Automatic hooking and unhooking method and system of gantry crane

Abstract

The application discloses a gantry crane automatic hooking and unhooking method and system, and relates to the technical field of gantry cranes. The gantry crane automatic hooking and unhooking method comprises the steps of rough positioning, multi-modal sensing, dynamic prediction and path planning, adaptive hooking and unhooking, and closed-loop confirmation. The gantry crane automatic hooking and unhooking method combines multi-modal sensing and a digital twin prediction model to achieve high-precision positioning and path planning in a dynamic environment, and realizes accurate docking of a hook and an angle piece. Dynamic prediction simulates a motion trajectory in real time based on a digital twin model, plans an optimal path to offset swing, adaptive operation allows fine adjustment of the end of the hook to adapt to slight deviations, and a micro-motion trajectory action when unhooking can break static friction and mechanical jamming to prevent hook residues, and closed-loop confirmation is cross-verified by multiple sensors to ensure successful operation.

Description

Technical Field

[0001] This invention relates to the field of gantry crane technology, specifically to an automatic hooking and unhooking method and system for gantry cranes. Background Technology

[0002] Gantry cranes, also known as portal cranes, are a type of bridge crane. Their core feature is that the metal frame is shaped like a "gate" and moves horizontally via rails. They are widely used in ports, freight yards, shipyards, and other settings for container handling and stacking operations.

[0003] The existing hooking and unhooking operations of gantry cranes still have the following shortcomings:

[0004] First, the positioning accuracy is insufficient. It relies heavily on a single vision or laser system for positioning, which is easily affected by factors such as ambient light, dust, and container tilting, resulting in large initial positioning deviations that require manual adjustment.

[0005] Second, the dynamic adaptability is poor. The lifting equipment is easily affected by wind load and swing inertia during the lifting process, and the existing system cannot predict the relative motion trajectory of the hook and the corner piece in real time. It is difficult to dynamically counteract the swing, and docking failure or mechanical collision is likely to occur.

[0006] Third, the operation is not safe and reliable, and lacks a multi-dimensional closed-loop verification mechanism. Whether the hook is in place and whether the hook is completely released depends on the feedback of a single sensor. It is easy to make misjudgments due to sensor failure, which may lead to safety accidents such as cargo falling and equipment damage.

[0007] Fourth, the reliance on manual intervention is high. In complex operating conditions, such as container tilting or mechanical jamming, manual intervention is required for adjustment, which is labor-intensive and increases the safety risks for operators. To address the shortcomings of existing technologies, this invention provides an automatic hooking and unhooking method and system for gantry cranes to solve the above problems. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides an automatic hooking and unhooking method and system for gantry cranes. By combining multimodal sensing with a digital twin prediction model, it achieves high-precision positioning and path planning in dynamic environments, enabling precise docking of the hook and corner fittings, significantly improving the success rate and adaptability of hooking and unhooking. Dynamic prediction, based on real-time simulation of the motion trajectory using a digital twin model, plans the optimal path to counteract sway; adaptive operation allows for fine-tuning of the hook end to accommodate minor deviations; the micro-motion trajectory during unhooking breaks static friction and mechanical jamming, preventing hook residue; closed-loop confirmation, through multi-sensor cross-verification, ensures successful operation. The entire process requires no manual intervention, greatly improving the automation level of hooking and unhooking.

[0009] To achieve the above objectives, the present invention provides the following technical solution: an automatic hooking and unhooking method for a gantry crane, comprising the following steps:

[0010] Step S1, Coarse Positioning: Initial positioning of the target container and its corner pieces is performed using a global vision system or laser scanning system to narrow down the search range for subsequent precise positioning;

[0011] Step S2, Multimodal perception: The target corner piece is fused and perceived using the sensor array integrated on the lifting device. The sensor array includes at least a binocular depth camera, a high-precision lidar, a thermal imaging camera, and an inertial measurement unit (IMU) to acquire color images, depth point clouds, temperature distribution of the corner piece, and real-time attitude and motion data of the lifting device itself.

[0012] Step S3, Dynamic Prediction and Path Planning: Construct a digital twin model based on a physics engine, using the precise pose of the corner piece, the pose of the hook, and the sway data of the lifting device obtained in S2 as input, and simulate and predict the relative motion trajectory of the hook and the corner piece in real time over a period of time; based on the predicted trajectory, calculate an optimal motion path for the hook that can dynamically offset the sway and achieve precise docking.

[0013] Step S4, Adaptive hook release: Control the hook to move along the optimal path generated in S3. When approaching the corner piece, allow the hook end effector to adaptively deflect within a small angle range to guide the hook to slide into the corner piece hole. When releasing the hook, after the force sensor confirms that the load has been removed, control the lifting device to perform a specific micro-motion trajectory to ensure that the hook and corner piece are completely separated.

[0014] Step S5, Closed-loop confirmation: After the hooking or unhooking operation, cross-verification is performed using feedback information from the vision sensor, force sensor, and position sensor. If the information from the three sensors is consistent, the operation is considered successful; otherwise, the process is terminated and an alarm is triggered.

[0015] Preferably, in step S2, the thermal imaging camera is used to detect the temperature of the corner piece area. When the temperature exceeds a preset threshold, the system issues an early warning signal and records the abnormal state in the operation log for subsequent equipment maintenance and safety checks.

[0016] Preferably, in step S2, the high-precision lidar is used to scan and acquire the precise 3D point cloud of the corner fitting hole, to identify the non-vertical orientation of the corner fitting hole caused by the tilt of the container, and to transmit the orientation angle data as a key input to the digital twin model.

[0017] Preferably, in step S3, the physics engine of the digital twin model calculates the impact of wind load and sway inertia on the lifting system in real time, and the mathematical expression of its dynamic prediction is a prediction equation based on the system state space:

[0018] X(t+Δt)=A·X(t)+B·U(t)+W(t);

[0019] Where X(t) is the current state vector (including position, velocity, and attitude angle), X(t+Δt) is the predicted state vector after time Δt, A is the system state matrix, B is the control input matrix, U(t) is the current control variable, and W(t) is the process noise including wind load disturbance.

[0020] Preferably, in step S3, when generating the optimal motion path of the hook, the objective function is to minimize the sum of the squared expected distances between the hook and the center point of the corner fitting hole during the prediction period.

[0021] Preferably, in step S4, the "adaptive micro-motion trajectory action" is a predefined composite action of micro-horizontal translation and vertical jitter, used to break the static friction or mechanical jamming that may exist between the hook and the corner hole.

