Ship intelligent workshop navigation vehicle hoisting path planning method

Abstract

The present application relates to the technical field of control system data processing, and especially relates to a ship intelligent workshop navigation vehicle hoisting path planning method, which comprises collecting navigation vehicle coordinates, obstacle distribution, load swing angle and material weight, inputting a physical constraint model to generate a structured environment state data set; wolf swarm algorithm parameters are initialized based on the structured environment state data set, the compressed investigation wolf walking step is calculated according to the obstacle distance, the industrial cloud platform generates a global task queue according to a dynamic parameter set, and the navigation vehicle end outputs an execution state data set; the walking step function is optimized by counting the number of emergency braking from the execution state data set, the investigation wolf proportion is optimized by comparing the task time, the step and swing mapping library is established by analyzing the historical swing data, and the physical constraint model is updated by feeding back the optimized parameter set. The present application realizes the cooperation of environment perception, dynamic path planning and closed loop verification, and reduces the accident risk and production delay in the narrow space of the ship workshop under the dynamic obstacle scene.

Description

Technical Field

[0001] This invention relates to the field of control system data processing technology, and in particular to a method for planning the hoisting path of a ship's intelligent workshop. Background Technology

[0002] The intelligent shipbuilding workshop is a core unit of modern shipbuilding that integrates the Internet of Things, artificial intelligence, and automation technologies. By integrating manufacturing execution systems, industrial networks, and intelligent equipment, it achieves digital control and collaborative optimization of the production process. Various devices in the workshop collect and interact with data in real time, and rely on digital twin models for dynamic simulation and decision-making reasoning, thereby autonomously scheduling resources and adjusting process parameters to improve manufacturing accuracy and production efficiency. Its highly integrated characteristics enable the production process to have self-adaptive and self-optimizing capabilities, significantly reducing reliance on manpower and operating costs, and supporting the evolution of shipbuilding towards flexibility and intelligence.

[0003] In the intelligent shipyard, overhead crane lifting operations rely on an intelligent system that integrates multiple types of sensors and advanced control units to achieve high-precision cargo handling and positioning. Through real-time data collection and analysis, the system automatically optimizes the lifting trajectory and load distribution, thereby improving operational efficiency and enhancing safety. Reducing manual intervention enables the lifting process to adapt to complex working conditions, lowers the probability of errors, and supports the continuity and coordination of the production process.

[0004] Existing intelligent shipbuilding workshop overhead crane lifting systems suffer from the following technical pain points in practical applications: In complex workshop environments, such as narrow workspaces or scenarios with dynamic obstacles, direct human involvement in overhead crane lifting operations can easily lead to collisions or falls due to blind spots or delayed judgment by the operator. For example, when moving heavy ship structural components, the operator cannot respond to changes in the position of the moving equipment in real time, resulting in a sharp increase in the risk of collisions between parts and equipment. Simultaneously, path planning relying on rules of thumb lacks a dynamic optimization mechanism based on intelligent data processing, causing path redundancy or task conflicts in batch material handling tasks. Specifically, this manifests as accumulated scheduling waiting time when multiple overhead cranes are working together, significantly extending the lifting cycle and delaying subsequent production stages. Under collaborative operation, high accident risk and production delays reinforce each other, threatening operational safety and reducing overall efficiency, failing to meet the requirements of intelligent manufacturing for control precision and data processing capabilities. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for planning overhead crane lifting paths in a ship intelligent workshop. This invention solves the technical problem that direct human involvement in overhead crane lifting operations and reliance on rules of thumb for path planning lead to high accident risks and deteriorated production delays and coordination in dynamic obstacles and narrow working environments within ship workshops.

[0006] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows: This invention provides a method for planning the overhead crane hoisting path in a ship's intelligent workshop, comprising: Step 1: Obtain the three-dimensional coordinates of the crane through the position sensor, scan the spatial distribution of dynamic obstacles through the laser sensor, monitor the load swing angle and hook height change through the tilt sensor, and record the material weight through the load sensor. Input the three-dimensional coordinates of the crane, the spatial distribution of dynamic obstacles, the load swing angle, the hook height change and the material weight into the physical constraint model, and output a structured environmental state dataset. The structured environmental state dataset includes the coordinate set, swing angle, hook height change value, material weight value, step size limit and safety distance matrix. Step 2: Receive the structured environmental state dataset output from Step 1. Initialize the wolf pack algorithm parameters based on the structured environmental state dataset. The wolf pack algorithm parameters include the total number of wolves, the proportion of scout wolves, and the step size factor. Calculate the obstacle distance based on the safety distance matrix in the structured environmental state dataset. Calculate the scout wolf's walking step length based on the obstacle distance. Calculate the wolf's running step length based on the swing angle and vehicle deceleration model in the structured environmental state dataset. Map the odor concentration gradient based on the hook height change value in the structured environmental state dataset. Use the odor concentration gradient to correct the alpha wolf's position. Synchronize the algorithm iteration cycle with the sensor sampling clock. Integrate the walking step length, running step length, and odor concentration gradient. Output a dynamic parameter set, which includes the walking step length, running step length, and odor concentration gradient. Step 3: Receive the dynamic parameter set output from Step 2. The industrial cloud platform pre-generates a global task queue and sends the initial parameters to the gantry via the 5G network. Based on the dynamic parameter set and the laser sensor, the coordinates of the dynamic obstacles are obtained through real-time scanning. The dynamic obstacle coordinates are encoded into the alpha wolf position offset based on the dynamic parameter set. A local optimized trajectory is generated based on the alpha wolf position offset. The alpha wolf position offset is sent back to the cloud to reconstruct the task queue. The walking mechanism is controlled to perform lateral or longitudinal movement. The lifting mechanism is controlled to dynamically adjust the lifting speed according to the swing angle. The trajectory execution status and the actual swing angle are recorded. The execution status dataset is output, which includes the trajectory completion degree and the actual swing angle. Step 4: Receive the execution status dataset output from Step 3, count the number of emergency braking operations of the vehicle from the execution status dataset, update the walk step length calculation function when the response delay exceeds the threshold, compare the actual task completion time with the preset threshold from the execution status dataset, optimize the reconnaissance wolf ratio when the difference is exceeded, analyze historical swing angle data from the execution status dataset to establish a wolf attack step length and swing mapping library, generate an optimization parameter set, the optimization parameter set includes the updated walk step length calculation function, the optimized reconnaissance wolf ratio, and the wolf attack step length and swing mapping library, feed the optimization parameter set back to Step 1, and update the physical constraint model.

[0007] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 1 includes: The vehicle's three-dimensional coordinates are continuously acquired through position sensors; A dynamic obstacle point cloud coordinate set is generated using a laser sensor; The load swing angle and hook height change values ​​are obtained through tilt sensors; The weight of the material is obtained through a load sensor; Input the three-dimensional coordinates of the crane, the point cloud coordinate set, the load swing angle, the hook height change value, and the material weight value into the physical constraint model; In the physical constraint model: the gantry control system maps the maximum deceleration to the upper limit of the algorithm step size, the swing angle threshold triggers the generation of the path curvature constraint function, the obstacle coordinate set is converted into a safety distance matrix, and the gantry three-dimensional coordinates, point cloud coordinate set, load swing angle, hook height change value, material weight value, step size upper limit and safety distance matrix are integrated to generate a structured environmental state dataset.

[0008] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 2 includes: Receive the structured environment state dataset output from step 1; Initialize the total number of wolves, the proportion of scout wolves, and the step size factor based on the structured environmental state dataset; Perform real-time parameter adjustment operations: Detecting wolf walking behavior: Obstacle distances are obtained based on the safe distance matrix in the structured environment state dataset. When the obstacle distance is less than the safe distance, the walking direction is reset and the compression step size is calculated. Wolf-like running behavior: Based on the swing angle in the structured environmental state dataset and combined with the vehicle deceleration model, when the swing angle exceeds the threshold in the physical constraint model, the reduced running step length is calculated. Alpha wolf decision-making mechanism: The odor concentration gradient is mapped based on the hook height change value in the structured environmental state dataset, and the alpha wolf position is corrected based on the odor concentration gradient; The compressed step size, reduced step size, and corrected alpha wolf position data are integrated and processed under the condition that the algorithm iteration cycle is synchronized with the sensor sampling clock, and the output includes a dynamic parameter set including the wandering step size, the running step size, and the odor concentration gradient.

[0009] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 3 includes: The industrial cloud platform receives the dynamic parameter set output in step 2 and pre-generates a global task queue based on the dynamic parameter set. Initial parameters are sent to the vehicle terminal via the 5G network; At the vehicle end: Input a dynamic parameter set, obtain dynamic obstacle coordinates based on real-time scanning of the laser sensor and integration of the dynamic parameter set, encode the dynamic obstacle coordinates into the alpha wolf position offset, generate a local optimized trajectory based on the alpha wolf position offset, and transmit the position offset back to the cloud. Reconstruct the task queue in the cloud; The traveling mechanism performs lateral or longitudinal movement, and the lifting mechanism dynamically adjusts the lifting speed according to the swing angle. It records the trajectory execution status and the actual swing angle, and outputs an execution status dataset including trajectory completion and actual swing angle.

