Self-adaptive control method for operation of car coupler and uncoupler robot under multiple working conditions

Abstract

The invention discloses a self-adaptive control method for operation of a coupler unhooking robot under multiple working conditions, and relates to the technical field of coupler unhooking, and the method comprises the steps: obtaining a target unhooking robot, and collecting and obtaining an operation environment data flow and a coupler state data flow; working condition sensing and coupler recognition are conducted, and the current working condition and coupler feature information are obtained; performing traversal retrieval and adaptive adjustment in the unhooking strategy database, and determining target unhooking strategy parameters; and executing unhooking operation according to the target unhooking strategy parameters, and performing adaptive closed-loop control and abnormal emergency control on the target unhooking robot based on the unhooking operation state data flow. The technical problems that in the prior art, a car coupler unhooking robot is difficult to adapt to complex and various working conditions, and consequently the unhooking operation accuracy, safety and efficiency are poor are solved, accurate sensing of the working conditions of the car coupler unhooking robot is achieved, and the technical effect of improving the unhooking operation accuracy, safety and efficiency is achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to coupler uncoupling, and in particular to an adaptive control method for coupler uncoupling robot operations under multiple working conditions. Background Art

[0002] Couplers are key components connecting train cars, and the accuracy and efficiency of their uncoupling operations directly impact the operational efficiency of the entire railway transportation system. With the continuous growth of railway transportation business volume and the increasing demand for transportation efficiency, the frequency of train operations has increased, and railway transportation scenarios are becoming more complex and diverse. There are many different operating conditions, such as varying line conditions, weather conditions, lighting, train types, track flatness, and coupler models. These conditions can include slight deviations in coupler position and potential mechanical failures. Traditional coupler uncoupling robots struggle to accurately and stably adapt to various operating conditions and different types of couplers in complex and changing operating environments. This means they are unable to dynamically adjust their operating strategies based on changing scenarios, making it difficult to ensure smooth uncoupling operations. These robots may even cause equipment damage or safety incidents, impacting the accuracy, safety, and efficiency of the operation.

[0003] At present, there is a technical problem in the relevant technologies that the coupler and uncoupling robots are difficult to adapt to complex and diverse working conditions, resulting in poor accuracy, safety and efficiency of the uncoupling operation. Summary of the Invention

[0004] This application solves the technical problem in the prior art that coupler unhooking robots are difficult to adapt to complex and diverse working conditions, resulting in poor accuracy, safety and efficiency of unhooking operations, by providing an adaptive control method for coupler unhooking robots under multiple working conditions. This application realizes accurate perception of the working conditions of the coupler unhooking robots, achieving the technical effect of improving the accuracy, safety and efficiency of unhooking operations.

[0005] The present application provides an adaptive control method for a coupler uncoupling robot operation under multiple working conditions, including: obtaining a target uncoupling robot, wherein the target uncoupling robot is integrated with a multi-source sensor device, and the operating environment data stream and the coupler status data stream are collected in real time by the multi-source sensor device; performing working condition perception and coupler identification on the operating environment data stream and the coupler status data stream to obtain the current operating condition and coupler feature information, wherein the coupler feature information includes the coupler type, coupler position information and abnormal operating conditions; mining an uncoupling strategy database based on the association of the target uncoupling robot, performing traversal retrieval and adaptive adjustment in the uncoupling strategy database based on the current operating condition and coupler feature information, and determining target uncoupling strategy parameters; controlling the target uncoupling robot to perform the uncoupling operation according to the target uncoupling strategy parameters, while monitoring and obtaining the uncoupling operation status data stream, and performing adaptive closed-loop control and abnormal emergency control on the target uncoupling robot based on the uncoupling operation status data stream.

[0006] In a possible implementation, the adaptive control method for coupler and uncoupling robot operations under multiple working conditions further performs the following processing: the operating environment data stream and the coupler status data stream are source-divided by the multi-source sensor device to obtain a multi-source uncoupling operation data stream, wherein the multi-source sensor device includes a 3D camera, a laser radar, and a force perception network; according to the acquisition characteristic information of the multi-source uncoupling operation data stream, a multi-source data filter is initialized, and the multi-source uncoupling operation data stream is filtered and pre-processed using the multi-source data filter to obtain a multi-source available uncoupling operation data stream; the multi-source available uncoupling operation data stream is synchronously aligned according to the sensor acquisition timing to obtain a multi-source standard uncoupling operation data stream; the multi-source standard uncoupling operation data stream is subjected to working condition perception and coupler identification to obtain the current working condition and coupler feature information.

[0007] In a possible implementation, the adaptive control method for the coupler uncoupling robot operation under multiple working conditions also performs the following processing: constructing a multimodal feature classification network based on the coupler type data set training, the backbone network structure of the multimodal feature classification network including the ResNet-50 deep residual network, the PointNet deep learning network and the image point cloud dual-branch architecture; using the multimodal feature classification network to perform feature extraction and fusion and coupler type matching on the point cloud data and image data in the multi-source standard uncoupling operation data stream to determine the coupler type; performing operating condition identification and coupler segmentation detection on the multi-source standard uncoupling operation data stream to obtain the current operating condition, coupler position information and abnormal operating conditions; performing feature integration on the coupler type, the coupler position information and the abnormal operating conditions to determine the coupler feature information.

[0008] In a possible implementation, the adaptive control method for coupler uncoupling robot operations under multiple working conditions further performs the following processing: building a deep visual recognition module, the cascade architecture of the deep visual recognition module including a YOLOv5-Hook network and a Mask R-CNN segmentation network; performing coupler segmentation detection and reconstruction position evaluation on the multi-source standard uncoupling operation data stream through the deep visual recognition module to obtain target coupler status and coupler position information; performing environmental condition extraction on the multi-source standard uncoupling operation data stream to obtain uncoupling environmental condition working conditions, and determining the current operating conditions based on the uncoupling environmental condition working conditions and the target coupler status; performing abnormal diagnosis based on the target coupler status and coupler position information to obtain abnormal operating conditions, and the abnormal operating conditions include coupler deformation detection information and operating position deviation.

[0009] In a possible implementation, the adaptive control method for the coupler uncoupling robot operation under multiple working conditions also performs the following processing: performing coupler identification detection and coupler segmentation identification on the multi-source standard uncoupling operation data stream through the deep vision recognition module to obtain a target coupler mask area; performing point cloud projection and three-dimensional reconstruction on the pixel point cloud data in the target coupler mask area to generate a three-dimensional model of the coupler status; constructing three-dimensional operation coordinates, performing coordinate mapping conversion on the three-dimensional coupler status model based on the three-dimensional operation coordinates, and outputting coupler position information; performing boundary detection and coupler status evaluation on the three-dimensional coupler status model to obtain a target coupler status.

