High-position stacker control method and system based on movement speed optimization

Abstract

The invention discloses a high-position stacker control method and system based on movement speed optimization. According to the method, firstly, a three-dimensional laser radar is used for obtaining spatial topological data in a high-position stacker operation roadway, and then an obstacle distribution map is established; and constructing a safe obstacle avoidance operation path in combination with the obtained operation task data. By constructing a simulation operation model, operation task data and a safe obstacle avoidance operation path are imported into the simulation operation model for stacking work simulation, and a motion speed combination of all motion shafts is determined. And in the operation process of the stacker, monitoring vibration spectrum data when the stacker executes the movement speed combination in real time, and performing abnormal vibration identification. And finally, optimizing the motion speed combination according to an abnormal vibration identification result to obtain a motion speed optimization strategy. The operation safety and stability of the high-position stacking machine can be effectively improved, the operation efficiency is improved, and the high-position stacking machine has wide application prospects in the fields of warehouse logistics and the like.

Description

Technical Field

[0001] This invention relates to the field of equipment control technology, and in particular to a high-level stacker crane control method and system based on motion speed optimization. Background Technology

[0002] High-bay stacker cranes are widely used in modern warehousing and logistics for automated storage and retrieval operations on high-rise shelves. To improve operational efficiency and safety, stacker cranes need to complete tasks at high speed and with precision in complex environments, avoiding collisions with obstacles. However, as operating speeds increase, the mechanical system may generate significant vibrations, leading to equipment instability, affecting operational accuracy, and even damaging goods.

[0003] Existing control methods for high-level stacker cranes primarily focus on static obstacle avoidance and path planning, but they still fall short in real-time obstacle avoidance and speed optimization for dynamic obstacles. Traditional path planning methods are slow to react to moving obstacles, making timely obstacle avoidance difficult. Furthermore, during high-speed operation, stacker cranes are prone to vibration due to uneven structural stress and frequent trajectory changes, which reduces work efficiency and increases the risk of equipment failure.

[0004] Current technologies for optimizing vibration control and path planning often consider these factors individually, lacking comprehensive optimization across multiple factors. Significant room for improvement remains, particularly in coordinating movement speed, path planning, and vibration control. A key technical challenge for improving the efficiency and safety of stacker cranes is how to optimize movement speed by combining dynamic obstacle avoidance and vibration monitoring while ensuring safety.

[0005] This invention proposes a high-level stacker crane control method and system based on motion speed optimization. By real-time monitoring of vibration spectrum and dynamic obstacle position, and combining task data and mechanical model to optimize motion speed combination, the stacker crane's working efficiency is improved while ensuring operational safety. Summary of the Invention

[0006] To address at least one of the aforementioned technical problems, this invention proposes a high-level stacker crane control method and system based on motion speed optimization.

[0007] The first aspect of this invention provides a high-level stacker crane control method based on motion speed optimization, comprising: Spatial topology data of the high-level stacker crane's operating aisle is acquired using a three-dimensional lidar, and an obstacle distribution map is established based on the spatial topology data. Obtain the operation task data of the high-level stacker crane, and construct a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and operation task data; A simulation operation model of the high-level stacker crane is constructed. The operation task data and the safe obstacle avoidance operation path are imported into the simulation operation model to simulate the stacking operation and determine the combination of motion speeds of each motion axis of the high-level stacker crane. The vibration spectrum data of the high-level stacker crane when it executes the motion speed combination is monitored in real time, and abnormal vibration is identified based on the vibration spectrum data to obtain the abnormal vibration identification result; Based on the abnormal vibration identification results, the motion speed combination is optimized to obtain a motion speed optimization strategy.

[0008] In this solution, the step of acquiring spatial topology data within the high-level stacker crane's operating aisle using a three-dimensional lidar, and then establishing an obstacle distribution map based on the spatial topology data, specifically involves: The high-level stacker crane's operating aisle is scanned using a 3D LiDAR to acquire a 3D point cloud data stream. The 3D point cloud data stream is then spatiotemporally aligned to obtain preprocessed 3D point cloud data. Based on the structural characteristics of fixed racks in the aisle of high-level stacker cranes, multi-scale feature analysis is performed on preprocessed three-dimensional point cloud data to extract the normal vector distribution characteristics, surface curvature and density gradient change parameters of the point cloud, and to construct the feature weight matrix of the fixed rack structure. The preprocessed 3D point cloud data is divided into local regions using a sliding window mechanism. Combined with historical topological data of the shelf installation location, the statistical distribution difference of point cloud features in each local region is calculated to generate a segmentation threshold that distinguishes between fixed structures and obstacles. A dynamic region growing algorithm based on feature weight matrix is ​​used to iteratively aggregate continuous point cloud clusters that meet the structural characteristics of fixed shelves according to the segmentation threshold, so as to obtain fixed shelf segmentation point cloud data. After the continuous point cloud clusters are iteratively aggregated to complete, the remaining point cloud is marked as an obstacle candidate set. The spatial position difference of the obstacle candidate set is calculated for the point cloud data of multiple consecutive frames to determine the motion continuity of the obstacle candidate set. When the obstacle candidate set has motion continuity, the residual point cloud data with motion continuity in multiple consecutive frames of point cloud data is labeled as moving obstacle segmentation point cloud data, and the remaining point cloud data is labeled as stationary obstacle segmentation point cloud data. Based on the continuous point cloud data of fixed racks, moving obstacles, and stationary obstacles, a dynamic spatial topology of the high-level stacker crane's operating aisle is constructed. Based on the dynamic spatial topology, the changes in obstacle positions are determined, and an obstacle distribution map is constructed.

[0009] In this solution, the step of acquiring the operational task data of the high-level stacker crane and constructing a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and operational task data specifically involves: Acquire the operation task data of the high-level stacker crane. The operation task data includes one or more target storage location coordinates, fork load weight, cargo access priority parameters, and cargo position in the forks. Generate an initial operation path tree based on the target storage location coordinates, cargo access priority parameters, and the position of stationary obstacles in the obstacle distribution map. The spatial change trajectory of moving obstacles in the working roadway is obtained according to the obstacle distribution map. Dynamic obstacle avoidance monitoring nodes are inserted into the initial working path tree according to the spatial change trajectory of moving obstacles. The dynamic obstacle avoidance monitoring nodes use a sliding window mechanism to predict the spatiotemporal conflict of moving obstacles. When a moving obstacle overlaps with the current path segment of the high-level stacker crane in time and space, the three adjacent path nodes at the end of the path segment are extracted to construct a local path replanning area. The path deviation compensation angle is calculated based on the projection component of the obstacle's motion velocity vector, and alternative path branches containing avoidance margin are generated. Based on the cargo access priority parameters in the task data, determine the cargo access urgency parameters and construct a dynamic priority evaluation model. If the path length increment of the alternative path branch exceeds the tolerance threshold corresponding to the urgency parameter, trigger the path tree pruning strategy to delete the alternative branches that conflict with the priority. When a moving obstacle is detected to have left the operating path coverage area of ​​the high-level stacker crane, the optimal convergence path is regenerated based on the spatial geometric relationship between the current position of the high-level stacker crane and the coordinates of the target cargo location using an inverse kinematics algorithm, and the rate of change of joint acceleration during the path convergence process is constrained within a preset safety range. By iteratively executing dynamic obstacle avoidance monitoring and path tree optimization operations until all target cargo location coordinates have been accessed and the path conflict detection result shows no collision risk, the final verified safe obstacle avoidance operation path is output.