[0022] Preferably, in step S4, the adaptive deflection angle range of the hook end effector is achieved through a hinge mechanism that satisfies the following torque balance equation:

[0023] M×g×L×sinθ≤τ_max;

[0024] Where M is the hook mass, g is the gravitational acceleration, L is the deflection arm length, θ is the deflection angle, and τ_max is the maximum allowable reset torque of the hinge mechanism, ensuring that it can automatically return to the neutral position after deflection.

[0025] Preferably, in step S5, the specific logic of the cross-validation is as follows: after the hook is successfully attached, the vision system needs to recognize that the hook has fallen into the hole, the force sensor needs to detect that the weight load has increased, and the position sensor needs to detect that the locking mechanism has reached the locking position; if all three states meet the expectations, it is determined that the hook is successfully attached.

[0026] Preferably, before step S1, a system self-test step is also included: verifying the validity of all sensor readings, the initialization status of the digital twin model, and the response status of the actuator to ensure that the system is in a ready state.

[0027] The second aspect of this invention discloses an automatic hooking and unhooking system for a gantry crane, used to realize the automatic hooking and unhooking method of the gantry crane, comprising:

[0028] Central processing unit;

[0029] The coarse positioning module, connected to the central processing unit, performs initial positioning of the target container and its corner pieces through a global vision system or a laser scanning system, thereby narrowing the search range for subsequent precise positioning.

[0030] A multimodal sensor array, connected to the central processing unit and mounted on the lifting device, includes a binocular depth camera, a high-precision lidar, a thermal imaging camera, and an IMU. It is used to perform fusion perception on the target corner piece using the sensor array integrated on the lifting device, and to acquire color images, depth point clouds, temperature distribution of the corner piece, and real-time attitude and motion data of the lifting device itself.

[0031] The digital twin prediction and path planning module, integrated into the central processing unit, is used to construct a digital twin model based on a physics engine. It takes the precise pose of the corner piece, the pose of the hook, and the sway data of the lifting device obtained by the multimodal sensor array as input, and simulates and predicts the relative motion trajectory of the hook and the corner piece in real time over a period of time. Based on the predicted trajectory, it calculates an optimal motion path for the hook that can dynamically offset the sway and achieve precise docking.

[0032] An adaptive actuator, connected to the central processing unit, includes a hook end effector capable of micro-deflection and a spreader drive system. It controls the hook to move along the optimal path generated by S3. When approaching the corner piece, the hook end effector is allowed to adaptively deflect within a small angle range to guide the hook into the corner piece hole. When unhooking, after the force sensor confirms that the load has been removed, the spreader is controlled to perform a specific micro-motion trajectory to ensure that the hook is completely separated from the corner piece.

[0033] The closed-loop verification module, connected to the central processing unit, includes the vision sensor, force sensor, and position sensor. After the hooking or unhooking operation, it performs cross-verification based on the feedback information from the vision sensor, force sensor, and position sensor. If the information from the three sensors is consistent, the operation is considered successful; otherwise, the process is terminated and an alarm is triggered.

[0034] The technical effects and advantages of this invention are as follows:

[0035] 1. This gantry crane's automatic hooking and unhooking method, through the combination of multimodal sensing and digital twin prediction models, achieves high-precision positioning and path planning in dynamic environments, enabling precise docking of the hook and corner fittings, and significantly improving the success rate and adaptability of hooking and unhooking. Dynamic prediction, based on real-time simulation of the motion trajectory using a digital twin model, plans the optimal path to counteract sway; adaptive operation allows for fine-tuning of the hook end to accommodate minor deviations; the micro-motion trajectory during unhooking breaks static friction and mechanical jamming, preventing hook residue; closed-loop confirmation is achieved through multi-sensor cross-verification to ensure successful operation. The entire process requires no manual intervention, significantly improving the automation level of hooking and unhooking.

[0036] 2. The automatic hooking and unhooking method of this gantry crane incorporates thermal imaging and attitude sensing functions, enhancing the system's ability to identify and respond to abnormal operating conditions, thereby improving equipment safety and reliability. The thermal imaging camera can monitor abnormal corner fitting temperatures in real time, and the lidar can identify non-perpendicular orientation of corner fitting holes caused by container tilting. The system can provide early warnings and adjust operating strategies to avoid equipment damage or operational failures due to excessive temperature or attitude deviations.

[0037] 3. The automatic hooking and unhooking method of this gantry crane significantly improves its adaptability and stability under complex working conditions through the synergistic application of a multimodal sensor array and a digital twin model. The fusion perception of sensors such as binocular depth cameras, high-precision lidar, and IMUs can simultaneously acquire visual, depth, attitude, and temperature data of the corner components, effectively avoiding the shortcomings of single sensors being affected by environmental interference. The digital twin model, combined with a physics engine, calculates the impact of wind load and swaying inertia on the spreader in real time, accurately predicting the relative motion trajectory. Compared with existing technologies that can only statically plan paths, this method can dynamically respond to complex scenarios such as spreader swaying and container tilting, increasing the hook docking success rate and reducing the operational failure rate caused by fluctuations in working conditions. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0040] Figure 2 This is a flowchart of the data acquisition and processing for multimodal sensing in this invention;

[0041] Figure 3 This is the logic diagram of dynamic prediction and path planning in this invention;

[0042] Figure 4 This is a flowchart illustrating the adaptive hook unhooking process of the present invention.

[0043] Figure 5 This is a logic diagram of closed-loop confirmation cross-validation for the present invention;

[0044] Figure 6 This is a flowchart of the system self-test of the present invention;

[0045] Figure 7 This is a diagram of the overall system architecture of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] This embodiment discloses an automatic hooking and unhooking method for a gantry crane, according to the appendix. Figure 1 To be continued Figure 7 As shown, it includes the following steps:

[0048] Step S1, Coarse Positioning: Initial positioning of the target container and its corner components is performed using a global vision system or laser scanning system to narrow down the search range for subsequent precise positioning. The coarse positioning process initially locks the position of the target container and its corner components. This stage does not pursue extremely high positioning accuracy, but rather eliminates interference from irrelevant areas through large-scale scanning and identification, limiting the search range for subsequent precise positioning to a smaller spatial range, laying the foundation for subsequent high-precision perception.

[0049] Step S2, Multimodal Perception: The target corner piece is fused and perceived using a sensor array integrated on the lifting device. This sensor array includes at least a binocular depth camera, a high-precision LiDAR, a thermal imaging camera, and an inertial measurement unit (IMU). It acquires color images, depth point clouds, temperature distribution data of the corner piece, and real-time attitude and motion data of the lifting device itself. In the multimodal perception stage, multiple types of sensors are integrated on the lifting device to form a sensor array, including a binocular depth camera, a high-precision LiDAR, a thermal imaging camera, and an IMU. The binocular depth camera acquires color images and three-dimensional spatial depth information of the corner piece; the high-precision LiDAR acquires precise three-dimensional point cloud data of the corner piece; the thermal imaging camera captures the temperature distribution of the corner piece area; and the IMU collects real-time attitude (such as tilt angle and deflection angle) and motion data (such as velocity and acceleration) of the lifting device itself. Through multi-sensor data fusion, a comprehensive and accurate perception of the corner piece and the lifting device's state is achieved.