[0010] In a possible implementation, the adaptive control method for the coupler uncoupling robot operation under multiple working conditions further performs the following processing: performing point cloud shape analysis and local curvature calculation based on the target coupler state to obtain local curvature information of the coupler; comparing and deforming the local curvature information of the coupler according to the coupler curvature threshold to obtain coupler deformation detection information; determining the coupler operation position threshold according to the target coupler state; performing deviation calculation on the coupler position information based on the coupler operation position threshold to determine the operation position deviation, and obtaining abnormal operation conditions based on the coupler deformation detection information and the operation position deviation.

[0011] In a possible implementation, the adaptive control method for coupler and uncoupling robot operations under multiple working conditions further performs the following processing: determining uncoupling coding element information based on the current working condition and coupler feature information; using the uncoupling coding element information to perform element encoding on the uncoupling strategy database to obtain an uncoupling strategy coding database; performing traversal, retrieval and matching within the uncoupling strategy database based on the current working condition and coupler feature information to obtain an uncoupling strategy fitness set; performing adaptive screening and adaptive adjustment on the uncoupling strategy database according to the uncoupling strategy fitness set to determine target uncoupling strategy parameters.

[0012] In a possible implementation, the adaptive control method for coupler and uncoupling robot operations under multiple working conditions further performs the following processing: adaptively sorting the uncoupling strategy database according to the uncoupling strategy fitness set to obtain an adaptive uncoupling strategy parameter sequence; optimizing the adaptive uncoupling strategy parameter sequence to determine initial uncoupling strategy parameters; performing simulation deviation analysis and adaptive adjustment based on the initial uncoupling strategy parameters to obtain target uncoupling strategy parameters.

[0013] In a possible implementation, the adaptive control method for coupler and unhooking robot operations under multiple working conditions further performs the following processing: obtaining an unhooking effect evaluation index set, performing simulation fitting on the unhooking strategy database based on the unhooking effect evaluation index set, and constructing an unhooking effect simulation module; performing simulation evaluation on the initial unhooking strategy parameters based on the unhooking effect simulation module to obtain an initial unhooking fitting effect; performing adaptive deviation adjustment on the initial unhooking strategy parameters based on the initial unhooking fitting effect to obtain target unhooking strategy parameters.

[0014] In a possible implementation, the adaptive control method for the coupler unhooking robot operation under multiple working conditions also performs the following processing: using the deviation information between the initial unhooking fitting effect and the preset unhooking effect as the unhooking optimization direction; adaptively adjusting and comparing the initial unhooking strategy parameters according to the unhooking optimization direction to determine the target unhooking strategy parameters.

[0015] The adaptive control method for coupler uncoupling robot operations under multiple working conditions proposed in this application is intended to obtain a target uncoupling robot, collect and obtain the working environment data stream and the coupler status data stream; perform working condition perception and coupler identification on the working environment data stream and the coupler status data stream to obtain the current working condition and coupler feature information; perform traversal retrieval and adaptive adjustment in the uncoupling strategy database to determine the target uncoupling strategy parameters; perform uncoupling operations according to the target uncoupling strategy parameters, and perform adaptive closed-loop control and abnormal emergency control on the target uncoupling robot based on the uncoupling operation status data stream. This solves the technical problem in the prior art that coupler uncoupling robots are difficult to adapt to complex and diverse working conditions, resulting in poor accuracy, safety, and efficiency in uncoupling operations, realizes accurate perception of the working conditions of the coupler uncoupling robot, and achieves the technical effect of improving the accuracy, safety, and efficiency of uncoupling operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below. In the present application, flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously according to needs. Meanwhile, other operations can be added to these processes, or one or more steps of operations can be removed from these processes.

[0017] Figure 1 A flowchart of a multi-working-condition car coupler uncoupling robot operation adaptive control method provided by the embodiments of the present application is shown.

[0018] Figure 2 A flowchart of working condition sensing and car coupler identification in a multi-working-condition car coupler uncoupling robot operation adaptive control method provided by the embodiments of the present application is shown. DETAILED DESCRIPTION

[0019] The foregoing description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described.

[0020] In order to make the purposes, technical solutions and advantages of the present application more clear, the following will further describe the present application in combination with the drawings, and the described embodiments should not be regarded as limiting the present application. All other embodiments obtained by those skilled in the art without making creative efforts belong to the scope of protection of the present application.

[0021] In the following description, "some embodiments" are described, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The term "first\second" is only to distinguish similar objects, and does not represent a specific order of the objects. The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, product or server including a series of steps does not have to be limited to those steps clearly listed, but can include other steps not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art in the technical field of the present application. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0022] The embodiments of the present application provide a multi-working-condition car coupler uncoupling robot operation adaptive control method, such as Figure 1The method comprises the following steps: In step S100, a target uncoupling robot is acquired, which is integrated with a multi-source sensor device, and through the multi-source sensor device, a working environment data stream and a coupler state data stream are acquired in real time.

[0023] Preferably, the target uncoupling robot for the current uncoupling operation is acquired, wherein the target uncoupling robot is integrated with a multi-source sensor device for sensing the state of the surrounding environment and the coupler. The multi-source sensor device can include a visual sensor such as a 3D camera for collecting image and video information of the working site, and further identifying the appearance features, position and surrounding obstacles of the coupler; a laser radar sensor for acquiring three-dimensional spatial information of the surrounding environment in real time, accurately measuring the distance between the coupler and the robot, and the position and shape of the surrounding objects; a force sensor arranged on the operating part of the robot for detecting the force applied by the robot during contact with the coupler and the reaction force received, helping the robot to determine whether the coupler is correctly grasped and whether abnormal resistance is encountered during uncoupling; a position sensor such as an inertial measurement unit (IMU) and an encoder for determining the position and attitude of the robot itself and the relative position of the coupler. Then, through the multi-source sensor device, a working environment data stream and a coupler state data stream are acquired in real time. Specifically, the working environment data stream refers to the on-site environment related data of the uncoupling robot operation, which can include environmental lighting intensity, temperature, humidity and other meteorological conditions, as well as information such as the terrain, track conditions and whether there are obstacles around the working site. For example, the visual sensor can capture whether other vehicles or personnel are close, and the laser radar can detect the track flatness and the position of the obstacles; the coupler state data stream refers to the state information of the coupler itself, including the type of coupler (such as automatic coupler, semi-automatic coupler, etc.), the connection state of the coupler (whether it is tightly connected, whether there are signs of looseness), the position information of the coupler (relative to the specific position and attitude of the robot), and whether there is an abnormal operation condition (such as coupler damage, deformation, etc.), and the force sensor can detect the tightness of the coupler connection, and the visual sensor can identify the type and appearance of the coupler.

[0024] In step S200, the working condition perception and coupler identification are performed on the working environment data stream and the coupler state data stream to obtain the current working condition and coupler feature information, wherein the coupler feature information includes the coupler type, the coupler position information and the abnormal operation condition.