[0010] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 4 includes: Receive the execution status dataset output from step 3; Extract the number of emergency braking operations of the vehicle from the execution status dataset, and update the walk step length calculation function when the response delay exceeds a preset threshold; Extract the actual task completion time from the execution status dataset and compare it with a preset threshold. When the difference exceeds the allowable range, optimize the reconnaissance wolf ratio. Historical swing angle data were extracted from the execution status dataset to establish a mapping library between the wolf's running stride and swing. The updated walking stride calculation function, the optimized reconnaissance wolf ratio, and the wolf sprint stride and swing mapping library are integrated into an optimized parameter set; Feedback optimization parameters to step 1, and update the path curvature constraint function and step size upper limit in the physical constraint model.

[0011] Furthermore, the ship intelligent workshop overhead crane hoisting path planning method of the present invention also includes: The rule for calculating the compressed step size in step 2 for detecting wolf movement is as follows: Step size = Dynamic coefficients obtained from the physical constraint model × × obstacle distance, k is the workshop environment factor obtained from the structured environment state dataset, and the compression step size is output to the dynamic parameter set of step 3; In step 3, the cloud receives compression step size data. When the compression step size data is less than the set threshold in the safety distance matrix, the cloud triggers a task queue priority reallocation operation and reconstructs the task queue based on the position offset.

[0012] Furthermore, the ship intelligent workshop overhead crane hoisting path planning method of the present invention also includes: The calculation rule for the reduction ratio of the wolf's running stride in step 2 is as follows: Reduction ratio = α × sin(swing angle) + β × deceleration coefficient in the vehicle deceleration model, where α and β are weighting parameters; The reduction step size is output to the dynamic parameter set in step 3; In step 4, the reduced step length data is received, and the reduced step length and historical swing angle data are input into the Wolf Rush step length and swing mapping library to generate a mapping relationship between the Wolf Rush step length change and the swing angle. The mapping relationship is sent to the crane mechanism, and the crane mechanism is controlled to adjust the lifting speed according to the mapping relationship.

[0013] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 3 of the cloud-based reconstruction of the task queue includes: Receive the position offset, receive the compression step size data output in step 2, and receive the collaborative optimization factor generated in step 4. Dynamic priority weights are calculated based on compressed step size data and collaborative optimization factors. Perform a position correction operation: The position offset is encoded as the alpha wolf position correction. The calculation of the alpha wolf position correction is constrained by the safety distance matrix. The constraint strength is positively correlated with the dynamic priority weight. The corrected position correction is then output. Send the corrected position adjustment to the Wolf Rush stride calculation step; In the process of calculating the stride length of a wolf during a running attack: Receive the corrected position correction amount, dynamically adjust the α and β weight parameters according to the position correction amount, calculate the reduction step size using the adjusted parameters, and output the reduction step size to the stability verification step. The dynamic priority weight W is calculated as follows: W = μ × compression step size + ν × collaborative optimization factor, where μ and ν are the balance coefficients obtained from the physical constraint model.

[0014] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, the calculation rule for optimizing the reconnaissance wolf ratio in step 4 is as follows: The scaling increment = γ × (actual task completion time - preset threshold), where γ is the scaling factor obtained from the physical constraint model; When the difference exceeds the allowable range, the scout wolf ratio will be updated to: original ratio + ratio increment; The updated scout wolf ratio and the corrected position amount are coupled to generate a collaborative optimization factor; The dynamic coefficients in the compression step size calculation formula are corrected based on the collaborative optimization factor; The corrected dynamic coefficients and the updated scout wolf ratio are integrated into the optimization parameter set.

[0015] Furthermore, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 2 further includes: Receive the set of optimized parameters from step 4; The updated reconnaissance wolf ratio, the updated wandering stride length calculation function, and the mapping relationship between wolf running stride length and swing were extracted from the optimized parameter set. The updated scout wolf ratio is used when initializing the wolf pack algorithm parameters; The updated walk step length calculation function is used when calculating the walk step length of the scout wolf. When calculating the wolf's running stride length, the mapping relationship between the wolf's running stride length and its swing is adopted. The walking step length, rushing step length and odor concentration gradient calculated after using updated parameters are integrated into the dynamic parameter set output; Among them, the optimized parameter set is generated in step 4 by statistically analyzing the number of emergency braking of the crane, comparing the actual completion time of the task and analyzing historical swing angle data, forming a closed-loop feedback mechanism from path execution to algorithm optimization, so that the wolf pack algorithm parameters can be adaptively adjusted according to the actual lifting operation effect.

[0016] Beneficial effects of this invention: This invention replaces manual monitoring with multi-sensor fusion for environmental perception. Laser sensors scan the spatial distribution of dynamic obstacles, combined with position sensors to construct the 3D coordinates of the overhead crane, creating a real-time workshop topology. Tilt sensors monitor the load swing angle, and data from load sensors and material weight are input into a physical constraint model to generate a structured environmental state dataset, eliminating collision risks caused by blind spots in human vision. The wolf pack algorithm parameters are dynamically initialized based on the environmental dataset. The scout wolf's walking stride is exponentially compressed according to obstacle distance, while the wolf's running stride is proportionally reduced according to the swing angle. Odor concentration gradients correct the alpha wolf's position, overcoming the global search efficiency limitations of rule-of-fact path planning. An industrial cloud platform is also included. The task queue is sent to the crane terminal via 5G network. The coordinates of dynamic obstacles are encoded as the position offset of the alpha wolf to trigger cloud rescheduling. The walking mechanism and the crane mechanism execute collaboratively based on the dynamic parameter set to solve the production delay caused by the conflict of multiple crane tasks. The execution status dataset drives closed-loop verification. The number of emergency braking is counted to update the walking step length function. The task completion time is compared to optimize the reconnaissance wolf ratio. The historical swing angle analysis establishes a step length and swing mapping library. The optimized parameter set feedback updates the physical constraint model, forming a self-reinforcing closed loop of perception, decision-making, execution and verification. This collaboratively reduces the accident rate and production delay rate in the narrow space and dynamic obstacle scenarios of the ship workshop. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0018] Figure 1 The flowchart illustrates a method for planning the hoisting path of a ship's intelligent workshop, as provided by this invention.

[0019] Figure 2 The diagram below illustrates the wolf pack algorithm hunting model for a method of overhead crane hoisting path planning in a ship intelligent workshop, as provided by this invention.

[0020] Figure 3 Experimental diagram of an improved wolf pack optimization algorithm for a ship intelligent workshop overhead crane hoisting path planning method provided by the present invention. Detailed Implementation

[0021] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0022] Please see Figures 1 to 3 The present invention provides a method for planning the overhead crane hoisting path in a ship intelligent workshop, comprising: Step 1: Obtain the three-dimensional coordinates of the crane through the position sensor, scan the spatial distribution of dynamic obstacles through the laser sensor, monitor the load swing angle and hook height change through the tilt sensor, and record the material weight through the load sensor. Input the three-dimensional coordinates of the crane, the spatial distribution of dynamic obstacles, the load swing angle, the hook height change and the material weight into the physical constraint model, and output a structured environmental state dataset. The structured environmental state dataset includes the coordinate set, swing angle, hook height change value, material weight value, step size limit and safety distance matrix. Step 2: Receive the structured environmental state dataset output from Step 1. Initialize the wolf pack algorithm parameters based on the structured environmental state dataset. The wolf pack algorithm parameters include the total number of wolves, the proportion of scout wolves, and the step size factor. Calculate the obstacle distance based on the safety distance matrix in the structured environmental state dataset. Calculate the scout wolf's walking step length based on the obstacle distance. Calculate the wolf's running step length based on the swing angle and vehicle deceleration model in the structured environmental state dataset. Map the odor concentration gradient based on the hook height change value in the structured environmental state dataset. Use the odor concentration gradient to correct the alpha wolf's position. Synchronize the algorithm iteration cycle with the sensor sampling clock. Integrate the walking step length, running step length, and odor concentration gradient. Output a dynamic parameter set, which includes the walking step length, running step length, and odor concentration gradient. Step 3: Receive the dynamic parameter set output from Step 2. The industrial cloud platform pre-generates a global task queue and sends the initial parameters to the gantry via the 5G network. Based on the dynamic parameter set and the laser sensor, the coordinates of the dynamic obstacles are obtained through real-time scanning. The dynamic obstacle coordinates are encoded into the alpha wolf position offset based on the dynamic parameter set. A local optimized trajectory is generated based on the alpha wolf position offset. The alpha wolf position offset is sent back to the cloud to reconstruct the task queue. The walking mechanism is controlled to perform lateral or longitudinal movement. The lifting mechanism is controlled to dynamically adjust the lifting speed according to the swing angle. The trajectory execution status and the actual swing angle are recorded. The execution status dataset is output, which includes the trajectory completion degree and the actual swing angle. Step 4: Receive the execution status dataset output from Step 3, count the number of emergency braking operations of the vehicle from the execution status dataset, update the walk step length calculation function when the response delay exceeds the threshold, compare the actual task completion time with the preset threshold from the execution status dataset, optimize the reconnaissance wolf ratio when the difference is exceeded, analyze historical swing angle data from the execution status dataset to establish a wolf attack step length and swing mapping library, generate an optimization parameter set, the optimization parameter set includes the updated walk step length calculation function, the optimized reconnaissance wolf ratio, and the wolf attack step length and swing mapping library, feed the optimization parameter set back to Step 1, and update the physical constraint model.

