Conveying and boxing system of self-adaptive stacking robot

Abstract

The invention relates to the technical field of intelligent stacking, and discloses a self-adaptive stacking robot conveying and boxing system which comprises a space analysis module, an adaptability analysis module, a path planning module, a stability analysis module and a stacking reconstruction module. Spatial topology information in the box body is obtained; based on the spatial topology information, carrying out stacking adaptability analysis on the article feature data of the to-be-boxed articles to obtain a boxing sequence of the to-be-boxed articles; based on the boxing sequence, path planning is conducted on the self-adaptive stacking robot, and a control instruction of the self-adaptive stacking robot is obtained; based on the execution result of the control instruction, stability analysis is conducted on the stacking state in the box body, and a stacking stability coefficient in the box body is obtained; based on the stacking stability coefficient, stacking reconstruction is conducted on the stacked objects lower than a preset stability threshold value; the conveying and boxing precision of the self-adaptive stacking robot can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent stacking technology, and in particular to a conveying and packing system for an adaptive stacking robot. Background Technology

[0002] Existing technologies for adaptive stacking robots in box-packing operations have significant limitations in the spatial analysis stage. They often rely solely on single-view images or simple dimensional measurements to determine the internal spatial state of the box, failing to perform accurate point cloud registration of multi-view depth information and lacking a step to construct a 3D voxel spatial framework based on point cloud representation. This results in the inability to accurately obtain the spatial topology information inside the box. Consequently, there are biases in the perception of the distribution of stacked items, and inaccurate judgments on the geometric attributes and spatial distribution relationships of effective space areas. This leads to mismatches between items to be packed and available space, not only reducing space utilization but also easily causing spatial conflicts during item placement and increasing the frequency of adjustments during the packing operation.

[0003] Meanwhile, existing technologies have significant shortcomings in controlling stacking stability. During the stability analysis phase, they typically rely solely on simple center-of-gravity estimation or visual observation, failing to obtain the three-dimensional spatial configuration of the stacked items through multi-view image analysis. They also lack comprehensive evaluation by incorporating pressure distribution information at the bottom of the container, resulting in an inability to accurately calculate center-of-gravity offset and pressure distribution stability indicators, making it difficult to obtain reliable stacking stability coefficients. Furthermore, when the stacking state falls below the stability threshold, existing technologies lack a systematic spatial pose interference analysis and reconstruction strategy deduction mechanism. This makes it impossible to formulate feasible stacking reconstruction schemes for already stacked items, making it difficult to adjust unstable stacking structures in a timely manner. This not only easily leads to problems such as item tipping and damage but also significantly reduces packing efficiency and increases operating costs due to frequent rework. Summary of the Invention

[0004] This invention provides an adaptive stacking robot conveying and packing system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a conveying and packing system for an adaptive stacking robot, characterized in that the system includes a spatial analysis module, an adaptive analysis module, a path planning module, a stability analysis module, and a stacking reconstruction module, wherein: The spatial analysis module is used to perform spatial analysis on the distribution of items already stacked in the box to obtain the spatial topology information inside the box; The adaptive analysis module is used to perform stacking adaptive analysis on the item feature data of the items to be packed based on the spatial topology information, so as to obtain the packing sequence of the items to be packed. The path planning module is used to plan the path for the adaptive stacking robot based on the packing sequence, and obtain the control commands for the adaptive stacking robot. The stability analysis module is used to perform stability analysis on the stacking state inside the box based on the execution result of the control command, and obtain the stacking stability coefficient inside the box. The stack reconstruction module is used to reconstruct stacked items that are below a preset stability threshold based on the stack stability coefficient.

[0006] In a preferred embodiment, when the spatial analysis module performs spatial analysis on the distribution of items already stacked in the container to obtain the spatial topology information inside the container, it is specifically used for: Point cloud registration is performed on the multi-view depth information inside the box to obtain the point cloud representation of the inside of the box. Based on the point cloud representation, the space inside the box is divided into meshes to obtain a three-dimensional voxel space framework inside the box. Spatial connectivity analysis was performed on the three-dimensional voxel space framework to obtain the effective spatial region inside the box. The spatial distribution relationship of the geometric attribute features of the effective spatial region is constructed to obtain the spatial topology information inside the box.

[0007] In a preferred embodiment, when the spatial analysis module performs the meshing of the space inside the box based on the point cloud representation to obtain a three-dimensional voxel spatial framework inside the box, it is specifically used for: Based on the internal dimensional parameters of the box, a spatial grid structure is constructed inside the box to obtain the reference coordinate system inside the box; Based on the reference coordinate system, the point cloud representation is spatially discretized to obtain the voxel set of the reference coordinate system; Spatial continuity is verified on the voxel set that is in an occupied state to obtain the three-dimensional voxel spatial framework inside the box.

[0008] In a preferred embodiment, when the spatial analysis module performs spatial discretization of the point cloud representation based on the reference coordinate system to obtain the voxel set of the reference coordinate system, it is specifically used for: Determine the meshing accuracy parameters and grid size parameters of the reference coordinate system; Based on the division accuracy parameters, the point cloud representation is divided into spatial grids to obtain the spatial voxels of the reference coordinate system; Point cloud density statistical analysis is performed on the spatial voxels to obtain the occupancy status attributes of the spatial voxels; The spatial voxels are binarized and labeled based on the occupancy status attribute to obtain the occupied voxels of the spatial voxels; Spatial topology optimization is performed on the occupied voxels to obtain the voxel set of the reference coordinate system.

[0009] In a preferred embodiment, when the adaptive analysis module performs stacking adaptive analysis on the item feature data of the items to be packed based on the spatial topology information to obtain the packing sequence of the items to be packed, it is specifically used for: Geometric feature parameters and physical attribute parameters are extracted from the product feature data and compiled into multidimensional parameters of the product to be packed. Spatial compatibility is evaluated on the available spatial regions in the spatial topology information to obtain the adaptation candidate regions inside the box; The space utilization efficiency of the candidate adaptation regions is evaluated to obtain the optimal score of the candidate adaptation regions; Based on the preferred score, the candidate regions are prioritized to obtain the packing sequence of the items to be packed.

[0010] In a preferred embodiment, when the adaptive analysis module performs a spatial compatibility assessment on the available spatial regions in the spatial topology information to obtain the adaptation candidate regions inside the box, it is specifically used for: A matching degree analysis is performed on the geometric feature parameters and the multidimensional parameters to obtain a compatibility evaluation index between the available space area and the items to be packed. Based on the compatibility evaluation index, the attitude feasibility of the box is verified to obtain the adaptation candidate region inside the box.

[0011] In a preferred embodiment, when the path planning module executes the path planning for the adaptive stacking robot based on the packing sequence to obtain the control commands for the adaptive stacking robot, it is specifically used for: The path sequence of the items in the packing sequence is analyzed to obtain the target path sequence of the adaptive stacking robot. Dynamic performance constraint detection is performed on the target path sequence to obtain the pulse control instruction set of the adaptive stacking robot; The integrity of the pulse control instruction set is checked to obtain the control instructions for the adaptive stacking robot.

[0012] In a preferred embodiment, when the stability analysis module performs stability analysis on the stacking state inside the box based on the execution result of the control command to obtain the stacking stability coefficient inside the box, it is specifically used for: Extract multi-view images of the stacked items inside the box; Contour analysis is performed on the multi-view appearance images to obtain the three-dimensional spatial configuration of the stacked items; Extract pressure distribution information from the bottom of the chamber; A centroid projection analysis is performed on the three-dimensional spatial configuration to obtain the centroid offset of the stacked items. The formula for calculating the centroid offset is as follows: ; In the formula, This is the centroid offset. For the first The quality of an item For the first The x-coordinate of the center of gravity of an item in the box coordinate system. For the first The vertical coordinate of the center of gravity of an item in the box coordinate system. Let x be the x-coordinate of the geometric center coordinates of the bottom of the box. The ordinate is the ordinate of the geometric center coordinates at the bottom of the box. The pressure distribution information is reconstructed to obtain a pressure distribution map at the bottom of the box. The pressure distribution map is matched and analyzed with a preset standard stable pressure distribution map to obtain the pressure distribution stability index at the bottom of the box. The stacking stability coefficient inside the box is obtained by fusing the center of gravity offset and the pressure distribution stability index.