[0025] Preferably, the working condition perception is performed on the working environment data stream. Specifically, the working condition of the working site is perceived based on the working environment data stream. For example, the lighting conditions of the working site are analyzed using visual sensor data to determine whether it is in a low-light environment (such as at night or in a tunnel); based on the lidar data, the terrain and obstacle distribution of the working site are perceived. If obstacles are detected near the track, such as fallen debris or other equipment, the robot needs to avoid these obstacles to ensure working safety, or adjust the working path to adapt to such working conditions; combined with environmental data such as temperature and humidity, it is determined whether the environment has an impact on the coupling and uncoupling operation. For example, in a low temperature environment, the coupler may freeze.

[0026] Preferably, coupler status data streams collected by multi-source sensor equipment are used to identify the coupler. Specifically, visual sensors use images to identify the coupler's appearance features, thereby determining the coupler type, such as whether it is a close-fitting coupler or a standard coupler. Data from lidar and position sensors are used to accurately determine the coupler's position, including its coordinates and angle in three-dimensional space. Data from devices such as force sensors is analyzed to determine whether the coupler is operating abnormally. For example, if a force sensor detects an abnormal force change within the normal operating range, it may indicate damage, deformation, or a stuck coupler. Through working condition perception analysis and coupler identification, current operating conditions (including environmental conditions, terrain conditions, obstacle conditions, etc.) and coupler feature information (including coupler type, coupler position information, and abnormal coupler operation) are obtained.

[0027] Further, such as Figure 2 As shown, step S200 also includes step S210, performing source diversion on the operating environment data stream and the coupler status data stream through the multi-source sensor device to obtain a multi-source unhooking operation data stream, wherein the multi-source sensor device includes a 3D camera, a laser radar and a force perception network; step S220, initializing a multi-source data filter according to the acquisition characteristic information of the multi-source unhooking operation data stream, and using the multi-source data filter to perform filtering preprocessing on the multi-source unhooking operation data stream respectively to obtain a multi-source available unhooking operation data stream; step S230, performing synchronous alignment processing on the multi-source available unhooking operation data stream according to the sensor acquisition timing to obtain a multi-source standard unhooking operation data stream; step S240, performing working condition perception and coupler identification on the multi-source standard unhooking operation data stream to obtain current working condition and coupler feature information.

[0028] Preferably, the multi-source sensor device splits the job environment data stream and the coupler status data stream according to their sources (i.e. different sensors) to more clearly process each type of data subsequently, obtaining a multi-source uncoupling operation data stream. Specifically, the multi-source sensor device includes a 3D camera, a laser radar, and a force perception network. The 3D camera provides three-dimensional image information of the job environment and the coupler, including the shape, size, spatial position, etc. of the object. The data collected by the 3D camera can identify the appearance features of the coupler, the relative position relationship with the surrounding objects, etc. The laser radar is used for distance measurement and can obtain three-dimensional spatial point cloud data of the surrounding environment in real time, which can accurately depict the terrain, obstacle distribution, and coupler position, etc. of the job site. The force perception network is used to detect the force and torque information received by the robot during contact and operation with the coupler. The data collected by the force perception network can determine the connection tightness of the coupler, whether there is abnormal resistance, etc.

[0029] Preferably, the data collected by different sensors has different characteristics. For example, the image data collected by the 3D camera may be affected by factors such as light and noise. The laser radar data may have measurement errors and data missing problems. The data collected by the force perception network may be disturbed by mechanical vibration and electromagnetic interference. According to the collection characteristics of different sensor data, appropriate filtering algorithms are selected and filter parameters are initialized. For example, for image data, median filtering, Gaussian filtering, etc. can be used to remove noise. For laser radar data, Kalman filtering, particle filtering, etc. can be used to improve the accuracy and stability of the data. For force data, low-pass filtering can be used to remove high-frequency interference. As shown in Table 1, the filter initialization parameters are shown as an example: Table 1 Multi-source sensor data filter initialization parameter table

[0030] Preferably, an initialized multi-source data filter is used to process the multi-source decoupling operation data stream to remove noise, interference and errors in the data to improve the quality and reliability of the data, thereby obtaining a multi-source available decoupling operation data stream. The multi-source available uncoupling operation data streams are adjusted and matched according to the actual time sequence of sensor data acquisition, aligning the data from different sensors in time. This ensures that the data from each sensor accurately reflects the operating environment and coupler status at the same moment, thereby processing the data stream to generate a multi-source standard uncoupling operation data stream. The system then comprehensively analyzes various information in the multi-source standard uncoupling operation data stream to determine the current operating environment and working conditions. This includes combining data from 3D cameras and lidar to determine the presence of obstacles at the work site, the flatness of the track, and the presence of good lighting conditions. Data from the force perception network is then used to determine whether there are any abnormal force changes during the operation, thereby inferring whether the coupler is in normal working condition. The data from the multi-source standard uncoupling operation data stream is also used to identify the specific characteristics of the coupler. Specifically, the 3D camera image data is used to identify the coupler type (e.g., automatic coupler, semi-automatic coupler, etc.). The lidar and 3D camera data are combined to accurately determine the coupler's position. Based on the data from the force perception network, the system determines whether the coupler is operating abnormally, such as if it is stuck or damaged.

[0031] Specifically, step S240 also includes step S241, constructing a multimodal feature classification network based on the training of the coupler type data set, and the backbone network structure of the multimodal feature classification network includes a ResNet-50 deep residual network, a PointNet deep learning network and an image point cloud dual-branch architecture; step S242, using the multimodal feature classification network to perform feature extraction, fusion and coupler type matching on the point cloud data and image data in the multi-source standard decoupling operation data stream to determine the coupler type; step S243, performing operating condition identification and coupler segmentation detection on the multi-source standard decoupling operation data stream to obtain the current operating condition, coupler position information and abnormal operating conditions; step S244, integrating the features of the coupler type, the coupler position information and the abnormal operating conditions to determine the coupler feature information.

[0032] Preferably, the car coupler type data set contains multi-modal data (such as images taken by a 3D camera and point cloud data obtained by a laser radar) of various different types of car couplers, wherein each data sample is labeled with the corresponding car coupler type, and then the multi-modal feature classification network is trained using the car coupler type data set, wherein the backbone network structure of the multi-modal feature classification network includes a ResNet-50 deep residual network, a PointNet deep learning network, and an image point cloud dual-branch architecture. Specifically, the ResNet-50 deep residual network is a deep convolutional neural network mainly used for processing image data. By introducing residual blocks, it solves the problems of gradient vanishing and gradient explosion in the training process of deep neural networks, and can effectively learn high-level semantic features in images, such as the appearance shape and texture of car couplers. The PointNet deep learning network is used to process point cloud data and learn global and local features of the point cloud, thereby capturing the geometric shape and structural information of the car coupler in three-dimensional space. The image point cloud dual-branch architecture is used to combine the ResNet-50 deep residual network and the PointNet deep learning network, fully utilizing the complementary information of the two different modal data, realizing feature fusion of image data and point cloud data, and improving the accuracy of car coupler type classification.