[0023] The intelligent shipbuilding workshop overhead crane hoisting path planning method provided by this invention achieves closed-loop optimization from environmental perception to path execution through multi-sensor fusion and intelligent algorithm collaboration. The core of the method lies in dynamically combining physical constraints with a wolf pack algorithm to address the path planning challenges in the confined space and dynamic obstacle scenarios of shipbuilding workshops. The overall process includes four key steps, which are closely connected through data flow, forming a self-reinforcing mechanism of perception, decision-making, execution, and verification.

[0024] In step 1, position sensors continuously acquire the 3D coordinates of the overhead crane in the workshop coordinate system. Laser sensors use time-of-flight ranging to scan the contours of dynamic obstacles, generating point cloud spatial distribution data with millimeter-level precision. Tilt sensors monitor changes in the pitch and roll angles of the hook load, while simultaneously recording hook height changes. Load sensors collect material weight values. After these raw data are input into the physical constraint model, the overhead crane control system maps the maximum deceleration parameter to the upper limit of the algorithm step size. The load swing angle safety threshold is associated with the path curvature constraint function to suppress resonance. The dynamic obstacle coordinate set is converted into a safe distance matrix using Euclidean distance calculation. Finally, a structured environmental state dataset is output, integrating the coordinate set, swing angle, hook height change value, material weight value, step size upper limit, and safe distance matrix, providing environmental topology and dynamic boundary benchmarks for path planning.

[0025] Step 2 receives the structured environmental state dataset output from Step 1 and initializes the wolf pack algorithm parameters based on the dataset, including the total number of wolves, the proportion of scout wolves, and the step size factor. During the scout wolf's roaming behavior, obstacle distances are calculated in real-time based on the safe distance matrix. When the distance is less than the safe threshold, the roaming direction is reset, and an exponential decay model is used to compress the step size to enhance obstacle avoidance response. For the wolf's attacking behavior, based on swing angle data combined with the vehicle's deceleration model, the attacking step size is proportionally reduced when the angle exceeds the limit to control the load swing amplitude. The alpha wolf decision-making mechanism maps the hook height change value to a nonlinear odor concentration gradient field, correcting the alpha wolf's position coordinates to optimize the global path direction. All parameters are synchronously integrated within the sensor sampling clock cycle, outputting a dynamic parameter set including roaming step size, attacking step size, and odor concentration gradient, achieving matching between algorithm iteration and real-time environmental changes.

[0026] Step 3 receives the dynamic parameter set output from Step 2. The industrial cloud platform pre-generates a global task queue based on the dynamic parameter set and sends the initial parameters to the edge of the gantry via the 5G network. The gantry encodes the dynamic obstacle coordinates obtained by the laser sensor in real time into the alpha position offset and generates a Bézier curve to optimize the trajectory by combining it with local environmental data. The position offset is sent back to the cloud to trigger task queue reconstruction. For example, when multiple gantry paths conflict, the Hungarian algorithm is used to dynamically reallocate priorities. The traveling mechanism executes lateral and longitudinal movement commands, and the lifting mechanism uses PID control to dynamically adjust the lifting speed based on the swing angle data. The trajectory tracking error and the actual swing angle are recorded throughout the process, and the execution status dataset, including trajectory completion and actual swing angle, is output for subsequent verification.

[0027] Step 4 receives the execution status dataset output from Step 3, counts the number of emergency braking operations of the vehicle from the dataset to evaluate obstacle avoidance response timeliness, and updates the walking step length calculation function to enhance dynamic obstacle avoidance capability when the number exceeds a threshold. It compares the actual task completion time with the preset production cycle time threshold; if the difference is too large, it optimizes the proportion of scout wolves to improve global search efficiency. It analyzes the spectral characteristics of historical swing angles, establishes a mapping relationship library between wolf attack step length and swing angle, and generates an optimization parameter set. The optimization parameter set is fed back to the physical constraint model from Step 1, updating the path curvature constraint function coefficients and the upper limit threshold of the step length, forming an adaptive adjustment closed loop from path execution to algorithm parameters.

[0028] This invention drives the wolf pack algorithm parameter iteration through environmental perception data, and performs reverse optimization of physical constraints based on the verification results, thereby collaboratively reducing accident risks and production delays in dynamic obstacles and narrow working scenarios in shipyards.

[0029] Specifically, the ship intelligent workshop overhead crane hoisting path planning method of the present invention includes step 1 as follows: The vehicle's three-dimensional coordinates are continuously acquired through position sensors; A dynamic obstacle point cloud coordinate set is generated using a laser sensor; The load swing angle and hook height change values ​​are obtained through tilt sensors; The weight of the material is obtained through a load sensor; Input the three-dimensional coordinates of the crane, the point cloud coordinate set, the load swing angle, the hook height change value, and the material weight value into the physical constraint model; In the physical constraint model: the gantry control system maps the maximum deceleration to the upper limit of the algorithm step size, the swing angle threshold triggers the generation of the path curvature constraint function, the obstacle coordinate set is converted into a safety distance matrix, and the gantry three-dimensional coordinates, point cloud coordinate set, load swing angle, hook height change value, material weight value, step size upper limit and safety distance matrix are integrated to generate a structured environmental state dataset.

[0030] Step 1 involves acquiring environmental data through multi-sensor fusion and generating a structured environmental state dataset, providing a foundation for path planning. Position sensors employ GNSS and UWB fusion positioning technology to continuously acquire millimeter-level three-dimensional coordinates of the overhead crane in the workshop coordinate system, updating the crane's position information in real time to support dynamic path adjustments. Laser sensors, based on time-of-flight ranging principles, scan the contours of dynamic obstacles such as mobile devices and material storage areas, generating a high-precision point cloud coordinate set to construct the real-time topological distribution of the workshop space and identify obstacle boundaries and safe zones. Tilt sensors monitor changes in the load's pitch and roll angles using MEMS inertial units, while simultaneously recording hook height changes, capturing the load's dynamic state to mitigate sway risks. Load sensors utilize strain gauge bridges to measure material weight, providing load mass data for calculating mechanical motion boundaries.

[0031] The collected 3D coordinates of the overhead crane, the point cloud coordinates of dynamic obstacles, the load swing angle, the hook height change value, and the material weight value are input into the physical constraint model for integrated processing. In the physical constraint model, the overhead crane control system maps the maximum deceleration parameter to the upper limit of the algorithm step size based on the motor torque characteristics, defining the mechanical constraints of the wolf pack algorithm's search range. The swing angle safety threshold triggers the generation of a path curvature constraint function, and cubic spline interpolation is used to optimize the turning trajectory to avoid resonance. The dynamic obstacle coordinate set is converted into a safety distance matrix through Euclidean distance calculation, quantifying the minimum safe distance between the obstacle and the overhead crane. Finally, all parameters are fused to generate a structured environmental state dataset, including the coordinate set, swing angle, hook height change value, material weight value, upper limit of step size, and safety distance matrix, forming a unified data interface for environmental perception and physical constraints.

[0032] The entire process involves updating the physical constraint model parameters in real time using sensor data, enabling the structured environmental state dataset to dynamically reflect changes in the workshop environment. Data streams from each sensor are collected in parallel and synchronously input into the model, eliminating blind spots in manual monitoring and providing an accurate environmental state benchmark for subsequent wolf pack algorithm initialization.

[0033] Specifically, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 2 includes: Receive the structured environment state dataset output from step 1; Initialize the total number of wolves, the proportion of scout wolves, and the step size factor based on the structured environmental state dataset; Perform real-time parameter adjustment operations: Detecting wolf walking behavior: Obstacle distances are obtained based on the safe distance matrix in the structured environment state dataset. When the obstacle distance is less than the safe distance, the walking direction is reset and the compression step size is calculated. Wolf-like running behavior: Based on the swing angle in the structured environmental state dataset and combined with the vehicle deceleration model, when the swing angle exceeds the threshold in the physical constraint model, the reduced running step length is calculated. Alpha wolf decision-making mechanism: The odor concentration gradient is mapped based on the hook height change value in the structured environmental state dataset, and the alpha wolf position is corrected based on the odor concentration gradient; The compressed step size, reduced step size, and corrected alpha wolf position data are integrated and processed under the condition that the algorithm iteration cycle is synchronized with the sensor sampling clock, and the output includes a dynamic parameter set including the wandering step size, the running step size, and the odor concentration gradient.

[0034] Step 2 receives the structured environmental state dataset output from Step 1. This dataset includes environmental state parameters such as coordinate set, swing angle, hook height change value, material weight value, step size limit, and safety distance matrix. Based on the structured environmental state dataset, the wolf pack algorithm parameters are initialized. The total number of wolves is dynamically set according to the workshop space area ratio, the proportion of reconnaissance wolves is allocated according to task complexity, and the step size factor is correlated with the maximum acceleration characteristics of the overhead crane, ensuring that the algorithm parameters match the physical environment constraints.

