution and sensing modules into a collaborative whole through an event message bus (using WiFi as an example of the main wireless communication carrier in the diagram) and industrial protocols such as Modbus.

[0099] Top-level interaction and decision-making begin with the human-machine interface interaction of the control tablet, which communicates via WiFi. The control tablet carries the overall scheduling function, responsible for receiving global status and allocating tasks. Based on this, the overall scheduling module executes the core AGV path planning and generates advanced operation instructions for units such as sorting robots, machine vision, and spill conveying devices. All instructions and status data are packaged and processed before being transmitted via the WiFi network.

[0100] The mid-level scheduling and centralized control are implemented by the sub-scheduling layer. This layer contains two key central control units: The sorting AGV master control unit directly manages the AGV power module, robotic arm power module, and vision module through the Modbus industrial bus, enabling precise low-level control of AGV movement, robot joint motion, and vision triggering.

[0101] The central control unit for the conveying device manages the execution units related to the leaky material conveying device via WiFi, including controlling the actions of the sorting module, receiving signals from the full-pile sensing module, and driving the transport power module.

[0102] The underlying execution and perception layers correspond to the functional modules shown in the diagram. The AGV-cut parts sorting task is completed collaboratively by the AGV and the robotic arm; the machine vision and full-pile perception systems, as independent perception units, provide workpiece identification and stacking status information, respectively; and the material leakage conveyor executes the final outflow of materials. All module status and perception data are packaged and fed back to the upper layer via WiFi.

[0103] The closed-loop data flow runs through all the arrows in the diagram: perception data (visual, full stack) is aggregated from bottom to top to the central scheduler; the central scheduler's decisions (path, task) are decomposed by the sub-schedules and then sent to the underlying modules for execution via Modbus or WiFi; the execution status and results are then fed back through the reverse path, forming a complete closed loop of "perception-decision-execution-feedback".

[0104] In conclusion, Figure 2 This paper precisely describes the technical architecture and complete data flow of this method, from the global command of the control tablet to the intelligent planning of the overall scheduling (task allocation, AGV path planning), to the centralized control decomposition of the sub-scheduling (sorting AGV overall control, conveying device overall control), and finally to the collaborative operation of all underlying modules such as AGV power, robotic arm power, vision, sorting, full-pile perception and transportation power through a hybrid Modbus and WiFi network.

[0105] Secondly, embodiments of this application also provide a sorting collaborative control device for multi-agent systems, combined with Figure 3 As shown, the device includes: a construction module 10, a generation module 20, a distribution module 30, an execution module 40, and a feedback module 50.

[0106] The construction module 10 constructs a composite data model by the central scheduling unit. The composite data model is jointly constructed through task definition, resource definition, and state diagram of the temporal occupancy relationship between records.

[0107] Within the rolling time window, the generation module 20 integrates safety constraints and energy consumption constraints to jointly solve the global scheduling instructions. The global scheduling instructions include task assignment, path segment occupancy time window and equipment execution sequence, and mark the commitment level and set the freeze boundary for the execution fragments in the instructions.

[0108] The dispatch module 30 dispatches global scheduling instructions through the event message bus.

[0109] The execution module 40 receives and executes instruction fragments assigned to each intelligent agent through its local control unit, and adjusts its trajectory or action within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information.

[0110] The feedback module 50 transmits the execution status of instruction segments, resource occupancy status, and abnormal events to the central scheduling unit in real time through the event message bus.

[0111] Thirdly, embodiments of this application provide an electronic device, combined with Figure 4 As shown, the electronic device includes a memory 131 and a processor 130. The memory 131 stores a computer program, and the processor 130 runs the computer program to make the electronic device perform the above-described method.

[0112] Furthermore, combined Figure 4 The electronic device shown also includes a bus 132 and a communication interface 133, with the processor 130, the communication interface 133 and the memory 131 connected via the bus 132.

[0113] The memory 131 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 133 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 132 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 It is indicated by a single double-headed arrow, but does not mean that there is only one bus or one type of bus.

[0114] Processor 130 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 130 or by instructions in software form. Processor 130 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 131, and processor 130 reads the information in memory 131 and, in conjunction with its hardware, completes the method of the aforementioned embodiments.

[0115] Fourthly, embodiments of this application provide a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the above-described method.

