Multi-vehicle adaptive unhooking robot working condition self-adaptive adjustment method and system

Abstract

The invention discloses a working condition self-adaptive adjustment method and system for a multi-vehicle adaptive unhooking robot, and relates to the technical field of self-adaptive adjustment methods. Determining an unhooking behavior event of the multi-vehicle adaptive unhooking robot relative to the unhooking feature according to the unhooking working area, the position of the unhooking feature and an unhooking working support arm of the multi-vehicle adaptive unhooking robot; and determining a self-adaptive unhooking event of the multi-vehicle adaptive unhooking robot according to each unhooking behavior event and the moving path of the multi-vehicle adaptive unhooking robot. Therefore, the self-adaptive unhooking behavior of the multi-vehicle adaptive unhooking robot is determined based on each working condition type and the corresponding self-adaptive unhooking event; the corresponding unhooking work efficiency is estimated based on all the self-adaptive unhooking behaviors, the remaining unhooking work systems of the multi-vehicle adaptive unhooking robot are determined according to the unhooking work efficiency, the corresponding unhooking positions and the remaining work content of the multi-vehicle adaptive unhooking robot, and the accuracy of the remaining unhooking work systems is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of adaptive adjustment methods, and in particular to a multi-vehicle adaptive hooking robot working condition adaptive adjustment method and system. BACKGROUND

[0002] With the development of technology, the multi-vehicle adaptive hooking robot is a robot system that can adapt to multiple vehicle types and automatically complete hooking operations. In the prior art, the multi-vehicle adaptive hooking robot performs hooking operations at the corresponding hooking position and around the hooking work area, without considering the movement path of the multi-vehicle adaptive hooking robot, affecting the accuracy of the adaptive hooking event of the multi-vehicle adaptive hooking robot, and without adaptive control of the hooking behavior, affecting the accuracy of the remaining hooking work system. SUMMARY

[0003] The present application aims to overcome the shortcomings of the prior art, and provides a multi-vehicle adaptive hooking robot working condition adaptive adjustment method and system.

[0004] The present application provides a multi-vehicle adaptive hooking robot working condition adaptive adjustment method, which comprises: determining a plurality of region nodes according to the movement path of the multi-vehicle adaptive hooking robot, and collecting a plurality of region work images according to the detection of the plurality of region nodes; determining a plurality of environment features, dynamic features and hooking features based on image recognition of each region work image, and determining a hooking work area according to the plurality of environment features, dynamic features and hooking features; determining a hooking behavior event of the multi-vehicle adaptive hooking robot relative to the hooking feature according to the hooking work area, the position of the hooking feature and the hooking work arm of the multi-vehicle adaptive hooking robot, determining an adaptive hooking event of the multi-vehicle adaptive hooking robot according to each hooking behavior event and the movement path of the multi-vehicle adaptive hooking robot; determining a working condition type of the multi-vehicle adaptive hooking robot according to each hooking position and the plurality of environment features, determining an adaptive hooking behavior of the multi-vehicle adaptive hooking robot based on each working condition type and the corresponding adaptive hooking event; estimating the corresponding hooking work efficiency based on each adaptive hooking behavior, and determining a remaining hooking work system of the multi-vehicle adaptive hooking robot according to the hooking work efficiency, the corresponding hooking position and the remaining work content of the multi-vehicle adaptive hooking robot.

[0005] The present application provides a multi-vehicle adaptive hooking robot working condition adaptive adjustment system, which is applied to the above-mentioned multi-vehicle adaptive hooking robot working condition adaptive adjustment method, and comprises:

[0006] A regional working image module is configured to determine a plurality of regional nodes according to a moving path of the multi-vehicle adaptive hooking robot, and to acquire a plurality of regional working images according to detection of the plurality of regional nodes;

[0007] A multi-modal data module is configured to determine a plurality of environment features, dynamic features and hooking features based on image recognition of the regional working images, and to determine a hooking working region according to the plurality of environment features, dynamic features and hooking features;

[0008] An adaptive hooking event module is configured to determine a hooking behavior event of the multi-vehicle adaptive hooking robot relative to the hooking feature according to the hooking working region, a position of the hooking feature and a hooking working branch arm of the multi-vehicle adaptive hooking robot, and to determine an adaptive hooking event of the multi-vehicle adaptive hooking robot according to the hooking behavior event and the moving path of the multi-vehicle adaptive hooking robot;

[0009] An adaptive hooking behavior module is configured to determine a working condition type of the multi-vehicle adaptive hooking robot according to the hooking position and the plurality of environment features, and to determine an adaptive hooking behavior of the multi-vehicle adaptive hooking robot based on the working condition type and the corresponding adaptive hooking event;

[0010] A remaining hooking working system module is configured to estimate a corresponding hooking working efficiency based on the adaptive hooking behavior, and to determine a remaining hooking working system of the multi-vehicle adaptive hooking robot according to the hooking working efficiency, the corresponding hooking position and a remaining working content of the multi-vehicle adaptive hooking robot.

[0011] Compared with the prior art, the present application has the following advantages:

[0012] In the embodiment of the present application, the method in the embodiment of the present application is used to determine a hooking behavior event of the multi-vehicle adaptive hooking robot relative to the hooking feature according to the hooking working region, a position of the hooking feature and a hooking working branch arm of the multi-vehicle adaptive hooking robot, and to determine an adaptive hooking event of the multi-vehicle adaptive hooking robot according to the hooking behavior event and the moving path of the multi-vehicle adaptive hooking robot, which introduces the hooking working region, and takes into account the overall consideration of the hooking behavior event and the moving path of the multi-vehicle adaptive hooking robot, thereby improving the accuracy of the adaptive hooking event of the multi-vehicle adaptive hooking robot.

[0013] Therefore, the working condition type of the multi-vehicle adaptive hooking robot is determined according to various hooking positions and multiple environmental characteristics, the adaptive hooking behavior of the multi-vehicle adaptive hooking robot is determined based on various working condition types and corresponding adaptive hooking events, the corresponding hooking work efficiency is estimated based on various adaptive hooking behaviors, and the remaining hooking work system of the multi-vehicle adaptive hooking robot is determined according to the hooking work efficiency, the corresponding hooking position and the remaining work content of the multi-vehicle adaptive hooking robot. The adaptive hooking behavior of the multi-vehicle adaptive hooking robot is introduced, the overall consideration of the hooking work efficiency, the corresponding hooking position and the remaining work content of the multi-vehicle adaptive hooking robot is realized, and the accuracy of the remaining hooking work system is improved. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application;

[0015] Figure 2 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application;

[0016] Figure 3 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application;

[0017] Figure 4 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application;

[0018] Figure 5 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application;

[0019] Figure 6 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application;

[0020] Figure 7 FIG. 1 is a flowchart of a multi-vehicle adaptive hooking robot working condition adaptive adjustment method in an embodiment of the present application; DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0022] Please refer to Figures 1 to 7 A multi-vehicle adaptive hooking robot working condition adaptive adjustment method applied to an adaptive adjustment scene; the multi-vehicle adaptive hooking robot working condition adaptive adjustment method comprises:

[0023] Step S11: Determine a plurality of region nodes according to the movement path of the multi-vehicle adaptive hooking robot, and collect a plurality of region working images according to the detection of the plurality of region nodes;

[0024] Step S12: Determine a plurality of environment features, dynamic features and hooking features based on image recognition of each region working image, and determine a hooking working region according to the plurality of environment features, dynamic features and hooking features;

[0025] Step S13: Determine a hooking behavior event of the multi-vehicle adaptive hooking robot relative to the hooking feature according to the hooking working region, the position of the hooking feature and the hooking working arm of the multi-vehicle adaptive hooking robot, and determine an adaptive hooking event of the multi-vehicle adaptive hooking robot according to each hooking behavior event and the movement path of the multi-vehicle adaptive hooking robot;

[0026] Step S14: Determine a working condition type of the multi-vehicle adaptive hooking robot according to each hooking position and the plurality of environment features, and determine an adaptive hooking behavior of the multi-vehicle adaptive hooking robot based on each working condition type and the corresponding adaptive hooking event;

[0027] Step S15: Estimate the corresponding hooking working efficiency based on each adaptive hooking behavior, and determine a remaining hooking working system of the multi-vehicle adaptive hooking robot according to the hooking working efficiency, the corresponding hooking position and the remaining working content of the multi-vehicle adaptive hooking robot.

[0028] Reference Figure 2 In step S11, the specific steps are as follows:

[0029] S111: Collect the current position and target position of the multi-vehicle adaptive hooking robot, determine the movement path of the multi-vehicle adaptive hooking robot according to the current position, target position and corresponding movement space, determine a plurality of hooking working regions based on the detection of the movement path of the multi-vehicle adaptive hooking robot, and mark the region nodes of each hooking working region;

[0030] S112: In each region node, trigger a plurality of surrounding cameras according to the node position of the region node, the plurality of cameras perform ring shooting on the hooking working region where the region node is located, and collect a plurality of sub-region images, and determine the corresponding region working image according to the plurality of sub-region images, the corresponding hooking position and the past positioning position of the multi-vehicle adaptive hooking robot.

