A method and system for planning the motion trajectory of an injection molding robot

Abstract

This application discloses a motion trajectory planning method and system for injection molding robots, relating to the field of mechanical automation technology. It aims to solve the problem of cumulative wear and degradation of the mold and end effector in existing injection molding robots, which affects component quality and the long-term operational stability of the automation system. The method includes: after each part-picking cycle operation, using a microscopic imaging sensor integrated on the injection molding robot, performing non-contact microscopic state sensing of key areas of the injection mold cavity and the contact surface of the injection molding robot's end effector to obtain current microscopic state data; comparing the current microscopic state data with historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the injection molding robot's end effector; dynamically correcting the motion trajectory of the injection molding robot based on the identified and quantified microscopic physical changes; and executing the dynamically corrected motion trajectory of the injection molding robot.

Description

Technical Field

[0001] This application relates to the field of mechanical automation technology, and in particular to a method and system for planning the motion trajectory of an injection molding robot. Background Technology

[0002] In modern automated production, especially in injection molding, robotic arms play a crucial role in removing newly molded plastic parts from the mold. These robotic assistants are meticulously designed to repeat precise movements at extremely high speeds, ensuring efficient production. The accuracy of the robotic arm's movements is essential for guaranteeing product quality and the reliable operation of the entire automated system.

[0003] However, during long-term, high-intensity continuous operation, the system faces some subtle challenges that gradually affect its performance. Over time, the surfaces of the mold and the robot's end effector (the part that grips the component) undergo microscopic changes that are difficult to detect with the naked eye due to the accumulation of minute substances in the environment and continuous wear. These changes subtly affect the robot's perception of its surroundings and its interaction with the component, potentially leading to a slow decline in product quality and increased equipment maintenance requirements. Therefore, addressing these hidden degradation phenomena is crucial to ensuring the long-term stability and efficiency of automated injection molding production. Summary of the Invention

[0004] This application provides a motion trajectory planning method and system for injection molding robots, aiming to solve the problem that the motion trajectory planning of injection molding robots in the prior art fails to fully consider and actively manage the cumulative wear and degradation of molds and end effectors caused by long-term, high-frequency, and weak mechanical interactions, thereby affecting the quality of parts and the long-term operational stability of the automation system.

[0005] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a motion trajectory planning method for an injection molding robot, comprising: after each part-picking cycle operation is completed by the injection molding robot, using a microscopic imaging sensor integrated on the injection molding robot, performing non-contact microscopic state perception on the key areas of the injection mold cavity and the contact surface of the injection molding robot end effector to obtain current microscopic state data; comparing the current microscopic state data with historical state data to identify and quantify the microscopic physical changes of the injection mold cavity and the injection molding robot end effector; dynamically correcting the motion trajectory of the injection molding robot based on the identified and quantified microscopic physical changes; the dynamic correction of the motion trajectory of the injection molding robot includes increasing the safety gap between the injection molding robot end effector and the injection mold cavity or adjusting the posture of the injection molding robot end effector; and executing the dynamically corrected motion trajectory of the injection molding robot.

[0006] Secondly, this application provides a motion trajectory planning system for an injection molding robot. The system includes: a microscopic state perception module, used to perform non-contact microscopic state perception of key areas of the injection mold cavity and the contact surface of the injection molding robot's end effector after each part-picking cycle operation, using a microscopic imaging sensor integrated on the injection molding robot to obtain current microscopic state data; a microscopic change recognition and quantification module, used to compare the current microscopic state data with historical state data to identify and quantify the microscopic physical changes of the injection mold cavity and the injection molding robot's end effector; a trajectory correction module, used to dynamically correct the motion trajectory of the injection molding robot based on the identified and quantified microscopic physical changes; the dynamic correction of the injection molding robot's motion trajectory includes increasing the safety gap between the injection molding robot's end effector and the injection mold cavity or adjusting the posture of the injection molding robot's end effector; and a trajectory execution module, used to execute the dynamically corrected motion trajectory of the injection molding robot.

[0007] Through the above technical solution, this application effectively solves the problem that traditional robotic arm motion trajectory planning methods in the prior art fail to fully consider and proactively manage the cumulative wear and degradation of molds and end effectors caused by long-term, high-frequency, and weak mechanical interactions. Traditional methods mainly focus on avoiding immediate collisions and adapting to the initial geometric deviations of components, but lack the ability to perceive and respond to microscopic changes that are difficult to detect with the naked eye. This leads to slight "micro-contacts" in the robotic arm during long-term operation, which in turn cause cumulative wear and deformation of the mold and end effector, ultimately affecting product quality and production efficiency. This application, by introducing a microscopic state perception and dynamic trajectory correction mechanism, achieves "proactive" management of microscopic changes on the surface of the mold and the robotic arm end effector. This method can identify and quantify potential wear risks in advance, and increase safety clearances or optimize posture by adjusting the motion trajectory, thereby minimizing or avoiding micro-contacts, effectively extending the service life of the mold and end effector, and significantly reducing component quality problems and equipment maintenance costs caused by wear. Compared with existing technologies, this application not only ensures immediate safety and efficiency in each cycle, but also has the ability to predict and avoid long-term cumulative micro-interactions, thereby significantly improving the long-term stability of injection molding production and the reliability of product quality, demonstrating remarkable and superior technical effects. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a motion trajectory planning method for an injection molding robot provided in this application. Detailed Implementation

[0009] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0010] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0011] In modern automated production, injection molding robots play a crucial role in removing newly molded plastic parts from molds. These robots are designed to repeat precise movements at extremely high speeds to ensure efficient production. However, in actual, long-term, high-intensity production practices, the accumulation of tiny suspended particles and microscopic mechanical wear of the mold surface lead to subtle changes in the optical properties and micro-geometry of the mold, introducing perceptual biases in the vision system and resulting in extremely slight "micro-contacts" between the robot and the mold or part. The cumulative micro-wear and deformation from these micro-contacts ultimately affect production quality and efficiency, creating a negative feedback loop. Traditional robot motion trajectory planning methods have failed to adequately consider and proactively manage this cumulative wear and degradation.

[0012] In view of the above problems, this application provides a motion trajectory planning method for injection molding robots, endowing the injection molding robot with a unique "self-inspection" and "environmental perception" capability. It no longer relies solely on preset trajectories or macroscopic judgments from external vision systems, but instead utilizes its integrated microscopic imaging sensors to perform high-resolution "physical examinations" of the mold cavity and the robot's end effector surface during the intervals of each work cycle. This real-time capture of micron-level physical changes allows the robot to dynamically perceive subtle deformations and changes in optical properties caused by long-term operation, environmental particle accumulation, or minor wear, and adjust its motion trajectory accordingly. This proactively avoids potential micro-contact risks, ensuring refined management of equipment component lifespan and product quality without affecting production cycle time.

[0013] The following specific embodiments will provide a detailed introduction and explanation of the injection molding robot motion trajectory planning method and system provided in this application.

[0014] Reference Figure 1 This application provides a method for planning the motion trajectory of an injection molding robot, which may include the following steps:

[0015] S1. After each part-picking cycle operation is completed by the injection molding robot, the microscopic imaging sensor integrated in the injection molding robot is used to perform non-contact microscopic state perception on the key areas of the injection mold cavity and the contact surface of the injection molding robot end effector to obtain the current microscopic state data.

