Port truck scheduling optimization system and method based on dynamic time matrix prediction

Abstract

This invention discloses a port-based truck scheduling optimization system and method based on dynamic time matrix prediction. By collecting TOS (Transportation Time of Sale) and vehicle monitoring data, a dynamic time matrix incorporating crane rhythm and lane impedance factors is constructed, and task results are used to correct the matrix. The system adaptively adjusts the truck resource pool threshold based on a task density function and employs tabu search and discrete particle swarm optimization algorithms for multi-objective scheduling optimization. When there is congestion in the yard or rhythm deviation, task priority and key road markers are considered to trigger task reordering, and algorithm parameters are automatically updated based on execution results. This system achieves a real-time closed loop of prediction-scheduling-feedback, improving port operation efficiency and resource utilization.

Description

Technical Field

[0001] This invention relates to the field of intelligent port scheduling and transportation optimization technology, specifically to a port truck scheduling optimization system and method based on dynamic time matrix prediction. Background Technology

[0002] With the continuous growth of container shipping volume and the increasing trend towards larger vessels, the complexity of equipment scheduling within terminals has significantly increased. Port operational efficiency largely depends on the coordinated operation of quay cranes, internal trucks, and yard cranes. Traditional scheduling methods typically rely on fixed rules or manual experience, lacking the ability to respond to real-time road conditions and dynamic task requirements, leading to problems such as operational delays, quay crane outages, and empty truck runs. Existing technologies suffer from the following technical shortcomings:

[0003] 1. The problem of static time matrix. The time matrix in the existing scheduling system is mostly calculated based on fixed parameters, which fails to reflect dynamic factors such as the terminal road network structure, road direction, number of turns and real-time congestion. It is also difficult to accurately predict the travel time of trucks in the terminal, resulting in a large deviation between the scheduling results and the actual operation status, which affects the efficiency of task connection.

[0004] 2. The scheduling algorithm and time prediction are not deeply integrated. Generally, the strategy is based on first-come, first-served or task priority matching, which lacks quantitative optimization of truck travel time, waiting time and route cost. It is difficult to achieve global scheduling with time optimization as the goal, resulting in high empty truck rate and low turnover efficiency.

[0005] 3. Insufficient coordination among multiple equipment operations. The existing system has failed to establish a time coordination mechanism between the gantry crane, internal trucks, and yard cranes. It cannot dynamically adjust the gantry crane operation rhythm and yard crane preparation time based on the estimated arrival time of the trucks, which can easily lead to truck waiting or gantry crane interruptions, resulting in unstable operation rhythm.

[0006] 4. Weak data support and real-time response capabilities. The system fails to effectively integrate multi-source information such as truck GPS tracks, task execution time, and road congestion status. In the face of equipment malfunctions or sudden traffic jams, it lacks an adaptive adjustment mechanism based on real-time data, resulting in slow response times and potential task delays.

[0007] In view of this, the present invention proposes a port container truck scheduling optimization system and method based on dynamic time matrix prediction. Summary of the Invention

[0008] The purpose of this invention is to provide a port container truck scheduling optimization system and method based on dynamic time matrix prediction, which aims to solve the problems of static time matrix, algorithm response lag and lack of time coordination among multiple devices in the existing port container truck scheduling system.

[0009] In a first aspect, the present invention provides a port container truck scheduling optimization system based on dynamic time matrix prediction, including a dynamic time matrix prediction module, a dynamic container truck resource pool configuration module, a loading and unloading task scheduling algorithm module, a task sequence optimization recommendation module, and an algorithm parameter configuration module:

[0010] The dynamic time matrix prediction module is used to receive historical and real-time data from the port operating system and vehicle monitoring terminal, extract road network nodes, vehicle trajectories and gantry crane operation status information, construct a travel time matrix that changes with the operating load, and introduce gantry crane rhythm coefficient and road impedance factor into the generation of the travel time matrix, and periodically update the travel time between any nodes.

[0011] The dynamic truck resource pool configuration module is used to receive the time matrix and task data, divide the truck resource pool according to the operation route, construct a task density function based on the task arrival rate and adaptively adjust the minimum and maximum truck thresholds, and output the dynamic resource pool status.

[0012] The loading and unloading ship task scheduling algorithm module takes the time matrix and resource pool status as input, establishes a two-way matching relationship between tasks and vehicles, and uses a hybrid heuristic algorithm to optimize job waiting, path distance and equipment coordination.

[0013] The task order optimization and recommendation module combines the yard congestion status and crane rhythm deviation information to perform local rearrangement and order correction on the preliminary scheduling results;

[0014] The algorithm parameter configuration module dynamically adjusts and optimizes function parameters based on runtime feedback and user settings, achieving closed-loop coordination of prediction, scheduling, and feedback.

[0015] As a preferred embodiment of the first aspect of the present invention, the dynamic time matrix prediction module includes:

[0016] Obtain the coordinates of road network nodes, edge connections, and travel directions; generate and store a road topology table containing node numbers, edge numbers, and direction attributes.

[0017] Obtain the operation cycle, operation time period and spatial location of the bridge crane, establish and store the rhythm parameter table of the bridge crane and adjacent nodes;

[0018] Calculate the road impedance value for each travel side (composed of road width, turning angle, and average dwell time), generate an impedance parameter table, and store it.

[0019] The rhythm parameter table and impedance parameter table are aligned and weighted according to the road topology table. The travel time of each node is calculated, and the travel time matrix containing the bridge crane rhythm coefficient and road impedance factor is output for the scheduling module to call.

[0020] As a preferred embodiment of the first aspect of the present invention, the periodic update of the driving time matrix includes:

[0021] Read the predicted driving time and the collected actual driving time from the previous cycle and calculate the difference to generate a difference record table;

[0022] Based on the difference record table, the bridge crane rhythm coefficient and road impedance factor are proportionally corrected to generate a correction parameter table;

[0023] Recalculate the node pair travel time according to the revised parameter table to generate a new travel time matrix file;

[0024] Write the new driving time matrix file to the database to replace the old version and output the revised driving time matrix for use by the resource pool configuration module.

[0025] As a preferred embodiment of the first aspect of the present invention, the dynamic card pool configuration module includes:

[0026] Read the task arrival time series data and count the number of tasks in each time period to generate a task distribution table;

[0027] Calculate the task arrival rate function based on the task distribution table to form a task density value table;

[0028] Based on the task density value, the minimum and maximum number of container trucks for each work route are calculated to generate a threshold table;

[0029] The threshold table is combined with the existing vehicle list to form a resource pool status table, and the resource pool status information is output for use by the task scheduling algorithm module.

