Anti-collision processing method and system of hook unhooking robot under complex working conditions

Abstract

This invention discloses a collision avoidance method and system for a de-hooking robot under complex working conditions. The invention relates to the technical field of collision avoidance methods. Based on the working area, the type of complex working condition, and the de-hooking robot, multiple de-hooking positions are determined within the working area. The de-hooking behavior of the robot at each de-hooking position is determined based on the current image corresponding to each de-hooking position and the robot's working mode. The collision area of ​​the de-hooking robot is predicted based on each de-hooking behavior and the current image corresponding to each de-hooking position. Multiple collision nodes are determined based on the identification of the collision areas. An optimized de-hooking behavior is determined based on the multiple collision nodes, the corresponding de-hooking behavior, and the robot's posture. Finally, a collision avoidance system for the de-hooking robot is determined based on the optimized de-hooking behavior, the original de-hooking behavior, and the robot's movement path, thus improving the accuracy of the collision avoidance system.

Description

Technical Field

[0001] This invention relates to the technical field of anti-collision processing methods, and in particular to an anti-collision processing method and system for a hook-unhooking robot under complex working conditions. Background Technology

[0002] With the development of technology, uncoupling robots are industrial robots specifically designed for automated uncoupling operations in scenarios such as railways, ports, and logistics. Their main function is to automatically unlock, separate, or detach at the connection points (such as couplers) of transport vehicles like trains, freight cars, and containers. In existing technologies, uncoupling robots rely on cameras to capture images of the uncoupling location and determine the uncoupling characteristics based on image recognition. However, this approach does not consider the impact of complex working conditions, making the uncoupling robot prone to collisions during the uncoupling process and affecting the accuracy of its uncoupling actions at various locations. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for preventing collisions of a hook-unhooking robot under complex working conditions.

[0004] This invention provides a collision avoidance method for a de-hooking robot under complex working conditions, comprising: determining multiple sub-working areas based on the detection of the working area of ​​the de-hooking robot, and marking previous working images of the multiple sub-working areas; determining multiple working condition data of the de-hooking robot based on the previous working images of the multiple sub-working areas, and determining the corresponding complex working condition type based on the multiple working condition data, the shape of the de-hooking robot, and the service life of the de-hooking robot; determining multiple de-hooking positions of the working area based on the working area, the complex working condition type, and the de-hooking robot, and determining the de-hooking behavior of the de-hooking robot at each de-hooking position based on the current image corresponding to the multiple de-hooking positions and the working mode of the de-hooking robot; predicting the collision area of ​​the de-hooking robot based on each de-hooking behavior and the current image corresponding to the de-hooking position, and determining multiple collision nodes based on the identification of the collision area; determining optimized de-hooking behavior based on the multiple collision nodes, the corresponding de-hooking behavior, and the posture of the de-hooking robot to eliminate multiple collision nodes; determining a collision avoidance system for the de-hooking robot based on the multiple optimized de-hooking behaviors, the original de-hooking behaviors, and the movement path of the de-hooking robot, and progressively improving the de-hooking efficiency of the de-hooking robot.

[0005] This invention provides a collision avoidance system for a decoy robot under complex working conditions. This system is applied to the aforementioned collision avoidance method for a decoy robot under complex working conditions. The collision avoidance system includes:

[0006] The previous working image module is used to determine multiple sub-working areas based on the detection of the working area of ​​the unhooking robot, and to mark the previous working images of multiple sub-working areas;

[0007] The multimodal data module is used to determine multiple working condition data of the unhooking robot based on previous working images of multiple sub-working areas, and to determine the corresponding complex working condition type based on multiple working condition data, the shape of the unhooking robot and the service life of the unhooking robot.

[0008] The unhooking behavior module is used to determine multiple unhooking positions in the work area based on the work area, the type of complex working conditions, and the unhooking robot, and to determine the unhooking behavior of the unhooking robot at each unhooking position according to the current image corresponding to the multiple unhooking positions and the working mode of the unhooking robot.

[0009] The optimization module is used to predict the collision area of ​​the unhooking robot based on the current image corresponding to each unhooking behavior and unhooking position, and to determine multiple collision nodes based on the identification of the collision area; and to determine the optimized unhooking behavior based on the multiple collision nodes, the corresponding unhooking behavior and the posture of the unhooking robot, so as to eliminate multiple collision nodes.

[0010] The anti-collision processing system module is used to determine the anti-collision processing system of the unhooking robot based on multiple optimized unhooking behaviors, the original unhooking behaviors, and the unhooking robot's movement path, and to improve the unhooking efficiency of the unhooking robot step by step.