[0010] In this solution, the construction of a simulation operation model for the high-level stacker crane involves importing the task data and safety obstacle avoidance path into the simulation operation model to simulate stacking operations and determine the motion speed combinations of each motion axis of the high-level stacker crane. Specifically: Based on the mechanical structure parameters of the high-level stacker crane, a three-dimensional physical model including a traveling mechanism, a lifting mechanism, and a fork extension mechanism is constructed according to the mechanical structure parameters. The three-dimensional physical model is then integrated using computer mechanics simulation software to construct a simulation operation model of the high-level stacker crane. Based on the trajectory curvature distribution characteristics extracted from the aforementioned safe obstacle avoidance operation path, and combined with the maximum acceleration constraint parameters of each motion axis of the high-level stacker crane, a dynamic constraint network containing the continuity equation of joint motion trajectory and emergency stop buffer boundary is constructed. Based on the fork load weight and cargo position distribution in the task data, the influence weight of the stacker crane's center of gravity offset on motion stability is calculated, and dynamic stability speed compensation coefficients for different path segments are generated in the dynamic constraint network. The safe obstacle avoidance operation path is discretized into a sequence of trajectory points containing timestamps. The path is divided into segments based on the curvature change detection results between trajectory points. Within each segment, the inverse dynamics algorithm is used to iteratively solve for the multi-axis velocity combination candidate set that meets the requirements of dynamic constraint network and dynamic stability velocity compensation coefficient. A particle swarm optimization algorithm is introduced, and the cargo stacking efficiency of the high-level stacker crane is used as the optimization objective of the particle swarm optimization algorithm. The algorithm searches the multi-axis speed combination candidate set and outputs the multi-axis speed combination candidate set with the highest cargo stacking efficiency, thus obtaining the motion speed combination of each motion axis of the high-level stacker crane.

[0011] In this solution, the vibration spectrum data of the high-level stacker crane during the execution of the motion speed combination is monitored in real time. Abnormal vibration is identified based on the vibration spectrum data to obtain the abnormal vibration identification result. Specifically: Vibration time-domain signals of the high-level stacker crane when executing the motion speed combination are obtained based on vibration sensors. After wavelet noise reduction processing of the vibration time-domain signals, time-frequency conversion operation is performed to generate vibration spectrum data. Based on the vibration spectrum data, the vibration frequency distribution characteristics are extracted. Based on the vibration frequency distribution characteristics, the vibration spectrum data is divided into frequency bands. The mean amplitude, peak frequency, and harmonic distortion rate in each frequency band are extracted as vibration state feature vectors. The vibration feature difference index is constructed by combining the theoretical safe spectrum corresponding to the current movement speed combination of the high-level stacker crane. A sliding time window is used to splice multiple consecutive sets of vibration state feature vectors in time series. A pre-trained stacker crane vibration modal classification model is used to perform pattern matching on the spliced ​​multidimensional feature matrix to identify abnormal frequency band distribution patterns that exceed the theoretical safe spectrum. A frequency domain fault fingerprint is generated based on the frequency components and energy diffusion trend of the abnormal frequency band distribution pattern. The frequency domain fault fingerprint is then input into a fault diagnosis model based on a convolutional neural network to locate the vibration source. The output includes abnormal vibration level, spectral offset, vibration location, and vibration amplitude.

[0012] In this solution, the optimization of the motion speed combination based on the abnormal vibration identification result to obtain a motion speed optimization strategy is specifically as follows: The ratio of the mean amplitude of the high-frequency vibration band to the mean amplitude of the low-frequency vibration band in the abnormal vibration identification results is extracted as the vibration energy distribution parameter. When the vibration energy distribution parameter exceeds the first preset safety threshold, the target motion axis that generates abnormal vibration is located, and the speed reduction is calculated based on the product of the trajectory curvature change rate of the current path segment of the target motion axis and the fork load weight. If the harmonic distortion rate of the high-frequency vibration band exceeds the second preset safety threshold, the acceleration time-domain signal of the target motion axis is integrated to obtain the accumulated vibration energy, and the acceleration attenuation coefficient is generated based on the ratio of the accumulated vibration energy to the maximum load capacity of the fork extension mechanism. The motion speed combination is adjusted according to the speed reduction amount and acceleration attenuation coefficient. The target motion axis with abnormal vibration and the safe obstacle avoidance operation path are decelerated until the running speed of the target motion axis reaches the speed reduction amount of the speed reduction compensation, thus obtaining the motion speed optimization strategy.

[0013] A second aspect of the present invention also provides a high-level stacker crane control system based on motion speed optimization. The system includes a memory and a processor. The memory includes a high-level stacker crane control method program based on motion speed optimization. When the processor executes the high-level stacker crane control method program based on motion speed optimization, it performs the following steps: Spatial topology data of the high-level stacker crane's operating aisle is acquired using a three-dimensional lidar, and an obstacle distribution map is established based on the spatial topology data. Obtain the operation task data of the high-level stacker crane, and construct a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and operation task data; A simulation operation model of the high-level stacker crane is constructed. The operation task data and the safe obstacle avoidance operation path are imported into the simulation operation model to simulate the stacking operation and determine the combination of motion speeds of each motion axis of the high-level stacker crane. The vibration spectrum data of the high-level stacker crane when it executes the motion speed combination is monitored in real time, and abnormal vibration is identified based on the vibration spectrum data to obtain the abnormal vibration identification result; Based on the abnormal vibration identification results, the motion speed combination is optimized to obtain a motion speed optimization strategy.

[0014] This invention discloses a control method and system for a high-level stacker crane based on motion speed optimization. The method first uses a 3D lidar to acquire spatial topology data within the high-level stacker crane's operating aisle, thereby establishing an obstacle distribution map. Combined with the acquired task data, a safe obstacle avoidance path is constructed. A simulation operation model is built, importing the task data and the safe obstacle avoidance path to simulate stacking operations and determine the motion speed combinations for each axis. During stacker crane operation, the vibration spectrum data when executing this motion speed combination is monitored in real time for abnormal vibration identification. Finally, based on the abnormal vibration identification results, the motion speed combinations are optimized to obtain an optimized motion speed strategy. This invention effectively improves the safety and stability of high-level stacker crane operation, increases operational efficiency, and has broad application prospects in warehousing and logistics. Attached Figure Description

[0015] Figure 1 A flowchart of a high-level stacker crane control method based on motion speed optimization according to the present invention is shown; Figure 2 A flowchart illustrating the abnormal vibration identification results obtained by this invention is shown. Figure 3 The flowchart illustrating the motion speed optimization strategy obtained by this invention is shown. Figure 4 A block diagram of a high-level stacker control system based on motion speed optimization according to the present invention is shown. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0018] Figure 1 A flowchart of a high-level stacker crane control method based on motion speed optimization according to the present invention is shown.

