Harbor quay crane-yard crane-AGV integrated scheduling system and method

Abstract

The invention relates to a harbor quay crane-field crane-AGV integrated scheduling system and method. The system comprises a total factor perception and communication module, a unified task modeling and decomposition module, a collaborative scheduling optimization engine, an elastic scheduling layer, a digital twinborn simulation sandbox module, a conflict prediction and dynamic resolution module and an instruction distribution and execution monitoring module. The scheduling method comprises the following steps: initializing a system and sensing a dynamic environment; performing rolling time domain optimization triggering and task set determination; generating and verifying an integrated cooperative scheduling scheme; dispatching instruction issuing and online dynamic fine tuning; and executing process monitoring and dynamic rescheduling. According to the method, global collaborative optimization of production tasks, logistics resources and energy supply is realized by constructing a unified space-time decision model and fusing digital twinborn verification, so that the ship berthing time is remarkably shortened, and the comprehensive utilization rate of equipment and the ability of the system to deal with uncertainty are improved.

Description

Technical Field

[0001] This invention relates to the technical field of port automation and intelligent scheduling, and in particular to an integrated scheduling system and method for port quay cranes, yard cranes and AGVs. Background Technology

[0002] Automated container terminals represent a core development direction for improving port operational efficiency and reducing labor costs. Their core operational processes involve three key pieces of equipment: quay cranes for loading and unloading containers between ships and the shore, yard cranes for storing and retrieving containers within the yard, and AGVs for horizontal transportation at the terminal. Traditional scheduling models generally employ a hierarchical or segmented architecture: first, a quay crane operation plan is formulated; then, AGVs are allocated to each operation line; and finally, yard crane operations are assigned to containers arriving at the yard. This "decoupled" scheduling has inherent flaws: a lack of global optimization, inconsistent scheduling objectives at different levels, and information lag, easily leading to local optima but overall low performance, and severe waiting between equipment; sluggish dynamic response, with slow response to disturbances such as equipment failures and operational delays, easily causing operational chain disruptions; passive conflict avoidance, as the spatiotemporal coupling relationship between AGV path planning and quay / yard crane operations is not proactively considered, often resulting in deadlocks at intersections and congestion in the work area; and insufficient resource coordination, with AGV charging and buffer zone management separated from production scheduling, affecting continuous operation capabilities.

[0003] While some existing technologies have attempted to jointly schedule pairs of devices, they have failed to place quay cranes, yard cranes, and AGVs within a unified spatiotemporal framework for millisecond-level collaborative decision-making. This makes it difficult to achieve global optimization across the entire "ship-horizontal transport-yard" chain, and lacks an efficient online adjustment and verification mechanism to cope with complex dynamic environments. Summary of the Invention

[0004] This invention aims to address the shortcomings of existing technologies by providing an integrated scheduling system and method for port quay cranes, yard cranes, and AGVs. By constructing a unified elastic spatiotemporal model, co-evolutionary intelligent optimization, and digital twin simulation verification closed loop, it achieves deep collaboration and global optimization of the three types of equipment, significantly improving the overall operational efficiency, resource utilization, and system robustness of the terminal.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A port quay crane-yard crane-AGV integrated scheduling system is deployed on a central control server at the terminal and connected to terminal equipment via the Industrial Internet of Things (IIoT), including:

[0007] The all-element perception and communication module is used to collect real-time equipment status information, container information, and operation plans from the upper-level system of quay cranes, yard cranes, and AGVs, and to realize bidirectional low-latency transmission of control commands and status feedback.

[0008] The unified task modeling and decomposition module is used to receive macro-level work plans and decompose them into atomic task units. Each atomic task unit defines a complete operation for transporting a specific container from one specified resource point to another, and is associated with time window, priority, and task type attributes.

[0009] The collaborative scheduling optimization engine, based on the real-time status provided by the full-element perception and communication module and the atomic task set provided by the unified task modeling and decomposition module, constructs a unified elastic spatiotemporal network model, establishes a multi-objective optimization model with minimizing ship berthing time as the main objective and taking into account the total energy consumption of AGVs and the balance of equipment load, and uses intelligent optimization algorithms to solve the problem, outputting an integrated collaborative scheduling pre-solution.

[0010] The digital twin simulation sandbox module is used to build a high-fidelity virtual image synchronized with the physical terminal. It receives the integrated collaborative scheduling pre-scheduling output from the collaborative scheduling optimization engine, performs accelerated simulation and deduction, evaluates key performance indicators and identifies potential spatiotemporal conflicts and risks, and feeds the evaluation results back to the collaborative scheduling optimization engine for scheme calibration and iterative optimization, forming a "optimization-simulation-verification" decision-making closed loop.