[0016] Injection molding robots are automated devices used on injection molding production lines to remove molded parts from injection molds. They typically consist of a robotic arm, an end effector, and a control system. Microscopic imaging sensors are sensors capable of acquiring images of the microscopic details of an object's surface. Their resolution is far higher than that of conventional industrial cameras, enabling them to capture surface changes that are imperceptible to the naked eye. The critical areas of the injection mold cavity refer to the surfaces inside the mold that directly contact the molded part, as well as edges, corners, and other areas that may have microscopic contact with the robot's end effector. The contact surface of the injection molding robot's end effector refers to the components used by the robot to grasp or absorb parts, such as suction cups or grippers, which frequently come into contact with the mold or part. Microscopic state data refers to images or 3D data acquired by microscopic imaging sensors that reflect the microscopic characteristics (such as roughness, deposits, wear marks, etc.) of the mold cavity and end effector surfaces.

[0017] Specifically, after each part-removal cycle, the injection molding robot uses microscopic imaging sensors integrated into the injection molding robot to perform non-contact microscopic state perception of key areas of the injection mold cavity and the contact surface of the injection molding robot's end effector in order to obtain current microscopic state data.

[0018] As one implementation method, the microscopic imaging sensor can employ a high-resolution confocal microscope to acquire high-precision three-dimensional topographic data by scanning the surfaces of the mold and the end effector. For example, after the robot arm completes the part removal action and exits the mold, the confocal microscope can scan the edges of the mold cavity, the area around the ejector pin hole, and the lip of the end effector suction cup along a preset path, generating a series of depth images. These images together constitute the current microscopic state data.

[0019] Another implementation involves using a structured light projection-based 3D scanner as the microscopic imaging sensor. This scanner projects light rays with a known pattern onto the surface and captures its deformation to calculate the surface's three-dimensional geometry. For example, after each part-removal cycle, the 3D scanner can quickly scan the mold cavity and end effector surface to obtain point cloud data, which serves as the current microscopic state data.

[0020] S2. Compare the current microscopic state data with the historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the end effector of the injection robot.

[0021] Historical state data refers to microscopic state data acquired and stored in the same way during previous production cycles, used to compare with current data to identify changes. Microscopic physical changes refer to physical alterations occurring at the microscopic level on the mold cavity and end effector surface, such as the adhesion of tiny particles, surface wear, scratches, and deformation.

[0022] As one implementation method, the currently acquired 3D topographic data can be compared point-to-point with pre-stored 3D reference data of the initial mold and end effector surfaces. Microscopic physical changes can be identified and quantified by calculating surface height deviations or roughness variations. For example, if the surface height in a certain area continues to increase, it may indicate the adhesion of tiny particles; if the surface roughness increases significantly, it may indicate wear.

[0023] Another approach is to use image processing algorithms to register and analyze the differences between the currently acquired point cloud data and historical point cloud data. This allows for the quantification of microscopic physical changes by calculating changes in the volume of local areas or changes in the surface normal direction. For example, by comparing point cloud data acquired at different times, minute collapses at the edge of a mold cavity or localized protrusions on the surface of an end effector can be detected.

[0024] S3. Dynamically correct the motion trajectory of the injection molding robot based on the identified and quantified microscopic physical changes.

[0025] Dynamically correcting the motion trajectory of an injection molding robot involves either increasing the safety clearance between the robot's end effector and the mold cavity or adjusting the posture of the robot's end effector. The safety clearance refers to the minimum distance maintained between the robot's end effector and the mold cavity to avoid physical contact. Posture refers to the position and orientation of the robot's end effector in space.

[0026] As one implementation method, if minute particles are detected adhering to a certain area of ​​the mold cavity, which may cause slight scratching when the robot arm passes through that area, the safety clearance of the robot arm's end effector when passing through that area can be increased. For example, the robot arm's trajectory in that area can be shifted upwards or outwards by 0.05 mm to ensure sufficient clearance.

[0027] Another approach is to adjust the end effector's suction cup lip if slight wear is detected, which may affect the stability of its gripping of the component. In this case, the end effector's orientation can be adjusted so that it enters the mold at a slight tilt angle. This optimizes the contact between the suction cup and the component, compensating for the impact of wear. For example, the suction cup's tilt angle can be adjusted by 0.1 degrees to ensure that the suction cup can more smoothly adhere to the component.

[0028] S4. Execute the dynamically corrected motion trajectory of the injection molding robot.

[0029] In one implementation, the corrected trajectory data is sent to the motion controller of the injection molding robot, which then drives the robot to perform the part-picking operation based on the new trajectory instructions. For example, if the trajectory is corrected to increase the safety clearance, the robot will move along a new trajectory that deviates slightly from the original path, thereby avoiding potential micro-contact with the mold cavity.

[0030] Another implementation is that if the trajectory is corrected to adjust the posture, the robot will enter the mold in a new posture to complete the gripping and removal of the parts.

[0031] The injection molding robot trajectory planning method proposed in this application utilizes microscopic imaging sensors integrated into the robot to perform non-contact microscopic state perception of key areas of the injection mold cavity and the contact surface of the robot's end effector after each part-picking cycle, acquiring current microscopic state data. This data can capture microscopic changes on the mold and end effector surfaces that are difficult to detect with traditional vision systems, such as the adhesion of tiny particles and surface wear. Subsequently, the current microscopic state data is compared with historical state data to identify and quantify these microscopic physical changes. This comparison mechanism enables the system to track and quantify the cumulative microscopic degradation process. Based on the identified and quantified microscopic physical changes, the system can dynamically correct the injection molding robot's trajectory, including increasing the safety clearance between the robot's end effector and the injection mold cavity or adjusting the posture of the end effector. This dynamic correction mechanism allows the robot to proactively adapt to microscopic changes in the mold and end effector, avoiding potential micro-contact and cumulative wear. Ultimately, the motion trajectory of the injection molding robot is dynamically corrected, thereby protecting the mold and end effector in actual production, extending their service life, and ensuring the long-term stability of product quality.

[0032] This application further proposes that, after each part-removal cycle of the injection molding robot, a microscopic imaging sensor integrated on the injection molding robot is used to perform non-contact microscopic state sensing on the key areas of the injection mold cavity and the contact surface of the injection molding robot's end effector to obtain current microscopic state data. Furthermore, the current microscopic state data is compared with historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the injection molding robot's end effector. Specifically, this includes: after each part-removal cycle, using the microscopic imaging sensor integrated on the injection molding robot, two consecutive image acquisitions are performed on the key areas of the injection mold cavity and the contact surface of the injection molding robot's end effector. A first image and a second image are obtained and used as current microscopic state data. The first image is acquired when the active illumination source inside the microscopic imaging sensor is turned on during the first acquisition, and the second image is acquired when the active illumination source is turned off during the second acquisition. The current microscopic state data is compared with historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the end effector of the injection molding robot. This includes: performing pixel-level difference operations on the first and second images to obtain a net difference image; and performing image analysis on the net difference image to identify and quantify the local high-density aggregation areas formed by tiny suspended particles to obtain microscopic physical changes.

[0033] Specifically, the two consecutive image acquisitions aim to effectively separate the real information caused by microscopic physical changes from environmental interference by comparing images under different lighting conditions. The first image was acquired with the active illumination source inside the microscopic imaging sensor on, and in this case, the image includes surface features under active illumination as well as the influence of ambient light. The second image was acquired with the active illumination source off, and in this case, the image mainly reflects ambient light and the surface's own reflective properties.

[0034] Furthermore, pixel-level difference operations are performed on the first and second images to eliminate or significantly reduce background interference such as ambient lighting, inherent surface texture, and sensor static noise, thereby highlighting image differences caused by microscopic physical changes such as tiny suspended particles. Through this difference processing, a net difference image can be obtained, which can more clearly reflect the actual microscopic physical changes.

