Automated palletizing methods, devices, terminals and storage media

Abstract

This application discloses an automatic palletizing method, apparatus, terminal, and storage medium. The method includes: acquiring package information of packages to be palletized using a data acquisition device; matching the optimal pallet type to the packages using historical pallet type data from the automatic palletizing system; performing point cloud segmentation on the pallet holding the packages using a preset segmentation model to obtain segmentation results; calculating pallet gaps and generating a palletizing path using the optimal pallet type and a preset path planning algorithm; controlling a flexible gripper device to grasp the packages to be palletized; real-time correcting the grasping posture based on feedback from a torque sensor; controlling a robotic arm to perform palletizing according to the palletizing path; when the segmentation result of the pallet matches the optimal pallet type, generating a 3D palletizing model; binding the 3D palletizing model to an AGV (Automated Guided Vehicle) under the pallet; and controlling the AGV to automatically enter the warehouse. This method improves the adaptability and space utilization of the palletizing system by matching pallet types and planning paths to complete palletizing and warehousing.

Description

Technical Field

[0001] This application relates to the field of warehousing and logistics, and in particular to an automated palletizing method, apparatus, terminal and storage medium. Background Technology

[0002] With the rapid development of the logistics industry, palletizing has become a crucial part of the logistics process. Efficient and accurate palletizing operations have a profound impact on the smooth operation of the entire logistics supply chain. Traditional manual palletizing methods can no longer meet the needs of large-scale and high-efficiency palletizing.

[0003] In traditional logistics palletizing scenarios, various methods have been used to solve the problem of goods stacking. On one hand, manual palletizing is the most basic method, where workers rely on their experience and physical strength to move and stack goods one by one onto pallets. On the other hand, early automated palletizing often utilized simple mechanical devices, such as programmed palletizers. These machines stack goods of specific sizes and shapes according to a pre-set program. There are also some semi-automatic palletizing devices that require manual assistance, such as adjusting the position and orientation of the goods, to cooperate with the mechanical device in completing the palletizing task.

[0004] However, existing palletizing methods have significant drawbacks. Manual palletizing is inefficient, easily affected by worker fatigue and mood, and cannot maintain consistent palletizing quality over long periods. Simple mechanical palletizing equipment lacks flexibility, struggles to adapt to goods of different shapes, sizes, and materials, and cannot dynamically adjust palletizing plans based on real-time goods information and warehouse conditions. Furthermore, they have limited capacity to handle complex stacking patterns and respond to changes in goods placement, resulting in low utilization of warehouse space and a higher risk of collisions and damage during the palletizing process. Summary of the Invention

[0005] The purpose of this application is to overcome the above-mentioned technical problems and provide an automatic palletizing method, apparatus, terminal and storage medium.

[0006] Firstly, this application provides an automatic palletizing method, employing the following technical means:

[0007] An automated palletizing method, based on an automated palletizing system, the automated palletizing system including a robotic arm and a flexible gripper device at the end of the robotic arm, the method comprising the following steps:

[0008] Obtain the package information of the package to be palletized, and combine the package information with the historical pallet type data of the automatic palletizing system to match the optimal pallet type for the current package to be palletized;

[0009] The pallet on which the package is stacked is segmented by point cloud using a preset segmentation model to obtain the segmentation result. The pallet gap is calculated and the stacking path is generated by combining the optimal stacking type and the preset path planning algorithm.

[0010] The flexible gripper device is controlled to grasp the package to be palletized, and the grasping posture is corrected in real time based on the feedback from the pre-set torque sensor. The robotic arm is then controlled to perform palletizing according to the palletizing path.

[0011] When the segmentation result of the pallet matches the optimal stacking type, a 3D palletizing model is generated, and the 3D palletizing model is bound to the AGV trolley under the pallet to control the AGV trolley to automatically enter the warehouse.

[0012] By adopting the above technical solution, the information of the packages to be palletized can be accurately obtained and the optimal pallet type can be matched. The pallet gap can be accurately calculated to generate a reasonable palletizing path. The flexible gripper device can be used to grab the packages and correct the gripping posture in real time to ensure accurate palletizing. When the pallet segmentation result matches the optimal pallet type, the palletizing 3D model and AGV can be bound and automatically put into the warehouse, thereby improving the efficiency and accuracy of palletizing.

[0013] Preferably, the step of acquiring package information of packages to be palletized based on a preset acquisition device and generating a unique ID for each package to be palletized, and matching the optimal pallet type for the current package to be palletized by combining the historical pallet type data of the automatic palletizing system, specifically includes the following steps:

[0014] The system acquires package information of packages to be palletized using a preset acquisition device, which includes a barcode scanner and a binocular depth camera. The barcode scanner scans the packages to be palletized to obtain the waybill information. The binocular depth camera acquires the surface material and size of the packages to be palletized. The system generates a unique ID for the packages to be palletized based on the waybill information and associates the surface material and size of the packages with the ID.

[0015] Based on the information on the waybill, determine the position of the pallet used to place the parcels to be stacked, and determine whether the pallet is empty;

[0016] If the pallet is empty, the historical successful pallet patterns of the automatic palletizing system are obtained from the cloud based on the surface material of the package. Candidate pallet patterns are selected from the historical successful pallet patterns based on the cosine similarity algorithm. The optimal pallet pattern is selected from the candidate pallet patterns based on the preset optimal pallet pattern matching rules. The package information and the optimal pallet pattern corresponding to the current package to be palletized are uploaded to the cloud.

[0017] If the pallet is not empty, obtain the optimal stacking type of the current pallet from the cloud.

[0018] By adopting the above technical solutions, the barcode scanning device and binocular depth camera can comprehensively and accurately acquire the waybill information, surface material, and size of the packages to be palletized, providing basic data for subsequent processing. Based on the waybill information, the pallet position can be accurately determined and its emptiness can be judged, so as to select the appropriate method to match the optimal pallet type for different situations. For empty pallets, the cosine similarity algorithm is used to filter candidate pallet types from historical successful pallet types, and then the optimal pallet type is determined according to preset rules and uploaded to the cloud, which helps to improve palletizing efficiency and rationality. For non-empty pallets, the optimal pallet type can be obtained directly from the cloud, which simplifies the process and saves time.

[0019] Preferably, the step of selecting the optimal stacking type from the candidate stacking types according to a preset optimal stacking type matching rule specifically includes the following steps:

[0020] A reinforcement learning environment is constructed, which includes a state space, an action space, and a reward function. The reward function is calculated based on the feasible path of the package to be palletized, the spatial score of the candidate pallet type, and the support score.