[0011] The conflict prediction and dynamic resolution module, based on the currently executing scheduling scheme and real-time device trajectory data, predicts the spatiotemporal conflicts between devices in the short time domain in the future in an online rolling manner, and uses speed adjustment, path node reservation or task sequence fine-tuning strategies for dynamic resolution;

[0012] The instruction distribution and execution monitoring module is used to parse the final scheduling scheme verified by the digital twin simulation sandbox module into a series of executable instruction sequences for specific devices and issue them out. At the same time, it monitors the instruction execution status and progress deviation in real time, and triggers a dynamic rescheduling process when the deviation exceeds the threshold or a specific event occurs.

[0013] The collaborative scheduling optimization engine specifically includes:

[0014] The spatiotemporal resource network modeling unit is used to discretize the physical area of ​​the terminal into spatial nodes and, in combination with the discretized time slices, construct a unified spatiotemporal network graph. Among them, network nodes represent the spatial location and operating status of equipment at a specific moment, and network edges represent the actions performed by the equipment when moving between different spatiotemporal nodes, including driving, loading and unloading, waiting, and charging. Container transportation tasks are modeled as the flow demand from the starting resource node to the target resource node on the spatiotemporal network.

[0015] A multi-objective robust optimization model unit is used to establish a mixed-integer programming model. The objective function includes at least minimizing the maximum completion time among all service vessels, minimizing the total driving energy consumption of all AGVs, and maximizing the load balance between quay cranes and yard cranes. In addition to equipment capacity and basic task logic constraints, the constraints also include: setting buffer capacity constraints for each quay crane and yard crane to limit the number of AGVs waiting below at the same time; modeling the battery power consumption and charging demand of AGVs as special tasks that can be inserted into the scheduling sequence to achieve coordination between production and energy supply; and introducing robustness constraints for critical operation durations to generate scheduling schemes that are insensitive to uncertainty.

[0016] The intelligent optimization solution unit employs a co-evolutionary algorithm based on spatiotemporal chromosome encoding and reinforcement learning to efficiently solve multi-objective robust optimization models. The algorithm maintains two co-evolutionary subpopulations for resource allocation and traffic scheduling, which optimize the quay crane / yard crane operation sequence and AGV path planning and fleet allocation, respectively. The subpopulations co-evolve by exchanging superior spatiotemporal scheduling segments and embed a variable neighborhood search strategy guided by a reinforcement learning agent for local deep optimization.

[0017] The unified elastic spatiotemporal network model is elastic, and the granularity of time slices can be dynamically adjusted according to the scheduling stage and the optimization accuracy requirements. A coarser time granularity is used in the offline macro-planning stage, and a finer time granularity is used in the online rolling optimization stage.

[0018] In the intelligent optimization solution unit, the reinforcement learning agent continuously learns from historical scheduling data and simulation feedback, and dynamically adjusts the neighborhood structure combination and calling order used in the variable neighborhood search strategy to adaptively improve the convergence speed and solution quality of the algorithm in solving different job scenarios.

[0019] The digital twin simulation sandbox module uses parallel computing and event-driven simulation technology to extrapolate scheduling schemes faster than actual operations. The module has an embedded key conflict rule library that can automatically identify potential deadlock, congestion, and equipment interference risks.

[0020] The digital twin simulation sandbox module adopts a federated twin architecture, which includes a high-fidelity core sandbox for final solution verification, and multiple lightweight fast sandboxes for initial screening and rapid evaluation of solutions within the optimization engine, in order to balance simulation accuracy and decision-making efficiency.

[0021] It also includes an elastic scheduling layer, located between the collaborative scheduling optimization engine and the instruction distribution and execution monitoring module. This layer receives the simulation-verified scheduling scheme and, based on the real-time perceived system load and device health status, slightly expands and contracts the execution time windows of some non-critical tasks in the scheme to further absorb micro-disturbances and enhance the compliance of execution.

[0022] The scheduling method based on the above-mentioned integrated scheduling system of port quay cranes-yard cranes-AGVs includes the following steps:

[0023] S1. System Initialization and Dynamic Environment Awareness:

[0024] Load the terminal digital map and equipment parameter library, and continuously obtain the real-time status of all controllable equipment and the list of pending tasks through the full-element perception and communication module;

[0025] S2. Rolling Time Domain Optimization Triggering and Task Set Determination:

[0026] A new round of optimization is initiated based on a preset fixed time period or triggered by equipment failure, serious deviation in task progress, or the addition of an emergency task; the set of all atomic tasks to be scheduled in the next optimization time domain is locked.

[0027] S3. Generation and Verification of Integrated Collaborative Scheduling Scheme:

[0028] S31. Call the collaborative scheduling optimization engine, based on the current dock status and the set of atomic tasks to be scheduled, run the intelligent optimization algorithm, solve the multi-objective optimization model, and obtain a preliminary integrated collaborative scheduling pre-scheduling plan;

[0029] S32. Input the preliminary integrated collaborative scheduling scheme into the digital twin simulation sandbox module, perform multi-speed accelerated simulation, evaluate the theoretical performance indicators, and detect potential hidden conflicts and execution risks in the scheme.