[0019] like Figure 1 As shown, the first aspect of the present invention provides a high-level stacker crane control method based on motion speed optimization, comprising: S102, acquire spatial topology data of the high-level stacker crane's operating aisle through three-dimensional lidar, and establish an obstacle distribution map based on the spatial topology data; S104, Obtain the operation task data of the high-level stacker crane, and construct a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and the operation task data; S106, Construct a simulation operation model of the high-level stacker crane, import the operation task data and the safe obstacle avoidance operation path into the simulation operation model to simulate stacking work, and determine the combination of motion speeds of each motion axis of the high-level stacker crane; S108, Real-time monitoring of vibration spectrum data when the high-level stacker crane executes the motion speed combination, and abnormal vibration identification based on the vibration spectrum data to obtain abnormal vibration identification results; S110, Based on the abnormal vibration identification results, the motion speed combination is optimized to obtain a motion speed optimization strategy.

[0020] It should be noted that, through the dynamic spatial scanning capability of 3D LiDAR, a precise 3D topological model of the shelving structure and obstacles within the aisle is constructed, significantly improving the real-time positioning accuracy of obstacles in complex environments; based on multi-scale feature analysis and dynamic region growing algorithms, fixed shelving and moving obstacles are effectively distinguished, enhancing the dynamic perception capability of temporary obstacles; combined with path planning technology based on dynamic priority evaluation and inverse kinematics, a dynamic balance between moving obstacle avoidance and emergency task execution is achieved, shortening the path replanning response time; through physical simulation and dynamic constraint modeling, the speed coordination control of each motion axis is optimized, improving operational efficiency while ensuring mechanical stability; using vibration spectrum analysis and deep learning models, high-frequency abnormal vibration characteristics are accurately identified, enabling early warning of potential mechanical structure failures; based on vibration source localization and adaptive adjustment of motion parameters, vibration energy propagation is effectively suppressed, reducing the risk of mechanical wear of key components.

[0021] According to an embodiment of the present invention, the step of acquiring spatial topology data within the operating aisle of the high-level stacker crane using a three-dimensional lidar, and establishing an obstacle distribution map based on the spatial topology data, specifically includes: The high-level stacker crane's operating aisle is scanned using a 3D LiDAR to acquire a 3D point cloud data stream. The 3D point cloud data stream is then spatiotemporally aligned to obtain preprocessed 3D point cloud data. Based on the structural characteristics of fixed racks in the aisle of high-level stacker cranes, multi-scale feature analysis is performed on preprocessed three-dimensional point cloud data to extract the normal vector distribution characteristics, surface curvature and density gradient change parameters of the point cloud, and to construct the feature weight matrix of the fixed rack structure. It should be noted that in the process of constructing the feature weight matrix of the fixed shelving structure, this invention adopts a parameter sensitivity evaluation method based on principal component analysis for the quantitative contribution allocation of normal vector distribution characteristics, surface curvature, and density gradient change parameters. First, through statistical analysis of standard point cloud samples of the shelving structure, the normal vector dispersion index is calculated: within the local neighborhood, the variance of the cosine of the angle between the normal vectors of each point and its neighboring points is used as the normal consistency coefficient, reflecting the regularity of the shelving planar structure; second, based on the moving least squares method to fit the local surface, the absolute difference between the Gaussian curvature and the average curvature is calculated as the surface curvature abrupt change intensity, used to characterize the geometric features at the connection between the shelving beams and uprights; for the density gradient change parameter, the difference operation of the number of unit point clouds after octree spatial partitioning is used to calculate the density change gradient magnitude along the vertical and horizontal directions of the shelving. Subsequently, the original values ​​of the three parameters are mapped to the [0,1] interval using the max-min normalization method to eliminate dimensional differences. Principal component analysis is used to decompose the normalized multidimensional features into covariance matrix, calculate the variance contribution rate of each parameter in the shelf feature expression, and then weight and fuse the contribution rate with the structural stability parameters in the historical topological data of the shelf installation location to generate a feature weight vector.

[0022] The preprocessed 3D point cloud data is divided into local regions using a sliding window mechanism. Combined with historical topological data of the shelf installation location, the statistical distribution difference of point cloud features in each local region is calculated to generate a segmentation threshold that distinguishes between fixed structures and obstacles. A dynamic region growing algorithm based on feature weight matrix is ​​used to iteratively aggregate continuous point cloud clusters that meet the structural characteristics of fixed shelves according to the segmentation threshold, so as to obtain fixed shelf segmentation point cloud data. After the continuous point cloud clusters are iteratively aggregated to complete, the remaining point cloud is marked as an obstacle candidate set. The spatial position difference of the obstacle candidate set is calculated for the point cloud data of multiple consecutive frames to determine the motion continuity of the obstacle candidate set. When the obstacle candidate set has motion continuity, the residual point cloud data with motion continuity in multiple consecutive frames of point cloud data is labeled as moving obstacle segmentation point cloud data, and the remaining point cloud data is labeled as stationary obstacle segmentation point cloud data. Based on the continuous point cloud data of fixed racks, moving obstacles, and stationary obstacles, a dynamic spatial topology of the high-level stacker crane's operating aisle is constructed. Based on the dynamic spatial topology, the changes in obstacle positions are determined, and an obstacle distribution map is constructed.

[0023] It should be noted that by constructing a feature weight matrix that integrates multi-dimensional geometric features such as normal vector distribution, surface curvature, and density gradient, the recognition accuracy of fixed shelf structures is significantly improved, avoiding shelf missegmentation caused by environmental noise or lighting changes. A dynamic region growing algorithm is used to iteratively aggregate continuous point cloud clusters that conform to shelf characteristics, effectively removing interfering point clouds attached to the shelf surface and ensuring the integrity and boundary clarity of fixed shelf segmentation. A differential calculation method based on residual point cloud motion continuity detection accurately distinguishes between moving obstacles (such as shuttles and AGV equipment) and stationary obstacles (such as temporarily stacked goods), eliminating instantaneous noise interference through multi-frame point cloud spatial position tracking. The resulting dynamic spatial topology can reflect the trajectory changes of moving obstacles and the distribution status of stationary obstacles in real time. The constructed obstacle distribution map provides millimeter-level precision environmental perception data support for path planning, while significantly reducing the risk of path collisions caused by point cloud misclassification in traditional methods, enhancing the operational safety and continuity of stacker cranes in complex dynamic environments. The structural features include the geometric shape of the fixed shelves; the obstacle distribution map is a dynamic diagram of the positional changes of obstacles in the working lane over time.