[0021] The candidate stack type is quantitatively evaluated, including feasible path quantitative analysis, spatial score calculation and support score calculation, to obtain feasible path score, spatial score and support score.

[0022] Based on the PPO reinforcement learning algorithm, a policy network is trained and iteratively optimized by combining the state space and the action space. Based on the trained policy network, candidate stacking patterns are simulated and executed. A comprehensive reward score is calculated based on the feasible path of the package to be stacked, the space score, and the support score. The candidate stacking pattern with the highest comprehensive reward score is output as the optimal stacking pattern.

[0023] By adopting the above technical solutions, a reinforcement learning environment including state space, action space, and reward function is constructed. This allows for the comprehensive consideration of the feasible paths of packages to be palletized, the spatial scores of candidate pallet types, and their support scores, thus more accurately measuring the suitability of each candidate pallet type. By training the policy network based on the PPO reinforcement learning algorithm and iteratively optimizing it, and simulating the execution of candidate pallet types, the comprehensive reward score is calculated to output the optimal pallet type. This effectively improves the accuracy and reliability of the selection process, providing a more reasonable pallet type selection for subsequent palletizing.

[0024] Preferably, the step of segmenting the pallet containing the packages using a preset segmentation model, calculating the pallet gaps and generating the palletizing path using the optimal stacking type and a preset path planning algorithm, specifically includes the following steps:

[0025] Acquire point cloud data of the pallet used to stack the parcels to be palletized, and segment the parcel outline on the pallet according to the preset YOLOv8 segmentation model and the pallet point cloud data to obtain the segmentation result, which includes available gaps.

[0026] Based on the optimal stacking type, the available gaps in the segmentation results are initially screened to exclude available gaps that conflict with the structure of the optimal stacking type.

[0027] Traverse all available gaps obtained after preliminary screening, calculate the minimum bounding rectangle of each gap, filter out all minimum bounding rectangles not smaller than the package size and summarize them into a gap candidate set, calculate the space utilization rate and optimal placement pose of each candidate gap in the gap candidate set, summarize all the optimal placement poses into a candidate placement pose list, and arrange the candidate placement pose list according to the space utilization rate.

[0028] By adopting the above technical solution, the YOLOv8 segmentation model is used to segment the package outline, which can accurately identify the distribution of packages and gaps on the pallet. Available gaps are initially screened based on the optimal stacking pattern, and conflicting gaps are eliminated to ensure that subsequent stacking conforms to the optimal stacking pattern. The available gaps after screening are traversed, the minimum bounding rectangle is calculated, and a gap candidate set is compiled. The space utilization rate and the best placement pose are calculated. The candidate placement poses are arranged according to the space utilization rate to form a list, which can effectively utilize the pallet space and provide a basis for selecting the appropriate stacking position, thereby improving the stacking efficiency and space utilization rate.

[0029] Preferably, the method further includes the following steps:

[0030] The current pose of the robotic arm and the candidate gaps and their optimal placement poses are obtained. Based on the pallet point cloud data, a pallet geometric model is constructed, and a workspace model carrying anti-collision area markers is obtained.

[0031] Multiple candidate paths satisfying preset constraints are generated based on the improved RRT* algorithm. The constraints include spatial constraints and acceleration constraints. The spatial constraint is that the candidate path does not collide with the tray geometry model or the anti-collision area. The acceleration constraint is that the corner acceleration of the candidate path does not exceed a preset threshold.

[0032] The candidate paths that meet the constraints are scored according to the preset path scoring rules, and the candidate path with the highest score is taken as the palletizing path.

[0033] By adopting the above technical solutions, a pallet geometric model is constructed based on pallet point cloud data, and a workspace model with anti-collision zone markings is obtained, which can more accurately reflect the palletizing environment. The improved RRT* algorithm is used to generate candidate paths that meet the spatial and acceleration constraints, which can avoid collisions and control corner acceleration, ensuring operational safety and stability. The candidate path with the highest score is selected as the palletizing path according to the preset path scoring rules, which can obtain the optimal palletizing path and improve palletizing efficiency and accuracy.

[0034] Preferably, the process of controlling the flexible gripper device to grasp the package to be palletized, and adjusting the grasping posture in real time based on feedback from a pre-set torque sensor, and controlling the robotic arm to perform palletizing according to the palletizing path, specifically includes the following steps:

[0035] The initial clamping force is obtained by calling the preset clamping force parameters according to the package material. The flexible gripper device includes a gripper and a pneumatic tendon that drives the gripper. A torque sensor is pre-installed on the flexible gripper device. A rotary encoder and an inertial measurement unit are pre-installed on the robotic arm.

[0036] The pneumatic tendon is controlled to drive the gripper to grasp the package to be stacked with the initial gripping force.

[0037] During the palletizing process of the robotic arm, the real-time clamping force is monitored by the torque sensor. When the deformation of the package to be palletized exceeds the preset deformation threshold, the real-time clamping force is dynamically adjusted.

[0038] The actual grasping posture deviation is estimated by obtaining the joint angle fed back by the rotary encoder and the end-effector acceleration collected by the inertial measurement unit. Based on the actual grasping posture deviation and the PID controller, a joint compensation command is generated, and the grasping posture of the robotic arm is corrected based on the joint compensation command.

[0039] When the torque sensor detects a preset contact force condition, it triggers a gripper release command to complete the placement of the package to be stacked.

[0040] By adopting the above technical solutions, the initial clamping force determined according to the package material can be adapted to the gripping needs of packages of different materials; real-time monitoring and dynamic adjustment of the clamping force can prevent packages from being damaged due to excessive force or falling due to insufficient clamping force; estimating and correcting the gripping posture deviation can improve the palletizing accuracy; and releasing the package when the contact force reaches the condition can ensure accurate placement of the package.

[0041] Preferably, after placing the packages to be stacked, the following steps are also included:

[0042] Obtain the current position of the package, calculate the center of gravity coordinates of the current package based on the package size, obtain the center of gravity coordinates of the remaining packages on the current tray, and calculate the overall center of gravity coordinates of all packages on the tray based on the center of gravity coordinates of the current package.

[0043] Obtain the stable center coordinates of the pallet, calculate the center of gravity offset by combining the overall center of gravity coordinates, and compare the center of gravity offset with the preset warning range;

[0044] When the center of gravity offset falls within the warning range, a center of gravity offset control strategy is executed by combining the overall center of gravity coordinates and the stable center coordinates.