[0030] S33. Make a decision based on the simulation evaluation results: If the results meet the preset performance and robustness thresholds, the final executable solution is confirmed; if not, the bottleneck information and risk points identified by the simulation are fed back to the collaborative scheduling optimization engine. After adjusting the model parameters or constraint weights, the process jumps to step S31 for iterative optimization until a satisfactory solution is obtained.

[0031] S4. Issuance of scheduling instructions and online dynamic fine-tuning:

[0032] The instruction distribution and execution monitoring module distributes the instruction sequence corresponding to the recent time period in the final executable plan to the corresponding devices; at the same time, the conflict prediction and dynamic resolution module continuously performs conflict prediction and online fine-tuning based on the plan and real-time trajectory during subsequent execution.

[0033] S5. Execution process monitoring and dynamic rescheduling:

[0034] When the actual execution progress of the system monitoring scheme reaches a new fixed optimization cycle, or when a critical equipment failure is detected, the cumulative deviation of the task exceeds the threshold, or a high-priority new task is inserted, dynamic rescheduling is triggered, and the process returns to step S2.

[0035] In step S2, the conditions and levels for triggering events are predefined, and different levels of events trigger dynamic rescheduling with different response speeds and optimization ranges.

[0036] In step S4, the online dynamic fine-tuning strategy adopts the principle of "speed adjustment as the main method and path node reservation as the auxiliary method". When there is sufficient buffer time and space between the equipment and the potential conflict point, a smooth speed adjustment strategy is adopted; when the conflict risk is high, the reservation mechanism for key intersections and loading / unloading positions is activated to avoid conflicts by controlling the arrival time of the equipment.

[0037] The beneficial effects of this invention are: by constructing a unified spatiotemporal decision-making model and integrating digital twin verification, this invention achieves global collaborative optimization of production tasks, logistics resources and energy supply, thereby significantly shortening ship berthing time and improving the overall utilization rate of equipment and the system's ability to cope with uncertainties. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the port quay crane-yard crane-AGV integrated scheduling system in this invention;

[0039] Figure 2 This is a flowchart illustrating the scheduling method of the port quay crane-yard crane-AGV integrated scheduling system in this invention.

[0040] The following will describe in detail, with reference to the accompanying drawings, embodiments of the present invention. Detailed Implementation

[0041] The principles and features of the present invention are described below with reference to the accompanying drawings. The embodiments given are for illustrative purposes only and are not intended to limit the scope of the invention. The invention is described more specifically in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0044] An integrated scheduling system for port quay cranes, yard cranes, and AGVs, such as Figure 1As shown, it is deployed on the central control server of the dock and connected to the dock equipment through the Industrial Internet of Things. It includes a full-element perception and communication module, a unified task modeling and decomposition module, a collaborative scheduling optimization engine, a flexible scheduling layer, a digital twin simulation sandbox module, a conflict prediction and dynamic resolution module, and an instruction distribution and execution monitoring module.

[0045] The all-element perception and communication module is used to collect real-time equipment status information, container information, and operation plans from the upper-level system of quay cranes, yard cranes, and AGVs, and to realize bidirectional low-latency transmission of control commands and status feedback.

[0046] The all-element sensing and communication module, acting as the system's "sensors," collects real-time, high-precision dynamic information from all controllable equipment via the IoT platform. This includes the position and status (working / idle / faulty) of the quay cranes and yard cranes, as well as the real-time position, speed, load, and battery charge (SOC) of AGVs. Simultaneously, it receives information from the Terminal Operating System (TOS) such as ship stowage diagrams, yard plans, and container attributes. This module ensures low-latency, highly reliable two-way transmission of all status information and control commands, providing a real-time data foundation for integrated scheduling.

[0047] The unified task modeling and decomposition module is used to receive macro-level work plans and decompose them into atomic task units. Each atomic task unit defines a complete operation for transporting a specific container from one specified resource point to another, and is associated with time windows, priority, and task type attributes.

[0048] The unified task modeling and decomposition module acts as the system's "translator," receiving macro-level loading / unloading instructions or yard overturning instructions from the TOS and decomposing them into standardized, schedulable atomic task units. For example, a "unloading" instruction is decomposed into a series of atomic tasks such as "quay crane A picks up container C at time t1 and places it in the buffer position below it," "AGVX arrives under quay crane A at time t2 to carry container C," "AGVX travels to yard area B," and "yard crane Y picks up container C from AGVX at time t3 and places it in the designated container position." Each atomic task is associated with attributes such as time window, priority, and task type (loading / unloading, transportation, charging), providing a unified input interface for subsequent optimization.