[0013] In a preferred embodiment, when the stacking reconstruction module performs stacking reconstruction on stacked items below a preset stability threshold based on the stacking stability coefficient, it is specifically used for: The stacking stability coefficient is compared with a preset stability threshold, and spatial pose interference analysis is performed on the stacked items that are less than the preset stability threshold to obtain the spatial constraint relationship of the stacked items. A feasibility assessment of the spatial constraints is conducted to obtain a reconstruction strategy for the stacked items. Based on the reconstruction strategy, the stacked items are reconstructed.

[0014] In a preferred embodiment, when the stacking reconstruction module performs the feasibility assessment of the spatial constraint relationship to obtain the reconstruction strategy for the stacked items, it is specifically used for: Spatial degree-of-freedom analysis is performed on the spatial constraint relationship to obtain a candidate set of reconstruction strategies for the stacked items; The candidate set of reconstruction strategies is deduced and verified to obtain a feasibility score for the implementation of the candidate set of reconstruction strategies; The feasibility scores are used to rank and optimize strategies to obtain the reconstruction strategies for the stacked items. Compared with the prior art, the present invention has the following beneficial effects: 1. This invention uses a spatial analysis module to perform point cloud registration on multi-view depth information inside the container, constructs a three-dimensional voxel spatial framework and analyzes spatial topology information. Combined with an adaptive analysis module, it evaluates the compatibility of item feature data with available space and prioritizes them to generate an optimized packing sequence. At the same time, the path planning module parses the target path based on the sequence and generates precise control commands, which effectively improves the utilization rate of container space, reduces item placement conflicts, and improves the overall efficiency of conveying and packing operations.

[0015] 2. This invention extracts the three-dimensional spatial configuration from multi-view apparent images through a stability analysis module, calculates the center of gravity offset and pressure distribution stability index by combining pressure distribution information, and obtains a reliable stacking stability coefficient. Furthermore, the stacking reconstruction module can perform spatial pose interference analysis on unstable stacks based on this coefficient and formulate feasible reconstruction strategies to achieve adaptive adjustment of the stacking structure, significantly enhancing the stability of the stacking state, reducing the risk of items tipping over, and ensuring the safety and continuity of packing operations. Attached Figure Description

[0016] Figure 1 This is a system architecture diagram of an adaptive stacking robot conveying and packing system provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0019] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0020] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0021] In practice, the server-side equipment deployed in an adaptive stacking robot's packing and delivery system may consist of one or more devices. This adaptive stacking robot packing and delivery system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node, providing an adaptive stacking robot packing and delivery system to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage each user terminal. Or, it can be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide an adaptive stacking robot packing and delivery system to each user terminal.

[0022] In terms of implementation, the adaptive stacking robot's conveying and packing system and the user terminal are mutually adapted. That is, if the adaptive stacking robot's conveying and packing system is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the adaptive stacking robot's conveying and packing system is implemented as a website, then the user terminal is implemented as a webpage; or if the adaptive stacking robot's conveying and packing system is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0023] like Figure 1The diagram shown is a system architecture diagram of an adaptive stacking robot conveying and packing system provided in an embodiment of the present invention.

[0024] The adaptive stacking robot conveying and packing system 100 of this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the adaptive stacking robot conveying and packing system 100 may include a space analysis module 101, an adaptability analysis module 102, a path planning module 103, a stability analysis module 104, and a stacking reconstruction module 105. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0025] In this embodiment of the invention, in an adaptive stacking robot's conveying and packing system, each of the above-mentioned modules can be implemented independently and can call other modules. Here, "calling" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. The adaptive stacking robot's conveying and packing system provided in this embodiment of the invention allows for adjustment of the system's applicability by adding modules and directly calling them, without modifying the program code. This enables cluster-based horizontal expansion, achieving the goal of quickly and flexibly expanding the adaptive stacking robot's conveying and packing system. In practical applications, the above modules can be set in the same device or different devices, or they can be set in virtual devices, such as service instances in a cloud server.

[0026] The following describes, with reference to specific embodiments, each component and its specific workflow of an adaptive stacking robot's conveying and packing system: The spatial analysis module 101 is used to perform spatial analysis on the distribution of items already stacked in the box to obtain the spatial topology information inside the box. In this embodiment of the invention, when the spatial analysis module performs spatial analysis on the distribution of items already stacked in the box to obtain the spatial topology information inside the box, it is specifically used for: Point cloud registration is performed on the multi-view depth information inside the box to obtain the point cloud representation of the inside of the box. Based on the point cloud representation, the space inside the box is divided into meshes to obtain a three-dimensional voxel space framework inside the box. Spatial connectivity analysis was performed on the three-dimensional voxel space framework to obtain the effective spatial region inside the box. The spatial distribution relationship of the geometric attribute features of the effective spatial region is constructed to obtain the spatial topology information inside the box.

[0027] When the spatial analysis module performs mesh generation of the space inside the box based on the point cloud representation to obtain a three-dimensional voxel space framework inside the box, it is specifically used for: Based on the internal dimensional parameters of the box, a spatial grid structure is constructed inside the box to obtain the reference coordinate system inside the box; Based on the reference coordinate system, the point cloud representation is spatially discretized to obtain the voxel set of the reference coordinate system; Spatial continuity is verified on the voxel set that is in an occupied state to obtain the three-dimensional voxel spatial framework inside the box.

[0028] When the spatial analysis module performs spatial discretization of the point cloud representation based on the reference coordinate system to obtain the voxel set of the reference coordinate system, it is specifically used for: Determine the meshing accuracy parameters and grid size parameters of the reference coordinate system; Based on the division accuracy parameters, the point cloud representation is divided into spatial grids to obtain the spatial voxels of the reference coordinate system; Point cloud density statistical analysis is performed on the spatial voxels to obtain the occupancy status attributes of the spatial voxels; The spatial voxels are binarized and labeled based on the occupancy status attribute to obtain the occupied voxels of the spatial voxels; Spatial topology optimization is performed on the occupied voxels to obtain the voxel set of the reference coordinate system.

[0029] Specifically, point cloud registration of multi-view depth information inside the enclosure is performed by setting up multiple depth sensors at different locations inside the enclosure to simultaneously collect depth data from various angles inside the enclosure. This depth data is then converted into point cloud data containing spatial coordinate information. The coordinates of points corresponding to the same physical location in different point cloud data are then compared, and the spatial angles and positions of each point cloud data are adjusted to ensure that all point cloud data are accurately aligned in the same spatial dimension, eliminating deviations caused by differences in viewing angles. Finally, the aligned point cloud data are merged into a complete dataset, which is the point cloud representation of the inside of the enclosure.

[0030] Specifically, based on the needs of the internal space analysis of the box, the fineness of the spatial mesh division under the reference coordinate system is determined. This fineness is the division accuracy parameter. Then, combined with the length, width, and height dimensions inside the box and the division accuracy parameter, the side length of each cubic mesh is calculated. The size of the cube corresponding to this side length is the mesh size parameter. According to the determined division accuracy parameter, the internal space of the box corresponding to the point cloud representation is divided according to the cubic mesh structure of the reference coordinate system. Each independent cubic unit formed after the division is the spatial voxel of the reference coordinate system.

[0031] Furthermore, examine each spatial voxel, count the number of points in the point cloud data contained within each spatial voxel, and compare the counted number of points with a preset occupancy determination standard. If the number of points in the spatial voxel reaches or exceeds the preset occupancy determination standard, the spatial voxel is determined to be in an occupied state. If the number of points in the spatial voxel does not reach the preset occupancy determination standard, the spatial voxel is determined to be in an unoccupied state. The occupancy or unoccupancy determination result for each spatial voxel is the occupancy status attribute of that spatial voxel.

[0032] Furthermore, based on the occupancy status attribute of each spatial voxel, the spatial voxels determined to be in an occupied state are specifically binarized and marked, that is, only two different identifiers are used to correspond to the occupied state and the unoccupied state respectively. All spatial voxels marked as occupied are collected, and these spatial voxels marked as occupied are the occupied voxels of the spatial voxels.

[0033] Furthermore, the spatial connectivity between each occupied voxel and its neighboring occupied voxels is examined. If an occupied voxel has no neighboring occupied voxels and the amount of point cloud data in the occupied voxel is much less than the amount of point cloud data in a normally occupied voxel, it is removed from the occupied voxels. For occupied voxels that are spatially adjacent and belong to the same item, they are integrated into a voxel group corresponding to a whole occupied area. After removing abnormal voxels and integrating adjacent voxels, the set of remaining occupied voxels is the voxel set of the reference coordinate system.