[0050] Step S3, Dynamic Prediction and Path Planning: Construct a physics engine-based digital twin model. Using the precise pose of the corner piece, hook pose, and spreader sway data obtained in S2 as input, the model simulates and predicts the relative motion trajectory of the hook and corner piece over a future period in real time. Based on the predicted trajectory, an optimal motion path for the hook that dynamically counteracts sway and achieves precise docking is calculated. The digital twin model constructed in the dynamic prediction and path planning stage can simulate the real physical environment and equipment motion state. The precise pose of the corner piece, hook pose, and spreader sway data obtained in the multimodal perception stage are input into the digital twin model. The model uses physics engine algorithms, such as dynamic equation calculation and kinematic simulation, to simulate and predict the relative motion trajectory of the hook and corner piece over a future period in real time. Based on the predicted relative motion trajectory, optimization algorithms, such as least squares and genetic algorithms, are used to calculate an optimal motion path that dynamically counteracts spreader sway and achieves precise docking of the hook and corner piece.

[0051] Step S4, Adaptive Hook and Hook Release: Control the hook to move along the optimal path generated in S3. When approaching the corner piece, allow the hook end effector to adaptively deflect within a small angle range to guide the hook into the corner piece hole. During release, after the force sensor confirms that the load has been removed, control the spreader to execute a specific micro-motion trajectory to ensure that the hook and corner piece are completely separated. In the adaptive hook and hook release stage, the hook is controlled to move according to the optimal path generated in the dynamic prediction and path planning stage. During the hook's approach to the corner piece, the hook end effector is allowed to adaptively deflect within a small angle range to adapt to possible slight positional or attitude deviations of the corner piece, guiding the hook to smoothly slide into the corner piece hole. During the release operation, first, the force sensor confirms that the load has been completely removed, then control the spreader to execute a specific micro-motion trajectory to ensure that the hook and corner piece are completely separated, avoiding hook residue or jamming.

[0052] Step S5, Closed-Loop Confirmation: After the hooking or unhooking operation, feedback information from the vision sensor, force sensor, and position sensor is cross-verified. If the three sensors are consistent, the operation is considered successful; otherwise, the process is stopped and an alarm is triggered. In the closed-loop confirmation phase, after the hooking or unhooking operation is completed, feedback information is obtained using the vision sensor, force sensor, and position sensor. The vision sensor observes whether the hook has accurately fallen into the corner fitting hole or has completely disengaged from it. The force sensor detects changes in load weight to determine whether the hooking or unhooking was successful. The position sensor detects whether the position of the locking mechanism or hook meets expectations. The feedback information from these three sensors is cross-verified. Only when all three sensors are consistent and meet the criteria for successful operation is the operation considered successful; otherwise, the operation process is immediately stopped and an alarm signal is issued, prompting personnel to troubleshoot the problem.

[0053] This method automates the entire process, covering all key aspects of gantry crane hook and unhooking operations. It achieves full automation from target positioning to operation confirmation, eliminating the need for manual intervention, significantly reducing the intensity of manual operations, minimizing the possibility of human error, and improving operational stability and reliability. Through a dual positioning approach of "coarse positioning + multimodal sensing," the search range is first narrowed down, followed by high-precision sensing, effectively improving the accuracy of diagonal component positioning and laying a solid foundation for subsequent precise docking.

[0054] The dynamic prediction and path planning stage can predict the relative motion trajectory of the hook and corner fitting in real time and generate the optimal motion path based on the prediction results, effectively dealing with the influence of dynamic factors such as spreader sway and wind load interference. The adaptive hooking and unhooking stage allows the hook end effector to adaptively deflect and the spreader to perform micro-motion trajectory movements, further enhancing the system's adaptability to equipment attitude deviations and mechanical jamming. The closed-loop confirmation stage, through multi-sensor cross-verification, can promptly detect abnormalities in the operation process, such as the hook not being accurately inserted into the hole or incomplete unhooking. By terminating the process and triggering alarm mechanisms, it avoids safety accidents such as equipment damage and cargo falling caused by operational errors, improving operational safety.

[0055] In step S2, the thermal imaging camera is used to detect the temperature of the corner area. When the temperature exceeds a preset threshold, the system issues an early warning signal and records the abnormal state in the operation log for subsequent equipment maintenance and safety checks.

[0056] Thermal imaging cameras operate based on the principle of thermal radiation. Any object with a temperature above absolute zero radiates infrared radiation. Corner components maintain a certain temperature range under normal operating conditions. However, when a corner component malfunctions, such as due to excessive friction or damage to internal components, it can cause an abnormal increase in localized temperature. The thermal imaging camera receives the infrared radiation emitted by the corner component, converts it into a temperature distribution image, and extracts the temperature data of the corner component area using image processing algorithms. The extracted temperature data is compared with a preset temperature threshold. When the temperature in the corner component area exceeds the preset threshold, the system's control unit triggers an early warning mechanism, issuing a warning signal such as an audible and visual alarm. Simultaneously, relevant information about this abnormal state, including the time of the anomaly, the corner component's location, and the temperature value, is recorded in the operation log. This provides subsequent equipment maintenance personnel with a basis for troubleshooting, facilitating the timely detection and repair of potential corner component faults. It also provides historical data support for safety inspections.

[0057] Therefore, this method can monitor the temperature changes of corner fittings in real time, promptly detect abnormal temperature increases caused by malfunctions, and issue early warning signals before serious damage occurs. This allows staff to take preventative measures, avoiding further escalation of the fault and reducing equipment maintenance costs and downtime. Recording operational logs of abnormal temperature conditions provides detailed fault information for equipment maintenance.

[0058] Abnormal temperatures in corner fittings can lead to a decline in their mechanical properties, such as reduced strength and decreased wear resistance. This can easily cause corner fitting breakage or hook detachment during hooking and unhooking operations, resulting in safety accidents. Thermal imaging cameras, with their temperature detection capabilities, can promptly identify such safety hazards and prevent dangerous operations through early warning mechanisms, ensuring the safety of the entire process.

[0059] In step S2, the high-precision lidar is used to scan and acquire the precise 3D point cloud of the corner fitting hole, to identify the non-vertical orientation of the corner fitting hole caused by the tilt of the container, and to transmit the orientation angle data as a key input to the digital twin model.