[0031] In the embodiments of the present application, the accurate coordinate position of the robot is obtained in real time through the positioning system (such as GPS, laser radar SLAM or visual positioning system) built in the robot, and the target position is obtained through the task scheduling system or manual input, which usually includes three-dimensional space coordinates and attitude information.

[0032] According to the current position, target position and corresponding movement space, the movement path of the multi-vehicle adaptive hooking robot is determined, and the movement space refers to the three-dimensional space range that the robot can operate, including physical limitations, safety boundaries and operation constraints; the path planning algorithm (such as RRT or dynamic window method) will comprehensively consider these factors to calculate the optimal path from the current position to the target position, and the planned movement path will be segmented into multiple continuous work areas according to certain rules (such as distance interval, functional area or environmental characteristics); each area represents a relatively independent hooking operation unit; a unique identification node is set for each work area, containing area ID, spatial coordinates, range boundary and other information.

[0033] Further, when the hooking robot reaches or approaches the preset area node, the system will automatically trigger the multiple cameras deployed around the node; the triggering mechanism is based on distance threshold judgment, and generally when the distance between the robot and the node is less than a preset value (such as 5 meters), the system will activate the camera network; the camera triggering needs to consider time synchronization to ensure that all cameras start collecting at the same time point, avoiding the time difference of image data affecting subsequent processing; at the same time, the triggering mechanism needs to have fault tolerance to prevent false triggering or missed triggering.

[0034] The multiple cameras perform ring-shaped shooting on the hooking work area where the area node is located, and the ring-shaped shooting refers to the synchronous shooting of multiple cameras from different angles (usually 360-degree full coverage) on the same work area; the cameras are usually distributed around the work area at fixed intervals (such as 45 degrees or 60 degrees) to form a non-blind observation network; the camera layout needs to consider the overlapping area to ensure that the fields of view of adjacent cameras have appropriate overlap, which is convenient for subsequent image stitching; at the same time, the focal length and shooting parameters of the cameras need to be adjusted according to the size of the work area to ensure that the image resolution meets the identification requirements.

[0035] Each camera is responsible for collecting sub-area images within its field of view; these sub-area images are raw image data that contain visual information of the work area from different angles; image collection usually uses high-resolution formats and includes timestamp and camera position information; image collection needs to consider changes in lighting conditions and needs to automatically adjust exposure parameters; at the same time, the image data volume is usually large, and efficient compression and transmission mechanisms are needed to ensure data real-time performance.

[0036] According to the plurality of sub-region images, the corresponding hooking position and the past positioning position of the multi-vehicle adaptive hooking robot, a corresponding region working image is determined, and the plurality of sub-region images are spliced into a complete region working image; the system will combine the hooking position (the specific position currently needing to be operated) and the past positioning position (the historical trajectory data) of the robot to optimize the image, and enhance the image quality of the key region; the image splicing needs accurate coordinate transformation and color correction to ensure seamless connection of the spliced image; meanwhile, the system will enhance the key region in the image according to the hooking position and the historical position of the robot, to improve the accuracy of subsequent feature extraction.

[0037] Reference Figure 3 In step S12, the specific steps are as follows:

[0038] S121: In each region working image, a plurality of sub-feature images are determined according to the recognition of the region working image, a corresponding image recognition mode is determined according to the position, image contour and image type weight of the plurality of sub-feature images, and different types of sub-feature images are recognized by using the corresponding image recognition mode;

[0039] S122: The corresponding image recognition mode is executed on the plurality of sub-feature images, and a corresponding feature combination is output, and there are a plurality of feature combinations in the region working image;

[0040] S123: A plurality of environment features, dynamic features and hooking features are determined according to the screening of the plurality of feature combinations, a preliminary region is constructed based on the plurality of environment features and dynamic features, and a hooking working region is determined according to the preliminary region and the hooking features.

[0041] In the embodiment of the present application, in each region working image, a plurality of sub-feature images are determined according to the recognition of the region working image, the system first pre-processes the region working image, including image enhancement, noise removal, contrast adjustment and other operations, to improve the image quality; the complete region working image is segmented into a plurality of sub-regions with independent features through an image segmentation algorithm (such as watershed segmentation, region growing, edge detection, etc.); each sub-feature image represents a potential object or region of interest, such as a car connector, a ground marker, an operating device, etc.; the system will assign a unique identifier to each sub-feature image, and record its position coordinates, size and other basic information in the original image; the selection of the image segmentation algorithm is based on the image characteristics and the processing speed requirement, and usually a combination of multiple algorithms is used to improve the segmentation accuracy; the minimum size of the sub-feature image has a threshold limit to avoid waste of computing resources caused by excessive segmentation, and at the same time, the texture, color, edge and other multi-dimensional features of the image are considered in the segmentation process to ensure the semantic integrity of the segmentation result.

[0042] For each sub-feature image, the system analyzes three key parameters: position information, image contour and image type weight; Position information: including the absolute coordinates and relative position relationship of the sub-feature image in the regional working image, which is used to judge the spatial importance of the feature and its correlation with other features; Image contour: The boundary shape, aspect ratio, roundness, complexity and other geometric features of the sub-feature image are obtained through the contour extraction algorithm. These features reflect the basic form of the target object; Image type weight: The system pre-establishes a feature type weight table, and different types of features have different processing priorities and importance coefficients; for example, the weight of the car connector feature is 0.9, while the weight of the ground marking feature is 0.6; Based on these three parameters, the system determines the recognition mode that best suits the sub-feature image through a decision matrix or machine learning model; Recognition modes include: template matching (applicable to fixed shape features), deep learning recognition (applicable to complex deformation features), edge feature analysis (applicable to features with obvious contours), texture analysis (applicable to objects with rich surface features), etc.

[0043] After assigning a determined recognition mode to each sub-feature image, the system performs multiple recognition tasks in parallel; different recognition modes have different input requirements and output formats: Template matching mode: input sub-feature image and template library, output matching score and best matching position; Deep learning mode: input normalized sub-feature image, output classification probability and feature vector; Edge feature analysis: input edge detection result, output geometric parameters; Texture analysis: input grayscale co-occurrence matrix, output texture feature parameters; Recognition results will undergo confidence evaluation, and only recognition results that reach the preset threshold will be accepted; The execution of recognition mode adopts multi-threading or distributed computing to improve processing efficiency; The system has mode switching capability, and when the recognition effect of a certain mode is not good, it can automatically try the backup mode; The recognition results contain rich metadata, such as processing time, confidence, feature parameters, etc., for use in subsequent steps.

[0044] Furthermore, corresponding image recognition modes are executed on the plurality of sub-feature images, and corresponding feature combinations are outputted, and there are a plurality of feature combinations in the regional working image, and a plurality of feature combinations are introduced.

[0045] At this time, the system calls the corresponding recognition algorithm for processing according to the recognition mode of each sub-feature image determined in S121; the recognition process includes feature extraction, pattern matching, classification judgment and other links, and different recognition modes have their specific processing procedures; the system will process multiple sub-feature images in parallel to improve the overall processing efficiency and ensure reasonable resource allocation between different recognition tasks; during the execution of each recognition mode, the system will record processing time, confidence, intermediate results and other key information for subsequent quality evaluation; parallel processing uses multi-threading or distributed computing architecture to dynamically allocate computing resources according to the complexity of the recognition mode; the recognition algorithm will be adaptively adjusted based on historical data, such as parameter fine-tuning of deep learning models, threshold optimization of template matching, etc.

[0046] After each sub-feature image is recognized, one or more feature vectors will be output, which contain the key information of the sub-feature; the system will combine related feature vectors to form a feature combination; the basis for feature combination includes spatial proximity, functional correlation, temporal consistency, etc.; each feature combination contains multiple original feature vectors, as well as their association relationships and combination weights; the feature combination will be checked for redundancy to remove duplicate or highly correlated features, ensuring the simplicity and effectiveness of the combination; the feature combination uses graph theory or clustering algorithms to effectively combine spatially adjacent or functionally related features; the timeliness of the features will be considered during the combination process, with the latest acquired feature data being preferentially selected; the system will calculate the comprehensive confidence of each feature combination, which will be an important basis for subsequent screening.

[0047] A complete area work image usually contains multiple feature combinations, each representing a specific scene element or functional area; the system will maintain the spatial and logical relationships between these feature combinations to form a complete scene understanding; the feature combinations will be verified for spatial consistency to ensure that their distribution in the area work image conforms to the physical laws of the actual scene; the system will regularly update the feature combinations to adapt to the dynamic changes of the scene and maintain the timeliness of the feature combinations; the management of feature combinations uses spatial indexing techniques such as R-tree or quad-tree to improve the efficiency of spatial queries and relationship analysis; the system will establish a semantic network between feature combinations to represent their functional dependencies and logical relationships; the dynamic update mechanism uses an incremental algorithm to only update the feature combinations that have changed, reducing computational overhead.

[0048] Therefore, multiple environmental features, dynamic features and hooking features are determined according to the screening of multiple feature combinations, a preliminary area is constructed based on the multiple environmental features and dynamic features, and a hooking work area is determined according to the preliminary area and the hooking features, which takes into account the overall consideration of the preliminary area and the hooking features, ensuring the accuracy of the hooking work area.

