Automatic intelligent system for stacker control
By constructing a stacker crane control system with state perception, digital twin and simulation deduction, multi-objective decision-making and distributed execution modules, the problems of insufficient global collaborative optimization and intelligent decision-making in existing stacker crane control technologies are solved, and efficient and reliable stacker crane control is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-04-10
AI Technical Summary
Existing stacker crane control systems lack global collaborative optimization, have poor dynamic adaptability, and insufficient intelligent decision-making capabilities. They are unable to achieve predictive maintenance, struggle to adapt to multi-task concurrent execution and path collision avoidance in complex warehousing environments, and have limited ability to perceive equipment status.
A state awareness and data acquisition module, a digital twin and simulation module, a multi-objective collaborative decision-making module, and a distributed execution control module are constructed to achieve global collaborative optimization of the stacker crane. The state awareness module collects data in real time, the digital twin module performs simulation and prediction, the multi-objective decision-making module generates the globally optimal control strategy, and the distributed execution module coordinates the actions of the stacker crane.
This has enabled the stacker crane control to shift from local independent control to global collaborative optimization, improving the system's intelligence level and equipment health status monitoring capabilities. It can promptly detect potential faults, adapt to different warehousing needs, and improve operational efficiency and equipment reliability.
Smart Images
Figure CN121832333A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of stacker crane control technology, and in particular to an automated intelligent system for stacker crane control. Background Technology
[0002] Automated warehousing and logistics systems are a core component of modern industry and supply chain management. They achieve efficient storage, picking, and handling of goods through automated equipment, significantly improving the accuracy and throughput of warehousing operations. Stacker cranes, as key operating equipment in automated storage and retrieval systems (AS / RS), undertake the core task of vertical and horizontal movement within the aisles to access and store goods.
[0003] The control strategy of the stacker crane directly determines the operational performance and responsiveness of the entire warehousing system. Traditional stacker crane control systems typically rely on preset fixed logic or scheduling algorithms based on simple rules, aiming to achieve motion control of a single device or a local area while satisfying basic access commands.
[0004] Existing technologies generally employ relatively independent control architectures, with stacker crane movement, lifting, and fork actions often managed by separate controllers, lacking a holistic perspective for collaborative optimization. This control approach struggles to meet the complex demands of large-scale, highly dynamic warehousing environments, such as concurrent multi-task execution, real-time collision avoidance, and optimal energy efficiency control. Furthermore, traditional control systems have limited ability to perceive the stacker crane's operating status, failing to perform real-time analysis and predictive maintenance of key parameters like motor load and mechanical vibration, leading to increased potential equipment failure risks and impacting long-term operational stability. Faced with the growing demands for intelligent and flexible warehousing, existing stacker crane control methods exhibit significant shortcomings in dynamic adaptability, collaborative efficiency, and intelligent decision-making capabilities. Summary of the Invention
[0005] The purpose of this invention is to provide an automated intelligent system for stacker crane control, in order to solve the problems of lack of global collaborative optimization, poor dynamic adaptability, insufficient intelligent decision-making level, and difficulty in achieving predictive maintenance in the existing stacker crane control technology.
[0006] To achieve the above objectives, the present invention provides the following technical solution: an automated intelligent system for stacker crane control, comprising a state perception and data acquisition module, a digital twin and simulation deduction module, a multi-objective collaborative decision-making module, and a distributed execution control module;
[0007] The system comprises several modules: a status perception and data acquisition module that collects real-time data on the stacker crane's operating status, warehousing environment, and upper-level scheduling instructions; a digital twin and simulation module that builds and maintains a virtual model synchronized with the physical stacker crane and warehousing environment based on the collected data, and performs offline simulation and performance prediction of control strategies; a multi-objective collaborative decision-making module that receives prediction information from the digital twin module and comprehensively considers multiple optimization objectives such as efficiency, energy consumption, and equipment health to generate a globally optimal control decision sequence; and a distributed execution control module that parses the control sequence output by the decision module into underlying drive instructions and coordinates the collaborative actions of the stacker crane's traveling mechanism, lifting mechanism, and fork mechanism.
[0008] Furthermore, the state perception and data acquisition module includes a sensor array mounted on the stacker crane body and a warehouse environment perception unit. The sensor array includes a first laser rangefinder for detecting the horizontal position of the stacker crane in the aisle, a second laser rangefinder for detecting the vertical height of the loading platform, a first current sensor for detecting the current of the travel drive motor, a second current sensor for detecting the current of the lifting drive motor, a third current sensor for detecting the current of the forklift motor, a first vibration sensor mounted on the travel wheel bearing housing, a second vibration sensor mounted on the lifting mechanism connection point, and an infrared temperature sensor for monitoring the temperature of electrical nodes. The warehouse environment perception unit includes a visual recognition device deployed at the aisle entrance for detecting the priority markers of tasks to be executed and a 3D LiDAR deployed in the aisle for detecting potential moving obstacles. The state perception and data acquisition module uploads all the collected sensor data and recognition results to the digital twin and simulation module in real time via industrial Ethernet.
