Intelligent forklift dynamic scheduling system and method based on multi-source data fusion

Abstract

The application discloses a kind of intelligent fork truck dynamic scheduling system and method based on multi-source data fusion, collect ground scale weighing, temporary storage container in situ detection, warehouse position occupation, detection platform use and work position calling signal and other multi-source data, adopt attention mechanism network weighted fusion, build environment and task comprehensive model.Based on fusion data generation detection, warehouse entry, warehouse exit and other scheduling tasks, make clear that detection scheduling is prior to handling scheduling, realize task priority allocation, AGV dynamic scheduling, optimal path planning and conflict avoidance;Through AGV state feedback adjustment task strategy, full-process key information is projected to monitoring large screen to realize visual management;System includes core inspection system, downstream AGV scheduling system and communication and integration module, each unit cooperates to complete data acquisition, fusion processing and other functions.The application improves the intelligentization and automation level of scheduling, reduces manual intervention, guarantees logistics operation efficient and accurate, and is suitable for heavy industry, machining and other large-scale production scenes.

Description

Technical Field

[0001] This invention belongs to the field of intelligent logistics and automated control technology, specifically relating to an intelligent forklift dynamic scheduling system and method based on multi-source data fusion. Background Technology

[0002] With the deepening development of Industry 4.0 and intelligent manufacturing technologies, especially in the current technological context of 2026, intelligent logistics systems have become a key link for modern manufacturing enterprises to enhance their core competitiveness. In densely populated manufacturing areas, heavy industry, machining, and large-scale production scenarios are placing increasingly higher demands on logistics automation. Traditional "human + forklift" or simple automated production lines are no longer sufficient to meet the needs of high-precision and high-flexibility production.

[0003] However, existing intelligent logistics systems still have significant limitations in practical applications, mainly in the following aspects:

[0004] System hierarchical fragmentation leads to low collaboration efficiency:

[0005] Existing logistics systems typically operate "inspection / monitoring" and "AGV scheduling" as two separate subsystems. The inspection system is responsible for data collection, while the AGV scheduling system is responsible for vehicle operation, with a lack of unified data flow and task flow interfaces between the two. This unclear hierarchical architecture leads to information lag, preventing the inspection system from sensing the AGV status in real time, and hindering the AGV scheduling system from obtaining detailed quality data from the production site, resulting in low overall collaborative efficiency.

[0006] Insufficient data utilization and low level of intelligent decision-making:

[0007] While existing inspection systems can collect multi-dimensional data (such as weighing data from scales, in-situ detection data from spherical monitors, and 3D vision inspection data), this data is often only used for archiving or simple alarms, failing to be fully integrated and used for downstream AGV scheduling decisions. For example, whether a visual inspection is qualified is usually a prerequisite for warehousing, but existing systems often cannot automatically convert this judgment result into AGV handling instructions, requiring manual intervention and increasing the complexity and error rate of the work process.

[0008] Lack of dynamic adaptability:

[0009] In complex production environments (such as mixed scenarios involving online production, warehousing, and outbound of semi-finished products), existing scheduling systems often employ static or pre-defined path planning, which struggles to adapt to dynamic changes resulting from the fusion of multi-source data. For instance, when inspection tasks conflict with handling tasks, the lack of a dynamic avoidance mechanism based on task priority (e.g., inspection takes precedence over handling) can easily lead to logistical congestion or task backlog.

[0010] Lagging level of visual management:

[0011] Existing monitoring methods typically only display the status of individual devices, failing to deeply integrate and project in real-time the results of inspections, the occupancy of temporary storage areas, the usage status of weighbridges and inspection platforms, and the operating trajectories of forklifts and AGVs. This makes it difficult for management to grasp the overall status of production logistics in real time, hindering their ability to make accurate production adjustment decisions based on real-time data.

[0012] Therefore, how to establish an efficient collaborative mechanism between the inspection system and the AGV scheduling system, and realize the effective integration of multi-source data, dynamic intelligent scheduling of tasks, and visualized management of the entire process, is a key technical problem that urgently needs to be solved in the field of intelligent logistics technology. Summary of the Invention

[0013] The purpose of this invention is to provide an intelligent forklift dynamic scheduling system and method based on multi-source data fusion, in order to solve the problems mentioned in the background art.

[0014] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic scheduling of intelligent forklifts based on multi-source data fusion, comprising the following steps:

[0015] S1. Collect weighing data from the weighbridge in the processing area, in-situ detection data of temporary storage containers based on spherical monitoring, occupancy data of temporary storage bins, and usage data of the testing platform;

[0016] S2. Collect the call signals from the control panels of each workstation;

[0017] S3. Use algorithms to fuse multi-source data, construct a comprehensive model of the environment and tasks, and determine whether the current composite container meets the entry conditions.