[0033] Preferably, image data of various car coupler types is obtained from the car coupler type data set, including car coupler images under different lighting conditions, shooting angles, and backgrounds. Each image is labeled with the corresponding car coupler type label, and the data set is divided into a training set, a validation set, and a test set in a ratio of 70%, 15%, and 15%. Preprocessing operations are performed on the images, such as adjusting the image size (usually 224x224 pixels to adapt to the input requirements of ResNet-50), normalizing (scaling the pixel values to the range [0, 1]), and data augmentation (such as random cropping, flipping, rotation, etc.) to increase the diversity of the data. Then, a ResNet-50 deep residual network is constructed, which includes multiple residual block groups, each containing multiple convolutional layers. A loss function is defined and the training set is used for iterative training. The loss is calculated and the model parameters are updated in each iteration to obtain the ResNet-50 deep residual network. The point cloud data is sampled and normalized to ensure that each point cloud sample contains the same number of points and the range of the point cloud data is within a suitable interval. The data set is divided into a training set, a validation set, and a test set in a ratio of 70%, 15%, and 15%. A PointNet deep learning network is constructed, including an input transformation network, a feature extraction network, and a classification network. A loss function is defined and the training set is used for training to obtain the PointNet deep learning network. Finally, the validation set and the test set are used to evaluate the trained ResNet-50 deep residual network and PointNet deep learning network.

[0034] Preferably, the point cloud data and image data in the multi-source standard uncoupling hook operation data stream are subjected to feature extraction and fusion by using a multi-modal feature classification network, that is, the point cloud data in the multi-source standard uncoupling hook operation data stream is input into a PointNet deep learning network to extract the features of the point cloud data, the image data is input into a ResNet-50 deep residual network to extract the features of the image data, then the extracted point cloud features and image features are fused through an image point cloud double-branch architecture to obtain multi-modal features; then the multi-modal features are input into a classifier of the multi-modal feature classification network to match the features with known types of couplers, so as to determine the type of the current coupler.

[0035] Preferably, the multi-source standard uncoupling hook operation data stream is subjected to operation condition recognition and coupler segmentation detection, that is, various information in the multi-source standard uncoupling hook operation data stream, such as ambient light, temperature, humidity, obstacle distribution, etc., is used to recognize the current operation condition through a pre-trained condition recognition model (such as a machine learning-based recognition model) to determine whether the operation environment is indoor or outdoor, whether there is adverse weather condition, etc.; the target detection and segmentation algorithm is used to accurately segment the part of the coupler from the multi-source standard uncoupling hook operation data stream, specifically, for image data, a target detection algorithm based on a convolutional neural network (such as Faster R-CNN, YOLO, etc.) is used, and for point cloud data, a point cloud segmentation algorithm based on deep learning (such as PointNet++, etc.) is used. Through the coupler segmentation detection, the position of the coupler in the image or point cloud can be determined, and then the position information of the coupler in the actual space can be calculated. By analyzing the data in the multi-source standard uncoupling hook operation data stream, it can be determined whether the coupler has abnormal operation conditions, such as whether the coupler is stuck or has abnormal stress condition through the data of the force perception network; whether the coupler is damaged or deformed through the image or point cloud data. Finally, the coupler type, coupler position information and abnormal operation conditions are integrated to form complete coupler feature information, which is helpful for realizing adaptive control of the uncoupling robot.

[0036] Further, the step S243 further comprises step A of building a deep visual recognition module, wherein the cascade architecture of the deep visual recognition module comprises a YOLOv5-Hook network and a Mask R-CNN segmentation network; step B of performing coupler segmentation detection and reconstructed position evaluation on the multi-source standard uncoupling hook operation data stream through the deep visual recognition module to obtain target coupler state and coupler position information; step C of extracting environment condition from the multi-source standard uncoupling hook operation data stream to obtain uncoupling environment condition, and determining the current operation condition according to the uncoupling environment condition and the target coupler state; and step D of performing abnormal diagnosis based on the target coupler state and coupler position information to obtain abnormal operation conditions, wherein the abnormal operation conditions comprise coupler deformation detection information and operation position deviation.

[0037] Preferably, a deep learning framework (such as PyTorch or TensorFlow) is used to define and configure corresponding parameters according to the network structure of YOLOv5 and Mask R-CNN, and the two networks are cascaded to form a deep visual recognition module. At the same time, appropriate training data (including images of couplers and corresponding annotation information) needs to be prepared to train and optimize the module so that it can accurately detect and segment couplers. The deep visual recognition module adopts a cascade architecture, including a YOLOv5-Hook network and a Mask R-CNN segmentation network. Specifically, the YOLOv5-Hook network is an efficient target detection algorithm that can quickly locate the approximate position and category of the coupler in the image data in the multi-source standard hook removal operation data stream; Mask R-CNN adds instance segmentation function on the basis of Faster R-CNN, which can perform more detailed processing on the coupler area detected by the YOLOv5-Hook network, not only accurately segmenting the coupler outline, but also assigning a category label to each pixel of the coupler, thereby achieving accurate recognition and segmentation of the coupler.

[0038] Preferably, the image data from the multi-source standard hook removal operation data stream is input into the deep visual recognition module. First, the YOLOv5-Hook network quickly scans the image to detect the approximate location and category of the coupler. The Mask R-CNN segmentation network then segments the detected coupler area, accurately outlining the coupler's contour and determining its specific location and shape in the image, thereby achieving accurate segmentation and detection of the coupler. The coupler's position in actual three-dimensional space is then reconstructed and evaluated by combining other information from the multi-source standard hook removal operation data stream (such as point cloud data and sensor location information). This includes combining lidar point cloud data and image data to more accurately determine the coupler's three-dimensional position and posture. Through the reconstructed position evaluation, the accurate position information (including precise coordinates and posture information) of the target coupler in the actual operating environment is obtained. The segmentation results of the Mask R-CNN segmentation network are analyzed and judged to determine the target coupler's status, mainly including whether the coupler's appearance is intact, damaged, or deformed.

