An automatic powder loading device based on visual recognition

Abstract

The present invention relates to the field of automated loading and logistics control technology, and particularly to an automatic powder loading device based on visual recognition. The device includes a visual recognition module, a flexible grasping module, a transfer module, and a stacking module; morphological, internal distribution, and material characteristic data of powder bags are obtained through multi-modal sensors to generate a grasping path and the grasping operation is completed by a flexible manipulator; the transfer module combines an optimized path and attitude monitoring to dynamically adjust the transportation state of the powder bags; the stacking module based on the stacking thermal distribution map monitors the force on the stacking layer in real time, generates adjustment suggestion data, and completes the optimization of the stacking position and angle through a self-correction unit to ensure the stability of the stacking structure; the present invention significantly improves the efficiency, stacking accuracy, and long-term stability of powder loading, and is applicable to efficient logistics management in complex loading scenarios.

Description

Technical Field

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

[0002] During the process of powder loading, traditional manual loading and semi-automatic loading methods are often inefficient, inaccurate, and pose relatively high safety risks. Automatic powder loading technology can significantly improve the loading efficiency, stacking accuracy, and overall automation level of the logistics process by integrating various sensor technologies and intelligent algorithms. In the prior art, powder loading usually adopts a semi-automatic system combining a mechanical grasping device with a fixed path planning. Although such technology partially realizes the automation of loading, it has obvious deficiencies in dealing with the complexity of powder bag shapes, the accuracy of grasping point selection, and the stacking stability. Specifically:

[0003] Most of the prior art is based on a fixed grasping path or mechanical limit of grasping points, making it difficult to adapt to the morphological changes of powder bags caused by differences in size, shape, or stacking state, and prone to grasping failures or bag slipping; traditional stacking methods often lack the ability to monitor the force distribution and balance of stacking layers in real time and make dynamic adjustments, resulting in uneven or unstable stacking; the prior art lacks the perception and adjustment of the real-time state of powder bags (such as center of gravity offset, attitude change, etc.) during the grasping and stacking processes, and cannot meet the requirements of complex loading scenarios. Summary of the Invention

[0004] In view of the many problems existing in the above prior art, the present invention provides an automatic powder loading device based on visual recognition. The present invention comprehensively perceives the morphological data, internal distribution, and material characteristics of powder bags using a multi-modal visual recognition module, combines dynamic path planning and inverse kinematics algorithms to generate accurate grasping paths, and completes the intelligent grasping and optimized placement of powder bags through a flexible manipulator. The stacking module dynamically adjusts the stacking position and angle through real-time force monitoring and analysis of the stacking thermal distribution map, and finally forms a stable stacking structure. The present invention significantly improves the loading efficiency and stacking stability, and meets the requirements of complex loading scenarios.

[0005] An automatic powder loading device based on visual recognition, comprising:

[0006] A visual recognition module, configured to collect the morphological data, internal distribution data, and surface characteristic data of powder bags, and generate morphological topology map data of powder bags through data fusion; the morphological topology map data is used to extract grasping node data and stacking surface feature data;

[0007] The flexible grasping module includes a flexible manipulator. The flexible manipulator 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 morphological changes of the powder bag, and generates optimized powder bag morphological data.

[0008] The transfer module includes an automatic guided robot. The automatic guided robot completes the transfer of the powder bag from the grasping point to the stacking point according to the optimized powder bag morphological data and the transfer path data, and outputs transfer status data. The transfer status data includes the attitude change information of the powder bag, the real-time center of gravity adjustment parameter, and the 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.

[0009] The stacking module includes a stacking device. The stacking device places the powder bag at the target stacking position according to the transfer status data and the stacking surface feature data, and generates stacking status data, which is used to evaluate the stability of the stacking structure.

[0010] Preferably, the visual recognition module includes an RGB-D sensor, an infrared sensor, and a ultrasonic sensor, where:

[0011] The RGB-D sensor is used to obtain the three-dimensional morphological data of the powder bag through optical imaging and depth information. The three-dimensional morphological data includes the length, width, height, and boundary features of the powder bag.

[0012] The infrared sensor is used to detect the distribution characteristics of the powder inside the powder bag. The distribution characteristics are calculated through the infrared thermal radiation signals in different temperature ranges.

[0013] The ultrasonic sensor emits sound waves to the surface of the powder bag and receives the reflected signals to obtain the density and edge deformation characteristics of the powder bag material.

[0014] Preferably, the visual recognition module includes a data fusion unit. The data fusion unit jointly processes the three-dimensional morphological data, the distribution characteristic data, and the material characteristic data, and generates morphological topology map data through a weighted average algorithm. The morphological topology map data contains the geometric positions and boundary connection information of the key nodes of the powder bag. The morphological topology map data is used to extract the grasping node data, which is calculated by a geometric graph analysis algorithm.

[0015] Preferably, the flexible manipulator includes a multi-degree-of-freedom joint structure, and the joint structure realizes multi-axis rotation and telescoping through servo drive; a pressure sensor and a contact sensor are arranged in the fingertip area of the flexible manipulator. The pressure sensor is used to monitor the applied force distribution during the grasping process in real time, and the contact sensor is used to detect the contact area and morphological changes of the grasping point to dynamically adjust the grasping angle and strength.

[0016] Preferably, the flexible manipulator further includes a path planning unit. The path planning unit generates grasping path data based on the grasping node data, and the grasping path data generates the motion trajectory of the flexible manipulator through the inverse kinematics algorithm; the path planning unit dynamically optimizes the motion trajectory according to the real-time feedback data, and the optimization objectives include the minimization of the grasping time, the force uniformity of the powder bag, and the morphological stability.

[0017] Preferably, 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 position of the target stacking point; the inclination sensor monitors the inclination angle of the powder bag in real time, calculates 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.

[0018] Preferably, the transfer status data generated by the transfer module includes the real-time attitude information of the powder bag, the center of gravity adjustment parameter, and the target position arrival time of the powder bag; the center of gravity adjustment parameter is 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 is used to guide the automatic guided robot to adjust the motion attitude in real time.

[0019] Preferably, the stacking device of the stacking module includes a stacking robotic arm driven by a servo motor. The stacking robotic arm accurately positions the powder bag according to the real-time received stacking surface feature data, and the stacking robotic arm driven by the servo motor adjusts the moving speed and placement angle of the robotic arm according to the coordinates of the target stacking point; the adjustment of the placement angle is calculated based on the distribution of the stacking contact points.

[0020] Preferably, the stacking module includes a thermal distribution monitoring unit. The thermal distribution monitoring unit 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 suggestion data.

[0021] 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 detecting the stacking stability scoring data in real time. The stacking stability scoring data is generated by analyzing the force concentration degree and overall balance of each contact point in the stacking thermal distribution map, and generates the adjusted stacking status data.