[0018] S4. Based on the fusion of data, intelligently process the handling and detection scheduling tasks, and determine the task priority and execution requirements;

[0019] S5. The inspection system sends inspection tasks to the AGV scheduling system.

[0020] S6. The AGV scheduling system receives detection tasks and schedules composite AGVs to execute the detection and scheduling tasks.

[0021] S7. The composite AGV moves to the inspection platform to collect visual inspection data of the composite robot;

[0022] S8. Determine whether to issue an inbound handling task based on the visual inspection results. At the same time, determine the target location of the handling task based on the on-site inspection results and optimize the inbound task scheduling by referring to the temporary storage warehouse space occupancy data.

[0023] S9. The inspection system sends the warehousing and handling task to the AGV scheduling system.

[0024] S10: The AGV scheduling system receives handling tasks and simultaneously collects the status data of the forklift AGV itself.

[0025] S11. Intelligently allocate inbound and handling tasks based on task requirements and forklift AGV status data;

[0026] S12. Use algorithms to plan the optimal path in real time;

[0027] S13. Conflict avoidance is achieved based on the type of scheduled task;

[0028] S14. Control the forklift AGV to perform inbound and transport tasks, and collect the execution status and running results in real time;

[0029] S15 The AGV scheduling system feeds back the execution status and operation results to the inspection system in real time. The inspection system adjusts the task generation strategy based on the feedback data, optimizes subsequent scheduling, and automatically projects the inspection results, temporary storage space occupancy, weighbridge usage, inspection platform usage, and the operation status and results of forklift AGVs and composite AGVs onto the monitoring screen to achieve real-time visual management.

[0030] Preferably, in step S3, a multi-source data weighted fusion judgment mechanism based on task type priority is used to fuse multi-source data, and different data source attention weights are set for different task types. The weighted fusion judgment mechanism includes: inputting the raw data collected by each subsystem into the fusion judgment module after normalization encoding, and calculating the comprehensive attention score according to the formula Score = w1 ×f1 + w2 × f2 + w3 × f3 + w4 × f4, where f1 is the normalized value of visual detection result, f2 is the normalized value of inventory status, f3 is the normalized value of call box signal, f4 is the normalized value of electronic tag matching result, w1 to w4 are weight coefficients and w1+w2+w3+w4=1; when Score≥0.7, it is determined that the entry condition is met; otherwise, it is determined that the condition is not met and corresponding prompt information is fed back.

[0031] Preferably, in step S8, the visual inspection result of the composite robot serves as the key criterion for determining whether to issue an inbound handling task. If the visual inspection is qualified, the inbound handling task is issued; if it is not qualified, it is not issued.

[0032] Preferably, in step S11, transportation tasks are intelligently allocated based on task priority, forklift AGV status and location, taking into account task time requirements, cargo weight and handling distance factors, and prioritizing the execution of high-priority tasks.

[0033] Preferably, in step S13, detection scheduling takes precedence over material handling scheduling to ensure the timely execution of detection tasks and avoid work conflicts between forklift AGVs and composite AGVs through task timing arrangement.

[0034] Preferably, the method further includes an online production process for semi-finished products. The intelligent logistics inspection system combines data from the logistics monitoring system, 3D vision inspection system, and electronic tag system to make a comprehensive judgment and then sends an inspection task request to the scheduling system. The comprehensive judgment adopts a multi-condition sequential verification and veto decision-making mechanism, including nine steps: input parameter verification, group validity verification, starting container positioning, material qualification judgment, target docking location search, target storage location idle judgment, starting resource readiness judgment, vehicle binding consistency judgment, and task creation and issuance. Among them, conditions C1 (material qualification condition), C2 (target storage location idle condition), and C3 (starting resource readiness condition) are necessary conditions. If any one of them is not met, the judgment will terminate immediately. If condition C4 (vehicle binding consistency condition) is not met, automatic repair will be attempted first.

[0035] Preferably, the method further includes a semi-finished product warehousing and handling process. The intelligent logistics inspection system combines data from the logistics monitoring system, 3D vision inspection system, electronic tag system, and in-situ detection system to make a comprehensive judgment and then sends a warehousing and handling task request to the scheduling system. The intelligent logistics inspection system combines data from each system to perform comprehensive verification of the semi-finished products, including inspection and testing of their qualification and weighing to verify weight information.

[0036] Preferably, the method further includes a semi-finished product outbound handling process. The intelligent logistics inspection system combines data from the logistics monitoring system, 3D vision inspection system, electronic tag system, and in-situ detection system to make a comprehensive judgment and then sends an outbound handling task request to the scheduling system.