[0049] At this time, the system first screens and classifies the multiple feature combinations generated in S122; the screening is based on preset evaluation criteria, including the confidence, integrity, timeliness, etc. of the feature combination; the environmental feature refers to relatively stable scene elements, such as fixed equipment, ground markings, building structures, etc.; the system determines the environmental feature by analyzing the static elements and spatial relationships in the feature combination; the dynamic feature refers to elements that change or move in the scene, such as personnel activities, vehicle movements, changes in illumination, etc.; the system identifies the dynamic feature by comparing the changes in the feature combination of consecutive time frames; the hooking feature refers to specific elements directly related to the hooking operation, such as car connectors, locking devices, operation handles, etc.; the system determines the hooking feature through shape matching, position analysis, and function identification; the feature screening adopts a multi-level filtering mechanism, first filtering low-quality features based on a confidence threshold, then removing redundant features based on correlation analysis, and finally retaining key features based on importance evaluation; the environmental feature recognition adopts spatial consistency verification to ensure that the identified environmental features remain stable in multiple image frames; the dynamic feature recognition adopts motion detection algorithms such as optical flow, background difference, etc. to analyze the time series changes of the feature combination; the hooking feature recognition adopts a special template library and deep learning model to accurately match different types of connection devices.

[0050] The feature screening adopts a multi-level filtering mechanism, first filtering low-quality features based on a confidence threshold, then removing redundant features based on correlation analysis, and finally retaining key features based on importance evaluation; the environmental feature recognition adopts spatial consistency verification to ensure that the identified environmental features remain stable in multiple image frames; the dynamic feature recognition adopts motion detection algorithms such as optical flow, background difference, etc. to analyze the time series changes of the feature combination; the hooking feature recognition adopts a special template library and deep learning model to accurately match different types of connection devices.

[0051] The system analyzes the spatial relationship between each preliminary region and the hooking feature to determine the region most suitable for hooking operation; the determination of the hooking work region considers the position distribution of the hooking feature, the functional attributes of the preliminary region, and the operation safety requirements; the system will conduct risk assessment on the candidate hooking work region, considering factors including obstacle distribution, personnel activity, equipment accessibility, etc.; the finally determined hooking work region will contain key information such as precise spatial boundaries, operation entry points, and safe evacuation paths; the determination of the hooking work region adopts a multi-objective optimization algorithm to balance operation efficiency, safety, and accessibility.

[0052] Reference Figure 4 In step S13, the specific steps are:

[0053] S131: Collecting the hook removal work area, determining multiple hook removal work nodes based on the identification of the hook removal work area, and determining a position event based on the positions of the multiple hook removal work nodes and the positions of the hook removal features;

[0054] S132: determining an action event based on multiple unhooking work nodes and the unhooking work arm of the multi-vehicle adaptive unhooking robot, and determining an unhooking behavior event of the multi-vehicle adaptive unhooking robot relative to the unhooking feature based on the position event, the action event, and the working mode of the multi-vehicle adaptive unhooking robot relative to the unhooking work area;

[0055] S133: Collect the moving path of the multi-vehicle adaptive unhooking robot, determine the activity range of the corresponding unhooking behavior based on the detection of each unhooking behavior event, and determine the adaptive unhooking event of the multi-vehicle adaptive unhooking robot based on the activity range of the unhooking behavior, the moving path of the multi-vehicle adaptive unhooking robot and the unhooking posture of the multi-vehicle adaptive unhooking robot.

[0056] In an embodiment of the present application, the unhooking work area is output by step S123, including the three-dimensional spatial coordinate range, the position of the unhooking features (such as connectors, locks, etc.), and environmental structure information (such as obstacles, ground markings, etc.); the system uses visual sensors (such as cameras, lidars) to collect images and point cloud data of the unhooking work area in real time, and combines historical data to perform regional modeling; the collected data is converted into a three-dimensional spatial model, and key feature points, boundaries, obstacles and other information are marked to form a computable spatial structure.

[0057] In the unhooking work area, the system automatically generates multiple unhooking work nodes based on the spatial layout, unhooking feature distribution and robot operation accessibility; these nodes are the key spatial position points for the robot to perform operations; node type: approach node: the position where the robot approaches the unhooking feature, usually 1-2 meters away from the target; alignment node: the position where the robot is aligned with the unhooking feature, usually 0.5-1 meter away from the target; operation node: the position where the robot performs the unhooking operation, usually in direct contact with the target; spatial sampling, clustering or feature density-based methods are used to ensure that the nodes are reasonably distributed and cover all operation positions; node generation algorithm: spatial sampling, clustering or feature density-based methods are used to ensure that the nodes are reasonably distributed and cover all operation positions.

[0058] According to the positions of the multiple hooking work nodes and the positions of the hooking features, a position event is determined, the position event describing a relative position relationship between the robot and the hooking features, for triggering a subsequent action; for example: a proximity event: the robot reaches a proximity node, triggering a speed reduction and a visual alignment; an alignment event: the robot reaches an alignment node, triggering a fine adjustment of the branch arm; an operation event: the robot reaches an operation node, triggering a hooking action; the system calculates the spatial relationship (distance, angle, direction) between each hooking work node and the hooking feature, and generates a corresponding position event; the system calculates the spatial relationship (distance, angle, direction) between each hooking work node and the hooking feature, and generates a corresponding position event.

[0059] Specifically, the hooking work area is a railway freight yard, and the robot needs to perform hooking operations on three different types of carriages (A type, B type, and C type); the hooking features are as follows: the A type carriage: the connector is located in the middle of the carriage, with a height of 1.2 meters; the B type carriage: the connector is located at the end of the carriage, with a height of 1.5 meters; the C type carriage: the connector is located at the bottom of the carriage, with a height of 0.8 meters; the initial position of the robot is the yard coordinates (100, 200), and the branch arm is fully retracted.

[0060] The system collects point cloud and image data of the three carriages through a laser radar and a camera; a three-dimensional area model is generated, and the connector positions of the A, B, and C type carriages and surrounding obstacles (such as signal lights and support columns) are labeled.

[0061] The hooking work nodes are determined as follows: the A type carriage: the proximity node N1 is (150, 250), 1.5 meters away from the connector; the alignment node N2 is (152, 252), 0.8 meters away from the connector; the operation node N3 is (153, 253), directly contacting the connector; the B type carriage: the proximity node N4 is (200, 300), 1.8 meters away from the connector; the alignment node N5 is (202, 302), 1.0 meters away from the connector; the operation node N6 is (203, 303), directly contacting the connector; the C type carriage: the proximity node N7 is (250, 350), 1.2 meters away from the connector; the alignment node N8 is (252, 352), 0.6 meters away from the connector; the operation node N9 is (253, 353), directly contacting the connector.

[0062] Position event generation: A-type carriage: event E1: robot reaches N1, triggers "approach A-type connector"; event E2: robot reaches N2, triggers "align A-type connector"; event E3: robot reaches N3, triggers "operate A-type connector"; B-type carriage: event E4: robot reaches N4, triggers "approach B-type connector"; event E5: robot reaches N5, triggers "align B-type connector"; event E6: robot reaches N6, triggers "operate B-type connector"; C-type carriage: event E7: robot reaches N7, triggers "approach C-type connector"; event E8: robot reaches N8, triggers "align C-type connector"; event E9: robot reaches N9, triggers "operate C-type connector"; the system generates 9 position events (E1-E9), each event contains: trigger condition (robot reaches specified node); event type (approach, align, operate); associated hook feature (A / B / C type connector); these position events will be used as input for the S132 step to generate action events and hook behavior events.

[0063] Further, according to the hook working nodes and the hook working arms of the multi-vehicle adaptive hook robot, the action events are determined based on the position events, the action events, and the working mode of the multi-vehicle adaptive hook robot relative to the hook working area, the hook behavior events of the multi-vehicle adaptive hook robot relative to the hook feature are determined based on the position events, the action events, and the working mode of the multi-vehicle adaptive hook robot relative to the hook working area, the overall consideration of the position events, the action events, and the working mode of the multi-vehicle adaptive hook robot relative to the hook working area is compatible, and the accuracy of the hook behavior events of the multi-vehicle adaptive hook robot relative to the hook feature is ensured.

[0064] At this time, the multiple hook working nodes (such as approach nodes, alignment nodes, and operation nodes) output by the S131 step are received, each node contains a spatial coordinate and an associated hook feature; the system calculates the reachability, pose, and motion trajectory of the arm at each node according to the arm structure (such as a multi-joint robot arm, a linear actuator, etc.) and the kinematic model of the robot; the arm actions include: stretching, rotating, clamping, and releasing, etc., each action corresponds to specific control parameters (such as angle, speed, torque);

[0065] For each hook working node, the system generates a corresponding action event to describe the specific action that needs to be performed by the arm at the node; the action event contains: action type (such as "arm stretching" and "claw closing"); action parameters (such as stretching length and clamping force); trigger condition (such as "reaching the alignment node"); execution time (such as "action lasts for 2 seconds"); force feedback control is introduced to limit the torque of clamping, releasing, and other actions to avoid damaging the equipment.

[0066] The hooking behavior event is determined based on the position event, the action event, and the working mode. The position event input is the position event generated in S131 (e.g., "arriving at the approaching node" and "arriving at the operating node"). The action event input is the action event generated in this step (e.g., "extending the arm" and "closing the gripper"). The system selects the working mode according to the current working condition (e.g., vehicle type, environmental condition, and task priority). For example, the standard mode is suitable for regular vehicles, the action is smooth, and efficiency is prioritized; the fine mode is suitable for precision vehicles, the action is slow, and precision is prioritized; and the emergency mode is suitable for abnormal situations, the action is fast, and safety is prioritized.