[0022] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0023] Through the multi-modal visual recognition technology, the present invention realizes the intelligent extraction and dynamic adjustment of the grasping points, solves the problem of inaccurate selection of grasping points in the prior art, and meets the requirements of the complexity of the powder bag form;

[0024] Through the real-time force monitoring and self-calibration technology, the present invention realizes the uniform distribution of the force on the stacking layers and the stability of the overall structure, and effectively solves the problems of uneven stacking and instability in the prior art;

[0025] Through the dynamic path planning and grasping optimization technology, the present invention realizes the efficient and stable operation of the whole process from grasping to stacking of the powder bags, and improves the overall loading efficiency;

[0026] Through the stacking thermal distribution map and the real-time feedback mechanism, the present invention realizes the intelligent optimization of the multi-layer stacking process, and further ensures the long-term stability of the stacking layers. Description of the Drawings

[0027] Figure 1 is the structural block diagram of the device of the present invention;

[0028] Figure 2 is the schematic flow diagram of the visual recognition module in the present invention;

[0029] Figure 3 is the working schematic diagram of the flexible grasping module in the present invention;

[0030] Figure 4 is the schematic flow diagram of the stacking module in the present invention. Detailed Embodiments

[0031] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, numerous specific details are set forth in order to provide a comprehensive understanding of the embodiments of the present disclosure.

[0032] As Figure 1 shown, an automatic powder loading device based on visual recognition includes:

[0033] A visual recognition module, configured to collect the form data, internal distribution data, and surface characteristic data of the powder bag, and generate the form topology map data of the powder bag through data fusion; the form topology map data is used to extract the grasping node data and the stacking surface feature data;

[0034] As Figure 2As shown, the core of the visual recognition module lies in using multi-modal sensors to comprehensively collect the characteristics of the powder bags. The types of data collected include morphological data, internal distribution data, and surface characteristic data. The acquisition of these data relies on the following sensors and technologies:

[0035] The 3D point cloud data of the powder bag is obtained through an 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 generate a 3D morphological model through the superposition processing of the depth map and the RGB image.

[0036] The infrared sensor uses the infrared thermal radiation characteristics of different temperature regions to detect the distribution of the powder inside the powder bag. Specifically, the stacking density and distribution uniformity of the powder in the packaging bag will cause differences in the infrared thermal radiation signal, and through these differences, it can be judged whether there is uneven stacking or voids inside the powder.

[0037] The ultrasonic sensor emits sound waves to the surface of the powder bag and receives the reflected signals to measure its material characteristics and edge deformation. This process uses the correlation between the sound wave propagation speed and the material density to infer the integrity of the powder bag surface and edge characteristics.

[0038] After the above data is collected, the data fusion unit jointly processes the multi-modal data to generate the morphological topology 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 topology model according to the importance through the weighted average algorithm.

[0039] The topology map is composed of nodes and edges. Among them, the nodes represent the important geometric feature points of the powder bag, such as edge points and surface key points; the edges represent the spatial connection relationships between the geometric feature points, such as the continuity of the surface contour or the proximity relationship of the edge points.

[0040] The generation formula of the morphological topology map can be expressed as:

[0041]

[0042] where, The generated morphological topology map data; The weight coefficient, which is used to represent the importance of different data sources (such as morphological data, distribution data, surface characteristic data); The set of geometric characteristics generated by different data sources; The number of data sources. By reasonably setting the weights , the system can dynamically adjust the contribution rates of different data sources to ensure that the generated topology map can comprehensively and accurately reflect the true state of the powder bag.

[0043] The grasping node data is extracted from the morphological topology map using a geometric graph analysis algorithm. The grasping node refers to the position of the optimal contact point when the flexible manipulator grasps the powder bag, and 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 surface of the grasping point is smooth).

[0044] The stacked surface feature data is obtained by analyzing the set of surface area nodes in the morphological topology map. These features include surface flatness, area size, and the spatial distribution density of surface nodes. The calculation of the stacked surface features helps the subsequent stacking module select the optimal placement surface and evaluate the stability of the stacked structure.

[0045] Preferably, the visual recognition module includes an RGB-D sensor, an infrared sensor, and a ultrasonic sensor, wherein:

[0046] The RGB-D sensor is used to obtain the three-dimensional morphological data of the powder bag through optical imaging and depth information. The three-dimensional morphological data includes the length, width, height, and boundary features of the powder bag;

[0047] The infrared sensor is used to detect the distribution characteristics of the powder inside the powder bag. The distribution characteristics are obtained by calculating the infrared thermal radiation signals in different temperature ranges;

[0048] The ultrasonic sensor emits sound waves to the surface of the powder bag and receives the reflected signals to obtain the density and edge deformation characteristics of the powder bag material.

[0049] The RGB-D sensor combines the functions of optical imaging (RGB) and depth perception (Depth) 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 measures the distance from each point on the surface of the powder bag to the sensor through structured light or time-of-flight technology to generate three-dimensional point cloud data.

[0050] The three-dimensional morphological data includes the length, width, height, and boundary features of the powder bag. The length, width, and height are calculated from the maximum and minimum values of the point cloud along the axes, and the boundary features are extracted by analyzing the distribution and connection relationship of the boundary points in the point cloud. The three-dimensional morphological data provides the basic geometric information for the extraction of grasping nodes and stacked surface features. By analyzing the point cloud data, the center of gravity position of the powder bag, the flatness of the edge points, and the areas suitable for grasping and stacking can be identified.

[0051] 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 the powder in different temperature zones causes changes in the reflected and transmitted radiation signals. By processing these signals, it is possible to identify whether the powder accumulation inside the powder bag is uniform.

[0052] The detection of the temperature zone signal is based on the relationship between the infrared radiation intensity and the temperature change. The intensity difference in the reflection signal is used to calculate the distribution characteristics through the following formula:

[0053] Where, represents the reflection coefficient, reflecting the powder distribution characteristics; represents the intensity of the detected reflection signal; represents the intensity of the incident signal. The change of the reflection coefficient can be used to judge the bulk density and internal void situation of the powder. The internal distribution characteristic data can guide the flexible manipulator to avoid unevenly stacked areas during grasping, and at the same time help optimize the force distribution when the powder bags are stacked.

[0054] The ultrasonic sensor measures the material density and edge deformation characteristics by emitting sound waves to the surface of the powder bag and receiving its reflection signal. When sound waves propagate in media with different densities, their speed and attenuation characteristics will 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. The density calculation formula:

[0055]

[0056] Where, represents the propagation speed of sound waves in the medium; represents the elastic modulus of the material; represents the density of the material. By measuring the sound wave speed , and combining with the elastic modulus of the material, the specific value of the material density can be calculated. The material density and edge deformation characteristics provide additional support for grasping and stacking, such as ensuring that the grasping force is suitable for the material strength of the powder bag and evaluating the impact of edge deformation on the stacking stability.

[0057] The visual recognition module realizes comprehensive data acquisition of the powder bag from geometric structure to internal characteristics and then to surface properties by integrating the different sensing capabilities of RGB-D sensors, infrared sensors, and ultrasonic sensors. Compared with a single-sensor system, this multi-modal 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 detection of the internal powder distribution by the infrared sensor makes up for the limitation that the visual sensor cannot penetrate the material, while the ultrasonic sensor provides high-precision measurements 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 points to ensure grasping stability, but also assist the transfer module to optimize the 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.

[0058] Example 1: The three-dimensional shape data of a certain powder bag is collected by an RGB-D sensor, and it is calculated that its length is 60 cm, width is 40 cm, and height is 20 cm. The point cloud analysis shows that the boundary points of the powder bag are evenly distributed and the shape is regular, and the center of gravity is located at (30, 20, 10) (three-dimensional coordinates). This data is used for the extraction of grasping nodes of the flexible manipulator, and the grasping points close to the center of gravity and with a flat surface are selected to ensure grasping stability.

