Automatic powder loading device based on visual identification

Abstract

The invention relates to the technical field of automatic loading and logistics control, in particular to an automatic powder loading device based on visual identification, which comprises a visual identification module, a flexible grabbing module, a transfer module and a stacking module, the form, internal distribution and material characteristic data of the powder bag are obtained through a multi-mode sensor, a grabbing path is generated, and grabbing operation is completed through a flexible manipulator; the transfer module dynamically adjusts the transportation state of the powder bag in combination with path optimization and posture monitoring; the stacking module monitors stress of a stacking layer in real time on the basis of the stacking thermal distribution diagram, adjustment suggestion data are generated, stacking position and angle optimization is completed through a self-correction unit, and the stability of a stacking structure is ensured; the powder loading efficiency, the stacking precision and the long-term stability are remarkably improved, and the method is suitable for efficient logistics management of complex loading scenes.

Description

Technical Field

[0001] The invention relates to the technical field of automated loading and logistics control, and in particular to an automatic powder loading device based on visual recognition. Background Art

[0002] In the process of powder loading, traditional manual loading and semi-automatic loading methods are often inefficient, inaccurate, and have high safety risks. Automatic powder loading technology can significantly improve loading efficiency, stacking accuracy, and the overall automation level of logistics processes by integrating a variety of sensor technologies and intelligent algorithms. In the prior art, powder loading usually uses a semi-automatic system with a mechanical gripping device combined with fixed path planning. Although this type of technology has partially achieved loading automation, it has obvious deficiencies in dealing with the complexity of powder bag shapes, the accuracy of gripping point selection, and stacking stability. Specifically: Existing technologies are mostly based on fixed grasping paths or mechanically restricted grasping points, which are difficult to adapt to the morphological changes of powder bags caused by differences in size, shape or stacking status, and can easily cause grasping failure or bag slippage; traditional stacking methods often lack the ability to monitor and dynamically adjust the force distribution and balance of the stacking layers in real time, resulting in uneven stacking or even instability; existing technologies lack perception and adjustment of the real-time status of powder bags (such as center of gravity offset, posture changes, etc.) during the grasping and stacking process, and cannot meet the needs of complex loading scenarios. Summary of the invention

[0003] In view of the many problems existing in the above-mentioned prior art, the present invention provides an automatic powder loading device based on visual recognition. The present invention utilizes a multimodal visual recognition module to comprehensively perceive the morphological data, internal distribution and material characteristics of the powder bag, combines dynamic path planning and inverse kinematics algorithm to generate an accurate grasping path, and completes the intelligent grasping and optimized placement of the powder bag through a flexible manipulator. The stacking module dynamically adjusts the stacking position and angle through real-time force monitoring and stacking thermal distribution map analysis, and finally forms a stable stacking structure. The present invention significantly improves the loading efficiency and stacking stability, and meets the needs of complex loading scenarios.

[0004] An automatic powder loading device based on visual recognition, comprising: A visual recognition module is used to collect the morphological data, internal distribution data and surface characteristic data of the powder bag, and generate the morphological topological map data of the powder bag through data fusion; the morphological topological map data is used to extract the grabbing node data and the stacking surface feature data; The flexible grasping module includes a flexible manipulator, which generates grasping path data according to the grasping node data and performs a grasping operation on the powder bag; during the grasping process, the flexible manipulator dynamically adjusts the grasping action by real-time monitoring the grasping force and the change of the powder bag shape, and generates optimized powder bag shape data; The transfer module includes an automatic guided robot, which completes the transfer of the powder bag from the grabbing point to the stacking point according to the optimized powder bag shape data and transfer path data, and outputs transfer status data, wherein the transfer status data includes posture change information of the powder bag, real-time center of gravity adjustment parameters and a time record of reaching the target stacking point, which is used to guide the stacking module to accurately adjust the position and angle of the powder bag; The stacking module comprises a stacking device, wherein the stacking device places the powder bag at a target stacking position according to the transport state data and the stacking surface characteristic data, and generates stacking state data for evaluating the stability of the stacking structure.

[0005] Preferably, the visual recognition module includes an RGB-D sensor, an infrared sensor and an ultrasonic sensor, wherein: The RGB-D sensor is used to obtain three-dimensional morphological data of the powder bag through optical imaging and depth information, and the three-dimensional morphological data includes the length, width, height and boundary features of the powder bag; The infrared sensor is used to detect the distribution characteristics of the powder inside the powder bag, and the distribution characteristics are obtained by calculating the infrared thermal radiation signals in different temperature ranges; The ultrasonic sensor acquires the density and edge deformation characteristics of the powder bag material by emitting sound waves to the surface of the powder bag and receiving reflected signals.

[0006] Preferably, the visual recognition module includes a data fusion unit, which jointly processes the three-dimensional morphological data, distribution characteristic data and material characteristic data, and generates morphological topological map data through a weighted average algorithm, wherein the morphological topological map data includes the geometric positions and boundary connection information of key nodes of the powder bag; the morphological topological map data is used to extract grasping node data, and the grasping node data is calculated by a geometric graph analysis algorithm.

[0007] Preferably, the flexible manipulator includes a multi-degree-of-freedom joint structure, and the joint structure realizes multi-axis rotation and extension through servo drive; pressure sensors and contact sensors are set in the fingertip area of ​​the flexible manipulator, and the pressure sensors are used to monitor the distribution of applied force during the grasping process in real time, and the contact sensors are used to detect the contact area and morphological changes of the grasping point, so as to dynamically adjust the grasping angle and strength.

[0008] Preferably, the flexible manipulator further includes a path planning unit, which generates grasping path data based on grasping node data, and the grasping path data generates a motion trajectory of the flexible manipulator through an inverse kinematics algorithm; the path planning unit dynamically optimizes the motion trajectory according to real-time feedback data, and the optimization objectives include minimizing the grasping time, uniformity of force on the powder bag, and morphological stability.

[0009] Preferably, the automatic guided robot of the transfer module includes a positioning system and an inclination sensor, wherein the positioning system generates a transfer path based on the optimized morphological data of the powder bag and the target stacking point position; the inclination sensor monitors the inclination angle of the powder bag in real time, calculates the adjustment parameters in combination with the center of gravity offset information of the powder bag, and realizes the transportation of the powder bag by adjusting the inclination angle of the automatic guided robot platform.

[0010] Preferably, the transfer status data generated by the transfer module includes real-time posture information of the powder bag, center of gravity adjustment parameters and target position arrival time of the powder bag; the center of gravity adjustment parameters are calculated based on the difference between the dynamic position of the center of gravity of the powder bag and the center of gravity position of the robot platform, and are used to guide the automatic guided robot to adjust its motion posture in real time.

[0011] Preferably, the stacking device of the stacking module includes a stacking robot arm driven by a servo motor, and the stacking robot arm accurately locates the position of the powder bag by receiving the stacking surface feature data in real time. The stacking robot arm driven by the servo motor adjusts the movement speed and placement angle of the robot arm according to the coordinates of the target stacking point; the adjustment of the placement angle is obtained based on the calculation of the stacking contact point distribution.

[0012] Preferably, the stacking module includes a thermal distribution monitoring unit, which detects the force distribution on the contact surface of the stacking layer through an infrared imaging device to generate a stacking thermal distribution map; the stacking thermal distribution map is used to analyze the force uniformity of the stacking layer and generate stacking adjustment recommendation data.

[0013] Preferably, the self-correction unit in the stacking module triggers the servo motor to adjust the stacking position and angle of the powder bag by real-time detection of the stacking stability score data, and the stacking stability score data is generated by analyzing the force concentration and overall balance of each contact point in the stacking thermal distribution diagram, and generating adjusted stacking state data.

[0014] Compared with the prior art, the advantages and beneficial effects of the present invention are: The present invention realizes intelligent extraction and dynamic adjustment of grabbing points through multimodal visual recognition technology, solves the problem of inaccurate grabbing point selection in the prior art, and adapts to the needs of complex powder bag shapes; The present invention realizes the uniformity of the force distribution of the stacked layers and the stability of the overall structure through real-time force monitoring and self-correction technology, and effectively solves the problems of uneven stacking and instability in the prior art. The present invention realizes efficient and stable operation of the whole process from grabbing to stacking of powder bags through dynamic path planning and grabbing optimization technology, thus improving the overall loading efficiency. The present invention realizes intelligent optimization of the multi-layer stacking process through stacking thermal distribution diagrams and a real-time feedback mechanism, and further ensures the long-term stability of the stacked layers. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a structural block diagram of the device of the present invention; Figure 2 This is a flow chart of the visual recognition module in the present invention; Figure 3 This is a working schematic diagram of the flexible grabbing module in the present invention; Figure 4 It is a schematic diagram of the stacking module process in the present invention. DETAILED DESCRIPTION

[0016] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure.