[0030] As a preferred embodiment of the first aspect of the present invention, the resource pool status adjustment process includes:

[0031] Read the task density value table and threshold table to generate an adjustment instruction table for each work route;

[0032] According to the adjustment instruction table, the number of trucks for the corresponding route is increased or decreased to form an adjustment result table;

[0033] Compare the resource pool status table with the adjustment result table and verify the vehicle load rate and standby time for each route;

[0034] The updated resource pool status table is output and sent to the task scheduling algorithm module for task matching in the next cycle.

[0035] As a preferred embodiment of the first aspect of the present invention, the task-vehicle bidirectional matching process includes:

[0036] Read the dynamic time matrix and resource pool status table to obtain the set of allocable tasks and the set of available vehicles;

[0037] Calculate the task-vehicle associated cost matrix based on task urgency, vehicle location, and travel time;

[0038] Perform a hybrid optimization of tabu search and discrete particle swarm optimization on the cost matrix to obtain a globally feasible matching solution set;

[0039] Select the correspondence with the minimum cost in the matching solution set and generate a task allocation instruction table to be output to the scheduling and execution module.

[0040] As a preferred embodiment of the first aspect of the present invention, the task and vehicle matching optimization process includes:

[0041] Tasks are processed in a priority hierarchy, and vehicles are grouped according to their idle time.

[0042] In each level, the matching weight is calculated based on the path length and resource pool constraints;

[0043] The weight parameters are adjusted iteratively by particle swarm optimization to converge to the optimal allocation solution;

[0044] Output the iteratively optimized task-vehicle matching table and send it to the task execution scheduling unit.

[0045] As a preferred embodiment of the first aspect of the present invention, the task priority optimization and recommendation process includes:

[0046] Receive preliminary scheduling results and yard congestion data to construct a task execution timetable;

[0047] Calculate the difference between the bridge crane rhythm deviation information and the task waiting time to identify reorderable task pairs;

[0048] Adjust the order of task execution based on the recognition results and generate a revised sequence list;

[0049] The output task execution sequence, after local rearrangement and order correction, is provided for use by the algorithm parameter configuration module.

[0050] As a preferred embodiment of the first aspect of the present invention, the task execution sequence correction includes:

[0051] Read the corrected sequence list and the crane rhythm data to establish a task time deviation model;

[0052] The mission start time and vehicle arrival time are synchronized and adjusted according to the deviation model.

[0053] Generate a synchronized and corrected task schedule;

[0054] The corrected task schedule is output for the algorithm parameter configuration module to update the optimization function parameters.

[0055] Secondly, the present invention provides a port truck scheduling optimization method based on dynamic time matrix prediction, comprising the following steps:

[0056] Historical and real-time operational data are collected from the port operating system and vehicle monitoring terminals to generate a basic dataset containing road network nodes, vehicle trajectories, and crane status.

[0057] Using the basic dataset, a travel time matrix incorporating the bridge crane rhythm coefficient and road impedance factor is constructed and periodic updates are performed to obtain the travel time prediction results between nodes;

[0058] Based on the prediction results and task arrival rate, establish a resource pool status table and calculate the minimum and maximum truck thresholds for each operation route;

[0059] Under resource pool constraints, a hybrid algorithm combining tabu search and discrete particle swarm optimization is used to establish a two-way matching relationship between tasks and vehicles and generate a task allocation instruction table.

[0060] By combining information on yard congestion and crane rhythm deviation, the preliminary scheduling results are rearranged and time-series corrected to form a task execution sequence.

[0061] The execution sequence and runtime feedback are input into the algorithm parameter configuration module to update and optimize function parameters and output an optimized scheduling scheme.

[0062] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0063] This invention achieves spatiotemporal adaptive optimization of container truck scheduling within ports by introducing a dynamic time matrix prediction and task feedback correction mechanism. Through a resource pool threshold function based on task density and a hybrid heuristic interaction algorithm, a global and local collaborative scheduling model is constructed, maintaining real-time response under high concurrency conditions. The task sequencing module uses yard congestion and crane rhythm as triggering conditions, significantly reducing waiting and empty runs. The parameter configuration module has self-learning capabilities, automatically adjusting optimization weights based on execution results. Compared to traditional static scheduling, this invention achieves an integrated closed loop of prediction, scheduling, and feedback, and enhances the system's scalability and robustness. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0065] Figure 1 This is a system architecture diagram of the port container truck scheduling optimization system of the present invention;

[0066] Figure 2 This is a block diagram of the core algorithm logic of the port container truck scheduling optimization system of the present invention;

[0067] Figure 3 This is a detailed map showing the dock road network and the areas where container trucks can travel.

[0068] Figure 4 This is a diagram showing the relationship between the card pool and the operating route in this invention;

[0069] Figure 5 This is a framework diagram of the port truck scheduling optimization system of the present invention;

[0070] Figure 6 This is a flowchart of the port container truck scheduling optimization method of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings.

[0072] Throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. The described embodiments are only a part of the embodiments of this application, not all of them. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0073] Example 1

[0074] Please see Figure 1-2 As shown, this embodiment provides a port truck scheduling optimization system based on dynamic time matrix prediction, which is used to achieve collaborative optimization of multi-vehicle task allocation and route planning in the complex operating environment of container terminals. Through multi-layer mechanisms such as time matrix prediction, task priority decision, dynamic scheduling and feedback learning, a real-time self-optimizing scheduling closed loop is constructed in the port operation process.

[0075] It should be noted that, as Figure 1As shown, the system is based on a microservice architecture, with each interface service having its own process and communicating using a lightweight mechanism (usually a HTTP-based RESTful API). It is built around the business functions corresponding to the business layer and is deployed independently through an automated deployment mechanism. Each functional module runs as an independent service, forming a distributed microservice system for intelligent port scheduling. This ensures that the scheduling algorithm can still run efficiently and the service can be smoothly expanded even in high-concurrency, multi-tasking, and complex road network environments.

[0076] Specifically, the system in this embodiment is divided into four main levels, including:

[0077] 1. The foundation layer is the underlying support environment for system operation, including components such as containerized deployment, databases, and big data computing engines.