[0024] According to an embodiment of the present invention, the step of acquiring the operational task data of the high-level stacker crane and constructing a safe obstacle avoidance operational path for the high-level stacker crane based on the obstacle distribution map and the operational task data specifically includes: Acquire the operation task data of the high-level stacker crane. The operation task data includes one or more target storage location coordinates, fork load weight, cargo access priority parameters, and cargo position in the forks. Generate an initial operation path tree based on the target storage location coordinates, cargo access priority parameters, and the position of stationary obstacles in the obstacle distribution map. The spatial change trajectory of moving obstacles in the working roadway is obtained according to the obstacle distribution map. Dynamic obstacle avoidance monitoring nodes are inserted into the initial working path tree according to the spatial change trajectory of moving obstacles. The dynamic obstacle avoidance monitoring nodes use a sliding window mechanism to predict the spatiotemporal conflict of moving obstacles. It should be noted that the initial operation path tree is a multi-branch path topology structure generated based on the target cargo location coordinates and task priorities of the high-level stacker crane. Its generation process includes: obtaining the target cargo location coordinates and priority parameters by parsing the operation instructions; sampling and expanding path branches in the stacker crane's movement space using a priority-guided improved RRT* algorithm; the multi-branch path topology structure refers to a set of multiple candidate paths organized in a tree structure during the high-level stacker crane's path planning. Specifically, the current position of the stacker crane is the root node, the target cargo location coordinates are the leaf nodes, and intermediate points generated during path planning are the branch nodes. All nodes are connected by directed edges, representing the stacker crane's movement trajectory. The dynamic obstacle avoidance monitoring node is a virtual monitoring unit embedded in the path tree, used to track the movement trajectory of moving obstacles in real time and predict their spatiotemporal conflicts with the stacker crane's path. Specifically, it includes extracting the spatial change trajectory of moving obstacles based on the obstacle distribution map, including their current position, speed, and direction. Through differential calculation of multi-frame point cloud data, it predicts the movement range of moving obstacles in the future time period. The predicted trajectory of moving obstacles is spatiotemporally matched with each path segment in the path tree to identify potential conflict paths. The start and end nodes of the conflicting path segments are determined as insertion points for dynamic obstacle avoidance monitoring nodes. Monitoring parameters are configured for each dynamic obstacle avoidance monitoring node, including monitoring range, monitoring period, and conflict threshold. Dynamic obstacle avoidance monitoring nodes are inserted between the start and end nodes of the conflicting path segments, forming sub-branches of the path tree. The monitoring nodes are associated with other nodes in the path tree to ensure the continuity and consistency of path planning.

[0025] When a moving obstacle overlaps with the current path segment of the high-level stacker crane in time and space, the three adjacent path nodes at the end of the path segment are extracted to construct a local path replanning area. The path deviation compensation angle is calculated based on the projection component of the obstacle's motion velocity vector, and alternative path branches containing avoidance margin are generated. Based on the cargo access priority parameters in the task data, determine the cargo access urgency parameters and construct a dynamic priority evaluation model. If the path length increment of the alternative path branch exceeds the tolerance threshold corresponding to the urgency parameter, trigger the path tree pruning strategy to delete the alternative branches that conflict with the priority. It should be noted that by mapping cargo access priority parameters to urgency parameters, a quantitative correlation is established between task execution timeliness and path optimization objectives, ensuring the shortest possible timeliness for high-priority task paths. The dynamic priority evaluation model achieves a dynamic balance between path length increments and task urgency through a tolerance threshold mechanism, avoiding time waste caused by excessive obstacle avoidance in low-priority task paths. The extraction of three adjacent path nodes at the end of the path segment to construct a local path replanning zone, where the three path nodes define the start, intermediate, and target points of the local path, ensures a geometrically smooth transition between the replanned path segment and the original path, avoiding sharp turns or abrupt path changes.

[0026] When a moving obstacle is detected to have left the operating path coverage area of ​​the high-level stacker crane, the optimal convergence path is regenerated based on the spatial geometric relationship between the current position of the high-level stacker crane and the coordinates of the target cargo location using an inverse kinematics algorithm, and the rate of change of joint acceleration during the path convergence process is constrained within a preset safety range. It should be noted that after a moving obstacle leaves the path coverage area, the stacker crane needs to quickly return to the original task path to avoid task interruption or delay. The inverse kinematics algorithm can directly calculate the optimal motion parameters for each motion axis based on the spatial geometric relationship between the current stacker crane position and the target storage location coordinates, significantly shortening the path recovery time. During path recovery, the rate of change of joint acceleration directly affects the stability of the stacker crane's movement and the safety of its mechanical structure. By constraining the rate of change of joint acceleration within a preset safety range, mechanical shocks and vibrations caused by rapid acceleration or deceleration are avoided, extending the equipment's service life.

[0027] By iteratively executing dynamic obstacle avoidance monitoring and path tree optimization operations until all target cargo location coordinates have been accessed and the path conflict detection result shows no collision risk, the final verified safe obstacle avoidance operation path is output.

[0028] It should be noted that by combining dynamic obstacle avoidance monitoring nodes with a sliding window mechanism, the spatiotemporal trajectory prediction of moving obstacles and real-time detection of path conflicts are achieved, enabling the stacker crane to have millisecond-level response capabilities in complex dynamic environments. The alternative branches generated based on the path deviation compensation angle quantify the avoidance margin through geometric projection relationships, ensuring a safe distance while avoiding severe path jitter. The dynamic priority evaluation model dynamically correlates the urgency of tasks with the path increment threshold, ensuring the shortest timeliness of high-priority task paths while allowing low-priority tasks to flexibly avoid obstacles. The inverse kinematics algorithm combined with the convergent path generation based on joint acceleration constraints balances path optimality and mechanical motion stability, avoiding structural damage caused by sudden stops and starts. The iterative optimization mechanism forms a multi-objective optimization solution that balances efficiency and safety through multiple rounds of path tree pruning and replanning.

[0029] According to an embodiment of the present invention, the construction of a simulation operation model for a high-level stacker crane involves importing the task data and the safe obstacle avoidance operation path into the simulation operation model to simulate stacking operations and determine the motion speed combination of each motion axis of the high-level stacker crane, specifically as follows: Based on the mechanical structure parameters of the high-level stacker crane, a three-dimensional physical model including a traveling mechanism, a lifting mechanism, and a fork extension mechanism is constructed according to the mechanical structure parameters. The three-dimensional physical model is then integrated using computer mechanics simulation software to construct a simulation operation model of the high-level stacker crane. Based on the trajectory curvature distribution characteristics extracted from the aforementioned safe obstacle avoidance operation path, and combined with the maximum acceleration constraint parameters of each motion axis of the high-level stacker crane, a dynamic constraint network containing the continuity equation of joint motion trajectory and emergency stop buffer boundary is constructed. It should be noted that the dynamic constraint network is a mathematical modeling framework based on path curvature characteristics and mechanical motion constraints. Its core is to coordinate the velocity control of each motion axis through trajectory continuity equations and emergency stop safety boundaries. The construction process includes: first, extracting the curvature distribution characteristics of the safe obstacle avoidance path, identifying high-curvature regions (such as turns and obstacle avoidance points) and low-curvature regions (such as straight sections); combining the maximum acceleration parameters of each motion axis, establishing joint motion trajectory continuity equations (such as cubic spline interpolation equations) to ensure that the velocity and acceleration of adjacent path segments are continuous and without abrupt changes; simultaneously, calculating the emergency stop buffer distance based on the maximum acceleration, setting dynamic safety boundaries in areas of abrupt path curvature changes to limit the rate of acceleration change and prevent mechanical impact; finally, integrating the continuity equations and safety boundaries of the path segment intervals into a constraint network.