[0017] like Figure 1 As shown, an automatic powder loading device based on visual recognition includes: A visual recognition module is used to collect the morphological data, internal distribution data and surface characteristic data of the powder bag, and generate the morphological topological map data of the powder bag through data fusion; the morphological topological map data is used to extract the grabbing node data and the stacking surface feature data; like Figure 2 As shown in the figure, the core of the visual recognition module is to use multimodal sensors to collect comprehensive characteristics of powder bags. The collected data types include morphological data, internal distribution data and surface characteristic data. The acquisition of these data depends on the following sensors and technologies: The 3D point cloud data of the powder bag is obtained through the RGB-D sensor. The RGB-D sensor can not only capture the surface color (RGB) information of the powder bag, but also obtain the length, width, height and boundary characteristics of the powder bag through depth perception technology (Depth). These data are processed by superimposing the depth map and the RGB image to generate a 3D morphological model.

[0018] The infrared sensor uses the infrared heat radiation characteristics of different temperature zones to detect the distribution of powder in the powder bag. Specifically, the stacking density and distribution uniformity of the powder in the packaging bag will lead to differences in infrared heat radiation signals, and these differences can be used to determine whether there is uneven stacking or gaps inside the powder.

[0019] The ultrasonic sensor transmits sound waves to the surface of the powder bag and receives the reflected signal to measure its material properties and edge deformation. This process uses the correlation between the propagation speed of sound waves and the density of the material to infer the integrity and edge characteristics of the powder bag surface.

[0020] After collecting the above data, the data fusion unit jointly processes the multimodal data to generate the morphological topological map data of the powder bag. The core principle of data fusion is to integrate the characteristic information of different data sources into a unified topological model according to importance through a weighted average algorithm.

[0021] The topological graph is composed of nodes and edges, where nodes represent important geometric feature points of the powder bag, such as edge points and surface key points; edges represent the spatial connection relationship between geometric feature points, such as the continuity of surface contours or the proximity relationship of edge points.

[0022] The generation formula of the morphological topology map can be expressed as:

[0023] in, Generated morphological topological map data; Weight coefficients, used to indicate the importance of different data sources (e.g., morphological data, distribution data, surface property data); A collection of geometric features generated from different data sources; The number of data sources. By setting the weights appropriately , the system can dynamically adjust the contribution rate of different data sources to ensure that the generated topology map can fully and accurately reflect the actual status of the powder bag.

[0024] The geometric graph analysis algorithm is used to extract the grasping node data from the morphological topology graph. The grasping node refers to the location of the optimal contact point when the flexible manipulator grasps the powder bag. The extraction criteria include the stability of the node (such as whether it is close to the center of gravity) and the flatness of the boundary (such as whether the grasping point surface is smooth).

[0025] The stacking surface feature data is obtained by analyzing the surface area node set in the morphological topology map. These features include surface flatness, area size, and spatial distribution density of surface nodes. The calculation of stacking surface features helps the subsequent stacking modules select the optimal placement surface and evaluate the stability of the stacking structure.

[0026] Preferably, the visual recognition module includes an RGB-D sensor, an infrared sensor and an ultrasonic sensor, wherein: The RGB-D sensor is used to obtain three-dimensional morphological data of the powder bag through optical imaging and depth information, and the three-dimensional morphological data includes the length, width, height and boundary features of the powder bag; The infrared sensor is used to detect the distribution characteristics of the powder inside the powder bag, and the distribution characteristics are obtained by calculating the infrared thermal radiation signals in different temperature ranges; The ultrasonic sensor acquires the density and edge deformation characteristics of the powder bag material by emitting sound waves to the surface of the powder bag and receiving reflected signals.

[0027] The RGB-D sensor combines optical imaging (RGB) and depth perception (Depth) functions to obtain the three-dimensional morphological data of the powder bag. Specifically, optical imaging is used to capture the color and surface texture information of the powder bag, and depth perception uses structured light or time-of-flight technology to measure the distance from each point on the surface of the powder bag to the sensor to generate three-dimensional point cloud data.

[0028] The 3D morphological data includes the length, width, height and boundary features of the powder bag. The length, width and height are calculated by the maximum and minimum axial values ​​of the point cloud, and the boundary features are extracted by analyzing the distribution and connection relationship of the boundary points in the point cloud. The 3D morphological data provides basic geometric information for the extraction of grasping nodes and stacking surface features. By analyzing the point cloud data, the center of gravity of the powder bag, the flatness of the edge points, and the area suitable for grasping and stacking can be identified.

[0029] The infrared sensor uses the difference in infrared thermal radiation to detect the distribution characteristics of the powder inside the powder bag. When the infrared beam passes through the powder bag, the distribution of powder in different temperature zones causes the reflected and transmitted radiation signals to change. By processing these signals, it is possible to identify whether the powder inside the powder bag is evenly deposited.

[0030] The detection of temperature zone signals is based on the relationship between the intensity of infrared radiation and temperature. The intensity difference in the reflected signal is calculated using the following formula to calculate the distribution characteristics:

[0031] in, It represents the reflection coefficient, reflecting the distribution characteristics of the powder; Indicates the detected reflected signal strength; Indicates the intensity of the incident signal. The change in reflection coefficient can be used to determine the packing density and internal voids of the powder. The internal distribution characteristic data can guide the flexible manipulator to avoid unevenly stacked areas when grasping, and help optimize the force distribution of powder bags when stacking.

[0032] Ultrasonic sensors measure material density and edge deformation characteristics by emitting sound waves to the surface of the powder bag and receiving its reflected signals. When sound waves propagate in media of different densities, their speed and attenuation characteristics change, and these changes can be used to infer the density of the powder bag material. In addition, the shape of the sound wave reflection signal can be used to analyze the degree of deformation of the powder bag edge. Density calculation formula:

[0033] in, It indicates the speed of sound wave propagation in the medium; Indicates the elastic modulus of the material; Indicates the density of a material. By measuring the speed of sound waves , combined with the elastic modulus of the material, the material density can be calculated The material density and edge deformation characteristics provide additional support for gripping and stacking, for example ensuring that the gripping force is appropriate to the material strength of the powder bag and assessing the impact of edge deformation on stacking stability.

[0034] The visual recognition module integrates the different sensing capabilities of RGB-D sensors, infrared sensors and ultrasonic sensors to achieve comprehensive data collection of powder bags from geometric structure to internal characteristics to surface properties. Compared with a single sensor system, this multimodal sensing method can provide more accurate and richer information, effectively reducing noise and data deviation. The three-dimensional point cloud data obtained by the RGB-D sensor can accurately describe the spatial form of the powder bag, the internal powder distribution detection of the infrared sensor makes up for the limitation that the visual sensor cannot penetrate the material, and the ultrasonic sensor provides high-precision measurement of surface density and deformation. The combination of the three enables the generated data to cover the external, internal and edge characteristics of the powder bag, comprehensively improving the recognition accuracy. The data generated by the visual recognition module can not only guide the flexible manipulator to select the best grasping point and ensure grasping stability, but also assist the transfer module to optimize path planning and provide key stacking surface feature data for the stacking module. This multi-level data support significantly improves the efficiency and safety of the entire loading process.

[0035] Example 1: The 3D morphological data of a powder bag is collected by RGB-D sensor, and its length, width and height are calculated to be 60 cm, 40 cm and 20 cm. Point cloud analysis shows that the boundary points of the powder bag are evenly distributed and regular in shape, and the center of gravity is located at (30, 20, 10) (3D coordinates). This data is used to extract the grasping nodes of the flexible manipulator, and a grasping point close to the center of gravity and with a flat surface is selected to ensure grasping stability.

[0036] Example 2: Scan the powder bag with an infrared sensor and detect the reflection coefficient of a certain area It is significantly lower than other areas, indicating that the powder accumulation in this area is sparse. Combined with the grasping path planning, the flexible manipulator avoids this area to grasp, preventing deformation caused by uneven powder accumulation during grasping.

[0037] Example 3: The surface density of a powder bag is measured by an ultrasonic sensor, and the material density is calculated to be At the same time, the ultrasonic signal shows that there is a slight deformation on the edge of the powder bag. Combined with the stacking surface feature extraction algorithm, the side of the powder bag with the smallest deformation is selected as the stacking surface to ensure stacking stability.

[0038] Preferably, the visual recognition module includes a data fusion unit, which jointly processes the three-dimensional morphological data, distribution characteristic data and material characteristic data, and generates morphological topological map data through a weighted average algorithm, wherein the morphological topological map data includes the geometric positions and boundary connection information of key nodes of the powder bag; the morphological topological map data is used to extract grasping node data, and the grasping node data is calculated by a geometric graph analysis algorithm.