[0067] The system combines the position event, the action event, and the working mode to generate the hooking behavior event. The hooking behavior event describes the complete behavior logic of the robot performing a specific action in a specific position and mode. The behavior event includes the behavior type (e.g., "approaching the hook", "aligning the hook", and "executing the hook"), the trigger condition (position event + action event), the working mode (e.g., "standard mode"), the behavior parameters (e.g., action speed and torque limit), and the subsequent behavior (e.g., "returning to the standby point after completion").

[0068] Specifically, the multi-vehicle adaptive hooking robot needs to perform hooking operations in the hooking working areas of three vehicle types (A, B, and C). The hooking features (connectors) of each vehicle type are different in position and structure, requiring different arm actions. The system uses two working modes: "standard mode" and "fine mode". The hooking working node input (from S131) is as follows: for the A-type vehicle, the approaching node N1 is (100, 200), the aligning node N2 is (101, 201), and the operating node N3 is (102, 202); for the B-type vehicle, the approaching node N4 is (150, 250), the aligning node N5 is (151, 251), and the operating node N6 is (152, 252); and for the C-type vehicle, the approaching node N7 is (200, 300), the aligning node N8 is (201, 301), and the operating node N9 is (202, 302).

[0069] Action event generation: Type A carriage: Action event Al: boom extension to length 1.2 m (trigger condition: reaching Nl); Action event A2: boom rotation to angle 30 degrees (trigger condition: reaching N2); Action event A3: gripper closing, torque 10 N-m (trigger condition: reaching N3); Type B carriage: Action event Bl: boom extension to length 1.5 m (trigger condition: reaching N4); Action event B2: boom rotation to angle 45 degrees (trigger condition: reaching N5); Action event B3: gripper closing, torque 15 N-m (trigger condition: reaching N6); Type C carriage: Action event Cl: boom extension to length 1.8 m (trigger condition: reaching N7); Action event C2: boom rotation to angle 60 degrees (trigger condition: reaching N8); Action event C3: gripper closing, torque 20 N-m (trigger condition: reaching N9).

[0070] Hooking behavior event generation: Type A carriage (standard mode): Behavior event BE1: approach hooking (trigger condition: position event El + action event Al); Behavior event BE2: align hooking (trigger condition: position event E2 + action event A2); Behavior event BE3: perform hooking (trigger condition: position event E3 + action event A3); Type B carriage (fine mode): Behavior event BE4: approach hooking (trigger condition: position event E4 + action event Bl, action speed reduced by 20%); Behavior event BE5: align hooking (trigger condition: position event E5 + action event B2, action speed reduced by 20%); Behavior event BE6: perform hooking (trigger condition: position event E6 + action event B3, torque limit increased by 10%); Type C carriage (standard mode): Behavior event BE7: approach hooking (trigger condition: position event E7 + action event Cl); Behavior event BE8: align hooking (trigger condition: position event E8 + action event C2); Behavior event BE9: perform hooking (trigger condition: position event E9 + action event C3).

[0071] Output result: The system generates 9 hooking behavior events (BE1-BE9), each event contains: trigger condition (position event + action event); behavior type (approach, align, perform); working mode (standard / fine); behavior parameters (speed, torque, etc.); these behavior events will be used as input for the adaptive hooking event determination in step S133.

[0072] Therefore, the moving path of the multi-vehicle adaptive hooking robot is collected, the activity range of the corresponding hooking behavior is determined according to the detection of each hooking behavior event, the adaptive hooking event of the multi-vehicle adaptive hooking robot is determined according to the activity range of the hooking behavior, the moving path of the multi-vehicle adaptive hooking robot and the hooking posture of the multi-vehicle adaptive hooking robot, the overall consideration of the activity range of the hooking behavior, the moving path of the multi-vehicle adaptive hooking robot and the hooking posture of the multi-vehicle adaptive hooking robot is compatible, the accuracy of the adaptive hooking event of the multi-vehicle adaptive hooking robot is ensured, at the same time, the hooking working area is introduced, the overall consideration of each hooking behavior event and the moving path of the multi-vehicle adaptive hooking robot is compatible, and the accuracy of the adaptive hooking event of the multi-vehicle adaptive hooking robot is improved.

[0073] At this time, the moving path of the multi-vehicle adaptive hooking robot is collected, the hooking behavior events (such as approaching hooking, aligning hooking and executing hooking) generated in S132, each event corresponds to specific trigger conditions and behavior parameters; the system calculates the effective activity range of the behavior in space according to the type and trigger condition of the behavior event; the activity range includes: spatial range: the three-dimensional space region (such as a spherical or cubic region with the hooking feature as the center) where the behavior event is effective; time range: the duration of the behavior event (such as within 5 seconds from triggering to completion); dynamic range: the activity range adjusted according to the speed and posture of the robot (such as expanding the range when moving at high speed and reducing the range when moving at low speed); range calculation: using spatial geometry algorithms (such as convex hull algorithm, bounding box algorithm) to calculate the spatial range; using time series analysis to calculate the time range; using dynamic programming algorithm to adjust the dynamic range.

[0074] According to the activity range, the moving path and the hooking posture, the adaptive hooking event is determined, the activity range: the spatial, temporal and dynamic range of each hooking behavior event; the moving path: the actual moving track and the planned path of the robot; the hooking posture: the posture of the arm, the state of the gripper, the moving direction and the like of the robot.

[0075] The system compares the moving path of the robot with the activity range in real time to determine whether the robot enters the effective range of a certain behavior event; when the robot enters the activity range, the system generates an adaptive hooking event according to the hooking posture and the behavior parameters; the adaptive hooking event includes: event type (such as adaptive approaching hooking and adaptive aligning hooking); trigger condition (such as entering the activity range and the posture meeting the requirements); execution action (such as adjusting the angle of the arm and reducing the moving speed); abnormal handling (such as stopping the action when the activity range is exceeded); event optimization: using reinforcement learning algorithm to optimize the adaptive hooking event, improving the success rate and efficiency of the event; introducing multi-objective optimization to balance the hooking speed, accuracy and safety.

[0076] Specifically, assume that the multi-car fitting uncoupling robot needs to perform uncoupling operations on three different types of carriages (A / B / C types) in the railway freight yard, collect the moving path, and path data: starting point: freight yard entrance (X=0, Y=0, Z=0); target point: C-type carriage position (X=100, Y=50, Z=0); path node: N1: entrance turning point (X=20, Y=10, Z=0); N2: near A-type carriage (X=40, Y=20, Z=0); N3: near B-type carriage (X=60, Y=30, Z=0); N4: near C-type carriage (X=80, Y=40, Z=0); speed curve: 0-20m: accelerate to 1m / s; 20-80m: constant speed 1m / s; 80-100m: decelerate to 0.2m / s.

[0077] Behavior events (take C-type carriage as an example): BE7: approach uncoupling (trigger condition: position event E7 action event C1); BE8: align uncoupling (trigger condition: position event E8 action event C2); BE9: execute uncoupling (trigger condition: position event E9 action event C3); activity range calculation: BE7 activity range: spatial range: spherical area with N4 as center and radius of 5 meters; time range: within 10 seconds from entering N4 to N8; dynamic range: range radius of 5 meters at speed of 1m / s, range radius of 2 meters at speed of 0.2m / s; BE8 activity range: spatial range: spherical area with N8 as center and radius of 1 meter; time range: within 5 seconds from reaching N8 to starting uncoupling; dynamic range: adjust range according to arm posture; BE9 activity range: spatial range: spherical area with N9 as center and radius of 0.5 meter; time range: within 3 seconds of uncoupling action duration; dynamic range: adjust range according to gripper state.

[0078] Path and activity range comparison: the robot moves to N4 (X=80, Y=40, Z=0) and enters the activity range of BE7; the system detects that the robot speed is 0.5m / s and the posture is that the arm is stretched by 30 degrees; adaptive uncoupling event generation: event AE1: adaptive approach uncoupling trigger condition: enter BE7 activity range and speed <1m / s; execute action: reduce speed to 0.2m / s and stretch arm to 45 degrees; exception handling: if it exceeds the BE7 range, stop action and alarm; the robot continues to move to N8 (X=90, Y=45, Z=0) and enters the activity range of BE8; the system detects that the robot speed is 0.2m / s and the posture is that the arm is aligned with the C-type connector.

[0079] Adaptive hooking event generation: Event AE2: Adaptive align hooking; Trigger condition: Enter BE8 active range and branch arm align connector; Perform action: Fine-tune branch arm angle to ensure alignment accuracy; Exception handling: If alignment fails, re-plan path; Robot moves to N9 (X=95, Y=48, Z=0) and enters the active range of BE9; System detects that the robot speed is 0 m / s and the posture is gripper closed; Adaptive hooking event generation: Event AE3: Adaptive perform hooking; Trigger condition: Enter BE9 active range and gripper closed; Perform action: Gripper open, complete hooking; Exception handling: If hooking fails, increase clamping force and retry.