[0059] Example 2: The powder bag is scanned by an infrared sensor, and it is detected that the reflection coefficient of a certain area is significantly lower than that of other areas, indicating that the powder accumulation in this area is relatively sparse. Combining with the grasping path planning, the flexible manipulator avoids this area for grasping to prevent deformation caused by uneven powder accumulation during the grasping process.

[0060] Example 3: The surface density of a certain 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 at the edge of the powder bag. Combining with the stacking surface feature extraction algorithm, the side with the smallest deformation of the powder bag is selected as the stacking surface to ensure stacking stability.

[0061] Preferably, the visual recognition module includes a data fusion unit. The data fusion unit jointly processes the three-dimensional shape data, distribution characteristic data, and material characteristic data, and generates morphological topology map data through a weighted average algorithm. The morphological topology map data contains the geometric positions and boundary connection information of the key nodes of the powder bag; the morphological topology map data is used to extract grasping node data, and the grasping node data is calculated by a geometric graph analysis algorithm.

[0062] Multimodal data refers to three-dimensional morphological data, distribution characteristic data, and material characteristic data obtained through different sensors. Due to the different emphases of various sensors on perceiving target characteristics, the collected data has its own advantages and limitations. For example, an RGB-D sensor can provide the geometric structure of the powder bag, but it cannot reflect the distribution state of the powder inside; an infrared sensor can detect the internal distribution, but the boundary characteristic information is weak; an ultrasonic sensor focuses on material characteristics and boundary deformation. To comprehensively reflect the physical state of the powder bag, these data must be jointly processed by a data fusion unit.

[0063] The data fusion unit integrates different data sources through a weighted average algorithm. The core idea of the weighted average algorithm is to assign different weights according to the reliability of each data source or the contribution degree to the target characteristics, and perform weighted processing on the characteristic values of the same target. The specific formula is:

[0064]

[0065] Among them, represents the fused characteristic data; represents the data from different sensors; represents the weight corresponding to the data, satisfying ; represents the number of sensors. By setting appropriate weights, the contribution of certain key data can be highlighted preferentially. For example, the three-dimensional morphological data important for the grasping operation may have a higher weight.

[0066] After completing the data fusion, the system generates morphological topology map data. The morphological topology map is a high-dimensional modeling of the geometric characteristics and boundary connection relationships of the powder bag, and its structure consists of nodes and edges: Nodes represent important geometric points of the powder bag, such as the center of gravity position, boundary feature points, or surface key points. Edges represent the connection relationships between nodes and are used to describe the contour, boundary continuity, and local structure of the powder bag surface.

[0067] Node set and edge set are generated based on the point cloud analysis of three-dimensional morphological data and material characteristic data. The connection relationship is determined by calculating the spatial distance and surface angle change of adjacent points in the point cloud to form a complete topology map structure.

[0068] The extraction of grasping node data depends on the geometric graph analysis algorithm, which determines the grasping points suitable for the flexible manipulator operation by analyzing the node characteristics (such as stability, smoothness, neighborhood distribution, etc.) in the morphological topology map.

[0069] Stability: The grasping node needs to be close to the center of gravity of the powder bag to ensure the balanced force during grasping.

[0070] Smoothness: The surface of the grasping point needs to be flat to avoid the sliding of the powder bag during the grasping process.

[0071] Neighborhood distribution: The density of the neighborhood distribution of the grasping nodes needs to be moderate to ensure the structural stability of the grasping area.

[0072] Calculation formula: The score of the grasping node is calculated by the following formula:

[0073]

[0074] where represents the score of the grasping node; represents the score of the geometric stability of the node; represents the score of the surface smoothness of the node; represents the score of the density of the neighborhood node distribution; represents the weight parameter, which is set according to the requirements of the grasping task.

[0075] The data fusion unit effectively integrates multi-modal data from different sensors through the weighted average algorithm, enabling the fused morphological topology map to comprehensively 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 data deviation or blind spot problems that may be brought by a single sensor. The accuracy of the morphological topology map directly affects the extraction accuracy of the grasping nodes and the characteristics of the stacking surface. 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 has high adaptability and stability, and significantly reducing the risk of grasping failure or powder bag sliding.

[0076] Example 1: A certain powder bag collects the following data through RGB-D sensors, infrared sensors, and ultrasonic sensors:

[0077] Three-dimensional morphological data: The geometric structure represented by the point cloud, and the center of gravity is initially calculated as (30, 20, 15);

[0078] Distribution characteristic data: The score of the internal distribution uniformity is 0.85 (full score 1.0);

[0079] Material characteristic data: The density is measured as 1.2 g / cm 3 , and the edge deformation detection is 0.2.

[0080] The data fusion unit calculates according to the weight Fuse the data to generate morphological topology map data, and display the geometric positions and boundary connection information of key nodes. Among them, the connection relationship of boundary points is calculated based on the distance threshold (within 5 mm) of adjacent points in the point cloud.

[0081] Example 2: Based on the morphological topology map data generated in Example 1, the scoring result of the geometric graph analysis algorithm shows that the grasping node is located at (28, 20, 14), and the score . This node is close to the center of gravity of the powder bag, with a smooth surface and a moderate neighborhood distribution density, making it suitable for the flexible manipulator to perform the grasping operation.

[0082] The flexible grasping module includes a flexible manipulator. The flexible manipulator 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 morphological changes of the powder bag, and generates optimized powder bag morphological data;

[0083] The grasping action of the flexible grasping module is executed 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 the grasping path data depends on the inverse kinematics algorithm, the core of which is to calculate the spatial motion trajectories of the joints of the flexible manipulator to ensure that the end of the manipulator can reach the target grasping node.

[0084] The grasping path data includes information such as the three-dimensional coordinates of the grasping point, the attitude angle of the end of the flexible manipulator, and the clamping force. Through the dynamic adjustment of these parameters, the flexible manipulator can accurately align with the grasping node and perform actions. After the path data is generated, it also needs to be optimized in combination with the on-site environment (such as the stacking angle of the powder bag, the distance between the manipulator and the powder bag, etc.) to reduce the movement time of the manipulator, avoid collision risks, and ensure the grasping efficiency.

[0085] During the grasping process, the flexible manipulator realizes real-time monitoring and dynamic adjustment through the built-in torque sensor and morphological sensor:

[0086] The torque sensor detects the grasping force exerted by the flexible manipulator during the grasping process and compares it with the target grasping force in real time. If the grasping force is too large, it may cause damage to the powder bag; if the grasping force is insufficient, it may lead to unstable grasping. Based on the real-time monitoring results, the manipulator can automatically adjust the grasping force to ensure the stability and safety of the grasping action.

[0087] The morphological sensor captures the deformation of the powder bag during the grasping process due to the applied force and updates the morphological data of the powder bag in real time. If it is detected that the powder bag has morphological changes due to internal non-uniformity, the system can automatically adjust the clamping position and angle of the grasping point to ensure that the grasping action is always within the allowable range of the morphology.