[0039] Multimodal data refers to three-dimensional morphological data, distribution characteristic data, and material characteristic data obtained by different sensors. Since various sensors have different focuses on perceiving target characteristics, the collected data have their own advantages and limitations. For example, RGB-D sensors can provide the geometric structure of powder bags, but cannot reflect the distribution state of the powder inside them; infrared sensors can detect internal distribution, but have weak information on boundary characteristics; ultrasonic sensors focus on material characteristics and boundary deformation. In order to fully reflect the physical state of the powder bag, these data must be jointly processed through a data fusion unit.

[0040] The data fusion unit integrates different data sources through the weighted average algorithm. The core idea of ​​the weighted average algorithm is to assign different weights to each data source according to its reliability or contribution to the target characteristics, and perform weighted processing on the characteristic values ​​of the same target. The specific formula is:

[0041] in, Represents the fused feature data; Represents data from different sensors; Represents the weight of the corresponding data, satisfying ; Indicates the number of sensors. By setting appropriate weights, the contribution of certain key data can be prioritized. For example, 3D morphological data that is important for grasping operations may have a higher weight.

[0042] After completing the data fusion, the system generates morphological topology data. The morphological topology is a high-dimensional modeling of the geometric characteristics and boundary connection relationship of the powder bag. Its structure consists of nodes and edges: nodes represent important geometric points of the powder bag, such as the center of gravity, boundary feature points or surface key points. Edges represent the connection relationship between nodes and are used to describe the contour, boundary continuity and local structure of the powder bag surface.

[0043] Node Collection and edge set The generation of point cloud is based on the analysis of 3D shape data and material property data. The complete topological structure is formed by calculating the spatial distance and surface angle change of adjacent points in the point cloud.

[0044] The extraction of grasping node data relies on the geometric graph analysis algorithm, which determines the grasping points suitable for flexible manipulator operation by analyzing the node characteristics (such as stability, smoothness, neighborhood distribution, etc.) in the morphological topology graph.

[0045] Stability: The grabbing node needs to be close to the center of gravity of the powder bag to ensure balanced force during grabbing.

[0046] Smoothness: The surface of the gripping point needs to be flat to prevent the powder bag from sliding during the gripping process.

[0047] Neighborhood distribution: The neighborhood distribution density of the grasping node needs to be moderate to ensure the structural stability of the grasping area.

[0048] Calculation formula: The crawl node score is calculated using the following formula:

[0049] in, Indicates the crawl node score; represents the node geometric stability score; represents the node surface smoothness score; Represents the neighborhood node distribution density score; Represents the weight parameter, which is set according to the requirements of the crawling task.

[0050] The data fusion unit effectively integrates the multimodal data of different sensors through a weighted average algorithm, so that the fused morphological topology map can fully reflect the geometric structure, internal distribution and surface characteristics of the powder bag. This efficient data integration method significantly improves the quality and reliability of the data, providing a stable foundation for subsequent operations. Through topological modeling, the system can clearly define the key geometric features and boundary connection information of the powder bag, eliminating the data deviation or blind spot problems that may be caused by a single sensor. The accuracy of the morphological topology map directly affects the extraction accuracy of the grasping nodes and stacking surface features. Based on the geometric graph analysis algorithm, the extraction of the grasping nodes fully considers the center of gravity position, surface smoothness and neighborhood distribution characteristics of the powder bag, ensuring that the grasping action of the flexible manipulator is highly adaptable and stable, significantly reducing the risk of grasping failure or powder bag sliding.

[0051] Example 1: A powder bag collects the following data through RGB-D sensor, infrared sensor and ultrasonic sensor: 3D morphological data: geometric structure represented by point cloud, with the center of gravity initially calculated as (30, 20, 15); Distribution characteristic data: The internal distribution uniformity score is 0.85 (out of 1.0); Material characteristic data: Density measurement is 1.2g / cm 3 , edge deformation detection is 0.2.

[0052] The data fusion unit is based on the weight The data is fused to generate morphological topology data, showing the geometric position of key nodes and boundary connection information. The connection relationship of boundary points is calculated by the distance threshold (within 5mm) of adjacent points in the point cloud.

[0053] Example 2: Based on the morphological topology data generated in Example 1, the scoring results of the geometric graph analysis algorithm show that the captured nodes are located at (28, 20, 14), and the score is The node is close to the center of gravity of the powder bag, has a smooth surface and a moderate neighborhood distribution density, which is suitable for the flexible manipulator to perform grasping operations.

[0054] The flexible grasping module includes a flexible manipulator, which generates grasping path data according to the grasping node data and performs a grasping operation on the powder bag; during the grasping process, the flexible manipulator dynamically adjusts the grasping action by real-time monitoring the grasping force and the change of the powder bag shape, and generates optimized powder bag shape data; The grasping action of the flexible grasping module is performed by the flexible manipulator. The flexible manipulator generates grasping path data based on the grasping node data obtained from the visual recognition module. The generation of grasping path data depends on the inverse kinematics algorithm, the core of which is to ensure that the end of the manipulator can reach the target grasping node by calculating the spatial motion trajectory of each joint of the flexible manipulator.

[0055] The grasping path data includes the three-dimensional coordinates of the grasping point, the posture angle of the end of the flexible manipulator, the clamping force and other information. Through the dynamic adjustment of these parameters, the flexible manipulator can accurately align the grasping node and perform the action. After the path data is generated, it is necessary to optimize it in combination with the on-site environment (such as the angle of the powder bag stacking, the distance between the manipulator and the powder bag, etc.) to reduce the manipulator's movement time, avoid collision risks, and ensure grasping efficiency.

[0056] During the grasping process, the flexible manipulator uses built-in torque sensors and shape sensors to achieve real-time monitoring and dynamic adjustment: The torque sensor detects the gripping force applied by the flexible manipulator during the gripping process and compares it with the target gripping force in real time. If the gripping force is too large, it may damage the powder bag; if the gripping force is insufficient, it may cause unstable gripping. Based on the real-time monitoring results, the manipulator can automatically adjust the gripping force to ensure the stability and safety of the gripping action.

[0057] The shape sensor updates the shape data of the powder bag in real time by capturing the deformation of the powder bag caused by the force during the grasping process. If the shape change of the powder bag caused by internal unevenness is detected, the system can automatically adjust the clamping position and angle of the grasping point to ensure that the grasping action is always carried out within the range allowed by the shape.

[0058] After the grasping is completed, the flexible grasping module optimizes the final powder bag morphology data and generates optimized powder bag morphology data. The optimized powder bag morphology data includes information such as the posture, surface flatness, and center of gravity of the powder bag. These data are provided by the torque sensor and morphology sensor of the flexible manipulator and are calculated and generated by the morphology optimization algorithm inside the system. The optimized powder bag morphology data provides a stable powder bag state input for the transfer module, ensuring that the powder bag will not slip or fall due to improper grasping during the transfer process, and at the same time provides a higher morphology adaptability for the stacking module.

[0059] Preferably, the flexible manipulator includes a multi-degree-of-freedom joint structure, and the joint structure realizes multi-axis rotation and extension through servo drive; pressure sensors and contact sensors are set in the fingertip area of ​​the flexible manipulator, and the pressure sensors are used to monitor the distribution of applied force during the grasping process in real time, and the contact sensors are used to detect the contact area and morphological changes of the grasping point, so as to dynamically adjust the grasping angle and strength.

[0060] The multi-degree-of-freedom joint structure of the flexible manipulator achieves multi-axis rotation and extension through servo drive, giving it flexible movement capabilities and the ability to adapt to powder bags of different sizes, shapes and positions.

[0061] The joint structure usually includes three rotating joints and one telescopic joint, which are used to control the three degrees of freedom (position adjustment) and gripping action (end control) of the end of the manipulator. Each joint is driven by an independent servo motor, which can accurately control the angle and speed according to the input command. During the grasping process, the manipulator can quickly move the end gripper to the target grasping point through the linkage of multi-degree-of-freedom joints, and adjust the grasping posture to meet the requirements of different powder bag positions and postures.

[0062] The fingertip area of ​​the flexible manipulator is equipped with pressure sensors to detect the force distribution applied to the powder bag during the grasping process and dynamically adjust it through the feedback control system. The pressure sensor senses the grasping force applied to the surface of the powder bag through piezoelectric materials or resistance changes, and transmits the data to the system in real time. The system compares the grasping force with the target value, calculates the difference and dynamically adjusts the output torque of the servo motor to ensure that the grasping force is stable within a safe range. Real-time grasping force monitoring can effectively prevent damage to the powder bag due to excessive force, and avoid grasping failure due to insufficient force.

[0063] The contact sensor is used to monitor the contact area and morphological changes of the grasping point to evaluate whether the grasping point is in full contact with the surface of the powder bag and detect the morphological changes of the powder bag during the grasping process. The size of the contact area between the fingertip and the powder bag is calculated by sensing the capacitance change or pressure distribution in the contact sensor. If the contact area is too small, it means that the grasping point may deviate from the expected position. The system will adjust the grasping angle and force to increase the contact area and ensure stable clamping. When the powder bag is deformed due to grasping force or center of gravity shift, the contact sensor will capture these changes and provide real-time feedback to the control system. The system combines pressure data and morphological change data to adjust the grasping strategy, such as changing the grasping point position or force to adapt to the deformation characteristics of the powder bag.