[0034] Furthermore, the spatial position of each occupied voxel in the voxel set of the reference coordinate system is examined one by one. It is checked whether there are spatial gaps between adjacent occupied voxels. If there are gaps, it is determined whether the gaps are in line with the reasonable gaps that may exist when actual items are stacked. If the gaps exceed the reasonable range, the spatial positions of adjacent occupied voxels are adjusted to make their connection conform to reality. At the same time, it is checked whether the distribution of occupied voxels is consistent with the actual possible stacking pattern of items inside the box. If there are unreasonable situations such as occupied voxels distributed outside the box boundary or floating in unsupported spatial positions, the state or position of these occupied voxels is corrected. After the above spatial continuity check and correction, the occupied and unoccupied voxels in the reference coordinate system are formed into a complete voxel framework according to the actual spatial distribution. This complete voxel framework is the three-dimensional voxel spatial framework inside the box.

[0035] Furthermore, to construct the spatial distribution relationship of the geometric attributes of the effective spatial region, it is necessary to measure the geometric attributes of each effective spatial region, such as its length, width, height, overall shape, and the coordinates of its center in the reference coordinate system. Then, by calculating the center distance and relative position between different effective spatial regions, the positional relationship and distribution pattern of each effective spatial region inside the box are clarified. These geometric attributes and spatial relationships are integrated into a structured information set, which is the spatial topology information inside the box.

[0036] Specifically, based on the internal dimensions of the box, a spatial grid structure is constructed inside the box. Taking a vertex on the bottom surface of the box as the origin, three mutually perpendicular axes are set along the length, width, and height directions of the box to form the three coordinate axes of the reference coordinate system. Then, according to the analysis requirements, equal intervals are set on the three coordinate axes, and planes perpendicular to the coordinate axes are drawn along the interval points. The cubic grid structure formed by the intersection of these planes covers the entire internal space of the box. The coordinate system on which this structure depends is the reference coordinate system inside the box.

[0037] Furthermore, based on the reference coordinate system, the point cloud representation is spatially discretized. First, the precision of dividing the reference coordinate system (i.e., the division accuracy parameter) and the actual size of each grid (i.e., the grid size parameter) need to be determined. Then, the space corresponding to the point cloud representation is divided into multiple cubic grids according to the division accuracy parameter. Each grid is a spatial voxel of the reference coordinate system. The number of point clouds contained in each spatial voxel is counted and compared with a preset occupancy judgment threshold. Voxels with a number that reaches or exceeds the threshold are judged as occupied, and those that do not reach the threshold are judged as unoccupied. Based on the judgment result, the spatial voxels are marked. Occupied voxels are marked as occupied voxels. Finally, the spatial connection relationship between occupied voxels is checked, isolated occupied voxels caused by point cloud noise are removed, and occupied voxels that are spatially adjacent and belong to the same item are merged. The set of occupied voxels after optimization is the voxel set of the reference coordinate system.

[0038] Furthermore, determining the grid division accuracy parameters and grid size parameters of the reference coordinate system requires setting the grid division accuracy parameters based on the minimum size of the items to be analyzed inside the box and the required level of detail in the spatial analysis. The higher the accuracy parameters, the finer the grid division. Then, the length of each grid in the three coordinate axes of the reference coordinate system is calculated based on the grid division accuracy parameters. This length is the grid size parameter, ensuring that the grid size accurately reflects the space occupied by the items and does not cause analysis errors due to being too large or too small.

[0039] Furthermore, based on the division accuracy parameters, the point cloud representation is divided into spatial grids. According to the grid size determined by the division accuracy parameters, multiple parallel planes are sequentially divided along the three coordinate axes in the reference coordinate system. The cubic grid formed by the intersection of these planes is the spatial voxel. The boundary coordinates of each spatial voxel are determined by its position in the reference coordinate system. All points in the point cloud representation are included in the corresponding spatial voxels, and the resulting grid is the spatial voxel of the reference coordinate system.

[0040] Furthermore, point cloud density statistical analysis is performed on the spatial voxels. It is necessary to count the number of point clouds contained in each spatial voxel and compare the number with a preset occupancy determination threshold. When the number of point clouds is greater than or equal to the threshold, it is determined that the spatial voxel is occupied by an item, and its occupancy status attribute is occupied. When the number of point clouds is less than the threshold, it is determined that the spatial voxel is not occupied by an item, and its occupancy status attribute is unoccupied.

[0041] Furthermore, based on the occupancy status attribute, the spatial voxels are binarized and marked. According to the occupancy status attribute of the spatial voxels, spatial voxels with the occupancy status attribute of "occupied" are marked as "occupied", and spatial voxels with the occupancy status attribute of "unoccupied" are marked as "unoccupied". All spatial voxels marked as "occupied" are collected, and the resulting set is the set of spatial voxels that are occupied.

[0042] Furthermore, spatial topology optimization of the occupied voxels requires checking the spatial connectivity between them, removing isolated occupied voxels caused by point cloud noise (i.e., voxels that have no adjacent relationship with other occupied voxels), and merging occupied voxels that are spatially adjacent and belong to the same item. This ensures that the optimized occupied voxels accurately reflect the actual stacking shape of the items inside the box. The set of optimized occupied voxels is the voxel set of the reference coordinate system.

[0043] In summary, point cloud registration is performed on the multi-view depth information inside the enclosure. For example, the ICP algorithm is used for point cloud registration to obtain the point cloud representation of the enclosure. This is done by collecting depth data from different directions using multiple depth sensors and converting it into point clouds. The spatial angles and positions of each point cloud are adjusted to make them accurately aligned. After eliminating viewpoint deviations, they are fused into a complete dataset, i.e., the point cloud representation.

[0044] In summary, the method of meshing the internal space of the box based on point cloud representation to obtain a three-dimensional voxel space framework inside the box involves first establishing a reference coordinate system based on the internal dimensional parameters of the box, then dividing the internal space of the box covered by the point cloud representation into multiple uniform cubic voxels according to the meshing rules determined by the analysis requirements, while simultaneously counting the occupancy status of each voxel, and finally integrating all voxels and their occupancy status to form a three-dimensional voxel space framework that can reflect the internal spatial structure of the box.

[0045] In summary, spatial connectivity analysis of the three-dimensional voxel spatial framework yields the effective space area inside the box. This involves examining the connection status between unoccupied voxels and surrounding voxels, determining whether a continuous region is formed, and filtering out unoccupied areas that are interconnected and have sufficient volume to accommodate the items to be packed.

[0046] In summary, constructing the spatial distribution relationship of the geometric attributes of the effective spatial region to obtain the spatial topological information inside the box involves measuring the geometric attributes of each region, calculating the clear distribution patterns such as the center distance and relative position between regions, and integrating this information into structured spatial topological information.

[0047] In summary, a spatial grid structure is constructed inside the box based on the internal dimensional parameters to obtain the reference coordinate system inside the box. This coordinate system is established by setting an origin and three mutually perpendicular axes, and then dividing the grid structure according to requirements. The coordinate system on which this structure depends is the reference coordinate system.

[0048] In summary, spatial discretization of point cloud representation based on a reference coordinate system yields a voxel set in the reference coordinate system. This involves dividing the internal space of the box corresponding to the point cloud representation into multiple uniform cubic units according to the grid structure of the reference coordinate system. The distribution of point cloud within each voxel is statistically analyzed to determine whether it is in an occupied state. Finally, all voxels are integrated according to their occupancy status, and the resulting set containing the occupancy information of each voxel is the voxel set in the reference coordinate system.

[0049] In summary, the spatial continuity verification of the voxel set in the occupied state is used to obtain the three-dimensional voxel spatial framework inside the box. This involves checking the spatial connection relationship between the voxels in the occupied state, checking for isolated occupied voxels or voxel distributions that do not conform to the actual stacking pattern of items, correcting unreasonable distributions, and ensuring that the distribution of occupied voxels is consistent with the actual spatial pattern inside the box. The complete voxel structure formed after verification is the three-dimensional voxel spatial framework inside the box.

[0050] In summary, dividing the point cloud representation into spatial grids based on the partitioning accuracy parameters to obtain spatial voxels in the reference coordinate system involves dividing the space corresponding to the point cloud representation into multiple uniform grid units according to the set partitioning accuracy. These units are the spatial voxels in the reference coordinate system.

[0051] In summary, point cloud density statistical analysis of spatial voxels yields their occupancy status attributes. This involves counting the number of point clouds contained within each spatial voxel and using preset criteria to determine whether the voxel is occupied by an item, thereby clarifying the occupancy status attributes of each spatial voxel.