[0088] After the grasping is completed, the flexible grasping module optimizes the final powder bag shape data and generates the optimized powder bag shape data. The optimized powder bag shape data includes information such as the posture of the powder bag, the surface flatness, and the center of gravity position. These data are jointly provided by the torque sensor and the shape sensor of the flexible manipulator and are calculated and generated through the shape optimization algorithm inside the system. The optimized powder bag shape data provides a stable powder bag state input for the transfer module, ensuring that the powder bag will not slip or tip over due to improper grasping during the transfer process, and at the same time providing higher shape adaptability for the stacking module.

[0089] Preferably, the flexible manipulator includes a multi-degree-of-freedom joint structure, and the joint structure realizes multi-axis rotation and telescoping through servo drive; a pressure sensor and a contact sensor are arranged in the fingertip area of the flexible manipulator. The pressure sensor is used to monitor the applied force distribution during the grasping process in real time, and the contact sensor is used to detect the contact area and shape change of the grasping point to dynamically adjust the grasping angle and force.

[0090] The multi-degree-of-freedom joint structure of the flexible manipulator realizes multi-axis rotation and telescoping through servo drive, endowing it with flexible movement ability, and can adapt to powder bags of different sizes, shapes and positions.

[0091] The joint structure usually includes three rotating joints and one telescoping joint, which are used to control the three degrees of freedom (position adjustment) of the end of the manipulator and the clamping action (end control). Each joint is driven by an independent servo motor, and the servo motor 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 the multi-degree-of-freedom joints, and at the same time adjust the grasping posture to meet the requirements of different powder bag positions and postures.

[0092] The fingertip area of the flexible manipulator is equipped with a pressure sensor, which is used to detect the force distribution applied to the powder bag during the grasping process and is dynamically adjusted through a 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 caused by excessive force application, and at the same time avoid grasping failure caused by insufficient force application.

[0093] The contact sensor is used to monitor the contact area and morphological changes at the grasping point to evaluate whether the grasping point is in full contact with the surface of the powder bag and to detect the morphological changes of the powder bag during the grasping process. By sensing the capacitance change or pressure distribution in the contact sensor, the size of the contact area between the fingertip and the powder bag is calculated. If the contact area is too small, it indicates that the grasping point may deviate from the expected position, and the system will adjust the grasping angle and force to increase the contact area and ensure stable clamping. When the powder bag deforms due to the grasping force or the shift of the center of gravity, the contact sensor will capture these changes and feed them back to the control system in real time. The system combines the pressure data and the morphological change data to adjust the grasping strategy, such as changing the position or force of the grasping point, to adapt to the deformation characteristics of the powder bag.

[0094] The flexible manipulator dynamically adjusts the grasping angle and force based on the real-time data provided by the pressure sensor and the contact sensor. When the contact sensor detects that the contact area is too small or the morphological change exceeds the threshold, the system calculates the normal direction of the surface at the grasping point and instructs the servo motor to adjust the joint angle so that the gripper is in perpendicular contact with the surface of the powder bag to maximize the contact stability. By comparing the difference between the current grasping force and the target grasping force, the amount of torque adjustment required for the servo motor is calculated, thereby dynamically changing the clamping force to ensure that the powder bag is not damaged during the grasping process and can be firmly clamped.

[0095] With 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 actions of the manipulator are rapid and accurate, effectively improving the grasping efficiency and adaptability. The real-time grasping force monitoring by the pressure sensor and the contact area detection by 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 the contact area are always within the optimal range and avoiding grasping failure or damage to the morphology of the powder bag. 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 enhancing the adaptability of the grasping action to powder bags with complex morphologies. This dynamic adjustment mechanism greatly enhances the processing ability of the grasping module for irregular powder bags.

[0096] Example 1: In the loading task, the grasping node coordinates of a certain powder bag are (25, 15, 10), and its surface material is relatively soft. After the flexible manipulator reaches the target grasping point through the grasping path planning, it starts to apply the clamping force. The grasping force detected by the pressure sensor is 12 N, while the system target grasping force is 10 N. The control system adjusts the output torque of the servo motor in real time according to the difference between the grasping force exceeding the target value, reducing the grasping force to 10 N to ensure stable grasping without damaging the powder bag.

[0097] Example 2: During the grasping process in 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 sliding risk at the grasping point. By analyzing the pressure distribution data of the contact sensor, the system calculated an adjustment angle of 5 degrees. The servo motor adjusted the joint angle to make the gripper contact the surface of the powder bag vertically while maintaining a uniform pressure distribution. Eventually, the contact area reached the expected value, and the grasping stability was ensured.

[0098] Example 3: During the grasping process, the contact sensor detected that the powder bag deformed in shape due to uneven internal powder distribution, resulting in a reduction in the contact area and a decrease in the clamping stability. The control system adjusted the grasping point position in combination with the shape change data, fine-tuning the grasping point from (25, 15, 10) to (24, 16, 11), and at the same time readjusted the clamping force to match the new contact shape, successfully completing the grasping task.

[0099] Preferably, as Figure 3 shown, the flexible manipulator further includes a path planning unit. The path planning unit generates grasping path data based on the grasping node data, and the grasping path data generates the motion trajectory of the flexible manipulator through the inverse kinematics algorithm; the path planning unit dynamically optimizes the motion trajectory according to the real-time feedback data, and the optimization objectives include the minimization of the grasping time, the force uniformity of the powder bag, and the shape stability.

[0100] The path planning unit takes the grasping node data provided by the visual recognition module as input and uses the inverse kinematics algorithm to generate the grasping path data. The grasping path data describes the motion trajectory of the flexible manipulator from the initial position to the target grasping point, including the angle changes, speeds, and final postures of each joint.

[0101] 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 ( ) and the posture angle ( ) of the target grasping point, the inverse kinematics algorithm generates the rotation angles and telescopic distances of all joints by solving the geometric structure equations of the flexible manipulator.

[0102] The grasping path data contains specific parameters such as the spatial coordinates of the grasping nodes, the joint angle sequence of the flexible manipulator, the clamping force, and the motion speed, providing a complete execution plan for the flexible manipulator.

[0103] During the grasping operation, the path planning unit dynamically optimizes the motion trajectory in combination with the real-time feedback data to ensure that the manipulator can adapt to the changes in the complex environment.

[0104] The real-time feedback data includes information such as the current position, speed, grasping force of the manipulator, and the morphological changes of the powder bag. These data are provided by the torque sensors, morphological sensors, and position encoders built into the manipulator, and are transmitted to the path planning unit for processing in real time. The optimization of the motion trajectory focuses on three objectives:

[0105] Minimization of grasping time: By optimizing the joint path of the manipulator, unnecessary motion time is reduced, and the loading efficiency is improved. Force uniformity of the powder bag: According to the morphology and force distribution of the powder bag, the grasping path is adjusted to ensure that the powder bag will not slide or be damaged due to uneven force during the grasping process. Morphological stability: Combining the real-time morphological data of the powder bag, the motion speed and clamping angle of the manipulator are optimized to prevent morphological deformation caused by too fast or too slow motion during the grasping process.

[0106] The path planning unit adopts an incremental path adjustment algorithm. By comparing the actual position of the manipulator with the target path in real time, a correction amount is generated and the motion trajectory is updated. Each updated correction amount ensures that the manipulator can quickly converge to the target path while maintaining a smooth grasping action.