[0035] Based on this, image analysis of the net differential images is performed to identify and quantify localized high-density aggregation areas formed by tiny suspended particles. These areas are often early signs of wear, deposits, or material degradation. By identifying and quantifying them, the degree and distribution of microscopic physical changes on the surface of injection mold cavities and injection robot end effectors can be accurately determined.

[0036] This application's solution effectively addresses the problem of traditional single-image acquisition being susceptible to interference from ambient light and surface reflection by introducing two consecutive image acquisitions, one with the active illumination source on and the other with it off. Specifically, when the active illumination source is on, the acquired first image contains detailed information about the target surface as well as the superposition effect of ambient light; while when the active illumination source is off, the acquired second image mainly reflects ambient light and the surface's own reflective characteristics. By performing pixel-level differential operations on these two images, image information directly related to microscopic physical changes revealed by active illumination can be accurately extracted, thereby minimizing background interference from ambient light, inherent surface texture, and sensor static noise. It is precisely because of this differential processing that the net differential image can more purely and accurately reflect microscopic physical changes such as tiny suspended particles on the surface of the injection mold cavity and the end effector of the injection molding robot, providing a high-quality data foundation for subsequent identification and quantification.

[0037] This application further proposes the following steps for image analysis of net differential images to identify and quantify local high-density aggregate regions formed by tiny suspended particles, thereby obtaining microscopic physical changes: performing image processing on the net differential image to obtain image data under different spectra or polarized light; fusing the image data under different spectra or polarized light to obtain a fused image; and analyzing pixel brightness, color distribution, and polarization information on the fused image to identify and quantify local high-density aggregate regions with specific optical characteristics, and determining these local high-density aggregate regions as being formed by tiny suspended particles, thereby obtaining microscopic physical changes.

[0038] Specifically, image processing techniques for net differential images to obtain image data under different spectra or polarizations refer to the re-acquisition or processing of net differential images under different wavelength ranges (e.g., visible light, near-infrared) or different polarization directions (e.g., horizontal polarization, vertical polarization, circular polarization) by adding switchable spectral or polarization filters to the front end of a microscopic imaging sensor, or by utilizing multispectral / hyperspectral imaging technology to obtain multi-channel image data. The aim is to capture the optical response characteristics of tiny suspended particles from multiple dimensions to provide richer information. The fusion processing of image data under different spectra or polarizations to obtain a fused image can be understood as using image fusion algorithms, such as pixel-level fusion, feature-level fusion, or decision-level fusion, to integrate image data acquired from different spectral or polarization channels. The fusion processing aims to comprehensively utilize the advantageous information of each channel, eliminate redundancy, enhance target features, and thus generate a fused image containing more comprehensive optical information. In practical applications, pixel brightness, color distribution, and polarization information are analyzed in fused images to identify and quantify localized high-density clusters with specific optical characteristics. These clusters are then identified as being formed by tiny suspended particles, revealing microscopic physical changes. Specifically, this involves extracting the brightness value, color components (such as RGB or HSV), polarization degree, and polarization angle from each pixel or region in the fused image. By setting specific thresholds or employing machine learning algorithms, regions exhibiting brightness, color, or polarization characteristics consistent with tiny suspended particles are identified. For example, tiny metal particles may exhibit high reflectivity under specific polarized light, while plastic particles may display a specific color under a specific spectrum. By comprehensively analyzing this multidimensional information, localized high-density clusters formed by tiny suspended particles can be identified more accurately and quantified, such as by calculating their area, density, or quantity.

[0039] Through the above technical solution, this application can more accurately and reliably identify and quantify localized high-density aggregation areas formed by tiny suspended particles on the surface of injection mold cavities and injection robot end effectors. Compared to methods relying solely on single image analysis, this solution, by introducing multispectral or polarized light information and performing fusion processing, greatly enriches the feature dimensions available for analysis, thereby effectively avoiding misjudgments or missed detections caused by factors such as changes in illumination, material reflection, or background noise. This refined ability to identify microscopic physical changes provides more accurate data support for subsequent dynamic correction of the injection robot's motion trajectory, ensuring the effectiveness and safety of trajectory correction, further reducing the risk of damage to the mold and robot end effector due to wear, and extending the equipment's service life.

[0040] In some preferred embodiments, when performing image processing on the net difference image, a switchable filter array equipped with a microscopic imaging sensor can be utilized. For example, in the first processing, the sensor switches to a 530nm narrowband filter to acquire image data under the green spectrum; in the second processing, it switches to a 650nm narrowband filter to acquire image data under the red spectrum; and in the third processing, it switches to a polarization filter to acquire image data under a specific polarization direction. Subsequently, these image data under different spectra and polarizations are input to an image fusion module, which can use a wavelet transform fusion algorithm to fuse these multi-channel images into a high-information fused image. On the fused image, by analyzing the RGB value and polarization degree of each pixel, for example, when the red component of a pixel is higher than a certain threshold and the polarization degree is within a specific range, it is marked as a potential region of tiny suspended particles. Further, isolated noise points are removed through morphological operations (such as opening and closing operations), and the area and density of the remaining connected regions are calculated, thereby identifying and quantifying local high-density clusters formed by tiny suspended particles. For example, when the area of ​​a certain region exceeds a preset threshold and the pixel density reaches a certain level, it is identified as a pre-wearing area that needs attention.

[0041] To improve the accuracy of identifying microscopic physical changes and the targeted nature of trajectory correction, the method further includes: acquiring the temperature, humidity, and batch information of the current production environment and the current injection molding raw material; querying a particle cluster wear effect correction factor corresponding to the current environmental conditions and raw material batch from a preset parameter table based on the temperature, humidity, and injection molding raw material batch information; applying the correction factor to a preset threshold to obtain an adjusted threshold; and comparing the information of local high-density aggregation areas with the adjusted threshold to mark wear acceleration hotspots.

[0042] Specifically, environmental sensors (such as temperature and humidity sensors) integrated into the injection molding production line collect real-time data on the ambient temperature and relative humidity of the injection molding workshop. Simultaneously, by reading records from the injection molding machine control system or material management system, the specific batch information of the injection molding raw materials currently in use is obtained. This information is a key external variable affecting the formation of microparticle clusters and wear effects.

[0043] The preset parameter table can be understood as a database of empirical or experimental correction factors storing the wear effects of microparticle clusters on molds or robotic end effectors under different environmental conditions (temperature and humidity ranges) and different batches of injection molding materials (e.g., different grades, different suppliers, and different production dates). This parameter table can be constructed and continuously updated through the accumulation of a large amount of experimental data, analysis of historical production data, or expert experience. For example, some highly filled materials are more likely to generate abrasive microparticles in high-temperature and high-humidity environments, and their corresponding correction factors will be higher.

[0044] The particulate cluster wear effect correction factor is a coefficient used to adjust the wear risk assessment threshold. Its value reflects the potential impact of tiny suspended particles forming localized high-density aggregation areas on the wear of molds or robotic end effectors under specific environmental and material conditions. For example, when the ambient temperature is high or the raw material batch contains more abrasive additives, the correction factor may be greater than 1, indicating an increased wear risk; conversely, it may be less than 1.

[0045] A preset threshold is an initial standard used to determine whether a localized high-density cluster poses a risk of wear, without considering environmental and material factors. This threshold is usually set based on general experience or experimental data under standard operating conditions.