[0060] High-precision lidar scans target objects by emitting laser beams. When the laser beam encounters a corner fitting hole, it is reflected. The lidar receiver receives the reflected laser signal and, based on the emission and reception time difference, combined with the laser's propagation speed, calculates the distance between the lidar and various points on the corner fitting hole. By processing and reconstructing a large amount of distance data from laser scanning points, a precise 3D point cloud model of the corner fitting hole is generated. This model clearly reflects the shape, size, and spatial location information of the corner fitting hole. When the container tilts, the orientation of the corner fitting hole deviates from the vertical direction, and the 3D point cloud model generated by the lidar accurately captures this orientation change. Through point cloud processing algorithms, such as plane fitting and feature extraction, the axial direction of the corner fitting hole is extracted from the 3D point cloud model, and the orientation angle data of the corner fitting hole is calculated. This orientation angle data is transmitted to a digital twin model, providing key parameters for the model to accurately simulate the relative positional relationship and motion trajectory of the hook and the corner fitting hole, ensuring the accuracy of path planning.

[0061] Compared to other sensors, LiDAR offers higher ranging accuracy and spatial resolution, accurately reflecting the fine structure and spatial position of corner fitting holes, providing high-precision data support for subsequent positioning and path planning. In practice, containers may tilt due to uneven placement or collisions during loading and unloading, causing corner fitting holes to be non-perpendicular, which greatly hinders hook docking. High-precision LiDAR can accurately identify this non-perpendicular orientation and calculate the orientation angle data, avoiding docking failures caused by corner fitting hole orientation deviations.

[0062] By inputting the angle data of the corner fitting hole into the digital twin model, the model can more realistically simulate the relative motion state between the hook and the corner fitting hole, thereby improving the rationality and accuracy of path planning and ensuring the smooth operation of the hook.

[0063] In step S3, the physics engine of the digital twin model calculates the impact of wind load and sway inertia on the lifting system in real time, and the mathematical expression of its dynamic prediction is a prediction equation based on the system state space:

[0064] X(t+Δt)=A·X(t)+B·U(t)+W(t);

[0065] Where X(t) is the current state vector (including position, velocity, and attitude angle), X(t+Δt) is the predicted state vector after time Δt, A is the system state matrix, B is the control input matrix, U(t) is the current control variable, and W(t) is the process noise including wind load disturbance.

[0066] The physics engine in the digital twin model is based on classical mechanics theories, such as Newton's laws of motion and fluid dynamics principles, to perform real-time calculations of wind loads and oscillation inertia experienced by the spreader system in actual working environments. Wind load calculations consider factors such as wind speed, wind direction, the spreader's frontal area, and air resistance coefficient, using fluid dynamics formulas to calculate the force exerted by the wind on the spreader. Oscillation inertia calculations, based on parameters such as the spreader's mass, velocity, and acceleration, utilize Newton's second law to calculate the inertial force generated during the spreader's oscillation. These calculated forces are incorporated as external disturbance factors into the motion state analysis of the spreader system.

[0067] X(t) is the system state vector at the current time t, containing key state parameters such as the position of the spreader and hook (e.g., x, y, z coordinates in a three-dimensional coordinate system), velocity (e.g., velocity components along the x, y, z axes), and attitude angles (e.g., pitch, yaw, roll). X(t+Δt) is the predicted system state vector at the future time t+Δt. A is the system state matrix, which is determined by the inherent characteristics of the spreader system, such as the mass distribution and moment of inertia of the spreader, reflecting the natural change law of the system state under the condition of no external control and disturbance. B is the control input matrix, reflecting the degree of influence of the control quantity on the change of the system state. Its matrix elements are determined according to parameters such as the transmission ratio and control efficiency of the control mechanism. U(t) is the control quantity at the current time t, such as the output torque of the spreader drive motor and the displacement control quantity. W(t) is the process noise, which mainly includes random disturbance factors such as wind load disturbance and mechanical vibration. Its statistical characteristics are determined by monitoring and analyzing the disturbance factors in the actual working environment, such as variance and mean.

[0068] At each sampling time t, the current system state vector X(t), control quantity U(t), and process noise W(t) are substituted into the prediction equation, and the system state vector X(t+Δt) at the future time t+Δt is calculated through matrix operations, thereby enabling the prediction of the future motion state of the lifting device and hook. By continuously updating the current state vector and control quantity, the motion trajectory of the system over a future period can be predicted in real time and continuously.

[0069] This method uses a physics engine to calculate the effects of wind load and sway inertia on the spreader system in real time. It accurately simulates the motion of the spreader in complex real-world working environments, avoiding prediction errors caused by ignoring external disturbances and improving the accuracy of dynamic prediction. A prediction equation based on the system's state space is presented, transforming the dynamic prediction process into a quantitative mathematical calculation. This provides the prediction results with a clear mathematical basis and repeatability, facilitating analysis and optimization of the prediction process. It also provides precise mathematical input for subsequent path planning algorithms.

[0070] The calculation process of this prediction equation is relatively simple and suitable for real-time operation. It can complete the prediction of the future system state in a short time, meet the real-time requirements of the automatic hooking and unhooking operation of the gantry crane, and ensure that the path planning can respond to changes in the system state in a timely manner.

[0071] In step S3, when generating the optimal motion path of the hook, the objective function is to minimize the sum of the squared expected distances between the hook and the center point of the corner fitting hole during the prediction period.

[0072] In the dynamic prediction and path planning phase, the position coordinates of the center points of the hook and corner fitting holes at various times within a future period (prediction period) are first predicted based on the digital twin model. Let there be n sampling times within the prediction period, denoted as t1, t2, ..., t... n At time t i (i = 1, 2, ..., n), the predicted hook position coordinates are (x_h, y_h, z_h), and the corner fitting hole center point coordinates are (x_c, y_c, z_c). Then at time t... i The distance d between the hook and the center point of the corner fitting hole can be calculated using the formula for the distance between two points in space:

[0073] The path planning process involves finding a sequence of control variables U(t1), U(t2), ..., U(t3) under the premise of satisfying system constraints, such as hook speed limits, acceleration limits, and drive motor output torque limits. n-1This allows the objective function J to reach its minimum value. By employing appropriate optimization algorithms, such as gradient descent, particle swarm optimization, or genetic algorithms, the objective function is solved to obtain the optimal control sequence. The optimal motion path of the hook is then generated based on this control sequence.

[0074] The optimization objective is to minimize the sum of squared expected distances between the hook and the center point of the corner fitting hole. This directly addresses the core requirement of precise docking between the hook and the corner fitting hole during hook-up operations. The objective function has a clear physical meaning and can ensure that the generated motion path makes the hook as close as possible to the center point of the corner fitting hole within the prediction period, thereby improving docking accuracy.