[0107] Through the inverse kinematics algorithm, the path planning unit can generate accurate motion trajectories for the flexible manipulator, ensuring that the manipulator can accurately reach the grasping node position and complete the grasping action with the best posture. The accuracy of the motion trajectory directly affects the success rate of grasping and the stability of the powder bag. Combining the real-time feedback data to dynamically optimize the motion trajectory, the path planning unit can adapt to the changes in the grasping environment in real time, including the morphological changes 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 at minimizing the grasping time and significantly improves the overall efficiency of the grasping operation by reducing the motion redundancy and unnecessary actions of the manipulator. This time optimization is of great significance in high-frequency loading tasks.

[0108] Example 1: In a certain loading task, the vision recognition module provides the grasping node data. The three-dimensional coordinates of the target grasping point are (50, 30, 20), and the attitude angle is (90°, 0°, 0°). The path planning unit uses the inverse kinematics algorithm to calculate the joint parameters of the flexible manipulator. Among them, the rotation angle of the shoulder joint is 45°, the rotation angle of the elbow joint is 30°, the clamping force of the end effector is 8N, and the grasping path data also includes the moving speed of the manipulator is 0.5 m / s. The flexible manipulator reaches the grasping point according to the generated path data and completes the grasping action.

[0109] Example 2: During the grasping process in Example 1, the torque sensor detected that the powder bag changed its shape due to the applied force, resulting in a 5-mm deviation in the position of the grasping node. At the same time, the shape sensor detected local deformation in the boundary area of the powder bag, affecting the grasping stability. The path planning unit combined this real-time feedback data to update the grasping path, adjusted the target position of the end effector to (50.5, 30.2, 20.1), and reduced the moving speed of the manipulator to 0.4 m / s to ensure the stability of the grasping action. The optimized motion trajectory enabled the successful completion of the grasping operation, and the shape of the powder bag remained stable.

[0110] Example 3: In another loading task, the path planning unit optimized the grasping time to 2 seconds in combination with the loading efficiency requirement. At the same time, during the grasping action, the joint path and grasping force were adjusted in real time to ensure uniform force distribution on the powder bag. The real-time feedback data showed that the uniformity score of the optimized grasping force distribution was 0.95 (full score 1.0), the shape stability score was 0.90, and the grasping time was reduced from the original 2.8 seconds to 2 seconds, significantly improving the loading efficiency.

[0111] The transfer module includes an automated guided vehicle (AGV). The AGV completes the transfer of the powder bag from the grasping point to the stacking point according to the optimized powder bag shape data and transfer path data, and outputs transfer status data. The transfer status data includes the attitude change information of the powder bag, real-time center of gravity adjustment parameters, and time records of reaching the target stacking point, which are used to guide the stacking module to accurately adjust the position and angle of the powder bag.

[0112] The automated guided vehicle (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 and includes the motion trajectory, turning angle, moving speed, etc. of the robot from the grasping point to the stacking point. The optimized path takes into account factors such as the distribution of site obstacles, path minimization, and energy consumption minimization. The AGV uses a combination of lidar (LIDAR) and ultra-wideband (UWB) positioning systems for joint positioning, real-time perceives environmental changes, and executes the path trajectory. Through the feedback control system, the AGV can automatically correct the deviation during the driving process to ensure the accuracy of path execution.

[0113] During the transfer process, the AGV monitors the attitude change of the powder bag in real time to ensure its stability during movement. The attitude change data is jointly obtained by the vision sensor and acceleration sensor installed on the top of the AGV. The vision sensor is used to capture the tilt angle and position change 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 attitude change exceeds the set threshold, the AGV corrects it by adjusting the speed and driving trajectory, such as decelerating or changing the direction of the tilt angle, to restore the stable state of the powder bag.

[0114] During the transfer process, the powder material bag may have its center of gravity shifted due to irregular shape or path turning. The center of gravity adjustment mechanism calculates the center of gravity position of the powder material bag in real time and dynamically adjusts the tilt angle of the AGV platform to achieve a redistribution of the center of gravity. By calculating the current center of gravity position based on the shape data and attitude change data of the powder material bag, and comparing it with the ideal center of gravity position, the power module of the AGV is instructed to adjust the tilt 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 material bag has shifted to the edge of the platform, the AGV will control the platform to tilt at a certain angle in the opposite direction through a servo motor to readjust the center of gravity to the center of the platform.

[0115] The transfer status data consists of the attitude change information of the powder material bag, the real-time center of gravity adjustment parameters, and the time record of reaching the target stacking point. The specific contents include: information such as the tilt angle and horizontal offset of the powder material bag during the transfer process, which is used to evaluate the stability of the transfer. Record the center of gravity offset of the powder material bag and the corresponding adjustment actions to provide the current balance state of the powder material bag for the stacking module. Record the time when the powder material bag reaches the target stacking point, which is used for the synchronization operation of the stacking module.

[0116] By combining the lidar and UWB positioning systems, the AGV can accurately execute the transfer path data in a complex site environment, avoid collisions and offsets, and ensure that the powder material bag can be safely transported from the grasping point to the stacking point in the shortest time and with the lowest energy consumption. Through attitude change monitoring and dynamic center of gravity adjustment, the AGV can correct the unstable state of the powder material bag in real time during the transfer process, such as tilting or sliding caused by irregular shape. This real-time adjustment mechanism effectively reduces the risk of damage to the powder material bag during the transfer process and improves the stability of the transfer. 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 attitude and center of gravity state of the powder material bag, further improving the overall loading efficiency of the entire system.

[0117] Example 1: In a certain loading task, the AGV receives the optimized path data from the grasping point to the stacking point. The total path length is 10 meters and includes two 90° turns. The AGV travels along the path through the lidar and UWB positioning systems. During the transfer process, due to unexpected obstacles in the site, the AGV adjusts the steering angle through the real-time path correction function, bypasses the obstacles and resumes the original path, ensuring that the powder material bag reaches the stacking point smoothly.

[0118] Example 2: During the transfer process of Example 1, the acceleration sensor detects that the powder material bag tilts forward due to sudden deceleration, and the tilt angle is 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 tilt angle of the platform to -3° through the servo motor to readjust the center of gravity of the powder material bag back to the center position of the platform and restore the stable state.

[0119] Example 3: After Example 2 is completed, the transfer module generates transfer status data, including the record of the change in the inclination angle of the powder bag (initially inclined at 8°, finally inclined at 1°), the center-of-gravity adjustment parameter (the platform is inclined at -3°), and the time record of reaching the stacking point (12 seconds). The stacking module calculates the adjustment amount of the stacking angle of the powder bag based on these data to ensure that the powder bag can be stably stacked at the target stacking position.

[0120] Preferably, 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 position of the target stacking point. The inclination sensor real-time monitors the inclination angle of the powder bag, calculates the adjustment parameter 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.

[0121] The positioning system is the core unit of the automatic guided robot and is responsible for generating the optimal transfer path based on the optimized morphological data of the powder bag and the position of the target stacking point. Input data includes the morphological data of the powder bag (including the center-of-gravity position, size, and morphological characteristics) and the three-dimensional coordinates of the target stacking point. The path generation is based on environmental perception and dynamic programming algorithms, and combines lidar (LIDAR) and ultra-wideband (UWB) positioning systems to perform real-time environmental modeling, generating 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.