[0045] By adopting the above technical solution, after the parcels to be palletized are placed, the overall center of gravity coordinates and center of gravity offset of all parcels on the pallet can be calculated. The center of gravity offset is compared with the preset warning range. When the offset falls within the warning range, the center of gravity offset control strategy is executed to ensure the stability of the parcels on the pallet and avoid problems such as parcels tipping over due to excessive center of gravity offset, thereby improving the safety and reliability of palletizing operations.

[0046] Secondly, this application provides an automatic palletizing device, which employs the following technical means:

[0047] An automated palletizing device, based on an automated palletizing system, the automated palletizing system including a robotic arm and a flexible gripper device at the end of the robotic arm, comprising the following modules:

[0048] The optimal pallet type matching module is used to obtain the package information of the package to be palletized, and combine the package information with the historical pallet type data of the automatic palletizing system to match the optimal pallet type for the current package to be palletized;

[0049] The palletizing path planning module is used to perform point cloud segmentation on the pallets on which packages are stacked using a preset segmentation model to obtain the segmentation result, and to calculate the pallet gap and generate the palletizing path by combining the optimal pallet type and the preset path planning algorithm.

[0050] The gripping posture correction module is used to control the flexible gripper device to grip the package to be palletized, and to correct the gripping posture in real time in combination with the feedback of the pre-set torque sensor, and to control the robotic arm to perform palletizing according to the palletizing path.

[0051] The automatic storage module is used to generate a 3D palletizing model when the segmentation result of the pallet matches the optimal palletizing type, bind the 3D palletizing model to the AGV trolley under the pallet, and control the AGV trolley to automatically store the pallet.

[0052] By adopting the above technical solutions, the optimal pallet type matching module can accurately obtain package information and generate a unique ID for each package. It combines historical pallet type data to match the optimal pallet type, improving the adaptability of package palletizing. The palletizing path planning module can perform point cloud segmentation on the pallet, calculate the pallet gap and generate the palletizing path by combining the optimal pallet type and path planning algorithm, improving space utilization. The gripping posture correction module corrects the gripping posture in real time through torque sensor feedback, ensuring the accuracy of the palletizing process and preventing package compression and deformation. When the pallet segmentation result matches the optimal pallet type, the automatic warehousing module generates a 3D palletizing model and binds it to the AGV cart to realize automatic warehousing and improve logistics efficiency.

[0053] Thirdly, this application provides a smart terminal, which adopts the following technical solution:

[0054] A smart terminal includes a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the automatic palletizing method as described above.

[0055] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0056] A computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the automated palletizing method as described above.

[0057] In summary, this application includes at least one of the following beneficial effects:

[0058] (1) The method of this application combines package information and historical pallet type data to match the optimal pallet type, which can dynamically adjust the palletizing scheme according to the real-time situation, thereby improving the utilization rate of warehouse space;

[0059] (2) The method of this application divides the pallet point cloud by a preset segmentation model and generates the optimal palletizing path by combining the optimal pallet type and path planning algorithm, thereby reducing the risk of collision and damage during the palletizing process;

[0060] (3) The method of this application uses a flexible gripper device to grab the package and combines the feedback of the torque sensor to correct the gripping posture in real time. It can adapt to goods of different shapes, sizes and materials, making the palletizing process more flexible, adapting to a variety of goods, and improving the gripping success rate. Attached Figure Description

[0061] Figure 1 This is an overall flowchart of the automatic palletizing method according to an embodiment of this application;

[0062] Figure 2 This is a flowchart of the stack type matching and path planning of the automatic palletizing method in the embodiments of this application;

[0063] Figure 3 This is a flowchart of the center of gravity offset determination process in the automatic palletizing method of this application embodiment;

[0064] Figure 4 This is a schematic diagram of the automatic palletizing device according to an embodiment of this application. Detailed Implementation

[0065] The technical solutions in the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments are only possible technical implementations of the present invention, but are not limited thereto. Other embodiments obtained by those skilled in the art in conjunction with the embodiments of the present invention without creative effort are also within the protection scope of the present invention.

[0066] This application mainly adopts a solution of obtaining package information, matching pallet type, and planning path to complete palletizing and warehousing, which achieves the effects of improving the adaptability of the palletizing system, space utilization, positioning accuracy, and reducing the failure rate. The following is a further detailed description of this application.

[0067] The automated palletizing method of this application is based on an automated palletizing system, which includes a robotic arm and a flexible gripper device at the end of the robotic arm, such as... Figure 1 As shown, the method includes the following steps:

[0068] S1. Obtain the package information of the packages to be palletized, and match the optimal pallet type for the current packages to be palletized by combining the package information with the historical pallet type data of the automatic palletizing system. This includes the following steps:

[0069] S11. Obtain package information of the package to be palletized according to the preset acquisition device, which includes a barcode scanning device and a binocular depth camera.

[0070] Based on the information obtained from the waybill by scanning the package to be palletized by the scanning device, in this embodiment, the scanning device is a P2 scanner, which includes an integrated laser ranging module. When the package to be palletized enters the detection area, the photoelectric sensor in the system triggers the P2 scanner to work.

[0071] The surface material and dimensions of the packages to be palletized are obtained based on a binocular depth camera. In this embodiment, the surface material of the packages includes cardboard boxes, woven bags, foam boxes, and hard plastics, and the dimensions of the packages include length (L), width (W), and height (H).

[0072] S12. Generate a unique ID for the parcel to be palletized based on the waybill information, summarize the parcel surface material and parcel size into parcel information and associate it with the generated ID, and upload the parcel information to the cloud.

[0073] S13, such as Figure 2 As shown, the position of the pallet for stacking packages is determined based on the waybill information. It is then determined whether the pallet for stacking packages is empty. If the pallet is empty, it means that the current package to be stacked is the first package, so the optimal stacking type needs to be matched.

[0074] S131. When the pallet is empty, retrieve all historical successful pallet types of the automatic palletizing system from the cloud based on the surface material of the package. The surface material of the package matched by the retrieved historical successful pallet types should be the same as the surface material of the current package.

[0075] S132. Extract features from the current packages to be palletized and construct a package feature vector for the packages to be palletized, i.e. ;

[0076] in, Material coding, The historical average density of the same material packaged together is obtained from the cloud and is expressed in kg / m³.

[0077] In this embodiment, the material code for the cardboard box is 1, the material code for the woven bag is 2, the material code for the foam box is 3, and the material code for the hard plastic is 4.