[0064] The flexible manipulator dynamically adjusts the gripping angle and force through real-time data provided by pressure sensors and contact sensors. When the contact sensor detects that the contact area is too small or the shape change exceeds the threshold, the system calculates the normal direction of the gripping point surface and instructs the servo motor to adjust the joint angle so that the gripper maintains vertical contact with the powder bag surface to maximize contact stability. By comparing the difference between the current gripping force and the target gripping force, the torque adjustment required by the servo motor is calculated, thereby dynamically changing the gripping force to ensure that the powder bag is not damaged during the gripping process and can be firmly clamped.

[0065] Through the multi-degree-of-freedom joint structure, the flexible manipulator can quickly adjust the position and posture of the gripper to adapt to powder bags of different shapes, sizes and positions. The high-precision control of the servo drive system ensures that the manipulator's movements respond quickly and accurately, effectively improving the grasping efficiency and adaptability. The real-time grasping force monitoring of the pressure sensor and the contact area detection of the contact sensor enable the grasping action to be dynamically adjusted according to the actual state of the powder bag, ensuring that the grasping force and contact area are always within the optimal range, avoiding grasping failure or destruction of the powder bag's morphology. The morphological change data provided by the contact sensor allows the flexible manipulator to adjust the grasping strategy according to the deformation of the powder bag, further improving the adaptability of the grasping action to powder bags with complex shapes. This dynamic adjustment mechanism greatly enhances the grasping module's ability to handle irregular powder bags.

[0066] Example 1: In the loading task, the grabbing node coordinates of a powder bag are (25, 15, 10), and its surface material is relatively soft. After the flexible manipulator reaches the target grabbing point through grabbing path planning, it begins to apply clamping force. The pressure sensor detects that the grabbing force is 12N, while the system target grabbing force is 10N. The control system adjusts the output torque of the servo motor in real time according to the difference between the grabbing force and the target value, reducing the grabbing force to 10N, ensuring stable grabbing and no damage to the powder bag.

[0067] Example 2: During the grasping process of Example 1, the contact sensor detected that the contact area between the fingertips of the flexible manipulator and the powder bag only covered 70% of the expected value, resulting in a possible slip risk at the grasping point. By analyzing the pressure distribution data of the contact sensor, the system calculated that the adjustment angle was 5 degrees. The servo motor adjusted the joint angle so that the gripper was in vertical contact with the surface of the powder bag while maintaining uniform pressure distribution. The final contact area reached the expected value and the grasping stability was guaranteed.

[0068] Example 3: During the grasping process, the contact sensor detected that the powder bag was deformed due to the uneven distribution of powder inside, resulting in a reduction in contact area and a decrease in clamping stability. The control system adjusted the grasping point position based on the morphological change data, fine-tuning the grasping point from (25, 15, 10) to (24, 16, 11), and readjusted the clamping force to match the new contact morphology, successfully completing the grasping task.

[0069] Preferably, Figure 3 As shown, the flexible manipulator further includes a path planning unit, which generates grasping path data based on grasping node data, and the grasping path data generates a motion trajectory of the flexible manipulator through an inverse kinematics algorithm; the path planning unit dynamically optimizes the motion trajectory according to real-time feedback data, and the optimization objectives include minimizing the grasping time, uniformity of force on the powder bag, and morphological stability.

[0070] The path planning unit uses the grasping node data provided by the visual recognition module as input and generates grasping path data using the inverse kinematics algorithm. The grasping path data describes the motion trajectory of the flexible manipulator from the initial position to the target grasping point, including the angle change, speed and final posture of each joint.

[0071] The inverse kinematics algorithm takes the target position and posture as input and calculates the specific parameters of each joint of the manipulator so that the end of the manipulator can accurately reach the target grasping point. For example, when the grasping node data provides the three-dimensional coordinates of the target grasping point ( ) and attitude angle( ), the inverse kinematics algorithm generates the rotation angles and extension distances of all joints by solving the geometric structure equations of the flexible manipulator.

[0072] The grasping path data includes the spatial coordinates of the grasping node, the joint angle sequence of the flexible manipulator, the clamping force, the movement speed and other specific parameters, providing a complete execution plan for the flexible manipulator.

[0073] During the grasping operation, the path planning unit dynamically optimizes the motion trajectory based on real-time feedback data to ensure that the robot can adapt to changes in complex environments.

[0074] Real-time feedback data includes the current position, speed, gripping force of the robot, and the shape change of the powder bag. These data are provided by the torque sensor, shape sensor and position encoder built into the robot, and transmitted to the path planning unit in real time for processing. The optimization of the motion trajectory is centered on three goals: Minimize the grasping time: By optimizing the joint path of the manipulator, unnecessary movement time is reduced and loading efficiency is improved. Uniform force of powder bags: Adjust the grasping path according to the shape and force distribution of the powder bags to ensure that the powder bags will not slide or be damaged due to uneven force during the grasping process. Shape stability: Combined with the real-time shape data of the powder bags, the movement speed and clamping angle of the manipulator are optimized to prevent shape deformation caused by too fast or too slow movement during the grasping process.

[0075] The path planning unit uses an incremental path adjustment algorithm to generate corrections and update the motion trajectory by comparing the actual position of the robot with the target path in real time. Each updated correction ensures that the robot can quickly converge to the target path while maintaining a smooth grasping action.

[0076] Through the inverse kinematics algorithm, the path planning unit can generate a precise motion trajectory for the flexible manipulator, ensuring that the manipulator can accurately reach the grasping node position and complete the grasping action in the best posture. The accuracy of the motion trajectory directly affects the success rate of grasping and the stability of the powder bag. By dynamically optimizing the motion trajectory in combination with real-time feedback data, the path planning unit can adapt to changes in the grasping environment in real time, including changes in the shape of the powder bag and the motion errors of the manipulator. The optimized trajectory not only improves the stability of the grasping action, but also significantly reduces the grasping failure rate caused by morphological instability or motion errors. The path planning unit aims to minimize the grasping time, and significantly improves the overall efficiency of the grasping operation by reducing the motion redundancy and unnecessary movements of the manipulator. This time optimization is of great significance in high-frequency loading tasks.

[0077] Example 1: In a loading task, the visual recognition module provides the grasping node data, the three-dimensional coordinates of the target grasping point are (50, 30, 20), and the posture angle is (90°, 0°, 0°). The path planning unit uses the inverse kinematics algorithm to calculate the joint parameters of the flexible manipulator, where the shoulder joint rotation angle is 45°, the elbow joint rotation angle is 30°, the clamping force of the end clamp is 8N, and the grasping path data also includes the manipulator's moving speed of 0.5 m / s. The flexible manipulator reaches the grasping point according to the generated path data and completes the grasping action.

[0078] Example 2: During the grasping process of Example 1, the torque sensor detected that the powder bag changed its shape due to the force, causing the grasping node position to shift by 5 mm. At the same time, the shape sensor detected that the boundary area of ​​the powder bag was locally deformed, affecting the grasping stability. The path planning unit combines these real-time feedback data to update the grasping path, adjust the target position of the end gripper to (50.5, 30.2, 20.1), and reduce the movement speed of the manipulator to 0.4 m / s to ensure the stability of the grasping action. The optimized motion trajectory enables the grasping operation to be completed smoothly, and the shape of the powder bag remains stable.

[0079] Example 3: In another loading task, the path planning unit optimizes the grasping time to 2 seconds in combination with the loading efficiency requirements. At the same time, the joint path and grasping force are adjusted in real time during the grasping action to ensure uniform force distribution of the powder bag. Real-time feedback data shows that the optimized grasping force distribution uniformity score is 0.95 (out of 1.0), the morphological stability score is 0.90, and the grasping time is reduced from the original 2.8 seconds to 2 seconds, which greatly improves the loading efficiency.

[0080] The transfer module includes an automatic guided robot, which completes the transfer of the powder bag from the grabbing point to the stacking point according to the optimized powder bag shape data and transfer path data, and outputs transfer status data, wherein the transfer status data includes posture change information of the powder bag, real-time center of gravity adjustment parameters and a time record of reaching the target stacking point, which is used to guide the stacking module to accurately adjust the position and angle of the powder bag; The automatic guided robot (AGV) performs the transfer operation according to the optimized powder bag shape data and transfer path data. The transfer path data is generated by the path planning module, which includes the robot's motion trajectory from the grasping point to the stacking point, steering angle, moving speed, etc. The optimized path takes into account factors such as the distribution of obstacles on the site, the shortest path, and the minimization of energy consumption. The AGV uses a laser radar (LIDAR) and ultra-wideband (UWB) positioning system for joint positioning, real-time perception of environmental changes and execution of path trajectories. Through the feedback control system, the AGV can automatically correct the deviation during driving to ensure the accuracy of path execution.