[0035] During real-time parameter adjustment, the reconnaissance wolf's roaming behavior uses the safe distance matrix in the structured environmental state dataset to obtain obstacle distances in real time. When the obstacle distance is less than the safe distance threshold, the roaming direction is reset, and a decay model is used to calculate the compression step size to enhance obstacle avoidance response. The wolf's attack behavior uses the swing angle data in the structured environmental state dataset, combined with the vehicle deceleration model. When the swing angle exceeds the threshold in the physical constraint model, the attack step size is reduced proportionally to suppress the load swing amplitude. The alpha wolf's decision-making mechanism uses the hook height change value in the structured environmental state dataset to generate an odor concentration gradient field through nonlinear mapping, correcting the alpha wolf's position coordinates to optimize the global path direction.

[0036] The compressed step size, reduced step size, and corrected alpha wolf position data are integrated and processed within the algorithm iteration cycle. The algorithm iteration cycle is synchronized with the sensor sampling clock, and a timed interrupt mechanism is used to ensure that parameter updates are aligned with the timing of sensor data acquisition. The output dynamic parameter set includes the walking step size, running step size, and odor concentration gradient. This dynamic parameter set serves as the input for step 3, supporting task queue generation on the industrial cloud platform and trajectory execution on the vehicle end. Each adjustment behavior is based on real-time feedback from environmental state data, forming a coherent logical chain from parameter initialization to dynamic optimization, enhancing the algorithm's adaptability to dynamic obstacle scenarios in confined spaces.

[0037] Specifically, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 3 includes: The industrial cloud platform receives the dynamic parameter set output in step 2 and pre-generates a global task queue based on the dynamic parameter set. Initial parameters are sent to the vehicle terminal via the 5G network; At the vehicle end: Input a dynamic parameter set, obtain dynamic obstacle coordinates based on real-time scanning of the laser sensor and integration of the dynamic parameter set, encode the dynamic obstacle coordinates into the alpha wolf position offset, generate a local optimized trajectory based on the alpha wolf position offset, and transmit the position offset back to the cloud. Reconstruct the task queue in the cloud; The traveling mechanism performs lateral or longitudinal movement, and the lifting mechanism dynamically adjusts the lifting speed according to the swing angle. It records the trajectory execution status and the actual swing angle, and outputs an execution status dataset including trajectory completion and actual swing angle.

[0038] Step 3 involves the industrial cloud platform receiving the dynamic parameter set output from Step 2. This set includes parameters such as walk step length, run step length, and odor concentration gradient. Based on this dynamic parameter set, the industrial cloud platform pre-generates a global task queue, optimizes task allocation order using a graph search algorithm, and generates a preliminary scheduling scheme considering the path topology between the current location of the vehicle and the target point. The initial parameters are then transmitted to the vehicle via the low-latency, high-reliability communication characteristics of the 5G network. Encryption protocols are used during network transmission to ensure data security and accurate parameter delivery.

[0039] At the vehicle end, a dynamic parameter set is input and integrated with real-time scanning data from laser sensors to dynamically acquire the coordinates of dynamic obstacles in the workshop environment. The laser sensors update obstacle position information through continuous ranging, and after fusing with the dynamic parameter set, the dynamic obstacle coordinates are encoded as a alpha position offset. The alpha position offset represents the directional adjustment amount in path planning. Based on the offset, a curve fitting algorithm is used to generate a locally optimized trajectory. The trajectory generation process considers the vehicle's kinematic constraints and load dynamics characteristics to avoid sharp turns or sudden speed changes. The position offset is transmitted back to the cloud via the communication module, triggering a cloud task queue reconstruction mechanism.

[0040] After receiving the position offset data in the cloud, the task queue is reconstructed based on the real-time workshop status. The reconstruction process employs a dynamic priority scheduling algorithm to reallocate multi-router task sequences to avoid conflicts. Once the task queue is updated, it is sent to the gantry crane, driving the traveling mechanism to perform lateral or longitudinal movement. The traveling mechanism controls the gantry crane to move along the optimized trajectory via servo motors. Simultaneously, the lifting mechanism dynamically adjusts its lifting speed based on the swing angle data, employing a closed-loop control strategy to suppress load sway. The entire trajectory execution status is recorded, including position deviation and speed curves, as well as actual swing angle data. An execution status dataset is output, including trajectory completion indicators and actual swing angle values, providing input for the verification and optimization in step 4.

[0041] Each sub-step is connected through data flow, forming a closed loop from global planning on the cloud platform to local execution on the vehicle. The logical chain is manifested as dynamic parameters driving trajectory generation, real-time feedback triggering queue reconstruction, and mechanism control ensuring motion stability.

[0042] Specifically, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 4 includes: Receive the execution status dataset output from step 3; Extract the number of emergency braking operations of the vehicle from the execution status dataset, and update the walk step length calculation function when the response delay exceeds a preset threshold; Extract the actual task completion time from the execution status dataset and compare it with a preset threshold. When the difference exceeds the allowable range, optimize the reconnaissance wolf ratio. Historical swing angle data were extracted from the execution status dataset to establish a mapping library between the wolf's running stride and swing. The updated walking stride calculation function, the optimized reconnaissance wolf ratio, and the wolf sprint stride and swing mapping library are integrated into an optimized parameter set; Feedback optimization parameters to step 1, and update the path curvature constraint function and step size upper limit in the physical constraint model.

[0043] Step 4 receives the execution status dataset output from Step 3. The execution status dataset includes parameters such as trajectory completion and actual swing angle, which are used for closed-loop verification and parameter optimization. The number of emergency braking operations of the vehicle is extracted from the execution status dataset. The number of emergency braking operations reflects the obstacle avoidance response time in path planning. When the response delay exceeds a preset threshold, a walk step length calculation function update mechanism is triggered. By adjusting the function parameters, the dynamic obstacle avoidance capability of the reconnaissance wolf is enhanced.

[0044] The actual task completion time is extracted from the execution status dataset and compared with the preset threshold for production cycle time. When the difference exceeds the allowable range, the proportion of scout wolves is optimized. Increasing the number of scout wolves improves global search efficiency and reduces task delays. Historical swing angle data is extracted from the execution status dataset, and the swing amplitude and frequency characteristics are analyzed. A mapping relationship library between the wolf's running stride length and the swing angle is established. The mapping relationship library records the correspondence between the stride length adjustment and the swing amplitude, which is used to guide the control of the crane mechanism.

[0045] The updated walking step length calculation function, the optimized reconnaissance wolf ratio, and the wolf sprint step length and swing mapping library are integrated into an optimization parameter set. This optimization parameter set is used as a feedback data packet and fed back to the physical constraint model in step 1 through the communication interface. The feedback process updates the path curvature constraint function coefficients and step length upper limit threshold in the physical constraint model, enabling the next path planning to inherit the previous optimization effect, forming an adaptive closed loop from execution verification to environmental perception.

[0046] Each sub-step is driven by the execution status dataset, with the logical chain manifested as data extraction, threshold judgment, function update, ratio optimization, mapping establishment, and integrated feedback. Technical features such as the execution status dataset, walk step length calculation function, and reconnaissance wolf ratio retain the original terminology to avoid introducing undefined concepts and enhance the coherence and feasibility of the solution. A closed-loop verification mechanism iteratively optimizes algorithm parameters using real-time data, improving the accuracy and safety of path planning in dynamic obstacle scenarios within a shipyard.

[0047] Specifically, the ship intelligent workshop overhead crane hoisting path planning method of the present invention further includes: The rule for calculating the compressed step size in step 2 for detecting wolf movement is as follows: Step size = Dynamic coefficients obtained from the physical constraint model × × obstacle distance, k is the workshop environment factor obtained from the structured environment state dataset, and the compression step size is output to the dynamic parameter set of step 3; In step 3, the cloud receives compression step size data. When the compression step size data is less than the set threshold in the safety distance matrix, the cloud triggers a task queue priority reallocation operation and reconstructs the task queue based on the position offset.

[0048] Step 2 involves calculating the compressed step size based on the wolf's wandering behavior. This calculation uses dynamic adjustments to the step size based on environmental data to enhance obstacle avoidance response. Specifically, dynamic coefficients are obtained from the physical constraint model. These coefficients, correlated with the vehicle's deceleration characteristics, are used to adjust the step size change rate. Obstacle distances are extracted from the safe distance matrix in the structured environmental state dataset, reflecting real-time obstacle proximity. The workshop environmental factor k is obtained from the structured environmental state dataset, characterizing the workshop layout complexity, such as aisle width or obstacle density, which affects the degree of step size decay. The compressed step size calculation uses an exponential decay model, multiplying the dynamic coefficient by e raised to the power of negative k and then by the obstacle distance to generate a step size value adapted to environmental risk. The calculated compressed step size is integrated into the dynamic parameter set and output to Step 3 to support subsequent path planning decisions.

[0049] In step 3, the cloud receives the compressed step size data from step 2 as part of the dynamic parameter set. The cloud compares the compressed step size data with a set threshold in the safety distance matrix, which is based on historical collision data or a preset safety standard. When the compressed step size data is less than the set threshold, it indicates a high obstacle risk, and the cloud triggers a task queue priority reassignment operation. Priority reassignment uses a dynamic scheduling algorithm, such as a weighted allocation based on urgency, to reorder the multi-vehicle task sequence. The cloud reconstructs the task queue based on position offsets, which are generated by encoding dynamic obstacle coordinates and reflect real-time environmental changes. The reconstructed task queue is then distributed to the vehicle via the 5G network, enabling timely path planning to adapt to dynamic obstacle scenarios, reducing collision risks and production delays.