[0049] The collaborative scheduling optimization engine, based on the real-time status provided by the all-element perception and communication module and the atomic task set provided by the unified task modeling and decomposition module, constructs a unified elastic spatiotemporal network model, establishes a multi-objective optimization model with minimizing ship berthing time as the main objective and taking into account the total energy consumption of AGVs and the balance of equipment load, and uses intelligent optimization algorithms to solve the problem, outputting an integrated collaborative scheduling pre-solution.

[0050] The unified elastic spatiotemporal network model is elastic, and the granularity of time slices can be dynamically adjusted according to the scheduling stage and the optimization accuracy requirements. A coarser time granularity is used in the offline macro-planning stage, and a finer time granularity is used in the online rolling optimization stage.

[0051] The collaborative scheduling optimization engine, serving as the "intelligent brain" of the system, is the core of this invention. It comprises a spatiotemporal resource network modeling unit, a multi-objective robust optimization model unit, and an intelligent optimization solution unit.

[0052] Spatiotemporal resource network modeling unit: The physical space of the terminal (quay crane bay locations, road nodes, and storage area bay locations) is discretized into a spatiotemporal grid over a future period, constructing a unified and flexible spatiotemporal network. Nodes represent the "position and state of equipment in a specific time slice," and edges represent "actions" (such as an AGV traveling from node A to node B, or a quay crane performing a loading / unloading operation). Tasks are transformed into "flow" from the starting spatiotemporal node to the ending spatiotemporal node. The "flexibility" of this model is reflected in its adjustable time granularity; coarse granularity (e.g., 1 minute) is used for rapid optimization during macro-planning, while fine granularity (e.g., 5 seconds) is used for precise positioning during rolling optimization.

[0053] Multi-objective robust optimization model unit: A mixed integer programming model is established with minimizing the maximum completion time of the ship as the core, while minimizing the total energy consumption of AGVs (related to travel distance and load), and maximizing the load balance between quay cranes and yard cranes. Its innovative constraints include: (1) Buffer capacity constraint: Set the maximum number of AGVs waiting below each quay crane and yard crane to prevent congestion at loading and unloading points; (2) Charging coordination constraint: Model the AGV power consumption as virtual load. When the power is below the threshold, generate a special task of "going to the charging station" and optimize and sort it together with the production task to achieve "charging along the way"; (3) Robustness constraint: Set a time interval for the single operation time of quay cranes and yard cranes instead of a fixed value so that the generated scheme can still remain feasible when the operation time fluctuates.

[0054] The intelligent optimization solution unit employs an innovative co-evolutionary algorithm based on spatiotemporal chromosome coding and reinforcement learning (RL). The algorithm initializes two sub-populations: a resource allocation population (chromosome-coded for the task execution order and time of quay cranes and field cranes) and a traffic scheduling population (chromosome-coded for AGV task allocation and path sequences). The two populations evolve independently and periodically exchange superior gene fragments (e.g., introducing an efficient quay crane operation time allocation pattern into the traffic scheduling population). In each generation's local optimization, a variable neighborhood search (VNS) process is embedded. The neighborhood structure combinations (such as swaps, insertions, and reversals) and their calling order used by the VNS are not fixed. Instead, a reinforcement learning agent dynamically adjusts the neighborhood structure combinations and calling order used in the variable neighborhood search strategy by continuously learning from historical scheduling data and simulation feedback, adaptively improving the algorithm's convergence speed and solution quality under different operation scenarios.

[0055] The digital twin simulation sandbox module is used to build a high-fidelity virtual image synchronized with the physical terminal. It receives the integrated collaborative scheduling pre-scheduling scheme output by the collaborative scheduling optimization engine, performs accelerated simulation and deduction, evaluates key performance indicators and identifies potential spatiotemporal conflicts and risks, and feeds the evaluation results back to the collaborative scheduling optimization engine for scheme calibration and iterative optimization, forming a "optimization-simulation-verification" decision-making closed loop.

[0056] The digital twin simulation sandbox module serves as the system's "pre-launch laboratory." It constructs a high-fidelity virtual environment that is geometrically and logically synchronized with the physical dock. For example, employing a federated twin architecture, it includes a high-fidelity core sandbox for ultra-real-time (e.g., 50x speed) accurate simulation of the final candidate solutions output by the optimization engine, evaluating KPIs such as completion time and equipment utilization, and utilizing a built-in conflict rule library to detect risks such as deadlock and collisions. It also includes multiple lightweight, fast sandboxes integrated within the optimization engine for rapid (e.g., 500x speed) approximate evaluation and initial screening of numerous intermediate solutions during algorithm iteration, significantly improving optimization iteration efficiency. Simulation results generate evaluation reports that are fed back to the optimization engine, driving it to adjust parameters and re-optimize, forming a closed loop.

[0057] The digital twin simulation sandbox module uses parallel computing and event-driven simulation technology to extrapolate scheduling schemes faster than actual operations. The module has an embedded key conflict rule library that can automatically identify potential deadlock, congestion, and equipment interference risks.