[0081] During the transfer process, the AGV monitors the posture changes of the powder bag in real time to ensure that it remains stable during movement. The posture change data is obtained by the visual sensor and acceleration sensor installed on the top of the AGV. The visual sensor is used to capture the tilt angle and position changes of the powder bag, and the acceleration sensor detects the impact of the inertial force caused by acceleration or deceleration on the powder bag. When the posture change exceeds the set threshold, the AGV makes corrections by adjusting the speed and driving trajectory, such as slowing down or changing the direction of the tilt angle, to restore the stability of the powder bag.

[0082] During the transportation process, the center of gravity of the powder bag may shift due to irregular shape or path turning. The center of gravity adjustment mechanism calculates the center of gravity position of the powder bag in real time and dynamically adjusts the inclination angle of the AGV platform to achieve the redistribution of the center of gravity. The current center of gravity position is calculated based on the powder bag shape data and posture change data. After comparing it with the ideal center of gravity position, the AGV power module is instructed to adjust the inclination angle of the platform to ensure that the center of gravity remains within a safe range. For example, when it is detected that the center of gravity of the powder bag shifts to the edge of the platform, the AGV will control the platform to tilt in the opposite direction at a certain angle through the servo motor to readjust the center of gravity to the center of the platform.

[0083] The transfer status data consists of the powder bag posture change information, real-time center of gravity adjustment parameters and the time record of reaching the target stacking point. The specific content includes: the tilt angle and horizontal offset of the powder bag during the transfer process, which is used to evaluate the stability of the transfer. The center of gravity offset of the powder bag and the corresponding adjustment action are recorded to provide the stacking module with the current balance state of the powder bag. The time when the powder bag reaches the target stacking point is recorded for the synchronous operation of the stacking module.

[0084] By combining LiDAR and UWB positioning systems, AGV can accurately execute transfer path data in complex site environments, avoid collisions and deviations, and ensure that powder bags can be safely transported from the grabbing point to the stacking point in the shortest time and with the lowest energy consumption. Through posture change monitoring and dynamic center of gravity adjustment, AGV can correct the unstable state of powder bags in real time during transportation, such as tilting or sliding due to irregular shapes. This real-time adjustment mechanism effectively reduces the risk of damage to powder bags during transportation and improves the stability of transportation. The generated transfer status data provides comprehensive and accurate input information for the stacking module, ensuring that the stacking module can adjust the stacking position and angle according to the current posture and center of gravity of the powder bags, further improving the overall loading efficiency of the entire system.

[0085] Example 1: In a loading task, the AGV receives optimized path data from the grabbing point to the stacking point. The total length of the path is 10 meters, including two 90° turns. The AGV drives along the path through the lidar and UWB positioning system. During the transfer process, due to unexpected obstacles on the site, the AGV adjusts the steering angle through the real-time path correction function, bypasses the obstacles and returns to the original path to ensure that the powder bags reach the stacking point smoothly.

[0086] Example 2: During the transportation process of Example 1, the acceleration sensor detected that the powder bag tilted forward due to sudden deceleration, and the tilt angle was 8°, exceeding the safety threshold of 5°. The system instructs the AGV to decelerate to 0.2 m / s in real time, and adjusts the platform tilt angle to -3° through the servo motor, readjusts the center of gravity of the powder bag back to the center of the platform, and restores the stable state.

[0087] Example 3: After Example 2 is completed, the transfer module generates transfer status data, including the change record of the powder bag tilt angle (initial tilt 8°, final tilt 1°), the center of gravity adjustment parameter (platform tilt -3°) and the time record of reaching the stacking point (12 seconds). The stacking module calculates the stacking angle adjustment amount of the powder bag based on these data to ensure that the powder bag can be stably stacked at the target stacking position.

[0088] Preferably, the automatic guided robot of the transfer module includes a positioning system and an inclination sensor, wherein the positioning system generates a transfer path based on the optimized morphological data of the powder bag and the target stacking point position; the inclination sensor monitors the inclination angle of the powder bag in real time, calculates the adjustment parameters in combination with the center of gravity offset information of the powder bag, and realizes the transportation of the powder bag by adjusting the inclination angle of the automatic guided robot platform.

[0089] The positioning system is the core unit of the automatic guided robot, responsible for generating the optimal transfer path based on the optimized powder bag shape data and the target stacking point position. Input data, powder bag shape data (including center of gravity position, size, shape characteristics) and the three-dimensional coordinates of the target stacking point. Path generation is based on environmental perception and dynamic planning algorithms, combined with laser radar (LIDAR) and ultra-wideband (UWB) positioning systems for real-time environmental modeling, to generate a transfer path that avoids obstacles, has the shortest path length and the lowest energy consumption. Output path, the transfer path is represented in the form of the robot's motion trajectory, including each position point on the motion path and the corresponding speed and steering angle. The path update frequency can be dynamically adjusted according to real-time environmental changes.

[0090] During the transportation process, the powder bag may change its posture or shift its center of gravity due to dynamic factors such as acceleration, deceleration, and turning. The inclination sensor detects the inclination angle of the powder bag in real time, and dynamically adjusts the platform inclination angle based on the center of gravity offset information collected by the sensor. The inclination sensor is installed on the robot platform to measure the inclination angle of the powder bag relative to the platform. If the inclination angle exceeds the set threshold (for example, ±5°), the system will trigger the center of gravity adjustment mechanism. By combining the powder bag morphology data and the real-time data of the inclination sensor, the offset between the current center of gravity position of the powder bag and the center of the platform is calculated, and adjustment parameters are generated to correct the inclination of the transport platform. The automatic guided robot adjusts the platform inclination angle through the servo motor to redistribute the center of gravity of the powder bag to the safe area of ​​the platform to avoid tipping or sliding.

[0091] The positioning system and the inclination sensor work together to ensure the safety and efficiency of the powder bags during transportation. The positioning system provides precise path execution, while the inclination sensor monitors posture changes in real time and dynamically adjusts the platform angle, ultimately ensuring that the powder bags can reach the target stacking point in a stable state. When turning or accelerating or decelerating, the feedback from the inclination sensor is combined with the speed and direction of the path planning to automatically adjust the transportation parameters. For example, when making a sharp turn, the system will reduce speed and adjust the inclination angle appropriately to ensure that the powder bags remain stable.

[0092] The transfer path generated by the positioning system can dynamically avoid obstacles and optimize energy consumption, ensuring that the robot can accurately complete the transportation task in a complex environment. The high efficiency of path planning not only improves the transfer speed, but also reduces unnecessary paths and energy waste. The inclination sensor monitors the posture changes and center of gravity offset of the powder bag in real time. By dynamically adjusting the platform inclination angle, it effectively solves the problems of tilting and sliding caused by inertia during transportation, ensuring the safety and stability of the powder bag during transportation. The generated inclination angle, center of gravity adjustment parameters and target position arrival information provide accurate input data for the subsequent stacking module. These data can help the stacking module further optimize the placement and angle of the powder bag, improve the stacking accuracy and overall stability.

[0093] Example 1: In a loading task, the automatic guided robot receives the three-dimensional coordinate data from the grasping point to the target stacking point. The path length is 8 meters, including two 45° turns. The positioning system uses the laser radar to scan the environment, recognizes that there is an obstacle on the path, and dynamically plans an obstacle avoidance path, increasing the path length to 8.5 meters and adjusting the speed to 0.4 m / s. The automatic guided robot drives along the planned path and accurately reaches the target stacking point to complete the transportation.

[0094] Example 2: During the transportation process of Example 1, the inclination sensor detected that the powder bag was tilted due to the inertial force during the acceleration phase, and the inclination angle reached 6°, exceeding the set threshold of 5°. The system calculated the center of gravity offset in real time as 10 cm, and adjusted the inclination angle of the transport platform to -3° through the servo motor, and finally restored the inclination angle of the powder bag to within 1°, and the center of gravity returned to the center of the platform.

[0095] Example 3: After Example 2 is completed, the last section of the transfer path includes a 90° turn. The robot automatically reduces its speed to 0.3 m / s during the turn. At the same time, the tilt sensor detects the risk of the powder bag rolling due to the turn. The platform makes a slight adjustment again, correcting the roll angle from 3° to 0°. The powder bag successfully reaches the target stacking point within 12 seconds, and generates transfer status data including tilt angle change, center of gravity adjustment parameters and arrival time for use by the stacking module.

[0096] Preferably, the transfer status data generated by the transfer module includes real-time posture information of the powder bag, center of gravity adjustment parameters and target position arrival time of the powder bag; the center of gravity adjustment parameters are calculated based on the difference between the dynamic position of the center of gravity of the powder bag and the center of gravity position of the robot platform, and are used to guide the automatic guided robot to adjust its motion posture in real time.