[0011] Compared with the prior art, the beneficial effects of the present invention are:

[0012] In this embodiment of the invention, the method determines multiple unhooking positions in the work area based on the work area, the type of complex working conditions, and the unhooking robot. The unhooking behavior of the unhooking robot at each unhooking position is determined according to the current image corresponding to the multiple unhooking positions and the working mode of the unhooking robot. The introduction of complex working conditions further controls the complex working conditions and takes into account the overall consideration of the current image corresponding to the multiple unhooking positions and the working mode of the unhooking robot, thereby improving the accuracy of the unhooking behavior of the unhooking robot at each unhooking position.

[0013] Therefore, based on the current image corresponding to each unhooking action and unhooking position, the collision area of ​​the unhooking robot is predicted, and multiple collision nodes are determined based on the identification of the collision area. Based on the multiple collision nodes, the corresponding unhooking actions, and the posture of the unhooking robot, the optimized unhooking actions are determined to eliminate multiple collision nodes. Based on the multiple optimized unhooking actions, the original unhooking actions, and the movement path of the unhooking robot, the anti-collision processing system of the unhooking robot is determined. Multiple collision nodes are introduced to further ensure the accuracy of the optimized unhooking actions. This achieves a holistic consideration of multiple optimized unhooking actions, the original unhooking actions, and the movement path of the unhooking robot, improves the accuracy of the anti-collision processing system of the unhooking robot, and gradually improves the unhooking efficiency of the unhooking robot. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the anti-collision handling method for a hook-unhooking robot under complex working conditions in an embodiment of the present invention.

[0015] Figure 2 This is a flowchart illustrating step S11 of the anti-collision handling method for a hook-unhooking robot under complex working conditions in an embodiment of the present invention.

[0016] Figure 3 This is a flowchart illustrating step S12 of the anti-collision handling method for a hook-unhooking robot under complex working conditions in an embodiment of the present invention.

[0017] Figure 4 This is a flowchart illustrating step S13 of the anti-collision handling method for a hook-unhooking robot under complex working conditions in an embodiment of the present invention.

[0018] Figure 5 This is a flowchart illustrating step S14 of the anti-collision handling method for a hook-unhooking robot under complex working conditions in an embodiment of the present invention.

[0019] Figure 6 This is a flowchart illustrating step S15 of the anti-collision handling method for a hook-unhooking robot under complex working conditions in an embodiment of the present invention.

[0020] Figure 7 This is a schematic diagram of the structural composition of the anti-collision system for the unhooking robot under complex working conditions in an embodiment of the present invention. Detailed Implementation

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

[0022] Please see Figures 1 to 7 A collision avoidance method for unhooking robots under complex working conditions is provided, applicable to collision avoidance scenarios for unhooking robots. The collision avoidance method for unhooking robots under complex working conditions includes:

[0023] Step S11: Determine multiple sub-working areas based on the detection of the unhooking robot's working area, and mark previous working images of multiple sub-working areas;

[0024] Step S12: Determine multiple working condition data of the unhooking robot based on previous working images of multiple sub-working areas, and determine the corresponding complex working condition type based on multiple working condition data, the shape of the unhooking robot and the service life of the unhooking robot.

[0025] Step S13: Based on the work area, the type of complex working condition, and the unhooking robot, determine multiple unhooking positions in the work area, and determine the unhooking behavior of the unhooking robot at each unhooking position according to the current image corresponding to the multiple unhooking positions and the working mode of the unhooking robot.

[0026] Step S14: Predict the collision area of ​​the unhooking robot based on the current image corresponding to each unhooking behavior and unhooking position, and determine multiple collision nodes based on the identification of the collision area; determine the optimized unhooking behavior based on the multiple collision nodes, the corresponding unhooking behavior and the posture of the unhooking robot, so as to eliminate multiple collision nodes.

[0027] Step S15: Determine the anti-collision system of the unhooking robot based on multiple optimized unhooking behaviors, the original unhooking behaviors, and the movement path of the unhooking robot, and improve the unhooking efficiency of the unhooking robot step by step.

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

[0029] S111: Collect the model of the unhooking robot, determine the working area of ​​the unhooking robot based on the matching of the unhooking robot's signals and the work schedule, and determine multiple sub-working areas based on the area location, area shape and work schedule of the unhooking robot's working area.