[0030] Based on the fork load weight and cargo position distribution in the task data, the influence weight of the stacker crane's center of gravity offset on motion stability is calculated, and dynamic stability speed compensation coefficients for different path segments are generated in the dynamic constraint network. It should be noted that the dynamic stability speed compensation coefficient is an adjustment parameter dynamically calculated based on the fork load weight and cargo position distribution, used to offset the impact of stacker crane center of gravity offset on motion stability. It is implemented as follows: First, based on the fork load weight and cargo position distribution on the forks, the real-time center of gravity offset of the stacker crane is calculated; then, combining the maximum load torque of the motion axis and the curvature characteristics of the path segment, a speed correction factor matching the current load state is generated; this coefficient reduces the motion axis speed in high-curvature path segments or when the center of gravity offset is large, while adjusting the coordination ratio of other axes, so that the stacker crane maintains torque balance during acceleration, turning, or emergency stop, thereby suppressing vibration and avoiding the risk of tipping over. This coefficient is embedded in the dynamic constraint network as one of the boundary conditions for inverse dynamics solution.

[0031] The safe obstacle avoidance operation path is discretized into a sequence of trajectory points containing timestamps. The path is divided into segments based on the curvature change detection results between trajectory points. Within each segment, the inverse dynamics algorithm is used to iteratively solve for the multi-axis velocity combination candidate set that meets the requirements of dynamic constraint network and dynamic stability velocity compensation coefficient. A particle swarm optimization algorithm is introduced, and the cargo stacking efficiency of the high-level stacker crane is used as the optimization objective of the particle swarm optimization algorithm. The algorithm searches the multi-axis speed combination candidate set and outputs the multi-axis speed combination candidate set with the highest cargo stacking efficiency, thus obtaining the motion speed combination of each motion axis of the high-level stacker crane.

[0032] It should be noted that after solving for the candidate set of multi-axis velocity combinations that satisfy the dynamic constraint network using the inverse kinematics algorithm, this invention introduces a particle swarm optimization algorithm to globally optimize the candidate set in order to further coordinate the collaborative efficiency of each motion axis and balance mechanical stability and operational timeliness. Specifically, although the inverse kinematics solution can ensure that the velocity combinations of each motion axis satisfy the geometric constraints and dynamic boundary conditions of path tracking, the efficiency of cargo stacking in warehousing operations is affected by multiple factors such as the timing matching degree of multi-axis motion, the uniformity of emergency stop buffer margin distribution, and the balance of energy consumption. A single inverse kinematics solution cannot take into account the multi-dimensional optimization objectives. Therefore, this method encodes the candidate velocity combinations as individual position vectors in the particle swarm, uses the reciprocal of the timing deviation of the fork reaching the target cargo position as the core index of the fitness function, and introduces a trajectory smoothness factor (based on the integral of the rate of change of acceleration) and an energy consumption balance factor (based on the power variance of each axis) to construct a multi-objective fitness evaluation model. During the algorithm iteration process, each particle updates its velocity component based on the weighted vector sum of its individual historical best position and the group's best position, adjusting the balance between global exploration and local exploitation capabilities through dynamic inertia weighting. When the algorithm converges, the position vector of the particle with the highest fitness value is output as the optimized multi-axis velocity combination. This combination, while ensuring the stable operation of the mechanical system, minimizes the stacking operation cycle and reduces the wear rate of joint mechanisms. By combining the particle swarm optimization algorithm with inverse kinematics solution, multi-dimensional collaborative optimization of the motion parameters of the high-level stacker crane under complex constraints is achieved, effectively solving the technical contradiction of balancing efficiency optimization and stability assurance in traditional methods.

[0033] It should be noted that a three-dimensional physical model and dynamic constraint network of the high-level stacker crane are constructed to simulate its operation under complex paths. First, a physical model including the traveling mechanism, lifting mechanism, and fork extension mechanism is constructed based on mechanical structural parameters to ensure simulation accuracy. Second, a dynamic constraint network is constructed by combining the path curvature distribution characteristics and the maximum acceleration constraints of each motion axis to address path continuity and emergency stop buffering issues. Then, the centroid offset is calculated based on the fork load weight and cargo position distribution to generate a dynamic stability-adjusting speed compensation coefficient, improving motion stability. Next, the path is discretized into a sequence of timestamp trajectory points, and an inverse dynamics algorithm is used to iteratively solve for the candidate set of speed combinations that satisfy the constraints. Finally, a particle swarm optimization algorithm is introduced, with cargo stacking efficiency as the optimization objective, to search for the optimal speed combination in the candidate set, achieving global optimization of multi-axis motion parameters. By combining curvature distribution characteristics with dynamic constraint networks, the stability of stacker crane movement under complex paths is ensured, reducing mechanical shocks caused by sudden stops and starts. The dynamic stability speed compensation coefficient calculated based on the centroid offset effectively balances the impact of load distribution on motion stability. The optimal speed combination is searched through particle swarm optimization algorithm to maximize cargo stacking efficiency and shorten task execution time.

[0034] Figure 2 A flowchart illustrating the abnormal vibration identification results obtained by the present invention is shown.

[0035] According to an embodiment of the present invention, the vibration spectrum data of the high-level stacker crane during real-time monitoring of the combined movement speeds is used to identify abnormal vibrations based on the vibration spectrum data, thereby obtaining an abnormal vibration identification result. Specifically: S202, Based on the vibration sensor, the vibration time-domain signal of the high-level stacker crane when executing the motion speed combination is obtained, and the vibration time-domain signal is subjected to wavelet noise reduction processing and then time-frequency conversion operation is performed to generate vibration spectrum data. S204, extract vibration frequency distribution features from the vibration spectrum data, divide the vibration spectrum data into frequency bands based on the vibration frequency distribution features, extract the mean amplitude, peak frequency and harmonic distortion rate in each frequency band as vibration state feature vectors, and construct vibration feature difference index by combining the theoretical safe spectrum corresponding to the current movement speed combination of the high-level stacker crane. S206 uses a sliding time window to splice multiple consecutive sets of vibration state feature vectors in time series, and uses a pre-trained stacker crane vibration mode classification model to perform pattern matching on the spliced ​​multidimensional feature matrix to identify abnormal frequency band distribution patterns that exceed the theoretical safe spectrum. S208, generate a frequency domain fault fingerprint based on the frequency components and energy diffusion trend of the abnormal frequency band distribution pattern, input the frequency domain fault fingerprint into the fault diagnosis model based on the convolutional neural network for vibration source localization, and output the abnormal vibration identification result including abnormal vibration level, spectral offset, vibration location and vibration amplitude.