[0116] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and apparatus described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0117] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0118] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0120] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A sorting collaborative control method for a multi-agent system, characterized in that, The method includes: A composite data model is constructed by a central scheduling unit. This composite data model is jointly constructed through task definition, resource definition, and a state diagram of the temporal occupancy relationship between records. Within the rolling time window, the central scheduling unit integrates safety constraints and energy consumption constraints for joint solution to generate a global scheduling instruction. The global scheduling instruction includes task assignment, path segment occupancy time window and equipment execution sequence, and marks the commitment level and sets the freeze boundary for the execution fragment in the instruction. The global scheduling command is issued via the event message bus; Each intelligent agent receives and executes instruction fragments assigned to it through its local control unit, and adjusts its trajectory or actions within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information. Each of the intelligent agents feeds back the execution status of the instruction fragment, the resource occupancy status, and abnormal events to the central scheduling unit in real time through the event message bus; The step of the central scheduling unit integrating safety constraints and energy consumption constraints to jointly solve for and generate a global scheduling instruction within a rolling time window includes: updating the state diagram based on received sensing events, and integrating safety domain constraints and energy consumption constraints into the composite data model to form a computable set of optimization constraints; within the rolling time window, jointly solving for task assignment, path segment occupancy time window, and device execution timing based on the composite data model and the set of constraints; generating the global scheduling instruction based on the solution results, marking the commitment level for each execution segment in the instruction, and setting a freeze boundary for the current instruction version based on the system state; wherein the step of jointly solving for task assignment, path segment occupancy time window, and device execution timing based on the composite data model and the set of constraints includes: the objective function for optimization is a multi-index interpretable scoring function; the multi-index includes at least system throughput, work-in-process quantity, mobile robot empty-run rate, and fixture switching frequency; The commitment level is used to identify the variability of the execution segment in subsequent optimization, including high commitment, medium commitment and low commitment; the freeze boundary is the dividing point on the time axis, and the execution segments before the freeze boundary are marked as high commitment to ensure execution stability.

2. The method of claim 1, wherein, The task definition includes task identifier, type, associated workstation, material identifier, and priority attribute; the resource definition abstracts mobile robot, industrial robot, vision window, fixture station, and buffer area into occupiable entities, whose attributes include resource identifier, type, capability set, and security domain label; the state graph records the entry timestamp, release timestamp, and reentrancy flag of the resource being occupied by the task through directed edges.

3. The method of claim 1, wherein, Within the rolling time window, the step of jointly solving for task assignment, path segment occupancy time window, and device execution timing based on the composite data model and the constraint set further includes: A conflict resolution strategy based on segment reservation and priority inheritance is adopted; The strategy includes: allocating exclusive time windows for the path segments and workspace segments of the mobile robot; dynamically allocating passage priorities based on task urgency and blocking cost when resource contention is detected; and passing high priority to low priority along the waiting chain if chained waiting is formed.

4. The method of claim 1, wherein, Each intelligent agent receives and executes instruction fragments assigned to it through its local control unit, and adjusts its trajectory or action within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information. The process also includes: When an abnormal event is received through the event message bus, it is determined whether the abnormal event affects the execution segment marked as high commitment within the frozen boundary; If so, a re-verification process is triggered for the affected execution fragment. The re-verification process includes performing reachability and mutual exclusion checks, and prioritizing the replacement of equivalent resources when the checks fail. If not, then subsequent scheduling instructions will be rearranged normally outside the frozen boundary.

5. A sorting collaborative control device for multi-agent systems, characterized in that, The apparatus is used to perform the method as described in any one of claims 1 to 4; the apparatus comprises: The construction module is used to build a composite data model by the central scheduling unit. The composite data model is jointly constructed through task definition, resource definition, and state diagram of time sequence occupancy relationship between records. The generation module is used by the central scheduling unit to jointly solve for safety constraints and energy consumption constraints within a rolling time window to generate a global scheduling instruction. The global scheduling instruction includes task assignment, path segment occupancy time window and device execution sequence, and marks the commitment level and sets the freeze boundary for the execution fragment in the instruction. The dispatch module is used to dispatch the global scheduling command via the event message bus; The execution module is used by each intelligent agent to receive and execute instruction fragments assigned to itself through a local control unit, and to adjust its trajectory or action within the spatiotemporal constraints promised by the instruction fragments based on local real-time perception information. The feedback module is used by each of the intelligent agents to feed back the execution status of instruction fragments, resource occupancy status and abnormal events to the central scheduling unit in real time through the event message bus.

6. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program and the processor running the computer program to cause the electronic device to perform the method of any one of claims 1 to 4.

7. A storage medium, characterized in that, The storage medium stores computer program instructions, which, when read and executed by a processor, perform the method described in any one of claims 1 to 4.

Images