[0122] During the transfer process, the powder bag may experience attitude changes or center-of-gravity offsets due to dynamic factors such as acceleration, deceleration, and turning. The inclination sensor real-time detects the inclination angle of the powder bag and, in combination with the center-of-gravity offset information collected by the sensor, dynamically adjusts the inclination angle of the platform. The inclination sensor is installed on the robot platform and is used to measure the inclination angle of the powder bag relative to the platform. If the inclination angle exceeds the set threshold (e.g., ±5°), the system will trigger the center-of-gravity adjustment mechanism. By combining the morphological data of the powder bag and the real-time data of the inclination sensor, the offset amount between the current center-of-gravity position of the powder bag and the center of the platform is calculated, and an adjustment parameter is generated to correct the inclination angle of the transport platform. The automatic guided robot adjusts the inclination angle of the platform through a servo motor to redistribute the center of gravity of the powder bag into the safe area of the platform, avoiding tipping or sliding.

[0123] The positioning system and the tilt sensor work together to ensure the safety and efficiency of the powder bag during transportation. The positioning system provides precise path execution, while the tilt sensor monitors the attitude changes in real time and dynamically adjusts the platform angle, ultimately ensuring that the powder bag can reach the target stacking point in a stable state. When turning or accelerating / decelerating, the feedback from the tilt sensor is combined with the speed and direction of the path planning to automatically adjust the transportation parameters. For example, during a sharp turn, the system will reduce the speed and appropriately adjust the tilt angle to ensure the powder bag remains stable.

[0124] 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 efficiency of the path planning not only increases the transfer speed but also reduces unnecessary paths and energy waste. The tilt sensor monitors the attitude changes and center of gravity offset of the powder bag in real time. By dynamically adjusting the platform tilt angle, it effectively solves problems such as tilting and sliding caused by inertial forces during transportation, ensuring the safety and stability of the powder bag during transportation. The generated tilt 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 position and angle of the powder bag, improving the stacking accuracy and overall stability.

[0125] Example 1: In a certain 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 and includes two 45° turns. The positioning system uses lidar to scan the environment, identifies an obstacle on the path, dynamically plans an obstacle avoidance path, increases the path length to 8.5 meters, and adjusts the speed to 0.4 m / s at the same time. The automatic guided robot travels along the planned path and accurately reaches the target stacking point to complete the transportation.

[0126] Example 2: During the transportation in Example 1, the tilt sensor detects that the powder bag tilts due to the inertial force during the acceleration stage, and the tilt angle reaches 6°, exceeding the set threshold of 5°. The system calculates the center of gravity offset in real time to be 10 cm and adjusts the tilt angle of the transportation platform to -3° through the servo motor, finally restoring the tilt angle of the powder bag to within 1° and the center of gravity back to the center position of the platform.

[0127] Example 3: After Example 2 is completed, the end section of the transfer path includes a 90° turn. The robot automatically reduces the speed to 0.3 m / s during the turn. At the same time, the tilt sensor monitors the risk of side tilt of the powder bag due to the turn, and the platform makes another fine adjustment to correct the side tilt 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 changes, center of gravity adjustment parameters, and arrival time for the stacking module to use.

[0128] Preferably, the transfer status data generated by the transfer module includes the real-time attitude information of the powder bag, the center of gravity adjustment parameter, and the target position arrival time of the powder bag; the center of gravity adjustment parameter is 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 is used to guide the automatic guided robot to adjust its motion attitude in real time.

[0129] The real-time attitude information of the powder bag refers to the position and angle changes of the powder bag relative to the platform during transportation, including parameters such as tilt angle, horizontal offset, and rotation attitude. The automatic guided robot obtains the real-time attitude information through visual sensors, inclinometers, and acceleration sensors installed on the platform. Among them, the visual sensor captures the three-dimensional spatial position of the powder bag, the inclinometer measures its tilt angle relative to the horizontal plane, and the acceleration sensor detects the change of inertial force during motion. The system compares the real-time collected attitude data with the ideal transportation state of the powder bag, calculates the attitude offset, and uses it as the reference input for subsequent adjustment of the motion attitude. For example, when the tilt angle deviates from the target value by more than the set threshold (such as ±5°), the system will trigger the center of gravity adjustment mechanism.

[0130] The center of gravity adjustment parameter is 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 guides the robot to adjust its motion attitude 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 sensor data. For example, according to the size, mass distribution, and current tilt angle of the powder bag, the three-dimensional coordinate position of its dynamic center of gravity is determined. By calculating the offset between the dynamic center of gravity of the powder bag and the center position of the robot platform, the center of gravity adjustment parameter is generated:

[0131] Among them, represents the center of gravity adjustment parameter, representing the difference between the current center of gravity and the ideal center of gravity; represents the current dynamic center of gravity coordinates of the powder bag; represents 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.

[0132] The target position arrival time refers to the total time for the powder bag to be transported from the grasping point to the stacking point, which is used to evaluate the transportation efficiency and guide the synchronous operation of the stacking module. The estimated arrival time of the target position is calculated through 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 the basis for evaluating the transportation efficiency, and at the same time helps 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.

[0133] By acquiring and analyzing real-time attitude information, the transfer module can comprehensively grasp the dynamic state of the powder bag during transportation, including changes such as tilt, offset, and rotation. This precise attitude monitoring provides sufficient data support for subsequent dynamic adjustments, effectively reducing the risk of the powder bag slipping or tipping during transportation. The real-time calculation and feedback of the center-of-gravity adjustment parameters guide the automatic guided robot to quickly respond to attitude changes. By adjusting the platform tilt angle or speed, the center-of-gravity adjustment mechanism can restore the balance state of the powder bag in the shortest time, improving the stability and safety of the transportation process. The attitude 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 bag, optimize the stacking effect, and improve the efficiency of the overall loading process.

[0134] Example 1: In a certain loading task, during the transportation of the powder bag by the automatic guided robot, the tilt sensor detects that the tilt angle of the powder bag reaches 7° due to the turning inertial force, exceeding the set safety threshold of 5°. The real-time attitude information shows that the horizontal offset of the powder bag is 3 cm. Based on these data, the system triggers the dynamic adjustment mechanism, adjusts the tilt angle of the robot platform to -2° through the servo motor, and reduces the speed to 0.3 m / s. Finally, the tilt angle is restored to within 1°, and the center of gravity of the powder bag returns to the center of the platform.

[0135] Example 2: In the above example, the dynamic center-of-gravity position of the powder bag is calculated as (25, 18, 10), while the ideal center-of-gravity position of the platform is (25, 20, 10). The center-of-gravity adjustment parameter The calculation result is (0, -2, 0), indicating that the center of gravity of the powder bag is offset by 2 cm. According to the adjustment parameter instruction, the system commands the servo motor to tilt the platform 1.5° in the positive Y direction, and finally adjusts the center-of-gravity offset to 0, restoring the balance state of the powder bag.

[0136] Example 3: During the above transportation process, the robot completes the transportation according to the instructions of the path planning module. The total path length is 10 meters, including two 90° turns. The system records that the time for the powder bag to reach the target stacking point is 12 seconds, and the transfer status data includes the final attitude information (tilt angle 0°) and the center-of-gravity adjustment parameter (no offset). The stacking module uses these data to adjust the placement angle of the powder bag to ensure the smoothness and stability of the stacking.