[0058] The digital twin simulation sandbox module adopts a federated twin architecture, which includes a high-fidelity core sandbox for final solution verification, and multiple lightweight fast sandboxes for initial screening and rapid evaluation of solutions within the optimization engine, in order to balance simulation accuracy and decision-making efficiency.

[0059] The conflict prediction and dynamic resolution module predicts spatiotemporal conflicts between devices in the short term online based on the current scheduling scheme and real-time device trajectory data, and dynamically resolves them using speed adjustment, path node reservation, or task sequence fine-tuning strategies.

[0060] The conflict prediction and dynamic resolution module acts as the system's "traffic coordinator." During the execution of the scheduling plan, this module continuously predicts potential conflicts within the next tens of seconds (such as two AGVs meeting at an intersection) based on the current predetermined trajectories and real-time reported positions and speeds of each device. The resolution strategy follows the "speed priority" principle: first, it attempts to make small speed adjustments (acceleration or deceleration) to subsequent devices to stagger their arrival times at the conflict point; if the speed adjustment space is insufficient or the conflict is complex, it "reserves" key path nodes or loading / unloading positions, that is, it locks a certain device to exclusively occupy the resource within a specific time window and instructs other devices to wait, thereby proactively avoiding conflicts.

[0061] The instruction distribution and execution monitoring module is used to parse the final scheduling scheme verified by the digital twin simulation sandbox module into a series of executable instruction sequences for specific devices and issue them out. At the same time, it monitors the instruction execution status and progress deviation in real time, and triggers a dynamic rescheduling process when the deviation exceeds the threshold or a specific event occurs.

[0062] The instruction distribution and execution monitoring module acts as both the system's "executor" and "supervisor." It parses the final solution validated in the twin sandbox into precise timestamp-based equipment-level instructions (e.g., "AGV-005 arrived at quay crane Q07 at 14:05:30, loading container number CONT123"), and distributes them via a reliable communication link. Simultaneously, this module rigorously monitors instruction execution, comparing planned and actual progress. If a critical equipment failure is detected, or the cumulative task delay exceeds a threshold (e.g., 5 minutes), or a higher-priority new vessel operation instruction is inserted, a dynamic rescheduling event is immediately triggered.

[0063] The elastic scheduling layer, located between the collaborative scheduling optimization engine and the instruction distribution and execution monitoring module, is used to receive the scheduling scheme verified by simulation. Based on the real-time perceived system load and device health status, it slightly elastically scales the execution time window of some non-critical tasks in the scheme (e.g., ±30 seconds) as a buffer to absorb the small delays that are difficult to predict in actual execution, avoid frequent rescheduling, and enhance the smoothness of execution.

[0064] Based on the above, a scheduling method for a port quay crane-yard crane-AGV integrated scheduling system is proposed, such as... Figure 2 As shown, it includes the following steps:

[0065] S1. System Initialization and Dynamic Environment Awareness:

[0066] Load the terminal digital map and equipment parameter library, and continuously obtain the real-time status of all controllable equipment and the list of pending tasks through the all-element perception and communication module.

[0067] S2. Rolling Time Domain Optimization Triggering and Task Set Determination:

[0068] A new round of optimization is initiated based on a preset fixed time period (e.g., 5 minutes) or triggered by equipment failure (e.g., quay crane failure), severe deviation in task progress, or the addition of an emergency task event; the set of all atomic tasks to be scheduled in the next optimization time domain is locked.

[0069] The conditions and levels for triggering events are predefined, and different levels of events trigger dynamic rescheduling with different response speeds and optimization ranges.

[0070] S3. Generation and Verification of Integrated Collaborative Scheduling Scheme:

[0071] S31. Call the collaborative scheduling optimization engine, based on the current dock status and the set of atomic tasks to be scheduled, run the intelligent optimization algorithm, solve the multi-objective optimization model, and obtain a preliminary integrated collaborative scheduling pre-scheduling plan;

[0072] S32. Input the preliminary integrated collaborative scheduling scheme into the digital twin simulation sandbox module, perform multi-speed accelerated simulation, evaluate the theoretical performance indicators, and detect potential hidden conflicts and execution risks in the scheme.

[0073] S33. Make a decision based on the simulation evaluation results: If the results meet the preset performance and robustness thresholds, the final executable solution is confirmed; if not, the bottleneck information and risk points identified by the simulation are fed back to the collaborative scheduling optimization engine. After adjusting the model parameters or constraint weights, the process jumps to step S31 for iterative optimization until a satisfactory solution is obtained.

[0074] S4. Issuance of scheduling instructions and online dynamic fine-tuning:

[0075] The instruction distribution and execution monitoring module distributes the instruction sequence corresponding to the recent time period in the final executable plan to the corresponding devices; at the same time, the conflict prediction and dynamic resolution module continuously performs conflict prediction and online fine-tuning based on the plan and real-time trajectory during subsequent execution.

