Dynamic task scheduling system for shoe-making flexible production line mechanical arm

Abstract

This invention discloses a dynamic task scheduling system for robotic arms in a flexible shoe manufacturing production line, aiming to solve the problems of low equipment utilization, easy line-wide shutdown due to single-point failures, and insufficient adaptability to shoe manufacturing processes caused by the fixed binding of robotic arms to workstations in existing shoe manufacturing production lines. This system includes a process-homogeneous adaptation hardware unit, a global status perception unit, and a core dynamic task scheduling engine. Through a unified quick-change standard and kinematic calibration, it enables robotic arms to perform cross-workstation operations within the same process. Combined with real-time status perception across the entire production line, it completes dynamic task scheduling and uninterrupted fault takeover. It is equipped with modules specific to shoe manufacturing processes, such as dynamic calibration, time-series coupled scheduling, and quality closed-loop feedback. This invention can significantly improve the overall utilization rate of robotic arms and the fault tolerance of the production line, adapting to mixed production lines of multiple shoe styles and small-batch customized production, significantly improving the flexibility, production efficiency, and product yield of shoe manufacturing production lines.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and flexible production control technology in the footwear industry, specifically a dynamic task scheduling system for robotic arms in flexible footwear production lines. Background Technology

[0002] This invention relates to the field of intelligent manufacturing and industrial robot scheduling technology in the shoe industry, specifically to a dynamic task scheduling system for robotic arms in flexible shoe production lines suitable for mixed production of multiple shoe styles and small-batch customized production.

[0003] The core of a flexible shoe production line is to achieve rapid switching and efficient collaboration in the production of multiple shoe styles through modular workstation design, industrial robot operation, and intelligent logistics transfer. Among them, the six-axis industrial robotic arm, as the core execution unit of the shoe manufacturing process, directly determines the production efficiency of the entire flexible production line through its operating efficiency, load balancing, and fault tolerance.

[0004] Currently, the industry has conducted extensive research and application on the modular design and automated operation of flexible shoe production lines. For example, Chinese utility model patent CN211672695U discloses a modular and reconfigurable flexible shoe production line. This solution uses multiple independently set modular shoe-making stations, each equipped with a six-axis robot to complete the shoe-making process at its station. AGV material transfer units facilitate the transfer of shoe lasts between stations, while also supporting modular reconfiguration of production line stations to adapt to the production needs of different shoe styles. This is currently a widely used technical solution in the field of flexible shoe production lines. Furthermore, Chinese invention patent CN108813828B discloses an automated shoe production line and its control method. Through a fully automated station layout, multi-robot collaborative operation, and AGV transfer logic optimization, it achieves automated production throughout the entire shoe-making process, improving the standardization level of the production line.

[0005] While the aforementioned existing technologies have achieved flexibility and automation in shoe production lines to some extent, there are still core technical pain points that cannot be resolved in practical applications, as follows: First, the fixed-bond architecture of the robotic arms and workstations results in extremely low overall equipment utilization. In the existing solution, each six-axis robotic arm can only execute the preset process tasks of its corresponding bound workstation and cannot complete the same type of process across workstations. However, in the mixed production line of shoe manufacturing, the operation time of each process varies greatly for different shoe styles. This easily leads to situations where some workstations' robotic arms are running at full capacity, causing severe congestion in the task queue, while robotic arms in adjacent workstations are idle for long periods. Actual production line verification shows that the overall utilization rate of robotic arms under this architecture is generally less than 45%, resulting in a significant waste of hardware capacity.

[0006] Secondly, the production line has extremely poor fault tolerance, posing a risk of a single point of failure causing a complete shutdown. In the existing solution, the operation of each workstation relies entirely on a single, dedicated robotic arm. If that robotic arm malfunctions and stops or issues a maintenance warning, the operation of the corresponding workstation is immediately interrupted. Even if there is an idle robotic arm at an adjacent workstation, it cannot take over the operation, directly causing the production cycle of the entire production line to be interrupted, order delivery cycles to be delayed, and the need for continuous production to be met.

[0007] Third, the lack of scheduling optimization tailored to the specific technological characteristics of the footwear industry easily leads to workpiece scrapping and disrupted production cycles. Footwear manufacturing processes are highly interconnected and subject to strict time constraints. For example, after applying adhesive, baking and pressing must be completed within a limited time; exceeding this time window will directly result in adhesive curing failure and workpiece scrapping. Furthermore, the quality of preceding processes directly affects the difficulty and duration of subsequent processes. Existing scheduling solutions only involve pre-set task allocation with fixed production cycles, failing to consider the temporal constraints and quality coupling characteristics of footwear manufacturing processes. This easily leads to problems such as idle processes, overdue workpiece scrapping, and defective products flowing into the next process, making it impossible to guarantee yield and production stability during mixed-line production.

[0008] Fourth, it cannot adapt to the mixed-line production needs of small-batch customized orders. The process complexity and single-workpiece operation time of customized orders are much higher than those of standard batch orders. Existing technology can only achieve passive real-time task allocation and cannot perform pre-load balancing and pre-scheduling based on the process complexity of the order. When high-complexity orders are put into production in a concentrated manner, it is easy to cause full-line scheduling congestion and production rhythm imbalance, and the delivery cycle of customized orders cannot be effectively controlled.

[0009] In summary, the existing robotic arm operation and scheduling solutions for flexible shoe production lines cannot simultaneously meet the core requirements of high equipment utilization, high production line fault tolerance, process adaptability, and order delivery stability under mixed-line production. This has become a key technical bottleneck restricting the flexible and intelligent upgrading of the shoe industry. Summary of the Invention

[0010] The purpose of this invention is to provide a dynamic task scheduling system for robotic arms in flexible shoe production lines. This invention breaks through the inherent technical bias of fixed binding of robotic arms to workstations in flexible shoe production lines. Under the premise of low-cost modification and complete reuse of existing production line hardware, it achieves a doubling of the comprehensive utilization rate of robotic arms and uninterrupted takeover of single-point failures in the production line. At the same time, it is deeply adapted to the special process characteristics of shoemaking and the production needs of small-batch, fast-response production, and simultaneously achieves the improvement of production efficiency, product yield, and reduction of process consumable costs.