[0037] Preferably, the system includes a core inspection system, a downstream AGV scheduling system, and a communication and integration module. The core inspection system includes a multi-source data acquisition unit, a data fusion processing unit, a task generation and distribution unit, a monitoring screen projection unit, and a monitoring management unit. The downstream AGV scheduling system includes an AGV status acquisition unit, a task receiving and parsing unit, a dynamic scheduling decision unit, an execution control unit, and a conflict avoidance unit.

[0038] The multi-source data acquisition unit is used to collect weighing data from the weighbridge in the processing area, in-situ detection data of the temporary storage container based on spherical monitoring, occupancy data of the temporary storage bin, usage data of the testing station, and call signals from the workstation control panel.

[0039] The data fusion processing unit is used to fuse multi-source data using algorithms and construct a comprehensive model of the environment and tasks;

[0040] The task generation and distribution unit is used to intelligently generate transportation tasks based on fused data and distribute them to the AGV scheduling system.

[0041] The dynamic scheduling decision unit is used to realize the dynamic scheduling and path planning of AGV based on real-time data and task requirements. Specifically, it adopts a path selection method based on predefined route topology, including six steps: route query, route uniqueness verification, target area determination, target area connection position query, idle connection position filtering, and optimal target selection. Dynamic obstacles are handled through a three-layer mechanism of target occupancy pre-inspection, initial resource verification, and underlying collaborative obstacle avoidance.

[0042] The conflict avoidance unit is used to avoid conflicts by adopting a conflict prevention strategy based on task status verification and resource status verification according to the type of scheduling task. This includes the detection and resolution of three types of scenarios: resource occupation conflict, vehicle binding conflict, and task legality conflict.

[0043] The monitoring screen projection unit is used to automatically project the test results, temporary storage space occupancy, weighbridge usage, testing platform usage, and the operating status and results of forklift AGVs and composite AGVs onto the monitoring screen.

[0044] Preferably, the multi-source data acquisition unit includes a weighbridge sensor, a spherical monitoring camera, a warehouse occupancy detection sensor, a detection platform sensor, and a control panel; electronic tags are installed on the outside of the logistics container for manual visual display, and the outbound quantity in the electronic tags is updated as the goods are shipped out.

[0045] Preferably, the AGV status acquisition unit includes a position sensor, a speed sensor, a power sensor, and a load sensor.

[0046] Preferably, the system is used to implement the control method according to any one of claims 1-8; wherein, the attention mechanism network is a multi-source data weighted fusion judgment mechanism based on task type priority, the optimal path planning adopts a path selection method based on predefined route topology and combined with real-time status verification to achieve dynamic obstacle avoidance, the conflict avoidance adopts a conflict prevention strategy based on task status verification and resource status verification, and the comprehensive judgment adopts a decision mechanism of multi-condition sequential verification and veto.

[0047] The technical effects and advantages of this invention are as follows:

[0048] 1. To achieve deep integration of inspection and scheduling and break down system hierarchical barriers, this invention constructs a direct interaction mechanism between the "core inspection system" and the "downstream AGV scheduling system". Through a unified data flow and task flow interface, it realizes closed-loop control from data acquisition, fusion, task generation to execution feedback, eliminating information silos.

[0049] 2. Intelligent decision-making based on multi-source data fusion improves automation and accuracy. By using an attention mechanism network to weightedly fuse multi-source data such as scales, spherical monitoring, and visual inspection, a comprehensive model is constructed, which realizes automatic judgment without the need for secondary manual confirmation, thus reducing the error rate.

[0050] 3. Clearly define "detection scheduling takes precedence over material handling scheduling" to ensure timely execution of critical tasks. Intelligent allocation is carried out by comprehensively considering the AGV's own status and task attributes, and the optimal path is planned in real time through algorithms to avoid conflicts.

[0051] 4. Full-process visualization and self-optimization enhance management transparency: The monitoring screen projection unit projects all key information such as detection results, warehouse occupancy, and equipment status in real time; the AGV scheduling system feeds back the execution results to the inspection system, which adjusts the task generation strategy accordingly, forming a self-optimizing intelligent cycle. Attached Figure Description

[0052] Figure 1 This is a flowchart of the online production process for semi-finished products according to the present invention;

[0053] Figure 2 This is a flowchart of the semi-finished product warehousing and handling process of the present invention;

[0054] Figure 3 This is a flowchart of the semi-finished product outbound handling process of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Specific Implementation Method 1: Online Production Process of Semi-finished Products (e.g.) Figure 1 (As shown)

[0057] This implementation method combines Figure 1 This document describes the specific implementation process of online production of semi-finished products, involving the collaborative work of intelligent logistics inspection system, scheduling system, composite AGV (with visual inspection function) and forklift AGV.