[0078] Docker & Kubernetes: Enable containerized management and elastic scaling of all algorithms and service modules;

[0079] Oracle & MySQL: Provide relational storage for business data and historical job data;

[0080] OLAP MDX & ETL Kettle: Perform data extraction (ETL), cleaning, and multidimensional analysis;

[0081] Spark Distributed Computing Engine: Handles highly parallel computing tasks for dynamic time matrix prediction and scheduling algorithms.

[0082] 2. The business layer is the core logic layer of the system, mainly responsible for algorithm execution, task management, and microservice registration. Its structure includes:

[0083] The business layer Spring Cloud framework serves as a business service framework, encapsulating the business logic and API calls of each functional module.

[0084] Consul: Used for microservice registration and health monitoring, enabling automatic service discovery and disaster recovery switching;

[0085] Scheduled task Jenkins management: Automatically schedules algorithm execution and task triggering to ensure the algorithm runs periodically;

[0086] Kafka: A distributed message queue used for scheduling task transmission and asynchronous result return in high-concurrency environments;

[0087] Scheduling Algorithm Python + Spark Library: This library carries the core algorithm logic, including dynamic time matrix prediction, card resource pool configuration, multi-objective scheduling, and task ordering optimization.

[0088] 3. The load balancing gateway layer achieves high availability and balanced access through multi-node load distribution and request management, including:

[0089] Nginx: A reverse proxy and load balancer server responsible for forwarding requests from the front end and external systems;

[0090] Spring Cloud Zuul: API Gateway Service, used for routing control and unified API management, enabling cross-module communication and security filtering.

[0091] 4. The application layer interacts with users and external systems, providing a visual user interface and interface services, and consists of the following:

[0092] Web-based React: A web-based visual interface built on the React framework, used for scheduling, configuration, and status monitoring;

[0093] WebSocket Real-Time Center: Enables real-time push and dynamic display of scheduling execution status;

[0094] Mobile App React Native: A mobile management client that enables task monitoring and anomaly alerts;

[0095] RESTful API: A standardized interface service that provides data access and algorithm invocation interfaces for external systems (such as TOS and CWP);

[0096] Echarts 4: Used to draw dynamic charts for real-time monitoring, scheduling results, and performance analysis.

[0097] Meanwhile, a unified security control module is configured on the right side of the system, spanning all layers, for: unified identity authentication and permission verification; encrypted data access and transmission; system logs and anomaly monitoring; and centralized management of security policies. Through multi-layered protection mechanisms, the system ensures communication security and data integrity in multi-tenant, multi-terminal access scenarios, and guarantees secure and encrypted transmission of data interaction and service calls.

[0098] The system architecture diagram illustrates the complete hierarchical structure of the system, from underlying computing resources to upper-layer application interfaces, and explains the physical dependencies and communication relationships between modules. The system achieves module decoupling through the Spring Cloud microservice architecture, message passing and asynchronous collaboration through Kafka, and all algorithm services are deployed independently in containers, thereby enabling high-concurrency processing and scalable deployment.

[0099] The core algorithm logic of the system is as follows Figure 2 As shown, the functional logic is structured in three layers: the data layer (interface layer), the algorithm layer, and the user interface layer. Wherein:

[0100] The data layer interfaces with the port operating system (TOS) and database system, and is responsible for receiving data such as work tasks, equipment status, and vehicle trajectories, and cleaning and structuring the data to provide standardized input for the algorithm layer.

[0101] The algorithm layer is the core of the system's decision-making, comprising a scheduling algorithm module, a path planning algorithm module, a truck and task data processing module, and a scheduling instruction data processing module. This layer achieves global optimization of task allocation and path planning through dynamic time matrix prediction and multi-objective optimization methods.

[0102] The user interface layer is for displaying and interacting with algorithm results, including a parameter configuration module, a scheduling monitoring module, and a scheduling analysis module. This layer allows users to select different scheduling strategy modes, monitor task execution status in real time, and generate scheduling result reports for analysis.

[0103] Figure 2 The algorithm logic architecture diagram illustrates the calling relationships and operational logic between algorithm modules within the system from a data flow perspective. The job and status data input into the TOS system are processed by the algorithm layer to generate scheduling instructions, which are then output to the user interface for result display and monitoring, thus forming a closed loop of data acquisition, model computation, and result output.

[0104] This can be understood as, Figure 1 Provides the hardware and software support environment for algorithm execution. Figure 2 The algorithm's computation and decision-making process in this environment is demonstrated, including: Figure 2 Can be regarded as Figure 1 The diagram showing the expanded functionality of the "Business Layer and Algorithm Engine" shows the hierarchical correspondence. Figure 1 base layer corresponding Figure 2 The data layer, together with the data layer, provides support for data acquisition, storage, and computation. Figure 1 Business layer corresponding Figure 2 The algorithm layer is the core decision-making center of the system; Figure 1 Application layer correspondence Figure 2 The user interface layer is used for visualization and interaction; Figure 1 The unified security control module runs through Figure 2 All layers ensure the security and stability of system operation. The load balancing gateway layer is located at the connection point between the two diagrams and is responsible for the secure transmission of data flow and service calls.

[0105] In terms of operational process, Figure 1 and Figure 2 This forms a closed-loop logic encompassing structural support, algorithm operation, and result output. First, in the data input phase, the system's foundation layer collects job data from the TOS and database, and performs ETL cleaning. Second, in the algorithm calculation phase, the business layer calls Python and the Spark engine to execute dynamic time matrix prediction and multi-objective scheduling. Next, in the result output phase, the calculation results are transmitted to the application layer's web and mobile interfaces via a RESTful API. Finally, a unified security control module implements access encryption and authentication, thus completing a secure closed loop throughout the entire process.

[0106] This system employs a highly available architecture at both the algorithm and database layers to ensure continuous operation and fault tolerance in production environments. The algorithm module is deployed across three servers, all running the same algorithm process. At any given time, only one server is in execution mode, responsible for algorithm computation and task scheduling, while the other two are in monitoring mode. The monitoring servers monitor the execution servers' operational status in real time through periodic heartbeat checks and a distributed lock mechanism. When an execution server experiences an anomaly (e.g., system failure, service interruption, or network failure), the servers in monitoring mode automatically initiate a lock-grabbing operation and take over task execution. This achieves automatic failover of the algorithm service, ensuring stable system operation even under single-point-of-failure conditions and maintaining the continuity and reliability of production tasks.

[0107] The database layer employs a high-availability solution combining MariaDB master-slave replication and Keepalived Virtual IP (VIP). The system consists of two physical databases and one virtual VIP. Client applications access the database service through the VIP, which normally points to the master database node. The master node synchronizes data to the standby node in real time using MariaDB's master-slave replication mechanism.