[0078] S133, Construct the historical stack-type feature vector, that is ;

[0079] in, This represents the average length of all packages in historically successful stacking type i. This is the arithmetic mean of the widths of all packages in historically successful stack type i. This represents the maximum height of a single package in historically successful stacking type i. The main material code, The average density of packages in historically successful stack type i is calculated based on the weight-to-volume ratio.

[0080] S134. Based on the preset cosine similarity algorithm, candidate stacking patterns are selected from historical successful stacking patterns;

[0081] ;

[0082] In this embodiment, S>0.75, and the top five candidate stack types with the highest similarity are calculated based on the cosine similarity algorithm. In other feasible implementation methods, the value of S can also be adjusted.

[0083] S135. Based on the preset optimal pallet type matching rules, the optimal pallet type is selected from the candidate pallet types, and the package information and the optimal pallet type corresponding to the current package to be palletized are uploaded to the cloud. The specific steps include the following.

[0084] Construct a reinforcement learning environment that includes a state space, an action space, and a reward function.

[0085] The state space includes the geometric features of candidate stack types, the attributes of packages to be stacked, and pallet space data; the geometric features of candidate stack types include the number of stack layers and the arrangement of packages; the attributes of packages to be stacked include the size, weight, and shape of the packages; and the pallet space data includes the shape and remaining space dimensions.

[0086] The action space is used to select one from the set of candidate pallet types as the current palletizing scheme;

[0087] The reward function calculates a comprehensive reward score based on the feasible path of the parcel to be palletized, the spatial score of the candidate pallet type, and the support score.

[0088] Right now, ;

[0089] in, , , These are respectively the feasible path score, spatial score, and support score. , , These are the corresponding weighting coefficients, which can be adjusted.

[0090] Next, the policy network (Actor network) and value network (Critic network) of the PPO algorithm are constructed, and the network parameters are randomly initialized.

[0091] The candidate stack type is quantitatively evaluated. The quantitative evaluation includes feasible path quantitative analysis, spatial score calculation and support score calculation, to obtain feasible path score, spatial score and support score.

[0092] In one specific implementation, feasible path quantification analysis includes the following steps:

[0093] For each candidate stack type, the shortest collision-free path length of the robotic arm from the starting position to the target package placement position is calculated, and the score is mapped according to the path length ratio. In a specific implementation, the shortest path length gets 1 point and the longest path length gets 0 points.

[0094] Determine whether the joint angles of the robotic arm are within the preset safe range when it performs actions at the target placement position. If they are outside the range, the feasible path score is 0.

[0095] Determine whether the flexible gripper of the robotic arm will collide with the preset package position during the grasping and placement process. If there is a risk of collision, the feasible path score is 0.

[0096] Normalize the above scores to the [0, 1] interval to obtain the final feasible path score.

[0097] In one specific implementation, a score is calculated based on the utilization rate of pallet space by the candidate stack type. The space utilization rate is calculated as the occupied pallet space volume / the total pallet volume. The space utilization rate is then mapped to the [0,1] interval as the space score.

[0098] In one specific implementation, the support score is obtained based on the stability of the candidate stack type. Stability includes the proportion of the support area under the package and whether the center of gravity projection is within the support surface. If the requirements are met, 1 point is awarded; otherwise, 0 points are awarded.

[0099] Based on the PPO reinforcement learning algorithm, the policy network is trained by combining the state space and action space and then iteratively optimized.

[0100] In one specific implementation, data sampling is performed based on the current policy network, candidate stack types are selected from the action space, and after execution, new states, rewards, and whether the process has ended are obtained. The state-action-reward-next state sequence data is recorded.

[0101] Based on the sequence data, the policy network parameters are updated using the PPO algorithm to optimize the policy for selecting candidate stack types. The value network parameters are also updated based on the sequence data. The data sampling and network update process is repeated multiple times to gradually optimize and iterate the policy network and value network.

[0102] Training stops when the reward change is less than the threshold or the preset number of training rounds is reached after multiple consecutive iterations, and the trained policy network is obtained.

[0103] Based on the trained policy network, candidate stacking patterns are simulated and executed. The comprehensive reward score is calculated based on the feasible path, spatial score, and support score of the packages to be stacked. The candidate stacking pattern with the highest comprehensive reward score is output as the optimal stacking pattern.

[0104] S14. If the pallet is not empty, retrieve the optimal stacking type of the current pallet from the cloud.

[0105] If the pallet is not empty, meaning the current package to be palletized is not the first package, then the optimal pallet type has already been matched when the first package was placed on the pallet, so the optimal pallet type matched for the pallet can be obtained directly.

[0106] S2. The pallets on which packages are stacked are segmented into point clouds using a preset segmentation model. The pallet gaps are calculated using a preset path planning algorithm, and a stacking path is generated. The specific steps include the following:

[0107] S21. Acquire point cloud data of the pallet used for stacking packages using a binocular depth camera. The pallet point cloud data includes already stacked packages and gaps. The binocular camera only scans the corresponding area of ​​the pallet, reducing computational load.

[0108] S22. Based on the preset YOLOv8 segmentation model and pallet point cloud data, segment the outline of the package on the pallet to obtain the segmentation result, which includes available gaps.

[0109] The segmentation results are subjected to an expansion operation. In this embodiment, the expansion operation is to increase the gap boundary by 3-5mm to ensure a safety margin for the operation of the robotic arm.

[0110] Specifically, the segmentation result is the bounding box and the segmentation mask image.

[0111] S23. Based on the structural characteristics and arrangement rules of the optimal stack type, perform preliminary screening of the available gaps in the segmentation results and eliminate available gaps that conflict with the structure of the optimal stack type.

[0112] In a feasible implementation, the rules for determining structural conflicts include whether the geometry or shape of the available gap does not conform to the stack arrangement rules, or whether the location of the available gap cannot meet the support stability requirements of the stack. For example, if the optimal stack requires the packages to be stacked horizontally (with the long side along the X-axis), then a conflict is considered when the width of the minimum bounding rectangle of the available gap is less than the length of the long side of the package; if the optimal stack is designed as a pyramid shape, i.e., the number of packages in the lower layer is greater than the number in the upper layer, then a conflict is considered when a available gap is located at the edge and cannot be fully supported by the lower layer of packages.

[0113] S24. Traverse all available gaps obtained after the initial screening, calculate the minimum bounding rectangle of each gap, and select all minimum bounding rectangles that are not smaller than the package size and summarize them into a gap candidate set.

[0114] S25. Calculate the space utilization rate and optimal placement pose of each candidate gap in the gap candidate set, and summarize all the optimal placement poses into a candidate placement pose list. The candidate placement pose list is arranged from high to low according to the space utilization rate.