[0097] The real-time posture information of the powder bag refers to its position and angle changes relative to the platform during transportation, including parameters such as tilt angle, horizontal offset and rotational posture. The automatic guided robot obtains real-time posture information through the visual sensor, inclination sensor and acceleration sensor installed on the platform. Among them, the visual sensor captures the three-dimensional spatial position of the powder bag, the inclination sensor measures its inclination angle relative to the horizontal plane, and the acceleration sensor detects the change of inertial force during movement. The system compares the posture data collected in real time with the ideal transportation state of the powder bag, calculates the posture offset, and uses it as a reference input for subsequent adjustment of the motion posture. For example, when the tilt angle deviates from the target value by more than the set threshold (such as ±5°), the system triggers the center of gravity adjustment mechanism.

[0098] The center of gravity adjustment parameters are used to describe the difference between the dynamic center of gravity of the powder bag and the ideal center of gravity of the automatic guided robot platform, and guide the robot to adjust its motion posture to maintain stability. The dynamic center of gravity position of the powder bag is calculated by combining the morphological data generated by the visual recognition module with the real-time data of the sensor. For example, the three-dimensional coordinate position of its dynamic center of gravity is determined based on the size, mass distribution and current tilt angle of the powder bag. The center of gravity adjustment parameters are generated by calculating the offset between the dynamic center of gravity of the powder bag and the center position of the robot platform:

[0099] in, Indicates the center of gravity adjustment parameter, which indicates the difference between the current center of gravity and the ideal center of gravity; Indicates the current dynamic center of gravity coordinates of the powder bag; Indicates the ideal center of gravity coordinates of the robot platform. The adjustment action means that when the center of gravity adjustment parameter exceeds the safety threshold, the robot adjusts the platform tilt angle or speed through the servo motor to dynamically restore the balance position of the center of gravity.

[0100] The target location arrival time refers to the total time it takes for the powder bag to be transported from the grabbing point to the stacking point. It is used to evaluate the transportation efficiency and guide the synchronous operation of the stacking module. The estimated arrival time at the target location is calculated by the path data generated by the path planning module and the real-time speed of the robot, and the actual time is recorded after the actual arrival. The recorded arrival time can be used as a basis for evaluating the transportation efficiency, and at the same time help the stacking module dynamically adjust the stacking sequence according to the actual arrival time of the powder bag to ensure the synchronization of subsequent operations.

[0101] By acquiring and analyzing real-time posture information, the transfer module can fully grasp the dynamic state of the powder bags during transportation, including changes such as tilt, offset and rotation. This precise posture monitoring provides sufficient data support for subsequent dynamic adjustments, effectively reducing the risk of powder bags slipping or tipping during transportation. Real-time calculation and feedback of the center of gravity adjustment parameters guide the automatic guided robot to respond quickly to posture changes. By adjusting the platform inclination angle or speed, the center of gravity adjustment mechanism can restore the balance state of the powder bags in the shortest time, improving the stability and safety of the transportation process. The posture information, center of gravity adjustment parameters and arrival time records in the transfer status data provide comprehensive input support for the stacking module. These data can guide the stacking module to adjust the placement angle and position of the powder bags, optimize the stacking effect, and improve the efficiency of the overall loading process.

[0102] Example 1: In a loading task, the automatic guided robot detected through the inclination sensor that the powder bag was tilted by 7° due to the turning inertia force, which exceeded the set safety threshold of 5°. The real-time posture information showed that the horizontal offset of the powder bag was 3 cm. The system triggered the dynamic adjustment mechanism based on these data, adjusted the tilt angle of the robot platform to -2° through the servo motor, and reduced the speed to 0.3 m / s, and finally restored the tilt angle to within 1°, and the center of gravity of the powder bag returned to the center of the platform.

[0103] Example 2: In the above example, the dynamic center of gravity of the powder bag is calculated to be (25, 18, 10), while the ideal center of gravity of the platform is (25, 20, 10). Center of gravity adjustment parameters The calculation result is (0, -2, 0), indicating that the center of gravity of the powder bag is offset by 2 cm. The system instructs the servo motor to tilt the platform 1.5° in the positive Y direction according to the adjustment parameters, and finally adjusts the center of gravity offset to 0, restoring the balance of the powder bag.

[0104] Example 3: During the above transportation process, the robot completes the transportation according to the instructions of the path planning module. The total length of the path is 10 meters, including two 90° turns. The system records that the time it takes for the powder bag to reach the target stacking point is 12 seconds, and the transfer status data includes the final posture information (tilt angle 0°) and the center of gravity adjustment parameters (no offset). The stacking module uses this data to adjust the placement angle of the powder bag to ensure the stability and stability of the stacking.

[0105] The stacking module comprises a stacking device, wherein the stacking device places the powder bag at a target stacking position according to the transport state data and the stacking surface characteristic data, and generates stacking state data for evaluating the stability of the stacking structure.

[0106] like Figure 4 As shown, the stacking module receives the transfer status data and the stacking surface feature data as the basis for the placement decision. The transfer status data provides the real-time posture information, center of gravity adjustment parameters and target position arrival time of the powder bag. The posture information is used to determine whether the current angle and position of the powder bag meet the stacking requirements; the center of gravity adjustment parameters reflect whether the powder bag is in a balanced state. The stacking surface feature data describes the geometric shape of the target stacking position, including flatness, surface node distribution density and spatial size. These data are used to select the stacking position that best suits the current powder bag. Stacking position calculation, the stacking device combines the input data, calculates the best matching point between the powder bag and the stacking surface through a geometric fitting algorithm, and generates placement coordinates and angle adjustment parameters. The placement coordinates ensure the accurate position of the powder bag in the stacking layer, and the angle adjustment parameters are used to correct the posture of the powder bag to match the flatness of the stacking surface.

[0107] The stacking device completes the powder bag placement operation through a stacking robot arm driven by a servo motor. The stacking robot arm adjusts the parameters according to the calculated placement coordinates and angles, and accurately controls the movement trajectory of the robot arm through the servo motor to move the powder bag to the target position. During the placement process, the stacking device monitors the contact status between the powder bag and the stacking surface in real time, and ensures that the powder bag is in full contact with the stacking surface and the force is evenly applied by fine-tuning the end posture of the robot arm. After the powder bag is safely placed in the target stacking position, the system will re-collect the status data of the stacking layer to evaluate the stacking effect.

[0108] After the stacking module is placed, it generates stacking status data to evaluate the stability of the stacking layer. The stacking status data includes the final position coordinates of the powder bag, the force distribution of the stacking layer, the flatness score, etc. The force distribution data is collected by the pressure sensor installed on the stacking surface, and the flatness score is calculated based on the deviation between the overall shape of the stacking layer and the design model. The stacking status data serves as the basis for the subsequent powder bag stacking operation to ensure the stability and uniformity of the entire stacking structure and avoid stacking instability caused by uneven stacking or concentrated force.

[0109] Preferably, the stacking device of the stacking module includes a stacking robot arm driven by a servo motor, and the stacking robot arm accurately locates the position of the powder bag by receiving the stacking surface feature data in real time. The stacking robot arm driven by the servo motor adjusts the movement speed and placement angle of the robot arm according to the coordinates of the target stacking point; the adjustment of the placement angle is obtained based on the calculation of the stacking contact point distribution.

[0110] The stacking surface feature data and the coordinates of the target stacking point are the key inputs for the stacking robot to perform the placement operation. The stacking surface feature data describes the geometric characteristics of the target stacking position, including flatness, surface node distribution density, and available space. These data are generated in real time through the visual recognition module and sensor network, providing the stacking robot with a detailed description of the stacking surface. The stacking robot combines the stacking surface feature data and the target point coordinates, and uses the servo control system to calculate the angle change of each joint and the end motion path to ensure that the end gripper can accurately reach the target position.

[0111] The servo motor is the core drive unit of the stacking robot, which achieves precise motion trajectory and end-point posture adjustment through closed-loop feedback control. The servo motor dynamically adjusts the movement speed of the robot arm according to the location of the target stacking point and real-time feedback data. For example, when approaching the target position, the speed will gradually decrease to ensure smooth placement. After receiving the command, the servo motor controls the posture of the end of the robot arm by adjusting the torque output by the motor and the joint position to make it reach the target position, and adjusts the placement angle to match the geometric characteristics of the stacking surface.

[0112] The placement angle is a key parameter that the stacking robot dynamically adjusts during the powder bag placement process, and its calculation is based on the distribution of the stacking contact points. The contact point refers to the key area where the bottom of the powder bag contacts the stacking surface. The distribution of the contact point reflects the flatness of the stacking surface and the uniformity of the force on the powder bag. The contact points are detected in real time through a sensor network and generated in combination with the powder bag morphology data. The stacking robot calculates the adjustment amount of the placement angle by analyzing the spatial distribution of the contact points. For example, when the contact points are unevenly distributed, the system will instruct the robot to tilt at a certain angle to increase the contact area and ensure uniform force distribution to prevent the powder bag from sliding or tipping over.