[0108] Keepalived periodically checks the health status of database nodes. When a primary node failure is detected, it automatically triggers a VIP migration mechanism, migrating the VIP to a backup node, allowing clients to continue accessing the database service without any disruption. In this architecture, the two MariaDB instances employ dual-master replication for rapid failover and a single-write strategy for write conflict isolation. Each node acts as both a primary and a secondary node, achieving bidirectional data synchronization and real-time consistency. Through this design, both the algorithm computation layer and the database storage layer possess automatic detection, automatic failover, and automatic recovery capabilities, significantly improving the system's fault tolerance and overall availability, ensuring continuous and stable operation in complex production environments.

[0109] In other words, Figure 1 This explains the system's engineering implementation and service deployment methods. Figure 2The internal logic and data flow mechanism of the algorithm are explained. These two aspects complement each other and are clearly layered, together forming an integrated system of "architectural support + algorithm execution." Through this architectural design, this system can achieve high-concurrency task processing, real-time path optimization, and safe and stable intelligent scheduling operation in complex port operation environments.

[0110] To further explain, the core functional modules of this system include a dynamic time matrix prediction module, a dynamic container truck resource pool configuration module, a loading and unloading task scheduling algorithm module, a task priority optimization and recommendation module, and an algorithm parameter configuration module, all of which are deployed on... Figure 1 In the business layer shown, and corresponding to Figure 2 The algorithm layer logic structure shown is the core computing unit for the system to achieve intelligent scheduling optimization. Specifically:

[0111] The dynamic time matrix prediction module runs in the business layer's Python + Spark algorithm environment, corresponding to... Figure 2 The algorithm layer includes path planning and time prediction. This module collects data from the base layer, such as the dock road network, vehicle trajectories, and crane locations, to construct a dynamic time matrix, providing real-time travel time predictions for the scheduling algorithm.

[0112] The dynamic card resource pool configuration module runs in the same business layer, corresponding to Figure 2 The task and vehicle data processing section. This module establishes a binding relationship between the truck resource pool and the work route based on the input data from the TOS system and database, sets minimum and maximum truck thresholds, and forms resource constraints for scheduling.

[0113] The ship loading and unloading task scheduling algorithm module is the core decision engine of the system, corresponding to... Figure 2 The scheduling algorithm section of the module takes a dynamic time matrix and resource pool information as input and uses a multi-objective optimization method to solve the task allocation and path planning of trucks, achieving the optimal scheduling scheme under the efficiency-first, benefit-first, and balanced modes.

[0114] Task priority optimization recommendation module Figure 2 The task execution logic optimization part in the algorithm layer. Based on the scheduling results, this module dynamically adjusts the task execution order according to the yard congestion status and the operation rhythm of the gantry crane, realizing time coordination and continuous operation optimization among multiple devices.

[0115] Algorithm parameter configuration module throughout Figure 1 The application layer and the business layer, corresponding to Figure 2 The user interface layer allows users to input scheduling strategies and parameter weights through the interface. The system then transmits the configuration information to the business layer via the load gateway layer, enabling adaptive adjustment of the algorithm's operating mode and strategy switching.

[0116] In summary, the five core modules of the system correspond structurally to Figure 1 The business layer logically corresponds to Figure 2 The algorithm layer relies on the foundational layer for data support and the application layer for display and control in terms of interaction. Figure 1 , Figure 2 Together with the core functional modules, they constitute the complete technical logic of the system: the operation process can be summarized into the following four stages:

[0117] The first stage is the data acquisition and preprocessing stage, in which the basic layer collects task data, equipment status and GPS trajectory information from TOS and database, and completes data cleaning and formatting through the ETL process.

[0118] The second stage is the algorithm calculation and optimization stage, in which the business layer calls the Python + Spark engine to perform dynamic time matrix prediction, truck resource pool management and multi-objective scheduling optimization.

[0119] The third stage is the result output and monitoring stage. The algorithm calculation results are pushed to the application layer interface via the RESTful API interface and displayed visually through the web and mobile terminals.

[0120] The fourth stage is the security control and feedback stage, in which a unified security control module realizes encrypted data transmission, identity verification and task log recording, and feeds back the execution results to the database, forming a continuous optimization closed loop.

[0121] Through the above architecture and module collaboration, this system achieves intelligent management of the entire process from data acquisition, algorithm calculation, scheduling decision-making to visual feedback. It has high concurrency processing capabilities, scalability and good fault tolerance, and can operate stably in complex port operation environments, significantly improving the operational efficiency and equipment utilization of container terminals.

[0122] Example 2

[0123] This embodiment, based on Embodiment 1, focuses on intelligent scheduling of internal container trucks using dynamic time matrix prediction and fused time prediction. It leverages existing data from the Guandong Container Terminal's TOS (Terminal Operating System) and CWP (Machinery Scheduling System), integrating operations research optimization, big data analysis, and heuristic algorithms to address the problems of static time matrices and insufficient multi-equipment coordination in traditional scheduling. The core content of this embodiment's port internal container truck scheduling optimization system based on dynamic time matrix prediction is further disclosed, such as... Figure 5 As shown, it includes a dynamic time matrix prediction module, a dynamic container truck resource pool configuration module, a loading and unloading task scheduling algorithm module, a task sequence optimization and recommendation module, and an algorithm parameter configuration module, forming a full-process intelligent scheduling chain from data perception, time prediction, task allocation to execution optimization.

[0124] The dynamic time matrix prediction module receives historical and real-time data from the port operating system (TOS) and vehicle monitoring terminals, extracts port road network nodes, vehicle trajectories and crane operation status information, and constructs a travel time matrix that varies with the operating load. The crane rhythm coefficient and road impedance factor are introduced into the travel time matrix, and the predicted travel time between any node pairs is periodically updated through a distributed computing engine, and the update results are output to downstream modules.