[0011] The technical solution adopted in this invention is as follows: A dynamic task scheduling system for robotic arms in a flexible shoe manufacturing production line, the flexible shoe manufacturing production line including multiple modular shoe manufacturing stations arranged along the material flow path, multiple six-axis industrial robotic arms, and AGV material transfer units for transferring shoe lasts. The initial working range of each six-axis industrial robotic arm covers at least one of the modular shoe manufacturing stations, and each six-axis industrial robotic arm is used to perform shoe manufacturing processes. The system includes: a) Process-integrated hardware unit, including a standardized quick-change connector, multiple sets of shoe-making process-specific end effectors, and a workstation quick-change table. The quick-change connector is installed at the end of each of the six-axis industrial robotic arms. The shoe-making process-specific end effectors are detachably installed on the workstation quick-change table. The workstation quick-change table is arranged next to each modular shoe-making workstation. All the six-axis industrial robotic arms have completed standardized kinematic parameter calibration and have the ability to perform all shoe-making processes. b) The global state perception unit includes a global vision positioning module deployed on the top of the flexible shoe production line, a workstation calibration component installed at each modular shoe production station, and an RFID storage chip built into the shoe last workpiece. The global vision positioning module is used to collect real-time data on the position of the shoe last workpiece, the posture of the robotic arm, and the position of the end effector throughout the entire production line. The workstation calibration component is used to provide a reference for the workstation's coordinate system. The RFID storage chip is used to store the process parameters and positioning reference data for the corresponding shoe model. c) A core dynamic task scheduling engine, which is communicatively connected to the hardware unit adapted to the aforementioned process, the global state perception unit, the six-axis industrial robotic arm, and the AGV material transfer unit, is configured as follows: i) Real-time data collection and updates of the load rate, remaining working time, and equipment health status of each six-axis industrial robotic arm, as well as the queue length of the tasks to be processed and the priority data of the corresponding orders for each modular shoemaking station. ii) When the length of the task queue to be processed at the target modular shoemaking station exceeds the preset congestion threshold and the real-time load rate of the six-axis industrial robot arm at the corresponding station continues to exceed the preset high load threshold, the system automatically searches for idle six-axis industrial robots whose work range can cover the target modular shoemaking station and whose real-time load rate is lower than the preset low load threshold, and generates cross-station work scheduling instructions. iii) Issue the cross-station operation scheduling command to the idle six-axis industrial robotic arm, control it to replace the end effector of the corresponding shoemaking process through the quick-change connector, enter the target modular shoemaking station to complete the corresponding process operation, and return to the initial station to wait after the operation is completed; iv) When any six-axis industrial robotic arm is detected to have a malfunction or health status warning, all pending tasks at the corresponding workstation of that robotic arm will be dynamically assigned to any available six-axis industrial robotic arm whose work range can cover that workstation, thus enabling uninterrupted takeover of operations during malfunctions.

[0012] Preferably, the core dynamic task scheduling engine is also equipped with a shoe last reference dynamic calibration module. The shoe last reference dynamic calibration module is configured to: when a six-axis industrial robot arm that receives a cross-workstation operation scheduling instruction enters the target modular shoemaking workstation, first scan the workstation calibration component of the workstation through the global vision positioning module to complete the real-time recalibration of the operation coordinate system, and then read the process parameters and positioning reference data in the RFID storage chip of the target shoe last workpiece to automatically correct the operation trajectory and ensure that the process accuracy of the cross-workstation operation is consistent with the six-axis industrial robot arm initially configured for the workstation.

[0013] Preferably, the core dynamic task scheduling engine is also equipped with a spatiotemporal segmentation obstacle avoidance scheduling module. The spatiotemporal segmentation obstacle avoidance scheduling module is configured to: pre-divide the working space of the flexible shoe production line into several grid units according to modular shoe workstations, and divide the time axis into several time slices according to a preset unit duration; when issuing cross-workstation operation scheduling instructions, pre-plan a non-interference motion path for the corresponding six-axis industrial robotic arm, and allocate a dedicated spatial grid unit and time slice to it, ensuring that only one working device is allowed to occupy the same spatial grid unit within the same time slice; at the same time, monitor the movement posture of the equipment in real time through the global vision positioning module, and immediately allocate backup paths and time slices for low-priority tasks when interference risks are predicted, so as to achieve zero-collision obstacle avoidance.

[0014] Preferably, the core dynamic task scheduling engine is further configured with a process timing coupling scheduling sub-engine. The process timing coupling scheduling sub-engine pre-stores the process timing constraint model corresponding to each shoe model. The process timing constraint model is marked with the preceding / following process relationship, the effective time window of the process, and the shortest and longest operation time of each process for the corresponding shoe model. The process timing coupling scheduling sub-engine is configured to: when allocating process tasks, predict the completion time of the preceding process based on the entire process timing chain, issue a pre-scheduling instruction in advance, control the corresponding six-axis industrial robotic arm to complete the end effector replacement and coordinate system calibration in advance, and immediately enter the workstation to perform the operation the moment the preceding process is completed; when the operation time of a certain process exceeds a preset threshold, automatically adjust the task allocation priority and scheduling sequence of the following process to ensure that all processes are completed within the effective time window of the process.

[0015] Preferably, the core dynamic task scheduling engine is also equipped with an order process feature pre-scheduling module. The order process feature pre-scheduling module is configured to: automatically parse the BOM files and process files of all shoe styles in the order when a production order is received, extract the process complexity features and generate the corresponding load coefficient of the order; based on the rated load capacity of the six-axis industrial robotic arms of the entire production line, split the order task into multiple task packages, pre-allocate the corresponding six-axis industrial robotic arm clusters to each task package, and reserve the corresponding operation time window and hardware resources; during mixed-line production, dynamically adjust the task production sequence based on order priority and load coefficient, stagger the production of high-load customized orders and low-load standard orders, so that the real-time load rate of the six-axis industrial robotic arms of the entire production line is maintained within the preset optimal range.

[0016] Preferably, the core dynamic task scheduling engine is also equipped with a quality-scheduling closed-loop feedback module. The quality-scheduling closed-loop feedback module is communicatively connected to the process quality inspection unit of each modular shoemaking station. The quality-scheduling closed-loop feedback module is configured to: receive workpiece operation quality scores and defect type data uploaded by the process quality inspection unit; for workpieces that are qualified but have quality scores below a preset threshold, allocate redundant operation time to their subsequent processes and prioritize scheduling six-axis industrial robotic arms that meet preset requirements for operation accuracy and operational stability; for workpieces with repairable defects, insert the rework task into the highest priority of the corresponding station's task queue and schedule the nearest idle six-axis industrial robotic arm to perform the rework; for non-conforming products, immediately schedule AGV material transfer units to transfer them to the waste area and delete the scheduling tasks of all subsequent processes of the workpiece to release scheduling resources.

[0017] Preferably, the core dynamic task scheduling engine is also configured with a hierarchical dual-queue scheduling mechanism. This mechanism divides all tasks to be executed into primary process tasks and secondary auxiliary tasks. The primary process tasks are high-precision processes such as shoemaking, grinding, gluing, pressing, and sole attaching. The secondary auxiliary tasks are low-precision auxiliary tasks such as workpiece loading, unloading, and transfer. The hierarchical dual-queue scheduling mechanism is configured to set up independent primary and secondary task queues for each six-axis industrial robot arm, strictly following the order of primary tasks first, then secondary tasks. Primary process tasks must not be interrupted by secondary auxiliary tasks during execution. Secondary auxiliary tasks can only be assigned to six-axis industrial robots that have no primary process tasks to be executed and whose real-time load rate is lower than a preset low load threshold.

[0018] Preferably, the end effector for the shoe manufacturing process includes a shoe upper grinding actuator, a shoe upper gluing actuator, a shoe upper and sole pressing actuator, and a shoe last clamping actuator. All end effectors are equipped with a unified standard docking interface that matches the quick-change connector, enabling rapid adaptation and replacement of any six-axis industrial robotic arm.

[0019] Preferably, when the core dynamic task scheduling engine performs a fault-free takeover, it prioritizes assigning tasks to idle six-axis industrial robotic arms that are adjacent to the faulty robotic arm and whose operating range completely covers the faulty workstation. At the same time, based on the order priority and process timing constraints of the tasks, it adjusts the task execution order to ensure that high-priority orders and tasks with adjacent effective process time windows are executed first.