[0046] The adjusted threshold is obtained by applying a correction factor to a preset threshold. The purpose is to allow the wear risk assessment criteria to dynamically adapt to changes in the actual production environment and raw materials. For example, if the correction factor is 1.2 and the preset threshold is 100 pixels, the adjusted threshold is 120 pixels. This means that under the current conditions, a larger local high-density cluster area is needed to be marked as a wear acceleration hotspot, or vice versa.

[0047] Marking wear-accelerating hotspots involves identifying and quantifying locally high-density clusters, then using an adjusted threshold to highlight areas exceeding that threshold. These areas are identified as requiring special attention and potentially leading to accelerated wear. These hotspots will serve as crucial data for subsequent trajectory re-optimization or maintenance decisions.

[0048] By employing the aforementioned technical solution, this application overcomes the limitation of traditional methods that rely on a fixed wear risk assessment threshold, effectively preventing misjudgments or omissions caused by changes in the environment or raw materials. Specifically, by dynamically adjusting the threshold, areas with genuine accelerated wear risk under specific production conditions can be identified more accurately, making subsequent corrections to the injection molding robot's motion trajectory more targeted and effective. This not only avoids unnecessary trajectory adjustments and improves production efficiency but also enables the timely detection of potential severe wear risks, extending the service life of molds and robot end effectors, and reducing maintenance costs.

[0049] As a specific implementation, suppose that during a certain injection molding cycle, a localized high-density cluster area of ​​150 pixels is identified through image analysis. The system first obtains the current workshop ambient temperature as 35℃ and humidity as 80%, and identifies the batch of injection molding raw material being used as "high glass fiber reinforced PA66, batch number 20230101". Based on this information, the system queries a preset parameter table and finds that, under an environment of "30-40℃, 70-90% humidity" and using "high glass fiber reinforced PA66" raw material, the wear effect correction factor for microparticle clusters is 1.3. If the preset general threshold is 120 pixels, then after adjusting the correction factor, the new threshold will become 120 * 1.3 = 156 pixels. Since the size of the currently identified localized high-density cluster area (150 pixels) is smaller than the adjusted threshold (156 pixels), this area will not be marked as a wear acceleration hotspot, avoiding unnecessary trajectory correction. Conversely, if the area contains 160 pixels, it will be marked as a wear acceleration hotspot, triggering a subsequent trajectory re-optimization process. In this way, the system can dynamically adjust the wear risk assessment criteria based on actual operating conditions, making the identification of wear acceleration hotspots more accurate.

[0050] In some embodiments, the above method further includes: continuously monitoring the microscopic state changes of the wear acceleration hotspot region; when any change in the size, shape, or risk level of the wear acceleration hotspot region is detected, initiating a trajectory re-optimization process, calculating a dynamic avoidance area based on the geometric information and risk level of the current wear acceleration hotspot region; using the dynamic avoidance area as a constraint, performing local path replanning on the original motion trajectory of the injection molding robot, and dynamically adjusting the speed of the injection molding robot when passing through the wear acceleration hotspot region according to the risk level and changing trend of the wear acceleration hotspot region.

[0051] Specifically, continuous monitoring of the microscopic changes in wear-accelerating hotspots refers to the uninterrupted or pre-set frequency of microscopic state sensing and data analysis of these specific areas after the initial identification and marking of wear-accelerating hotspots. This monitoring process can utilize the same microscopic imaging sensors as the initial sensing to obtain the latest microscopic state data of the wear-accelerating hotspots. Changes in microscopic state may include, but are not limited to, an increase in the area and depth of the wear region, changes in surface roughness, or alterations in particle aggregation density.

[0052] When a change is detected in any of the dimensions, shape, or risk level of a wear-accelerated hotspot, the system will automatically initiate a trajectory re-optimization process. Dimensional changes can refer to increases or decreases in the length, width, or area of ​​the wear region; shape changes can refer to alterations in the geometric contour of the wear region, such as changing from a circle to an irregular shape; and risk level changes refer to a reassessment of the hazard level of the hotspot based on the severity of wear, its rate of development, and the assessment of its potential impact on the injection molding process. Once these changes are detected, the system will calculate a dynamic avoidance zone based on the latest geometric information of the current wear-accelerated hotspot (such as precise boundaries, center position, and dimensions) and the updated risk level. This dynamic avoidance zone is a spatially constructed safety buffer zone surrounding the wear-accelerated hotspot, and its size and shape are adjusted in real time according to the actual changes in the hotspot.

[0053] Furthermore, using the dynamic avoidance zone as a constraint, the original motion trajectory of the injection molding robot is locally replanned. This means that when the robot performs a part-picking operation, its motion path will be modified to ensure that the robot's end effector does not enter or approach the dynamic avoidance zone in an unsafe manner. Local path replanning can be implemented using various path planning algorithms, such as those based on RRT (Fast Random Tree), PRM (Probabilistic Path Map), or A* algorithm, to maintain the efficiency of the original trajectory as much as possible while satisfying the avoidance constraint. In addition, the speed of the injection molding robot when passing through the wear acceleration hotspot is dynamically adjusted according to the risk level and trend of the wear acceleration hotspot. For example, if the risk level is high or the wear trend is accelerating, the speed of the robot passing through the area may be reduced to increase reaction time or reduce potential impact; conversely, if the risk level decreases or the wear trend slows down, the speed may be appropriately increased to optimize production efficiency.

[0054] Through the above technical solution, this application enables dynamic and adaptive management of the motion trajectory of the injection molding robot. Compared with solutions that only perform one-time trajectory correction, this application can effectively address the dynamic evolution of microscopic wear on the injection mold cavity and the robot's end effector surface, significantly improving the safety and reliability of the injection molding robot during long-term operation. Specifically, the continuous monitoring mechanism ensures real-time updates of wear information, avoiding potential risks caused by information lag; the trajectory re-optimization process allows the robot to precisely adjust its motion path and speed according to the latest wear state, thereby minimizing the possibility of collisions or friction between the robot and the mold cavity, extending the service life of the mold and the robot's end effector, and ultimately improving the overall stability and economic benefits of the injection molding production line.

[0055] In some preferred embodiments, it is assumed that after a certain injection molding cycle, the system identifies and marks a wear acceleration hotspot with a diameter of approximately 0.5 mm in a certain edge area of ​​the injection mold cavity using the method described above. Based on this, the motion trajectory of the injection molding robot is initially corrected, increasing the safety clearance in that area. However, in subsequent continuous production processes, through continuous monitoring, the system detects after the 1000th cycle that the size of the wear acceleration hotspot area has expanded to approximately 0.8 mm in diameter, and its shape has become more irregular, while its risk level has also increased from "medium" to "high".

[0056] At this point, the system immediately initiates a trajectory re-optimization process. First, based on the updated geometry of the 0.8 mm diameter and irregular shape, and the "higher" risk level, the system recalculates a larger, more closely shaped dynamic avoidance zone that better fits the current wear area. For example, this avoidance zone might be an elliptical space extending 2 mm around the wear area. Subsequently, the injection molding robot's control system uses this new elliptical avoidance zone as a constraint to locally replan the portion of the robot's original trajectory that passes through this area. For example, the robot might be guided to bypass the area with a wider arc. Simultaneously, because the risk level has been raised to "higher," the system dynamically adjusts the robot's speed when passing through this wear acceleration hotspot, reducing it from 0.5 m / s to 0.3 m / s to further reduce potential collision risks and provide the system with more reaction time. Through this continuous monitoring and dynamic adjustment, even as wear continues to develop, the injection molding robot can always maintain a safe and efficient operating state.