[0125] This can be understood as providing spatiotemporal cost information to extract road nodes, vehicle locations, and operational status information from historical and real-time driving data collected from the port operating system (TOS) and vehicle monitoring terminals, constructing a dynamic time matrix model. Based on the Spark distributed computing engine, it performs parallel modeling and prediction of driving times for different operating periods, route segments, and equipment states, outputting a driving time matrix between any operating units. In traditional time matrix methods, only an electronic map of the yard area is typically constructed, and tasks are allocated based on the start and end locations of truck tasks. However, the real-time location and operational status of the gantry cranes are not effectively updated and utilized, resulting in significant deficiencies in the spatial accuracy and real-time performance of the time matrix, making it difficult to accurately reflect traffic characteristics in dynamic operating environments. The dynamic time matrix prediction module, by introducing a real-time positioning and status update mechanism, utilizes GPS-collected truck location data and gantry crane location change information in the database to update the electronic map and road network status in real time, thereby significantly improving the accuracy and timeliness of time matrix prediction and enhancing the responsiveness to spatiotemporal changes in complex operating scenarios.

[0126] It should be noted that by organizing and analyzing historical data of container trucks, the computational features affecting truck travel time are extracted, improving the accuracy of the truck time matrix. This leads to more effective task scheduling, increased truck turnaround efficiency, and a reduced ratio of gantry cranes to container trucks. Through data analysis, a dynamic time matrix (travel time between locations) is provided to the Truck Scheduling System (TOS). This dynamic time matrix is ​​provided via a RESTful algorithm service interface, with an algorithm computation service built in the backend. The dynamic time matrix needs to consider the actual road network structure at the terminal, lane directions, the real-time operating location of the gantry cranes, the number of turns, and the frequency of road congestion. The time matrix provides the arrival time of container trucks at pick-up / delivery points. Accurate arrival times provide information for the truck scheduling program to assess the costs and consequences of each allocation scheme, such as… Figure 3 The diagram shows the port road network, with detailed markings of areas accessible to container trucks. This serves as the foundational input for subsequent scheduling calculations, reflecting changes in traffic congestion and dynamic capacity within the port area, enabling real-time estimation of vehicle travel times for future operations.

[0127] This can be understood as establishing a time-varying matrix model using static road network data from the port (including latitude and longitude coordinates, lane directions, intersections, etc.) and real-time operational data (including the location of internal trucks, the location of crane operations, and road congestion status). The Spark distributed computing engine calculates the travel time between any nodes, and the matrix is ​​dynamically corrected using real-time congestion coefficients. The matrix update cycle is 3 seconds, and the prediction error is controlled within ±3 seconds, accurately reflecting crane interference and time-varying changes in the road network. Compared to traditional fixed-time matrices, this scheme significantly improves time prediction accuracy, provides accurate input for scheduling algorithms, and avoids truck waiting and task delays caused by time deviations.

[0128] The dynamic truck resource pool configuration module receives the dynamic time matrix prediction results and job task data, divides the truck resource pool according to the job route, constructs a task density function based on the task arrival rate, calculates the minimum and maximum truck thresholds every 15 seconds, forms resource pool status information that is dynamically adjusted with the task load, and outputs this status as a constraint condition to the scheduling algorithm module.

[0129] In other words, resource pools are divided according to crane operation routes, and minimum and maximum truck number thresholds are set for each route. The system automatically adjusts the thresholds based on the real-time task density function, adding trucks when the task volume increases and releasing vehicles when the task volume decreases, thus achieving load balancing. In traditional resource pool configuration methods, the number of trucks corresponding to each operation route is usually fixed. This static configuration cannot reflect the dynamic changes in task volume. When some operation routes have fewer tasks, truck resources are idle for a long time, while operation routes with more tasks suffer from insufficient trucks, leading to task delays and increased waiting time, thereby reducing overall scheduling efficiency. The dynamic truck resource pool configuration module, by introducing a real-time task awareness and threshold adaptive adjustment mechanism, can automatically adjust the number of trucks according to changes in task load, achieving dynamic balance and efficient utilization of resources, significantly improving the system's responsiveness and operational continuity.

[0130] In addition, a fixed number of trucks is set for key operating routes to ensure operational continuity. Actual testing showed that the resource pool mechanism keeps the truck utilization rate deviation between operating routes within 5%. During ship loading and unloading scheduling, a multi-objective optimization algorithm uses "empty driving distance" as one of the optimization indicators (with a weight of 0.3 in the balanced mode), combined with a dynamic time matrix to generate the optimal route plan. This can reduce the average empty driving distance of internal trucks by more than 10%, corresponding to a 10%–15% reduction in fuel consumption, significantly reducing terminal operating costs.

[0131] This can be understood as follows: based on the port area's operational road network structure and task priorities, a container truck resource pool is constructed, and minimum and maximum available quantity thresholds are set; the output resource pool status information serves as a constraint condition for the loading and unloading task scheduling algorithm module, thereby realizing the dynamic allocation and adaptive scheduling of vehicle resources.

[0132] Specifically, trucks and work routes are bound to a fixed truck pool. Trucks within a pool only serve the bound work routes. Business personnel can configure truck pools based on future work plans. A truck pool can be associated with multiple work routes and trucks. A truck work route and truck pairing can only be assigned to one truck pool, such as... Figure 4 As shown, minimum and maximum thresholds for the number of trucks simultaneously operating on a work route are set. The system dynamically schedules tasks within these threshold ranges, maintaining a dynamic balance of tasks on the work route and reducing congestion. When the number of trucks operating simultaneously is less than the minimum number, the next scheduled task must meet the minimum number of trucks. When the minimum number is less than or equal to the number of trucks operating simultaneously but less than the maximum number, the system gradually increases the number of scheduled tasks based on round-robin scheduling. When the number of trucks operating simultaneously equals the maximum number, the system temporarily suspends scheduling. A fixed number of trucks can be set for key routes to ensure the efficiency and continuity of operations on these routes.

[0133] The loading and unloading task scheduling algorithm module takes the dynamic time matrix and resource pool status as unified inputs, and uses a heuristic optimization algorithm that combines tabu search and discrete particle swarm optimization to establish a multi-objective model. It performs task allocation and path planning under efficiency-first, benefit-first, and balance modes, and outputs preliminary scheduling results and task allocation table.

[0134] This can be understood as generating the optimal scheduling scheme for loading and unloading operations based on dynamic time matrix prediction results and truck resource pool status information. It employs a multi-objective optimization algorithm, with the optimization objectives of minimizing operation waiting time, shortest travel distance, and maximum equipment utilization, comprehensively considering task execution order, vehicle path planning, and resource allocation constraints. Parallel optimization is performed using a Python + Spark algorithm engine, outputting task allocation instructions and optimal path planning results for each truck, achieving real-time task scheduling and global coordination for terminal operations. In traditional truck scheduling modes, a "first-come, first-served" or "first-served" allocation principle is typically used. The system only performs one-way matching based on task arrival order or vehicle idle status—that is, tasks find available vehicles or vehicles find tasks to be executed. This approach ignores the global matching relationship between tasks and vehicles, easily leading to problems such as vehicle waiting, uneven resource utilization, and path conflicts. This module introduces a "two-way matching" mechanism between tasks and vehicles, comprehensively considering task urgency, vehicle location, road congestion, and resource pool constraints. Optimal matching is achieved through a multi-objective optimization algorithm, making the scheduling process more adaptive and globally coordinated.