[0020] Preferably, the system is suitable for flexible shoe production lines that support mixed production of multiple shoe styles and small-batch customized production, and can be adapted to the entire shoe manufacturing process of sports shoes, leather shoes, and casual shoes.

[0021] This invention breaks through the inherent technical bias of "fixed binding of robotic arms and workstations" in existing flexible shoe production lines. It achieves comprehensive innovation targeting the unique process characteristics and production pain points of the shoe industry, making significant and substantial progress compared to existing technologies. Furthermore, it realizes several groundbreaking and beneficial effects that could not have been foreseen by those skilled in the art based on existing technologies. The overall content is as follows: First, without adding any new core production hardware, this invention achieves a significant leap in the overall utilization rate of robotic arms, resulting in unexpected capacity gains. In existing technologies, due to the limitations of workstation-bound architectures, the overall utilization rate of robotic arms in mixed-line production is generally less than 45%. Those skilled in the art have long held the technical prejudice that "high-precision processes in shoemaking must be completed by robotic arms bound to specific workstations, and it is impossible to improve equipment utilization through cross-workstation operations." However, this invention, through process-homogeneous hardware design and a dynamic scheduling core architecture, achieves a stable increase in the overall utilization rate of robotic arms in mixed-line production to over 85% without fully reusing existing production line robotic arm hardware and with modification costs less than 10% of adding new equipment. This nearly doubles the rate, completely unleashing the capacity potential of existing hardware—a technical effect that those skilled in the art could not have predicted through simple combinations of existing technologies.

[0022] Secondly, it achieves uninterrupted operation takeover for single-point failures, completely solving the long-standing industry problem of "single-point failure leading to a complete line shutdown." In existing technologies, the failure of a single robotic arm inevitably leads to the interruption of the corresponding process, and may even cause chaos in the entire line's rhythm. This invention not only achieves dynamic and proximity-based allocation of faulty tasks, but also optimizes scheduling by combining shoemaking process timing constraints and order priorities. This reduces unplanned downtime caused by single-equipment failures by more than 90%, while completely avoiding the problem of workpieces being scrapped beyond the process time window during fault takeover. It achieves production line-level fault tolerance without adding redundant backup equipment, an effect that was also unpredictable by those skilled in the art.

[0023] Third, it achieves simultaneous optimization of production efficiency, product yield, and material costs, breaking the conventional wisdom in the field that "efficiency improvement and quality assurance are mutually exclusive." This invention deeply couples the scheduling system with the specific process characteristics of shoemaking. Through innovative designs such as time-series coupled scheduling and closed-loop quality feedback, it stabilizes the production line yield at over 99.5%, while reducing the consumption of process consumables such as glue by 30% and shortening the delivery cycle of small-batch customized orders by more than 40%. It simultaneously improves production efficiency and achieves quality improvement and cost reduction, resulting in unexpected technical effects of multi-dimensional synergistic gains. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a flowchart illustrating the cross-workstation operation scheduling logic of the present invention. Detailed Implementation

[0025] This specific embodiment is used to illustrate in detail the dynamic task scheduling system for the robotic arm of the flexible shoe production line of the present invention. The described embodiments are only used to explain the technical solution of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, parameter optimizations, and structural improvements made within the core concept and principles of the present invention should fall within the scope of protection of the present invention.

[0026] This system is applicable to the entire shoe manufacturing process of athletic shoes, leather shoes, and casual shoes. It can be perfectly adapted to flexible shoe production line scenarios such as mixed production of multiple shoe styles and small-batch customized production. It solves the core pain points of the industry, such as low equipment utilization, single-point failure leading to the shutdown of the entire line, and disconnect between process constraints and scheduling, which are caused by the fixed binding of robotic arms and workstations in existing shoe production lines.

[0027] See Figure 1 and 2The flexible shoe-making production line upon which this invention is based includes multiple modular shoe-making workstations arranged along the material flow path, multiple six-axis industrial robotic arms, and AGV material transfer units for transferring shoe lasts. The modular shoe-making workstations are arranged sequentially along the AGV's closed-loop flow path, and are divided into four stages according to the entire shoe-making process: upper pretreatment, upper grinding, upper gluing, shoe last baking, upper and sole pressing, finished product shaping, and finished product inspection. Each workstation is equipped with a precise AGV docking position and a work reference positioning structure. The working range of a single workstation can be covered by the initial working radius of at least one six-axis industrial robotic arm. Each six-axis industrial robotic arm is fixed to the production line floor with anchor bolts, and its initial working range covers at least one modular shoe-making workstation to perform the corresponding shoe-making process. The controllers of all six-axis industrial robotic arms are connected to the production line's central control system via industrial Ethernet, supporting real-time status data uploads and motion control command issuance.

[0028] The core of this system consists of three basic units: a process-homogeneous adaptation hardware unit, a global status perception unit, and a core dynamic task scheduling engine. These three units communicate in real time via industrial Ethernet with a communication latency of no more than 1ms, meeting the control requirements for real-time scheduling of the production line.

[0029] The process-homogeneous adaptation hardware unit is the hardware foundation for realizing cross-workstation operation of robotic arms with the same process. The core includes a unified standard quick-change connector, multiple sets of shoemaking process-specific end effectors, and workstation quick-change table.

[0030] The quick-connect coupling adopts a mechanically locking universal docking structure, consisting of a main coupling on the robotic arm side and a slave coupling on the actuator side. The main coupling is fixedly installed on the end output shaft of each six-axis industrial robotic arm via a flange, while the slave coupling is uniformly fixedly installed on the tail mounting base of all shoemaking process-specific end actuators. All main and slave couplings adhere to a completely standardized mechanical docking dimensions, electrical signal interfaces, and pneumatic control interfaces. After docking, mechanical locking, electrical signal connection, and synchronous pneumatic path conduction are achieved. The repeatability after locking is no less than ±0.02mm, meeting the operational requirements of high-precision shoemaking processes. The locking and unlocking actions of the quick-connect coupling are controlled by the robotic arm controller via electrical signals, and a position detection sensor provides real-time feedback on the docking status, ensuring a safe and reliable docking process.

[0031] The shoe manufacturing process is equipped with specialized end effectors that correspond to the core processes of the entire shoe manufacturing workflow. These include four types: shoe upper grinding actuators, shoe upper gluing actuators, shoe upper and sole bonding actuators, and shoe last clamping actuators. All end effectors are equipped with a standardized slave connector at the tail, allowing for quick docking and replacement with the master connector of any six-axis industrial robotic arm. The shoe upper grinding actuator includes a servo drive motor, a replaceable grinding head, and a two-dimensional force control sensor. It is used for contour grinding and edge removal of the shoe upper and sole. The force control sensor can collect the contact force during grinding in real time, achieving constant force grinding in conjunction with the robotic arm's force control mode. The shoe upper gluing actuator includes a metering glue valve, a glue gun body, a glue path controller, and a laser contour sensor. It is used for gluing the bonding surfaces of the shoe upper and sole. The metering glue valve precisely controls the glue dispensing amount, and the laser contour sensor... It can correct the glue application trajectory in real time to adapt to the contour changes of different shoe styles; the shoe upper and sole pressing actuator includes a contour pressing head, pressure sensor, and pressure regulating cylinder, which is used to perform the bonding and pressing process of the shoe upper and sole. The pressing pressure can be precisely controlled by the pressure regulating cylinder to adapt to the pressing process requirements of different shoe sizes and materials; the shoe last clamping actuator includes pneumatic grippers, positioning pins, and clamping force sensors, which is used to perform the loading, unloading, transfer, and positioning clamping processes of shoe last workpieces, and can achieve stable clamping and precise positioning of shoe lasts.