[0075] This objective function considers the distance between the hook and the center point of the corner fitting hole at various times during the prediction period, rather than just focusing on the distance at the final docking time. This ensures that the hook maintains a relatively close distance to the center point of the corner fitting hole throughout the entire movement process, avoiding collisions and jamming caused by excessive distance during movement, thus improving the stability and safety of the movement process.

[0076] In step S4, the "adaptive micro-motion trajectory action" is a predefined composite action of micro-horizontal translation and vertical jitter, which is used to break the static friction or mechanical jamming that may exist between the hook and the corner hole.

[0077] During the unhooking process, after the load is removed, static friction may occur between the hook and the corner fitting hole due to factors such as the roughness of the contact surface and pressure, or mechanical jamming may occur due to manufacturing errors, wear, etc., making it impossible for the hook to detach smoothly from the corner fitting hole.

[0078] The algorithm pre-designs a composite micro-motion trajectory combining minute horizontal translation and vertical jitter based on parameters such as the structural dimensions and material properties (e.g., coefficient of friction) of the hook and corner fitting hole. The minute horizontal translation refers to controlling the spreader to move the hook in a small reciprocating motion in the horizontal direction, typically a distance of a few millimeters to tens of millimeters, with the specific value determined based on the actual equipment. The vertical jitter refers to controlling the spreader to move the hook in a small up-and-down jitter in the vertical direction, with a similarly small amplitude, generally within a few millimeters, and a low frequency to avoid excessive impact on the equipment. During the unhooking operation, once the force sensor confirms that the load has been removed, the system controls the spreader to move according to the pre-defined micro-motion trajectory. The horizontal translation changes the contact position and contact pressure between the hook and corner fitting hole, disrupting the balance of static friction; the vertical jitter generates a periodic impact force, impacting the contact area between the hook and corner fitting hole, breaking the mechanical jamming, thus allowing the hook to smoothly separate from the corner fitting hole.

[0079] To address common issues like static friction and mechanical jamming during unhooking, a composite micro-motion trajectory system effectively resolves these problems, ensuring complete separation of the hook from the corner fitting hole and preventing equipment malfunctions or safety accidents caused by incomplete unhooking. The parameters of the micro-motion trajectory, such as horizontal translation distance, vertical vibration amplitude, and frequency, are pre-designed based on the actual equipment conditions and physical characteristics. The small amplitude and low frequency of the movements prevent excessive impact and damage to the lifting gear, hook, and corner fitting, thus extending the equipment's lifespan. The system is automatically triggered and executed without manual intervention, further enhancing the automation of the unhooking operation and reducing manual operation costs and the risk of human error.

[0080] In step S4, the adaptive deflection angle range of the hook end effector is achieved through a hinge mechanism, which satisfies the following torque balance equation:

[0081] M×g×L×sinθ≤τ_max;

[0082] Where M is the hook mass, g is the gravitational acceleration, L is the deflection arm length, θ is the deflection angle, and τ_max is the maximum allowable reset torque of the hinge mechanism, ensuring that it can automatically return to the neutral position after deflection.

[0083] The end effector of the hook is connected to the main body of the lifting device via a hinge mechanism. This mechanism allows the actuator to rotate within a certain angle range around the hinge point, thereby achieving adaptive deflection. When the hook approaches the corner fitting, if there is a slight positional or attitude deviation, the actuator can adjust its own attitude by deflecting to align the hook with the hole in the corner fitting and guide the hook to slide in.

[0084] Parameter definitions: M is the mass of the hook (unit: kg), g is the acceleration due to gravity (taken as 9.8 m / s²). 2 L is the length of the deflection arm (i.e., the distance from the hinge point to the center of gravity of the hook, in meters), θ is the deflection angle (in rad), and τ_max is the maximum allowable reset torque of the hinge mechanism (in N·m), which is determined by the performance of the spring, hydraulic or pneumatic reset components of the hinge mechanism.

[0085] When the actuator deflects at an angle θ, the weight of the hook will generate a torque about the hinge point, with a magnitude of M×g×L×sinθ (gravity is in the vertical direction, and the lever arm is L×sinθ). This torque attempts to keep the actuator in its deflected state, while the reset torque τ_max attempts to pull the actuator back to its neutral position. To ensure automatic recovery after deflection, the torque generated by gravity must not exceed the maximum permissible reset torque, i.e., M×g×L×sinθ≤τ_max. The maximum permissible deflection angle θ_max can be calculated using this equation, and in practical applications, the deflection angle must be controlled within θ_max.

[0086] The maximum deflection angle is clearly defined through the torque balance equation, preventing actuator failure or structural damage due to excessive deflection and ensuring safe and reliable deflection action. Based on the balanced design of reset torque and gravitational torque, automatic reset of the actuator after deflection can be achieved without an additional drive device, simplifying the mechanism structure and reducing equipment costs and failure risks. Within the allowable deflection range, the actuator can flexibly adjust its posture to adapt to minor deviations in the corner fittings, improving the success rate of hook-to-corner fitting hole alignment and reducing operational failures caused by deviations.

[0087] In step S5, the specific logic of the cross-validation is as follows: after the hook is successfully attached, the vision system needs to recognize that the hook has fallen into the hole, the force sensor needs to detect that the weight load has increased, and the position sensor needs to detect that the locking mechanism has reached the locking position; if all three states meet the expectations, it is determined that the hook is successfully attached.

[0088] After the hooking operation is completed, the vision sensor captures an image of the corner fitting hole area and uses an image recognition algorithm to determine whether the hook has completely fallen into the hole (e.g., detecting the relative position of the hook outline and the edge of the hole); the force sensor is installed on the load-bearing part of the hook or spreader to detect the load weight in real time. After the hooking is successful, due to the loading of the container weight, the sensor reading will jump from the unloaded value (e.g., the hook's own weight M×g) to the load value (M×g + container weight); the position sensor (e.g., limit switch, encoder) is installed at the locking mechanism to detect whether the mechanism has moved to the preset locking position (e.g., the locking pin is inserted into place).

[0089] The system performs a logical AND operation on the feedback signals from the three sensors: the hook is considered successfully engaged only when the visual signal shows "hook in hole", the force sensor signal shows "load increased to preset value", and the position sensor signal shows "locking mechanism in position". If any signal does not meet expectations (such as the visual sensor not recognizing the hook or the force sensor showing no load change), the system is considered to have failed, the process is stopped, and an alarm is triggered.