[0135] This can be understood as sending a pre-preparation instruction to the yard crane 30 seconds before the estimated arrival time of the container truck, ensuring that the yard crane enters a ready-to-work state in advance. When the difference between the remaining time of the gantry crane task and the estimated arrival time of the next container truck is detected to be greater than 10 seconds, the system automatically calls upon backup container truck resources to execute the task to prevent gaps in service. This mechanism reduces the gantry crane downtime rate by more than 20%, shortens the average queuing time of container trucks to less than 3 minutes, and increases the container dispatch instruction execution rate to over 95%, significantly improving equipment coordination efficiency.

[0136] Specifically, the efficient operation of container terminals relies on reasonable truck scheduling. To maximize efficiency and reduce waiting time and delays, a dynamic truck scheduling algorithm based on multi-objective optimization is designed, comprehensively considering task priority, equipment status, driver working hours, and congestion factors. The algorithm objectives are as follows:

[0137] Minimize task delays: Ensure tasks are completed in the shortest possible time, reducing truck waiting time and task delays.

[0138] Optimize equipment utilization: maximize the utilization rate of trucks and reduce idle time.

[0139] Improve the re-entry and re-exit rate of container trucks;

[0140] Reduce the average turnaround time and turnaround distance of container trucks;

[0141] Balance driver workload: allocate driver tasks reasonably to avoid excessive fatigue.

[0142] The overall steps of the scheduling algorithm are as follows:

[0143] (1) Data preprocessing: Obtain and organize task, truck, driver and external environment data from the database. Clean the data and handle missing and outlier values. For specific business situations, lock and assign tasks.

[0144] (2) Confirm task priority ranking: Rank tasks according to their priority and expected completion time. High-priority and urgent tasks should be ranked first. The container dispatch situation in the yard and the user situation should also be considered.

[0145] (3) Task allocation strategy: Based on the current location and status of the trucks, and considering both the planned task time and the truck's travel cost, optimize and adjust the truck's travel distance and shortest execution time to adjust the task allocation. Consider the type of truck, the different operating processes of loading and unloading and gantry cranes (double-box crane, double spreader).

[0146] (4) Dynamic scheduling and adjustment: Real-time monitoring of task execution and truck status. Dynamically adjust the scheduling plan according to actual conditions (such as task delays or equipment failures).

[0147] (5) Feedback mechanism: Collect data on scheduling execution, analyze and evaluate it. Continuously optimize the scheduling algorithm to improve scheduling efficiency.

[0148] The task order optimization and recommendation module receives the preliminary results from the scheduling algorithm module, combines the yard congestion status and crane rhythm deviation information, performs local rearrangement and order optimization of task execution order, generates the final task execution sequence, and outputs it to the algorithm parameter configuration module and result interface.

[0149] This can be understood as a mechanism used to reorder task execution sequences after the scheduling algorithm outputs an initial plan, taking into account yard congestion, crane operation rhythm, and equipment feedback. It performs logical optimization and dynamic correction of the task execution order, generating an optimized recommended sequence by analyzing dependencies and time conflicts between tasks. In traditional scheduling models, only the task sequence itself is considered, i.e., trucks are allocated according to a preset order, without considering task priority or route priority. This task sequence optimization and recommendation module introduces a task priority and route priority mechanism into intelligent scheduling, prioritizing the allocation of suitable vehicles to special containers or high-priority tasks, and prioritizing trucks for key routes, thereby achieving efficient matching of tasks and vehicles and improving overall scheduling efficiency and resource utilization.

[0150] This can be understood as truck scheduling needing the capability to optimize the sequence of loading and unloading tasks. Based on actual terminal operations, unloading tasks have a high degree of flexibility. Tasks assigned to unloading trucks are often temporarily changed by the crane operator below the quay crane. Only after the crane operator confirms the container is placed on the truck is the truck and the task strictly bound one-to-one. For already dispatched containers, truck scheduling considers congestion in the container area (container area congestion module) and rationally adjusts the loading sequence, assigning later-sequence tasks to trucks first (requiring a module to search for interchangeable containers). Trucks with later-sequence tasks arrive at the quay crane first. By designing a truck scheduling algorithm that optimizes the sequence, task execution efficiency can be further improved, delays and waiting times reduced, thereby enhancing the overall operational efficiency of the terminal.

[0151] The algorithm parameter configuration module takes the task execution sequence and real-time running feedback as inputs and provides users with an interface for strategy input and parameter adjustment. Users set the scheduling mode and target weight through the interface, and the system dynamically adjusts the optimization function parameters and iteration number accordingly. The updated parameters are then synchronized back to the scheduling algorithm module to achieve adaptive strategy adjustment and continuous optimization output.

[0152] Specifically, users can choose between efficiency-first, benefit-first, or balanced modes based on port operation needs, and adjust the optimization objective functions and constraint coefficients of each algorithm module in real time according to the configuration results. The parameter configuration results are synchronized to the algorithm engine in the business layer, forming a closed-loop control mechanism from user input to algorithm execution, enabling adaptive adjustment of scheduling strategies. Algorithm parameters include the following categories:

[0153] Supports the setup of horizontal transport mechanical pools;

[0154] Setting the weight for re-entry and re-exit rates;

[0155] Horizontal transport machinery driving parameters (including driving speed, waiting threshold, path avoidance strategy, etc.);

[0156] Setting the number of iterations for the scheduling heuristic search;

[0157] Setting efficiency parameters for bridge crane yards and water transport machinery;

[0158] The algorithm supports setting target values ​​for operating parameters in single / double box cranes, double spreaders, and simultaneous loading and unloading operations. It provides efficiency-first, benefit-first, and balanced modes, and can output the optimal solution for each mode. The overall implementation should ensure that input parameters are configurable and adjustable, and output results are displayed and evaluable.

[0159] Example 3

[0160] This embodiment, based on embodiment 2, provides an exemplary description of the collection of two types of core data based on the port operating system (TOS): static road network data and dynamic operation data.