[0039] Preferably, relevant information about environmental conditions is extracted from the multi-source standard unhooking operation data stream, such as using the lighting information in the image data to judge the brightness of the environment, whether it is day or night, whether there is sufficient light, etc.; the temperature, humidity, air pressure and other data obtained by the sensor are used to understand the climatic conditions of the environment; the lidar data is analyzed to determine whether there are obstacles at the work site, the flatness of the track, etc.; these extracted environmental information are integrated to obtain the unhooking environmental condition working conditions; then, the unhooking environmental condition working conditions and the target coupler status are combined to comprehensively judge the current operating conditions of unhooking. Specifically, if the unhooking environmental condition working conditions are displayed as low light, obstacles on the track, and the target coupler state is deformed, then the current operating conditions are described as "complex operating conditions with deformation of the coupler in a low-light, obstacle-filled environment." Finally, the target coupler status is analyzed in detail. The coupler segmentation results obtained through the Mask R-CNN segmentation network are compared with the shape and appearance of a normal coupler to determine whether the coupler is deformed and the degree and location of the deformation. At the same time, the actual position information of the coupler is compared with the preset standard position information to calculate the operating position deviation. This helps to promptly discover and solve problems in the unhooking operation and ensure the safety of the unhooking operation.

[0040] Furthermore, step B also includes step B1, performing coupler identification detection and coupler segmentation identification on the multi-source standard uncoupling operation data stream through the deep visual recognition module to obtain a target coupler mask area; performing point cloud projection and three-dimensional reconstruction on the pixel point cloud data in the target coupler mask area to generate a three-dimensional model of the coupler status; constructing three-dimensional operation coordinates, performing coordinate mapping conversion on the three-dimensional coupler status model based on the three-dimensional operation coordinates, and outputting coupler position information; performing boundary detection and coupler status evaluation on the three-dimensional coupler status model to obtain a target coupler status.

[0041] Preferably, the deep visual recognition module receives image data from a multi-source standard hook removal operation data stream, and the YOLOv5-Hook network quickly scans the image, detects the approximate position and category information of the coupler in the image, and quickly locates the area where the coupler is located; then, the Mask R-CNN segmentation network performs more detailed processing on the coupler area, accurately outlines the outline of the coupler by classifying and segmenting each pixel in the image, and assigns specific labels to the pixels of the coupler, distinguishing the coupler from the background and other objects, and finally obtains the target coupler mask area, which is a binary image of the same size as the original image, clearly showing the precise position and shape of the coupler in the image, wherein the pixels of the coupler are identified.

[0042] Preferably, after obtaining the target coupler mask area, combined with the point cloud data in the multi-source standard hooking operation data stream (such as the point cloud from the lidar), for each pixel in the mask area, the corresponding point cloud data point is found (the correspondence between the pixel and the point cloud is achieved through methods such as internal and external parameter calibration of the sensor), and then the point cloud data is projected into a coordinate system related to the image plane to establish a mapping between the two-dimensional image pixels and the three-dimensional point cloud data; then, based on the mapped point cloud data, a three-dimensional reconstruction algorithm (such as a point cloud-based gridding algorithm, a surface fitting algorithm, etc.) is used to construct a three-dimensional model of the coupler status to accurately reflect information such as the shape, size and spatial position of the coupler, and intuitively display the status of the coupler in three-dimensional space.

[0043] Preferably, the working three-dimensional coordinates are constructed according to the actual conditions of the working site and the installation position and parameters of the sensors, and then the generated three-dimensional model of the coupler status is converted from its own local coordinate system to the working three-dimensional coordinate system. Through coordinate transformation operations (such as translation, rotation, scaling, etc.), the coordinates of the three-dimensional model of the coupler status are matched with the working three-dimensional coordinate system. After the coordinate mapping conversion, the position information of the coupler in the working three-dimensional coordinate system (including the three-dimensional coordinates of the coupler center point, the orientation of the coupler, etc.) is accurately determined, and the coupler position information is output. Then, the boundary detection of the three-dimensional model of the coupler status is performed, that is, the outer contour and boundary of the coupler model are identified, that is, the normal vector and curvature of each point of the three-dimensional model are calculated, and the edge and boundary points of the coupler are determined, so as to accurately outline the external shape and boundary of the coupler, so as to more clearly understand the structure and form of the coupler; finally, the coupler status is evaluated, including comparing the actual size of the coupler with the standard size, checking whether the key parts of the coupler (such as the connection point, coupler tongue, etc.) are deformed, worn or damaged, etc., and finally the target coupler status (such as normal, slightly deformed, seriously damaged, etc.) is obtained, providing important status information for the coupler uncoupling operation.

[0044] Furthermore, step D also includes step D1, performing point cloud shape analysis and local curvature calculation based on the target coupler state to obtain local curvature information of the coupler; step D2, comparing and deforming the local curvature information of the coupler according to the coupler curvature threshold to obtain coupler deformation detection information; step D3, determining the coupler operating position threshold according to the target coupler state; step D4, performing deviation calculation on the coupler position information based on the coupler operating position threshold to determine the operating position deviation, and obtaining abnormal operating conditions based on the coupler deformation detection information and the operating position deviation.

[0045] Preferably, after obtaining the target coupler state, point cloud data related to the coupler is combined to perform point cloud shape analysis to study the overall shape and structural characteristics of the coupler point cloud, including observing the distribution pattern and density of the point cloud. Curvature calculation is then performed for each local region of the coupler point cloud, including calculating the curvature of each point or local region in the coupler point cloud data to reflect the curvature of the coupler surface at that location. Local curvature information for different locations on the coupler is then obtained, more accurately reflecting the shape characteristics of the coupler surface. The coupler curvature threshold is a pre-set standard value based on the coupler's design standards and curvature data under normal operating conditions, defining the normal curvature range of the coupler at different locations. The calculated local curvature information of the coupler is then compared with the coupler curvature threshold to determine whether the curvature is within the corresponding threshold range. If the curvature of a region exceeds the coupler curvature threshold, deformation is considered to have occurred, thereby obtaining coupler deformation detection information, including the locations that have been deformed and the degree of deformation (e.g., slight deformation, severe deformation), etc.

[0046] Preferably, based on the design requirements of the coupler and the position accuracy standards during normal operation, combined with the target coupler state, the allowable position range of the coupler during operation, that is, the coupler operation position threshold, is determined, which stipulates the allowable deviation range of the coupler's coordinate position and posture in space. For example, the coupler has corresponding position thresholds in the horizontal direction, vertical direction and rotation angle to ensure that the coupler is in a suitable position when performing the uncoupling operation; then the coupler position information is compared with the coupler operation position threshold to calculate the deviation between the actual position of the coupler and the allowable position range, such as calculating the deviation value of the coupler center point coordinate in each coordinate axis direction and the deviation value of the coupler posture (rotation angle, etc.), and then obtaining specific operation position deviation information, clarifying the degree of difference between the current position of the coupler and the ideal operation position, which is used to judge whether the coupler can successfully perform the uncoupling operation and whether position adjustment is required. Finally, the coupler deformation detection information and the operating position deviation are integrated to determine whether there is an abnormal operating situation. For example, if the coupler is severely deformed and the operating position deviation is large, it is judged that the current operation is abnormal and corresponding measures need to be taken, such as stopping the operation, repairing the coupler, or adjusting the robot's operating strategy, etc., which helps to ensure the safety and smooth progress of the coupler uncoupling operation.