[0057] In some embodiments, this application further proposes that the above-mentioned continuous monitoring of the micro-state changes of the wear-accelerating hotspot area includes: after each part-picking cycle of the injection molding robot, using a micro-imaging sensor integrated on the injection molding robot, multi-angle and multi-distance image acquisition is performed on the key area of ​​the injection mold cavity and the contact surface of the injection molding robot end effector to obtain multi-view image data; three-dimensional reconstruction is performed on the multi-view image data to obtain three-dimensional point cloud data of the injection mold cavity and the surface of the injection molding robot end effector; surface normal calculation and curvature analysis are performed on the three-dimensional point cloud data to identify and remove occluded or distorted areas caused by pits, scratches or foreign matter attachment, and the three-dimensional point cloud data after removing occluded or distorted areas is meshed to obtain comprehensive and distortion-free micro-state data, and the micro-state data reflects the micro-state changes of the continuously monitored wear-accelerating hotspot area.

[0058] Specifically, multi-angle, multi-distance image acquisition refers to acquiring a series of images from different perspectives and distances by adjusting the relative position and angle between the microscopic imaging sensor and the surface being measured. Its purpose is to capture the complete geometric information and texture details of the surface being measured, avoiding blind spots and information gaps that may exist in a single perspective, and providing a rich data foundation for subsequent 3D reconstruction.

[0059] Among them, 3D reconstruction converts these multi-view image data into point cloud data in three-dimensional space. Point cloud data consists of a series of points with three-dimensional coordinates, which can accurately represent the geometry of the injection mold cavity and the surface of the injection robot end effector.

[0060] In practical applications, surface normal calculation and curvature analysis are crucial steps in evaluating the local geometric features of point cloud data. A surface normal is the normal vector at a point on a surface, reflecting the orientation of the surface at that point; curvature describes the degree of bending of the surface at that point. By analyzing these geometric features, smooth wear areas can be distinguished from local abrupt changes or irregularities caused by pits, scratches, or foreign object attachments. For example, pits and scratches typically cause significant changes in local curvature, while foreign object attachments may create anomalies in the normal direction. Identifying and removing these occluded or distorted areas aims to cleanse the point cloud data and ensure the accuracy of subsequent analyses.

[0061] Furthermore, meshing the 3D point cloud data after removing occluded or distorted regions involves converting the discrete point cloud data into a structured mesh model. A mesh model typically consists of vertices, edges, and faces, facilitating the calculation and analysis of parameters such as surface area, volume, and shape changes, thereby obtaining comprehensive and distortion-free microscopic state data. This processed microscopic state data can more realistically and accurately reflect the microscopic state changes in wear-accelerated hotspot areas, providing a reliable basis for subsequent trajectory re-optimization.

[0062] Through the above technical solution, this application provides a more accurate and robust method for monitoring the microscopic condition of wear-accelerated hotspot areas. Compared to monitoring based solely on two-dimensional images or unfiltered three-dimensional data, this solution effectively avoids misjudgments and omissions by eliminating areas obstructed or distorted by pits, scratches, or foreign matter attachments, ensuring that the acquired microscopic condition data truly reflects the actual wear situation. Therefore, the size, shape, and risk level of wear-accelerated hotspots can be more accurately assessed, providing high-quality input for the re-optimization of the injection molding robot's motion trajectory. This further improves the accuracy of trajectory correction, effectively extends the service life of the injection mold and the robot's end effector, reduces maintenance costs, and increases production efficiency.

[0063] In some preferred embodiments, it is assumed that a critical area within the injection mold cavity is marked as a wear acceleration hotspot. While continuously monitoring this hotspot area, a microscopic imaging sensor acquires images of the area from multiple angles and distances, such as perpendicular to the surface, tilted at 30 degrees, tilted at 60 degrees, etc., and at different focal lengths. These multi-view images are then used to construct a 3D point cloud model of the area. In the point cloud model, if a tiny scratch or an attached dust particle exists, its local normal vector and curvature values ​​will differ significantly from the surrounding smooth wear area. For example, the curvature of the scratch edge will increase sharply, while dust particles may cause a sudden deviation in the local normal direction. By using a preset threshold, the system can automatically identify and remove these abnormal points caused by scratches or dust, thus obtaining a pure 3D point cloud data containing only actual wear characteristics. Subsequently, this pure point cloud data is meshed to generate an accurate surface model that accurately reflects the true geometric changes of the wear acceleration hotspot area, such as wear depth and area expansion, providing reliable geometric information and risk assessment basis for subsequent trajectory re-optimization.

[0064] In some embodiments, this application further proposes the above-mentioned surface normal calculation and curvature analysis of three-dimensional point cloud data to identify and remove occlusion or distortion regions caused by pits, scratches, or foreign object attachment, including: obtaining the normal vector and curvature value of each point in the three-dimensional point cloud data; calculating the mean angle and curvature variance between the normal vectors of each point in the three-dimensional point cloud data and its local neighborhood points based on the normal vector and curvature value; identifying regions where the mean angle of the normal vector exceeds a threshold and the curvature variance is lower than a threshold as micro-wear feature regions; identifying regions where the mean angle of the normal vector exceeds a threshold and the curvature variance exceeds a threshold as non-wear surface defect regions; and removing the identified non-wear surface defect regions from the three-dimensional point cloud data as occlusion or distortion regions caused by pits, scratches, or foreign object attachment.

[0065] Specifically, obtaining the normal vector and curvature value of each point in a 3D point cloud can be achieved through point cloud processing algorithms. For example, the K-Nearest Neighbor (K-NN) search algorithm can be used to determine the local neighborhood of each point, and the surface normal vector within that neighborhood can be calculated using Principal Component Analysis (PCA). Simultaneously, the curvature value is calculated to characterize the degree of curvature of the local surface. The purpose is to provide basic geometric information for subsequent feature analysis. Calculating the mean angle and curvature variance between each point in the 3D point cloud and the normal vectors of its local neighbors, based on the normal vector and curvature value, can be understood as a further quantification of the local surface geometric features. The mean angle reflects the consistency or degree of variation in the local surface orientation, while the curvature variance reflects the fluctuation of the local surface curvature. The purpose is to distinguish different types of surface features using these statistics. In practical applications, regions where the mean angle exceeds a threshold and the curvature variance is below a threshold are identified as areas with minor wear characteristics. This indicates that the surface normal direction in this region exhibits significant changes (such as at edges or with minor depressions), but its local curvature remains relatively stable, which is typically a characteristic of initial wear or material removal. For example, when slight material peeling or minute pits appear on the surface of a mold cavity, its normal vector will undergo local deflection, but the overall curvature change may not be drastic. Furthermore, regions where the mean angle between normal vectors exceeds a threshold value and the curvature variance exceeds a threshold value are identified as non-wear surface defect regions. This indicates that these regions not only show significant changes in the surface normal direction but also exhibit dramatic fluctuations in local curvature, which is more consistent with the characteristics of non-wear, structural defects such as scratches, deep pits, or foreign matter attachment. For example, a sharp scratch or a particle attached to the surface can simultaneously cause dramatic changes in the normal vector and significant fluctuations in curvature. Therefore, the identified non-wear surface defect regions are treated as occlusion or distortion regions caused by pits, scratches, or foreign matter attachment and are removed from the 3D point cloud data. The purpose is to ensure that subsequent wear monitoring and trajectory planning are based solely on real and valid surface condition data, avoiding interference introduced by non-wear defects.