[0080] Output result: The system generates 3 adaptive hooking events (AE1-AE3), each event contains: trigger condition (enter active range + posture / velocity requirement); Perform action (speed adjustment, branch arm control, gripper operation); Exception handling strategy (stop, alarm, retry).

[0081] Reference Figure 5 In step S14, the specific steps are:

[0082] S141: Collect each hooking position, and determine the corresponding hooking environment area according to each hooking position and the corresponding hooking feature, determine a plurality of environment parameters according to the real-time detection of the hooking environment area, and determine a plurality of environment features according to the positions of the plurality of environment parameters, the parameter types, and the shape of the hooking environment area;

[0083] S142: Determine the working condition environment of the multi-vehicle adaptive hooking robot according to the plurality of environment features and the surrounding space of each hooking position, and mark the working condition type of the multi-vehicle adaptive hooking robot according to the working condition environment;

[0084] S143: Determine the first behavior trajectory according to each working condition type and hooking position, determine the second behavior trajectory according to each working condition type and the corresponding adaptive hooking event, and determine the adaptive hooking behavior of the multi-vehicle adaptive hooking robot according to the first behavior trajectory, the second behavior trajectory, and the hooking behavior matching table.

[0085] In the embodiment of the present application, each hooking position is collected, and the corresponding hooking environment area is determined according to each hooking position and the corresponding hooking feature, a plurality of environment parameters are determined according to the real-time detection of the hooking environment area, and a plurality of environment features are determined according to the positions of the plurality of environment parameters, the parameter types, and the shape of the hooking environment area. The overall consideration of the positions of the plurality of environment parameters, the parameter types, and the shape of the hooking environment area is compatible, and the accuracy of the plurality of environment features is guaranteed.

[0086] At this time, the system dynamically constructs a three-dimensional space region, i.e., a "hooking environment region", based on the spatial coordinates of the hooking position and the geometric structure (such as size, shape, operation space requirement) of the hooking feature; the region shape can be spherical, cubic, cylindrical, etc., depending on the type of hooking feature and operation requirement; for example: A-type connector requires a larger operation space, and the region is set as a spherical region with a radius of 1.5 meters; C-type connector requires more delicate operation, and the region is set as a cubic region with a length of 0.8 meters, a width of 0.8 meters, and a height of 1.2 meters.

[0087] The system scans the hooking environment region in real time through multi-sensor fusion (such as RGB-D camera, laser radar, infrared sensor, ultrasonic sensor); the detection frequency is usually 10-30 Hz, ensuring the real-time and accuracy of the data; the system extracts the following types of environment parameters from the sensor data: spatial parameters: obstacle position (such as other vehicles, equipment, walls); ground flatness (calculate the inclination angle through point cloud data); spatial height limit (such as top pipeline, ceiling height); physical parameters: temperature (affecting the lubrication of the robot arm and the performance of the sensor); humidity (affecting the clarity of the vision system); light intensity (affecting the imaging quality of the camera); dynamic parameters: speed and direction of moving objects (such as people walking, vehicle moving); wind speed (affecting the stability of the robot arm).

[0088] The system generates multiple environment features through feature extraction algorithm according to the spatial distribution of environment parameters, parameter type and geometric shape of hooking environment region; the feature generation method includes: clustering obstacle position into "dense area" or "sparse area"; calculating the mean and variance of ground flatness to determine whether it is "flat" or "unflat"; determining whether it is "strong light" or "weak light" according to the light intensity threshold (such as 500 lux); typical environment features: obstacle density (high / medium / low); ground flatness (flat / unflat); light condition (strong light / weak light / no light); dynamic interference degree (high / medium / low); spatial openness (open / semi-open / closed).

[0089] Specifically, in a certain railway freight station, a multi-car adaptive hooking robot needs to perform hooking operation on three sections of different types of train cars: hooking position 1: A-type connector, coordinates (10, 5, 1.2); hooking position 2: B-type connector, coordinates (15, 8, 1.0); hooking position 3: C-type connector, coordinates (20, 12, 1.5).

[0090] The system collects three hook removal locations and determines the environmental areas respectively: Type A connector: spherical area, radius 1.5 meters; Type B connector: cylindrical area, radius 1.0 meters, height 1.5 meters; Type C connector: cubic area, length and width 0.8 meters, height 1.2 meters; The system scans the three areas using lidar and camera and obtains the following parameters: Area 1 (Type A): Obstacle location: 2 toolboxes are located at (9.5, 4.8, 0) and (10.5, 5.2, 0); Ground flatness: 2° inclination angle; Light intensity: 800 lux; Dynamic objects: None; Area 2 (Type B): Obstacle location: 1 mobile cart at (14.8, 7.9, 0), speed 0.5 m / s; Ground flatness: 5° inclination angle; Light intensity: 300 lux; Dynamic objects: Mobile cart; Area 3 (Type C): Obstacle location: None; Ground flatness: 1° inclination angle; Light intensity: 100 lux; Dynamic objects: None.

[0091] The system generates environmental features based on the parameters: Area 1: Obstacle density: medium (2 obstacles); Ground flatness: flat (tilt angle <3°); Lighting conditions: strong light (>500lux); Dynamic interference level: low (no moving objects); Area 2: Obstacle density: low (1 obstacle); Ground flatness: uneven (tilt angle >3°); Lighting conditions: weak light (<500lux); Dynamic interference level: medium (with slow-moving objects); Area 3: Obstacle density: low (no obstacles); Ground flatness: flat (tilt angle <3°); Lighting conditions: no light (<200lux); Dynamic interference level: low (no moving objects); Output results: The system generates the following environmental feature vectors for the three unhooking positions: Position 1: [medium, flat, strong light, low]; Position 2: [low, uneven, weak light, medium]; Position 3: [low, flat, no light, low].

[0092] Furthermore, the working environment of the multi-vehicle adaptive unhooking robot is determined based on multiple environmental characteristics and the surrounding space of each unhooking position, and the working condition type of the multi-vehicle adaptive unhooking robot is marked according to the working environment, which is compatible with the overall consideration of multiple environmental characteristics and the surrounding space of each unhooking position, and ensures the accuracy of the working environment of the multi-vehicle adaptive unhooking robot.

[0093] At this time, multiple environmental features are generated by step S141, and each hook removal position corresponds to a set of environmental feature vectors, for example: Position 1: [Obstacle density: medium, ground flatness: flat, lighting conditions: strong light, dynamic interference level: low]; Position 2: [Obstacle density: low, ground flatness: uneven, lighting conditions: weak light, dynamic interference level: medium]; Position 3: [Obstacle density: low, ground flatness: flat, lighting conditions: no light, dynamic interference level: low].

[0094] The system analyzes the spatial structure around each hooking position, including: spatial freedom: whether there is enough operating space (such as whether the top and sides are limited); obstacle distribution: the location, size, and whether the obstacle is movable; ground conditions: ground slope, friction coefficient, whether there are steps or potholes; dynamic factors: whether there are moving vehicles, personnel, or other equipment; the system fuses environmental features with surrounding space information to construct a multi-dimensional work condition environment description; the work condition environment includes: static environment: elements that do not change (such as walls, fixed equipment); dynamic environment: elements that change (such as moving vehicles, changes in lighting); operating constraints: restrictions on robot motion (such as minimum operating distance, maximum allowed speed).

[0095] The system presets multiple work condition types, each corresponding to specific environmental conditions and operating requirements; common work condition types include: standard work condition: good environment, no significant obstacles, sufficient light, flat ground; complex work condition: presence of obstacles or dynamic interference, weak light, slightly uneven ground; severe work condition: dense obstacles, extremely poor light, severely uneven ground, high-frequency dynamic interference.

[0096] The system automatically labels the work condition type based on the comprehensive score of the work condition environment; an example of the scoring rule: standard work condition: obstacle density: low; ground flatness: flat; lighting conditions: strong light; dynamic interference degree: low; complex work condition: obstacle density: medium or low; ground flatness: flat or uneven; lighting conditions: weak light; dynamic interference degree: medium; severe work condition: obstacle density: high; ground flatness: uneven; lighting conditions: no light; dynamic interference degree: high; labeling process: the system quantitatively scores each environmental feature (such as low = 1 point, medium = 2 points, high = 3 points); calculate the total score of the work condition environment; map the total score to the corresponding work condition type according to the total score range.

[0097] Specifically, the multi-vehicle adapted hooking robot needs to perform the hooking task at three hooking positions (P1, P2, P3); the environmental features of each position are as follows: P1: obstacle density: medium; ground flatness: flat; lighting conditions: strong light; dynamic interference degree: low; P2: obstacle density: low; ground flatness: uneven; lighting conditions: weak light; dynamic interference degree: medium; P3: obstacle density: low; ground flatness: flat; lighting conditions: no light; dynamic interference degree: low.

[0098] Working condition environment determination: P1: surrounding space: 2 fixed obstacles, flat ground, no dynamic interference; working condition environment: static environment, less operation constraints; P2: surrounding space: 1 movable obstacle, 5-degree inclined ground, low-speed moving vehicle; working condition environment: dynamic environment, obstacle avoidance and low-speed operation required; P3: surrounding space: no obstacle, flat ground, but poor lighting; working condition environment: static environment, but relying on non-vision sensor.