[0058] S1. Model Selection and Parameter Distribution: Operators select the model of the semi-finished product to be produced through the intelligent logistics inspection system, and the system retrieves the corresponding 3D vision inspection standard and production parameters for that model.

[0059] S2. Calling Inspection Task: The intelligent logistics inspection system combines data from the logistics monitoring system (weighbridge data), the 3D vision inspection system (camera status), and the electronic tag system to make a comprehensive judgment and then sends a high-priority inspection task request to the scheduling system, reflecting the principle of "inspection scheduling takes precedence over handling scheduling".

[0060] The comprehensive judgment adopts a multi-condition sequential verification and veto decision-making mechanism, which includes nine steps: input parameter verification, group validity verification, initial container positioning, material qualification judgment (condition C1), target docking location search, target storage location idle judgment (condition C2), initial resource readiness judgment (condition C3), vehicle binding consistency judgment (condition C4), and task creation and issuance. Among them, conditions C1, C2, and C3 are necessary conditions. If any one of them is not met, the judgment will be terminated immediately and the reason for failure will be returned. If the vehicle binding consistency condition (condition C4) is not met, automatic repair will be attempted first. If the repair is successful, it is considered as passing.

[0061] S3. Task Creation and Allocation: The intelligent logistics inspection system creates inspection tasks and sends them to the downstream AGV scheduling system. The AGV scheduling system dispatches the composite AGV to the inspection station based on the current position and status of the composite AGV.

[0062] S4. Robot in Position: The composite AGV moves to the inspection station and confirms its position using laser or vision positioning, ready to perform the inspection task. At this time, the semi-finished product has been manually or transported to the inspection station by the production line.

[0063] S5. Inspection: The 3D vision system on the composite AGV performs an all-round scan of the semi-finished product, collects visual inspection data, and measures appearance defects and dimensions.

[0064] S6. Information Update and Flow: Based on the inspection results, the system automatically updates the electronic tag information (including ID, model, and result) and container status. If the inspection is qualified, the intelligent logistics inspection system determines that the warehousing conditions are met and generates an warehousing handling task request, scheduling a forklift AGV to perform subsequent handling; if it is unqualified, an alarm is triggered and no handling task is issued.

[0065] Specific Implementation Method Two: Semi-finished Product Warehousing and Handling Process (e.g.) Figure 2 (As shown)

[0066] This implementation method combines Figure 2 Explain the specific implementation process for handling semi-finished products upon warehousing.

[0067] S1. Model Selection: Operators select the model of the semi-finished product to be put into the warehouse in the intelligent logistics inspection system.

[0068] S2, Call for Handling: The system combines data from logistics monitoring (weighbridge), 3D vision (quality), electronic tags (identity), and on-site detection (location) to make a comprehensive judgment and send an inbound handling task request to the dispatch system.

[0069] S3. Multi-source data comprehensive verification: Before the task is issued, the data fusion processing unit performs weighted fusion of multi-source data, verifies the 3D vision detection results, weighing data of the weighbridge and the status of the weighbridge. If the data is inconsistent (such as the weight is inconsistent or the vision is unqualified), the task is intercepted and an alarm is triggered.

[0070] In step S3, the specific process of multi-source data comprehensive verification is as follows: the data fusion processing unit performs weighted fusion of multi-source data and calculates the comprehensive score according to Score = w1×f1 + w2×f2 + w3×f3 + w4×f4, where w1=0.40 (visual detection result), w2=0.30 (weighing data of the weighbridge), w3=0.20 (on-site status), and w4=0.10 (electronic tag matching); when the Score≥0.7 and all necessary conditions are met, the verification is passed; otherwise, the task is intercepted and an alarm is triggered.

[0071] S4. Warehouse Space Verification: The system queries the temporary storage warehouse space occupancy data and, combined with the attributes of the incoming semi-finished products (such as size and weight), intelligently allocates the optimal target warehouse space.

[0072] S5. Task Issuance: Create an inbound transport AGV task and issue it to the AGV scheduling system.

[0073] S6. Task Execution: The forklift AGV receives the task and proceeds to the pickup point. The forklift AGV collects its location and load status in real time, and uses a path selection method based on a predefined route topology to plan the optimal path in real time, combined with real-time status verification to achieve dynamic obstacle avoidance. Specifically, the system uses a two-level topology structure of "region-connection point" to establish a warehouse map model. The warehouse space is divided into several functional areas (such as workstations, temporary storage areas, and inspection areas), each containing several connection points (i.e., AGV docking points). Areas are connected by predefined routes to form a directed topology graph. Path selection consists of six steps: route query, route uniqueness verification, target area determination, and path selection. The system queries target area docking stations, filters available docking stations, and selects the optimal target. It handles dynamic obstacles through a three-layer mechanism: the first layer is target occupancy pre-check, verifying the real-time occupancy status of the target docking station before task issuance; the second layer is initial resource verification, checking whether the initial docking station is already bound to the container to be transported; the third layer is low-level collaborative obstacle avoidance, issuing the transport task to an external scheduling system (such as Hikvision RCS system), which is responsible for the AGV's low-level path planning and dynamic obstacle avoidance during actual operation (such as LiDAR obstacle detection, speed adjustment, and real-time path adjustment). After path selection, the system performs container position consistency verification and vehicle binding status verification.