[0050] Throughout the process, the compression step size calculation and cloud response form a closed loop. The step size adjustment is based on physical constraints and environmental factors, and the cloud decision is based on real-time data threshold comparison. The logical chain reflects the coordination from local step size optimization to global task scheduling, enhancing path safety and efficiency in the narrow space of the shipyard.

[0051] Specifically, the ship intelligent workshop overhead crane hoisting path planning method of the present invention further includes: The calculation rule for the reduction ratio of the wolf's running stride in step 2 is as follows: Reduction ratio = α × sin(swing angle) + β × deceleration coefficient in the vehicle deceleration model, where α and β are weighting parameters; The reduction step size is output to the dynamic parameter set in step 3; In step 4, the reduced step length data is received, and the reduced step length and historical swing angle data are input into the Wolf Rush step length and swing mapping library to generate a mapping relationship between the Wolf Rush step length change and the swing angle. The mapping relationship is sent to the crane mechanism, and the crane mechanism is controlled to adjust the lifting speed according to the mapping relationship.

[0052] Step 2, the calculation rule for the reduction ratio of the wolf-like running attack involves dynamically adjusting the step size based on real-time sensor data to optimize path execution stability. The weight parameters α and β are obtained from the physical constraint model. The α parameter is related to the sway angle and affects the weight, while the β parameter is related to the deceleration coefficient and affects the weight; both work together to adjust the reduction ratio. The sway angle data is monitored by a tilt sensor, reflecting the load sway amplitude, while the deceleration coefficient is derived from the vehicle's deceleration model, characterizing braking characteristics. The reduction ratio calculation integrates the sine value of the sway angle and the deceleration coefficient to generate a step size adjustment amount adapted to the current operating conditions, aiming to suppress load sway and improve motion smoothness.

[0053] The calculated reduction step size is output to the dynamic parameter set in step 3. The dynamic parameter set integrates parameters such as walking step size, rushing step size, and odor concentration gradient, serving as input for the industrial cloud platform to generate a global task queue. The reduction step size data is transmitted to the gantry crane via a 5G network, supporting local trajectory optimization and real-time control decisions.

[0054] In step 4, the reduced step size data is received, and the reduced step size and historical swing angle data are input into the Wolf Rush step size and swing mapping library. Historical swing angle data is extracted from the execution status dataset, recording past swing patterns. The mapping library uses regression analysis or machine learning algorithms to generate a mapping relationship between the Wolf Rush step size change and the swing angle. This relationship captures the quantitative impact of step size adjustment on swing suppression. The mapping relationship is sent to the crane control unit, and the crane dynamically adjusts its lifting speed according to the mapping relationship. Through a closed-loop control strategy, continuous optimization of load stability is achieved, reducing the risk of swing during lifting. The entire process forms a data flow from step size calculation to execution feedback, enhancing operational safety and efficiency in dynamic obstacle scenarios in shipyards.

[0055] Specifically, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 3 of the cloud-based reconstruction of the task queue includes: Receive the position offset, receive the compression step size data output in step 2, and receive the collaborative optimization factor generated in step 4. Dynamic priority weights are calculated based on compressed step size data and collaborative optimization factors. Perform a position correction operation: The position offset is encoded as the alpha wolf position correction. The calculation of the alpha wolf position correction is constrained by the safety distance matrix. The constraint strength is positively correlated with the dynamic priority weight. The corrected position correction is then output. Send the corrected position adjustment to the Wolf Rush stride calculation step; In the process of calculating the stride length of a wolf during a running attack: Receive the corrected position correction amount, dynamically adjust the α and β weight parameters according to the position correction amount, calculate the reduction step size using the adjusted parameters, and output the reduction step size to the stability verification step. The dynamic priority weight W is calculated as follows: W = μ × compression step size + ν × collaborative optimization factor, where μ and ν are the balance coefficients obtained from the physical constraint model.

[0056] Step 3 involves cloud-based reconstruction of the task queue, which integrates multiple data sources and calculates dynamic weights to optimize task scheduling response. The cloud receives position offsets, compressed step size data output from Step 2, and collaborative optimization factors generated in Step 4. Position offsets are generated from dynamic obstacle coordinate encoding, compressed step size data is output from the scout wolf's movement behavior, and collaborative optimization factors are generated in Step 4 based on the execution state dataset. Dynamic priority weights are calculated based on the compressed step size data and collaborative optimization factors. The compressed step size data reflects the urgency of obstacle avoidance, and the collaborative optimization factors characterize historical performance optimization effects. The dynamic priority weight W is calculated using a linear weighted model, satisfying W = μ × compressed step size + ν × collaborative optimization factor. μ and ν are balance coefficients obtained from the physical constraint model, used to adjust the contribution ratio of the compressed step size and the collaborative optimization factor.

[0057] When performing a position correction operation, the position offset is encoded as the alpha wolf's position correction value. The encoding process uses a coordinate transformation algorithm to map the two-dimensional or three-dimensional offset into a position correction value that the algorithm can process. A safety distance matrix is ​​applied to constrain the calculation of the alpha wolf's position correction value. The safety distance matrix is ​​obtained from a structured environment state dataset, defining a minimum safety interval. The constraint strength is positively correlated with the dynamic priority weight; the higher the dynamic priority weight, the stronger the constraint, thus achieving path safety in high-risk scenarios. The corrected position correction value is then output.

[0058] The corrected position adjustment is sent to the Wolf Rush step length calculation step, transmitting data via an internal communication interface. During the Wolf Rush step length calculation, the corrected position adjustment is received, and the weighting parameters α and β are dynamically adjusted based on the position adjustment. The position adjustment indicates the degree of environmental change, while the α and β parameters control the weights of the swing angle and deceleration coefficient in the reduction ratio calculation, respectively. The reduced step length is calculated using the adjusted parameters. The reduced step length calculation integrates real-time sensor data and adjusted parameters, and is output to the stability verification step for subsequent swing suppression control.

[0059] The entire process achieves coordination from task redistribution to path execution through a closed-loop data flow. Dynamic priority weights adjust constraint strength, and position correction feedback adjusts algorithm parameters, enhancing the system's adaptability to dynamic obstacles and reducing collision risks and production delays in shipyard hoisting operations.

[0060] Specifically, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, the calculation rule for optimizing the reconnaissance wolf ratio in step 4 is as follows: The scaling increment = γ × (actual task completion time - preset threshold), where γ is the scaling factor obtained from the physical constraint model; When the difference exceeds the allowable range, the scout wolf ratio will be updated to: original ratio + ratio increment; The updated scout wolf ratio and the corrected position amount are coupled to generate a collaborative optimization factor; The dynamic coefficients in the compression step size calculation formula are corrected based on the collaborative optimization factor; The corrected dynamic coefficients and the updated scout wolf ratio are integrated into the optimization parameter set.

[0061] Step 4 involves optimizing the calculation rules for the reconnaissance wolf ratio by dynamically adjusting algorithm parameters based on the deviation between the actual task completion time in the execution status dataset and a preset threshold, in order to improve path planning efficiency. The actual task completion time is extracted from the execution status dataset, which records the actual time taken for the overhead crane to complete the hoisting task. This time is compared with a preset threshold obtained from the physical constraint model, where the preset threshold represents the ideal task completion time benchmark. When the difference between the actual task completion time and the preset threshold exceeds the allowable range, the reconnaissance wolf ratio optimization mechanism is triggered.

[0062] The scaling increment is calculated by multiplying a scaling factor γ obtained from the physical constraint model by the difference between the actual task completion time and a preset threshold. The scaling factor γ is used to adjust the sensitivity of the scaling increment to adapt to different workshop operation rhythms. When the difference exceeds the allowable range, the reconnaissance wolf ratio is updated to the original ratio plus the scaling increment, increasing the number of reconnaissance wolves to enhance global search capabilities and reduce task delays.

[0063] The updated scout wolf ratio and the corrected position quantity are coupled, with the corrected position quantity derived from the alpha wolf position offset in step 3 or the position correction operation in step 4. The coupling process employs a linear weighting or nonlinear mapping method to generate a collaborative optimization factor. The collaborative optimization factor characterizes the synergistic effect between the scout wolf ratio and position adjustment and is used for subsequent parameter correction.

[0064] The dynamic coefficients in the compression step size calculation formula are corrected based on the collaborative optimization factor. These dynamic coefficients, obtained from the physical constraint model, affect the calculation of the reconnaissance wolf's walking step size. The correction process adjusts the magnitude of the dynamic coefficients to match the current environmental requirements. The corrected dynamic coefficients and the updated reconnaissance wolf ratio are integrated into the optimization parameter set. This optimization parameter set is sent as a feedback data packet to the physical constraint model in step 1 to update the model parameters, forming a closed-loop optimization cycle from execution verification to perception input. The entire process uses real-time data to drive parameter adjustments, enhancing the algorithm's adaptability to dynamic obstacle scenarios in ship workshops.