[0090] Cross-validation using three different types of sensors avoids misjudgments caused by a single sensor malfunction (e.g., a visual sensor misjudging a successful connection due to obstruction can be supplemented by a force sensor for verification), significantly improving the accuracy of hook-up success determination. If a determination fails, the cause of the fault can be quickly located by analyzing the signal status of a single sensor (e.g., no change in the force sensor may indicate the hook is not properly engaged, or the position sensor not being in position may indicate a malfunction in the locking mechanism), facilitating subsequent troubleshooting. Strict judgment criteria ensure that hook-up success is only confirmed when all safety conditions are fully met, preventing containers from falling during lifting due to insecure hook-ups and improving operational safety.

[0091] Before step S1, a system self-test step is also included: verifying the validity of all sensor readings, the initialization status of the digital twin model, and the response status of the actuator to ensure that the system is in a ready state.

[0092] Sensor reading verification: The system sends test signals to each sensor (binocular camera, lidar, IMU, force sensor, etc.), receives sensor feedback data, and determines whether the data is within the normal range (e.g., lidar ranging error ≤ ±2mm, IMU attitude angle error ≤ ±0.1°). If it exceeds the range, the sensor is marked as abnormal.

[0093] Digital twin model initialization verification: Check whether the model has correctly loaded the 3D model parameters (such as size, mass, and moment of inertia) of the lifting device, hook, and corner piece, whether the physics engine parameters (such as drag coefficient and friction coefficient) are set correctly, and whether the simulation results of the model in the initial state are consistent with the static state of the actual equipment (such as the model attitude angle being 0 when the lifting device is horizontal).

[0094] Actuator response verification: Send small control commands (such as driving the spreader to move horizontally by 50mm or the articulating mechanism to deflect by 2°) to the spreader drive system, articulation mechanism, and locking mechanism, and check whether the actuators act according to the commands, whether the feedback position and angle signals match the commands, and determine whether the mechanism response is normal.

[0095] Identifying potential faults in sensors, models, or actuators before formal operation prevents operational failures or safety incidents due to equipment malfunctions, reducing subsequent downtime for maintenance. This ensures system reliability and reduces operational risks.

[0096] The second aspect of this invention discloses an automatic hooking and unhooking system for a gantry crane, used to realize the automatic hooking and unhooking method of the gantry crane, comprising:

[0097] Central Processing Unit (CPU): As the core of the system, the CPU is responsible for receiving data from various modules, performing algorithm calculations (such as data fusion and path planning), sending control commands, and coordinating the synchronous work of various modules. It is equivalent to the "brain" of the system.

[0098] The coarse positioning module, connected to the central processing unit, performs initial positioning of the target container and its corner pieces through a global vision system or a laser scanning system, thereby narrowing the search range for subsequent precise positioning.

[0099] A multimodal sensor array, connected to the central processing unit and mounted on the lifting device, includes a binocular depth camera, a high-precision lidar, a thermal imaging camera, and an IMU. It is used to perform fusion perception on the target corner piece using the sensor array integrated on the lifting device, and to acquire color images, depth point clouds, temperature distribution of the corner piece, and real-time attitude and motion data of the lifting device itself.

[0100] The digital twin prediction and path planning module, integrated into the central processing unit, is used to construct a digital twin model based on a physics engine. It takes the precise pose of the corner piece, the pose of the hook, and the sway data of the lifting device obtained by the multimodal sensor array as input, and simulates and predicts the relative motion trajectory of the hook and the corner piece in real time over a period of time. Based on the predicted trajectory, it calculates an optimal motion path for the hook that can dynamically offset the sway and achieve precise docking.

[0101] An adaptive actuator, connected to the central processing unit, includes a hook end effector capable of micro-deflection and a spreader drive system. It controls the hook to move along the optimal path generated by S3. When approaching the corner piece, the hook end effector is allowed to adaptively deflect within a small angle range to guide the hook into the corner piece hole. When unhooking, after the force sensor confirms that the load has been removed, the spreader is controlled to perform a specific micro-motion trajectory to ensure that the hook is completely separated from the corner piece.

[0102] The closed-loop verification module, connected to the central processing unit, includes the vision sensor, force sensor, and position sensor. After the hooking or unhooking operation, it performs cross-verification based on the feedback information from the vision sensor, force sensor, and position sensor. If the information from the three sensors is consistent, the operation is considered successful; otherwise, the process is terminated and an alarm is triggered.

[0103] Example 1: Automatic hook-on operation under standard working conditions;

[0104] Operating parameters

[0105] Container specifications: 20-foot standard container; corner fitting hole size (length × width × depth): 80mm × 80mm × 120mm; corner fitting weight: 5kg.

[0106] Hook parameters: mass M = 10 kg, deflection arm length L = 0.3 m, maximum reset torque τ_max of hinge mechanism = 50 N·m;

[0107] Environmental parameters: wind speed v = 2 m / s (light breeze), no obvious vibration, ambient temperature 25℃;

[0108] Digital twin model parameters: System state matrix A = [[1,0.01,0],[0,1,0.01],[0,0,1]] (sampling time 0.01s, simplified two-dimensional position-velocity model), control input matrix B = [[0.00005],[0.01],[0]], process noise W(t) has a mean of 0 and a variance of [[0.0001,0,0],[0,0.0001,0],[0,0,0.0001]] (wind load disturbance is relatively small).

[0109] Operation process and calculation

[0110] System self-check:

[0111] Sensor calibration: Binocular camera resolution 1920×1080, image clarity ≥90%; LiDAR ranging error ±1mm, meeting requirements; Force sensor range 0-5000kg, zero drift ≤0.1%FS;

[0112] Model initialization: Input a 3D model of the container corner fittings (dimensions 80mm×80mm×120mm), with a spreader mass of 500kg and a moment of inertia of 100kg·m. 2 Initialize the state vector X(0) = [10m, 0m / s, 0rad] (initial position x = 10m, velocity 0, attitude angle 0);

[0113] Actuator response: The control spreader was moved horizontally by 100mm, and the actual movement distance was 99.8mm, with an error of ≤0.2%, indicating a normal response.

[0114] Coarse positioning:

[0115] The global vision system captures images of the area, identifies the container's location, and has a positioning error of ±50mm, thus narrowing the subsequent precise positioning search range to a 100mm×100mm×100mm cube area.

[0116] Multimodal perception:

[0117] A binocular camera acquires color images of the corner fitting, and depth data is used to calculate the coordinates of the center point of the corner fitting hole (x_c, y_c, z_c) = (10.02m, 0.5m, 3.2m).

[0118] The lidar scans the corner hole, generates a 3D point cloud, and calculates the hole axis orientation angle to be 0° (vertical orientation), with no tilt.

[0119] The thermal imaging camera detected a corner piece temperature of 26℃, which is lower than the preset threshold of 50℃, and no warning was issued.

[0120] IMU acquired the following attitude data for the lifting device: attitude angle 0.1°, velocity 0.05 m / s, acceleration 0.02 m / s². 2 .