[0009] Furthermore, the digital twin and simulation module internally constructs geometric, kinematic, dynamic, and energy consumption models corresponding to the physical entity. The digital twin and simulation module first receives real-time data from the state perception and data acquisition module and drives the virtual model to maintain state synchronization with the physical stacker crane. The digital twin and simulation module includes a strategy simulation engine, which performs advanced simulation in the virtual environment based on the current system state and candidate control strategies provided by the multi-objective collaborative decision-making module. The simulation process simulates the stacker crane starting from its current position, fully executing the action sequence defined by the candidate control strategies, and simultaneously calculating multiple performance indicators during the simulation process. These performance indicators include at least the estimated task completion time, estimated energy consumption, and estimated stress loads on mechanical components. All simulation results and performance indicators are encapsulated and sent to the multi-objective collaborative decision-making module for decision analysis.
[0010] Furthermore, the core of the multi-objective collaborative decision-making module is a multi-objective optimization function based on weighted summation. The input variable of the multi-objective optimization function is the candidate control strategy, and the output is the comprehensive evaluation value of the strategy. The multi-objective optimization function is composed of a linearly weighted sub-function of efficiency objective, energy consumption objective, and equipment health objective. The efficiency objective sub-function optimizes by minimizing the total task execution time, and its calculation relies on the estimated task completion time provided by the digital twin module. The energy consumption objective sub-function optimizes by minimizing the total electrical energy used by the stacker crane during task execution, and its calculation integrates the estimated energy consumption of walking, lifting, and fork movements provided by the digital twin module. The equipment health objective sub-function optimizes by minimizing the cumulative fatigue damage of mechanical components, and its calculation is based on the estimated stress load provided by the digital twin module and combined with the stress life characteristics of the material to estimate the degree of damage.
[0011] Furthermore, the multi-objective collaborative decision-making module has a set of weight coefficients pre-set, which correspond to the relative importance of the three objectives of efficiency, energy consumption and equipment health. The decision-making process of the multi-objective collaborative decision-making module is as follows: receiving a task set from the upper-level system, generating multiple control strategy sequences, using the digital twin module to simulate and deduce each strategy and obtain its performance index, and then selecting the control strategy sequence that makes the comprehensive evaluation value the highest by calculating the value of the multi-objective optimization function as the final decision output.
[0012] Furthermore, the distributed execution control module includes a trajectory planning unit, a motion control unit, and an execution status monitoring unit. The trajectory planning unit is responsible for decomposing the target position and action commands in the decision sequence into the motion trajectories of the stacker crane's traveling axis, lifting axis, and fork axis. This trajectory planning ensures that the movement of each axis is impact-free while meeting the hardware limitations of maximum speed and acceleration. The motion control unit generates corresponding pulse commands or analog signals based on the planned trajectory to drive the traveling servo driver, lifting servo driver, and fork servo driver, respectively. The execution status monitoring unit continuously receives real-time feedback signals from the underlying sensors, including the actual position, speed, and drive current of each axis, and compares these actual values with the command values of the motion control unit. When the deviation exceeds the preset safety threshold, the monitoring unit immediately triggers the abnormal handling program and sends an alarm message to the upper layer of the system, while taking degraded operation or safe shutdown measures according to the preset strategy.
[0013] Furthermore, it also includes a health prediction and maintenance reminder submodule; the health prediction and maintenance reminder submodule continuously acquires vibration spectrum data from the first vibration sensor and the second vibration sensor from the status perception module, and extracts their characteristic frequency amplitudes; the health prediction and maintenance reminder submodule has a built-in fault recognition model based on a deep convolutional neural network. This model takes the vibration spectrum as input, extracts deep features through multi-layer convolution and pooling operations, and finally outputs the health status score and potential fault mode probability of the stacker crane's rotating components, including the traveling wheel bearings and the lifting wire rope drum; when the health status score is lower than a preset threshold or the probability of a specific fault mode exceeds a set threshold, the health prediction and maintenance reminder submodule automatically generates a pre-maintenance work order and pushes it to the warehouse management system.
[0014] Furthermore, the weighting coefficients are designed to be dynamically adjustable. The system introduces an operating mode switching logic, which can receive external instructions from the warehouse management system or automatically switch the system's optimization focus mode according to a preset time strategy. The system predefines at least three operating modes: peak efficiency mode, balanced operation mode, and energy-saving maintenance mode. In peak efficiency mode, the weight of the efficiency objective sub-function is set to the highest value, while the weights of energy consumption and equipment health are reduced accordingly. In balanced operation mode, the weights of the three objectives are set to equal or empirically calibrated balance values. In energy-saving maintenance mode, the weights of the energy consumption objective sub-function and the equipment health objective sub-function are increased, while the efficiency weight is reduced accordingly. The dynamic adjustment of the weighting coefficients is achieved by modifying the corresponding coefficient values in the multi-objective optimization function.
[0015] Furthermore, the simulation process of the digital twin and simulation deduction module introduces a precision adaptive mechanism; the precision adaptive mechanism dynamically adjusts the complexity of the simulation model and the deduction step size according to the criticality of the task to be decided and the real-time computing load of the system; the precision adaptive mechanism is managed by a resource scheduler, which monitors the utilization of the central processing unit in real time and selects simulation configuration parameters according to a preset rule base.