[0065] Specifically, in the ship intelligent workshop overhead crane hoisting path planning method of the present invention, step 2 further includes: Receive the set of optimized parameters from step 4; The updated reconnaissance wolf ratio, the updated wandering stride length calculation function, and the mapping relationship between wolf running stride length and swing were extracted from the optimized parameter set. The updated scout wolf ratio is used when initializing the wolf pack algorithm parameters; The updated walk step length calculation function is used when calculating the walk step length of the scout wolf. When calculating the wolf's running stride length, the mapping relationship between the wolf's running stride length and its swing is adopted. The walking step length, rushing step length and odor concentration gradient calculated after using updated parameters are integrated into the dynamic parameter set output; Among them, the optimized parameter set is generated in step 4 by statistically analyzing the number of emergency braking of the crane, comparing the actual completion time of the task and analyzing historical swing angle data, forming a closed-loop feedback mechanism from path execution to algorithm optimization, so that the wolf pack algorithm parameters can be adaptively adjusted according to the actual lifting operation effect.

[0066] Step 2 receives the optimized parameter set from Step 4. This optimized parameter set is generated in Step 4 by statistically analyzing the number of emergency braking operations of the overhead crane, comparing the actual task completion time, and analyzing historical swing angle data. It includes the updated reconnaissance wolf ratio, the updated wandering step length calculation function, and the mapping relationship between wolf attack step length and swing. As the output of the closed-loop feedback, the optimized parameter set enables the wolf pack algorithm parameters to be adaptively adjusted based on the actual lifting operation results.

[0067] The updated reconnaissance wolf ratio, updated walking stride length calculation function, and wolf sprint stride length-swing mapping relationship were extracted from the optimized parameter set. The updated reconnaissance wolf ratio reflects the optimization requirements for mission completion efficiency, the updated walking stride length calculation function integrates obstacle avoidance response improvements, and the wolf sprint stride length-swing mapping relationship includes stride length adjustment rules driven by historical swing data. The extraction process is based on parameter identifier matching to ensure data accuracy and consistency.

[0068] When initializing the wolf pack algorithm parameters, an updated scout wolf ratio is used. This ratio sets the proportion of scout wolves in the wolf pack and affects the overall search capability. Using the updated ratio ensures the algorithm's initial state matches the latest operating environment, improving the adaptability of the path planning starting point.

[0069] An updated walking step length calculation function is used when calculating the scout wolf's walking step length. This function is dynamically generated based on a physical constraint model and a safety distance matrix. The updated function incorporates emergency braking statistics and enhances the step length compression logic to cope with dynamic obstacles. The calculation process calls the function output in real time to optimize the walking behavior.

[0070] When calculating the wolf's sprint stride length, a mapping relationship between the sprint stride length and the swing angle is adopted. This mapping relationship establishes a correlation rule between the sprint stride length and the swing angle, which is generated based on historical swing angle analysis. After adopting the mapping relationship, the sprint stride length is dynamically adjusted according to real-time swing data, controlling the load swing amplitude and improving motion stability.

[0071] The walking step length, raid step length, and odor concentration gradient, calculated after parameter updates, are integrated into the output of a dynamic parameter set. The walking step length represents the distance traveled by the scout wolf, the raid step length represents the distance traveled by the predator wolf, and the odor concentration gradient represents the position correction amount of the alpha wolf. The integration process is completed within the algorithm iteration cycle, and the dynamic parameter set serves as the input to step 3, supporting task queue generation and trajectory execution.

[0072] The entire process forms a closed-loop feedback mechanism from path execution to algorithm optimization. The optimized parameter set is used to update the algorithm parameters in step 2, enabling the wolf pack algorithm to continuously adapt to the actual lifting operation results, thereby reducing the risk of accidents and production delays in the narrow space of ship workshops with dynamic obstacles.

[0073] This invention replaces direct manual monitoring with a multi-sensor fusion environmental perception system, addressing the collision risks caused by operator blind spots and judgment delays. Position sensors continuously acquire millimeter-level 3D coordinates of the crane, laser sensors scan dynamic obstacles to generate point cloud spatial distribution, tilt sensors monitor load swing angles and hook height changes, and load sensors record material weight data. This sensor data is input into a physical constraint model to generate a structured environmental state dataset. This dataset integrates coordinate sets, swing angles, hook height changes, material weight values, step size limits, and a safety distance matrix to construct the real-time topology and dynamic boundaries of the workshop, eliminating blind spots in manual monitoring.

[0074] The wolf pack algorithm parameters are dynamically initialized based on a structured environmental state dataset, overcoming the limitations of path planning that relies on rules of thumb. The scout wolf's walking stride length employs an exponential compression strategy based on obstacle distance; when the obstacle distance is less than a safe threshold, the direction is reset and the compressed stride length is calculated, enhancing dynamic obstacle avoidance response. The wolf's sprint stride length is reduced proportionally to the swing angle; combined with a vehicle deceleration model, the sprint stride length is adjusted when the angle exceeds the limit, suppressing load swing amplitude. The alpha wolf decision-making mechanism maps the hook height change value to the odor concentration gradient, correcting the alpha wolf's position and optimizing the global path direction, improving search efficiency in confined spaces.

[0075] The industrial cloud platform pre-generates a global task queue based on a dynamic parameter set and distributes the parameters to the gantry crane via a 5G network. Dynamic obstacle coordinates are encoded in real-time as the alpha wolf position offset, triggering cloud-based rescheduling. A locally optimized trajectory is generated using Bézier curves. The walking mechanism performs lateral and longitudinal movements, while the lifting mechanism dynamically adjusts its lifting speed based on the swing angle, resolving production delays caused by conflicts between multiple gantry crane tasks. The execution status dataset drives closed-loop verification, statistically analyzes the number of emergency braking events to update the walking step length function, compares the actual task completion time to optimize the alpha wolf ratio, and analyzes historical swing data to establish a step length-swing mapping library.

[0076] The optimized parameter set is fed back to the physical constraint model to update the path curvature constraint function and step size upper limit, forming a self-reinforcing closed loop of environmental perception, dynamic programming, and execution verification. By driving algorithm parameter iteration through sensor data, the execution results optimize the physical constraints in reverse, collaboratively reducing the accident rate and production delay rate in the narrow space and dynamic obstacle scenarios of shipyards, realizing a technological transformation from manual intervention to intelligent adaptation.

[0077] The specific implementation of this invention relates to the actual operation process of a crane hoisting path planning method in a ship intelligent workshop. In view of the problem that direct manual operation in narrow spaces and dynamic obstacle scenarios can easily lead to collisions and production delays, this invention achieves automated path planning through multi-sensor fusion and intelligent algorithm collaboration.

[0078] In a shipyard environment, position sensors continuously acquire the 3D coordinates of the overhead crane, laser sensors scan dynamic obstacles to generate a point cloud coordinate set, tilt sensors monitor the load swing angle and hook height changes, and load sensors record the material weight. After these data are input into a physical constraint model, the overhead crane control system maps the maximum deceleration to an upper limit of the algorithm step size, triggers the generation of a path curvature constraint function based on the swing angle threshold, and converts the obstacle coordinate set into a safety distance matrix. Finally, a structured environmental state dataset is generated, including the coordinate set, swing angle, hook height change, material weight, upper limit of step size, and safety distance matrix.

[0079] The wolf pack algorithm parameters are initialized based on a structured environmental state dataset, including the total number of wolves, the proportion of scout wolves, and the step size factor. The scout wolf's walking behavior calculates obstacle distances based on a safe distance matrix; when the distance is less than a safe threshold, its direction is reset and a compressed step size is calculated. The wolf's attacking behavior combines a swing angle and a vehicle deceleration model; when the angle exceeds the limit, a reduced step size is calculated. The alpha wolf decision-making mechanism corrects the alpha wolf's position by mapping the hook height change value to the odor concentration gradient. All parameters are synchronously integrated within the sensor sampling period, outputting a dynamic parameter set including walking step size, attacking step size, and odor concentration gradient.

[0080] The industrial cloud platform receives a dynamic parameter set and pre-generates a global task queue, then sends the parameters to the gantry crane via a 5G network. The gantry crane uses laser sensors to scan in real-time and acquires the coordinates of dynamic obstacles from the dynamic parameter set, encoding these coordinates as the alpha wolf position offset to generate a locally optimized trajectory. This position offset is then transmitted back to the cloud to trigger task queue reconstruction. The traveling mechanism performs lateral and longitudinal movements, while the lifting mechanism dynamically adjusts its lifting speed based on the swing angle, recording the trajectory execution status and the actual swing angle, and outputting the execution status dataset.

[0081] The system statistically analyzes the number of emergency braking events from the execution status dataset. When the response delay exceeds a threshold, it updates the walking step length calculation function. It compares the actual completion time with a preset threshold, and optimizes the reconnaissance wolf ratio when deviations are found. It analyzes historical swing angle data to establish a wolf attack step length and swing mapping library. It generates an optimized parameter set including the updated walking step length calculation function, the optimized reconnaissance wolf ratio, and the wolf attack step length and swing mapping library. This set is fed back to the physical constraint model to update the path curvature constraint function and step length upper limit, forming a closed-loop adaptive mechanism from environmental perception to algorithm optimization.

[0082] This invention uses sensor data to drive the wolf pack algorithm parameter iteration and the execution result feedback to optimize physical constraints. It enables autonomous planning and continuous optimization of the overhead crane hoisting path in dynamic obstacle and narrow space scenarios, effectively reducing the risk of accidents and production delays in ship workshops.