[0115] S26. Obtain the current pose and candidate gaps of the robotic arm and its optimal placement pose, construct a pallet geometric model based on the pallet point cloud data, and obtain a workspace model carrying anti-collision area markers.

[0116] In this embodiment, the anti-collision area is mainly the gantry frame. It is necessary to ensure that the robotic arm does not hit the gantry frame during its movement. The workspace model with anti-collision area markings is the gantry frame safety boundary.

[0117] S27. Generate multiple candidate paths that satisfy preset constraints based on the improved RRT* algorithm.

[0118] The constraints include spatial constraints and acceleration constraints;

[0119] The spatial constraint is that the candidate path does not collide with the tray geometry model or the anti-collision area. A random tree is generated based on the current pose of the robotic arm, which wraps the spherical area around the optimal placement pose. Sampling points are randomly generated in the workspace. The node closest to the sampling point is selected from the tree. A new node is generated by expanding the step size δ from the nearest node toward the sampling point. In a specific embodiment, δ can be 80mm. The new node pose is checked to see if it collides with the gantry. If it does, the node is discarded.

[0120] In one specific implementation, an 80mm safety buffer zone is set around the gantry.

[0121] The acceleration constraint is that the corner acceleration of the candidate path does not exceed a preset threshold.

[0122] To estimate the joint angular acceleration at the path corner, in one specific implementation, three consecutive nodes are selected, and the joint angle corresponding to each node is determined by a rotary encoder installed on the robotic arm. Combining the displacement changes and time intervals between nodes, the rate of change of the joint angle is calculated, and thus the corner acceleration is obtained.

[0123] The calculated corner acceleration is compared with a preset threshold. If the acceleration exceeds the threshold, it means that the robotic arm is moving too fast, which may cause the package to slip due to inertia. In this case, the path node or path segment will be discarded and re-sampled or replanned. If the acceleration is within the threshold range, the node or path segment will be retained.

[0124] For a complete candidate path, perform the above acceleration calculation and judgment segment by segment to ensure that the corner accelerations on the entire path meet the requirements. If there are any parts that do not meet the requirements, optimize and adjust the path or directly exclude the path.

[0125] S28. Score candidate paths that meet the constraints according to the preset path scoring rules. The scoring content includes spatial safety score and acceleration smoothness score. Spatial safety is judged based on the minimum distance between the path and obstacles. The greater the distance, the higher the spatial safety score. Acceleration smoothness is judged based on the reciprocal of the maximum corner acceleration of the path. The smaller the acceleration, the higher the acceleration smoothness score.

[0126] The final score is obtained by weighting the pre-set spatial safety score and acceleration smoothness score. The weighting coefficients can be adjusted.

[0127] S29. The candidate path with the highest score is taken as the palletizing path and converted into a robotic arm joint angle-time series.

[0128] S3. Control the flexible gripper device to grasp the package to be palletized, and adjust the gripping posture in real time based on feedback from the pre-set torque sensor. Control the robotic arm to perform palletizing according to the palletizing path. The specific steps include the following:

[0129] S31. The initial clamping force is obtained by calling the preset clamping force parameters according to the material of the package. The flexible gripper device includes a gripper and a pneumatic tendon that drives the gripper. A torque sensor is pre-installed on the flexible gripper device, and a rotary encoder and an inertial measurement unit are pre-installed on the robotic arm.

[0130] Torque sensors are used to detect the force on the end effector of the robotic arm, and monitor the three-dimensional force (Fx, Fy, Fz) and torque (Mx, My, Mz) at the contact point between the gripper and the package in real time; rotary encoders are used to provide feedback on joint angles and read the real-time angles of each joint of the robotic arm; inertial measurement units are used to detect the end effector vibration state and collect the three-axis acceleration and angular velocity of the robotic arm.

[0131] S32, control the pneumatic tendon to drive the gripper to grasp the package to be stacked with initial clamping force.

[0132] In this embodiment, the initial clamping force of the woven bag is The initial clamping force of the foam box is The initial clamping force of the rigid box is .

[0133] S33. During the palletizing process of the robotic arm, the real-time clamping force is monitored by a torque sensor. The deformation of the package to be stacked is estimated by the rate of change of torque. When the deformation exceeds a preset deformation threshold, i.e. The clamping force is dynamically adjusted in real time.

[0134] The formula for dynamic adjustment is: ;in These are the preset control parameters.

[0135] Maintain pressure change rate .

[0136] S34. Obtain the joint angles fed back by the rotary encoder and the end effector acceleration collected by the inertial measurement unit. Fuse the data through Kalman filtering to estimate the actual grasping pose deviation between the actual pose and the target pose.

[0137] Actual grasping pose deviation includes position deviation ( , , ), angle deviation ( , , ).

[0138] S35. Generate joint compensation commands based on the actual grasping posture deviation and the PID controller, and correct the grasping posture of the robotic arm based on the joint compensation commands.

[0139] In this embodiment, the process of calculating the compensation amount based on the actual grasping posture deviation and the PID controller is as follows: the control amount is synthesized based on the PID controller and the preset control parameters, and the joint compensation command is generated based on the control amount.

[0140] In this embodiment, the control parameters are Kp=0.8, Ki=0.2, and Kd=0.1.

[0141] The joint motors of the robotic arm receive joint compensation commands, adjust the joint angles, and achieve posture correction.

[0142] S36. When the torque sensor detects the preset contact force condition, it triggers the gripper release command to complete the placement of the package to be stacked.

[0143] In this embodiment, when the torque sensor detects the Z-axis contact force... Fz When the temperature exceeds 5N, a gripper release command is triggered, completing the package placement. In this embodiment, Fz The condition >5N is determined when the package contacts the surface of the pallet.

[0144] S37. Obtain the current position of the package, and calculate the coordinates of the package's center of gravity based on the package dimensions, such as... Figure 3 As shown.

[0145] The coordinates of the barycenter are, ;

[0146] ;

[0147] Place pose: (x,y, ), To place the rotation angle.

[0148] Obtain the center-of-gravity coordinates of the remaining packages on the current pallet, and calculate the overall center-of-gravity coordinates of all packages on the pallet by combining the center-of-gravity coordinates of the current package. Specifically, this involves accumulating the center-of-gravity contribution of all packages.

[0149] Obtain the stable center coordinates of the pallet, and calculate the center of gravity offset by combining the overall center of gravity coordinates. The offset is the Euclidean distance between the overall center of gravity coordinates and the stable center coordinates of the pallet.