[0121] Dynamic prediction and path planning:

[0122] State prediction: At the current time t = 0, X(0) = [10m, 0m / s, 0rad], control quantity U(0) = 200N (drive motor output force), substituting into the prediction equation X(0.01) = A·X(0) + B·U(0) + W(0); we get: X(0.01) = [[1,0.01,0],[0,1,0.01],[0,0,1]]·

[0123] [

[10] ,[0],[0]]+[[0.00005],[0.01],[0]]·200+[[0.001],[0.001],[0]]

[0124] (W(0) takes values ​​near the mean) = [[10.01], [2.001], [0]];

[0125] That is, at t = 0.01s, the position of the lifting device is x = 10.01m, the speed is 2.001m / s, and the attitude angle is 0rad; the optimization objective function is: prediction period T = 0.5s (50 sampling times), objective function J = Σ[(x_h - 10.02)] 2 +(y_h-0.5) 2 (i = 1-50) is solved by gradient descent to obtain the optimal control quantity sequence U(t) and generate the optimal path: the lifting device moves at a constant speed of 2m / s and reaches x = 10.02m after 0.05s, aligning with the corner fitting hole.

[0126] Adaptive hook:

[0127] Deflection angle calculation: From the torque balance equation M×g×L×sinθ≤τ_max, substituting M=10kg, g=9.8m / s 2 , L=0.3m, τ_max=50N·m;

[0128] We get 10×9.8×0.3×sinθ≤50→sinθ≤50 / (29.4)≈1.7;

[0129] Since the physical upper limit of sinθ is 1, the theoretical maximum allowable value of θ is θ_max = arcsin(1) = 90°.

[0130] Considering mechanical structure, safety clearance and reset reliability, this embodiment limits the actual usable deflection angle to ±10°, satisfying sin10°≈0.174 and stress moment 29.4×0.174≈5.1N·m≤50N·m.

[0131] When the hook approaches the corner piece, there is a 0.5° attitude deviation. The actuator automatically deflects by 0.5° to adjust the hook attitude and guide the hook to slide into the corner piece hole.

[0132] Closed-loop confirmation:

[0133] Visual sensor recognition: The hook falls completely into the corner fitting hole, and the gap between the edge of the hole and the hook is ≤5mm;

[0134] Force sensor detection: The load jumped from 10 × 9.8 = 98 N to (10 + 5 + 2000) × 9.8 = 19750.2 N (container mass 2000 kg), which is in line with expectations;

[0135] Position sensor detection: The limit switch of the locking mechanism is triggered, and the display shows that the locked position has been reached;

[0136] Cross-validation passed, the buckle was successfully attached.

[0137] Example 2: Automatic hook-up operation under container tilting conditions;

[0138] Operating parameters

[0139] Container condition: Due to uneven ground, the container is tilted, and the angle θ of the corner fitting hole axis is 5° (deviating from the vertical direction);

[0140] Other parameters: Same as in Example 1 (hook M = 10kg, L = 0.3m, τ_max = 50N·m; wind speed v = 3m / s; model parameters remain unchanged).

[0141] Key Calculation and Operation Differences

[0142] Multimodal perception:

[0143] The laser radar scans the 3D point cloud of the corner fitting hole, extracts the hole axis direction through a plane fitting algorithm, calculates the orientation angle θ = 5°, and inputs this data into the digital twin model.

[0144] Dynamic prediction and path planning:

[0145] The digital twin model is input with a corner hole tilt angle of 5°, and the coordinates of the corner hole center point in the prediction equation are adjusted over time (due to the tilt, the hole center has a slight offset in the vertical direction).

[0146] The objective function is optimized to J = Σ[(x_h - x_c)]. 2 +(y_h-y_c) 2 +(z_h-z_c) 2 +(θ_h-5°)2](Add attitude angle deviation term to ensure that the hook attitude matches the hole tilt angle;

[0147] The optimal path was calculated: while the spreader moves horizontally, it causes the hook to deflect by 5°, and it reaches the docking position after 0.06s.

[0148] Adaptive hook:

[0149] The actuator deflection angle is 5°. Substituting into the torque balance equation: 10×9.8×0.3×sin5°≈10×9.8×0.3×0.087≈2.56N·m≤50N·m, which meets the reset requirement;

[0150] The hook slides into the corner fitting hole with a 5° deflection, and the connection is successful.

[0151] Example 3: Unhooking operation;

[0152] Operating parameters

[0153] Hook status: The hook is attached to the corner fitting of the container, with a load capacity of 2015kg (2000kg container + 5kg corner fitting + 10kg hook).

[0154] Unhooking environment: wind speed v = 1.5 m / s, the spreader has placed the container in the designated position, and the ground support is stable;

[0155] Micro-motion trajectory parameters: horizontal translation distance ±3mm, translation frequency 1Hz; vertical jitter amplitude ±2mm, jitter frequency 2Hz.

[0156] Operation process and calculation

[0157] Load confirmation:

[0158] The force sensor detected that the load weight dropped from 19750.2N (2015×9.8) to 98N (hook weight only), confirming that the load had been removed.

[0159] Adaptive decoupling:

[0160] The system triggers a micro-motion trajectory action: the control spreader first moves horizontally to the right by 3mm, then moves horizontally to the left by 3mm (reciprocating once, taking 1 second); at the same time, it shakes vertically upward by 2mm, then shakes downward by 2mm (reciprocating twice, taking 1 second).

[0161] Mechanical Analysis: The static friction force between the hook and the corner fitting hole is f_s = μ_s × F_N, where the coefficient of friction μ_s = 0.3 (steel-to-steel contact), and the normal force F_N = 98N (hook weight). Therefore, f_s = 0.3 × 98 = 29.4N. The horizontal force generated by the fretting motion is F_x = ma. The lifting device mass is 500kg, the horizontal acceleration is a = Δv / Δt, and the translation time of 3mm is 0.5s (one-way). Therefore, Δv = Δx / (Δt) 2 / 2)=0.003m / (0.5s 2 / 2)=0.024m / s, a=Δv / Δt=0.024m / s / 0.5s=0.048m / s 2 F_x = 500kg × 0.048m / s 2 =24N; The impact force generated by vertical shaking F_z = Δp / Δt, the momentum change Δp = m × Δv_z, Δv_z = 0.002m × 2 (round trip) / 0.25s = 0.016m / s, Δp = 500kg × 0.016m / s = 8kg·m / s, F_z = 8kg·m / s / 0.1s = 80N;

[0162] The combined force effect: the horizontal force of 24N is close to the static friction force of 29.4N, and the vertical impact force of 80N breaks the mechanical jamming contact state. The two work together to break the static friction and jamming, and the hook is successfully released from the corner hole.