[0016] Furthermore, the motion control unit adopts a cross-coupling control algorithm to achieve synchronization between the walking mechanism and the lifting mechanism. The cross-coupling control algorithm calculates the position error between the walking axis and the lifting axis in real time. It not only performs independent proportional, integral, and derivative adjustments on each axis, but also introduces a compensation term based on the difference in position error between the two axes. After being adjusted by a coupling gain coefficient, the compensation term is added to the control output of the walking axis and the lifting axis respectively, thereby suppressing the following error during the dual-axis motion process.
[0017] Furthermore, the calculation of the equipment health objective sub-function involves material fatigue analysis; the fatigue damage degree of the component is estimated through the following process: obtaining the number of stress cycles experienced at a specific stress level, which is statistically obtained by the digital twin module in the simulation; querying the pre-stored standard material stress life characteristic database to obtain the number of cycles required for the material to fail at that stress level; calculating the cumulative fatigue damage degree based on the ratio of the number of stress cycles to the number of cycles required for failure; the value of the equipment health objective sub-function is inversely proportional to the total damage degree.
[0018] Compared with the prior art, the beneficial effects of the present invention are:
[0019] 1. This invention achieves a fundamental shift in stacker crane control from local independent control to global collaborative optimization by constructing a complete closed-loop system that integrates state perception, digital twin simulation, multi-objective decision-making, and distributed execution. The introduction of the digital twin module enables the control strategy to be fully verified and its performance predicted in virtual space. Combined with multi-objective optimization decision-making, the system can seek the global optimal solution in multiple dimensions such as efficiency, energy consumption, and equipment health, significantly improving the overall performance and intelligence level of stacker crane operations.
[0020] 2. This invention integrates predictive maintenance functions based on vibration analysis and deep learning models, which can monitor the health status of key mechanical components of the stacker crane in real time and identify early faults. This forward-looking maintenance strategy can promptly detect potential faults, guide preventive maintenance, effectively avoid unplanned downtime, extend equipment service life, and improve the operational reliability and availability of the entire warehousing system.
[0021] 3. This invention designs dynamically adjustable weighting coefficients and operating modes, enabling the system to flexibly adapt to different warehousing operation needs and operational strategies. Whether during peak business periods requiring high throughput or during regular or off-peak periods where cost and equipment protection are emphasized, the system can automatically adjust the control focus by switching optimization modes, thereby enhancing the system's environmental adaptability and economy.
[0022] 4. This invention employs a cross-coupling control algorithm at the control execution level, which effectively improves the accuracy and stability of the multi-axis coordinated motion of the stacker crane. By compensating for the inter-axis following error, it reduces overshoot and oscillation during the positioning process, thereby not only improving the accuracy of a single storage and retrieval operation, but also reducing component wear caused by mechanical impact, further supporting the achievement of equipment health goals.
[0023] 5. This invention cleverly balances the contradiction between the accuracy of simulation and the real-time nature of decision response through the precision adaptive mechanism of the digital twin module; the system can intelligently adjust the simulation granularity according to the urgency of the task and the available computing resources, ensuring that high-quality control decisions can still be generated in a timely manner under complex working conditions, thus guaranteeing the stable and efficient operation of the system in a highly dynamic warehousing environment. Attached Figure Description
[0024] Figure 1 This is a schematic diagram of the overall technical architecture of the automated intelligent system for stacker crane control proposed in this invention. Detailed Implementation
[0025] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.
[0026] Example 1
[0027] like Figure 1 As shown, an automated intelligent system for stacker crane control consists of four core components: a state perception and data acquisition module, a digital twin and simulation module, a multi-objective collaborative decision-making module, and a distributed execution control module. The modules interact with each other through an industrial Ethernet network to form a complete closed-loop control system from data perception to intelligent decision-making to precise execution.