[0161] Static road network data describes the spatial structure and operational area distribution within the port area, including the coordinates of wharf roads and yard roads (linked to X / Y coordinates via POINT_ID), lane direction attributes (one-way or two-way), intersection locations (linked to POINT_ST start point and POINT_END end point), and container area distribution information (defined by YAA_AREANO container area number and YAA_AREAX / YAA_AREAY coordinates). This data constitutes the topology of driving paths within the port area, providing a spatial reference for path planning and time matrix construction.

[0162] Dynamic operation data describes the real-time operational status of port area operations, including the real-time location of internal trucks (TRK_LOCATION field of the PS_CONTAINER_TRUCKS table), the real-time operation location of gantry cranes (linked to the operation road coordinates via the CRANE_NO field), and road congestion status (obtained through aggregation analysis of points in the truck GPS trajectory that show a dwell time exceeding 5 minutes and are not in operation). This data reflects the dynamic changes in port area traffic flow and operation flow, and is used to update travel time weights in real time.

[0163] By fusing the two types of data, the system dynamically constructs a travel time matrix between port nodes under the Spark distributed computing framework, enabling real-time prediction and updating of travel time for any origin-destination pair, and providing dynamic traffic time constraints and decision-making basis for the loading and unloading task scheduling algorithm.

[0164] Based on the dynamic time matrix prediction results and the status information of the container truck resource pool, multi-objective optimization scheduling of loading and unloading operations is performed.

[0165] Truck resource pools are divided according to the crane operation routes (identified by the TPL_NO flag in the MCS_TRUCK_POOLS table). Each resource pool is bound to 1-3 operation routes (associated with the ORT_ROUTENO flag in the MCS_OP_ROUTES table), and a minimum number of trucks (ORT_MINNUM) and a maximum number of trucks (ORT_MAXNUM) are configured for each operation route. When the actual number of trucks in the pool is less than the minimum value, the system automatically supplements tasks from idle vehicles; when the number of trucks exceeds the maximum value, the system suspends the allocation of new tasks. For key operation routes, a fixed number of trucks is set to ensure the stability of continuous operation, realizing hierarchical management and dynamic constraints of vehicle resources, and providing feasible boundary conditions for subsequent scheduling algorithms.

[0166] A hybrid heuristic algorithm structure of "tabo search + discrete particle swarm optimization (DPSO)" is adopted. The dynamic time matrix is ​​used as the core input. Combined with the task data to be assigned (task_id, orig_loc start position, dest_loc destination position in the MCS_OPERATE_INSTRUCTIONS table) and truck status data (TRK_MCHSTATUS mechanical status, TRK_FREETIME unloading time), a multi-objective optimization model is established.

[0167] Based on the multi-objective optimization model, three switchable optimization modes are also provided:

[0168] Efficiency-first mode: The objective is to minimize the arrival delay time of the trucks, with a weight of 0.6;

[0169] Efficiency-first mode: The goal is to maximize the re-entry and re-exit rate, with reference to the no-load distance (TDL_NOLOAD_DISTANCE) and a weight of 0.5;

[0170] Balanced mode: Combines three indicators: delay time, re-entry and re-exit rate, and empty driving distance, with weights of 0.4, 0.3, and 0.3, respectively.

[0171] The algorithm runs in parallel in a Spark distributed computing environment. It explores the global path through tabu search and uses the discrete particle swarm optimization algorithm to iteratively optimize the local solution. Finally, it outputs the optimal mapping between each truck and the task (trkno-task_id mapping), the corresponding path planning, and the estimated job time (pst_orig_move, pft_orig_move).

[0172] This module supports real-time scheduling calculations for more than 200 trucks, with an average algorithm response time of less than 3 seconds. It can maintain stable decision-making and fast response even in large-scale parallel operation scenarios, and is the core algorithm engine for the system to achieve intelligent scheduling optimization.

[0173] Example 4

[0174] like Figure 6 As shown in the example, the parts not detailed in this embodiment are as shown in Example 2. This embodiment provides a port truck scheduling optimization method based on dynamic time matrix prediction. Using the port operating system (TOS) and vehicle monitoring data as core inputs, it achieves a real-time closed loop of predictive scheduling feedback and re-optimization through dynamic time matrix prediction, multi-objective scheduling optimization, and task prioritization, thereby improving port operation efficiency and resource utilization. The method includes the following steps:

[0175] S101: Collect work tasks, equipment status and internal container truck running trajectory data from the port operating system and vehicle monitoring terminal; extract terminal road network nodes, lane directions and intersection relationships; establish a node connection table; calculate vehicle trajectory speed, dwell time and work identification; remove outliers and generate a spatiotemporal feature sample set.

[0176] S102: Model the spatiotemporal feature sample set based on the distributed computing engine, using the average travel time of any node pair and its congestion weight as the core parameters, update the predicted values ​​according to the time window, and generate a dynamic time matrix that changes with time.

[0177] S103: Divide the truck resource pool according to the work route, set the minimum and maximum truck thresholds, automatically replenish the trucks when the number of trucks in the truck resource pool is lower than the minimum value, and pause task allocation when the number of trucks exceeds the maximum value, and output the resource pool configuration status.

[0178] S104: Using the dynamic time matrix and resource pool status as input, a hybrid algorithm of tabu search and discrete particle swarm optimization is used to perform multi-objective optimization in efficiency-first, benefit-first, or balance mode to generate task allocation and path planning results.

[0179] S105: Based on the yard congestion status and the gantry crane operation rhythm, optimize the task execution order of the scheduling results and output the adjusted execution sequence;

[0180] S106: Based on the weights and policy parameters input by the user, adjust the algorithm's objective function and number of iterations to achieve adaptive switching and continuous optimization of the scheduling strategy, and build an intelligent scheduling closed loop of time prediction, resource constraints and task optimization.