[0074] S7. Task Completion Feedback: The forklift AGV moves the semi-finished product to the target warehouse and reports the task completion status.

[0075] S8. Inventory Synchronization Update: The system confirms the return of empty containers and the entry of fully loaded containers into the warehouse, synchronously updates the inventory data and electronic tag status, and projects the entry results and warehouse occupancy status onto the monitoring screen.

[0076] Specific Implementation Method Three: Semi-finished Product Outbound Handling Process (e.g.) Figure 3 (As shown)

[0077] This implementation method combines Figure 3 Explain the specific implementation process for handling semi-finished products out of the warehouse.

[0078] S1. Outbound Request: Operators select the model and quantity of semi-finished products to be outbound in the intelligent logistics inspection system.

[0079] S2, Outbound Call: The system combines inventory data, electronic tag information, and on-site detection data to send an outbound handling task request to the scheduling system.

[0080] S3. Warehouse Location: The system queries the temporary storage warehouse, locates the warehouse where the target semi-finished product is located, and confirms the feasibility of outbound shipment.

[0081] S4. Task Creation: Create an outbound transport AGV task and send it to the AGV scheduling system.

[0082] S5. Task Execution: Dispatch forklift AGVs to the target warehouse, pick up semi-finished products and transport them to the outbound point (such as the welding area).

[0083] S6. Task Complete: The forklift AGV completes unloading and reports the task completion status.

[0084] S7. Closed-loop management and updates: The system confirms that the fully loaded container has been dispatched and tracks the return location of the empty container (ensuring that the empty container is in place, forming a closed-loop logistics system); updates inventory data (remaining inventory, dispatch time) and projects the updated information onto the screen.

[0085] Specifically, the electronic tags installed on the outside of the logistics containers are used to display the quantity and status of the containers that have been shipped out, allowing on-site personnel to manually and visually verify them, thus achieving human-machine collaborative management.

[0086] System Deployment and Implementation

[0087] To achieve intelligent scheduling of the entire process of online production, warehousing, and outbound of the aforementioned semi-finished products, this invention deploys an intelligent forklift dynamic scheduling system based on multi-source data fusion. The system's physical architecture consists of a core inspection system, a downstream AGV scheduling system, a communication and integration module, and a field execution and perception layer.

[0088] 1. Core Inspection System (Decision Brain): Deployed on the server side, this system is responsible for data fusion, task generation, and global monitoring, and specifically includes the following functional units:

[0089] Multi-source data acquisition unit: Collects real-time data from the industrial gateway, including weighing data from the weighbridge in the processing area, in-situ detection data of temporary storage containers based on spherical monitoring, data from the temporary storage bin occupancy detection sensor, usage status data of the testing station, and call signals from the control panels of each workstation.

[0090] Data fusion processing unit: This unit employs a multi-source data weighted fusion judgment mechanism based on task type priority to perform weighted fusion of the aforementioned multi-source heterogeneous data, constructing an environment state model and a task state model. This unit is responsible for executing the algorithm in step 3, focusing on data features related to the current production task (such as online production, warehousing, and outbound). Specifically, the raw data collected by each subsystem is normalized and encoded before being input into the fusion judgment module: visual detection results are encoded as qualified (1) / unqualified (0), inventory status is encoded as idle (1) / occupied (0), call box signal is encoded as triggered (1) / not triggered (0), and electronic tag data is encoded as matched (1) / not matched (0). The system calculates the comprehensive attention score according to the formula Score = w1×f1 + w2×f2 + w3×f3 + w4×f4, where w1 to w4 are weight coefficients and satisfy w1+w2+w3+w4=1. The comprehensive score threshold θ=0.7 is set. When Score≥θ, the task triggering condition is determined to be met. When Score<θ, the condition is determined not to be met and the corresponding prompt information is fed back.

[0091] Task generation and distribution unit: Based on the fused data model, it intelligently determines whether the composite container meets the warehousing conditions (such as weight verification and visual inspection qualification), and automatically generates inspection tasks or handling tasks according to the business process (S1-S15) and distributes them to the downstream AGV scheduling system.