[0052] In summary, the method of binarizing and marking spatial voxels based on their occupancy status attributes to obtain the occupied voxels is to specifically mark the spatial voxels determined to be occupied based on their occupancy status, and collect all voxels marked as occupied to form a set of occupied voxels.

[0053] In summary, spatial topology optimization of occupied voxels to obtain the voxel set of the reference coordinate system involves sorting out the spatial connections between occupied voxels, removing unreasonable isolated voxels, and integrating adjacent and related voxels. The optimized set is the voxel set of the reference coordinate system.

[0054] The adaptive analysis module 102 is used to perform stacking adaptive analysis on the item feature data of the items to be packed based on the spatial topology information, so as to obtain the packing sequence of the items to be packed. In this embodiment of the invention, when the adaptive analysis module performs stacking adaptive analysis on the item feature data of the items to be packed based on the spatial topology information to obtain the packing sequence of the items to be packed, it is specifically used for: Geometric feature parameters and physical attribute parameters are extracted from the product feature data and compiled into multidimensional parameters of the product to be packed. Spatial compatibility is evaluated on the available spatial regions in the spatial topology information to obtain the adaptation candidate regions inside the box; The space utilization efficiency of the candidate adaptation regions is evaluated to obtain the optimal score of the candidate adaptation regions; Based on the preferred score, the candidate regions are prioritized to obtain the packing sequence of the items to be packed.

[0055] When the adaptive analysis module performs spatial compatibility evaluation on the available spatial regions in the spatial topology information to obtain the adaptation candidate regions inside the box, it is specifically used for: A matching degree analysis is performed on the geometric feature parameters and the multidimensional parameters to obtain a compatibility evaluation index between the available space area and the items to be packed. Based on the compatibility evaluation index, the attitude feasibility of the box is verified to obtain the adaptation candidate region inside the box.

[0056] Specifically, the geometric feature parameters and physical property parameters extracted from the product feature data are compiled into multi-dimensional parameters of the product to be packed. This is achieved by using a high-precision 3D laser scanner to completely scan the product. The scanner emits laser beams from multiple angles and receives reflected signals. Based on the signal return time, the 3D coordinates of each point on the surface of the product are calculated to form a point cloud model containing millions of points.

[0057] Furthermore, geometric feature parameters of the object are extracted from the point cloud model. Specifically, the maximum length, maximum width, and maximum height of the object in three-dimensional space are determined by identifying the extreme points in the point cloud model. The overall shape of the object is determined by analyzing the distribution density and continuity of the point cloud. For regular-shaped objects, it is determined whether they belong to the type of cuboid, cube, cylinder, or sphere. For irregular-shaped objects, the position and size of the protruding and concave parts are marked.

[0058] Furthermore, the weight of the item is precisely measured using an electronic scale, down to the gram level; multiple pressure tests are conducted on different locations on the surface of the item using a hardness tester, and the average value is taken as the material hardness parameter of the item; a coefficient of friction tester is used to slide the surface of the item relative to a standard test plate, and the required force value is recorded and converted into a surface smoothness parameter.

[0059] Furthermore, these geometric feature parameters and physical property parameters are organized according to a preset structured format. Geometric feature parameters are listed as a separate category, including sub-items such as length, width, height, shape type, and dimensions of special parts. Physical property parameters are listed as another category, including sub-items such as weight, material hardness, and surface smoothness. The two categories of parameters together constitute the multi-dimensional parameters of the items to be packed.

[0060] Furthermore, a spatial compatibility assessment is performed on the available spatial areas in the spatial topology information to obtain the suitable candidate areas inside the box. First, detailed data of all available spatial areas are retrieved from the spatial topology information, including the three-dimensional dimensions, overall shape, center coordinates, distribution of surrounding items, and load-bearing capacity of the supporting structure of each area.

[0061] Furthermore, the geometric feature parameters of the items to be packed are compared with the three-dimensional dimensions and shape of the available space area. Each item is checked to see if its length is less than or equal to the length of the available space area, its width is less than or equal to the width of the available space area, and its height is less than or equal to the height of the available space area. At the same time, the shape of the item is observed to see if it matches the shape of the available space area. For example, whether a cylindrical item is suitable to be placed in a cylindrical space area, and whether the protruding part of an irregularly shaped item can be accommodated by the concave part of the space area.

[0062] Furthermore, based on geometric comparison, the physical property parameters in the multi-dimensional parameters are used for further evaluation. This includes checking whether the weight of the item is within the maximum load-bearing capacity of the support structure below the available space area to avoid damage to the support structure due to overweight. The hardness of the item's material is also checked. If the material is soft, it is necessary to confirm whether there are any hard objects around the available space area that may cause it to be squeezed and deformed. If the material is hard, it is necessary to confirm whether it will damage any fragile items already placed nearby. The surface smoothness of the item is also analyzed. If the surface smoothness is high, it is necessary to assess whether it is easy to slide due to slight vibration after placement and whether additional fixing measures are needed.

[0063] Furthermore, by considering factors such as geometric size matching, shape adaptability, weight adaptability, material compatibility, and surface characteristic adaptability, a comprehensive compatibility evaluation index is generated. Based on this index, all available space areas are screened, and areas that meet the preset qualification standards are selected. These areas are the candidate adaptation areas inside the box.

[0064] Furthermore, the space utilization efficiency of the candidate adaptation areas is evaluated to obtain the optimal score for each candidate adaptation area. This involves conducting a multi-dimensional efficiency analysis for each candidate adaptation area. First, the space utilization rate is calculated by dividing the volume of the items to be packed by the volume of the candidate adaptation area. The percentage obtained is the space utilization rate. A higher utilization rate indicates less wasted space in that area, and it accounts for the highest weight in the score.

[0065] Furthermore, the usability of the remaining space is analyzed. After simulating the placement of items to be packed, the three-dimensional dimensions and shape of the remaining space are measured. If the remaining space is a regular cuboid with a large volume, it indicates that other items can be placed there later, and a higher score can be given. If the remaining space is a scattered small area or an irregular shape, it is difficult to reuse, and a lower score is given.

[0066] Furthermore, the rationality of the area's location is evaluated. If the suitable candidate area is located in a corner or edge of the box, and does not obstruct the placement path of other items after items are placed there, while also leaving more space in the central area, bonus points can be awarded. If the area is located in the center of the box, but can serve as a stable support base for other items after items are placed there, bonus points can also be awarded appropriately. If the location of the area would require subsequent items to be placed around it or would make stacking difficult, points will be deducted accordingly. Simultaneously, the overall stability of the box is considered, analyzing the impact of the area on the overall center of gravity of the box after items are placed there. If it brings the center of gravity closer to the center of the box, which is beneficial to overall stability, the score can be increased; if it causes an excessive shift in the center of gravity, the score will be decreased.

[0067] Furthermore, specific scoring rules and weights are set according to the above factors. Space utilization rate has the highest weight, remaining space availability has a relatively high weight, and location rationality and overall stability also have corresponding weights. The comprehensive score of each candidate region is calculated according to these weights, and this score is the optimal score of the candidate region.

[0068] Furthermore, based on the preferred scores, the candidate matching regions are prioritized to obtain the packing sequence of the items to be packed. This involves organizing the preferred scores of all candidate matching regions and arranging them in descending order of score. For regions with the same score, their space utilization rate is further compared, with regions having higher utilization rates ranked higher. If the space utilization rates are also the same, the rationality of their location is compared, with regions located in corners or on the edges given priority.

[0069] Furthermore, after sorting, the candidate area with the highest score is determined as the first placement position for the items to be packed, corresponding to the first placement order; the area with the second highest score is determined as the second placement position, corresponding to the second placement order, and so on, forming a clear order of placement. Simultaneously, the placement posture of the item at each position is recorded, i.e., the angle and orientation at which the item is placed in that area, ensuring consistency with the previous posture feasibility verification results. These sequentially arranged placement positions and their corresponding postures are integrated to form a sequence containing placement order, specific location, and posture information; this sequence is the packing sequence for the items to be packed.

[0070] Specifically, a matching degree analysis is performed on the geometric feature parameters and the multidimensional parameters to obtain a compatibility evaluation index between the available space area and the items to be packed. This involves extracting key data of the items to be packed from the geometric feature parameters, including specific values ​​and shape descriptions of length, width, and height, and comparing these values ​​with the length, width, height, and shape descriptions of the available space area item by item. The ratios of each dimension of the item to the corresponding dimension of the space area are calculated, such as the ratio of item length to space length, item width to space width, and item height to space height. The higher the average of these three ratios and the closer they are to one, the higher the size matching degree. For shape matching, if the item is a regular cuboid and the space area is also a cuboid, the shape matching degree is high. If the item is a cylinder and the space area is a cube, it is necessary to check whether the diameter of the cylinder is smaller than the side length of the cube and whether its height is smaller than the height of the cube to determine the degree of shape fit.