[0083] The technical features of this invention are explained below: The structured environmental state dataset is an integrated dataset generated by processing environmental data collected from multiple sensors through a physical constraint model. It includes parameters such as coordinate set, swing angle, hook height change value, material weight value, step size limit and safety distance matrix, providing environmental topology and dynamic boundary benchmarks for path planning.

[0084] The change in hook height is monitored by tilt sensors to measure the height displacement during load lifting and lowering, reflecting the change in the vertical position of the hook. This is used to assess the load space status and participate in the calculation of odor concentration gradient mapping.

[0085] The upper limit of the step size is generated by the vehicle control system based on the maximum deceleration parameter mapping. It defines the mechanical constraint boundary of the search range of the wolf pack algorithm and limits the maximum value of the algorithm step size to achieve motion safety.

[0086] The safety distance matrix is ​​generated by transforming the dynamic obstacle coordinate set through Euclidean distance calculation, quantifying the minimum safe distance between obstacles and the vehicle, and providing spatial relationship data support for obstacle avoidance decisions.

[0087] The wolf pack algorithm is a swarm intelligence optimization algorithm that simulates the hunting behavior of wolf packs. In this invention, it is used to solve the path planning problem and achieves global optimization by simulating the social hierarchy and cooperation mechanism of wolf packs.

[0088] The wolf pack algorithm parameters include adjustable variables such as the total number of wolves, the proportion of scout wolves, and the step size factor. These parameters are dynamically initialized based on a structured environmental state dataset to match the algorithm parameters with the physical environment constraints.

[0089] The total number of wolves is dynamically set according to the proportion of the workshop space area, which determines the total number of artificial wolves in the algorithm and affects the coverage and computational complexity of the global search.

[0090] The proportion of reconnaissance wolves is allocated based on the complexity of the mission. Setting the percentage of reconnaissance wolves in the wolf pack affects the balance between overall exploration capabilities and local development capabilities.

[0091] The step size factor is related to the maximum acceleration characteristics of the vehicle. Adjusting the search step size of the algorithm affects the convergence speed and accuracy.

[0092] The scout wolf's walking step length is dynamically generated based on the obstacle distance calculated from the safety distance matrix. When the distance is less than the safety threshold, an exponential decay model is used to compress the step length, thereby enhancing obstacle avoidance response capabilities.

[0093] The stride length of the wolf attack is calculated based on the swing angle data and the vehicle deceleration model. When the angle exceeds the limit, the stride length is reduced proportionally to suppress the load swing amplitude.

[0094] The odor concentration gradient is generated by nonlinear mapping of the hook height change value, forming a gradient field that guides the alpha wolf's position update and optimizes the global path direction.

[0095] The alpha wolf's position represents the location information of the optimal solution within the wolf pack. It is continuously updated through scent concentration gradient correction, guiding the entire wolf pack to move towards the optimal area.

[0096] The dynamic parameter set integrates algorithm output parameters such as walk step size, rush step size, and odor concentration gradient, and serves as the input data interface for the industrial cloud platform to generate task queues.

[0097] The walking stride represents the distance the scout wolf moves during the search process. It is dynamically adjusted according to the distance to obstacles to achieve adaptive stride control.

[0098] The stride length during a sprint represents the length of a wolf's stride when moving toward its prey. It is dynamically adjusted according to the swing angle to balance the requirements of movement efficiency and stability.

[0099] The global task queue is pre-generated by the industrial cloud platform based on a dynamic parameter set. Initial parameters are sent to the gantry terminal via the 5G network to schedule multi-gantry task sequences, avoid path conflicts and production delays, and support dynamic priority reallocation.

[0100] Dynamic obstacle coordinates are acquired in real time through laser sensor scanning and integration, representing the real-time position data of obstacles such as mobile devices and material stacking areas in the workshop environment. This data is encoded as the alpha wolf position offset to trigger path adjustment.

[0101] The alpha wolf position offset is generated by dynamic obstacle coordinate encoding, representing the direction correction in path planning. It is used to generate a locally optimized trajectory and transmit it back to the cloud to reconstruct the task queue, reflecting the impact of environmental changes on the path.

[0102] The local optimized trajectory is generated using a curve fitting algorithm based on the offset of the alpha wolf's position. Taking into account the vehicle's kinematic constraints and load dynamics, the turning path is optimized to avoid sharp turns or sudden speed changes, thereby improving motion smoothness.

[0103] The walking mechanism executes the lateral or longitudinal movement commands of the gantry crane, and controls the gantry crane to move along a locally optimized trajectory through servo motors to achieve precise position control and support dynamic obstacle avoidance and path tracking.

[0104] The lifting mechanism dynamically adjusts the lifting speed based on the swing angle data, and adopts a closed-loop control strategy to suppress load swing, thereby achieving stability in the lifting process and reducing the risk of swing.

[0105] The execution status dataset records the trajectory execution status and actual swing angle, including parameters such as trajectory completion and actual swing angle, which are used for subsequent closed-loop verification and parameter optimization.

[0106] The trajectory completion index evaluates the effectiveness of path execution, including positional deviation, velocity curve, and task completion progress, and is used to measure the accuracy and efficiency of path planning.

[0107] The actual swing angle is monitored by a tilt sensor, which records the changes in the load pitch and roll angles to reflect the load dynamics and is used to analyze the swing amplitude and frequency.

[0108] The number of emergency brakings is counted from the execution status dataset to reflect the timeliness of the vehicle's obstacle avoidance response. When the number of emergency brakings exceeds the threshold, the walk step length calculation function is updated to enhance the dynamic obstacle avoidance capability.

[0109] The wandering step length calculation function is dynamically generated based on the obstacle distance and safety distance matrix. It is used to calculate the wandering step length of the reconnaissance wolf. After the update, emergency braking data is incorporated to enhance the step length compression logic.

[0110] The ratio of scout wolves is optimized and adjusted based on the comparison between the actual task completion time and the preset threshold. When the difference exceeds the allowable range, the number of scout wolves is increased to improve the overall search efficiency and reduce task delays.

[0111] Historical swing angle data is extracted from the execution status dataset, recording past swing patterns for analyzing spectral characteristics and establishing a mapping relationship between step size and swing.

[0112] The Wolf Rush Stride and Swing Mapping Library is generated using regression analysis or machine learning algorithms to establish a mapping relationship between the change in rush stride length and the swing angle, which is used to guide the control of lifting speed of crane mechanisms.

[0113] The optimized parameter set integrates and updates the walking step length calculation function, the optimized reconnaissance wolf ratio, and the wolf running step length and swing mapping library, which are sent to the physical constraint model as feedback data packets to support adaptive parameter updates.

[0114] The path curvature constraint function is generated in the physical constraint model by triggering the swing angle safety threshold. It optimizes the path curvature through spline interpolation, avoids resonance, and improves path safety.

[0115] The physical constraint model constructs an environmental constraint framework by integrating multi-sensor data. It receives the 3D coordinates of the crane from position sensors, the point cloud coordinate set generated by laser sensors, the load swing angle and hook height change values ​​monitored by tilt sensors, and the material weight value recorded by load sensors. The model maps the maximum deceleration parameter of the crane control system to the upper limit of the algorithm step size, converts the swing angle safety threshold into a path curvature constraint function, and converts the obstacle coordinate set into a safety distance matrix through Euclidean distance calculation. Finally, it fuses all input data to generate a structured environmental state dataset, outputting an integrated data package including the coordinate set, swing angle, hook height change value, material weight value, upper limit of step size, and safety distance matrix, providing environmental topology and dynamic boundary benchmarks for path planning.

[0116] The vehicle deceleration model establishes a braking characteristic mapping relationship based on vehicle dynamics principles, receiving deceleration parameters from the physical constraint model and sway angle data provided by tilt sensors. By analyzing the relationship between the vehicle's motor torque characteristics and load dynamics, the model calculates a deceleration coefficient suitable for different operating conditions. This coefficient is used to adjust the reduction ratio of the wolf-like stride length. The model outputs the deceleration coefficient to the wolf-like stride behavior calculation module, which dynamically adjusts the stride length based on real-time sway angle data to suppress load sway amplitude and achieve motion stability and safety.