[0092] Monitoring screen projection unit: Automatically projects real-time test results (qualified / unqualified), temporary storage warehouse occupancy, weighbridge and testing platform usage status, as well as the running trajectory and status of forklift AGVs and composite AGVs onto the monitoring screen, realizing full-process visual management.

[0093] 2. Downstream AGV scheduling system (execution and command)

[0094] This system is responsible for receiving instructions from the core inspection system and controlling the field robots to execute them. It includes the following units:

[0095] AGV Status Acquisition Unit: Real-time acquisition of data from position sensors, speed sensors, power sensors, and load sensors of forklift AGVs and composite AGVs to ensure that scheduling decisions are based on the actual status of the vehicles.

[0096] Task receiving and parsing unit: Receives task instructions (such as inspection tasks, inbound handling tasks, and outbound handling tasks) from the inspection system and parses task attributes (such as priority and target location).

[0097] Dynamic Scheduling Decision Unit: Based on real-time data and task requirements, this unit enables dynamic scheduling and path planning for AGVs. Specifically, it employs a path selection method based on predefined route topology. The system establishes a warehouse map model using a two-level topology structure of "region-connection point." The warehouse space is divided into several functional areas, each containing several connection points. These areas are connected by predefined routes to form a directed topology graph. Path selection involves six steps: route query, route uniqueness verification, target area determination, target area connection point query, idle connection point selection, and optimal target selection. The system handles dynamic obstacles through a three-layer mechanism: the first layer is target occupancy pre-check, verifying the target before task issuance. The system performs three layers of checks: 1) Real-time occupancy status of the docking station; 2) Initial resource verification, checking whether the initial docking station is bound to the container to be transported; 3) Low-level collaborative obstacle avoidance, distributing the transport task to an external scheduling system (such as Hikvision RCS system), which is responsible for the AGV's low-level path planning and dynamic obstacle avoidance during actual operation (such as LiDAR obstacle detection, speed adjustment, real-time path adjustment, etc.); This system and the external scheduling system form a division of labor: This system is responsible for task decision-making and target allocation at the business level, while the external scheduling system is responsible for path execution and real-time obstacle avoidance at the equipment level; After path selection is completed, the system performs container position consistency verification and vehicle binding status verification.

[0098] The conflict avoidance unit is used to avoid conflicts based on the scheduling task type and employs a conflict prevention strategy based on task status verification and resource status verification. The conflict avoidance includes defining three conflict scenarios: resource occupancy conflict (the target docking station is already occupied by another container), vehicle binding conflict (the binding relationship between the container and the docking station is inconsistent between the system's database and the external scheduling system), and task legitimacy conflict (the business preconditions of the task to be executed are not met). During the task creation phase, three checks are performed sequentially: target storage location occupancy check, material qualification check, and initial resource readiness check. During the task issuance phase, two additional checks are performed: vehicle binding consistency check and automatic binding conflict repair. When a vehicle binding conflict occurs, the system automatically performs an unbinding and rebinding operation. The system sets a priority attribute for each task; when multiple tasks are waiting to be issued simultaneously, they are processed sequentially according to their creation time, with the task created earlier receiving priority in obtaining execution resources.

[0099] Execution control unit: Sends specific motion control commands to forklift AGVs and composite AGVs.

[0100] 3. On-site hardware deployment and connection (physical layer)

[0101] To support the above specific implementation methods, the field hardware is deployed and connected in the following manner:

[0102] Perception layer deployment:

[0103] Floor scale sensors and 3D vision inspection cameras are deployed in the processing area to collect weight and quality data of semi-finished products.

[0104] Spherical surveillance cameras and occupancy detection sensors are deployed in the temporary storage and testing areas to monitor the container's position and occupancy status in real time.

[0105] Electronic labels are installed on the outside of logistics containers to display container ID, model, and quantity information.

[0106] Execution layer deployment:

[0107] Deploy a composite AGV (composite robot) equipped with a vision acquisition module to perform inspection tasks at the inspection station (Implementation Method 1).

[0108] Deploy forklift AGVs equipped with LiDAR, cameras, and load detection devices to perform the tasks of warehousing (implementation 2) and warehousing (implementation 3) of semi-finished products.

[0109] Network connection:

[0110] All sensors (weighbridge, camera, and detection platform sensors) are connected to the multi-source data acquisition unit via industrial Ethernet or I / O modules.

[0111] A TCP / IP communication connection is established between the core inspection system and the downstream AGV scheduling system to achieve real-time interaction between task flow and status flow.

[0112] The AGV scheduling system communicates in real time with forklift AGVs and composite AGVs via wireless APs, issuing path instructions and receiving vehicle status.

[0113] 4. System Operating Logic

[0114] In this embodiment, the system achieves a closed loop of "perception-decision-execution" through the above deployment:

[0115] Input: Operators select the model through the workstation control panel (triggering S1), or the system automatically triggers the task according to the production cycle.