[0071] Furthermore, physical property data are extracted from multidimensional parameters. The weight of the item is compared with the maximum load-bearing capacity of the space area to calculate the weight ratio. The smaller the ratio, the better the load-bearing adaptability. The hardness of the item's material is compared with the hardness of the materials of surrounding items. If the hardness of the two is similar, the material compatibility is good. If the difference is large, it is necessary to assess whether there is a risk of damage. The smoothness of the item's surface is compared with the smoothness of the contact surface of the space area. If both are low in smoothness, the coefficient of friction is high and the placement stability is good. If one of them is high in smoothness, the risk of sliding needs to be considered.

[0072] Furthermore, the size matching degree, shape matching degree, load-bearing adaptability, material compatibility and stability adaptability are weighted and calculated according to their respective weights to obtain a value within a range. This value is the compatibility evaluation index between the available space area and the items to be packed.

[0073] Furthermore, based on the compatibility evaluation index, the attitude feasibility of the box interior is verified to obtain the adaptation candidate region inside the box interior. First, a qualified threshold for the compatibility evaluation index is set, and all available space regions with scores reaching or exceeding the threshold are screened out as preliminary candidate regions.

[0074] Furthermore, for each preliminary candidate area, the items to be packed are simulated to be placed in a variety of possible postures. The types of postures are determined according to the shape of the items. Cuboid items usually have a variety of basic postures with different faces as the bottom, cylindrical items have two basic postures, and irregularly shaped items are placed according to their symmetry features and the position of the protruding parts.

[0075] Furthermore, in each pose, the object model is placed into the spatial region model using 3D modeling software. It is then checked whether each vertex of the object is inside the spatial region and whether the surface of the object overlaps with the boundary of the spatial region or the surface of other objects placed around it. If any overlap exists, the pose is deemed infeasible.

[0076] Furthermore, the coordinates of the object's center of gravity in this posture are calculated, and it is checked whether the center of gravity is within the support surface range of the spatial region. The support surface range is the area formed by the inward contraction of the bottom boundary of the spatial region by a certain distance. If the center of gravity is within the support surface range, the object is not easy to tip over after being placed, and its stability is good; if the center of gravity exceeds the support surface range, the stability is poor, and this posture is not feasible.

[0077] Furthermore, for the poses that pass the boundary check and the center of gravity check, they are recorded as feasible poses. Each preliminary candidate region is determined as an adaptation candidate region as long as there is at least one feasible pose. Finally, all regions that meet the conditions together form the adaptation candidate regions inside the box.

[0078] In summary, spatial compatibility assessment of available spatial areas in spatial topology information is used to obtain suitable candidate areas inside the box. This involves combining the geometric features and physical properties of the items to be packed with information such as the size, shape, and load-bearing capacity of available spatial areas. The assessment evaluates the degree of compatibility between the two in terms of geometric matching, material compatibility, and stability, and selects areas that meet the preset qualification standards as suitable candidate areas.

[0079] In summary, evaluating the spatial utilization efficiency of suitable candidate areas and obtaining the optimal score for suitable candidate areas involves analyzing each candidate area from multiple dimensions, such as spatial utilization rate, availability of remaining space, location rationality, and overall stability. A comprehensive score is calculated based on the weight of each factor, thereby quantifying the spatial utilization efficiency of the area.

[0080] In summary, the candidate regions are prioritized based on the optimal score to obtain the packing sequence of the items to be packed. The candidate regions are sorted from high to low score. If the scores are the same, the better region is selected first. The placement order, specific location and item posture of each region are determined and integrated to form a clear packing sequence.

[0081] In summary, by performing matching degree analysis on geometric feature parameters and multidimensional parameters, a compatibility evaluation index for the available space area and the items to be packed is obtained. This involves comparing the geometric dimensions, shape, and other characteristics of the items, as well as their physical properties such as weight and material hardness, with the dimensions, shape, and load-bearing capacity of the available space area to evaluate the degree of size matching, shape fit, and load-bearing capacity. The results are then weighted and calculated to form a quantitative compatibility evaluation index.

[0082] In summary, based on compatibility evaluation indicators, the attitude feasibility of the box interior is verified to obtain the adaptation candidate regions inside the box. A qualified threshold is set to screen out the initial candidate regions. The object is simulated to be placed in each region in various possible postures. It is checked whether it overlaps with the boundary or other objects, whether the center of gravity is within the support range, and the feasible postures are recorded. The region with at least one feasible posture is determined as the adaptation candidate region.

[0083] The path planning module 103 is used to perform path planning for the adaptive stacking robot based on the packing sequence, and obtain the control commands for the adaptive stacking robot. In this embodiment of the invention, when the path planning module executes the path planning for the adaptive stacking robot based on the packing sequence to obtain the control instructions for the adaptive stacking robot, it is specifically used for: The path sequence of the items in the packing sequence is analyzed to obtain the target path sequence of the adaptive stacking robot. Dynamic performance constraint detection is performed on the target path sequence to obtain the pulse control instruction set of the adaptive stacking robot; The integrity of the pulse control instruction set is checked to obtain the control instructions for the adaptive stacking robot.

[0084] Specifically, the path analysis of the placement order of items in the packing sequence is performed to obtain the target path sequence of the adaptive stacking robot. This involves first accurately extracting the placement order of each item from the packing sequence, as well as the three-dimensional coordinates of the placement position of each item, while simultaneously recording the three-dimensional coordinates of the robot's current picking position.

[0085] Furthermore, for each item, a path planning model is built in a three-dimensional virtual space, with the pickup location as the starting point and the placement location as the ending point. By importing the three-dimensional coordinate data of the items already placed inside the box and the boundary coordinate data of the box walls, all possible collision points on the path are comprehensively investigated.

[0086] Furthermore, a point-by-point inspection method is employed. Starting from the starting point, each potential path point is sequentially checked to determine whether it overlaps with already placed items or box walls. If overlap exists, the path direction is adjusted until a continuous route that completely avoids all obstacles is planned. Along this route, according to the robot's motion characteristics, a key node is set at fixed intervals. Each node not only marks the robot's three-dimensional coordinates at that location but also clearly records the robot's turning angle, ensuring that the robot can smoothly complete turning or attitude adjustments when passing through nodes, avoiding item swaying or path deviation due to sudden angle changes.

[0087] Furthermore, the paths of all items are sequentially connected according to the placement order of the packing sequence to form a continuous sequence containing the starting point, all key nodes, the ending point, and detailed parameters of each route segment. This sequence is the target path sequence of the adaptive stacking robot.

[0088] Furthermore, dynamic performance constraint detection is performed on the target path sequence to obtain the pulse control instruction set of the adaptive stacking robot. The hardware performance parameters of the adaptive stacking robot are collected in advance, including core indicators such as the maximum speed of the drive motor, the upper limit of the output torque, the maximum overall moving speed of the robot, the maximum acceleration, the minimum turning radius, and the maximum angle of a single turn.

[0089] Furthermore, following the sequence of the target path, the route segments between each key node are analyzed segment by segment, calculating the actual length of each segment, the theoretically required moving speed, and the turning angle. If the calculated moving speed exceeds the robot's maximum moving speed, the speed of that segment is adjusted to the robot's maximum permissible speed. Based on the robot's maximum acceleration, the specific durations of the acceleration and deceleration phases are planned in detail to ensure that the robot can smoothly accelerate from its current speed to its maximum speed when entering that segment and smoothly decelerate to the preset speed when approaching the next key node. If the calculated turning angle exceeds the robot's maximum turning angle in a single turn, this turning action is broken down into multiple consecutive small-angle turns. Each small angle is controlled within the robot's permissible range, and a brief pause is set between two adjacent small-angle turns to ensure the stability of the robot's turning process and avoid body swaying caused by continuous turning.

[0090] Furthermore, based on the adjusted speed, angle, and duration, a corresponding pulse signal for the drive motor is generated. Each pulse signal contains three key parameters: pulse frequency, number of pulses, and duration. The pulse frequency determines the motor's rotation speed, the number of pulses corresponds to the motor's rotation angle, and the duration controls the motor's working time, ensuring that the robot can accurately complete the movement or turning action of that section of the route.