[0137] The stacking module includes a stacking device. The stacking device places the powder bag at the target stacking position according to the transfer status data and the stacking surface feature data, and generates stacking status data for evaluating the stability of the stacking structure.

[0138] Such as Figure 4As shown, the stacking module receives transfer status data and stacking surface feature data as the basis for placement decisions. The transfer status data provides real-time attitude information of the powder bag, center of gravity adjustment parameters, and the time to reach the target position. The attitude 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 dimensions, etc. These data are used to select the most suitable stacking position for the current powder bag. For stacking position calculation, the stacking device combines the input data and calculates the best matching point between the powder bag and the stacking surface through a geometric fitting algorithm, generating placement coordinates and angle adjustment parameters. The placement coordinates ensure the precise position of the powder bag in the stacking layer, while the angle adjustment parameters are used to correct the attitude of the powder bag to match the flatness of the stacking surface.

[0139] The stacking device completes the placement operation of the powder bag through a stacking robotic arm driven by a servo motor. The stacking robotic arm precisely controls the movement trajectory of the robotic arm through the servo motor according to the calculated placement coordinates and angle adjustment parameters, and moves the powder bag to the target position. During the placement process, the stacking device real-time monitors the contact state between the powder bag and the stacking surface, and ensures that the powder bag is in full contact with the stacking surface and evenly stressed by fine-tuning the end attitude of the robotic arm. After the powder bag is safely placed at the target stacking position, the system will re-collect the state data of the stacking layer to evaluate the stacking effect.

[0140] After the placement is completed, the stacking module generates stacking state data for evaluating the stability of the stacking layer. The stacking state data includes the final position coordinates of the powder bag, the force distribution of the stacking layer, flatness score, etc. The force distribution data is collected by pressure sensors 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 state data serves as the basis for subsequent powder bag stacking operations to ensure the stability and uniformity of the entire stacking structure and avoid stacking instability caused by uneven stacking or concentrated stress.

[0141] Preferably, the stacking device of the stacking module includes a stacking robotic arm driven by a servo motor. The stacking robotic arm precisely locates the position of the powder bag by receiving the stacking surface feature data in real time. The stacking robotic arm driven by the servo motor adjusts the movement speed and placement angle of the robotic 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.

[0142] The stacked surface feature data and the coordinates of the target stacking points are the key inputs for the stacking robotic arm to perform the placement operation. The stacked surface feature data describes the geometric characteristics of the target stacking position, including flatness, surface node distribution density, available space, etc. These data are generated in real time by the vision recognition module and the sensor network, providing a detailed description of the stacked surface for the stacking robotic arm. The stacking robotic arm combines the stacked surface feature data and the target point coordinates, and uses the servo control system to calculate the angular changes of each joint and the end motion path, ensuring that the end gripper can accurately reach the target position.

[0143] The servo motor is the core driving unit of the stacking robotic arm, which realizes precise motion trajectory and end attitude adjustment through closed-loop feedback control. The servo motor dynamically adjusts the moving speed of the robotic arm according to the position of the target stacking point and the real-time feedback data. For example, when approaching the target position, the speed will gradually decrease to ensure smooth placement. After receiving the instruction, the servo motor controls the attitude of the end of the robotic arm by adjusting the torque and joint position output by the motor, so that it reaches the target position and adjusts the placement angle to match the geometric characteristics of the stacked surface.

[0144] The placement angle is a key parameter dynamically adjusted by the stacking robotic arm during the placement of the powder bag, 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 stacked surface, and the distribution of the contact points reflects the flatness of the stacked surface and the force uniformity of the powder bag. The contact points are detected in real time by the sensor network and generated in combination with the powder bag morphology data. The stacking robotic arm 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 robotic arm to tilt at a certain angle to increase the contact area and ensure uniform force distribution, preventing the powder bag from sliding or tipping over.

[0145] During the placement process, the stacking robotic arm dynamically adjusts the motion trajectory and the placement angle through a real-time feedback mechanism. The feedback data includes the joint positions of the stacking robotic arm, the real-time attitude of the powder bag, and the force distribution information of the stacked surface pressure sensor. When it is detected that there is a deviation between the actual state and the target state, the servo motor corrects the joint angle and the end position in real time according to the deviation amount, ensuring the precise placement of the powder bag.

[0146] The stacking robotic arm combines the high-precision control of the servo motor and the stacked surface feature data, and can realize the precise adjustment of the position and angle of the powder bag. By dynamically adjusting the placement angle to match the geometric characteristics of the stacked surface, it ensures that the powder bag is in full contact with the stacked surface during the stacking process, improving the stability of the stacking.

[0147] Based on the placement angle adjustment mechanism of the stacked contact point distribution, the stacking robotic arm can optimize the force distribution of the powder bags during the placement process, effectively reducing the risk of sliding or tipping caused by uneven stacking, and significantly improving the overall stability of the stacked layers. The stacking robotic arm can flexibly adapt to the changes in the stacking surface and the morphological characteristics of the powder bags by adjusting the motion trajectory and posture in real-time feedback, achieving efficient processing of complex stacking tasks.

[0148] Example 1: In a certain loading task, the stacking module received the coordinates of the target stacking point as (50, 30, 10), and the flatness score of the stacking surface feature data was 0.85. The stacking robotic arm is driven by a servo motor and moves to the stacking point according to the target coordinates. When approaching the target position, the robotic arm detects that the pressure in the left area is relatively high and the contact area in the right area is insufficient based on the contact point distribution. The system instructs the robotic arm to adjust the placement angle to a 2° right tilt, finally achieving uniform force and stable placement of the powder bag.

[0149] Example 2: During the process of Example 1, the stacking robotic arm detected the posture change of the powder bag caused by the center of gravity shift during the placement operation, with a 1° right tilt. The servo motor corrected the end position in real-time, adjusted the tilt angle of the robotic arm to a 1° left tilt to balance the posture of the powder bag, and at the same time reduced the end speed to 0.1 m / s to ensure a smooth completion of the placement process. After placement, the data of the pressure sensors on the stacking surface showed uniform contact point distribution, and the stacking stability score was 0.95.

[0150] Example 3: In a multi-layer stacking task, when the stacking robotic arm stacked the second layer of powder bags, it was found that the force on the left front area of the first-layer stacking surface was too large, resulting in an overall tilt of 1.5°. The stacking module adjusted the placement angle of the second layer of powder bags to a 1° left tilt according to the feedback data, and at the same time increased the contact area in the left rear area to balance the stacking structure. Finally, the overall stability of the multi-layer stacking was achieved.

[0151] Preferably, the stacking module includes a thermal distribution monitoring unit. The thermal distribution monitoring unit detects the force distribution on the contact surface of the stacked layer through an infrared imaging device to generate a stacked thermal distribution map; the stacked thermal distribution map is used to analyze the force uniformity of the stacked layer and generate stacked adjustment recommendation data.