[0150] S38. Compare the center of gravity offset with the preset warning range.

[0151] In one feasible implementation, three threshold ranges are set, including a safe range, a warning range, and a danger range, and the center of gravity offset is compared with each of these three threshold ranges.

[0152] A center of gravity offset of ≤50mm is considered safe, and parcels can continue to be stacked while the status of parcels on the tray is updated in real time.

[0153] A center of gravity offset between 50mm and 80mm falls within the warning range. When the center of gravity offset falls within the warning range, a center of gravity offset control strategy is implemented by combining the overall center of gravity coordinates and the stable center coordinates.

[0154] The center of gravity offset control strategy aims to prioritize placing subsequent packages in the area opposite to the center of gravity and increase lateral pressure by 10%-20% according to the clamping force to suppress the tendency to tip over.

[0155] If the center of gravity offset is greater than 80mm, it is considered a dangerous range. The system will immediately trigger emergency procedures, stop the current pallet stacking, and control the AGV under the pallet to be dispatched to the manual re-inspection area.

[0156] S4. When the pallet segmentation result matches the optimal pallet type, generate a 3D palletizing model, bind the 3D palletizing model to the AGV cart under the pallet, and control the AGV cart to automatically enter the warehouse. The specific steps include the following:

[0157] S41. Perform real-time analysis on the segmentation results to determine whether the package layout in the segmentation results matches the optimal stacking type. Specifically, the system determines whether the available space on the current pallet, the package layout, etc., match the conditions required by the optimal stacking type.

[0158] For example, check whether the size, shape, and position of the available gaps can be used to place the remaining packages in the optimal stacking configuration, and whether the position and orientation of the already placed packages conform to the specifications of the optimal stacking configuration.

[0159] If the conditions are met, the segmentation result matches the optimal stacking type.

[0160] By analyzing the segmentation results in real time and determining whether they match the optimal pallet type, it can be ensured that the palletizing meets expectations.

[0161] S42. Obtain the package information and placement pose of all packages on the pallet, and generate a 3D palletizing model of the current pallet and all packages using the WebGL engine.

[0162] Specifically, the system collects the position coordinates, size, and orientation information of all packages on the current pallet, as well as relevant data of the pallet itself, and converts this data into a format suitable for WebGL processing, including converting spatial coordinate data into three-dimensional vectors and converting the geometric shape information of packages and the pallet into polygonal mesh data.

[0163] The WebGL engine is used to create a 3D scene. Based on the converted data, 3D models of packages and pallets are created and drawn into the scene to form a complete 3D palletizing model. The current palletizing status is displayed in real time. The 3D palletizing model is generated by the WebGL engine to realize the visualization of the palletizing status.

[0164] S43. Associate identification information with the palletizing 3D model and obtain information about the AGV cart under the pallet, including the AGV cart's number, location, and operating status.

[0165] Establish a correspondence in the system database and bind the identification information with the information of the AGV, so that the system can match the palletizing 3D model with a specific AGV, providing a basis for subsequent control of the AGV.

[0166] Based on the package information, navigation information is generated for the AGV vehicle. The navigation information includes the location coordinates of the target warehouse and the driving path planning instructions. After receiving this information, the AGV vehicle drives autonomously according to the planned path.

[0167] The system controls AGVs to transport pallets carrying stacked packages to the target warehouse according to navigation information, completing the automatic warehousing operation, which facilitates subsequent cargo management and transportation and improves logistics efficiency.

[0168] Based on the same inventive concept described above, this application also discloses an automatic palletizing device. The automatic palletizing system includes a robotic arm and a flexible gripper device at the end of the robotic arm, such as... Figure 4 As shown, it includes the following modules:

[0169] The optimal pallet type matching module is used to obtain the package information of the package to be palletized, and combine the package information with the historical pallet type data of the automatic palletizing system to match the optimal pallet type for the current package to be palletized;

[0170] The palletizing path planning module is used to perform point cloud segmentation on pallets containing packages using a preset segmentation model to obtain segmentation results. It then combines the optimal pallet type and a preset path planning algorithm to calculate pallet gaps and generate palletizing paths.

[0171] The gripping posture correction module is used to control the flexible gripper device to grip the package to be palletized. It combines the feedback from the pre-set torque sensor to correct the gripping posture in real time and control the robotic arm to perform palletizing according to the palletizing path.

[0172] The automatic storage module generates a 3D palletizing model when the pallet segmentation result matches the optimal palletizing pattern. It then binds the 3D palletizing model to the AGV cart under the pallet and controls the AGV cart to automatically store the pallet.

[0173] In a specific feasible implementation, the optimal stack type matching module includes the following units:

[0174] The first optimal pallet type matching unit is used to obtain the package information of the package to be palletized according to the preset acquisition device. The acquisition device includes a barcode scanning device and a binocular depth camera. The barcode scanning device scans the package to be palletized to obtain the waybill information. The binocular depth camera obtains the package surface material and package size of the package to be palletized. The waybill information generates a unique ID of the package to be palletized and summarizes the package surface material and package size into package information and associates it with the ID.

[0175] The second optimal pallet type matching unit is used to determine the position of the pallet for stacking packages based on the waybill information and to determine whether the pallet is empty.

[0176] If the pallet is empty, the system retrieves the historical successful pallet patterns from the cloud based on the surface material of the package, selects candidate pallet patterns from the historical successful pallet patterns using the cosine similarity algorithm, selects the optimal pallet pattern from the candidate pallet patterns based on the preset optimal pallet pattern matching rules, and uploads the package information and the optimal pallet pattern corresponding to the current package to be palletized to the cloud.

[0177] If the pallet is not empty, retrieve the optimal stacking configuration for the current pallet from the cloud.

[0178] In one specific feasible implementation, the second optimal stack type matching unit includes the following sub-units:

[0179] The optimal stacking type matching subunit is used to construct a reinforcement learning environment that includes a state space, an action space, and a reward function. The reward function is calculated based on the feasible path of the package to be stacked, the spatial score of the candidate stacking type, and the support score.

[0180] The candidate stack type is quantitatively evaluated. The quantitative evaluation includes feasible path quantitative analysis, spatial score calculation and support score calculation to obtain feasible path score, spatial score and support score.

[0181] Based on the PPO reinforcement learning algorithm, a policy network is trained and iteratively optimized by combining the state space and action space. Based on the trained policy network, candidate stacking patterns are simulated and executed. The comprehensive reward score is calculated based on the feasible path, space score, and support score of the packages to be stacked. The candidate stacking pattern with the highest comprehensive reward score is output as the optimal stacking pattern.