[0099] Working condition type marking: P1: score: medium (2) + flat (1) + strong light (1) + low (1) = 5 points; working condition type: standard working condition; P2: score: low (1) + uneven (2) + weak light (2) + medium (2) = 7 points; working condition type: complex working condition; P3: score: low (1) + flat (1) + no light (3) + low (1) = 6 points; working condition type: complex working condition (special treatment required due to no light); the system generates the following working condition type markings for the three hooking positions: P1: standard working condition; P2: complex working condition; P3: complex working condition; these working condition types will be used as inputs for the S143 step to generate adaptive hooking behavior.

[0100] Therefore, the first behavior trajectory is determined according to each working condition type and hooking position, the second behavior trajectory is determined according to each working condition type and the corresponding adaptive hooking event, and the adaptive hooking behavior of the multi-vehicle adaptive hooking robot is determined according to the first behavior trajectory, the second behavior trajectory, and the hooking behavior matching table, which is compatible with the overall consideration of the first behavior trajectory, the second behavior trajectory, and the hooking behavior matching table, ensuring the accuracy of the adaptive hooking behavior of the multi-vehicle adaptive hooking robot.

[0101] At this time, the first behavior trajectory is determined according to each working condition type and hooking position: generated by S141, such as P1, P2, and P3; each working condition type: marked by S142, such as P1 being "standard working condition" and P2 and P3 being "complex working condition"; the first behavior trajectory is a preliminary motion sequence generated based on static working condition types and hooking positions; it does not consider real-time dynamic events and only generates a basic path and motion logic according to preset rules; each trajectory point contains: position coordinates (x, y, z); motion type (such as approach, alignment, grasp, release); speed and direction control parameters; optionally, the system calls a preset trajectory template according to the spatial coordinates of the hooking position and the working condition type; for example: standard working condition: use high-speed straight trajectory, fast approach; complex working condition: use low-speed curved trajectory, obstacle avoidance first.

[0102] Determine the second behavior trajectory according to each working condition type and the corresponding adaptive hooking event. Each working condition type is, for example, "standard working condition" or "complex working condition". The adaptive hooking event is generated by S133, for example, AE1, AE2, or AE3. The second behavior trajectory is a dynamically adjusted trajectory generated on the basis of the first behavior trajectory in combination with real-time event triggering. It contains event-driven action changes, such as pausing or detouring when encountering an obstacle, switching to infrared mode when there is insufficient light, and automatically retrying when hooking fails. Optionally, the system maps the adaptive hooking event to the trajectory node to form a "condition-action" branch. Each branch point contains a trigger condition (such as "detecting an obstacle"), an action response (such as "pausing and re-planning the path"), and a recovery strategy (such as "continuing the original trajectory after detouring").

[0103] Determine the adaptive hooking behavior according to the first behavior trajectory, the second behavior trajectory, and the hooking behavior matching table. The first behavior trajectory is a basic action sequence. The second behavior trajectory is a dynamically adjusted action sequence. The hooking behavior matching table is a pre-defined behavior rule library containing optimal behaviors under different working conditions and event combinations. Collect the hooking behavior matching table, which is shown in Table 1:

[0104] Table 1 Hooking Behavior Matching Table

[0105] Working condition type Event type Behavior suggestion Speed control Torque control Standard working condition Normal approach Fast approach High speed Medium Standard working condition Failed to unhook Retry once Medium speed High Complex working condition Encounter obstacle Avoid obstacle by detour Low speed Low Complex working condition Insufficient light Switch sensor Medium speed Medium

[0106] The system fuses the first and second behavior trajectories to generate a final behavior sequence. For each trajectory point, query the matching table to determine the optimal behavior parameters. The final behavior contains an action sequence (such as "approach→align→grab→release"), control parameters (speed, torque, sensor mode), and an exception handling mechanism (such as retry, alarm, and emergency stop).

[0107] Specifically, there are three hooking positions: P1, P2, and P3. The working condition type is P1: standard working condition, P2: complex working condition, and P3: complex working condition. The adaptive hooking event is AE1: normal approach at P1, AE2: encountering an obstacle at P2, and AE3: insufficient light at P3.

[0108] Generate the first behavior trajectory: P1 (standard working condition): trajectory: straight-line fast approach→align→grab→release; P2 (complex working condition): trajectory: curved low-speed approach→align→grab→release; P3 (complex working condition): trajectory: curved low-speed approach→align→grab→release.

[0109] Generate the second behavior trajectory: P1 (AE1: normal approach): no adjustment, follow the first trajectory; P2 (AE2: encountering an obstacle): adjust to: approach→detect obstacle→detour→re-approach→align→grab→release; P3 (AE3: insufficient light): adjust to: approach→switch to infrared sensor→align→grab→release.

[0110] Query hooking behavior matching table: P1: match "standard working condition + normal approach" -> behavior: fast approach, high speed, medium torque; P2: match "complex working condition + encounter obstacles" -> behavior: detour to avoid obstacles, low speed, low torque; P3: match "complex working condition + insufficient light" -> behavior: switch sensors, medium speed, medium torque.

[0111] The system generates three adaptive hooking behaviors: P1 behavior: action sequence: fast approach -> alignment -> grab -> release; control parameters: high speed, medium torque; exception handling: none; P2 behavior: action sequence: low-speed approach -> detour -> re-approach -> alignment -> grab -> release; control parameters: low speed, low torque; exception handling: alarm when obstacle avoidance fails; P3 behavior: action sequence: low-speed approach -> switch infrared sensors -> alignment -> grab -> release; control parameters: medium speed, medium torque; exception handling: retry when sensor switching fails; these adaptive hooking behaviors will serve as final control instructions, input into the robot execution system, to achieve efficient and safe multi-vehicle adaptive hooking operations.

[0112] Reference Figure 6 In step S15, the specific steps are:

[0113] S151: In each adaptive hooking behavior, determine the working route of the hooking working boom of the multi-vehicle adaptive hooking robot according to the adaptive hooking behavior, estimate the corresponding hooking work efficiency according to the working route of the hooking working boom, the working beat of the hooking working boom, and the action dimension of the adaptive hooking behavior;

[0114] S152: Collect multiple hooking positions, determine the hooking work level of the multi-vehicle adaptive hooking robot according to the multiple hooking positions and the corresponding hooking work efficiency, and simultaneously collect the remaining work content of the multi-vehicle adaptive hooking robot, determine multiple sub-work lists according to the recognition of the remaining work content of the multi-vehicle adaptive hooking robot, and determine the corresponding remaining hooking behavior and the corresponding time node based on the detection of the multiple sub-work lists;

[0115] S153: Determine the remaining hooking work system of the multi-vehicle adaptive hooking robot according to each remaining hooking behavior, the corresponding time node, and the hooking work level of the multi-vehicle adaptive hooking robot, and realize adaptive control of the multi-vehicle adaptive hooking robot on the remaining hooking position under the triggering of the remaining hooking work system.

[0116] In the embodiments of the present application, in each adaptive hooking behavior, the working route of the hooking working branch arm of the multi-vehicle adaptive hooking robot is determined according to the adaptive hooking behavior, the hooking working efficiency corresponding to the adaptive hooking behavior is estimated according to the working route of the hooking working branch arm, the working rhythm of the hooking working branch arm and the action dimension of the adaptive hooking behavior, the overall consideration of the adaptive hooking behavior is compatible, and the accuracy of the working route of the hooking working branch arm of the multi-vehicle adaptive hooking robot is ensured.

[0117] At this time, each adaptive hooking behavior (from S143), for example: P1: fast approach→alignment→grasp→release; P2: low-speed approach→circumvention→re-approach→alignment→grasp→release; P3: low-speed approach→infrared sensor switching→alignment→grasp→release.

[0118] The working route is the motion trajectory of the end effector (claw) in three-dimensional space when the hooking working branch arm executes the hooking behavior; each action corresponds to a space path point, and all path points constitute a complete working route; the path point contains: coordinates (x, y, z); attitude (pitch angle, yaw angle, roll angle); speed and acceleration constraints; the system selects different path planning algorithms according to the action type and the working condition type: linear action (such as fast approach): linear interpolation algorithm is used to generate a straight line path; curved action (such as circumvention): a smooth curve path is generated by using a Bezier curve or a spline interpolation; sensor switching action: sensor switching action points are inserted in the path to ensure that the branch arm position is stable when the sensor switches.

[0119] The working rhythm refers to the execution time interval of each action; the length of the rhythm directly affects the hooking working efficiency: short rhythm→high efficiency; long rhythm→low efficiency; the system sets the rhythm parameters according to the action type and the working condition type: standard working condition (such as P1): short rhythm (such as 1 second / action); complex working condition (such as P2, P3): long rhythm (such as 2 seconds / action); the rhythm parameter can be dynamically adjusted: if an obstacle is detected, the rhythm is automatically lengthened; if the path optimization is successful, the rhythm is automatically shortened; the working efficiency calculation formula is:

[0120]

[0121] Total rhythm time = sum of all action rhythms; action dimension score = space dimension score + control dimension score.