[0163] Closed-loop confirmation:

[0164] Visual sensor recognition: The hook is completely detached from the corner fitting hole, with no residue inside the hole;

[0165] Force sensor detection: The load remains stable at 98N (hook weight only), with no fluctuations;

[0166] Position sensor detection: Locking mechanism resets to initial position;

[0167] Cross-validation passed, indicating successful decoupling.

[0168] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic hook-and-unhooking method for a gantry crane, characterized in that, Includes the following steps: Step S1, Coarse Positioning: Initial positioning of the target container and its corner pieces is performed using a global vision system or laser scanning system to narrow down the search range for subsequent precise positioning; Step S2, Multimodal perception: The target corner piece is fused and perceived using the sensor array integrated on the lifting device. The sensor array includes at least a binocular depth camera, a high-precision lidar, a thermal imaging camera, and an inertial measurement unit (IMU) to acquire color images, depth point clouds, temperature distribution of the corner piece, and real-time attitude and motion data of the lifting device itself. Step S3, Dynamic Prediction and Path Planning: Construct a digital twin model based on a physics engine, using the precise pose of the corner piece, the pose of the hook, and the sway data of the lifting device obtained in S2 as input, and simulate and predict the relative motion trajectory of the hook and the corner piece in real time over a period of time; based on the predicted trajectory, calculate an optimal motion path for the hook that can dynamically offset the sway and achieve precise docking. Step S4, Adaptive hook release: Control the hook to move along the optimal path generated in S3. When approaching the corner fitting, allow the hook end effector to adaptively deflect within a small angle range to guide the hook to slide into the corner fitting hole. When releasing the hook, after the force sensor confirms that the load has been removed, control the lifting device to perform a specific micro-motion trajectory action. Step S5, Closed-loop confirmation: After the hooking or unhooking operation, cross-verification is performed using feedback information from the vision sensor, force sensor, and position sensor. If the information from the three sensors is consistent, the operation is considered successful; otherwise, the process is terminated and an alarm is triggered.

2. The automatic hooking and unhooking method for a gantry crane according to claim 1, characterized in that, In step S2, the thermal imaging camera is used to detect the temperature of the corner area. When the temperature exceeds a preset threshold, the system issues an early warning signal and records the abnormal state in the operation log.

3. The automatic hooking and unhooking method for a gantry crane according to claim 2, characterized in that, In step S2, the high-precision lidar is used to scan and acquire the precise 3D point cloud of the corner fitting hole, to identify the non-vertical orientation of the corner fitting hole caused by the tilt of the container, and to transmit the orientation angle data as a key input to the digital twin model.

4. The automatic hooking and unhooking method for a gantry crane according to claim 3, characterized in that, In step S3, the physics engine of the digital twin model calculates the impact of wind load and sway inertia on the lifting system in real time, and the mathematical expression of its dynamic prediction is a prediction equation based on the system state space: X(t+Δt)=A·X(t)+B·U(t)+W(t); Where X(t) is the current state vector, X(t+Δt) is the predicted state vector after time Δt, A is the system state matrix, B is the control input matrix, U(t) is the current control quantity, and W(t) is the process noise including wind load disturbance.

5. The automatic hooking and unhooking method for a gantry crane according to claim 4, characterized in that, In step S3, when generating the optimal motion path of the hook, the objective function is to minimize the sum of the squared expected distances between the hook and the center point of the corner fitting hole during the prediction period.

6. The automatic hooking and unhooking method for a gantry crane according to claim 1, characterized in that, In step S4, the "adaptive micro-motion trajectory action" is a predefined composite action of micro-horizontal translation and vertical jitter, which is used to break the static friction or mechanical jamming that may exist between the hook and the corner hole.

7. The automatic hooking and unhooking method for a gantry crane according to claim 1, characterized in that, In step S4, the adaptive deflection angle range of the hook end effector is achieved through a hinge mechanism, which satisfies the following torque balance equation: M×g×L×sinθ≤τ_max; Where M is the hook mass, g is the gravitational acceleration, L is the deflection arm length, θ is the deflection angle, and τ_max is the maximum allowable reset torque of the hinge mechanism.

8. The automatic hooking and unhooking method for a gantry crane according to claim 1, characterized in that, In step S5, the specific logic of the cross-validation is as follows: after the hook is successfully attached, the vision system needs to recognize that the hook has fallen into the hole, the force sensor needs to detect that the weight load has increased, and the position sensor needs to detect that the locking mechanism has reached the locking position; if all three states meet the expectations, it is determined that the hook is successfully attached.

9. The automatic hooking and unhooking method for a gantry crane according to claim 1, characterized in that, Before step S1, a system self-test step is also included: verifying the validity of all sensor readings, the initialization status of the digital twin model, and the response status of the actuator.

10. An automatic hooking and unhooking system for a gantry crane used to implement the automatic hooking and unhooking method of any one of claims 1-9, characterized in that, include: Central processing unit; The coarse positioning module, connected to the central processing unit, performs initial positioning of the target container and its corner pieces through a global vision system or a laser scanning system, thereby narrowing the search range for subsequent precise positioning. A multimodal sensor array, connected to the central processing unit and mounted on the lifting device, includes a binocular depth camera, a high-precision lidar, a thermal imaging camera, and an IMU. It is used to perform fusion perception on the target corner piece using the sensor array integrated on the lifting device, and to acquire color images, depth point clouds, temperature distribution of the corner piece, and real-time attitude and motion data of the lifting device itself. The digital twin prediction and path planning module, integrated into the central processing unit, is used to construct a digital twin model based on a physics engine. It takes the precise pose of the corner piece, the pose of the hook, and the sway data of the lifting device obtained by the multimodal sensor array as input, and simulates and predicts the relative motion trajectory of the hook and the corner piece in real time over a period of time. Based on the predicted trajectory, it calculates an optimal motion path for the hook that can dynamically offset the sway and achieve precise docking. An adaptive actuator, connected to the central processing unit, includes a hook end effector capable of micro-deflection and a spreader drive system. It controls the hook to move along the optimal path generated by S3. When approaching the corner piece, the hook end effector is allowed to adaptively deflect within a small angle range to guide the hook into the corner piece hole. When unhooking, after the force sensor confirms that the load has been removed, the spreader is controlled to perform a specific micro-motion trajectory to ensure that the hook is completely separated from the corner piece. The closed-loop verification module, connected to the central processing unit, includes the vision sensor, force sensor, and position sensor. After the hooking or unhooking operation, it performs cross-verification based on the feedback information from the vision sensor, force sensor, and position sensor. If the information from the three sensors is consistent, the operation is considered successful; otherwise, the process is terminated and an alarm is triggered.

Images