[0182] In one specific implementation scheme, the palletizing path planning module includes the following units:

[0183] The first palletizing path planning unit is used to acquire the point cloud data of the pallet used to stack the parcels to be palletized, and to segment the parcel outline on the pallet according to the preset YOLOv8 segmentation model and the pallet point cloud data to obtain the segmentation result, which includes available gaps.

[0184] The second palletizing path planning unit is used to perform preliminary screening of available gaps in the segmentation results based on the optimal pallet type, and to exclude available gaps that conflict with the structure of the optimal pallet type.

[0185] The third palletizing path planning unit is used to traverse all available gaps obtained after the initial screening, calculate the minimum bounding rectangle of each gap, filter out all minimum bounding rectangles not smaller than the package size and summarize them into a gap candidate set, calculate the space utilization rate and optimal placement pose of each candidate gap in the gap candidate set, summarize all optimal placement poses into a candidate placement pose list, and arrange the candidate placement pose list according to the space utilization rate.

[0186] The fourth palletizing path planning unit is used to obtain the current pose of the robotic arm and the candidate gaps and their optimal placement pose, construct the pallet geometric model based on the pallet point cloud data, and obtain the workspace model carrying anti-collision area markers.

[0187] The fifth palletizing path planning unit is used to generate multiple candidate paths that meet preset constraints based on the improved RRT* algorithm. The constraints include spatial constraints and acceleration constraints. The spatial constraint is that the candidate path does not collide with the pallet geometry model or the anti-collision area. The acceleration constraint is that the corner acceleration of the candidate path does not exceed a preset threshold.

[0188] The sixth palletizing path planning unit is used to score the candidate paths that meet the constraints according to the preset path scoring rules, and to take the candidate path with the highest score as the palletizing path.

[0189] In one specific implementation scheme, the grasping posture correction module includes the following units:

[0190] The first gripping posture correction unit is used to obtain the initial gripping force by calling the preset gripping force parameters according to the package material. The flexible gripper device includes a gripper and a pneumatic tendon that drives the gripper. A torque sensor is pre-installed on the flexible gripper device, and a rotary encoder and an inertial measurement unit are pre-installed on the robotic arm.

[0191] The second gripping posture correction unit is used to control the pneumatic tendon to drive the gripper to grip the package to be stacked with an initial gripping force.

[0192] The third gripping posture correction unit is used to monitor the real-time gripping force through a torque sensor during the palletizing process of the robotic arm. When the deformation of the package to be palletized exceeds the preset deformation threshold, the real-time gripping force is dynamically adjusted.

[0193] The fourth grasping posture correction unit is used to obtain the joint angles fed back by the rotary encoder and the end-effector acceleration collected by the inertial measurement unit to estimate the actual grasping posture deviation, generate joint compensation commands based on the actual grasping posture deviation and the PID controller, and correct the grasping posture of the robotic arm based on the joint compensation commands.

[0194] The fifth gripping posture correction unit is used to trigger the gripper release command when the torque sensor detects the preset contact force condition, thereby completing the placement of the package to be palletized.

[0195] In one specific implementation, an automated palletizing device further includes the following modules:

[0196] The center of gravity offset warning module allows users to obtain the current position of the package, calculate the center of gravity coordinates of the current package based on the package size, obtain the center of gravity coordinates of the other packages on the current tray, and calculate the overall center of gravity coordinates of all packages on the tray based on the center of gravity coordinates of the current package.

[0197] Obtain the stable center coordinates of the pallet, calculate the center of gravity offset by combining the overall center of gravity coordinates, and compare the center of gravity offset with the preset warning range;

[0198] When the center of gravity offset falls within the warning range, a center of gravity offset control strategy is executed by combining the overall center of gravity coordinates and the stable center coordinates.

[0199] Based on the same inventive concept described above, this application also discloses a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set can be loaded and executed by a processor to implement the automatic palletizing method provided in the above method embodiments.

[0200] Based on the same inventive concept described above, this application also discloses a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the automatic palletizing method as described above.

[0201] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code.

[0202] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An automatic palletizing method, characterized in that, An automated palletizing system, comprising a robotic arm and a flexible gripper device at the end of the robotic arm, is applied to such a system, comprising the following steps: Obtain package information for the parcels to be palletized, and combine this package information with historical pallet type data from the automated palletizing system to match the optimal pallet type for the current parcels to be palletized; specifically, The system acquires package information of packages to be palletized using a preset acquisition device, which includes a barcode scanner and a binocular depth camera. The barcode scanner scans the packages to be palletized to obtain the waybill information. The binocular depth camera acquires the surface material and size of the packages to be palletized. The system generates a unique ID for the packages to be palletized based on the waybill information and associates the surface material and size of the packages with the ID. Based on the information on the waybill, determine the position of the pallet used to place the parcels to be stacked, and determine whether the pallet is empty; If the pallet is empty, the historical successful pallet patterns of the automatic palletizing system are obtained from the cloud based on the surface material of the package. Candidate pallet patterns are selected from the historical successful pallet patterns based on the cosine similarity algorithm. The optimal pallet pattern is selected from the candidate pallet patterns based on the preset optimal pallet pattern matching rules. The package information and the optimal pallet pattern corresponding to the current package to be palletized are uploaded to the cloud. If the pallet is not empty, obtain the optimal stacking type of the pallet from the cloud; The pallet on which the package is stacked is segmented by point cloud using a preset segmentation model to obtain the segmentation result. The pallet gap is calculated and the stacking path is generated by combining the optimal stacking type and the preset path planning algorithm. The flexible gripper device is controlled to grasp the package to be palletized, and the grasping posture is corrected in real time based on the feedback from the pre-set torque sensor. The robotic arm is then controlled to perform palletizing according to the palletizing path. When the segmentation result of the pallet matches the optimal stacking type, a 3D palletizing model is generated, and the 3D palletizing model is bound to the AGV trolley under the pallet to control the AGV trolley to automatically enter the warehouse.