[0076] The online dynamic fine-tuning strategy adopts the principle of "speed adjustment as the main method and path node reservation as the auxiliary method". When there is sufficient buffer time and space between the equipment and the potential conflict point, a smooth speed adjustment strategy is adopted. When the risk of conflict is high, the reservation mechanism for waiting resources at key intersections and loading and unloading positions is activated to avoid conflict by controlling the arrival time of the equipment.

[0077] S5. Execution process monitoring and dynamic rescheduling:

[0078] The actual execution progress of the system monitoring scheme is monitored. When a new fixed optimization cycle is reached, or when a critical equipment failure is detected, the cumulative deviation of the task exceeds the threshold, or a high-priority new task is inserted, dynamic rescheduling is triggered, and the process returns to step S2 to start a new round of global or local rescheduling, thereby achieving continuous adaptive optimization of the system. Specific Implementation Example 1:

[0080] Integrated scheduling in peak operation scenarios for large single vessels:

[0081] A 20,000 TEU ultra-large container ship, "A", arrived at a certain port. After system initialization, it received an operation plan containing 1,500 loading and unloading tasks.

[0082] The system detected that 8 quay cranes, 30 yard cranes, and 80 AGVs were available, and set the optimization time range to 40 minutes.

[0083] The collaborative scheduling optimization engine operates by automatically identifying the ship's structure and dividing the hull into three working zones: forward, mid-course, and aft. Corresponding yard zones and AGV circulation paths are assigned to the quay cranes in each zone, forming "regionalized small loops" that significantly reduce long-distance AGV crossings. Simultaneously, the optimization model inserts fast-charging tasks for AGVs with battery levels below 30% during task breaks. The generated preliminary solution, simulated using digital twins, predicts a 18% reduction in completion time compared to traditional methods.

[0084] The plan was issued and implemented. During the process, a potential congestion at a certain intersection predicted by the simulation was resolved by the conflict resolution module by introducing a 5-second wait instruction to some AGVs. The system completed peak operations with high throughput. Specific Implementation Example 2:

[0086] Multi-ship operations and dynamic priority scheduling scenarios:

[0087] Berth 1 has a high-priority mainline vessel "B", and berth 2 has two feeder vessels "C" and "D". The system needs to handle the operations of the three vessels simultaneously.

[0088] The system assigns differentiated weights to the missions of different ships (e.g., the mission weight of B-ship is 1.0, and that of C / D-ship is 0.7).

[0089] Within a rolling optimization cycle, under the premise of satisfying all basic operational constraints, the optimization engine naturally tilts resources towards the high-weight "B-round" tasks, meaning more AGVs are allocated to the quay cranes serving "B-round" tasks, and the priority of yard crane operations in related yards is also increased. Digital twin simulations verify that this solution can ensure the key "B-round" shifts while also taking into account the progress of "C-round" and "D-round" tasks.

[0090] When the "B round" operation is ahead of schedule, the system will automatically adjust the weights in the next optimization cycle to release more resources to feeder vessels, thus achieving a dynamic balance between fairness and efficiency. Specific Implementation Example 3:

[0092] Emergency Response and Reconfiguration in the Event of Sudden Equipment Failure:

[0093] During the operation, the quay crane "Q05" serving the "E" vessel experienced a sudden main hoisting failure, and was expected to be shut down for repairs for 2 hours.

[0094] The monitoring module detected that "Q05" had no job output for a continuous period of time, which was determined to be a major fault, and immediately triggered an event-driven rescheduling.

[0095] In the new optimization window, the system marks the 80 unfinished tasks originally assigned to "Q05" as "to be reallocated". The optimization engine quickly resolves the problems starting from the current instantaneous state.

[0096] A possible new solution:

[0097] 1) Assign some tasks to the adjacent quay cranes "Q04" and "Q06";

[0098] 2) Adjust the AGV paths and destination stacking areas involved;

[0099] 3) To maintain a balance in yard operations, generate a small number of necessary new "in-yard container relocation" tasks.

[0100] Digital twin simulation quickly assessed the recovery time of the new solution.

[0101] The new scheduling plan was generated and distributed within 3 minutes, and the dock operations quickly returned to an orderly and efficient state after a brief fluctuation, demonstrating the system's strong robustness.

[0102] This invention can achieve a leap in global optimality: through unified spatiotemporal modeling and integrated solution, it realizes millisecond-level coordination of ship, vehicle, and site resources, fundamentally reducing waiting time at the system level and significantly improving theoretical optimality.

[0103] This invention exhibits extremely strong dynamic robustness: by combining rolling optimization, digital twin pre-verification, and event-driven rescheduling, the system can quickly absorb various disturbances, possesses powerful self-healing and adaptive capabilities, and ensures stable execution of the job plan.