[0032] The quick-change station is fixedly installed on the side of the AGV docking position of each modular shoemaking station. The installation height matches the end-effector working height of the six-axis industrial robotic arm. The quick-change station table surface is equipped with positioning slots that match the shape of the end effector for fixing and placing the dedicated end effector of the corresponding station. Each slot is equipped with an arrival detection sensor and an RFID identification tag for real-time detection of the end effector's position status and model information. All arrival detection sensor signals are uploaded to the core dynamic task scheduling engine in real time.

[0033] To enable all six-axis industrial robotic arms to perform the same processes, this unit performs unified kinematic parameter calibration on all six-axis industrial robotic arms within the production line. The calibration process is based on a unified world coordinate system for the production line: First, using the calibration board coordinate system of the first workstation on the production line as a reference, a unified three-dimensional world coordinate system for the production line is established, where the X-axis extends along the length of the production line, the Y-axis extends along the width of the production line, and the Z-axis is perpendicular to the ground and upwards. Then, using a global vision positioning module, the DH parameters (link length, link torsion angle, joint offset, and joint rotation angle) of each six-axis industrial robotic arm are uniformly calibrated to eliminate errors in the robotic arm's installation position, link machining, and joint wear. A transformation matrix between the base coordinate system and the world coordinate system for each robotic arm is established, so that the motion parameters of all robotic arms are mapped to a unified world coordinate system. Finally, through standard trajectory testing, it is ensured that the motion accuracy and repeatability of all robotic arms meet the operational requirements of the shoemaking process, enabling all six-axis industrial robotic arms to have the basic ability to perform all the same shoemaking processes.

[0034] The global state perception unit is the data foundation for the core dynamic task scheduling engine to achieve precise scheduling. The core includes a global vision positioning module deployed on the top of the flexible shoe production line, a workstation calibration component installed at each modular shoemaking workstation, and an RFID storage chip built into the shoe last workpiece.

[0035] The global vision positioning module consists of multiple industrial area array cameras evenly deployed along the mounting brackets at the top of the production line, with a spacing of 3 meters. The fields of view of all cameras are stitched together to completely cover the entire working area of ​​the production line. The frame rate of each camera is no less than 30fps, and the image resolution is no less than 2 million pixels. All cameras have been uniformly calibrated, and their coordinate systems have been transformed to the production line's unified world coordinate system. The global vision positioning module has a built-in edge processing chip, which can perform target detection and 3D pose calculation algorithms in real time. It can collect and output real-time data on the position and posture of shoe lasts, the end effector posture and motion trajectory of the six-axis industrial robot, the in-situ status and position of the end effector, and the position and motion status data of the AGV material transfer unit. The data update frequency is no less than 10Hz. All data is uploaded to the core dynamic task scheduling engine in real time, providing full-dimensional real-time status perception data for scheduling decisions.

[0036] The workstation calibration component is fixedly installed at the reference position of the AGV docking position in each modular shoemaking workstation. It includes a checkerboard vision calibration board and a workstation RFID reader. The coordinate system of the vision calibration board is completely coincident with the working coordinate system of the corresponding workstation, serving as the working reference coordinate system for the workstation and used for coordinate system recalibration when the robotic arm operates across workstations. The reading range of the workstation RFID reader covers the AGV docking position of the workstation and can read the RFID storage chip data in the shoe last workpiece docked at the workstation in real time. The read data is uploaded to the core dynamic task scheduling engine in real time.

[0037] The RFID storage chip uses a passive UHF RFID tag, which is embedded in the internal cavity of the shoe last workpiece. It does not affect the shape and size of the shoe last or the shoemaking process. Each RFID storage chip has a unique ID number corresponding to a single shoe last workpiece. The data stored in the chip includes: the corresponding shoe model number, shoe size specification, shoe process parameters (grinding trajectory, glue application path, pressing pressure and duration), three-dimensional coordinates of the positioning reference point, process sequence constraint parameters, the order number, order priority, and records of completed processes. All data can be read and updated through the workstation RFID reader, providing workpiece-level dedicated data support for the correction of the robotic arm's work trajectory and scheduling decisions.

[0038] The core dynamic task scheduling engine is the central control hub of this system. Its hardware platform utilizes an industrial-grade edge computing server, installed within the main control cabinet of the production line. It achieves bidirectional real-time communication via industrial Ethernet with process-compatible hardware units, the overall status perception unit, the controllers of all six-axis industrial robotic arms, and the main control system of the AGV material handling unit. The engine's software architecture employs a layered design, consisting of a data layer, a logic layer, and an execution layer. The data layer is responsible for the real-time acquisition, cleaning, and storage of status data across the entire production line. The logic layer integrates all scheduling algorithms and functional modules, responsible for generating scheduling decisions based on real-time data. The execution layer is responsible for converting scheduling instructions into executable control commands for the equipment, distributing them to the corresponding hardware devices, and providing feedback on the execution results.

[0039] The core scheduling logic of the core dynamic task scheduling engine is as follows: First, the system collects and updates the status data of the entire production line in real time at a frequency of 100Hz to ensure the real-time nature of scheduling decisions. The collected data includes: real-time load rate, remaining operating time, and equipment health status data for each six-axis industrial robotic arm. The real-time load rate is the maximum ratio of the real-time output torque of each joint motor to its rated torque, directly output by the robotic arm controller. The remaining operating time is calculated based on preset process parameters, subtracting the already executed time from the preset total time of the currently executing task. Equipment health status data includes the temperature, vibration value, reducer operating status, and fault codes of each joint motor, uploaded in real time by the robotic arm controller. Simultaneously, the collected data also includes the queue length of tasks awaiting processing at each modular shoemaking station and the priority data of corresponding orders. The queue length represents the number of tasks waiting to be executed at that station. Order priorities are divided into 1-5 levels, with level 1 being the highest priority, corresponding to urgent and customized orders, and level 5 being the lowest priority, corresponding to regular batch orders.

[0040] Second, the engine automatically triggers cross-workstation job scheduling and generates cross-workstation job scheduling instructions. The engine has built-in cross-workstation scheduling trigger logic with preset congestion thresholds of 3, high load thresholds of 90%, duration thresholds of 5 seconds, and low load thresholds of 30%. When the queue length of tasks awaiting processing at the target modular shoemaking workstation exceeds the preset congestion threshold, and the real-time load rate of the corresponding workstation's six-axis industrial robot arm continuously exceeds the preset high load threshold for a duration exceeding the duration threshold, the engine automatically triggers the cross-workstation scheduling process. First, it searches for idle six-axis industrial robots whose work area can cover the target modular shoemaking workstation and whose real-time load rate is lower than the preset low load threshold. The logic for determining work area coverage is as follows: based on the calibrated reachable workspace model of the robot arm, it determines whether the target workstation's work area is completely within the robot arm's reachable workspace and does not pass through the robot arm's singularity region, ensuring that the robot arm can successfully complete the target process. After finding a qualified idle robot arm, the engine plans a complete work path based on the type of the target process, the location of the required end effector, and the robot arm's current position, generating the corresponding cross-workstation job scheduling instructions.