[0152] The force distribution on the contact surface of the stacked layers reflects the force uniformity of the stacked structure and is an important indicator for evaluating the stacking stability. The thermal distribution monitoring unit detects the temperature distribution on the contact surface of the stacked layers through an infrared imaging device and correlates the temperature changes with the force distribution. Areas with greater force on the contact surface will generate local temperature rises 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 layers in the form of a visual thermal map. After processing the original temperature data generated by the infrared imaging device, a stacked thermal distribution map is formed. In the map, the high-temperature areas correspond to the force-concentrated areas, and the low-temperature areas correspond to the areas with less force or no force.

[0153] The thermal distribution map is a visual expression of the force distribution on the stacked layers. By analyzing the range, distribution, and uniformity of the force areas in the map, the stability of the stacked structure is evaluated. Force uniformity analysis: The system calculates the force uniformity score based on the temperature differences in each area of the thermal distribution map. 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 may indicate a risk of stacking instability. By comparing the area and temperature gradient of the high-temperature areas, the system determines whether the force is concentrated on certain specific points or areas. If the area of the high-temperature area is small and the gradient is large, it indicates that there is force concentration and the stacked layer may experience local instability; if the high-temperature areas are evenly distributed, the overall force on the stacked layer is more balanced.

[0154] Based on the analysis results of the thermal distribution map, the system generates stacked adjustment recommendation data to guide the stacking module to optimize the placement angle and position of the subsequent powder bags. The stacked adjustment recommendation data includes information such as the target placement angle of the powder bag, the placement position offset, and the preferred placement area. For example, when there is a force-concentrated area in a certain stacked 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 stacked layer and the target uniformity score to ensure that the adjustment actions can effectively improve the force distribution.

[0155] By continuously monitoring the temperature changes on the contact surface through the infrared imaging device, the thermal distribution monitoring unit can accurately capture the force distribution state of the stacked layers, 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.

[0156] The stacked thermal distribution map provides an intuitive display of the force uniformity of the stacked layers. Combining with the stacked adjustment recommendation data, the stacking module can optimize the force state of the stacked layers by adjusting the placement position and angle, effectively improving the overall stability of the stacked structure.

[0157] 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 powder bags can be dynamically adjusted to avoid the cumulative effect of stacking instability.

[0158] Example 1: In a certain 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 (5°C higher than the average temperature) in the left front area, and a low-temperature area (3°C lower than the average temperature) in the right rear area. The system analysis results show that the force is concentrated in the left front area, and the stacking uniformity score is 75%. According to the thermal distribution map, the stacking module determines that it is necessary to optimize the placement strategy of subsequent powder bags.

[0159] Example 2: On the basis of Example 1, the thermal distribution monitoring unit generates stacking adjustment suggestion 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 robotic arm performs adjustment operations according to the suggestion data and successfully places the second powder bag in the right rear area. The thermal distribution map shows that the overall force distribution uniformity score has increased to 90%.

[0160] Example 3: In a multi-layer stacking task, when the stacking module places the powder bag on the third layer, it detects through the thermal distribution monitoring unit that the right side area of the second layer stacking surface is stressed concentratedly and there is local inclination. The system generates adjustment suggestions according to the thermal distribution map, instructing the stacking robotic arm to place the third layer powder bag in the left area and adjusting the placement angle to 2° to the left. After placement, the overall stability score of the stacking structure has increased from 82% to 96%.

[0161] 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 detecting the stacking stability score data in real time. The stacking stability score data is generated by analyzing the force concentration degree and overall balance of each contact point in the stacking thermal distribution map, and generates the adjusted stacking state data.

[0162] The stacking stability score data is an evaluation index for the dynamic stability of the stacking structure, which is calculated based on the force distribution and overall balance in the stacking heat map. The force concentration degree is a key index to evaluate 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 heat map, it is judged whether there is local force concentration in the stacking layer. If the area of the high-temperature area is small and the temperature gradient is large, it indicates that the force is concentrated and the stacking structure may be unstable. The overall balance is an index to evaluate 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 indicates that there is a risk of instability in the stacking layer. The stacking stability score data is calculated by comprehensively considering the force concentration degree and overall balance, and the scoring range is from 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.

[0163] When the stacking stability score is lower than the threshold, the self-correcting unit triggers the servo motor to adjust the stacking position and angle of the powder bags to restore the stability of the stacking layer. By changing the position or placement angle of the powder bags 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 robotic arm with the servo motor. According to the low-temperature area (less force) in the stacking heat map, the powder bags are 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 tilt angle of the powder bags, 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 robotic arm to tilt the powder bag by a certain angle in the opposite direction.

[0164] After the adjustment is completed, the self-correcting unit generates the adjusted stacking state data, which is used 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, updated force distribution, and optimized stacking stability score of the powder bags. By evaluating the adjustment effect, the system can dynamically optimize the next stacking operation and gradually improve the overall stability of the stacking structure.

[0165] By real-time detecting the stacking stability score data, the self-correcting unit can quickly respond to the instability of the stacking layer and trigger adjustment actions to restore the balance of the stacking structure. This dynamic optimization ability significantly improves the safety and long-term stability of the stacking layer.

[0166] Combined with the stacking heat map and contact point distribution information, the position and angle adjustments under the control of the servo motor have high precision. The adjustment actions can not only balance the force distribution but also effectively reduce the risk of center of gravity offset, thus ensuring the neatness and stability of the stacking layer.

[0167] The self - calibration unit implements a closed - loop control process of real - time detection, dynamic adjustment, and optimization feedback during the stacking operation, making the entire stacking process more intelligent. At the same time, by reducing the additional adjustment requirements caused by stacking instability, the overall loading efficiency is improved.

[0168] Example 1: In a certain loading task, the self - calibration unit detected through the stacking thermal distribution map generated by the thermal distribution monitoring unit that the left - front area of the first - layer stacking surface had concentrated stress, and there was almost no stress in the right - rear area. The stacking stability score was 72, lower than the set threshold of 80. The self - calibration unit triggered the servo motor to control the stacking robotic arm to move the powder bag 3 cm to the right - rear area and at the same time adjusted the placement angle to a 1.5° right - tilt. After adjustment, the stacking thermal distribution map showed uniform stress distribution, and the stability score increased to 92.

[0169] Example 2: In a multi - layer stacking task, the self - calibration unit detected that the center of gravity of the second - layer stacking structure was offset by 3 cm (deviating from the central axis), and the stacking stability score was 65. According to the analysis results of the thermal distribution map, the system instructed the stacking robotic arm to place the third - layer powder bag in the area in the opposite direction of the offset and at the same time adjusted the placement angle to a 2° left - tilt to balance the overall structure. After adjustment, the stacking stability score increased to 90, and the center of gravity returned to the central axis.

[0170] Example 3: In the above example, after the stacking adjustment was completed, the self - calibration unit generated the adjusted stacking state data, including the final position (50, 30, 15) of the powder bag, the placement angle of 1.5° right - tilt, the updated stress distribution 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.

[0171] The above are only the embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within 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 three-dimensional 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 three-dimensional shape data includes the length, width, height and boundary feature data of the powder bag; The internal distribution data includes infrared thermal radiation signal data in different temperature ranges; The surface characteristic data obtains density and edge deformation characteristic data of the powder bag material; 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 through optical imaging and depth information; The infrared sensor is used to detect the distribution characteristics of the powder inside the powder bag and obtain the internal distribution data; The ultrasonic sensor obtains surface characteristic data 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 three-dimensional 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