[0066] Through the above technical solution, this application enables more accurate identification and analysis of the microscopic state of the injection mold cavity and the surface of the injection molding robot end effector. This solution, by introducing the mean of the included angle of the normal vector and the variance of curvature as discrimination criteria, significantly improves the ability to distinguish between minute wear features and non-wear surface defects. Therefore, it effectively avoids misjudging actual early wear signs as obstructed or distorted areas that need to be removed, thus ensuring the accuracy and sensitivity of wear monitoring. This refined data processing method provides more reliable input for subsequent wear acceleration hotspot marking and trajectory re-optimization, thereby improving the overall accuracy and safety of the injection molding robot's motion trajectory planning and effectively extending the service life of the mold and the robot.

[0067] In some preferred embodiments, a specific example is illustrated below. Suppose that 3D point cloud data of a key area of ​​an injection mold cavity is obtained through multi-angle, multi-distance image acquisition and 3D reconstruction. To identify and remove occluded or distorted areas, the normal vector and curvature value are first calculated for each point in the point cloud data. For example, a local neighborhood radius can be set, and PCA analysis can be performed on each point within its neighborhood to obtain the normal vector, while simultaneously calculating its Gaussian curvature or mean curvature. Next, based on these normal vectors and curvature values, the mean angle between each point and the normal vectors of points in its local neighborhood and the mean curvature variance are calculated. For example, for point P, the mean angle between its normal vector and the normal vectors of all other points in its neighborhood, as well as the variance of the curvature values ​​of all points in the neighborhood, can be calculated. Subsequently, thresholds for the mean angle of the normal vectors (e.g., 15 degrees) and the mean curvature variance (e.g., 0.05) are set. The system will traverse all points in the point cloud and classify them according to the calculated statistics. If the mean angle of the normal vector at a point exceeds 15 degrees, but its curvature variance is less than 0.05, the region containing that point is identified as a minor wear feature area, such as a slight dent or material spalling. These areas will be retained for subsequent wear monitoring. Conversely, if the mean angle of the normal vector at a point exceeds 15 degrees, and its curvature variance also exceeds 0.05, the region containing that point is identified as a non-wear surface defect area, such as a sharp scratch or a foreign particle attached to the surface. These areas are explicitly identified as occlusion or distortion areas caused by dents, scratches, or foreign object attachment and are removed from the 3D point cloud data. In this way, the final 3D point cloud data will be cleaner, containing only the true surface geometry and the wear features that need to be monitored, thus providing a more accurate basis for the motion trajectory planning of the injection molding robot.

[0068] In some embodiments, the above-mentioned surface normal calculation and curvature analysis of the 3D point cloud data to identify and remove occlusion or distortion regions caused by pits, scratches, or foreign object attachments specifically includes: obtaining density information of each point in the 3D point cloud data; identifying low-density regions in the 3D point cloud data based on the density information; performing connectivity analysis on the low-density regions to identify isolated points or sparse point clusters in the low-density regions; obtaining the normal vector and curvature value of the isolated points or sparse point clusters; comparing the variation range of the normal vector and the fluctuation range of the curvature value of the isolated points or sparse point clusters with a preset threshold; identifying isolated points or sparse point clusters whose variation range of the normal vector and the fluctuation range of the curvature value are both within the preset threshold as low-density noise points caused by non-solid attachments; and removing the identified low-density noise points caused by non-solid attachments from the 3D point cloud data as occlusion or distortion regions caused by pits, scratches, or foreign object attachments.

[0069] Specifically, obtaining the density information of each point in 3D point cloud data refers to quantifying its density by calculating the number of points within its local neighborhood. For example, the K-Nearest Neighbors (K-NN) algorithm or radius search algorithm can be used to determine the point density within a certain distance or a certain number of neighboring points around each point. Low-density regions can be understood as areas with point density below a preset threshold; these regions may correspond to surface defects, noise points, or sparse foreign object attachments. Further, connectivity analysis is performed on low-density regions to cluster discrete low-density points into meaningful regions or clusters. This can be achieved through clustering algorithms (such as DBSCAN, K-means, etc.) or graph-based connectivity analysis to identify isolated points or sparse point clusters. An isolated point is a single point with almost no other points in its local neighborhood, while a sparse point cluster is a set of points with a density far lower than the surrounding normal area.

[0070] After identifying isolated points or sparse point clusters, it is necessary to obtain the normal vectors and curvature values ​​of these points. The normal vector reflects the orientation of the surface where the point is located, while the curvature value describes the degree of curvature of the surface at that point. These geometric features help to further distinguish different types of low-density regions. Subsequently, the range of variation of the normal vectors and the range of fluctuation of the curvature values ​​of isolated points or sparse point clusters are compared with preset thresholds. The range of variation of the normal vectors refers to the degree of difference between the normal vectors of each point within the isolated point or sparse point cluster; the range of fluctuation of the curvature values ​​refers to the degree of dispersion of the curvature values ​​of these points. The preset thresholds are determined based on experience or experiments and are used to distinguish between normal surface features and anomalous features. Thus, isolated points or sparse point clusters whose range of variation of the normal vectors and the range of fluctuation of the curvature values ​​are both within the preset thresholds are identified as low-density noise points caused by non-solid attachments. This means that although these points have low density, their local geometric features (normals and curvature) are relatively stable, and they do not exhibit obvious structural defects (such as pits or scratches). This is more consistent with the characteristics of low-density noise points caused by sensor noise, environmental interference, or very small, non-structural foreign objects (such as dust particles). Finally, the identified low-density noise points caused by non-solid attachments are treated as occlusion or distortion areas caused by pits, scratches, or foreign object attachments and are removed from the 3D point cloud data. This step ensures the accuracy of subsequent data processing and avoids interference from these noise points in surface reconstruction and microstructure analysis.

[0071] Through the above technical solution, this application can more accurately identify and remove low-density noise points or sparse point clusters caused by non-solid attachments in 3D point cloud data. Compared with methods that rely solely on surface normals and curvature analysis, this solution shows significant advantages in handling subtle, unstructured distortions, effectively avoiding the interference of these noise points on subsequent 3D reconstruction and microscopic state data analysis. Therefore, more comprehensive and distortion-free microscopic state data can be obtained, significantly improving the accuracy and reliability of surface state monitoring of injection mold cavities and injection molding robot end effectors, providing a more solid data foundation for the dynamic correction of injection molding robot motion trajectories.

[0072] In some preferred embodiments, after acquiring the 3D point cloud data of the injection mold cavity and the surface of the injection robot end effector, the density of each point is first calculated. For example, a search radius R can be set, and the number of neighboring points within this radius can be counted. If the number of neighboring points of a point is lower than a preset threshold N (e.g., N=5), it is marked as a low-density point. Subsequently, DBSCAN clustering analysis is performed on all low-density points to form clusters of connected low-density points, and isolated points (i.e., clusters containing only one point) and sparse point clusters (clusters containing a small number of points) are identified. For these isolated points and sparse point clusters, the normal vector and curvature value of each point within them are calculated, and the average angle variation range of these normal vectors and the variance of the curvature values ​​are further calculated. Assuming the preset threshold for the normal vector variation range is 5 degrees and the threshold for the curvature fluctuation range is 0.01, if the average angle variation range of the normal vector of an isolated point or sparse point cluster is less than 5 degrees and the curvature variance is less than 0.01, it is determined to be a low-density noise point caused by non-solid attachments. For example, a cluster of sparse points, where the normal directions of the internal points are almost identical and the surface curvature varies little, is likely not a real surface defect, but rather caused by tiny dust particles or transient sensor errors. These identified noise points are then removed from the original 3D point cloud data to ensure the accuracy of subsequent meshing and microscopic state analysis.