Claims

1. A method for planning the overhead crane hoisting path in a ship's intelligent workshop, characterized in that, include: Step 1: Obtain the three-dimensional coordinates of the crane through the position sensor, scan the spatial distribution of dynamic obstacles through the laser sensor, monitor the load swing angle and hook height change through the tilt sensor, and record the material weight through the load sensor. Input the three-dimensional coordinates of the crane, the spatial distribution of dynamic obstacles, the load swing angle, the hook height change and the material weight into the physical constraint model, and output a structured environmental state dataset. The structured environmental state dataset includes the coordinate set, swing angle, hook height change value, material weight value, step size limit and safety distance matrix. Step 2: Receive the structured environmental state dataset output from Step 1. Initialize the wolf pack algorithm parameters based on the structured environmental state dataset. The wolf pack algorithm parameters include the total number of wolves, the proportion of scout wolves, and the step size factor. Calculate the obstacle distance based on the safety distance matrix in the structured environmental state dataset. Calculate the scout wolf's walking step length based on the obstacle distance. Calculate the wolf's running step length based on the swing angle and vehicle deceleration model in the structured environmental state dataset. Map the odor concentration gradient based on the hook height change value in the structured environmental state dataset. Use the odor concentration gradient to correct the alpha wolf's position. Synchronize the algorithm iteration cycle with the sensor sampling clock. Integrate the walking step length, running step length, and odor concentration gradient. Output a dynamic parameter set, which includes the walking step length, running step length, and odor concentration gradient. Step 3: Receive the dynamic parameter set output from Step 2. The industrial cloud platform pre-generates a global task queue and sends the initial parameters to the gantry via the 5G network. Based on the dynamic parameter set and the laser sensor, the coordinates of the dynamic obstacles are obtained through real-time scanning. The dynamic obstacle coordinates are encoded into the alpha wolf position offset based on the dynamic parameter set. A local optimized trajectory is generated based on the alpha wolf position offset. The alpha wolf position offset is sent back to the cloud to reconstruct the task queue. The walking mechanism is controlled to perform lateral or longitudinal movement. The lifting mechanism is controlled to dynamically adjust the lifting speed according to the swing angle. The trajectory execution status and the actual swing angle are recorded. The execution status dataset is output, which includes the trajectory completion degree and the actual swing angle. Step 4: Receive the execution status dataset output from Step 3, count the number of emergency braking operations of the vehicle from the execution status dataset, update the walking step length calculation function when the response delay exceeds the threshold, compare the actual task completion time with the preset threshold from the execution status dataset, optimize the reconnaissance wolf ratio when the difference is exceeded, analyze historical swing angle data from the execution status dataset to establish a wolf attack step length and swing mapping library, generate an optimization parameter set, the optimization parameter set includes the updated walking step length calculation function, the optimized reconnaissance wolf ratio and the wolf attack step length and swing mapping library, feed the optimization parameter set back to Step 1, and update the physical constraint model; Step 2 includes: Receive the structured environment state dataset output from step 1; Initialize the total number of wolves, the proportion of scout wolves, and the step size factor based on the structured environmental state dataset; Perform real-time parameter adjustment operations: Detecting wolf walking behavior: Obstacle distances are obtained based on the safe distance matrix in the structured environment state dataset. When the obstacle distance is less than the safe distance, the walking direction is reset and the compression step size is calculated. Wolf-like running behavior: Based on the swing angle in the structured environmental state dataset and combined with the vehicle deceleration model, when the swing angle exceeds the threshold in the physical constraint model, the reduced running step length is calculated. Alpha wolf decision-making mechanism: The odor concentration gradient is mapped based on the hook height change value in the structured environmental state dataset, and the alpha wolf position is corrected based on the odor concentration gradient; The compressed step size, reduced step size, and corrected alpha wolf position data are integrated and processed under the condition that the algorithm iteration cycle is synchronized with the sensor sampling clock, and the output includes a dynamic parameter set including the wandering step size, the running step size, and the odor concentration gradient.

2. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 1, characterized in that, Step 1 includes: The vehicle's three-dimensional coordinates are continuously acquired through position sensors; A dynamic obstacle point cloud coordinate set is generated using a laser sensor; The load swing angle and hook height change values ​​are obtained through tilt sensors; The weight of the material is obtained through a load sensor; Input the three-dimensional coordinates of the crane, the point cloud coordinate set, the load swing angle, the hook height change value, and the material weight value into the physical constraint model; In the physical constraint model: the gantry control system maps the maximum deceleration to the upper limit of the algorithm step size, the swing angle threshold triggers the generation of the path curvature constraint function, the obstacle coordinate set is converted into a safety distance matrix, and the gantry three-dimensional coordinates, point cloud coordinate set, load swing angle, hook height change value, material weight value, step size upper limit and safety distance matrix are integrated to generate a structured environmental state dataset.

3. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 2, characterized in that, Step 3 includes: The industrial cloud platform receives the dynamic parameter set output in step 2 and pre-generates a global task queue based on the dynamic parameter set. Initial parameters are sent to the vehicle terminal via the 5G network; At the vehicle end: Input a dynamic parameter set, obtain the coordinates of dynamic obstacles based on real-time scanning of the laser sensor and integration of the dynamic parameter set, encode the dynamic obstacle coordinates into the alpha wolf position offset, generate a local optimized trajectory based on the alpha wolf position offset, and transmit the position offset back to the cloud. Reconstruct the task queue in the cloud; The traveling mechanism performs lateral or longitudinal movement, and the lifting mechanism dynamically adjusts the lifting speed according to the swing angle. It records the trajectory execution status and the actual swing angle, and outputs an execution status dataset including the trajectory completion degree and the actual swing angle.

4. The method for planning the overhead crane hoisting path in a ship intelligent workshop according to claim 3, characterized in that, Step 4 includes: Receive the execution status dataset output from step 3; Extract the number of emergency braking operations of the vehicle from the execution status dataset, and update the walk step length calculation function when the response delay exceeds a preset threshold; Extract the actual task completion time from the execution status dataset and compare it with a preset threshold. When the difference exceeds the allowable range, optimize the reconnaissance wolf ratio. Historical swing angle data were extracted from the execution status dataset to establish a mapping library between the wolf's running stride and swing. The updated walking stride calculation function, the optimized reconnaissance wolf ratio, and the wolf sprint stride and swing mapping library are integrated into an optimized parameter set; Feedback optimization parameters to step 1, and update the path curvature constraint function and step size upper limit in the physical constraint model.

5. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 4, characterized in that, Also includes: The rule for calculating the compressed step size in step 2 for detecting wolf movement is as follows: Step size = Dynamic coefficients obtained from the physical constraint model × × obstacle distance, k is the workshop environment factor obtained from the structured environment state dataset, and the compression step size is output to the dynamic parameter set of step 3; In step 3, the cloud receives compression step size data. When the compression step size data is less than the set threshold in the safety distance matrix, the cloud triggers a task queue priority reallocation operation and reconstructs the task queue based on the position offset.

6. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 5, characterized in that, Also includes: The calculation rule for the reduction ratio of the wolf's running stride in step 2 is as follows: Reduction ratio = α × sin(swing angle) + β × deceleration coefficient in the vehicle deceleration model, where α and β are weighting parameters; The reduction step size is output to the dynamic parameter set in step 3; In step 4, the reduced step length data is received, and the reduced step length and historical swing angle data are input into the Wolf Rush step length and swing mapping library to generate a mapping relationship between the Wolf Rush step length change and the swing angle. The mapping relationship is sent to the crane mechanism, and the crane mechanism is controlled to adjust the lifting speed according to the mapping relationship.

7. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 6, characterized in that, The cloud-based reconstruction task queue in step 3 includes: Receive the position offset, receive the compression step size data output in step 2, and receive the collaborative optimization factor generated in step 4. Dynamic priority weights are calculated based on compression step size data and collaborative optimization factors. Perform a position correction operation: The position offset is encoded as the alpha wolf position correction. The calculation of the alpha wolf position correction is constrained by the safety distance matrix. The constraint strength is positively correlated with the dynamic priority weight. The corrected position correction is then output. Send the corrected position adjustment to the Wolf Rush stride calculation step; In the process of calculating the stride length of a wolf during a running attack: Receive the corrected position correction amount, dynamically adjust the α and β weight parameters according to the position correction amount, calculate the reduction step size using the adjusted parameters, and output the reduction step size to the stability verification step. The dynamic priority weight W is calculated as follows: W = μ × compression step size + ν × collaborative optimization factor, where μ and ν are the balance coefficients obtained from the physical constraint model.

8. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 7, characterized in that, The calculation rule for optimizing the reconnaissance wolf ratio in step 4 is as follows: The scaling increment = γ × (actual task completion time - preset threshold), where γ is the scaling factor obtained from the physical constraint model; When the difference exceeds the allowable range, the scout wolf ratio will be updated to: original ratio + ratio increment; The updated scout wolf ratio and the corrected position amount are coupled to generate a collaborative optimization factor; The dynamic coefficients in the compression step size calculation formula are corrected based on the collaborative optimization factor; The corrected dynamic coefficients and the updated reconnaissance wolf ratio are integrated into the optimization parameter set.

9. The method for planning the overhead crane hoisting path in a ship's intelligent workshop according to claim 8, characterized in that, Step 2 also includes: Receive the set of optimized parameters from step 4; The updated reconnaissance wolf ratio, the updated wandering stride length calculation function, and the mapping relationship between wolf running stride length and swing were extracted from the optimized parameter set. The updated scout wolf ratio is used when initializing the wolf pack algorithm parameters; The updated walk step length calculation function is used when calculating the walk step length of the scout wolf. When calculating the wolf's running stride length, the mapping relationship between the wolf's running stride length and its swing is adopted. The walking step size, rushing step size, and odor concentration gradient calculated after using updated parameters are integrated into the dynamic parameter set output. Among them, the optimized parameter set is generated in step 4 by statistically analyzing the number of emergency braking of the crane, comparing the actual completion time of the task and analyzing historical swing angle data, forming a closed-loop feedback mechanism from path execution to algorithm optimization, so that the wolf pack algorithm parameters can be adaptively adjusted according to the actual lifting operation effect.

Images