[0116] Processing: The core inspection system integrates electronic tags, weighbridges, and visual data to determine the status of materials.

[0117] Scheduling: For inspection tasks, schedule composite AGVs; for material handling tasks, schedule forklift AGVs. The system uses the conflict avoidance mechanism in S13 to ensure the safety and efficiency of both types of vehicles when operating in combination.

[0118] Feedback: After the AGV completes its operation, its status is reported to the inspection system, the inventory database is updated, and the monitoring screen is refreshed simultaneously.

[0119] When a task fails to be sent to the database, the system automatically clears the created task record to ensure data consistency. Specifically, if a task has been successfully written to the database but fails to be sent to an external system, the system automatically deletes the task record to avoid creating "created but not executed" pending tasks. Simultaneously, the system uses a transaction management mechanism to ensure the atomicity of multi-step operations—if any step fails, all executed database operations are automatically rolled back.

[0120] The applicant further declares that while the above embodiments illustrate the implementation method and apparatus structure of the present invention, the present invention is not limited to the above-described embodiments, meaning that the present invention must rely on the above methods and structures to be implemented. Those skilled in the art should understand that any improvements to the present invention, equivalent substitutions for the selected implementation methods, additions to steps, and selections of specific methods all fall within the protection and disclosure scope of the present invention.

[0121] This invention is not limited to the above-described embodiments. All methods that employ similar structures and approaches to achieve the objectives of this invention are within the scope of protection of this invention.

Claims

1. A dynamic scheduling system and method for intelligent forklifts based on multi-source data fusion, characterized in that, Includes the following steps: S1. Collect weighing data from the weighbridge in the processing area, in-situ detection data of temporary storage containers based on spherical monitoring, occupancy data of temporary storage bins, and usage data of the testing platform; S2. Collect the call signals from the control panels of each workstation; S3. Use algorithms to fuse multi-source data, construct a comprehensive model of the environment and tasks, and determine whether the current composite container meets the entry conditions. S4. Based on the fusion of data, intelligently process the handling and detection scheduling tasks, and determine the task priority and execution requirements; S5. The inspection system sends inspection tasks to the AGV scheduling system. S6. The AGV scheduling system receives detection tasks and schedules composite AGVs to execute the detection and scheduling tasks. S7. The composite AGV moves to the inspection platform to collect visual inspection data of the composite robot; S8. Determine whether to issue an inbound handling task based on the visual inspection results. At the same time, determine the target location of the handling task based on the on-site inspection results and optimize the inbound task scheduling by referring to the temporary storage warehouse space occupancy data. S9. The inspection system sends the warehousing and handling task to the AGV scheduling system. S10: The AGV scheduling system receives handling tasks and simultaneously collects the status data of the forklift AGV itself. S11. Intelligently allocate inbound and handling tasks based on task requirements and forklift AGV status data; S12. Use an algorithm to plan the optimal path in real time; S13. Implement conflict avoidance based on the type of scheduled task; S14. Control the forklift AGV to perform inbound and material handling tasks, and collect the execution status and running results in real time; S15 The AGV scheduling system will provide real-time feedback on the execution status and operation results to the inspection system. The inspection system will adjust the task generation strategy based on the feedback data and automatically project the inspection results, the temporary storage space occupancy status, and the operation status of forklift AGVs and composite AGVs onto the monitoring screen.

2. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: In step S3, a multi-source data weighted fusion judgment mechanism based on task type priority is used to fuse multi-source data, and different data source attention weights are set for different task types. The weighted fusion judgment mechanism includes: inputting the raw data collected by each subsystem into the fusion judgment module after normalization and encoding; calculating the comprehensive attention score according to the formula Score = w1 × f1 + w2 × f2 + w3 × f3 + w4 × f4, where f1 is the normalized value of the visual detection result, f2 is the normalized value of the inventory status, f3 is the normalized value of the call box signal, f4 is the normalized value of the electronic tag matching result, w1 to w4 are weight coefficients and w1+w2+w3+w4=1; when Score≥0.7, it is determined that the entry condition is met; otherwise, it is determined that the condition is not met and corresponding prompt information is fed back.

3. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: In step S8, the visual inspection results of the composite robot serve as the key criterion for determining whether to issue an inbound handling task. If the visual inspection is qualified, the inbound handling task is issued; if it is not qualified, it is not issued.

4. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: In step S11, transportation tasks are intelligently allocated based on task priority, forklift AGV status and location, taking into account task time requirements, cargo weight, and handling distance factors, and prioritizing the execution of high-priority tasks.

5. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: In step S13, inspection scheduling takes priority over material handling scheduling to ensure the timely execution of inspection tasks. Through task timing arrangement, work conflicts between forklift AGVs and composite AGVs are avoided.

6. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: The method also includes an online production process for semi-finished products. The intelligent logistics inspection system combines data from the logistics monitoring system, 3D vision inspection system, and electronic tag system to make a comprehensive judgment and then sends an inspection task request to the scheduling system. The comprehensive judgment adopts a multi-condition sequential verification and veto decision-making mechanism, including nine steps: input parameter verification, group validity verification, starting container positioning, material qualification judgment, target docking location search, target storage location idle judgment, starting resource readiness judgment, vehicle binding consistency judgment, and task creation and issuance. Among them, conditions C1 (material qualification condition), C2 (target storage location idle condition), and C3 (starting resource readiness condition) are necessary conditions. If any one of them is not met, the judgment will terminate immediately. If condition C4 (vehicle binding consistency condition) is not met, automatic repair will be attempted first.

7. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: The method also includes a semi-finished product warehousing and handling process. The intelligent logistics inspection system combines data from the logistics monitoring system, 3D vision inspection system, electronic tag system, and in-situ detection system to make a comprehensive judgment and then sends a warehousing and handling task request to the scheduling system. The intelligent logistics inspection system combines data from each system to perform comprehensive verification of the semi-finished products, including inspection and testing of their qualification and weighing to verify weight information.

8. The intelligent forklift dynamic scheduling method based on multi-source data fusion according to claim 1, characterized in that: The method also includes a semi-finished product outbound handling process. The intelligent logistics inspection system combines data from the logistics monitoring system, 3D vision inspection system, electronic tag system, and in-situ detection system to make a comprehensive judgment and then sends an outbound handling task request to the scheduling system.

9. A dynamic scheduling system for intelligent forklifts based on multi-source data fusion, characterized in that: It includes a core inspection system, a downstream AGV scheduling system, and a communication and integration module. The core inspection system includes a multi-source data acquisition unit, a data fusion processing unit, a task generation and distribution unit, a monitoring screen projection unit, and a monitoring and management unit. The downstream AGV scheduling system includes an AGV status acquisition unit, a task receiving and parsing unit, a dynamic scheduling decision unit, an execution control unit, and a conflict avoidance unit. The multi-source data acquisition unit is used to collect weighing data from the weighbridge in the processing area, in-situ detection data of the temporary storage container based on spherical monitoring, occupancy data of the temporary storage bin, usage data of the testing station, and call signals from the workstation control panel. The data fusion processing unit is used to fuse multi-source data using algorithms and construct a comprehensive model of the environment and tasks; The task generation and distribution unit is used to intelligently generate transportation tasks based on fused data and distribute them to the AGV scheduling system. The dynamic scheduling decision unit is used to realize the dynamic scheduling and path planning of AGV based on real-time data and task requirements. Specifically, it adopts a path selection method based on predefined route topology, including six steps: route query, route uniqueness verification, target area determination, target area connection position query, idle connection position filtering, and optimal target selection. Dynamic obstacles are handled through a three-layer mechanism of target occupancy pre-inspection, initial resource verification, and underlying collaborative obstacle avoidance. The conflict avoidance unit is used to avoid conflicts by adopting a conflict prevention strategy based on task status verification and resource status verification according to the type of scheduling task. This includes the detection and resolution of three types of scenarios: resource occupation conflict, vehicle binding conflict, and task legality conflict. The monitoring screen projection unit is used to automatically project the test results, temporary storage space occupancy, weighbridge usage, testing platform usage, and the operating status and results of forklift AGVs and composite AGVs onto the monitoring screen.

10. The intelligent forklift dynamic scheduling system based on multi-source data fusion according to claim 9, characterized in that: The multi-source data acquisition unit includes a weighbridge sensor, a spherical monitoring camera, a warehouse occupancy detection sensor, a detection platform sensor, and a control panel; electronic tags are installed on the outside of the logistics container for manual visual display, and the outbound quantity in the electronic tags is updated as the goods are shipped out.

11. The intelligent forklift dynamic scheduling system based on multi-source data fusion according to claim 9, characterized in that: The AGV status acquisition unit includes a position sensor, a speed sensor, a power sensor, and a load sensor.

12. The intelligent forklift dynamic scheduling system based on multi-source data fusion according to any one of claims 9-12, characterized in that: The system is used to implement the control method according to any one of claims 1-8; wherein, the attention mechanism network is a multi-source data weighted fusion judgment mechanism based on task type priority, the optimal path planning adopts a path selection method based on predefined route topology and combined with real-time status verification to achieve dynamic obstacle avoidance, the conflict avoidance adopts a conflict prevention strategy based on task status verification and resource status verification, and the comprehensive judgment adopts a decision mechanism of multi-condition sequential verification and veto.

Images