[0113] During the placement process, the stacking robot dynamically adjusts the motion trajectory and placement angle through a real-time feedback mechanism. The feedback data includes the joint position of the stacking robot, the real-time posture of the powder bag, and the force distribution information of the stacking surface pressure sensor. When a deviation between the actual state and the target state is detected, the servo motor corrects the joint angle and end position in real time according to the deviation to ensure the accurate placement of the powder bag.

[0114] The stacking robot combines the high-precision control of the servo motor and the characteristic data of the stacking surface to achieve precise adjustment of the position and angle of the powder bag. By dynamically adjusting the placement angle to match the geometric characteristics of the stacking surface, it ensures that the powder bag is in full contact with the stacking surface during the stacking process, thereby improving the stability of the stacking.

[0115] Based on the placement angle adjustment mechanism of the stacking contact point distribution, the stacking robot can optimize the force distribution of the powder bag during the placement process, effectively reduce the risk of sliding or tipping caused by uneven stacking, and significantly improve the overall stability of the stacking layer. The stacking robot adjusts the motion trajectory and posture through real-time feedback, and can flexibly adapt to the changes in the stacking surface and the morphological characteristics of the powder bag, realizing efficient processing of complex stacking tasks.

[0116] Example 1: In a loading task, the stacking module receives the coordinates of the target stacking point as (50, 30, 10), and the stacking surface feature data shows that its flatness score is 0.85. The stacking robot arm is driven by a servo motor and moves to the stacking point according to the target coordinates. When approaching the target position, the robot arm detects that the pressure in the left area is higher and the contact area in the right area is insufficient based on the distribution of contact points. The system instructs the robot arm to adjust the placement angle to 2° to the right, ultimately achieving uniform force and stable placement of the powder bag.

[0117] Example 2: During the process of Example 1, the stacking robot arm detected a change in the posture of the powder bag due to the center of gravity shift during the placement operation, and tilted 1° to the right. The servo motor corrected the end position in real time, adjusted the tilt angle of the robot arm to 1° to the left, balanced the posture of the powder bag, and reduced the end speed to 0.1 m / s to ensure that the placement process was completed smoothly. After placement, the stacking surface pressure sensor data showed that the contact points were evenly distributed, and the stacking stability score was 0.95.

[0118] Example 3: In a multi-layer stacking task, when the stacking robot was stacking the second layer of powder bags, it was found that the left front area of ​​the first layer of stacking surface was subjected to excessive force, resulting in an overall tilt of 1.5°. The stacking module adjusted the placement angle of the second layer of powder bags to 1° to the left based on the feedback data, and at the same time increased the contact area in the left rear area to balance the stacking structure. Ultimately, the overall stability of the multi-layer stacking was achieved.

[0119] Preferably, the stacking module includes a thermal distribution monitoring unit, which detects the force distribution on the contact surface of the stacking layer through an infrared imaging device to generate a stacking thermal distribution map; the stacking thermal distribution map is used to analyze the force uniformity of the stacking layer and generate stacking adjustment recommendation data.

[0120] The force distribution on the contact surface of the stacked layer reflects the force uniformity of the stacked structure and is an important indicator for evaluating the stability of the stack. The thermal distribution monitoring unit detects the temperature distribution on the contact surface of the stacked layer through infrared imaging equipment, and associates the temperature change with the force distribution. Areas of the contact surface with greater force will produce local temperature rise due to friction and extrusion, while areas with less force or no contact will have lower temperatures. The infrared imaging device captures the temperature distribution of different areas on the contact surface and presents the force state of the stacked layer in the form of a visual heat map. The raw temperature data generated by the infrared imaging device is processed to form a stack thermal distribution map. The high temperature areas in the map correspond to areas with concentrated force, and the low temperature areas correspond to areas with less force or no force.

[0121] The thermal distribution diagram is a visual expression of the force distribution of the stacked layer. The stability of the stacked structure is evaluated by analyzing the range, distribution and uniformity of the force area in the diagram. Force uniformity analysis: The system calculates the force uniformity score based on the temperature difference between each area in the thermal distribution diagram. The uniformity score reflects the rationality of the force distribution on the contact surface. A higher score indicates a more uniform force distribution and a more stable stacked layer; a lower score indicates that there may be a risk of stacking instability. By comparing the area and temperature gradient of the high-temperature area, the system determines whether the force is concentrated at certain specific points or areas. If the high-temperature area is small and the gradient is large, it means that there is force concentration and the stacked layer may be locally unstable; if the high-temperature area is evenly distributed, the overall force of the stacked layer is relatively balanced.

[0122] Based on the results of the thermal distribution map analysis, the system generates stacking adjustment recommendation data to guide the stacking module to optimize the placement angle and position of subsequent powder bags. The stacking adjustment recommendation data includes information such as the target placement angle, placement position offset, and priority placement area of ​​the powder bag. For example, when there is a concentrated force area in a stacking layer, the system will recommend placing the next powder bag in an area with relatively less force, or adjusting the placement angle to disperse the force. The recommendation data is calculated based on the current force distribution of the stacking layer and the target uniformity score to ensure that the adjustment action can effectively improve the force distribution.

[0123] By using infrared imaging equipment to monitor the temperature changes on the contact surface in real time, the thermal distribution monitoring unit can accurately capture the stress distribution state of the stacked layer, providing reliable data support for subsequent analysis. This detection method based on the relationship between temperature and force is not only sensitive, but also has the advantages of non-contact and real-time.

[0124] The stacking thermal distribution diagram provides an intuitive display of the uniformity of the stress on the stacked layers. Combined with the stacking adjustment recommendation data, the stacking module can optimize the stress state of the stacked layers by adjusting the placement position and angle, effectively improving the overall stability of the stacking structure.

[0125] Through continuous monitoring and analysis, the thermal distribution monitoring unit can provide real-time adjustment suggestions during the stacking process, ensuring that even in multi-layer stacking tasks, the placement strategy of the powder bags can be dynamically adjusted to avoid the cumulative effect of stacking instability.

[0126] Example 1: In a loading task, the stacking module detects the contact surface of the first layer of powder bags through the thermal distribution monitoring unit. The thermal distribution map generated by the infrared imaging device shows that there is an obvious high temperature area in the left front area (the temperature is 5°C higher than the average value), and the right rear area is a low temperature area (3°C lower than the average value). The system analysis results show that the left front area is under concentrated force, and the stacking uniformity score is 75%. Based on the thermal distribution map, the stacking module determines that the placement strategy of subsequent powder bags needs to be optimized.

[0127] Example 2: Based on Example 1, the thermal distribution monitoring unit generates stacking adjustment recommendation data, suggesting that the next powder bag be placed in the right rear area, and at the same time adjusting the placement angle to 3° to the right to balance the overall force distribution. The stacking robot performs the adjustment operation according to the recommended data and successfully places the second powder bag in the right rear area. The thermal distribution diagram shows that the overall force distribution uniformity score is improved to 90%.

[0128] Example 3: In a multi-layer stacking task, when the stacking module placed the powder bag on the third layer, the thermal distribution monitoring unit detected that the right area of ​​the second layer stacking surface was under concentrated force and partially tilted. The system generated adjustment suggestions based on the thermal distribution map, and instructed the stacking robot to place the third layer of powder bags on the left area and adjust the placement angle to 2° to the left. After the placement was completed, the overall stability score of the stacking structure increased from 82% to 96%.

[0129] Preferably, the self-correction unit in the stacking module triggers the servo motor to adjust the stacking position and angle of the powder bag by real-time detection of the stacking stability score data, and the stacking stability score data is generated by analyzing the force concentration and overall balance of each contact point in the stacking thermal distribution diagram, and generating adjusted stacking state data.

[0130] The stacking stability score data is an indicator for evaluating the dynamic stability of the stacking structure, which is calculated based on the force distribution and overall balance in the stacking thermal distribution diagram. The force concentration is a key indicator for evaluating whether the force on the contact surface of the stacking layer is uniform. By analyzing the size, position and temperature gradient of the high temperature area in the stacking thermal distribution diagram, it is judged whether there is local force concentration in the stacking layer. If the high temperature area is small and the temperature gradient is large, it means that the force is concentrated and the stacking structure may be unstable. The overall balance is an indicator for evaluating whether the center of gravity position of the stacking layer deviates from the ideal position. By calculating the force distribution and position offset of each contact point of the stacking layer, the center of gravity offset of the stacking structure is judged. If the center of gravity offset exceeds the set threshold, it means that the stacking layer is at risk of instability. The stacking stability score data is calculated based on the force concentration and overall balance. The score range is 0 to 100. The higher the score, the better the stability of the stacking layer. If the score is lower than the set threshold (such as 80 points), the stacking adjustment is triggered.