[0073] In some embodiments, this application further proposes to simultaneously acquire the posture sensor data of the injection molding robot itself when acquiring the three-dimensional point cloud data; to perform posture correction on the three-dimensional point cloud data based on the posture sensor data; and to perform surface normal calculation and curvature analysis on the corrected three-dimensional point cloud data in order to identify and remove occluded or distorted areas caused by pits, scratches or foreign objects.

[0074] Specifically, synchronously acquiring the injection molding robot's own attitude sensor data refers to simultaneously using attitude sensors integrated on the injection molding robot body (such as inertial measurement units, encoders, or laser trackers) to record the robot's position and attitude information in space in real time, while microscopic imaging sensors acquire multi-view image data to generate 3D point cloud data. This attitude sensor data can include the angles of each joint of the robot, the 3D coordinates of the end effector, and its attitude in space (such as pitch, yaw, and roll angles). The purpose is to provide accurate reference coordinate system information for subsequent 3D point cloud data.

[0075] In this context, attitude correction of 3D point cloud data based on attitude sensor data can be understood as performing geometric transformations or coordinate system conversions on the original 3D point cloud data using synchronously acquired attitude sensor data. For example, each point cloud data point can be transformed from the coordinate system of the robot's end effector to the global coordinate system, eliminating point cloud data offsets or rotations caused by changes in robot posture or minor vibrations. Specifically, based on the rotation matrix and translation vector provided by the attitude sensor, each point in the 3D point cloud can be inversely transformed, mapping it to a unified and stable reference coordinate system, thereby eliminating geometric distortions introduced by robot posture instability or measurement errors. The aim is to improve the intrinsic accuracy and consistency of the 3D point cloud data, making it more accurately reflect the true microscopic state of the injection mold cavity and the surface of the injection robot's end effector.

[0076] In practical applications, performing surface normal calculations and curvature analysis on the corrected 3D point cloud data to identify and remove occluded or distorted areas caused by pits, scratches, or foreign object attachments refers to performing the same surface normal calculation and curvature analysis steps as described in the above embodiments on the more accurate and stable 3D point cloud data after attitude correction. Since the input 3D point cloud data has eliminated the influence of robot arm attitude errors, subsequent analysis will more accurately identify surface anomalies caused by real physical defects (such as pits, scratches) or foreign object attachments, rather than artifacts caused by measurement errors. The purpose is to ensure that the identified and removed occluded or distorted areas are real, thereby obtaining more reliable microscopic state data.

[0077] Through the above technical solution, this application can significantly improve the accuracy and reliability of 3D point cloud data. By eliminating the influence of the injection molding robot's own posture changes on data acquisition, the obtained microscopic state data is more realistic and accurate, thus enabling more precise identification and quantification of microscopic physical changes in the injection mold cavity and the surface of the injection molding robot's end effector. This high-precision microscopic state data provides a more reliable basis for subsequent trajectory correction, effectively avoiding misjudgments or over-corrections caused by inaccurate data, thereby improving the overall robustness and safety of the injection molding robot's motion trajectory planning and extending the service life of the mold and end effector.

[0078] In some preferred embodiments, the injection molding robot can be equipped with high-precision six-axis force / torque sensors and / or inertial measurement units (IMUs). While the microscopic imaging sensors acquire images at different angles and distances to construct 3D point cloud data, these attitude sensors simultaneously record the real-time position and attitude information of the robot's end effector. For example, the IMU can provide real-time angular velocity and acceleration data, and the robot's attitude can be accurately estimated using data fusion algorithms (such as Kalman filtering). After acquiring the raw 3D point cloud data, a transformation matrix can be constructed using this synchronized attitude data to accurately transform each point cloud data from its local coordinate system at the time of acquisition to a predefined global reference coordinate system. For example, if the robot experiences a slight pitch or yaw during acquisition, the attitude correction algorithm calculates the corresponding rotation and translation based on the sensor data and applies it inversely to the point cloud data, thereby eliminating these geometric distortions caused by the robot's attitude. The corrected 3D point cloud data will exhibit higher geometric consistency and lower noise levels, enabling subsequent surface normal calculations and curvature analysis to more accurately distinguish between real surface defects (such as tiny pits or scratches) and measurement errors. For example, an area that might have been misjudged as a surface protrusion may be accurately identified as a tiny foreign object attached to a flat surface after attitude correction, thus avoiding unnecessary trajectory adjustments or more precisely guiding avoidance strategies.

[0079] This application also provides a motion trajectory planning system for an injection molding robot, comprising: a microscopic state perception module, used to perform non-contact microscopic state perception of key areas of the injection mold cavity and the contact surface of the injection molding robot end effector after each part-picking cycle operation, using a microscopic imaging sensor integrated on the injection molding robot to obtain current microscopic state data; a microscopic change recognition and quantification module, used to compare the current microscopic state data with historical state data to identify and quantify the microscopic physical changes of the injection mold cavity and the injection molding robot end effector; a trajectory correction module, used to dynamically correct the motion trajectory of the injection molding robot based on the identified and quantified microscopic physical changes; the dynamic correction of the injection molding robot's motion trajectory includes increasing the safety gap between the injection molding robot end effector and the injection mold cavity or adjusting the posture of the injection molding robot end effector; and a trajectory execution module, used to execute the dynamically corrected motion trajectory of the injection molding robot.

[0080] The injection molding robot motion trajectory planning system proposed in this application integrates functional modules such as microscopic state perception, microscopic change identification and quantification, trajectory correction, and trajectory execution. This constructs an intelligent system capable of real-time monitoring and proactively adapting to microscopic changes in the injection mold cavity and the robot's end effector. The system aims to solve the problem of cumulative microscopic damage caused by microparticle adhesion and long-term mechanical wear, which traditional methods cannot effectively address. By dynamically adjusting the robot's motion trajectory, it effectively avoids potential micro-contacts, thereby extending the service life of the mold and end effector and ensuring long-term product quality stability.

[0081] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for planning the motion trajectory of an injection molding robot, characterized in that, The method includes: After each part-picking cycle operation is completed, the injection molding robot uses microscopic imaging sensors integrated on the injection molding robot to perform non-contact microscopic state perception of key areas of the injection mold cavity and the contact surface of the injection molding robot end effector in order to obtain current microscopic state data. The current microscopic state data is compared with historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the end effector of the injection robot. The motion trajectory of the injection molding robot is dynamically corrected based on the identified and quantified microscopic physical changes. The dynamic correction of the motion trajectory of the injection molding robot includes increasing the safety clearance between the end effector of the injection molding robot and the cavity of the injection mold or adjusting the posture of the end effector of the injection molding robot. The dynamically corrected motion trajectory of the injection molding robot is executed. After each part-removal cycle, the injection molding robot uses microscopic imaging sensors integrated into the robot to perform non-contact microscopic state sensing of key areas of the injection mold cavity and the contact surface of the robot's end effector, in order to obtain current microscopic state data, including: After each part-removal cycle, the injection molding robot uses a microscopic imaging sensor integrated into the robot to perform two consecutive image acquisitions on the key areas of the injection mold cavity and the contact surface of the robot's end effector, obtaining a first image and a second image. The first image and the second image are used as the current microscopic state data. The first image is acquired when the active illumination source inside the microscopic imaging sensor is turned on during the first acquisition, and the second image is acquired when the active illumination source is turned off during the second acquisition. The step of comparing the current microscopic state data with historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the end effector of the injection molding robot includes: Perform pixel-level difference operations on the first image and the second image to obtain a net difference image; Image analysis is performed on the net difference image to identify and quantify the local high-density aggregation regions formed by tiny suspended particles, thereby obtaining the microscopic physical changes.