[0028] The status awareness and data acquisition module serves as the system's data source, with its hardware deployment encompassing both the stacker crane itself and the storage environment. A high-precision sensor array is installed on the stacker crane itself. Specifically, a first laser rangefinder sensor is rigidly mounted on the side of the stacker crane's traveling mechanism near the track. This sensor continuously emits a laser beam towards a reflector at the end of the aisle, accurately calculating the stacker crane's absolute horizontal position within the aisle by measuring the laser's round-trip time difference. This position data is transmitted in real-time via a fieldbus. A second laser rangefinder sensor is installed on the fixed support of the stacker crane's lifting mechanism's loading platform. This sensor is vertically aligned with a fixed reflector at the top, used to accurately detect the vertical height of the loading platform. To monitor the load and energy efficiency of the drive system, a first current sensor is connected in series on the power input line of the traveling drive motor, a second current sensor is connected in series on the power input line of the lifting drive motor, and a third current sensor is connected in series on the power input line of the fork extension motor. These current sensors all utilize the Hall effect principle, enabling them to capture the current waveform during motor operation in real time. To assess the health of the mechanical structure, a first vibration sensor, a triaxial accelerometer, is installed at the drive wheel bearing housing of the traveling mechanism, capable of collecting vibration acceleration signals in three directions. A second vibration sensor, also a triaxial accelerometer, is installed at the critical hinge points of the lifting mechanism or at the wire rope guide wheel bearing housing. Furthermore, several non-contact infrared temperature sensors are deployed on the heat sink surfaces of key power components in the traveling drive, lifting drive, and main control cabinet to monitor the thermal state of the electrical system. The warehouse environment sensing unit includes a visual recognition device deployed on a pillar at the aisle entrance. This device comprises a high-resolution industrial camera and a matching ring light source. Its lens is aimed at the area that inbound or outbound pallets must pass through to capture barcodes or QR codes affixed to the pallets, and the built-in image processing chip parses the task priority information. Simultaneously, a 3D LiDAR is installed on the top or front crossbeam of the stacker crane. This LiDAR scans a fan-shaped area in front of and to the sides at a specific frequency, using point cloud data processing algorithms to detect in real time whether there are unexpected obstacles in the aisle, such as fallen goods or personnel. The state perception and data acquisition module integrates a data acquisition and preprocessing subunit. This subunit is responsible for filtering, denoising, scaling, and formatting all raw sensor signals. It also packages all processed sensor data and visual recognition results into standard data frames through an industrial Ethernet switch and uploads them to the data receiving interface of the digital twin and simulation module in real time at a frequency of one hundred times per second.
[0029] The digital twin and simulation module runs on the system's high-performance industrial server. This module first maintains a high-fidelity virtual model that corresponds one-to-one with the physical stacker crane and the warehousing environment. This virtual model contains four sub-models: a geometric model that accurately describes the stacker crane's three-dimensional dimensions, rack structure, and aisle layout; a kinematic model that defines the motion relationships and constraints of the stacker crane's three degrees of freedom: walking, lifting, and forks; a dynamic model that includes parameters such as the mass, inertia, friction coefficient, and transmission stiffness of each component to simulate the forces and deformations during actual movement; and an energy consumption model that constructs a mathematical model of the system's energy flow based on motor characteristic curves, transmission efficiency, and standby power consumption. After the module starts, its data synchronization engine continuously receives real-time data streams from the state perception and data acquisition module, including the stacker crane's horizontal position, loading platform height, motor currents, vibration signals, temperature readings, and environmental perception results. The data synchronization engine uses this real-time data to drive and update the corresponding state variables in the virtual model, ensuring that the virtual model and the physical entity maintain a high degree of synchronization in key states such as position, speed, and load, with synchronization errors controlled within millimeter and millisecond levels. The core functional component of the digital twin and simulation module is the strategy simulation engine. The strategy simulation engine is activated when the multi-objective collaborative decision-making module provides a set of candidate control strategies. The simulation begins with the current state of the synchronized virtual model. The simulation process mimics the stacker crane strictly following the instruction sequence defined by the candidate control strategies, such as accelerating from its current position to the target column, synchronously lifting the loading platform to the target layer, and then extending the forks for storage and retrieval operations. During the simulation, the engine calls the dynamic model to calculate the force and stress distribution of each component, especially performing dynamic stress load analysis on mechanical components such as the traveling axles, lifting wire ropes, and key bearings, recording their stress time history. Simultaneously, the engine calls the energy consumption model to integrally calculate the total electrical energy consumed by the traveling drive motor, lifting drive motor, and fork motor throughout the entire simulation task. The engine contains a high-precision clock to track the virtual time required from the start of the task until the forks are fully retracted and the loading platform returns to its standby position, serving as the estimated task completion time. All calculated performance metrics, including estimated task completion time, estimated energy consumption, maximum stress of key mechanical components, and number of stress cycles, are encapsulated in a structured data packet. This data packet is sent in real-time to the performance metric input buffer of the multi-objective collaborative decision-making module via an internal communication interface.
[0030] The multi-objective collaborative decision-making module is the core of the system's intelligent decision-making. The core algorithm of this module is based on a weighted summation multi-objective optimization function. The input to this function is a candidate control strategy, and the output is the comprehensive evaluation value of that strategy. Specifically, the optimization function consists of three linearly weighted objective sub-functions. The efficiency objective sub-function aims to minimize the total task execution time, and its calculation directly relies on the task's estimated completion time provided by the digital twin module. The energy consumption objective sub-function aims to minimize the total electrical energy used by the stacker crane during task execution, and its calculation integrates the estimated energy consumption of walking, lifting, and fork movements provided by the digital twin module. The equipment health objective sub-function aims to minimize the cumulative fatigue damage of key mechanical components, and its calculation is based on the estimated stress load data provided by the digital twin module. The calculation of the equipment health objective sub-function involves material fatigue analysis, and its principle is based on the stress-life characteristics of materials. For key components, their fatigue damage degree D can be estimated using the following formula: Where, n i Indicates at stress level S i The number of stress cycles experienced, which was obtained by the digital twin module during simulation; N i Indicates the material at stress level S i The number of cycles required for damage to occur is determined by querying a pre-stored database of standard material SN curves. The value of the equipment health objective sub-function is typically inversely proportional to the total damage degree D. The multi-objective collaborative decision-making module has a pre-set set of weighting coefficients, namely the efficiency weight W. t Energy consumption weight W e Equipment health weight W h The sum of these three weighting coefficients is 1, and their initial values are set by the system administrator according to the warehouse operation strategy. The decision-making process begins with receiving a set of tasks from the upper-level warehouse management system. The task parser of the decision module first analyzes the task requirements, including the target storage location coordinates, task type, and priority identifier. Subsequently, the strategy generator generates multiple feasible control strategy sequences based on the current stacker crane status and task constraints. These candidate strategy sequences differ in acceleration, velocity curves, and action sequence. Each candidate strategy sequence is immediately sent to the digital twin and simulation module for simulation. After completing the simulation, the digital twin module returns the corresponding performance index package. The evaluation calculator of the decision module then uses these performance indices to calculate the efficiency objective function value F for each candidate strategy. t Energy consumption objective function value F e Equipment health objective sub-function value F h Next, the overall evaluation value V of the candidate strategy is calculated: V = W t *F t +W e *F e +Wh *F h The evaluation calculator iterates through all candidate strategies and calculates their respective comprehensive evaluation values V. Finally, the decision selector selects the control strategy sequence that yields the highest comprehensive evaluation value V as the final decision output. This optimal control decision sequence is then sent to the instruction receiving queue of the distributed execution control module through the decision output interface.