2. The automatic palletizing method according to claim 1, characterized in that, The step of selecting the optimal stacking type from the candidate stacking types according to the preset optimal stacking type matching rules specifically includes the following steps: A reinforcement learning environment is constructed, which includes a state space, an action space, and a reward function. The reward function is calculated based on the feasible path of the package to be palletized, the spatial score of the candidate pallet type, and the support score. The candidate stack type is quantitatively evaluated, including feasible path quantitative analysis, spatial score calculation and support score calculation, to obtain feasible path score, spatial score and support score. Based on the PPO reinforcement learning algorithm, a policy network is trained and iteratively optimized by combining the state space and the action space. Based on the trained policy network, candidate stacking patterns are simulated and executed. A comprehensive reward score is calculated based on the feasible path of the package to be stacked, the space score, and the support score. The candidate stacking pattern with the highest comprehensive reward score is output as the optimal stacking pattern.

3. The automatic palletizing method according to claim 1, characterized in that, The process of segmenting the pallets containing the packages using a preset segmentation model, calculating the pallet gaps and generating the palletizing path using the optimal stacking type and a preset path planning algorithm, specifically includes the following steps: Acquire pallet point cloud data for pallets used to stack the parcels to be palletized, segment the parcel outlines on the pallet according to the preset YOLOv8 segmentation model and the pallet point cloud data, and obtain the segmentation result, which includes available gaps. Based on the optimal stacking type, the available gaps in the segmentation results are initially screened to exclude available gaps that conflict with the structure of the optimal stacking type. Traverse all available gaps obtained after preliminary screening, calculate the minimum bounding rectangle of each gap, filter out all minimum bounding rectangles not smaller than the package size and summarize them into a gap candidate set, calculate the space utilization rate and optimal placement pose of each candidate gap in the gap candidate set, summarize all the optimal placement poses into a candidate placement pose list, and arrange the candidate placement pose list according to the space utilization rate.

4. The automatic palletizing method according to claim 3, characterized in that, It also includes the following steps: The current pose of the robotic arm and the candidate gaps and their optimal placement poses are obtained. Based on the pallet point cloud data, a pallet geometric model is constructed, and a workspace model carrying anti-collision area markers is obtained. Multiple candidate paths satisfying preset constraints are generated based on the improved RRT* algorithm. The constraints include spatial constraints and acceleration constraints. The spatial constraint is that the candidate path does not collide with the tray geometry model or the anti-collision area. The acceleration constraint is that the corner acceleration of the candidate path does not exceed a preset threshold. The candidate paths that meet the constraints are scored according to the preset path scoring rules, and the candidate path with the highest score is taken as the palletizing path.

5. The automatic palletizing method according to claim 1, characterized in that, The process of controlling the flexible gripper device to grasp the package to be palletized, and adjusting the grasping posture in real time based on feedback from a pre-set torque sensor, and controlling the robotic arm to perform palletizing according to the palletizing path, specifically includes the following steps: The initial clamping force is obtained by calling the preset clamping force parameters according to the package material. The flexible gripper device includes a gripper and a pneumatic tendon that drives the gripper. A torque sensor is pre-installed on the flexible gripper device. A rotary encoder and an inertial measurement unit are pre-installed on the robotic arm. The pneumatic tendon is controlled to drive the gripper to grasp the package to be stacked with the initial gripping force. During the palletizing process of the robotic arm, the real-time clamping force is monitored by the torque sensor. When the deformation of the package to be palletized exceeds the preset deformation threshold, the real-time clamping force is dynamically adjusted. The actual grasping posture deviation is estimated by obtaining the joint angle fed back by the rotary encoder and the end-effector acceleration collected by the inertial measurement unit. Based on the actual grasping posture deviation and the PID controller, a joint compensation command is generated, and the grasping posture of the robotic arm is corrected based on the joint compensation command. When the torque sensor detects a preset contact force condition, it triggers a gripper release command to complete the placement of the package to be stacked.

6. The automatic palletizing method according to claim 5, characterized in that, After the placement of the parcels to be stacked is completed, the following steps are also included: Obtain the current position of the package, calculate the center of gravity coordinates of the current package based on the package size, obtain the center of gravity coordinates of the remaining packages on the current tray, and calculate the overall center of gravity coordinates of all packages on the tray based on the center of gravity coordinates of the current package. Obtain the stable center coordinates of the pallet, calculate the center of gravity offset by combining the overall center of gravity coordinates, and compare the center of gravity offset with the preset warning range; When the center of gravity offset falls within the warning range, a center of gravity offset control strategy is executed by combining the overall center of gravity coordinates and the stable center coordinates.

7. An automatic palletizing device, characterized in that, Based on an automated palletizing system, which is based on the automated palletizing method of claim 1, the automated palletizing system includes a robotic arm and a flexible gripper device at the end of the robotic arm, and includes the following modules: The optimal pallet type matching module is used to acquire package information of packages to be palletized, and combine the package information with historical pallet type data of the automatic palletizing system to match the optimal pallet type for the current package to be palletized. Specifically, it acquires package information of packages to be palletized according to a preset acquisition device, which includes a barcode scanner and a binocular depth camera. The barcode scanner scans the package to be palletized to obtain the waybill information, and the binocular depth camera acquires the surface material and size of the package to be palletized. Based on the waybill information, it generates a unique ID for the package to be palletized and summarizes the surface material and size of the package. The package information is associated with the ID; the position of the pallet for stacking the package to be palletized is determined based on the waybill information, and it is determined whether the pallet is empty; if the pallet is empty, the historical successful pallet patterns of the automatic palletizing system are obtained from the cloud based on the surface material of the package, and candidate pallet patterns are selected from the historical successful pallet patterns based on the cosine similarity algorithm; the optimal pallet pattern is selected from the candidate pallet patterns based on the preset optimal pallet pattern matching rule, and the package information and the optimal pallet pattern corresponding to the current package to be palletized are uploaded to the cloud; if the pallet is not empty, the optimal pallet pattern of the current pallet is obtained from the cloud. The palletizing path planning module is used to perform point cloud segmentation on the pallets on which packages are stacked using a preset segmentation model to obtain the segmentation result, and to calculate the pallet gap and generate the palletizing path by combining the optimal pallet type and the preset path planning algorithm. The gripping posture correction module is used to control the flexible gripper device to grip the package to be palletized, and to correct the gripping posture in real time in combination with the feedback of the pre-set torque sensor, and to control the robotic arm to perform palletizing according to the palletizing path. The automatic storage module is used to generate a 3D palletizing model when the segmentation result of the pallet matches the optimal palletizing type, bind the 3D palletizing model to the AGV trolley under the pallet, and control the AGV trolley to automatically store the pallet.

8. A smart terminal, characterized in that, The method includes a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the at least one instruction, at least one program, code set, or instruction set is loaded and executed by the processor to implement the automated palletizing method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by a processor to implement the automatic palletizing method as described in any one of claims 1 to 6.

Images