[0041] Third, control the entire process of cross-workstation operations. The engine issues cross-workstation operation scheduling instructions to the selected idle six-axis industrial robotic arm, controlling it to execute the operation according to the following process: First, the robotic arm moves to the quick-change station of the corresponding end effector, completes the unlocking and placement of the current end effector, and then docks and locks with the end effector corresponding to the target process. After the positioning detection sensor confirms the docking is complete, it sends a feedback signal to the engine. Next, the robotic arm moves to the calibration plate position of the target modular shoemaking station, completes the real-time recalibration of the operation coordinate system, and at the same time reads the process parameters and positioning reference data in the RFID storage chip of the target shoe last workpiece through the workpiece RFID reader, automatically correcting the operation trajectory. Then, the robotic arm enters the operation area of ​​the target workpiece and executes the corresponding shoemaking process according to the corrected operation trajectory, and provides real-time feedback on the execution status to the engine during the operation. Finally, after the operation is completed, the robotic arm puts the end effector back into the positioning slot of the corresponding quick-change station, returns to its initial workpiece to enter the standby state, and sends a work completion signal to the engine. The engine updates the process completion status of the corresponding workpiece and transfers it to the next process.

[0042] Fourth, it enables uninterrupted operation takeover in the event of equipment failure. The engine monitors the health status data and fault codes of all six-axis industrial robotic arms in real time. When any six-axis industrial robotic arm is detected to have stopped due to a fault (fault code uploaded by the controller) or when health status parameters exceed preset warning thresholds (such as motor temperature exceeding 85°C or vibration exceeding safety thresholds), the fault takeover process is immediately triggered. First, the engine immediately suspends the current execution task of the faulty robotic arm, removes the unfinished task and all pending tasks corresponding to that workstation from the original task queue, and generates a fault takeover task list. Then, the engine prioritizes searching for available six-axis industrial robotic arms adjacent to the faulty robotic arm and whose operating range completely covers the faulty workstation, and completes task allocation according to task priority. The weighting formula for task priority is: ; In the formula: This represents the overall priority weight of the task; the larger the value, the higher the priority. The order priority weights are as follows: Level 1 orders have a weight of 5, and Level 5 orders have a weight of 1. The urgency weight of the time window is calculated using the following formula: ,in This is the deadline for the effective time window of the process corresponding to this task. This is the current system time.

[0043] The engine assigns tasks to available robotic arms in descending order of overall priority weight, ensuring that high-priority orders and tasks near the process time window are executed first, avoiding order delays and workpiece scrapping, and enabling uninterrupted takeover of faulty workstations, thus completely solving the pain point of the entire line stopping due to the failure of a single piece of equipment in the existing technology.

[0044] The shoe last benchmark dynamic calibration module is integrated into the core dynamic task scheduling engine. It is used to solve the coordinate system deviation problem between the robotic arm and the target workstation when performing cross-workstation operations, and to ensure that the process accuracy of cross-workstation operations is completely consistent with that of the robotic arm at the original workstation.

[0045] The sanding and gluing processes in shoemaking require extremely high trajectory accuracy, with a maximum allowable deviation of no more than ±0.1mm. However, there are installation position errors and workstation reference deviations between the robotic arm operating across workstations and the target workstation. Directly using a preset trajectory will result in accuracy exceeding the tolerance. This module completely solves this problem through real-time coordinate system recalibration and trajectory correction.

[0046] The specific implementation process of this module is as follows: When the six-axis industrial robotic arm receives the cross-workstation operation scheduling command and enters the target modular shoemaking workstation, it first moves to a preset calibration position directly above the calibration plate of that workstation. The preset calibration position is 300mm above the calibration plate, ensuring that the calibration plate is completely within the field of view of the global vision positioning module. Subsequently, the global vision positioning module acquires high-definition images of the calibration plate and calculates the precise three-dimensional coordinates and attitude of the calibration plate in the world coordinate system through a corner detection algorithm. At the same time, it calculates the real-time pose of the robotic arm end effector in the world coordinate system. Then, the module calculates the real-time transformation matrix between the target workstation's operating coordinate system and the robotic arm's base coordinate system. The formula for calculating the transformation matrix is: ; In the formula: This is the transformation matrix from the robot arm's base coordinate system to the target workstation's operating coordinate system; This is the pre-calibration transformation matrix from the world coordinate system to the robot arm's base coordinate system; This is the real-time transformation matrix from the target workstation coordinate system (calibration plate coordinate system) to the world coordinate system.

[0047] Based on this transformation matrix, the module performs coordinate transformation on the preset process trajectory, converting the trajectory from the workstation coordinate system to the robot arm base coordinate system. Subsequently, the module reads the positioning reference point coordinates and process parameters in the RFID storage chip of the target shoe last workpiece through the workstation RFID reader, and performs secondary correction on the transformed trajectory to eliminate the positional deviation of the shoe last clamping. Ultimately, it ensures that the accuracy error of the robot arm's trajectory does not exceed ±0.1mm, which is completely consistent with the robot arm's initial configuration accuracy at the workstation, meeting the operational requirements of high-precision shoe manufacturing processes.

[0048] The spatiotemporal segmentation obstacle avoidance scheduling module is integrated into the core dynamic task scheduling engine. It is used to solve the motion interference problem of robotic arms operating across workstations and AGV transfer in the narrow space of the shoe production line, and to achieve zero-collision obstacle avoidance for all equipment on the production line.

[0049] The workstations on the flexible shoe production line are close together and the working space is narrow. The movement path of the robotic arm working across workstations overlaps with the transfer path of the AGV. Existing reactive obstacle avoidance algorithms are prone to sudden stops and collisions. This module avoids motion interference from the source by using a pre-allocation mechanism of spatial gridding and time partitioning.

[0050] The specific implementation of this module is as follows: First, the module pre-divides the entire working space of the flexible shoe production line into cubic grid units with a side length of 100mm according to the unified world coordinate system of the production line. Each grid unit has a unique spatial number, and the coordinate range of the grid unit is fixed, which can be accurately mapped to the corresponding space under the world coordinate system. Then, the module divides the system's time axis into continuous time slices in 100ms units. Each time slice has a unique time number, and all scheduling commands of the engine are synchronized in units of time slices to ensure that the action time of all equipment on the production line is synchronized.

[0051] When the engine issues a cross-workstation job scheduling command, the module first uses A The path planning algorithm plans a collision-free motion path for the robotic arm from its current position to the target workstation. All grid cells along the path, as well as those that need to be occupied during the operation, are marked as unoccupied space resources. Subsequently, the module allocates corresponding continuous time slices to these unoccupied grid cells, ensuring that only one operating device (a six-axis industrial robotic arm or an AGV material transfer unit) is allowed to occupy the same grid cell within the same time slice. The transfer path and space occupation of the AGV are also planned and allocated uniformly by this module.

[0052] Meanwhile, the module monitors the motion posture and position of all equipment on the entire production line in real time through the global vision positioning module, updates the space occupancy status of the equipment every 100ms, and predicts whether there is an interference risk of two devices occupying the same grid cell within the next three time slices. If an interference risk is predicted, the module immediately suspends the execution of low-priority tasks, replans interference-free motion paths for low-priority tasks, and allocates new grid cells and time slices to ensure zero-collision operation of all equipment on the production line and avoid production line downtime and workpiece scrapping caused by equipment interference.