[0031] The distributed execution control module is deployed within the programmable logic controller (PLC) and motion controller of the stacker crane body. This module receives the optimal control decision sequence from the multi-objective collaborative decision-making module. The module contains three functional units. The trajectory planning unit is responsible for decomposing the abstract target position and action commands in the decision sequence into specific, continuous, and time-parameterized motion trajectories for the stacker crane's traveling axis, lifting axis, and fork axis. The trajectory planning employs an S-curve acceleration / deceleration algorithm to ensure that the speed, acceleration, and jerk of each axis change continuously, thus avoiding rigid impacts, while strictly constraining the maximum operating speed and acceleration of each axis to not exceed the physical limits of its mechanical structure and actuators. The planned trajectory includes the target position, target speed, and target acceleration of each axis within each control cycle. The motion control unit then runs the control algorithm at a fixed control cycle based on the target trajectory output by the trajectory planning unit. For the traveling axis, lifting axis, and fork axis, the motion control unit generates corresponding pulse sequences or analog voltage signals, which are output to the traveling servo driver, lifting servo driver, and fork servo driver, respectively, driving the corresponding servo motors to accurately track the target trajectory. The execution status monitoring unit operates continuously. It receives actual position feedback from the underlying encoder via a high-speed input channel and reads actual current feedback from the motor driver via an analog input module. An internal comparator compares the actual position of each axis with the commanded position issued by the motion control unit in real time to calculate the position tracking error; simultaneously, it compares the actual current with the commanded current. Preset safety thresholds are set for both position and current errors. If the real-time position tracking error of any axis continuously exceeds the threshold for a certain period, or if the current abnormally exceeds the threshold, the monitoring unit will immediately trigger the anomaly handling procedure. The anomaly handling procedure first sends an alarm message containing the error code and level to the upper-level system through the alarm output interface; simultaneously, according to the preset safety strategy, it may instruct the motion control unit to perform a smooth deceleration stop, or switch to a low-speed degraded operation mode, until manual intervention is required.
[0032] As an advanced function of the system, the vibration data collected by the first and second vibration sensors in the state perception and data acquisition module are used not only for the synchronization of the digital twin model but also specifically for the online assessment and prediction of equipment health status. The system adds an independent health prediction and maintenance reminder submodule, which runs on the data analysis server. This submodule's data access point continuously acquires the raw vibration acceleration waveform data from the first and second vibration sensors from the state perception module. The data preprocessing stage first performs bandpass filtering on the waveform data to remove high-frequency noise and extremely low-frequency interference, and then converts the time-domain signal into a frequency-domain spectrum using a fast Fourier transform. The feature extractor extracts the amplitude of several predefined characteristic frequencies from the spectrum; these characteristic frequencies correspond to the natural frequencies or fault characteristic frequencies of key rotating components such as the traveling wheel bearing and the lifting drum bearing. The core of the health prediction and maintenance reminder submodule is a pre-trained deep convolutional neural network model. The input layer of this model receives standardized vibration spectrum data. The network structure consists of multiple alternating convolutional and pooling layers. The convolutional layers use small kernels to extract local feature patterns from the spectrum, while the pooling layers perform feature dimensionality reduction and enhance translation invariance. After multiple nonlinear transformations, the fully connected layers map the learned high-level features to the output layer. The output layer uses the Softmax activation function to output a health status score and the probability of occurrence of several specific fault modes, including bearing outer race damage, inner race damage, rolling element damage, and drum imbalance. The health status score is a value between zero and one, with a value closer to one indicating a better health status. This submodule includes a decision logic unit that continuously monitors the health status score and the probability of each fault mode. When the health status score falls below a preset threshold of 0.85, or the probability of any fault mode exceeds a set threshold of 0.7, the decision logic unit automatically triggers the work order generator. The work order generator creates a structured pre-maintenance work order, which includes equipment identification, suspected faulty parts, severity level, recommended inspection measures, and timestamp. The work order is then pushed to the maintenance management interface of the warehouse management system in real time through the system-integrated message middleware, reminding maintenance personnel to arrange preventive maintenance.