[0131] When the stacking stability score is lower than the threshold, the self-correction unit triggers the servo motor to adjust the stacking position and angle of the powder bag to restore the stability of the stacking layer. By changing the position or placement angle of the powder bag on the stacking surface, the force distribution is optimized and the center of gravity offset of the stacking layer is reduced, ultimately improving the overall balance. The position adjustment is achieved by driving the stacking robot arm with a servo motor. According to the low-temperature area (less force) in the stacking thermal distribution map, the powder bag is moved to the high-temperature area with uneven force distribution to balance the force. The angle adjustment is based on the contact point distribution information. By changing the inclination angle of the powder bag, the contact area with the stacking surface is increased, thereby optimizing the local force distribution. For example, when it is detected that the force is concentrated on one side of the powder bag, the system will instruct the robot arm to tilt the powder bag in the opposite direction at a certain angle.

[0132] After the adjustment is completed, the self-correction unit generates adjusted stacking state data to record the optimization effect of the stacking structure and provide input support for subsequent stacking operations. The adjusted stacking state data includes the final position coordinates, placement angle, force distribution update of the powder bag, and the optimized stacking stability score. By evaluating the adjustment effect, the system can dynamically optimize the next stacking operation and improve the overall stability of the stacking structure layer by layer.

[0133] By detecting the stack stability score data in real time, the self-correction unit can quickly respond to the instability of the stack layer and trigger adjustment actions to restore the balance of the stack structure. This dynamic optimization capability significantly improves the safety and long-term stability of the stack layer.

[0134] Combined with the stacking thermal distribution map and contact point distribution information, the position and angle adjustment under servo motor control is highly accurate. The adjustment action can not only balance the force distribution, but also effectively reduce the risk of center of gravity shift, thereby ensuring the neatness and stability of the stacking layer.

[0135] The self-correction unit realizes a closed-loop control process of real-time detection, dynamic adjustment and optimization feedback in the stacking operation, making the entire stacking process more intelligent. At the same time, by reducing the need for additional adjustments caused by stacking instability, the overall loading efficiency is improved.

[0136] Example 1: In a certain loading task, the self-correction unit detected through the stacking thermal distribution map generated by the thermal distribution monitoring unit that the left front area of ​​the first stacking surface was under concentrated force, and the right rear area was almost free of force. The stacking stability score was 72, which was lower than the set threshold of 80. The self-correction unit triggered the servo motor to control the stacking robot arm to move the powder bag 3 cm to the right rear area, and at the same time adjusted the placement angle to 1.5° to the right. After adjustment, the stacking thermal distribution map showed that the force distribution was uniform, and the stability score was improved to 92.

[0137] Example 2: In the multi-layer stacking task, the self-correction unit detected that the center of gravity of the second layer of stacking structure was offset by 3 cm (deviated from the central axis), and the stacking stability score was 65. According to the results of the thermal distribution map analysis, the system instructed the stacking robot to place the third layer of powder bags in the area in the opposite direction of the offset, and adjusted the placement angle to 2° to the left to balance the overall structure. After the adjustment, the stacking stability score increased to 90, and the center of gravity returned to the central axis.

[0138] Example 3: In the above example, after the stacking adjustment is completed, the self-correction unit generates adjusted stacking state data, including the final position of the powder bag (50, 30, 15), the placement angle of 1.5° to the right, the force distribution update result, and the optimized stacking stability score of 92. The adjusted stacking state data is used to guide the stacking operation of the next powder bag to ensure the layer-by-layer optimization of the entire stacking process.

[0139] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. An automatic powder loading device based on visual recognition, characterized in that: include: The visual recognition module is used to collect the morphological data, internal distribution data and surface characteristic data of the powder bag, and generate the morphological topological map data of the powder bag through data fusion; The morphological topological map data is used to extract the captured node data and the stacking surface feature data; The flexible grasping module includes a flexible manipulator, which generates grasping path data according to the grasping node data and performs a grasping operation on the powder bag; during the grasping process, the flexible manipulator dynamically adjusts the grasping action by real-time monitoring the grasping force and the change of the powder bag shape, and generates optimized powder bag shape data; The transfer module includes an automatic guided robot, which completes the transfer of the powder bag from the grabbing point to the stacking point according to the optimized powder bag shape data and transfer path data, and outputs transfer status data, wherein the transfer status data includes posture change information of the powder bag, real-time center of gravity adjustment parameters and a time record of reaching the target stacking point, which is used to guide the stacking module to accurately adjust the position and angle of the powder bag; The stacking module comprises a stacking device, wherein the stacking device places the powder bag at a target stacking position according to the transport state data and the stacking surface characteristic data, and generates stacking state data for evaluating the stability of the stacking structure.

2. The automatic powder loading device based on visual recognition according to claim 1 is characterized in that: The visual recognition module includes an RGB-D sensor, an infrared sensor and an ultrasonic sensor, wherein: The RGB-D sensor is used to obtain three-dimensional morphological data of the powder bag through optical imaging and depth information, and the three-dimensional morphological data includes the length, width, height and boundary features of the powder bag; The infrared sensor is used to detect the distribution characteristics of the powder inside the powder bag, and the distribution characteristics are obtained by calculating the infrared thermal radiation signals in different temperature ranges; The ultrasonic sensor acquires the density and edge deformation characteristics of the powder bag material by emitting sound waves to the surface of the powder bag and receiving reflected signals.

3. The automatic powder loading device based on visual recognition according to claim 2 is characterized in that: The visual recognition module includes a data fusion unit, which jointly processes the three-dimensional morphological data, distribution characteristic data and material characteristic data, and generates morphological topological map data through a weighted average algorithm. The morphological topological map data includes the geometric positions and boundary connection information of key nodes of the powder bag; the morphological topological map data is used to extract the grasping node data, and the grasping node data is calculated by the geometric graph analysis algorithm.

4. The automatic powder loading device based on visual recognition according to claim 1 is characterized in that: The flexible manipulator includes a multi-degree-of-freedom joint structure, which realizes multi-axis rotation and extension through servo drive; pressure sensors and contact sensors are set in the fingertip area of ​​the flexible manipulator, the pressure sensors are used to monitor the distribution of applied force during the grasping process in real time, and the contact sensors are used to detect the contact area and morphological changes of the grasping point to dynamically adjust the grasping angle and strength.

5. The automatic powder loading device based on visual recognition according to claim 4 is characterized in that: The flexible manipulator further includes a path planning unit, which generates grasping path data based on grasping node data, and the grasping path data generates a motion trajectory of the flexible manipulator through an inverse kinematics algorithm; the path planning unit dynamically optimizes the motion trajectory according to real-time feedback data, and the optimization objectives include minimizing the grasping time, uniform force on the powder bag, and morphological stability.

6. The automatic powder loading device based on visual recognition according to claim 1 is characterized in that: The automatic guided robot of the transfer module includes a positioning system and an inclination sensor. The positioning system generates a transfer path based on the optimized morphological data of the powder bag and the target stacking point position; the inclination sensor monitors the inclination angle of the powder bag in real time, calculates the adjustment parameters in combination with the center of gravity offset information of the powder bag, and realizes the transportation of the powder bag by adjusting the inclination angle of the automatic guided robot platform.

7. The automatic powder loading device based on visual recognition according to claim 6 is characterized in that: The transfer status data generated by the transfer module includes real-time posture information of the powder bag, center of gravity adjustment parameters and target position arrival time of the powder bag; the center of gravity adjustment parameters are calculated based on the difference between the dynamic position of the center of gravity of the powder bag and the center of gravity position of the robot platform, and are used to guide the automatic guided robot to adjust its motion posture in real time.

8. The automatic powder loading device based on visual recognition according to claim 1 is characterized in that: The stacking device of the stacking module includes a stacking robot arm driven by a servo motor. The stacking robot arm accurately locates the position of the powder bag by receiving the stacking surface feature data in real time. The stacking robot arm driven by the servo motor adjusts the movement speed and placement angle of the robot arm according to the coordinates of the target stacking point; the adjustment of the placement angle is obtained based on the distribution calculation of the stacking contact points.

9. The automatic powder loading device based on visual recognition according to claim 8 is characterized in that: The stacking module includes a thermal distribution monitoring unit, which detects the force distribution on the contact surface of the stacking layer through an infrared imaging device to generate a stacking thermal distribution map; the stacking thermal distribution map is used to analyze the force uniformity of the stacking layer and generate stacking adjustment recommendation data.

10. The automatic powder loading device based on visual recognition according to claim 9 is characterized in that: The self-correction unit in the stacking module triggers the servo motor to adjust the stacking position and angle of the powder bag by real-time detection of the stacking stability score data. The stacking stability score data is generated by analyzing the force concentration and overall balance of each contact point in the stacking thermal distribution diagram, and the adjusted stacking state data is generated.

Images