[0053] The process timing coupling scheduling sub-engine is integrated into the core dynamic task scheduling engine. It is designed specifically for the strong sequential correlation and strict process time window constraints unique to shoe manufacturing processes, solving the problems of process waiting and workpiece scrapping due to exceeding the time window that are prone to occur in existing scheduling schemes.

[0054] In the shoe manufacturing process, after the glue is applied, baking must be completed within a preset time window. After baking, the sole must be pressed together within a preset time window. Exceeding the time window will cause the glue to cure and fail, and the workpiece will be scrapped. This is a unique process characteristic of the shoe manufacturing industry, and the general scheduling scheme does not cover this constraint at all.

[0055] The specific implementation of this sub-engine is as follows: First, the sub-engine's database pre-stores the process timing constraint models corresponding to each shoe model. These models are built based on the corresponding shoe model's process documents and are labeled with the preceding process number, the succeeding process number, the effective process time window, the shortest operation time, and the longest operation time for each process. The effective process time window is represented by a binary tuple. express, This is the shortest waiting time for this process after the preceding process is completed, in order to meet the requirements of process leveling and cooling. This is the maximum allowable interval between the completion of the preceding process and the completion of the current process. Workpieces exceeding this interval are scrapped. The core scheduling logic of this sub-engine is based on the critical path method and real-time process status feedback. Specifically, when the preceding process of a workpiece is completed, the sub-engine immediately records the completion time of that process and, based on the process sequence constraint model, calculates the effective time window deadline for this process. ,in The completion time of the preceding process is used; subsequently, based on the longest operation time of this process, the latest start time of this process is calculated. ,in This is the longest operation time for this process.

[0056] To achieve seamless workflow transitions, the sub-engine sends a pre-scheduling command to the corresponding six-axis industrial robotic arm 10 seconds before the latest start time of the current workflow. This command controls the robotic arm to complete preparatory work such as end effector replacement and coordinate system calibration in advance. At the optimal time point after the completion of the preceding workflow, the robotic arm immediately enters the workstation to perform its task. This avoids the time wasted waiting due to the robotic arm arriving early and completely eliminates the risk of missing the process time window. When the actual operation time of a workflow exceeds the preset maximum operation time, resulting in a workflow delay, the sub-engine immediately recalculates the effective time window and latest start time of all subsequent workflows. It dynamically adjusts the allocation priority of subsequent tasks, prioritizing tasks with deadlines approaching the end time window for idle robotic arms. This ensures that all workflows are completed within the effective process time window, completely resolving the issue of workpiece scrap caused by workflow timing constraints.

[0057] The order process feature pre-scheduling module is integrated into the core dynamic task scheduling engine. It is designed for the "small order, fast response" production characteristics of the footwear industry and solves the problems of pre-scheduling imbalance and production line congestion when small batch customized orders are mixed on the production line.

[0058] In flexible production lines for shoe manufacturing, the process complexity of customized orders is much higher than that of standard batch orders. The processing time for a single workpiece can be more than three times that of a standard order. When high-complexity orders are put into production in a concentrated manner, it will lead to production line load imbalance, scheduling congestion, and extended order delivery cycle. Existing passive real-time scheduling solutions cannot solve this problem.

[0059] The specific implementation of this module is as follows: First, when a production order is connected to the production line MES system, the module automatically reads the core data of the order, including the order number, order priority, shoe model number, total order quantity, and delivery time. Then, the module automatically parses the BOM file and process file of the corresponding shoe model and extracts the process complexity characteristics of the shoe model, including the total number of processes, the proportion of special processes, the average operation time of a single process, and the total operation time of a single workpiece. Among them, special processes include non-standard processes such as logo grinding and special color matching glue application. The proportion of special processes is the ratio of the number of special processes to the total number of processes.

[0060] Based on the extracted process complexity features, the module calculates the overall load factor for this order using the following formula: ; In the formula: This is the order load factor; the larger the value, the greater the load pressure the order places on the production line. This represents the total number of workpieces in the order. This represents the average total operation time for a single workpiece. The special process coefficient is calculated using the following formula: The proportion of special processes; is the production line baseline load constant, and is the rated daily operating capacity of a single six-axis industrial robotic arm.

[0061] After calculating the order load factor, the pre-scheduling engine divides the order tasks into multiple independent task packages based on the number of six-axis industrial robotic arms and their rated load capacity across the entire production line. The load factor of each task package does not exceed 80% of the rated load of a single robotic arm. Subsequently, a corresponding cluster of six-axis industrial robotic arms is allocated to each task package, and corresponding operation time windows and hardware resources are reserved in advance to avoid load imbalance caused by concentrated task deployment.

[0062] During mixed-line production, the module dynamically adjusts the production sequence of tasks based on order priority and load factor, and adopts a staggered production strategy to alternate between customized orders with high load factor and standard batch orders with low load factor. This ensures that the average real-time load rate of all six-axis industrial robotic arms on the entire production line is always maintained in the optimal operating range of 70%-85%. Within this range, the robotic arms have the highest operating efficiency and the lowest failure risk, which maximizes equipment utilization and avoids problems such as untimely scheduling response and production line congestion caused by excessive load, and significantly shortens the delivery cycle of small batch customized orders.

[0063] The quality-scheduling closed-loop feedback module is integrated into the core dynamic task scheduling engine to solve the closed-loop failure problem of process quality and scheduling being disconnected in the existing technology, and to realize real-time linkage between shoe manufacturing process quality and task scheduling.

[0064] In the shoe manufacturing process, the quality of the preceding process directly affects the difficulty and time of the subsequent process. For example, if the shoe upper is not properly sanded, it will increase the time of the gluing process, and may even require secondary sanding. The existing scheduling scheme is based only on preset process parameters and does not take into account the actual quality situation, which can easily lead to congestion in the task queue and defective products flowing into the next process.

[0065] The specific implementation of this module is as follows: The module communicates with the process quality inspection unit of each modular shoemaking station. The process quality inspection unit includes an industrial vision inspection camera, a force control sensor, and a dimensional inspection device, which is installed next to the work area of ​​each station. It can perform full-dimensional quality inspection of the workpiece immediately after the process is completed. After the process is completed, the quality inspection unit immediately performs the inspection, generates a quality score of 0-100, identifies the defect type, and uploads the inspection data to this module in real time. The module executes the corresponding closed-loop scheduling strategy based on the quality level.

[0066] For qualified workpieces with a quality score ≥90, the module normally transfers them to the next process and assigns subsequent tasks according to the preset scheduling logic. For workpieces with a quality score between 80 and 89, which are qualified but low, the module immediately adds 10%-20% of the operation redundancy time to the next process of that workpiece. At the same time, when assigning tasks, it prioritizes scheduling a six-axis industrial robot arm with higher operation accuracy and higher equipment health status score to perform the subsequent process, so as to avoid secondary defects in the subsequent process due to the quality deviation of the previous process. For repairable defective workpieces with a quality score between 60 and 79, the module immediately generates... The module generates a corresponding rework task, inserts it into the highest priority task queue of the corresponding workstation, and simultaneously schedules the nearest idle six-axis industrial robotic arm to immediately execute the rework process. After rework, a quality inspection is performed again, and only after passing the inspection is the workpiece transferred to the next process. For unqualified workpieces with a quality score of <60, the module immediately issues a transfer instruction to the AGV material transfer unit to transfer the workpiece to the scrap area. At the same time, all subsequent process tasks for the workpiece are deleted from the scheduling queue, releasing the corresponding scheduling hardware resources and avoiding invalid tasks from occupying production line capacity. This achieves real-time closed-loop control of process quality and task scheduling.