[0033] The weighting coefficients in the multi-objective collaborative decision-making module are designed to be dynamically adjustable to adapt to different operational needs. The system introduces an operating mode switching logic. This logic can receive manual instructions from the warehouse management system or automatically switch the system's optimization focus mode based on preset time-based automated scripts. The system predefines at least three operating modes. The peak efficiency mode is suitable for peak inbound and outbound business periods; in this mode, the operating mode switching logic will adjust the weight W of the efficiency objective sub-function. t The weight W of the energy consumption objective sub-function is set to 0.6. eThe weight W of the device health objective sub-function is set to 0.2. h Set to 0.2. This weighting configuration allows the system to prioritize the strategy that completes the task fastest during decision-making, slightly sacrificing energy consumption and equipment wear. The balanced operation mode is suitable for regular operating periods, where the weights of the three objectives are set equal, i.e., W. t W is 0.33. e W is 0.33. h It can be 0.33, or calibrated to a set of balanced values based on historical operating data, such as W. t W is 0.35. e W is 0.3. h The weight is 0.35. The energy-saving maintenance mode is suitable for nighttime or off-peak business periods, or when equipment health monitoring indicators indicate a need for special attention. In this case, the weights of the energy consumption target sub-function and the equipment health target sub-function are significantly increased, for example, by setting W... e W is 0.4. h W is 0.4. t The weighting is set to 0.2, prioritizing energy consumption reduction and equipment wear mitigation, with a slight relaxation of efficiency requirements. Dynamic adjustment of the weighting coefficients is achieved by directly modifying the corresponding coefficient values in the multi-objective optimization function. After the mode switching command takes effect, all subsequent decisions will be calculated based on the new weighting combination.
[0034] The simulation process of the digital twin and simulation derivation module incorporates an accuracy-adaptive mechanism to balance computational accuracy and response speed. This mechanism is managed by a resource scheduler. The resource scheduler monitors the CPU utilization of the running server in real time. Simultaneously, it receives task criticality indicators from the multi-objective collaborative decision-making module. Internally, the resource scheduler maintains a rule base that defines the simulation configuration parameters to be used in different scenarios. These parameters primarily include model complexity and simulation step size. Model complexity is divided into two levels: high-fidelity and simplified. The high-fidelity model uses complete dynamic equations, considering elastic deformation and nonlinear friction; the simplified model uses rigid body kinematic equations, ignoring some dynamic effects. The simulation step size is divided into small and large step sizes, with a small step size of 0.01 seconds and a large step size of 0.1 seconds. When the task to be decided is marked as a high-priority urgent task, and the resource scheduler detects that the CPU utilization is below 70%, the rule base instructs the use of a high-fidelity dynamic model with a small simulation step size of 0.01 seconds for accurate derivation to obtain the most reliable performance prediction. When a task has a normal priority and the CPU utilization is above 70%, the rule base instructs the use of a simplified kinematic model and a large simulation step size of 0.1 seconds for rapid inference. Although the accuracy is slightly reduced, this significantly shortens the simulation computation time and ensures timely decision-making. Before each simulation task begins, the resource scheduler selects the most suitable set of simulation configuration parameters from the rule base based on the current system state and task attributes, and passes it to the policy simulation engine. The policy simulation engine then dynamically loads the corresponding model and sets the integration step size based on the received configuration parameters, thereby completing the adaptive simulation inference process.
[0035] The motion control unit of the distributed execution control module employs a cross-coupling control algorithm to improve synchronization accuracy when controlling the coordinated movement of the traveling and lifting mechanisms. This algorithm runs in real-time within each control cycle of the motion controller. The algorithm first independently calculates the position error E of the traveling axis. x Position error E with lifting shaft y Position error is defined as the difference between the commanded position and the actual feedback position. Traditional independent proportional-integral-derivative (PID) control separately affects E... x and E y Perform calculations to generate independent control outputs U x and U y Based on this, the cross-coupling control algorithm introduces an inter-axis synchronization error E. sync Its calculation is E sync Equal to E x Subtract E y The synchronization error E sync The circuit passes through a coupling controller, typically a proportional element, with a gain coefficient of K. cThe output of the coupled controller produces a compensation quantity Delta. U This compensation amount, Delta U It's not simply added to a certain axis, but rather directional compensation is performed: the Delta in the positive direction is... U Control output U attached to the travel axis x Up, and simultaneously the negative Delta U Control output U attached to the lifting shaft y Above. The specific calculation is as follows:
[0036] U xfinal =U x +K c *E sync
[0037] U yfinal =U y -K c *E sync
[0038] Among them, U xfinal and U yfinal This is the final control quantity after cross-coupling compensation, output to its respective servo drive. The core of this control strategy is that when the travel axis lags behind the lift axis, E... sync A negative value results in a compensation term that enhances the control output of the traveling axis while weakening the control output of the lifting axis. This causes the traveling axis to accelerate to catch up, while the lifting axis decelerates appropriately to wait, thus reducing the relative position error between the two axes. Conversely, a positive value has the same effect. Through this real-time, interconnected compensation mechanism, the following error between the traveling and lifting mechanisms on the composite motion trajectory is effectively suppressed, ensuring precise synchronization of the loading platform during horizontal and vertical movement. Ultimately, this improves the stacker crane's positioning accuracy and motion stability in three-dimensional space. Coupling gain coefficient K c The value needs to be tuned through on-site debugging to achieve the best balance between response speed and system stability.