[0122] The action dimension refers to the complexity of the action, including: spatial dimension (such as straight line, curve, and round); control dimension (such as speed, torque, and sensor switching); the system scores each action dimension: straight line action: low score (such as 1 point); curve action: medium score (such as 2 points); round action: high score (such as 3 points); sensor switching action: high score (such as 3 points); the higher the score, the more complex the action and the lower the efficiency; the system calculates the work efficiency of each hooking behavior according to the action dimension score and the total beat time; for example: P1: total beat time = 4 seconds, action dimension score = 4 points → work efficiency = 1 / 4 x 4 = 1; P2: total beat time = 12 seconds, action dimension score = 12 points → work efficiency = 1 / 12 x 12 = 1; P3: total beat time = 10 seconds, action dimension score = 10 points → work efficiency = 1 / 10 x 10 = 1.

[0123] Further, a plurality of hooking positions are collected, and a hooking work level of the multi-vehicle adaptive hooking robot is determined according to the plurality of hooking positions and the corresponding hooking work efficiency, and at the same time, the remaining work content of the multi-vehicle adaptive hooking robot is collected, and a plurality of sub-work lists are determined according to the recognition of the remaining work content of the multi-vehicle adaptive hooking robot, and the corresponding remaining hooking behaviors and the corresponding time nodes are determined based on the detection of the plurality of sub-work lists, which is compatible with the overall consideration of the detection of the plurality of sub-work lists, and ensures the accuracy of the corresponding remaining hooking behaviors and the corresponding time nodes.

[0124] At this time, a plurality of hooking positions are collected, and the plurality of hooking positions: for example, P1, P2, P3, each position contains spatial coordinates and working condition types; the corresponding hooking work efficiency is calculated by S151, for example: P1: work efficiency 1.0; P2: work efficiency 0.58; P3: work efficiency 0.7.

[0125] The work level is a comprehensive evaluation of the difficulty of executing the hooking position, which is used to guide the subsequent task scheduling; the level division standard: high efficiency level (A level): efficiency ≥ 0.9, indicating that the action is simple, the execution is fast, and the resource consumption is low; medium efficiency level (B level): 0.6 ≤ efficiency < 0.9, indicating that the action is moderate and has certain complexity; low efficiency level (C level): efficiency < 0.6, indicating that the action is complex, the execution is slow, and the resource consumption is high; the system automatically maps to the corresponding level according to the work efficiency of the hooking position; for example: P1: efficiency 1.0 → A level; P2: efficiency 0.58 → C level; P3: efficiency 0.7 → B level.

[0126] The system obtains a list of uncompleted hooking tasks of the robot from the task management module; each task contains: hooking position (such as P4, P5, P6); working condition type (such as standard, complex, obstacle-intensive); priority (high, medium, low); estimated completion time; the sub-work list is a classification and grouping of the remaining work content, which is convenient for scheduling and execution; the classification basis: grouping according to position area (such as area A, area B); grouping according to working condition type (such as standard working condition group, complex working condition group); grouping according to priority (such as high priority group, medium priority group); the system divides the remaining work content into multiple sub-work lists according to the preset rules.

[0127] The system generates corresponding hooking behaviors for the tasks in each sub-work list; the behavior generation method: searching for a behavior matching table, matching the best behavior template according to the working condition type and position characteristics; for example: P4: standard working condition → fast approach alignment grasping release; P5: complex working condition → low-speed approach detour re-approach alignment grasping release; P6: obstacle-intensive working condition → low-speed approach switching infrared sensor alignment grasping release.

[0128] The time node is the planned execution time of each hooking behavior; the calculation method: based on the current time, task priority, and estimated execution time, dynamically scheduling the time node; high-priority tasks are arranged first, and low-priority tasks are executed later; for example: current time: 10:00:00; P4: high priority, estimated execution time 5 seconds → time node 10:00:00; P5: medium priority, estimated execution time 10 seconds → time node 10:00:10; P6: low priority, estimated execution time 8 seconds → time node 10:00:25.

[0129] Therefore, according to the various remaining hooking behaviors, the corresponding time nodes, and the hooking work level of the multi-vehicle adaptive hooking robot, the remaining hooking work system of the multi-vehicle adaptive hooking robot is determined, and the self-adaptive control of the multi-vehicle adaptive hooking robot to the remaining hooking position is realized under the triggering of the remaining hooking work system, which is compatible with the overall consideration of the various remaining hooking behaviors, the corresponding time nodes, and the hooking work level of the multi-vehicle adaptive hooking robot, ensuring the accuracy of the remaining hooking work system of the multi-vehicle adaptive hooking robot, at the same time, the self-adaptive hooking behavior of the multi-vehicle adaptive hooking robot is introduced, realizing the overall consideration of the hooking work efficiency, the corresponding hooking position, and the remaining work content of the multi-vehicle adaptive hooking robot, and improving the accuracy of the remaining hooking work system.

[0130] At this time, each remaining hooking behavior: generated from S152 step, for example: P4: fast approach→alignment→grab→release; P5: low-speed approach→circumnavigation→re-approach→alignment→grab→release; P6: low-speed approach→switch infrared sensor→alignment→grab→release; corresponding time nodes: for example: P4: 10:00:00; P5: 10:00:10; P6: 10:00:25; hooking work level: determined by S152, for example: P4: A level (high efficiency); P5: B level (medium efficiency); P6: C level (low efficiency); the remaining hooking work system is the overall scheduling and execution framework of the robot for the remaining hooking task; including the following elements: behavior sequence: specific action sequence of each hooking position; time node: starting time of each behavior; work level: used for dynamically adjusting resource allocation and priority; exception handling: different exception response strategies are set for tasks of different levels.

[0131] Work system construction method: the system generates a task execution queue according to time node sorting; dynamically adjusts resources according to work level: A-level tasks: preferentially allocate high-speed channels to reduce waiting time; B-level tasks: use standard resources with appropriate delay; C-level tasks: use low-speed channels and allow high-priority tasks to preempt; exception handling strategy: A-level tasks: retry immediately after failure; B-level tasks: delay retry after failure; C-level tasks: log after failure, do not retry.

[0132] The system automatically starts the work system when the first time node is reached (e.g., 10:00:00); trigger condition: time node arrival; the robot is in an idle state; the work environment is normal; execution flow: the system extracts the task of the current time node (e.g., P4) from the work system; allocate resources according to the work level of the task (A level); execute the hooking behavior sequence: fast approach→alignment→grab→release; after completion, the system checks the next time node (10:00:10) and starts P5; and so on until all tasks are completed; dynamic adjustment: if P4 execution times out, the system automatically moves the time nodes of P5 and P6 backward; if P6 is preempted by a high-priority task, the system puts it into the waiting queue and reschedules when resources are idle; exception handling: if P5 execution fails, the system retries after a delay of 5 seconds according to the B-level strategy; if P6 execution fails, the system logs and skips the task.

[0133] Please refer to Figure 7 , Figure 7 is a structural composition diagram of the multi-vehicle adaptive hooking robot working condition self-adaptive adjustment system in the embodiment of the application; the multi-vehicle adaptive hooking robot working condition self-adaptive adjustment system comprises:

[0134] The regional working image module 21 is configured to determine a plurality of regional nodes according to the moving path of the multi-vehicle adaptive hooking robot, and to acquire a plurality of regional working images according to detection of the plurality of regional nodes.

[0135] The multi-modal data module 22 is configured to determine a plurality of environment features, dynamic features and hooking features based on image recognition of the respective regional working images, and to determine a hooking working region according to the plurality of environment features, dynamic features and hooking features.

[0136] The adaptive hooking event module 23 is configured to determine a hooking behavior event of the multi-vehicle adaptive hooking robot relative to the hooking features according to the hooking working region, the positions of the hooking features and a hooking working branch arm of the multi-vehicle adaptive hooking robot, and to determine an adaptive hooking event of the multi-vehicle adaptive hooking robot according to the respective hooking behavior events and the moving path of the multi-vehicle adaptive hooking robot.

[0137] The adaptive hooking behavior module 24 is configured to determine a working condition type of the multi-vehicle adaptive hooking robot according to the respective hooking positions and the plurality of environment features, and to determine an adaptive hooking behavior of the multi-vehicle adaptive hooking robot based on the respective working condition types and the corresponding adaptive hooking events.

[0138] The remaining hooking working system module 25 is configured to estimate a corresponding hooking working efficiency based on the respective adaptive hooking behaviors, and to determine a remaining hooking working system of the multi-vehicle adaptive hooking robot according to the hooking working efficiency, the corresponding hooking positions and a remaining working content of the multi-vehicle adaptive hooking robot.

[0139] Any combination of the technical features of the above embodiments is possible. In order to make the description concise, not all combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not exist, they should be considered as the scope of the present disclosure.

Claims

1. A method for adaptively adjusting the working conditions of a multi-vehicle adaptive unhooking robot, characterized in that: include: Determine multiple regional nodes according to the moving path of the multi-vehicle adaptive hook removal robot, and collect multiple regional working images according to the detection of the multiple regional nodes; Determine multiple environmental features, dynamic features, and hook removal features based on image recognition of working images in each area, and determine the hook removal working area based on the multiple environmental features, dynamic features, and hook removal features; Determine the hooking behavior event of the multi-vehicle adaptive hooking robot relative to the hooking feature according to the hooking work area, the position of the hooking feature, and the hooking work support arm of the multi-vehicle adaptive hooking robot; and determine the adaptive hooking event of the multi-vehicle adaptive hooking robot according to each hooking behavior event and the moving path of the multi-vehicle adaptive hooking robot; Determine the working condition type of the multi-vehicle adaptive unhooking robot according to each unhooking position and multiple environmental characteristics, and determine the adaptive unhooking behavior of the multi-vehicle adaptive unhooking robot based on each working condition type and the corresponding adaptive unhooking event; Based on the estimation of the corresponding unhooking work efficiency of each adaptive unhooking behavior, the remaining unhooking work system of the multi-vehicle adaptive unhooking robot is determined according to the unhooking work efficiency, the corresponding unhooking position and the remaining work content of the multi-vehicle adaptive unhooking robot.

2. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 1 is characterized in that: The method of determining a plurality of regional nodes according to the moving path of the multi-vehicle adaptive hook-removing robot and collecting a plurality of regional working images according to the detection of the plurality of regional nodes includes: Collect the current position and target position of the multi-vehicle adaptive dehooking robot, determine the movement path of the multi-vehicle adaptive dehooking robot based on the current position, target position and corresponding movement space, determine multiple dehooking operation areas based on the detection of the movement path of the multi-vehicle adaptive dehooking robot, and mark the area node of each dehooking operation area; In each regional node, multiple cameras in the surrounding area are triggered according to the node position of the regional node. The multiple cameras take circular shots of the unhooking operation area where the regional node is located, and collect multiple sub-area images. The corresponding regional working image is determined based on the multiple sub-area images, the corresponding unhooking positions and the previous positioning positions of the multi-vehicle adaptive unhooking robots.

3. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 1 is characterized in that: The method of determining a plurality of environmental features, dynamic features, and hook removal features based on image recognition of working images of each area, and determining the hook removal working area according to the plurality of environmental features, dynamic features, and hook removal features, includes: In each regional working image, multiple sub-feature images are determined based on the recognition of the regional working image, and corresponding image recognition modes are determined based on the positions, image contours and image type weights of the multiple sub-feature images. Different types of sub-feature images are recognized using corresponding image recognition modes. executing corresponding image recognition modes on the plurality of sub-feature images and outputting corresponding feature combinations, wherein the regional working image has a plurality of feature combinations; Based on the screening of multiple feature combinations, multiple environmental features, dynamic features and hook removal features are determined, a preliminary area is constructed based on the multiple environmental features and dynamic features, and the hook removal working area is determined based on the preliminary area and the hook removal features.

4. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 1 is characterized in that: The method of determining a hook removal behavior event of the multi-vehicle adaptive hook removal robot relative to the hook removal feature according to the hook removal work area, the position of the hook removal feature, and the hook removal work support arm of the multi-vehicle adaptive hook removal robot, and determining an adaptive hook removal event of the multi-vehicle adaptive hook removal robot according to each hook removal behavior event and the moving path of the multi-vehicle adaptive hook removal robot, includes: Collecting the hook removal work area, determining a plurality of hook removal work nodes according to the identification of the hook removal work area, and determining a position event according to the positions of the plurality of hook removal work nodes and the positions of the hook removal features; The action events are determined according to multiple unhooking work nodes and the unhooking work arms of the multi-vehicle adaptive unhooking robot, and the unhooking behavior events of the multi-vehicle adaptive unhooking robot relative to the unhooking feature are determined based on the position events, action events and the working mode of the multi-vehicle adaptive unhooking robot relative to the unhooking work area.

5. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 4 is characterized in that: The method further includes determining a hook removal behavior event of the multi-vehicle adaptive hook removal robot relative to the hook removal feature according to the hook removal work area, the position of the hook removal feature, and the hook removal work support arm of the multi-vehicle adaptive hook removal robot, and determining an adaptive hook removal event of the multi-vehicle adaptive hook removal robot according to each hook removal behavior event and the moving path of the multi-vehicle adaptive hook removal robot. The moving paths of the multi-vehicle adaptive unhooking robots are collected, and the activity range of the corresponding unhooking behavior is determined based on the detection of each unhooking behavior event. The adaptive unhooking events of the multi-vehicle adaptive unhooking robots are determined based on the activity range of the unhooking behavior, the moving paths of the multi-vehicle adaptive unhooking robots and the unhooking postures of the multi-vehicle adaptive unhooking robots.

6. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 1 is characterized in that: The method of determining the working condition type of the multi-vehicle adaptive unhooking robot according to each unhooking position and multiple environmental features, and determining the adaptive unhooking behavior of the multi-vehicle adaptive unhooking robot based on each working condition type and the corresponding adaptive unhooking event, includes: Collecting each hook removal position, and determining a corresponding hook removal environment area according to each hook removal position and the corresponding hook removal feature, determining a plurality of environmental parameters according to real-time detection of the hook removal environment area, and determining a plurality of environmental features according to the position of the plurality of environmental parameters, the type of the parameters and the shape of the hook removal environment area; The working environment of the multi-vehicle adaptive hook removal robot is determined according to multiple environmental features and the surrounding space of each hook removal position, and the working condition type of the multi-vehicle adaptive hook removal robot is marked according to the working environment.

7. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 6 is characterized in that: The method further includes determining the working condition type of the multi-vehicle adaptive unhooking robot according to each unhooking position and multiple environmental features, and determining the adaptive unhooking behavior of the multi-vehicle adaptive unhooking robot based on each working condition type and the corresponding adaptive unhooking event. The first behavior trajectory is determined according to each working condition type and the unhooking position, the second behavior trajectory is determined according to each working condition type and the corresponding adaptive unhooking event, and the adaptive unhooking behavior of the multi-vehicle adaptive unhooking robot is determined according to the first behavior trajectory, the second behavior trajectory and the unhooking behavior matching table.

8. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 1 is characterized in that: The method of estimating the corresponding hook removal work efficiency based on each adaptive hook removal behavior and determining the remaining hook removal work system of the multi-vehicle adaptive hook removal robot according to the hook removal work efficiency, the corresponding hook removal position and the remaining work content of the multi-vehicle adaptive hook removal robot includes: In each adaptive unhooking behavior, the working route of the unhooking working arm of the multi-vehicle adaptive unhooking robot is determined according to the adaptive unhooking behavior, and the corresponding unhooking work efficiency is estimated according to the working route of the unhooking working arm, the working rhythm of the unhooking working arm and the action dimension of the adaptive unhooking behavior.

9. The method for self-adapting working conditions of a multi-vehicle adaptive unhooking robot according to claim 8 is characterized in that: The method of estimating the corresponding hook removal work efficiency based on each adaptive hook removal behavior and determining the remaining hook removal work system of the multi-vehicle adaptive hook removal robot according to the hook removal work efficiency, the corresponding hook removal position and the remaining work content of the multi-vehicle adaptive hook removal robot further includes: Collect multiple hook removal positions, determine the hook removal work level of the multi-vehicle adaptive hook removal robot based on the multiple hook removal positions and the corresponding hook removal work efficiency, and at the same time, collect the remaining work content of the multi-vehicle adaptive hook removal robot, determine multiple sub-work lists based on the identification of the remaining work content of the multi-vehicle adaptive hook removal robot, and determine the corresponding remaining hook removal behaviors and corresponding time nodes based on the detection of the multiple sub-work lists; The remaining unhooking work system of the multi-vehicle adaptive unhooking robot is determined according to the remaining unhooking behaviors, the corresponding time nodes and the unhooking work level of the multi-vehicle adaptive unhooking robot, and the multi-vehicle adaptive unhooking robot realizes adaptive control of the remaining unhooking positions under the triggering of the remaining unhooking work system.

10. A multi-vehicle adaptive unhooking robot working condition adaptive adjustment system, characterized in that: The multi-vehicle adaptive unhooking robot working condition self-adaptive adjustment system is applied to the multi-vehicle adaptive unhooking robot working condition self-adaptive adjustment method according to any one of claims 1 to 9, and the multi-vehicle adaptive unhooking robot working condition self-adaptive adjustment system includes: The regional working image module is used to determine multiple regional nodes according to the moving path of the multi-vehicle adaptive hook removal robot, and collect multiple regional working images according to the detection of the multiple regional nodes; A hook removal work area module is used to determine multiple environmental features, dynamic features and hook removal features based on image recognition of work images in each area, and determine the hook removal work area based on the multiple environmental features, dynamic features and hook removal features; An adaptive hook removal event module is used to determine the hook removal behavior event of the multi-vehicle adaptive hook removal robot relative to the hook removal feature based on the hook removal work area, the location of the hook removal feature, and the hook removal work support arm of the multi-vehicle adaptive hook removal robot, and to determine the adaptive hook removal event of the multi-vehicle adaptive hook removal robot based on each hook removal behavior event and the movement path of the multi-vehicle adaptive hook removal robot; An adaptive unhooking behavior module is used to determine the working condition type of the multi-vehicle adaptive unhooking robot according to each unhooking position and multiple environmental characteristics, and to determine the adaptive unhooking behavior of the multi-vehicle adaptive unhooking robot based on each working condition type and the corresponding adaptive unhooking event; The remaining unhooking work system module is used to estimate the corresponding unhooking work efficiency based on each adaptive unhooking behavior, and determine the remaining unhooking work system of the multi-vehicle adaptive unhooking robot according to the unhooking work efficiency, the corresponding unhooking position and the remaining work content of the multi-vehicle adaptive unhooking robot.

Images