[0067] The task-level dual-queue scheduling mechanism is integrated into the core dynamic task scheduling engine to solve the problem of process interruption caused by task priority conflicts, and to ensure the continuity and stability of high-precision shoe manufacturing processes.

[0068] In the shoe manufacturing process, high-precision processes such as polishing, gluing, and pressing require extremely high continuity of operations. If the process is interrupted, it can lead to problems such as glue curing, over-polishing, and uneven pressing, directly causing the workpiece to be scrapped. The existing scheduling scheme schedules process tasks and auxiliary tasks with the same priority, which can easily lead to the problem of process tasks being interrupted.

[0069] The specific implementation of this mechanism is as follows: First, the mechanism strictly divides all robotic arm tasks to be executed within the production line into two levels: Level 1 process tasks and Level 2 auxiliary tasks. Level 1 process tasks are core high-precision processes in shoemaking, including shoe upper grinding, shoe upper gluing, shoe upper and sole pressing, and shoe upper shaping. These tasks have extremely high requirements for operational continuity and trajectory accuracy, and cannot be interrupted once they start. Level 2 auxiliary tasks are low-precision auxiliary tasks, including workpiece loading and unloading, workpiece transfer, end effector replacement, and workstation cleaning. These tasks have low requirements for continuity and can be interrupted and resumed at any time.

[0070] The mechanism sets up two independent task queues for each six-axis industrial robotic arm: a primary task queue and a secondary task queue. Both queues adopt a first-in-first-out (FIFO) sorting rule, and the scheduling priority of the primary task queue is always higher than that of the secondary task queue. The mechanism's execution rules are as follows: the robotic arm's task execution order strictly follows the principle of "first-level, then second-level," prioritizing the execution of all tasks in the first-level task queue. Tasks in the second-level task queue will only be executed when the first-level task queue is empty. Once a first-level process task is assigned to the robotic arm and begins execution, it must not be interrupted by any second-level auxiliary task until the first-level task is fully completed. Second-level auxiliary tasks can only be assigned to six-axis industrial robotic arms with empty first-level task queues and a real-time load rate below 30%, avoiding consuming the robotic arm's core operating time. When the second-level task queue at a workstation becomes congested, the mechanism prioritizes scheduling idle robotic arms in the vicinity without first-level tasks, and will never interrupt a robotic arm currently executing a first-level task. For workstations with frequent loading and unloading tasks, a low-cost SCARA robotic arm can be used to exclusively execute second-level auxiliary tasks, while the six-axis industrial robotic arm focuses on executing first-level process tasks, further improving the operational efficiency and yield rate of core processes.

[0071] The complete system workflow is as follows: This embodiment illustrates the complete system workflow in a multi-shoe-style mixed-line production scenario. The production line is equipped with 8 modular shoe-making workstations, 8 six-axis industrial robotic arms, and 4 AGV material transfer units. Simultaneously, it is producing one batch order of standard athletic shoes with a priority of level 5 and one urgent order of customized leather shoes with a priority of level 1.

[0072] After system startup, the first step is to complete the unified kinematic calibration of all six-axis industrial robotic arms, establish a unified world coordinate system for the production line, initiate real-time data acquisition by the global state perception unit, and start the core dynamic task scheduling engine. Upon order integration, the order process feature pre-scheduling module analyzes the process complexity of two orders, calculating that the load factor for customized orders is 2.8 times that of standard orders. The module employs a staggered production strategy, alternately allocating tasks from the two orders to the production line, maintaining the average load rate of the robotic arms across the entire production line at approximately 75%.

[0073] During production, the robotic arm at grinding station No. 3 triggered a fault warning due to motor overheating. The core dynamic task scheduling engine immediately triggered the fault takeover process, sorting the pending tasks at the station according to priority, and prioritizing the high-priority tasks of customized orders to the adjacent idle robotic arms No. 2 and No. 4, achieving fault takeover without interruption and preventing production line shutdown.

[0074] During the gluing process, the queue length of tasks waiting to be processed at gluing station 2 reached 4, exceeding the congestion threshold, and the corresponding robotic arm's load rate consistently exceeded 90%. The engine triggered cross-station scheduling, detecting that the robotic arm at pressing station 5 was idle with a load rate of only 22%, and its working range could cover gluing station 2. A cross-station operation instruction was then issued. Upon receiving the instruction, robotic arm 5 completed a quick change of the gluing actuator, entered station 2, recalibrated its coordinate system, read the shoe last RFID data to correct the gluing trajectory, and successfully completed the gluing operation. After completing the operation, it returned to its initial station. The entire process did not affect the production line's cycle time.

[0075] After the adhesive coating process is completed, the process timing coupling scheduling sub-engine calculates the effective time window for the baking process based on the process timing constraint model. It issues a pre-scheduling instruction to the corresponding robotic arm 10 seconds in advance to ensure that the workpiece completes baking within the effective time window, preventing any workpiece from being scrapped. After baking, the process quality inspection unit detects that the baking temperature of a certain workpiece is too low, which is a repairable defect. The quality-scheduling closed-loop feedback module immediately generates a rework task for secondary baking, inserts it into the highest priority scheduling, and schedules an idle robotic arm to complete the rework, preventing defective products from flowing into the pressing process.

[0076] Throughout the entire production process, the task-level dual-queue scheduling mechanism strictly ensures the absolute priority of the primary process tasks. All gluing, grinding, and pressing processes are not interrupted by auxiliary tasks. The yield rate of the production line is stable at 99.6%, the comprehensive utilization rate of the robotic arm reaches 86%, and the delivery cycle of customized orders is shortened by 42% compared with the existing technology, fully realizing the expected technical effect of this invention.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic task scheduling system for robotic arms in a flexible shoe manufacturing production line, wherein the flexible shoe manufacturing production line includes multiple modular shoe manufacturing stations arranged along the material flow path, multiple six-axis industrial robotic arms, and AGV material transfer units for transferring shoe lasts, wherein the initial working range of each six-axis industrial robotic arm covers at least one of the modular shoe manufacturing stations, and each six-axis industrial robotic arm is used to perform shoe manufacturing processes, characterized in that... The system includes: a) Process-integrated hardware unit, including a standardized quick-change connector, multiple sets of shoe-making process-specific end effectors, and a workstation quick-change table. The quick-change connector is installed at the end of each of the six-axis industrial robotic arms. The shoe-making process-specific end effectors are detachably installed on the workstation quick-change table. The workstation quick-change table is arranged next to each modular shoe-making workstation. All the six-axis industrial robotic arms have completed standardized kinematic parameter calibration and have the ability to perform all shoe-making processes. b) The global state perception unit includes a global vision positioning module deployed on the top of the flexible shoe production line, a workstation calibration component installed at each modular shoe production station, and an RFID storage chip built into the shoe last workpiece. The global vision positioning module is used to collect real-time data on the position of the shoe last workpiece, the posture of the robotic arm, and the position of the end effector throughout the entire production line. The workstation calibration component is used to provide a reference for the workstation's coordinate system. The RFID storage chip is used to store the process parameters and positioning reference data for the corresponding shoe model. c) A core dynamic task scheduling engine, which is communicatively connected to the hardware unit adapted to the aforementioned process, the global state perception unit, the six-axis industrial robotic arm, and the AGV material transfer unit, is configured as follows: i) Real-time data collection and updates of the load rate, remaining working time, and equipment health status of each six-axis industrial robotic arm, as well as the queue length of the tasks to be processed and the priority data of the corresponding orders for each modular shoemaking station. ii) When the length of the task queue to be processed at the target modular shoemaking station exceeds the preset congestion threshold and the real-time load rate of the six-axis industrial robot arm at the corresponding station continues to exceed the preset high load threshold, the system automatically searches for idle six-axis industrial robots whose work range can cover the target modular shoemaking station and whose real-time load rate is lower than the preset low load threshold, and generates cross-station work scheduling instructions. iii) Issue the cross-station operation scheduling command to the idle six-axis industrial robotic arm, control it to replace the end effector of the corresponding shoemaking process through the quick-change connector, enter the target modular shoemaking station to complete the corresponding process operation, and return to the initial station to wait after the operation is completed; iv) When any six-axis industrial robotic arm is detected to have a malfunction or health status warning, all pending tasks at the corresponding workstation of that robotic arm will be dynamically assigned to any available six-axis industrial robotic arm whose work range can cover that workstation, thus enabling uninterrupted takeover of operations during malfunctions.