[0039] The system described in this embodiment achieves a comprehensive intelligent upgrade of stacker crane control through the precise collaborative work of the aforementioned modules, moving from local perception to global optimization, from passive execution to proactive decision-making, and from periodic maintenance to predictive maintenance. The status perception and data acquisition module provides a comprehensive and real-time data foundation; the digital twin and simulation module constructs a reliable virtual testing environment; the multi-objective collaborative decision-making module realizes intelligent decision-making with comprehensive optimization; and the distributed execution control module ensures the accurate implementation of decisions. Additional functions such as health prediction, dynamic weight adjustment, accuracy adaptive simulation, and cross-coupling control further enhance the system's overall performance in terms of efficiency, energy consumption, equipment health, and environmental adaptability. The entire system constitutes a highly integrated, intelligent, efficient, stable, and reliable stacker crane automation control solution.
[0040] The above specific embodiments are merely several preferred embodiments of the present invention. Based on the technical solutions of the present invention and the relevant teachings of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.
Claims
1. An automated intelligent system for controlling a stacker crane, characterized in that: It includes a state awareness and data acquisition module, a digital twin and simulation module, a multi-objective collaborative decision-making module, and a distributed execution control module; The system comprises several modules: a status perception and data acquisition module that collects real-time data on the stacker crane's operating status, warehousing environment, and upper-level scheduling instructions; a digital twin and simulation module that builds and maintains a virtual model synchronized with the physical stacker crane and warehousing environment based on the collected data, and performs offline simulation and performance prediction of control strategies; a multi-objective collaborative decision-making module that receives prediction information from the digital twin module and comprehensively considers multiple optimization objectives such as efficiency, energy consumption, and equipment health to generate a globally optimal control decision sequence; and a distributed execution control module that parses the control sequence output by the decision module into underlying drive instructions and coordinates the collaborative actions of the stacker crane's traveling mechanism, lifting mechanism, and fork mechanism. Furthermore, the state perception and data acquisition module includes a sensor array mounted on the stacker crane body and a warehouse environment perception unit. The sensor array includes a first laser rangefinder for detecting the horizontal position of the stacker crane in the aisle, a second laser rangefinder for detecting the vertical height of the loading platform, a first current sensor for detecting the current of the travel drive motor, a second current sensor for detecting the current of the lifting drive motor, a third current sensor for detecting the current of the forklift motor, a first vibration sensor mounted on the travel wheel bearing housing, a second vibration sensor mounted on the lifting mechanism connection point, and an infrared temperature sensor for monitoring the temperature of electrical nodes. The warehouse environment perception unit includes a visual recognition device deployed at the aisle entrance for detecting the priority markers of tasks to be executed and a 3D LiDAR deployed in the aisle for detecting potential moving obstacles. The state perception and data acquisition module uploads all the collected sensor data and recognition results to the digital twin and simulation module in real time via industrial Ethernet.
2. The automated intelligent system for stacker crane control according to claim 1, characterized in that: The digital twin and simulation module internally constructs geometric, kinematic, dynamic, and energy consumption models corresponding to the physical entity. The digital twin and simulation module first receives real-time data from the state perception and data acquisition module and drives the virtual model to maintain state synchronization with the physical stacker crane. The digital twin and simulation module has a strategy simulation engine, which performs advanced simulation in the virtual environment based on the current system state and candidate control strategies provided by the multi-objective collaborative decision-making module. The simulation process simulates the stacker crane starting from its current position, fully executing the action sequence defined by the candidate control strategy, and simultaneously calculating multiple performance indicators during the simulation process. The performance indicators include at least the estimated task completion time, estimated energy consumption, and estimated stress load on mechanical components. All simulation results and performance indicators are encapsulated and sent to the multi-objective collaborative decision-making module for decision analysis.
3. The automated intelligent system for stacker crane control according to claim 1, characterized in that: The core of the multi-objective collaborative decision-making module is a multi-objective optimization function based on weighted summation. The input variable of the multi-objective optimization function is the candidate control strategy, and the output is the comprehensive evaluation value of the strategy. The multi-objective optimization function is composed of a linearly weighted sub-function of efficiency objective, energy consumption objective, and equipment health objective. The efficiency objective sub-function optimizes by minimizing the total task execution time, and its calculation relies on the estimated task completion time provided by the digital twin module. The energy consumption objective sub-function optimizes by minimizing the total electrical energy used by the stacker crane during task execution, and its calculation integrates the estimated energy consumption of walking, lifting, and fork movements provided by the digital twin module. The equipment health objective sub-function optimizes by minimizing the cumulative fatigue damage of mechanical components, and its calculation is based on the estimated stress load provided by the digital twin module and combined with the stress life characteristics of the material to estimate the degree of damage.