[0091] Furthermore, all pulse signals are arranged in the order of the target path sequence to form a complete set, which is the pulse control instruction set of the adaptive stacking robot.

[0092] Furthermore, the integrity of the pulse control instruction set is checked to obtain the control instructions for the adaptive stacking robot. First, a detailed checklist is formulated based on the key nodes in the target path sequence and the preset action requirements.

[0093] Furthermore, the list clearly outlines the action type for each key node, such as movement or turning; action parameters, such as movement speed, turning angle, and action duration; and the corresponding instruction number, ensuring that the action of each node has clear verification standards.

[0094] Furthermore, following the order of the inspection checklist, each instruction in the pulse control instruction set is checked one by one to see if there is an instruction that completely matches a certain node action in the checklist, ensuring that there are no missing node actions and no duplicate instructions.

[0095] Furthermore, the pulse parameters of each instruction are rigorously verified to check whether the pulse frequency is within the allowable operating frequency range of the drive motor, avoiding motor overload due to excessively high frequency or affecting the accuracy of the action due to excessively low frequency; check whether the rotation angle corresponding to the number of pulses is consistent with the planned steering angle to ensure the accuracy of the steering action; verify whether the duration is sufficient to meet the moving distance of the route segment to prevent failure to reach the target node due to insufficient time or inefficiency due to excessive time.

[0096] Furthermore, if any instruction is found to have parameters that exceed the limit or do not match the list during the inspection process, the system immediately returns to the dynamic performance constraint detection step, readjusts the parameters, and generates the corresponding pulse signal. Once all instructions have passed inspection, the pulse control instruction set is organized into a standard format recognizable by the robot, clearly defining the execution order of each instruction and its corresponding triggering conditions, such as automatically triggering the next instruction after the previous one is completed. This rigorously inspected and standardized instruction set constitutes the control instructions for the adaptive stacking robot.

[0097] In summary, path analysis of the placement order of items in the packing sequence yields the target path sequence for the adaptive stacking robot. This involves analyzing the path the robot must traverse from picking up to placing items based on the order and specific location of the items, clarifying the direction, distance, and other information of each path segment, and forming the target path sequence that the robot must follow.

[0098] In summary, dynamic performance constraint detection of the target path sequence yields the pulse control instruction set for the adaptive stacking robot. This involves combining the robot's dynamic performance constraints, checking whether the speed, acceleration, and other parameters of each segment of the target path meet the constraints, and converting the path information into pulse instructions that can control the robot's movement, thus forming the corresponding instruction set.

[0099] In summary, performing a completeness check on the pulse control instruction set to obtain the control instructions for the adaptive stacking robot involves checking whether the pulse control instruction set covers all necessary control information, ensuring that the instructions are complete and conflict-free, and confirming that the instructions are complete and valid as the robot's control instructions.

[0100] The stability analysis module 104 is used to perform stability analysis on the stacking state inside the box based on the execution result of the control command, and obtain the stacking stability coefficient inside the box. In this embodiment of the invention, when the stability analysis module performs stability analysis on the stacking state inside the box based on the execution result of the control command to obtain the stacking stability coefficient inside the box, it is specifically used for: Extract multi-view images of the stacked items inside the box; Contour analysis is performed on the multi-view appearance images to obtain the three-dimensional spatial configuration of the stacked items; Extract pressure distribution information from the bottom of the chamber; A centroid projection analysis is performed on the three-dimensional spatial configuration to obtain the centroid offset of the stacked items. The formula for calculating the centroid offset is as follows: ; In the formula, This is the centroid offset. For the first The quality of an item For the first The x-coordinate of the center of gravity of an item in the box coordinate system. For the first The vertical coordinate of the center of gravity of an item in the box coordinate system. Let x be the x-coordinate of the geometric center coordinates of the bottom of the box. The ordinate is the ordinate of the geometric center coordinates at the bottom of the box. The pressure distribution information is reconstructed to obtain a pressure distribution map at the bottom of the box. The pressure distribution map is matched and analyzed with a preset standard stable pressure distribution map to obtain the pressure distribution stability index at the bottom of the box. The stacking stability coefficient inside the box is obtained by fusing the center of gravity offset and the pressure distribution stability index.

[0101] Specifically, to extract multi-view images of the stacked items inside the box, a high-definition camera is installed on the front, back, left, right, and top sides of the box. The camera lens parameters are uniform and the shooting ranges are interconnected to ensure complete coverage of the entire outer surface of the stacked items. All cameras are connected through the same synchronous control device. When the device sends a trigger signal, all cameras shoot simultaneously to avoid image deviations caused by time differences. A ring-shaped uniform lighting device is installed inside the box to eliminate the influence of shadows and reflections on image details. After shooting, the images are sorted according to their orientation to form a multi-view image that can comprehensively reflect the appearance of the stacked items.

[0102] Furthermore, contour analysis is performed on the multi-view appearance images to obtain the three-dimensional spatial configuration of the stacked items. First, noise reduction is performed on each image to remove pixel interference. Then, the stacked item area is separated by threshold segmentation. Feature points on the item contours are extracted using an edge detection tool, and the pixel coordinates of these points are recorded. Then, based on the camera's installation position and angle parameters, the corresponding feature points in the images from different viewpoints are paired. The spatial coordinates of these points in the three-dimensional coordinate system are calculated using the triangulation principle. All three-dimensional coordinate points are connected to form a three-dimensional frame. The internal structural details of the frame are supplemented, and finally, a complete three-dimensional spatial configuration of the stacked items is constructed.

[0103] Furthermore, to extract the pressure distribution information at the bottom of the box, an array of multiple pressure sensors is laid at the bottom of the box. The sensors are evenly distributed in a matrix. Each sensor can convert the sensed pressure into an electrical signal and transmit it to the data acquisition module. After the items are stacked, the data acquisition module reads the pressure values ​​of all sensors and arranges these values ​​in order according to the actual position of the sensors at the bottom, forming a raw data set reflecting the pressure conditions in different areas of the bottom of the box. This set is the pressure distribution information at the bottom of the box.

[0104] Further, a center of gravity projection analysis is performed on the three-dimensional spatial configuration to obtain the center of gravity offset of the stacked items. First, the mass of each item and its center of gravity coordinates are obtained from the three-dimensional spatial configuration, where the center of gravity coordinates include the horizontal and vertical coordinates. The product of the mass of each item and its own center of gravity horizontal coordinate is calculated and summed, and then divided by the total mass of all items to obtain the horizontal coordinate of the overall center of gravity of the stack. The vertical coordinate of the overall center of gravity is calculated using the same method. The length and width of the bottom of the box are measured to determine the horizontal and vertical coordinates of its geometric center. The difference between the horizontal coordinate of the overall center of gravity and the horizontal coordinate of the geometric center of the bottom of the box is calculated and squared. The difference between the vertical coordinate of the overall center of gravity and the vertical coordinate of the geometric center of the bottom of the box is calculated and squared. The two squared results are added together and the square root is calculated. The obtained value is the center of gravity offset of the stacked items.

[0105] Furthermore, the pressure distribution information is reconstructed to obtain a pressure distribution map of the bottom of the chamber. This involves correlating the pressure data of each sensor in the pressure distribution information with its position on the bottom of the chamber, establishing a planar coordinate system based on the actual dimensions of the bottom of the chamber, marking the sensor positions and pressure values ​​in the coordinate system, and supplementing the blank areas between sensors with pressure values ​​through linear interpolation based on the pressure data of adjacent sensors to make the pressure distribution show continuous changes. Then, the pressure values ​​are mapped to different colors according to a gradient from low to high, with the color becoming darker as the pressure increases. Finally, a pressure distribution map of the bottom of the chamber is formed that intuitively displays the pressure distribution.

[0106] Furthermore, the pressure distribution map is matched and analyzed with a preset standard stable pressure distribution map to obtain the pressure distribution stability index at the bottom of the box. First, the preset standard stable pressure distribution map is retrieved. This map shows a uniform distribution characteristic with the center of the bottom of the box as the symmetrical point. The pressure is highest in the central area and gradually decreases towards the edge. After aligning the actual pressure distribution map with the standard map, the same grid cells are divided. The pressure difference of each cell is calculated. The sum of the absolute values ​​of all differences is calculated. The relative deviation rate is obtained by dividing the sum by the total pressure value of the standard map. The relative deviation rate is subtracted from one and multiplied by one hundred. The resulting value is the pressure distribution stability index at the bottom of the box.