2. The method for planning the motion trajectory of an injection molding robot according to claim 1, characterized in that, The step of performing image analysis on the net difference image to identify and quantify the local high-density aggregation regions formed by tiny suspended particles, and obtaining the microscopic physical changes, includes: The net difference image is processed using image technology to obtain image data under different spectra or polarized light. The image data under different spectra or polarized light are fused to obtain a fused image; On the fused image, pixel brightness, color distribution, and polarization information are analyzed to identify and quantify local high-density clustered regions with specific optical characteristics. These local high-density clustered regions with specific optical characteristics are identified as local high-density clustered regions formed by tiny suspended particles, thus obtaining the microscopic physical changes.

3. The method for planning the motion trajectory of an injection molding robot according to claim 2, characterized in that, The method further includes: Obtain the current production environment's temperature and humidity, as well as the batch information of the current injection molding raw materials; Based on the temperature, humidity, and injection molding raw material batch information, the particle cluster wear effect correction factor corresponding to the current environmental conditions and raw material batch is queried from the preset parameter table. The correction factor is applied to a preset threshold to obtain the adjusted threshold; and The information on local high-density clustered areas is compared with the adjusted threshold to mark wear-accelerated hotspots.

4. The method for planning the motion trajectory of an injection molding robot according to claim 3, characterized in that, The method further includes: Continuously monitor the microstructure changes in wear-accelerated hotspot areas; When a change is detected in any of the size, shape, or risk level of the wear acceleration hotspot area, the trajectory re-optimization process is initiated to calculate the dynamic avoidance area based on the geometric information and risk level of the current wear acceleration hotspot area. Using the dynamic avoidance area as a constraint, the original motion trajectory of the injection molding robot is locally replanned, and the speed of the injection molding robot when passing through the wear acceleration hotspot area is dynamically adjusted according to the risk level and changing trend of the wear acceleration hotspot area.

5. The method for planning the motion trajectory of an injection molding robot according to claim 4, characterized in that, The continuous monitoring of the microscopic changes in the wear-accelerated hotspot region includes: After each part-picking cycle is completed, the microscopic imaging sensor integrated in the injection molding robot is used to acquire multi-angle and multi-distance images of key areas of the injection mold cavity and the contact surface of the end effector of the injection molding robot, thereby obtaining multi-view image data. The multi-view image data is reconstructed in three dimensions to obtain three-dimensional point cloud data of the injection mold cavity and the surface of the injection manipulator end effector. The surface normals and curvature of the three-dimensional point cloud data are calculated to identify and remove occluded or distorted areas caused by pits, scratches or foreign objects. The three-dimensional point cloud data after removing occluded or distorted areas is then meshed to obtain comprehensive and distortion-free microscopic state data. The microscopic state data reflects the microscopic state changes of the continuously monitored wear-accelerated hotspot areas.

6. The method for planning the motion trajectory of an injection molding robot according to claim 5, characterized in that, The step of performing surface normal calculation and curvature analysis on the three-dimensional point cloud data to identify and remove occluded or distorted areas caused by pits, scratches, or foreign object attachments includes: Obtain the normal vector and curvature value of each point in the three-dimensional point cloud data; Based on the normal vector and the curvature value, calculate the mean angle and curvature variance between the normal vectors of each point in the 3D point cloud data and its local neighborhood points; Regions where the mean angle between the normal vectors exceeds a threshold value and the curvature variance is below a threshold value are identified as regions with minor wear characteristics. Regions where the mean angle between the normal vectors exceeds the threshold value and the curvature variance exceeds the threshold value are identified as non-wear surface defect regions. The identified non-wear surface defect areas are considered as occlusion or distortion areas caused by pits, scratches, or foreign matter attachments and are removed from the 3D point cloud data.

7. The method for planning the motion trajectory of an injection molding robot according to claim 5, characterized in that, The step of performing surface normal calculation and curvature analysis on the three-dimensional point cloud data to identify and remove occluded or distorted areas caused by pits, scratches, or foreign object attachments includes: Obtain the density information of each point in the three-dimensional point cloud data; Based on the density information, identify low-density regions in the three-dimensional point cloud data; Connectivity analysis is performed on the low-density region to identify isolated points or sparse clusters of points within the low-density region. Obtain the normal vector and curvature value of the isolated point or sparse point cluster; The range of variation of the normal vector and the range of fluctuation of the curvature value of the isolated point or sparse point cluster are compared with a preset threshold. Isolated points or sparse point clusters whose normal vector variation range and curvature value fluctuation range are both within a preset threshold are identified as low-density noise points caused by non-solid attachments. Low-density noise points caused by identified non-solid attachments are identified as occlusion or distortion areas caused by pits, scratches, or foreign object attachments and are removed from the 3D point cloud data.

8. The method for planning the motion trajectory of an injection molding robot according to claim 5, characterized in that, The step of performing surface normal calculation and curvature analysis on the three-dimensional point cloud data to identify and remove occluded or distorted areas caused by pits, scratches, or foreign object attachments includes: While acquiring the three-dimensional point cloud data, the posture sensor data of the injection molding robot itself is acquired simultaneously; Based on the attitude sensor data, attitude correction is performed on the three-dimensional point cloud data; and on the corrected three-dimensional point cloud data, surface normal calculation and curvature analysis are performed to identify and remove occluded or distorted areas caused by pits, scratches or foreign object attachments.

9. A motion trajectory planning system for an injection molding robot, characterized in that, The system includes: The micro-state perception module is used to obtain current micro-state data by using micro-imaging sensors integrated in the injection molding robot to perform non-contact micro-state perception on key areas of the injection mold cavity and the contact surface of the end effector of the injection molding robot after each part picking cycle operation. The micro-change identification and quantification module is used to compare the current micro-state data with historical state data to identify and quantify the micro-physical changes of the injection mold cavity and the end effector of the injection robot. The trajectory correction module is used to dynamically correct the motion trajectory of the injection molding robot based on the identified and quantified microscopic physical changes; the dynamic correction of the motion trajectory of the injection molding robot includes increasing the safety clearance between the end effector of the injection molding robot and the cavity of the injection mold or adjusting the posture of the end effector of the injection molding robot. The trajectory execution module is used to execute the dynamically corrected motion trajectory of the injection molding robot. After each part-removal cycle, the injection molding robot uses microscopic imaging sensors integrated into the robot to perform non-contact microscopic state sensing of key areas of the injection mold cavity and the contact surface of the robot's end effector, in order to obtain current microscopic state data, including: After each part-removal cycle, the injection molding robot uses a microscopic imaging sensor integrated into the robot to perform two consecutive image acquisitions on the key areas of the injection mold cavity and the contact surface of the robot's end effector, obtaining a first image and a second image. The first image and the second image are used as the current microscopic state data. The first image is acquired when the active illumination source inside the microscopic imaging sensor is turned on during the first acquisition, and the second image is acquired when the active illumination source is turned off during the second acquisition. The step of comparing the current microscopic state data with historical state data to identify and quantify the microscopic physical changes in the injection mold cavity and the end effector of the injection robot includes: Perform pixel-level difference operations on the first image and the second image to obtain a net difference image; Image analysis is performed on the net difference image to identify and quantify the local high-density aggregation regions formed by tiny suspended particles, thereby obtaining the microscopic physical changes.

Images