[0036] It should be noted that in the construction of the fault diagnosis model based on convolutional neural networks, the network adopts a parallel coding structure with multi-scale feature fusion, specifically including three parts: an input layer, a feature extraction module, and a multi-task classification head. The input layer receives a 128×128 pixel single-channel grayscale image of the frequency domain fault fingerprint, which represents the spectral distribution of vibration energy in the range of 0-2000Hz. The feature extraction module consists of five sets of cascaded convolutional blocks, each containing a 3×3 convolutional kernel to achieve local feature capture, batch normalization layer to eliminate data offset, ReLU activation function to introduce nonlinearity, and cross-layer skip connections to fuse shallow high-frequency details with deep semantic features. The multi-task classification head uses a spatial pyramid pooling layer to compress the feature dimension and then connects two fully connected branches in parallel. The first branch outputs the probability distribution of fault types (including normal state, bearing wear, gear misalignment, etc.) through a softmax function, and the second branch uses a bilinear interpolation deconvolution layer to reconstruct the three-dimensional spatial heat map of the vibration source. The training data consisted of vibration spectrum data collected from a high-level stacker crane experimental platform under various load conditions and speed combinations. After wavelet denoising and spectrum normalization, the data was converted into grayscale images. Data augmentation was achieved through random rotation of ±5 degrees, horizontal translation of 10% pixels, and Gaussian noise injection. Each image was labeled with a fault type code and the normalized position coordinates of the vibration source in the stacker crane coordinate system. Test data was independently collected from different models of stacker cranes, including spectrum samples of unknown fault modes not included in the training, used to verify the model's cross-equipment generalization ability. The network training employed the Adam optimizer, using a weighted sum of cross-entropy loss and mean squared error loss as the multi-task optimization objective. An early-stop strategy dynamically controlled the training rounds, ultimately forming a spatial perception diagnostic model capable of simultaneously identifying fault types and locating vibration sources.

[0037] For example: Training dataset composition (partial example data): ; Test dataset composition (partial example data): ; It should be noted that wavelet denoising and time-frequency conversion techniques effectively remove environmental noise interference, significantly improving the signal-to-noise ratio and feature extraction accuracy of vibration spectrum data. The vibration state analysis mechanism based on frequency band division and multi-dimensional feature vector construction can accurately capture abnormal spectral distribution patterns such as high-frequency harmonic distortion and amplitude abrupt changes, enhancing the sensitivity to early, weak abnormal vibrations. Utilizing sliding time window splicing and vibration modal classification models, dynamic tracking and pattern matching of continuous vibration characteristics are achieved, overcoming the limitations of traditional single-frame spectrum analysis and accurately identifying transient abnormal vibration events. Fault diagnosis technology that integrates frequency domain fault fingerprinting and convolutional neural networks enables multi-dimensional localization of vibration sources. The quantitative output of spectral offset and vibration location in the abnormal vibration identification results provides data support for targeted maintenance, reducing the risk of overturning and damage to goods caused by vibration. The real-time comparison mechanism between dynamic difference indicators and theoretical safe spectrum enables the system to adaptively adjust thresholds, effectively suppressing false alarm rates.

[0038] Figure 3 A flowchart illustrating the motion speed optimization strategy obtained by the present invention is shown.

[0039] According to an embodiment of the present invention, the optimization of the motion speed combination based on the abnormal vibration identification result to obtain a motion speed optimization strategy specifically includes: S302, extract the ratio of the average amplitude of the high-frequency vibration band to the average amplitude of the low-frequency vibration band in the abnormal vibration identification results as the vibration energy distribution parameter; S304, when the vibration energy distribution parameter exceeds the first preset safety threshold, locate the target motion axis that generates abnormal vibration, and calculate the speed reduction based on the product of the trajectory curvature change rate of the current path segment of the target motion axis and the fork load weight. It should be noted that high-frequency vibrations are usually caused by sudden changes in dynamic load or mechanical anomalies of the motion axis. Precisely locating the target axis can avoid the efficiency loss of global speed reduction. The rate of change of trajectory curvature quantifies the dynamic adjustment requirements of the path curvature. The larger the value, the more frequent the acceleration adjustment of the motion axis, leading to a surge in inertial force and mechanical stress. The load weight of the fork directly affects the inertial torque of the motion axis. The product of the two (rate of change of curvature × load weight) comprehensively characterizes the intensity of vibration excitation, establishing a quantitative basis for speed reduction from both kinematic and dynamic dimensions. This calculation strategy correlates curvature complexity with load inertia through physical mechanisms, enabling the speed reduction to dynamically adapt to the current working conditions—a large speed reduction is prioritized to ensure mechanical safety when a high curvature path is superimposed with a heavy load, while a small speed reduction is maintained to preserve operational efficiency when a low curvature path is lightly loaded. This not only suppresses the structural impact caused by sudden acceleration changes, but also achieves a precise balance between vibration suppression and operational efficiency through parametric modeling, breaking through the limitations of traditional experience-based speed adjustment.

[0040] S306, if the harmonic distortion rate of the high-frequency vibration band exceeds the second preset safety threshold, the acceleration time-domain signal of the target motion axis is integrated to obtain the accumulated vibration energy, and an acceleration attenuation coefficient is generated based on the ratio of the accumulated vibration energy to the maximum load capacity of the fork extension mechanism. It should be noted that when the harmonic distortion rate of the high-frequency vibration band exceeds the second preset safety threshold, the excessive high-frequency harmonic distortion rate indicates the presence of abnormal periodic impact or nonlinear resonance in the vibration signal, which may be caused by abnormal gear meshing, local bearing damage, or instantaneous load imbalance. Integrating the acceleration time-domain signal can quantify the cumulative effect of vibration energy within a specific time period (E=∫a(t)²dt), avoiding the instantaneous peak value from masking the risk of continuous damage. By comparing the accumulated energy with the maximum load-bearing capacity of the fork extension mechanism (α=E / Fmax), the relative impact intensity of vibration energy on the mechanical structure can be dynamically assessed. When the accumulated energy approaches or exceeds the structural load-bearing limit, an acceleration attenuation coefficient is generated to proportionally reduce the acceleration of the moving shaft.

[0041] S308, adjust the motion speed combination according to the speed reduction amount and acceleration attenuation coefficient, decelerate the target motion axis with abnormal vibration and the safe obstacle avoidance operation path until the running speed of the target motion axis reaches the speed reduction amount of the speed reduction compensation, and obtain the motion speed optimization strategy.

[0042] It should be noted that since the power of each motion axis of the high-level stacker crane is limited, high power operation will cause vibration of the stacker crane, which will lead to instability of the goods. Therefore, by coordinating the adjustment of the speed reduction and acceleration attenuation coefficient, the propagation of high-frequency vibration energy is precisely suppressed, reducing the risk of fatigue damage to mechanical components. The directional deceleration strategy based on vibration source location ensures the safety of the obstacle avoidance path while avoiding the loss of operational efficiency caused by global speed reduction. The dynamic adaptation mechanism between the acceleration attenuation coefficient and the load-bearing capacity of the forks ensures that the mechanical stress is always within the safe range of elastic deformation, preventing structural failure.

[0043] According to an embodiment of the present invention, it further includes: Based on the task data, determine the priority parameters for each stored and retrieved item. Based on the priority parameters, perform a conflict analysis on the urgency of the storage and retrieval tasks for each stored and retrieved item, mark the stored and retrieved items with conflicting urgency, and construct a concurrent storage and retrieval task dataset, including priority parameters, task generation timestamps, preset latest execution times, and target location coordinates. Based on the difference between the current system time and the task generation timestamp, calculate the real-time waiting time for each task, and generate a task timeliness decay coefficient according to the preset latest execution time. If multiple tasks are found to have the same priority parameter, the time decay function is triggered to perform time-sensitive weighting on the priority parameters of each task and generate a dynamic priority queue. When the difference in task timeliness decay coefficients in the dynamic priority queue exceeds a set threshold, the queue is reversed and a new priority sequence is generated; otherwise, the original priority parameters are maintained. The order of accessing and storing goods is determined based on the new priority sequence.