[0107] Furthermore, the center of gravity offset and the pressure distribution stability index are fused to obtain the stacking stability coefficient inside the box. First, the weight of the center of gravity offset is set to sixty, and the weight of the pressure distribution stability index is set to forty. The center of gravity offset is converted into a stability score. A maximum allowable offset is preset. When the offset is zero, the score is one hundred. When it is equal to the maximum allowable value, the score is zero. The intermediate value is calculated according to a linear ratio. The pressure distribution stability index is directly used as another part of the score. The score corresponding to the center of gravity offset is multiplied by sixty, and the pressure distribution stability index is multiplied by forty. The result of the sum of the two is the stacking stability coefficient inside the box.

[0108] Specifically, in the formula for calculating the center of gravity offset, This represents the offset of the center of gravity of the stacked items, which is the horizontal distance of the overall center of gravity of the stack from the geometric center of the bottom of the box. This indicates that the relevant values ​​of all items are summed. Representing the The quality of an item Representing the The x-coordinate of the center of gravity of an item in the box coordinate system. Representing the The vertical coordinate of the center of gravity of an item in the box coordinate system. This represents the sum of the products obtained by multiplying the mass of each item by the x-coordinate of its center of gravity. This represents the total mass obtained by adding up the masses of all items. This means that the sum of the products of the masses of all items and their x-coordinates, divided by the total mass, gives the x-coordinate of the overall center of gravity of the stacked items. The x-coordinate of the geometric center of the bottom of the box in the coordinate system. The vertical coordinate of the geometric center of the bottom of the box is represented in the coordinate system. This means subtracting the x-coordinate of the geometric center of the bottom of the box from the x-coordinate of the overall center of gravity of the stack, and then squaring the difference. This means subtracting the ordinate of the geometric center of the bottom of the box from the ordinate of the overall centroid of the stack, and then squaring the difference. The final result is... This refers to the offset of the center of gravity of the stacked items.

[0109] In summary, contour analysis of multi-view images yields the three-dimensional spatial configuration of stacked objects. This is achieved by analyzing the contour features of objects in images from different perspectives, extracting the spatial position and morphological information of the contours, and integrating them to form a three-dimensional spatial structure that can fully represent the stacked object form.

[0110] In summary, the center of gravity projection analysis of a three-dimensional spatial configuration to obtain the center of gravity offset of a stack of items is based on the position and weight information of each item in the three-dimensional spatial configuration. The center of gravity position of the entire stack is calculated, and then compared with the center position of the box to obtain the degree and direction of the center of gravity deviation, i.e., the center of gravity offset.

[0111] In summary, reconstructing the pressure distribution information to obtain a pressure distribution map at the bottom of the chamber involves processing the collected pressure distribution data to restore and construct the specific distribution of pressure at each location at the bottom of the chamber, presenting the pressure distribution in a visually intuitive graphical form.

[0112] In summary, by fusing parameters of the center of gravity offset and the pressure distribution stability index, the stacking stability coefficient inside the box is obtained. This coefficient is a comprehensive calculation that integrates the equilibrium state reflected by the center of gravity offset and the force balance reflected by the pressure distribution stability index into a single coefficient that can measure the overall stacking stability.

[0113] The stacking reconstruction module 105 is used to reconstruct stacked items that are below a preset stability threshold based on the stacking stability coefficient.

[0114] In this embodiment of the invention, when the stacking reconstruction module performs stacking reconstruction on stacked items below a preset stability threshold based on the stacking stability coefficient, it is specifically used for: The stacking stability coefficient is compared with a preset stability threshold, and spatial pose interference analysis is performed on the stacked items that are less than the preset stability threshold to obtain the spatial constraint relationship of the stacked items. A feasibility assessment of the spatial constraints is conducted to obtain a reconstruction strategy for the stacked items. Based on the reconstruction strategy, the stacked items are reconstructed.

[0115] When the stacking reconstruction module performs the feasibility assessment of the spatial constraint relationship to obtain the reconstruction strategy for the stacked items, it is specifically used for: Spatial degree-of-freedom analysis is performed on the spatial constraint relationship to obtain a candidate set of reconstruction strategies for the stacked items; The candidate set of reconstruction strategies is deduced and verified to obtain a feasibility score for the implementation of the candidate set of reconstruction strategies; The feasibility scores of the implementation are used to rank the strategies for optimal implementation, thereby obtaining the reconstruction strategies for the stacked items.

[0116] Specifically, the stacking stability coefficient is compared with a preset stability threshold, and spatial pose interference analysis is performed on the stacked items that are less than the preset stability threshold to obtain the spatial constraint relationship of the stacked items. First, the preset stability threshold is retrieved. This threshold is the lowest coefficient value that can ensure stacking safety, determined through multiple experiments. The calculated stacking stability coefficient is then compared with this threshold to filter out the stacked items whose coefficient is less than the threshold.

[0117] Furthermore, for each item in the stacked items, the actual position coordinates and placement posture parameters are obtained through its three-dimensional spatial configuration. The spatial positional relationship between any two items is checked one by one to determine whether there are overlapping areas or situations where the distance is too close. The combination of items that hinder movement and the specific direction of the obstruction are recorded. At the same time, the restriction range of the box boundary on the movement of items is marked. These obstruction relationships and restriction ranges are integrated to form the spatial constraint relationship of the stacked items.

[0118] Furthermore, a spatial degree-of-freedom analysis is performed on the spatial constraints to obtain a candidate set of reconstruction strategies for the stacked items. Based on the obstruction relationships and limitations recorded in the spatial constraints, the possible directions of movement and rotational angles of each item are analyzed without interference from other items or the box boundaries, determining the translational and rotational degrees of freedom of each item in three-dimensional space. Based on these degrees of freedom, multiple adjustment schemes are designed. Each scheme includes the items to be moved, the specific direction and distance of movement, the specific angle of rotation, and the order of adjustment, ensuring that no new spatial interference occurs during the adjustment process. All adjustment schemes that conform to the spatial constraints are summarized to form a candidate set of reconstruction strategies for the stacked items.

[0119] Specifically, the reconstruction strategy candidate set is deduced and verified to obtain the implementation feasibility score of the reconstruction strategy candidate set. This involves simulating the movement and rotation of objects in a three-dimensional spatial configuration according to the adjustment steps of each strategy in the reconstruction strategy candidate set, checking in real time whether there is interference between objects and between objects and boxes after each adjustment step, and recording whether there are any stutters or failures to complete the adjustment as planned during the adjustment process.

[0120] Furthermore, the expected stacking stability coefficient after the implementation of each strategy is calculated to evaluate its effect on improving stability. Based on the interference, implementation smoothness, and stability improvement effect, a quantitative score is assigned to each candidate strategy. The strategy with less interference, smoother implementation, and more significant stability improvement receives a higher score. This score is the implementation feasibility score of the reconstruction strategy candidate set.

[0121] Furthermore, the feasibility scores are used to optimize and rank strategies to obtain reconstruction strategies for the stacked items. This involves arranging the feasibility scores of all strategies in the candidate reconstruction strategy set in descending order to form an ordered list. During the ranking process, if strategies with the same score appear, the strategy with fewer adjustment steps is selected first. The first strategy in the ranked list, i.e., the strategy with the highest score and the fewest adjustment steps, is taken as the reconstruction strategy for the stacked items.

[0122] Furthermore, based on the aforementioned reconstruction strategy, the stacked items are reconstructed by controlling a stacking robot to operate on designated items sequentially according to the adjustment order specified in the determined reconstruction strategy. The robot translates the items to new positions according to the movement direction and distance set in the strategy, adjusts the placement posture of the items according to the set rotation angle, and after each item is adjusted, images are captured in real time by a camera to update the three-dimensional spatial configuration, confirming that the adjusted item does not interfere with other items or the box boundary. After all items are adjusted, the stacking stability coefficient is recalculated to ensure it reaches a preset stability threshold, thus completing the stacking reconstruction of the items already stacked inside the box.

[0123] In summary, the stacking stability coefficient is compared with a preset stability threshold, and spatial pose interference analysis is performed on stacked items that are below the threshold to obtain the spatial constraint relationship of the stacked items. The target with insufficient stability is screened out by numerical comparison, and then the position and orientation of the items in the target are analyzed to clarify the obstruction relationship between items and the box boundary restriction, and integrated to form information reflecting the spatial constraint.

[0124] In summary, the feasibility assessment of spatial constraints leads to a reconstruction strategy for stacked items. This involves analyzing the movement and rotational degrees of freedom of the items based on the constraints, designing multiple adjustment schemes and verifying their feasibility, and finally selecting the optimal scheme as the reconstruction strategy through scoring and ranking.