[0047] Step S300: mining a hook removal strategy database based on the target hook removal robot association, performing traversal retrieval and adaptive adjustment in the hook removal strategy database based on the current working conditions and coupler feature information, and determining target hook removal strategy parameters.

[0048] Preferably, the target uncoupling robot is associated with an uncoupling strategy database, and uncoupling strategy information related to the target uncoupling robot is extracted from the database through data mining technology, wherein the uncoupling strategy database stores a large amount of uncoupling strategy related data, including information on various possible working conditions, coupler characteristics and corresponding uncoupling strategy parameters, etc.; then the current working conditions (uncoupling environmental condition, target coupler state, etc.) and coupler characteristic information (type, size, structure, connection method, etc.) are used to search and match in the uncoupling strategy database, to find the uncoupling strategy that best matches the current actual situation, and then the retrieved uncoupling strategy is adjusted according to the actual working conditions to better adapt to the specific working conditions and coupler characteristics, for example, if the current coupler position deviates from the standard position recorded in the database, the motion path parameters of the uncoupling robot may need to be adjusted; or if the current environment is dark, the parameters of the robot vision system may need to be adjusted to ensure accurate identification of the coupler. Finally, the uncoupling strategy parameters suitable for the current target uncoupling robot are determined to guide the operation of the uncoupling robot, including the robot's motion trajectory, action sequence, force control, vision recognition parameters, etc., so that the uncoupling robot can complete the uncoupling operation in the best way, improving the efficiency and safety of the operation.

[0049] Further, step S300 further comprises step S310 of determining uncoupling coding element information according to the current working conditions and coupler characteristic information; step S320 of element coding the uncoupling strategy database using the uncoupling coding element information to obtain an uncoupling strategy coding database; step S330 of performing traversal search and matching in the uncoupling strategy database based on the current working conditions and coupler characteristic information to obtain an uncoupling strategy adaptation degree set; and step S340 of performing adaptation screening and adaptive adjustment on the uncoupling strategy database according to the uncoupling strategy adaptation degree set to determine target uncoupling strategy parameters.

[0050] Preferably, according to the current working conditions, lighting conditions (bright, dim, night, etc.), temperature, humidity, whether there are obstacles, the state of the track, etc., and the coupler characteristic information, including the type of coupler (such as automatic coupler, semi-automatic coupler, specific model coupler, etc.), coupler position information (coordinates and attitude in three-dimensional space), whether the coupler has deformation or damage, etc. Key elements that can have a significant impact on uncoupling strategy are extracted and organized and coded to form uncoupling coding element information; then each record in the uncoupling strategy database is coded according to the uncoupling coding element information, specifically, the description information about working conditions and coupler characteristics in the database is converted according to the rules of uncoupling coding element information to make it into a unified coding form, and an uncoupling strategy coding database is obtained.

[0051] Preferably, in the uncoupling strategy coding database, according to the current actual working condition and the coding of the coupler feature information, the similarity or matching degree of the coding is compared to determine the matching degree of the uncoupling strategy corresponding to the record to the current actual situation, for example, whether the coding of the current coupler type is consistent with the coding of the coupler type in a record in the database, whether the coding of the working environment matches, and the like, and then the matching degree value of each record is calculated to reflect the matching degree of the uncoupling strategy corresponding to the record to the current actual situation, and the matching degree values of all records are calculated and composed into an uncoupling strategy matching degree set, wherein the matching degree value can be a numerical value (such as a decimal between 0 and 1, 0 indicating complete mismatch and 1 indicating complete match).

[0052] Preferably, according to the uncoupling strategy matching degree set, the uncoupling strategy records with higher matching degrees are screened out from the uncoupling strategy database, and for the screened candidate uncoupling strategies, adaptive adjustment is performed according to the actual working condition, including fine-tuning some parameters in the strategy, such as the movement speed of the robot, the gripping force, the operation sequence, and the like, to better adapt to the current specific working condition and coupler feature, and then the target uncoupling strategy most suitable for the current actual situation is determined from the candidate uncoupling strategies, and the related parameters of the strategy are extracted, i.e., the target uncoupling strategy parameters are determined, which are used to guide the specific operation of the target uncoupling robot, to ensure that the robot can efficiently and safely complete the uncoupling operation.

[0053] Further, step S340 further includes step S341 of performing adaptive sorting on the uncoupling strategy database according to the uncoupling strategy matching degree set to obtain an adaptive uncoupling strategy parameter sequence; step S342 of performing optimization on the adaptive uncoupling strategy parameter sequence to determine initial uncoupling strategy parameters; and step S343 of performing simulation deviation analysis and adaptive adjustment based on the initial uncoupling strategy parameters to obtain target uncoupling strategy parameters.

[0054] Preferably, according to the matching degree values in the uncoupling strategy matching degree set, all uncoupling strategies in the uncoupling strategy database are adaptively sorted, and the strategies with higher matching degrees are placed in front and the strategies with lower matching degrees are placed in back to obtain an adaptive uncoupling strategy parameter sequence, each uncoupling strategy containing corresponding parameters (such as movement trajectory parameters, operation force parameters, time parameters, and the like of the robot); and then the adaptive uncoupling strategy parameter sequence is evaluated and screened, such as setting additional screening conditions for optimization, such as the complexity of the strategy (a strategy that is too complex may be difficult to execute in actual operation), the execution time requirement (some operations may have strict requirements on uncoupling time), and the like, and then the uncoupling strategy most suitable for the current operation requirement and actual situation is selected from the adaptive uncoupling strategy parameter sequence, and the corresponding parameters are extracted to determine the initial uncoupling strategy parameters.

[0055] Preferably, the initial unhooking strategy parameters are simulated by means of computer simulation or the like, i.e. the target unhooking robot is simulated to perform unhooking operations according to the initial unhooking strategy parameters, and the deviation between the actual simulation results and the expected results is recorded during the simulation, for example, whether the motion trajectory of the simulated robot when performing the unhooking operation accurately reaches the position of the coupler, whether the force when grabbing the coupler is appropriate, whether the unhooking time meets the requirements, and the like, to find the deficiencies in the initial unhooking strategy parameters; and the initial unhooking strategy parameters are adjusted and optimized according to the results of the simulation deviation analysis, which can include adjusting the motion speed of the robot, changing the setting of the grabbing force, optimizing the operation sequence, and the like, to reduce the deviation and make the unhooking strategy more in line with the actual operation requirements; and finally the target unhooking strategy parameters are obtained, which are used to actually control the target unhooking robot to perform unhooking operations, to ensure the smooth and efficient completion of the operations.