[0104] This invention is highly efficient in resource utilization: by incorporating AGV charging and buffer zone management into the main optimization process, it achieves deep synergy between production and energy and space resources, and significantly reduces empty running rate and idle rate.

[0105] This invention enables proactive conflict resolution: by combining in-model constraints with forward-looking predictions at the execution layer, it transforms passive response into proactive avoidance, greatly improving operational safety and process smoothness.

[0106] This invention promotes scientific decision-making: by introducing a digital twin simulation sandbox, a reliable "testing ground" is provided for scheduling schemes, upgrading decision-making from experience-driven to data and model-driven, thereby improving the return on investment.

[0107] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any improvements made using the inventive concept and technical solution of the present invention, or direct application to other occasions without modification, are all within the protection scope of the present invention.

Claims

1. A port quay crane-yard crane-AGV integrated scheduling system, deployed on a central control server at the terminal and connected to terminal equipment via the Industrial Internet of Things, characterized in that, include, The all-element perception and communication module is used to collect real-time equipment status information, container information, and operation plans from the upper-level system of quay cranes, yard cranes, and AGVs, and to realize bidirectional low-latency transmission of control commands and status feedback. The unified task modeling and decomposition module is used to receive macro-level work plans and decompose them into atomic task units. Each atomic task unit defines a complete operation for transporting a specific container from one specified resource point to another, and is associated with time window, priority, and task type attributes. The collaborative scheduling optimization engine, based on the real-time status provided by the full-element perception and communication module and the atomic task set provided by the unified task modeling and decomposition module, constructs a unified elastic spatiotemporal network model, establishes a multi-objective optimization model with minimizing ship berthing time as the main objective and taking into account the total energy consumption of AGVs and the balance of equipment load, and uses intelligent optimization algorithms to solve the problem, outputting an integrated collaborative scheduling pre-solution. The digital twin simulation sandbox module is used to build a high-fidelity virtual image synchronized with the physical terminal. It receives the integrated collaborative scheduling pre-scheduling output from the collaborative scheduling optimization engine, performs accelerated simulation and deduction, evaluates key performance indicators and identifies potential spatiotemporal conflicts and risks, and feeds the evaluation results back to the collaborative scheduling optimization engine for scheme calibration and iterative optimization, forming a "optimization-simulation-verification" decision-making closed loop. The conflict prediction and dynamic resolution module, based on the currently executing scheduling scheme and real-time device trajectory data, predicts the spatiotemporal conflicts between devices in the short time domain in the future in an online rolling manner, and uses speed adjustment, path node reservation or task sequence fine-tuning strategies for dynamic resolution; The instruction distribution and execution monitoring module is used to parse the final scheduling scheme verified by the digital twin simulation sandbox module into a series of executable instruction sequences for specific devices and issue them out. At the same time, it monitors the instruction execution status and progress deviation in real time, and triggers a dynamic rescheduling process when the deviation exceeds the threshold or a specific event occurs.

2. The port quay crane-yard crane-AGV integrated scheduling system according to claim 1, characterized in that, The collaborative scheduling optimization engine specifically includes: The spatiotemporal resource network modeling unit is used to discretize the physical area of ​​the terminal into spatial nodes and, in combination with the discretized time slices, construct a unified spatiotemporal network graph. Among them, network nodes represent the spatial location and operating status of equipment at a specific moment, and network edges represent the actions performed by the equipment when moving between different spatiotemporal nodes, including driving, loading and unloading, waiting, and charging. Container transportation tasks are modeled as the flow demand from the starting resource node to the target resource node on the spatiotemporal network. A multi-objective robust optimization model unit is used to establish a mixed-integer programming model. The objective function includes at least minimizing the maximum completion time among all service vessels, minimizing the total driving energy consumption of all AGVs, and maximizing the load balance between quay cranes and yard cranes. In addition to equipment capacity and basic task logic constraints, the constraints also include: setting buffer capacity constraints for each quay crane and yard crane to limit the number of AGVs waiting below at the same time; modeling the battery power consumption and charging demand of AGVs as special tasks that can be inserted into the scheduling sequence to achieve coordination between production and energy supply; and introducing robustness constraints for critical operation durations to generate scheduling schemes that are insensitive to uncertainty. The intelligent optimization solution unit employs a co-evolutionary algorithm based on spatiotemporal chromosome encoding and reinforcement learning to efficiently solve multi-objective robust optimization models. The algorithm maintains two co-evolutionary subpopulations for resource allocation and traffic scheduling, which optimize the quay crane / yard crane operation sequence and AGV path planning and fleet allocation, respectively. The subpopulations co-evolve by exchanging superior spatiotemporal scheduling segments and embed a variable neighborhood search strategy guided by a reinforcement learning agent for local deep optimization.