[0044] It's important to note that in automated warehousing systems, when a high-level stacker crane simultaneously receives multiple storage and retrieval tasks with the same urgency, traditional static priority allocation mechanisms cannot effectively resolve task conflicts. This leads to a decision-making deadlock, causing path planning redundancy, decreased operational efficiency, and even task timeouts. Especially in scenarios with a surge in goods storage and retrieval requests, fixed priority parameters ignore the time sensitivity differences between tasks and cannot dynamically reflect the urgency of tasks nearing their deadlines, resulting in delays of time-sensitive tasks or frequent conflicts in the robotic arm's movement paths. By introducing a task timeliness decay function and a dynamic priority reassignment mechanism, a concurrent storage and retrieval task dataset is first constructed based on goods attributes, a preset latest execution time, and the task generation sequence. The timeliness decay coefficient for each task is calculated in real time. When multiple tasks have the same priority parameter, the priority is weighted and adjusted based on the dynamic relationship between waiting time and deadline within the time window, generating a dynamic queue reflecting real-time urgency. Furthermore, a threshold for the difference in decay coefficients triggers a priority reordering and path tree pruning strategy. This achieves adaptive quantification of task urgency, prioritizing tasks nearing their timeout and significantly reducing task delay rates.

[0045] Figure 4 A block diagram of a high-level stacker control system based on motion speed optimization according to the present invention is shown.

[0046] A second aspect of the present invention also provides a high-level stacker crane control system 4 based on motion speed optimization. The system includes a memory 41 and a processor 42. The memory includes a high-level stacker crane control method program based on motion speed optimization. When the processor executes the high-level stacker crane control method program based on motion speed optimization, it performs the following steps: Spatial topology data of the high-level stacker crane's operating aisle is acquired using a three-dimensional lidar, and an obstacle distribution map is established based on the spatial topology data. Obtain the operation task data of the high-level stacker crane, and construct a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and operation task data; A simulation operation model of the high-level stacker crane is constructed. The operation task data and the safe obstacle avoidance operation path are imported into the simulation operation model to simulate the stacking operation and determine the combination of motion speeds of each motion axis of the high-level stacker crane. The vibration spectrum data of the high-level stacker crane when it executes the motion speed combination is monitored in real time, and abnormal vibration is identified based on the vibration spectrum data to obtain the abnormal vibration identification result; Based on the abnormal vibration identification results, the motion speed combination is optimized to obtain a motion speed optimization strategy.

[0047] This invention discloses a control method and system for a high-level stacker crane based on motion speed optimization. The method first uses a 3D lidar to acquire spatial topology data within the high-level stacker crane's operating aisle, thereby establishing an obstacle distribution map. Combined with the acquired task data, a safe obstacle avoidance path is constructed. A simulation operation model is built, importing the task data and the safe obstacle avoidance path to simulate stacking operations and determine the motion speed combinations for each axis. During stacker crane operation, the vibration spectrum data when executing this motion speed combination is monitored in real time for abnormal vibration identification. Finally, based on the abnormal vibration identification results, the motion speed combinations are optimized to obtain an optimized motion speed strategy. This invention effectively improves the safety and stability of high-level stacker crane operation, increases operational efficiency, and has broad application prospects in warehousing and logistics.

[0048] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0049] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0050] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0051] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0052] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0053] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A high-level stacker crane control method based on motion speed optimization, characterized in that, Includes the following steps: Spatial topology data of the high-level stacker crane's operating aisle is acquired using a three-dimensional lidar, and an obstacle distribution map is established based on the spatial topology data. Obtain the operation task data of the high-level stacker crane, and construct a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and operation task data; A simulation operation model of the high-level stacker crane is constructed. The operation task data and the safe obstacle avoidance operation path are imported into the simulation operation model to simulate the stacking operation and determine the combination of motion speeds of each motion axis of the high-level stacker crane. The vibration spectrum data of the high-level stacker crane when it executes the motion speed combination is monitored in real time, and abnormal vibration is identified based on the vibration spectrum data to obtain the abnormal vibration identification result; Based on the abnormal vibration identification results, the motion speed combination is optimized to obtain a motion speed optimization strategy.

2. The high-level stacker crane control method based on motion speed optimization according to claim 1, characterized in that, The process of acquiring spatial topology data within the high-level stacker crane's operating aisle using 3D lidar and establishing an obstacle distribution map based on this spatial topology data specifically involves: The high-level stacker crane's operating aisle is scanned using a 3D LiDAR to acquire a 3D point cloud data stream. The 3D point cloud data stream is then spatiotemporally aligned to obtain preprocessed 3D point cloud data. Based on the structural characteristics of fixed racks in the aisle of high-level stacker cranes, multi-scale feature analysis is performed on preprocessed three-dimensional point cloud data to extract the normal vector distribution characteristics, surface curvature and density gradient change parameters of the point cloud, and to construct the feature weight matrix of the fixed rack structure. The preprocessed 3D point cloud data is divided into local regions using a sliding window mechanism. Combined with historical topological data of the shelf installation location, the statistical distribution difference of point cloud features in each local region is calculated to generate a segmentation threshold that distinguishes between fixed structures and obstacles. A dynamic region growing algorithm based on feature weight matrix is ​​used to iteratively aggregate continuous point cloud clusters that meet the structural characteristics of fixed shelves according to the segmentation threshold, so as to obtain fixed shelf segmentation point cloud data. After the continuous point cloud clusters are iteratively aggregated to complete, the remaining point cloud is marked as an obstacle candidate set. The spatial position difference of the obstacle candidate set is calculated for the point cloud data of multiple consecutive frames to determine the motion continuity of the obstacle candidate set. When the obstacle candidate set has motion continuity, the residual point cloud data with motion continuity in multiple consecutive frames of point cloud data is labeled as moving obstacle segmentation point cloud data, and the remaining point cloud data is labeled as stationary obstacle segmentation point cloud data. Based on the continuous point cloud data of fixed racks, moving obstacles, and stationary obstacles, a dynamic spatial topology of the high-level stacker crane's operating aisle is constructed. Based on the dynamic spatial topology, the changes in obstacle positions are determined, and an obstacle distribution map is constructed.