2. The system according to claim 1, characterized in that, The core dynamic task scheduling engine is also equipped with a shoe last reference dynamic calibration module. The shoe last reference dynamic calibration module is configured to: when a six-axis industrial robot arm that receives a cross-workstation operation scheduling instruction enters the target modular shoe manufacturing workstation, it first scans the workstation calibration component of the workstation through the global vision positioning module to complete the real-time recalibration of the operation coordinate system, and then reads the process parameters and positioning reference data in the RFID storage chip of the target shoe last workpiece to automatically correct the operation trajectory and ensure that the process accuracy of the cross-workstation operation is consistent with the six-axis industrial robot arm initially configured for the workstation.

3. The system according to claim 1 or 2, characterized in that, The core dynamic task scheduling engine is also equipped with a spatiotemporal segmentation obstacle avoidance scheduling module. The spatiotemporal segmentation obstacle avoidance scheduling module is configured to: pre-divide the working space of the flexible shoe production line into several grid units according to modular shoe workstations, and divide the time axis into several time slices according to a preset unit duration; when issuing cross-workstation operation scheduling instructions, pre-plan a non-interference motion path for the corresponding six-axis industrial robotic arm, and allocate a dedicated spatial grid unit and time slice to it, ensuring that only one working device is allowed to occupy the same spatial grid unit within the same time slice; at the same time, monitor the movement posture of the equipment in real time through the global vision positioning module, and immediately allocate backup paths and time slices for low-priority tasks when interference risks are predicted, so as to achieve zero-collision obstacle avoidance.

4. The system according to claim 1, characterized in that, The core dynamic task scheduling engine is also equipped with a process timing coupling scheduling sub-engine. The process timing coupling scheduling sub-engine has a pre-stored process timing constraint model corresponding to each shoe model. The process timing constraint model is marked with the preceding / following process relationship, effective time window of the process, and the shortest and longest operation time of each process of the corresponding shoe model. The process timing coupling scheduling sub-engine is configured to: when allocating process tasks, predict the completion time of the preceding process based on the full process timing chain, issue a pre-scheduling instruction in advance, control the corresponding six-axis industrial robotic arm to complete the end effector replacement and coordinate system calibration in advance, and immediately enter the workstation to perform the operation the moment the preceding process is completed; when the operation time of a certain process exceeds the preset threshold, automatically adjust the task allocation priority and scheduling sequence of the following process to ensure that all processes are completed within the effective process time window.

5. The system according to claim 1, characterized in that, The core dynamic task scheduling engine is also equipped with an order process feature pre-scheduling module. This module is configured to: automatically parse the BOM files and process files of all shoe styles within the order when a production order is received, extract process complexity features, and generate the corresponding load coefficient for the order; based on the rated load capacity of the six-axis industrial robotic arms of the entire production line, split the order tasks into multiple task packages, pre-allocate the corresponding six-axis industrial robotic arm clusters to each task package, and reserve the corresponding operation time window and hardware resources; during mixed-line production, dynamically adjust the task deployment sequence based on order priority and load coefficient, stagger the deployment of high-load customized orders and low-load standard orders, so that the real-time load rate of the six-axis industrial robotic arms of the entire production line is maintained within the preset optimal range.

6. The system according to claim 1, characterized in that, The core dynamic task scheduling engine is also equipped with a quality-scheduling closed-loop feedback module. This module is communicatively connected to the process quality inspection unit of each modular shoemaking station. The quality-scheduling closed-loop feedback module is configured to: receive workpiece operation quality scores and defect type data uploaded by the process quality inspection unit; for workpieces that are qualified but have quality scores below a preset threshold, allocate redundant operation time to their subsequent processes and prioritize scheduling six-axis industrial robotic arms whose operation accuracy and stability meet preset requirements; for repairable defective workpieces, insert the rework task into the highest priority of the corresponding station's task queue and schedule the nearest available six-axis industrial robotic arm to perform the rework; for non-conforming products, immediately schedule AGV material transfer units to transfer them to the waste area and delete all subsequent process scheduling tasks for that workpiece, releasing scheduling resources.

7. The system according to claim 1, characterized in that, The core dynamic task scheduling engine is also equipped with a hierarchical dual-queue scheduling mechanism. This mechanism divides all tasks to be executed into primary process tasks and secondary auxiliary tasks. Primary process tasks are high-precision processes such as shoemaking, grinding, gluing, pressing, and sole attaching. Secondary auxiliary tasks are low-precision auxiliary tasks such as workpiece loading, unloading, and transfer. The hierarchical dual-queue scheduling mechanism is configured to set up independent primary and secondary task queues for each six-axis industrial robot arm, strictly following the order of primary tasks first, then secondary tasks. Primary process tasks must not be interrupted by secondary auxiliary tasks during execution. Secondary auxiliary tasks can only be assigned to six-axis industrial robots that have no primary process tasks to be executed and whose real-time load rate is below a preset low load threshold.

8. The system according to claim 1, characterized in that, The dedicated end effector for the shoe manufacturing process includes a shoe upper grinding actuator, a shoe upper gluing actuator, a shoe upper and sole pressing actuator, and a shoe last clamping actuator. All end effectors are equipped with a unified standard docking interface that matches the quick-change connector, enabling rapid adaptation and replacement of any six-axis industrial robotic arm.

9. The system according to claim 1, characterized in that, When performing fault-free takeover, the core dynamic task scheduling engine prioritizes assigning tasks to idle six-axis industrial robotic arms that are adjacent to the faulty robotic arm and whose work area completely covers the faulty workstation. At the same time, based on the order priority and process timing constraints of the tasks, it adjusts the task execution order to ensure that high-priority orders and tasks with adjacent effective process time windows are executed first.

Images