[0125] In summary, the reconstruction strategy for stacked items with insufficient stability within the box involves controlling related operations according to a defined strategy, adjusting the position and orientation of the items sequentially, confirming in real time that there is no interference, and ultimately ensuring that the stacking stability reaches a preset threshold to complete the re-stacking of the items.

[0126] In summary, spatial degree-of-freedom analysis of spatial constraints yields a candidate set of reconstruction strategies for stacked items. Based on the obstruction relationships and restrictions recorded in the spatial constraints, the range of movement and rotation of items without generating new interference is analyzed. Based on this, multiple schemes including adjustments to objects, operation methods, and sequences are designed, and a set of candidate reconstruction strategies for selection is formed.

[0127] In summary, the feasibility score of each reconstruction strategy candidate set is obtained by simulating the implementation process of each candidate strategy, checking for interference and smoothness in the adjustment of items, evaluating the effect on improving stability, and assigning quantitative scores accordingly.

[0128] In summary, the feasibility score is used to optimize and rank the strategies to obtain the reconstruction strategies for stacked items. The candidate strategies are ranked from high to low. When the scores are the same, the strategy with fewer adjustment steps is selected first. Finally, the strategy ranked first is selected as the reconstruction strategy for stacked items.

[0129] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0130] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A conveying and packing system for an adaptive stacking robot, characterized in that, The system includes a spatial analysis module, an adaptive analysis module, a path planning module, a stability analysis module, and a stacking reconstruction module, wherein: The spatial analysis module is used to perform spatial analysis on the distribution of items already stacked in the box to obtain the spatial topology information inside the box; The adaptive analysis module is used to perform stacking adaptive analysis on the item feature data of the items to be packed based on the spatial topology information, so as to obtain the packing sequence of the items to be packed. The path planning module is used to plan the path for the adaptive stacking robot based on the packing sequence, and obtain the control commands for the adaptive stacking robot. The stability analysis module is used to perform stability analysis on the stacking state inside the box based on the execution result of the control command, and obtain the stacking stability coefficient inside the box. The stacking reconstruction module is used to reconstruct stacked items that are below a preset stability threshold based on the stacking stability coefficient.

2. The adaptive stacking robot conveying and packing system as described in claim 1, characterized in that, When the spatial analysis module performs spatial analysis on the distribution of items already stacked in the container to obtain the spatial topology information inside the container, it is specifically used for: Point cloud registration is performed on the multi-view depth information inside the box to obtain the point cloud representation of the inside of the box. Based on the point cloud representation, the space inside the box is divided into meshes to obtain a three-dimensional voxel space framework inside the box. Spatial connectivity analysis was performed on the three-dimensional voxel space framework to obtain the effective spatial region inside the box. The spatial distribution relationship of the geometric attribute features of the effective spatial region is constructed to obtain the spatial topology information inside the box.

3. The adaptive stacking robot conveying and packing system as described in claim 2, characterized in that, When the spatial analysis module performs mesh generation of the space inside the box based on the point cloud representation to obtain a three-dimensional voxel space framework inside the box, it is specifically used for: Based on the internal dimensional parameters of the box, a spatial grid structure is constructed inside the box to obtain the reference coordinate system inside the box; Based on the reference coordinate system, the point cloud representation is spatially discretized to obtain the voxel set of the reference coordinate system; Spatial continuity is verified on the voxel set that is in an occupied state to obtain the three-dimensional voxel spatial framework inside the box.

4. The adaptive stacking robot conveying and packing system as described in claim 3, characterized in that, When the spatial analysis module performs spatial discretization of the point cloud representation based on the reference coordinate system to obtain the voxel set of the reference coordinate system, it is specifically used for: Determine the meshing accuracy parameters and grid size parameters of the reference coordinate system; Based on the division accuracy parameters, the point cloud representation is divided into spatial grids to obtain the spatial voxels of the reference coordinate system; Point cloud density statistical analysis is performed on the spatial voxels to obtain the occupancy status attributes of the spatial voxels; The spatial voxels are binarized and labeled based on the occupancy status attribute to obtain the occupied voxels of the spatial voxels; Spatial topology optimization is performed on the occupied voxels to obtain the voxel set of the reference coordinate system.

5. The adaptive stacking robot conveying and packing system as described in claim 1, characterized in that, When the adaptive analysis module performs stacking adaptive analysis on the item feature data of the items to be packed based on the spatial topology information to obtain the packing sequence of the items to be packed, it is specifically used for: Geometric feature parameters and physical attribute parameters are extracted from the product feature data and compiled into multidimensional parameters of the product to be packed. Spatial compatibility is evaluated on the available spatial regions in the spatial topology information to obtain the adaptation candidate regions inside the box; The space utilization efficiency of the candidate adaptation regions is evaluated to obtain the optimal score of the candidate adaptation regions; Based on the preferred score, the candidate regions are prioritized to obtain the packing sequence of the items to be packed.

6. The adaptive stacking robot conveying and packing system as described in claim 5, characterized in that, When the adaptive analysis module performs spatial compatibility evaluation on the available spatial regions in the spatial topology information to obtain the adaptation candidate regions inside the box, it is specifically used for: A matching degree analysis is performed on the geometric feature parameters and the multidimensional parameters to obtain a compatibility evaluation index between the available space area and the items to be packed. Based on the compatibility evaluation index, the attitude feasibility of the box is verified to obtain the adaptation candidate region inside the box.

7. The adaptive stacking robot conveying and packing system as described in claim 1, characterized in that, When the path planning module executes the path planning for the adaptive stacking robot based on the packing sequence to obtain the control commands for the adaptive stacking robot, it is specifically used for: The path sequence of the items in the packing sequence is analyzed to obtain the target path sequence of the adaptive stacking robot. Dynamic performance constraint detection is performed on the target path sequence to obtain the pulse control instruction set of the adaptive stacking robot; The integrity of the pulse control instruction set is checked to obtain the control instructions for the adaptive stacking robot.

8. The adaptive stacking robot conveying and packing system as described in claim 1, characterized in that, When the stability analysis module executes the stability analysis based on the execution result of the control command to obtain the stacking stability coefficient inside the box, it is specifically used for: Extract multi-view images of the stacked items inside the box; Contour analysis is performed on the multi-view appearance images to obtain the three-dimensional spatial configuration of the stacked items; Extract pressure distribution information from the bottom of the chamber; A centroid projection analysis is performed on the three-dimensional spatial configuration to obtain the centroid offset of the stacked items. The formula for calculating the centroid offset is as follows: ； In the formula, This is the centroid offset. For the first The quality of an item For the first The x-coordinate of the center of gravity of an item in the box coordinate system. For the first The vertical coordinate of the center of gravity of an item in the box coordinate system. Let x be the x-coordinate of the geometric center coordinates of the bottom of the box. The ordinate is the ordinate of the geometric center coordinates at the bottom of the box. The pressure distribution information is reconstructed to obtain a pressure distribution map at the bottom of the box. The pressure distribution map is matched and analyzed with a preset standard stable pressure distribution map to obtain the pressure distribution stability index at the bottom of the box. The stacking stability coefficient inside the box is obtained by fusing the center of gravity offset and the pressure distribution stability index.

9. The adaptive stacking robot's conveying and packing system as described in claim 1, characterized in that, When the stack reconstruction module performs stack reconstruction on stacked items below a preset stability threshold based on the stack stability coefficient, it is specifically used for: The stacking stability coefficient is compared with a preset stability threshold, and spatial pose interference analysis is performed on the stacked items that are less than the preset stability threshold to obtain the spatial constraint relationship of the stacked items. A feasibility assessment of the spatial constraints is conducted to obtain a reconstruction strategy for the stacked items. Based on the reconstruction strategy, the stacked items are reconstructed.

10. The adaptive stacking robot's conveying and packing system as described in claim 9, characterized in that, When the stacking reconstruction module performs the feasibility assessment of the spatial constraint relationship to obtain the reconstruction strategy for the stacked items, it is specifically used for: Spatial degree-of-freedom analysis is performed on the spatial constraint relationship to obtain a candidate set of reconstruction strategies for the stacked items; The candidate set of reconstruction strategies is deduced and verified to obtain a feasibility score for the implementation of the candidate set of reconstruction strategies; The feasibility scores of the implementation are used to rank the strategies for optimal implementation, thereby obtaining the reconstruction strategies for the stacked items.

Images