[0056] Further, step S343 further includes step E, obtaining an unhooking effect evaluation index set, simulating and fitting the unhooking strategy database based on the unhooking effect evaluation index set, and constructing an unhooking effect simulation module; step F, simulating and evaluating the initial unhooking strategy parameters based on the unhooking effect simulation module, and obtaining an initial unhooking fitting effect; and step G, adaptively adjusting the initial unhooking strategy parameters based on the initial unhooking fitting effect, and obtaining target unhooking strategy parameters.

[0057] Preferably, the unhooking effect evaluation index set is constituted by analyzing historical unhooking operation data, including the probability of unhooking success, the time required for unhooking, the damage degree to the coupler and related equipment during the unhooking process, the energy consumption of the unhooking operation, and the like, and then the unhooking strategies in the unhooking strategy database are simulated and evaluated according to the unhooking effect evaluation index set, i.e. each strategy is simulated to run under different hypothetical conditions (corresponding to different operation conditions and coupler characteristics) by means of simulation software, and the performance of each strategy is evaluated according to the unhooking effect evaluation index set, for example, for a certain unhooking strategy, the unhooking process under specific lighting conditions and coupler types is simulated, and the time required for completing unhooking and the damage degree to the coupler are calculated; and the entire simulation process and performance evaluation are constructed as an unhooking effect simulation module, which can quickly simulate the unhooking process according to the input unhooking strategy parameters and operation conditions, coupler characteristics, and the like, and output the corresponding unhooking effect evaluation results.

[0058] Preferably, the initial unhooking strategy parameters are simulated and evaluated using an unhooking effect simulation module. Specifically, the initial unhooking strategy parameters are input into the unhooking effect simulation module, and relevant information such as the current working conditions and coupler characteristics are input at the same time. The target unhooking robot is simulated to perform unhooking operations according to the initial unhooking strategy parameters, and the unhooking process is monitored according to the unhooking effect evaluation index set, and the actual values ​​of various indicators are recorded, thereby obtaining the initial unhooking fitting effect, that is, the actual performance of the strategy under the current working conditions; then, based on the initial unhooking fitting effect, the problems and deviations of the initial unhooking strategy parameters in the actual simulation are analyzed, and the initial unhooking strategy parameters are adjusted accordingly. For example, if the motion trajectory is found to be inaccurate, the motion parameters of the robot are adjusted; if the gripping force is inappropriate, the set value of the gripping force is changed; after multiple adaptive deviation adjustments and simulation evaluations, the initial unhooking strategy parameters are continuously adjusted until a satisfactory unhooking effect is achieved, and then the target unhooking strategy parameters are determined for actually controlling the target unhooking robot to perform unhooking operations, so as to achieve efficient, accurate and safe unhooking operations.

[0059] The preferred step G also includes step G1, taking the deviation information between the initial unhooking fitting effect and the preset unhooking effect as the unhooking optimization direction; step G2, adaptively adjusting and comparing the initial unhooking strategy parameters according to the unhooking optimization direction to determine the target unhooking strategy parameters.

[0060] Preferably, the preset unhooking effect is an ideal effect of the unhooking operation set in advance based on actual operation requirements, safety standards, etc. For example, the preset unhooking effect may stipulate that the probability of successful unhooking should reach more than 95%, the unhooking time should be controlled within 30 seconds, and there should be no obvious damage to the coupler; by comparing the initial unhooking fitting effect and the preset unhooking effect, the difference between the two in various indicators is calculated, and the optimization direction of unhooking is clarified based on the deviation information.

[0061] Preferably, the initial unhooking strategy parameters are reasonably adjusted according to the unhooking optimization direction and the actual working conditions and coupler characteristics. For example, the optimization direction is to increase the probability of successful unhooking, adjust the position parameters and force parameters of the robot to grab the coupler, or change the operation sequence of unhooking, etc.; if it is to shorten the unhooking time, speed up the robot's movement speed parameters, etc.; then the new unhooking strategy parameters after adaptive adjustment are simulated and evaluated again (using the unhooking effect simulation module) to obtain a new unhooking fitting effect, compare the new fitting effect with the preset unhooking effect, analyze the index changes, and continuously find a better combination of unhooking strategy parameters through multiple adjustments and comparisons. When the unhooking strategy parameters are found to make the unhooking fitting effect as close as possible to the preset unhooking effect, and all indicators meet the requirements of the actual operation, they are determined as target unhooking strategy parameters, which are used to actually control the target unhooking robot to perform unhooking operations, so as to achieve more efficient, more accurate and safer unhooking operations.

[0062] Step S400, controlling the target unhooking robot to perform unhooking operations according to the target unhooking strategy parameters, while monitoring and obtaining the unhooking operation status data stream, and performing adaptive closed-loop control and abnormal emergency control on the target unhooking robot based on the unhooking operation status data stream.

[0063] Preferably, the target unhooking strategy parameters are input into the control system of the target unhooking robot, and the target unhooking robot performs the unhooking operation according to the target unhooking strategy parameters, which may include moving to the vicinity of the coupler according to a predetermined motion trajectory, and then grabbing the coupler according to the set force and method to perform the unhooking operation; at the same time, sensors installed on the target unhooking robot (such as 3D cameras, lidars, force perception networks, etc.) are used to collect various status information of the target unhooking robot during the operation in real time, that is, to obtain the unhooking operation status data stream, including the actual motion position and posture information of the target unhooking robot (obtained by position sensors and inertial measurement units), the force conditions when the target unhooking robot contacts the coupler (provided by the force perception network), the real-time status image of the coupler (taken by a 3D camera), the robot's operation steps and time records, etc., so as to comprehensively and accurately reflect the real-time progress and status of the robot's unhooking operation.

[0064] Preferably, the actual operation situation reflected in the unhooking operation status data stream is compared and analyzed with the ideal state set by the target unhooking strategy parameters. If a deviation is found between the actual situation and the ideal state (for example, the robot's motion trajectory deviates from the predetermined path, or the force of grabbing the coupler does not match the set value), the robot's operating parameters and actions are automatically adjusted; the robot's operating status is continuously monitored (the unhooking operation status data stream is obtained), the actual state is compared with the target state, and then the control instructions are adjusted according to the comparison result (the robot's operation is adjusted), and the adjusted results are fed back for re-monitoring. Through closed-loop control, the deviation of the robot during the unhooking operation can be discovered and corrected in time to ensure the accuracy and stability of the operation.

[0065] Preferably, the data stream of the unhooking operation status is analyzed and judged in real time. When an abnormal situation is found in the data stream (such as the force perception network detects an abnormally large force value, indicating that the coupler is stuck or the robot is subjected to a strong collision; or the 3D camera captures a serious deformation of the coupler, etc.), the corresponding emergency control measures are immediately initiated, which may include immediately stopping the current operation of the robot to prevent further damage or danger; or attempting to take repair or adjustment actions (such as adjusting the robot's posture or strength, trying to re-grasp the coupler, etc.) to resolve the abnormal situation, so that the unhooking operation can continue or stop safely, thereby ensuring the smooth, safe and efficient unhooking operation of the target unhooking robot.