4. An automated intelligent system for stacker crane control according to claim 3, characterized in that, The multi-objective collaborative decision-making module has a set of weight coefficients pre-set, which correspond to the relative importance of the three objectives of efficiency, energy consumption and equipment health. The decision-making process of the multi-objective collaborative decision-making module is as follows: it receives a task set from the upper-level system, generates multiple control strategy sequences, uses a digital twin module to simulate and deduce each strategy and obtain its performance index, and then selects the control strategy sequence that gives the highest comprehensive evaluation value as the final decision output by calculating the value of the multi-objective optimization function.
5. An automated intelligent system for stacker crane control according to claim 1, characterized in that: The distributed execution control module includes a trajectory planning unit, a motion control unit, and an execution status monitoring unit. The trajectory planning unit is responsible for decomposing the target position and action command in the decision sequence into the motion trajectory of the stacker crane's traveling axis, lifting axis, and fork axis. This trajectory planning ensures that the movement of each axis is shock-free, while meeting the hardware limitations of maximum speed and acceleration. The motion control unit generates corresponding pulse commands or analog signals according to the planned trajectory, which drive the traveling servo driver, lifting servo driver, and fork servo driver respectively. The execution status monitoring unit continuously receives real-time feedback signals from the underlying sensors, including the actual position, speed, and drive current of each axis, and compares these actual values with the command values of the motion control unit. When the deviation exceeds the preset safety threshold, the monitoring unit immediately triggers the abnormal handling program and sends alarm information to the upper layer of the system. At the same time, it takes degraded operation or safe shutdown measures according to the preset strategy.
6. An automated intelligent system for stacker crane control according to claim 1, characterized in that: It also includes a health prediction and maintenance reminder submodule; the health prediction and maintenance reminder submodule continuously acquires vibration spectrum data from the first vibration sensor and the second vibration sensor from the status sensing module, and extracts their characteristic frequency amplitude; The health prediction and maintenance reminder submodule incorporates a fault recognition model based on a deep convolutional neural network. This model takes the vibration spectrum as input and extracts deep features through multi-layer convolution and pooling operations. The final output is the health status score and potential fault mode probability of the stacker crane's rotating components, including the traveling wheel bearings and the lifting wire rope drum. When the health status score is lower than a preset threshold or the probability of a specific fault mode exceeds a set threshold, the health prediction and maintenance reminder submodule automatically generates a pre-maintenance work order and pushes it to the warehouse management system.
7. An automated intelligent system for stacker crane control according to claim 4, characterized in that: The weighting coefficients are designed to be dynamically adjustable; the system introduces an operating mode switching logic, which can receive external instructions from the warehouse management system or automatically switch the system's optimization focus mode according to a preset time strategy. The system has at least three predefined operating modes: peak efficiency mode, balanced operation mode, and energy-saving maintenance mode. In peak efficiency mode, the weight of the efficiency target sub-function is set to the highest value, and the weights of energy consumption and equipment health are reduced accordingly. In balanced operation mode, the weights of the three objectives are set to equal or balanced values calibrated through experience; in energy-saving maintenance mode, the weights of the energy consumption objective sub-function and the equipment health objective sub-function are increased, while the efficiency weight is reduced accordingly. The dynamic adjustment of the weight coefficients is achieved by modifying the corresponding coefficient values in the multi-objective optimization function.
8. An automated intelligent system for stacker crane control according to claim 2, characterized in that: The simulation process of the digital twin and simulation deduction module introduces a precision adaptive mechanism. The precision adaptive mechanism dynamically adjusts the complexity of the simulation model and the deduction step size according to the criticality of the task to be decided and the real-time computing load of the system. The precision adaptive mechanism is managed by a resource scheduler, which monitors the utilization of the central processing unit in real time and selects simulation configuration parameters according to a preset rule base.
9. An automated intelligent system for stacker crane control according to claim 5, characterized in that: The motion control unit uses a cross-coupling control algorithm to synchronize the walking mechanism and the lifting mechanism. The cross-coupling control algorithm calculates the position error between the walking axis and the lifting axis in real time. It not only performs independent proportional, integral and derivative adjustments on each axis, but also introduces a compensation term based on the difference in position error between the two axes. After the compensation term is adjusted by a coupling gain coefficient, it is added to the control output of the walking axis and the lifting axis respectively, thereby suppressing the following error in the dual-axis motion process.
10. An automated intelligent system for stacker crane control according to claim 3, characterized in that, The calculation of the equipment health objective sub-function involves material fatigue analysis; the fatigue damage of the component is estimated through the following process: obtaining the number of stress cycles experienced at a specific stress level, which is statistically obtained by the digital twin module in the simulation; Query the pre-stored standard material stress-life characteristic database to obtain the number of cycles required for the material to fail at this stress level; The cumulative fatigue damage is calculated based on the ratio of the number of stress cycles to the number of cycles required for failure. The value of the equipment health target sub-function is inversely proportional to the total damage degree.