[0181] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A port truck scheduling optimization system based on dynamic time matrix prediction, characterized in that, This includes a dynamic time matrix prediction module, a dynamic container truck resource pool configuration module, a loading and unloading task scheduling algorithm module, a task priority optimization and recommendation module, and an algorithm parameter configuration module. The dynamic time matrix prediction module receives historical and real-time data from the port operating system and vehicle monitoring terminals, extracts road network node, vehicle trajectory, and crane operation status information, obtains road network node coordinates, edge connections, and travel directions, and generates a road topology table; obtains crane operation cycles, operation periods, and spatial locations, and establishes a rhythm parameter table for cranes and adjacent nodes; calculates the road width, turning angle, and average dwell time for each travel edge, forming an impedance parameter table; aligns the rhythm parameter table and impedance parameter table according to the road topology table, performs field alignment and weighted synthesis, calculates the travel time between any nodes, outputs a travel time matrix containing crane rhythm coefficients and road impedance factors, and periodically updates the travel time between any nodes; The dynamic truck resource pool configuration module is used to receive the time matrix and task data, divide the truck resource pool according to the operation route, construct a task density function based on the task arrival rate and adaptively adjust the minimum and maximum truck thresholds, and output the dynamic resource pool status. The loading and unloading ship task scheduling algorithm module takes the time matrix and resource pool status as input, establishes a two-way matching relationship between tasks and vehicles, and uses a hybrid heuristic algorithm to establish a multi-objective model. The optimization objectives are minimum operation waiting time, shortest travel distance and maximum equipment utilization. Taking into account task execution order, vehicle path planning and resource allocation constraints, the module outputs preliminary scheduling results and task allocation table. The task order optimization and recommendation module combines the yard congestion status and crane rhythm deviation information, and takes into account the task priority and work route priority mechanism to perform local rearrangement and timing correction of the task execution order in the preliminary scheduling results. The algorithm parameter configuration module dynamically adjusts and optimizes the function parameters based on runtime feedback and user settings. Users set the scheduling mode and target weight through the interface, and the optimization function parameters and iteration count are dynamically adjusted accordingly. The updated parameters are then synchronized back to the scheduling algorithm module to achieve closed-loop collaboration between prediction, scheduling, and feedback.

2. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 1, characterized in that, The periodic updates of the driving time matrix include: Read the predicted driving time and the collected actual driving time from the previous cycle and calculate the difference to generate a difference record table; Based on the difference record table, the bridge crane rhythm coefficient and road impedance factor are proportionally corrected to generate a correction parameter table; Recalculate the travel time between nodes according to the corrected parameter table to generate a new travel time matrix file; Write the new driving time matrix file to the database to replace the old version and output the revised driving time matrix for use by the resource pool configuration module.

3. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 1, characterized in that, The dynamic card collection resource pool configuration module includes: Read the task arrival time series data and count the number of tasks in each time period to generate a task distribution table; Calculate the task arrival rate function based on the task distribution table to form a task density value table; Based on the task density value, the minimum and maximum number of container trucks for each work route are calculated to generate a threshold table; The threshold table is combined with the existing vehicle list to form a resource pool status table, and the resource pool status information is output for use by the task scheduling algorithm module.

4. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 3, characterized in that, The process of adjusting the resource pool status based on the resource pool status table includes: Read the task density value table and threshold table to generate an adjustment instruction table for each work route; According to the adjustment instruction table, the number of trucks for the corresponding route is increased or decreased to form an adjustment result table; Compare the resource pool status table with the adjustment result table and verify the vehicle load rate and standby time for each route; The updated resource pool status table is output and sent to the task scheduling algorithm module for task matching in the next cycle.

5. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 1, characterized in that, The task and vehicle two-way matching process includes: Read the dynamic time matrix and resource pool status table to obtain the set of allocable tasks and the set of available vehicles; Calculate the task-vehicle associated cost matrix based on task urgency, vehicle location, and travel time; Perform a hybrid optimization of tabu search and discrete particle swarm optimization on the cost matrix to obtain a globally feasible matching solution set; Select the correspondence with the minimum cost in the matching solution set and generate a task allocation instruction table to be output to the scheduling and execution module.

6. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 5, characterized in that, The optimization process for the two-way matching of tasks and vehicles includes: Tasks are processed in a priority hierarchy, and vehicles are grouped according to their idle time. In each level, the matching weight is calculated based on the path length and resource pool constraints; The weight parameters are adjusted iteratively by particle swarm optimization to converge to the optimal allocation solution; Output the iteratively optimized task-vehicle matching table and send it to the task execution scheduling unit.

7. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 1, characterized in that, The task priority optimization and recommendation process includes: Receive preliminary scheduling results and yard congestion data to construct a task execution timetable; Calculate the difference between the bridge crane rhythm deviation information and the task waiting time to identify tasks that need to have their execution order adjusted. Adjust the order of task execution based on the recognition results and generate a revised sequence list; The output task execution sequence, after local rearrangement and order correction, is provided for use by the algorithm parameter configuration module.

8. The port truck scheduling optimization system based on dynamic time matrix prediction according to claim 7, characterized in that, The task execution sequence correction includes: Read the corrected sequence list and the crane rhythm data to establish a task time deviation model; The mission start time and vehicle arrival time are synchronized and adjusted according to the deviation model. Generate a synchronized and corrected task schedule; The corrected task schedule is output for the algorithm parameter configuration module to update the optimization function parameters.

9. A port truck scheduling optimization method based on dynamic time matrix prediction, used to execute the port truck scheduling optimization system based on dynamic time matrix prediction as described in any one of claims 1-8, characterized in that, Includes the following steps; Historical and real-time operational data are collected from the port operating system and vehicle monitoring terminals to generate a basic dataset containing road network nodes, vehicle trajectories, and crane status. A travel time matrix incorporating the bridge crane rhythm coefficient and road impedance factor is constructed using the basic dataset, and periodic updates are performed to obtain the inter-node travel time prediction results; specifically: Obtain the coordinates of road network nodes, edge connections, and travel directions to generate a road topology table; obtain the operation cycle, operation time period, and spatial location of the bridge crane to establish a rhythm parameter table for the bridge crane and adjacent nodes; calculate the road width, turning angle, and average dwell time of each travel edge to form an impedance parameter table; align the fields of the rhythm parameter table and the impedance parameter table according to the road topology table and perform weighted synthesis to calculate the travel time between any nodes, and output a travel time matrix containing the bridge crane rhythm coefficient and road impedance factor; Based on the prediction results and task arrival rate, establish a resource pool status table and calculate the minimum and maximum truck thresholds for each operation route; Under resource pool constraints, a hybrid algorithm combining tabu search and discrete particle swarm optimization is used to establish a two-way matching relationship between tasks and vehicles and generate a task allocation instruction table. By combining the information on yard congestion and crane rhythm deviation, the order of task execution is rearranged and the timing is corrected based on the preliminary scheduling results to form a task execution sequence. The task execution sequence and running feedback are input into the algorithm parameter configuration module to update and optimize function parameters and output an optimized scheduling scheme.

Images