[0066] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. The adaptive control method for coupler and uncoupler robot operation under multiple working conditions is characterized by: The method comprises: Acquire a target unhooking robot, wherein the target unhooking robot is integrated with a multi-source sensor device and acquires an operating environment data stream and a coupler status data stream in real time through the multi-source sensor device; Performing working condition perception and coupler identification on the working environment data stream and the coupler status data stream to obtain current working conditions and coupler feature information, wherein the coupler feature information includes coupler type, coupler position information, and abnormal working conditions; Based on the target unhooking robot association, a unhooking strategy database is mined, and based on the current working conditions and coupler feature information, traversal retrieval and adaptive adjustment are performed in the unhooking strategy database to determine the target unhooking strategy parameters; The target unhooking robot is controlled to perform unhooking operations according to the target unhooking strategy parameters, while monitoring and obtaining the unhooking operation status data stream, and adaptive closed-loop control and abnormal emergency control are performed on the target unhooking robot based on the unhooking operation status data stream.

2. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 1, characterized in that: The current operating condition and coupler characteristic information are obtained, including: The operating environment data stream and the coupler status data stream are source-splitted by the multi-source sensor device to obtain a multi-source unhooking operation data stream, wherein the multi-source sensor device includes a 3D camera, a laser radar, and a force perception network; Initializing a multi-source data filter according to the acquisition characteristic information of the multi-source decoupling operation data stream, and using the multi-source data filter to perform filtering preprocessing on the multi-source decoupling operation data stream respectively to obtain a multi-source available decoupling operation data stream; Performing synchronization alignment processing on the multi-source available dehooking operation data streams according to the sensor acquisition timing to obtain a multi-source standard dehooking operation data stream; The multi-source standard uncoupling operation data stream is subjected to working condition perception and coupler identification to obtain current working condition and coupler feature information.

3. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 2, characterized in that: The current operating condition and coupler characteristic information are obtained, including: A multimodal feature classification network is constructed based on the coupler type dataset training. The backbone network structure of the multimodal feature classification network includes a ResNet-50 deep residual network, a PointNet deep learning network, and an image point cloud dual-branch architecture. Using the multimodal feature classification network to perform feature extraction and fusion and coupler type matching on the point cloud data and image data in the multi-source standard uncoupling operation data stream to determine the coupler type; Performing operation condition identification and coupler segmentation detection on the multi-source standard uncoupling operation data stream to obtain current operation conditions, coupler position information, and abnormal operation conditions; The coupler type, the coupler position information and the abnormal operation condition are feature-integrated to determine the coupler feature information.

4. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 3, characterized in that: The obtaining of the current operating conditions, coupler position information and abnormal operating conditions includes: Building a deep visual recognition module whose cascade architecture includes a YOLOv5-Hook network and a Mask R-CNN segmentation network; Performing coupler segmentation detection and reconstruction position evaluation on the multi-source standard uncoupling operation data stream through the deep vision recognition module to obtain target coupler status and coupler position information; Extracting environmental conditions from the multi-source standard uncoupling operation data stream to obtain uncoupling environmental conditions and working conditions, and determining a current working condition based on the uncoupling environmental conditions and working conditions and the target coupler state; An abnormality diagnosis is performed based on the target coupler state and coupler position information to obtain an abnormal operating condition, wherein the abnormal operating condition includes coupler deformation detection information and operating position deviation.

5. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 4, characterized in that: The obtaining of target coupler status and coupler position information includes: Performing coupler recognition detection and coupler segmentation identification on the multi-source standard hook removal operation data stream through the deep vision recognition module to obtain a target coupler mask area; Performing point cloud projection and three-dimensional reconstruction on the pixel point cloud data in the target coupler mask area to generate a three-dimensional model of the coupler state; Constructing a three-dimensional operation coordinate, performing coordinate mapping conversion on the three-dimensional model of the coupler state based on the three-dimensional operation coordinate, and outputting coupler position information; Boundary detection and coupler state evaluation are performed on the three-dimensional coupler state model to obtain a target coupler state.

6. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 4, characterized in that: The abnormal operation condition is obtained, including: Performing point cloud shape analysis and local curvature calculation based on the target coupler state to obtain local curvature information of the coupler; Comparing the local curvature information of the coupler according to the coupler curvature threshold to determine deformation, thereby obtaining coupler deformation detection information; determining a coupler operation position threshold according to the target coupler state; The coupler position information is subjected to deviation calculation based on the coupler operating position threshold to determine the operating position deviation, and the abnormal operating condition is obtained according to the coupler deformation detection information and the operating position deviation.

7. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 1, characterized in that: Determining target decoupling strategy parameters includes: Determine the coupling uncoupling coding element information according to the current working condition and coupler characteristic information; Using the decoupling coding element information to perform element coding on the decoupling strategy database to obtain a decoupling strategy coding database; Based on the current working condition and coupler feature information, a traversal search and matching is performed in the unhooking strategy database to obtain an unhooking strategy fitness set; The unhooking strategy database is subjected to adaptation screening and adaptive adjustment according to the unhooking strategy adaptability set to determine target unhooking strategy parameters.

8. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 7, characterized in that: Determining target decoupling strategy parameters includes: Adapting and sorting the unhooking strategy database according to the unhooking strategy adaptability set to obtain an adapted unhooking strategy parameter sequence; Optimizing the adaptive unhooking strategy parameter sequence to determine initial unhooking strategy parameters; Based on the initial unhooking strategy parameters, simulation deviation analysis and adaptive adjustment are performed to obtain target unhooking strategy parameters.

9. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 8, characterized in that: The target unhooking strategy parameters are obtained, including: Obtaining a set of evaluation indicators for the effect of hook removal, performing simulation fitting on the hook removal strategy database based on the set of evaluation indicators for the effect of hook removal, and constructing a hook removal effect simulation module; Performing simulation evaluation on the initial unhooking strategy parameters based on the unhooking effect simulation module to obtain an initial unhooking fitting effect; Based on the initial unhooking fitting effect, the initial unhooking strategy parameters are adaptively biased to obtain target unhooking strategy parameters.

10. The adaptive control method for coupler and uncoupler robot operations under multiple working conditions according to claim 9, characterized in that: The target unhooking strategy parameters are obtained, including: The deviation information between the initial hook removal fitting effect and the preset hook removal effect is used as the hook removal optimization direction; According to the unhooking optimization direction, the initial unhooking strategy parameters are adaptively adjusted and compared to find the best, and the target unhooking strategy parameters are determined.

Images