3. The high-level stacker crane control method based on motion speed optimization according to claim 1, characterized in that, The process of acquiring the operational task data of the high-level stacker crane and constructing a safe obstacle avoidance path for the high-level stacker crane based on the obstacle distribution map and operational task data specifically involves: Acquire the operation task data of the high-level stacker crane. The operation task data includes one or more target storage location coordinates, fork load weight, cargo access priority parameters, and cargo position in the forks. Generate an initial operation path tree based on the target storage location coordinates, cargo access priority parameters, and the position of stationary obstacles in the obstacle distribution map. The spatial change trajectory of moving obstacles in the working roadway is obtained according to the obstacle distribution map. Dynamic obstacle avoidance monitoring nodes are inserted into the initial working path tree according to the spatial change trajectory of moving obstacles. The dynamic obstacle avoidance monitoring nodes use a sliding window mechanism to predict the spatiotemporal conflict of moving obstacles. When a moving obstacle overlaps with the current path segment of the high-level stacker crane in time and space, the three adjacent path nodes at the end of the path segment are extracted to construct a local path replanning area. The path deviation compensation angle is calculated based on the projection component of the obstacle's motion velocity vector, and alternative path branches containing avoidance margin are generated. Based on the cargo access priority parameters in the task data, determine the cargo access urgency parameters and construct a dynamic priority evaluation model. If the path length increment of the alternative path branch exceeds the tolerance threshold corresponding to the urgency parameter, trigger the path tree pruning strategy to delete the alternative branches that conflict with the priority. When a moving obstacle is detected to have left the operating path coverage area of ​​the high-level stacker crane, the optimal convergence path is regenerated using an inverse kinematics algorithm based on the spatial geometric relationship between the current position of the high-level stacker crane and the coordinates of the target cargo location, and the rate of change of joint acceleration during the path convergence process is constrained within a preset safety range. By iteratively executing dynamic obstacle avoidance monitoring and path tree optimization operations until all target cargo location coordinates have been accessed and the path conflict detection result shows no collision risk, the final verified safe obstacle avoidance operation path is output.

4. The high-level stacker crane control method based on motion speed optimization according to claim 1, characterized in that, The construction of the high-level stacker crane simulation operation model involves importing the operation task data and safety obstacle avoidance operation path into the simulation operation model to simulate stacking operations and determine the motion speed combination of each motion axis of the high-level stacker crane, specifically: Based on the mechanical structure parameters of the high-level stacker crane, a three-dimensional physical model including a traveling mechanism, a lifting mechanism, and a fork extension mechanism is constructed according to the mechanical structure parameters. The three-dimensional physical model is then integrated using computer mechanics simulation software to construct a simulation operation model of the high-level stacker crane. Based on the trajectory curvature distribution features extracted from the aforementioned safe obstacle avoidance operation path, and combined with the maximum acceleration constraint parameters of each motion axis of the high-level stacker crane, a dynamic constraint network containing the continuity equation of joint motion trajectory and emergency stop buffer boundary is constructed. Based on the fork load weight and cargo position distribution in the task data, the influence weight of the stacker crane's center of gravity offset on motion stability is calculated, and dynamic stability speed compensation coefficients for different path segments are generated in the dynamic constraint network. The safe obstacle avoidance operation path is discretized into a sequence of trajectory points containing timestamps. The path is divided into segments based on the curvature change detection results between trajectory points. Within each segment, the inverse dynamics algorithm is used to iteratively solve for the multi-axis velocity combination candidate set that meets the requirements of dynamic constraint network and dynamic stability velocity compensation coefficient. Particle swarm optimization (PSO) algorithm is introduced, and the cargo stacking efficiency of the high-level stacker crane is used as the optimization objective of the PSO algorithm. The algorithm searches the multi-axis speed combination candidate set and outputs the multi-axis speed combination candidate set with the highest cargo stacking efficiency, thus obtaining the motion speed combination of each motion axis of the high-level stacker crane.

5. The high-level stacker crane control method based on motion speed optimization according to claim 1, characterized in that, The system monitors the vibration spectrum data of the high-level stacker crane when it executes the combined motion speeds, and identifies abnormal vibrations based on this vibration spectrum data to obtain abnormal vibration identification results. Specifically: Vibration time-domain signals of the high-level stacker crane when executing the motion speed combination are obtained based on vibration sensors. After wavelet denoising processing of the vibration time-domain signals, time-frequency conversion is performed to generate vibration spectrum data. Based on the vibration spectrum data, the vibration frequency distribution characteristics are extracted. Based on the vibration frequency distribution characteristics, the vibration spectrum data is divided into frequency bands. The mean amplitude, peak frequency, and harmonic distortion rate in each frequency band are extracted as vibration state feature vectors. The vibration feature difference index is constructed by combining the theoretical safe spectrum corresponding to the current movement speed combination of the high-level stacker crane. A sliding time window is used to splice multiple consecutive sets of vibration state feature vectors in time series. A pre-trained stacker crane vibration modal classification model is used to perform pattern matching on the spliced ​​multidimensional feature matrix to identify abnormal frequency band distribution patterns that exceed the theoretical safe spectrum. A frequency domain fault fingerprint is generated based on the frequency components and energy diffusion trend of the abnormal frequency band distribution pattern. The frequency domain fault fingerprint is then input into a fault diagnosis model based on a convolutional neural network to locate the vibration source. The output includes abnormal vibration level, spectral offset, vibration location, and vibration amplitude.

6. The high-level stacker crane control method based on motion speed optimization according to claim 1, characterized in that, The optimization of the motion speed combination based on the abnormal vibration identification result yields a motion speed optimization strategy, specifically as follows: The ratio of the mean amplitude of the high-frequency vibration band to the mean amplitude of the low-frequency vibration band in the abnormal vibration identification results is extracted as the vibration energy distribution parameter. When the vibration energy distribution parameter exceeds the first preset safety threshold, the target motion axis that generates abnormal vibration is located, and the speed reduction is calculated based on the product of the trajectory curvature change rate of the current path segment of the target motion axis and the fork load weight. If the harmonic distortion rate of the high-frequency vibration band exceeds the second preset safety threshold, the acceleration time-domain signal of the target motion axis is integrated to obtain the accumulated vibration energy, and an acceleration attenuation coefficient is generated based on the ratio of the accumulated vibration energy to the maximum load capacity of the fork extension mechanism. The motion speed combination is adjusted according to the speed reduction amount and acceleration attenuation coefficient. The target motion axis with abnormal vibration and the safe obstacle avoidance operation path are decelerated until the running speed of the target motion axis reaches the speed reduction amount of the speed reduction compensation, thus obtaining the motion speed optimization strategy.

7. A high-level stacker crane control system based on motion speed optimization, characterized in that, The high-level stacker crane control system based on motion speed optimization includes a storage device and a processor. The storage device includes a high-level stacker crane control method program based on motion speed optimization. When the high-level stacker crane control method program based on motion speed optimization is executed by the processor, it performs the following steps: Spatial topology data of the high-level stacker crane's operating aisle is acquired using a three-dimensional lidar, and an obstacle distribution map is established based on the spatial topology data. Obtain the operation task data of the high-level stacker crane, and construct a safe obstacle avoidance operation path for the high-level stacker crane based on the obstacle distribution map and operation task data; A simulation operation model of the high-level stacker crane is constructed. The operation task data and the safe obstacle avoidance operation path are imported into the simulation operation model to simulate the stacking operation and determine the combination of motion speeds of each motion axis of the high-level stacker crane. The vibration spectrum data of the high-level stacker crane when it executes the motion speed combination is monitored in real time, and abnormal vibration is identified based on the vibration spectrum data to obtain the abnormal vibration identification result; Based on the abnormal vibration identification results, the motion speed combination is optimized to obtain a motion speed optimization strategy.

Images