3. The port quay crane-yard crane-AGV integrated scheduling system according to claim 2, characterized in that, The unified elastic spatiotemporal network model is elastic, and the granularity of time slices can be dynamically adjusted according to the scheduling stage and the optimization accuracy requirements. A coarser time granularity is used in the offline macro-planning stage, and a finer time granularity is used in the online rolling optimization stage.

4. The port quay crane-yard crane-AGV integrated scheduling system according to claim 2, characterized in that, In the intelligent optimization solution unit, the reinforcement learning agent continuously learns from historical scheduling data and simulation feedback, and dynamically adjusts the neighborhood structure combination and calling order used in the variable neighborhood search strategy to adaptively improve the convergence speed and solution quality of the algorithm in solving different job scenarios.

5. The port quay crane-yard crane-AGV integrated scheduling system according to claim 1, characterized in that, The digital twin simulation sandbox module uses parallel computing and event-driven simulation technology to extrapolate scheduling schemes faster than actual operations. The module has an embedded key conflict rule library that can automatically identify potential deadlock, congestion, and equipment interference risks.

6. The port quay crane-yard crane-AGV integrated scheduling system according to claim 5, characterized in that, The digital twin simulation sandbox module adopts a federated twin architecture, which includes a high-fidelity core sandbox for final solution verification, and multiple lightweight fast sandboxes for initial screening and rapid evaluation of solutions within the optimization engine, in order to balance simulation accuracy and decision-making efficiency.

7. The port quay crane-yard crane-AGV integrated scheduling system according to claim 1, characterized in that, It also includes an elastic scheduling layer, located between the collaborative scheduling optimization engine and the instruction distribution and execution monitoring module. This layer receives the simulation-verified scheduling scheme and, based on the real-time perceived system load and device health status, slightly expands and contracts the execution time windows of some non-critical tasks in the scheme to further absorb micro-disturbances and enhance the compliance of execution.

8. A scheduling method for a port quay crane-yard crane-AGV integrated scheduling system according to any one of claims 1-7, characterized in that, Includes the following steps: S1. System Initialization and Dynamic Environment Awareness: Load the terminal digital map and equipment parameter library, and continuously obtain the real-time status of all controllable equipment and the list of pending tasks through the full-element perception and communication module; S2. Rolling Time Domain Optimization Triggering and Task Set Determination: A new round of optimization is initiated based on a preset fixed time period or triggered by equipment failure, serious deviation in task progress, or the addition of an emergency task; the set of all atomic tasks to be scheduled in the next optimization time domain is locked. S3. Generation and Verification of Integrated Collaborative Scheduling Scheme: S31. Call the collaborative scheduling optimization engine, based on the current dock status and the set of atomic tasks to be scheduled, run the intelligent optimization algorithm, solve the multi-objective optimization model, and obtain a preliminary integrated collaborative scheduling pre-scheduling plan; S32. Input the preliminary integrated collaborative scheduling scheme into the digital twin simulation sandbox module, perform multi-speed accelerated simulation, evaluate the theoretical performance indicators, and detect potential hidden conflicts and execution risks in the scheme. S33. Make a decision based on the simulation evaluation results: If the results meet the preset performance and robustness thresholds, the final executable solution is confirmed; if not, the bottleneck information and risk points identified by the simulation are fed back to the collaborative scheduling optimization engine. After adjusting the model parameters or constraint weights, the process jumps to step S31 for iterative optimization until a satisfactory solution is obtained. S4. Dispatch instruction issuance and online dynamic fine-tuning: The instruction distribution and execution monitoring module distributes the instruction sequence corresponding to the recent time period in the final executable plan to the corresponding devices; at the same time, the conflict prediction and dynamic resolution module continuously performs conflict prediction and online fine-tuning based on the plan and real-time trajectory during subsequent execution. S5. Execution process monitoring and dynamic rescheduling: When the actual execution progress of the system monitoring scheme reaches a new fixed optimization cycle, or when a critical equipment failure is detected, the cumulative deviation of the task exceeds the threshold, or a high-priority new task is inserted, dynamic rescheduling is triggered, and the process returns to step S2.

9. The scheduling method of a port quay crane-yard crane-AGV integrated scheduling system according to claim 8, characterized in that, In step S2, the conditions and levels for triggering events are predefined, and different levels of events trigger dynamic rescheduling with different response speeds and optimization ranges.

10. The scheduling method of a port quay crane-yard crane-AGV integrated scheduling system according to claim 9, characterized in that, In step S4, the online dynamic fine-tuning strategy adopts the principle of "speed adjustment as the main method and path node reservation as the auxiliary method". When there is sufficient buffer time and space between the equipment and the potential conflict point, a smooth speed adjustment strategy is adopted; when the conflict risk is high, the reservation mechanism for key intersections and loading / unloading positions is activated to avoid conflicts by controlling the arrival time of the equipment.

Images