[0030] S112: Determine the corresponding static and dynamic nodes based on the detection of each sub-working area, and collect the node positions corresponding to the static nodes and the node positions corresponding to the dynamic nodes. Based on the tracing of the node positions corresponding to the static nodes and the node positions corresponding to the dynamic nodes, determine the previous working images of multiple sub-working areas.

[0031] In the embodiments of this application, the specific model information of the unhooking robot is obtained, including the robot's physical dimensions, working range, load capacity, sensor configuration and other technical parameters; the model information is usually stored in the robot's control system and can be obtained by reading the device ID or querying the device database; different models of unhooking robots have different working characteristics and limitations, and this information directly affects the subsequent division of the working area.

[0032] Robot signals include real-time position signals, status signals (working / standby / fault, etc.), and task execution signals; the work schedule contains information such as the scheduled task time, location, and content, and is usually generated by the production management system; by matching the robot's real-time signals with the work schedule, the task that the robot should currently perform and its corresponding work area can be determined; the matching process needs to consider time synchronization and position verification to ensure that the robot is in the correct execution position.

[0033] Based on the location, shape, and work schedule of the unhooking robot's work area, multiple sub-work areas are determined. Location refers to the coordinates and relative positions of the work area within the overall space; shape refers to the geometric shape, size, boundary features, and other physical attributes of the work area; the work schedule includes information such as task priority, time window, and task type. Sub-work area division needs to consider: the robot's working radius and accessibility, the spatial distribution characteristics of tasks, the distribution of environmental obstacles, and the logical sequence of the workflow. The division principle is to ensure that the tasks within each sub-area are similar and that the transition paths between sub-areas are optimized.

[0034] Furthermore, based on the detection of each sub-work area, corresponding static and dynamic nodes are determined. Static nodes refer to objects or reference points that are fixed in position and do not move within the sub-work area, such as equipment brackets, fixed fixtures, walls, and columns; these nodes maintain their positions unchanged during long-term operation. Dynamic nodes refer to objects or reference points that change position within the sub-work area, such as moving workpieces, temporarily placed tools, operators, and other mobile devices; these nodes need to have their position changes tracked in real time. Through multi-sensor fusion technology such as LiDAR, vision sensors, and ultrasonic sensors, the sub-work area is scanned and identified to distinguish between static and dynamic elements. Algorithms such as point cloud clustering, feature extraction, and motion detection are used to classify the detected elements into static or dynamic nodes.

[0035] Collect the node positions corresponding to static nodes and dynamic nodes, store the collected location data in a database, and establish a node-location-time association; establish a time-series database to record the historical location information of all nodes; use timestamps to associate the location data of static and dynamic nodes; reconstruct the scene state at a specific time point based on the historical node location data; combine the actual images captured by the vision system to generate working images containing node annotations; the images should include metadata such as node identifiers, location coordinates, and timestamps; classify and store images: classify working images according to dimensions such as sub-working areas, time, and working condition type; establish an image indexing system to facilitate subsequent querying and analysis.

[0036] Establish a time-series database to record the historical location information of all nodes; use timestamps to associate the location data of static and dynamic nodes; reconstruct the scene state at a specific time point based on the historical node location data; combine actual images captured by the vision system to generate working images containing node annotations; the images should contain metadata such as node identifiers, location coordinates, and timestamps; classify the working images according to dimensions such as sub-work areas, time, and working condition type; establish an image indexing system to facilitate subsequent querying and analysis.

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

[0038] S121: Collect previous working images of multiple sub-working areas, determine multiple working condition features based on the recognition of the previous working images of multiple sub-working areas, and determine the working condition area based on the position, feature shape and movement path of the unhooking robot relative to the sub-working area of ​​multiple working condition features.

[0039] S122: Determine multiple working condition data of the unhooking robot based on the detection of the working area. At the same time, collect the shape and service life of the unhooking robot. Construct corresponding multimodal data based on multiple working condition data, the shape and service life of the unhooking robot.

[0040] S123: Determine multi-level working condition combinations based on the detection of multi-modal data, and determine the corresponding complex working condition types based on the identification of multi-level working condition combinations.

[0041] In the embodiments of this application, previous working images of multiple sub-working areas are collected, and multiple working condition features are determined based on the recognition of the previous working images of multiple sub-working areas. The working condition area is determined based on the position, feature shape and movement path of the unhooking robot relative to the sub-working area of ​​the multiple working condition features. This approach takes into account the overall consideration of the position, feature shape and movement path of the unhooking robot relative to the sub-working area of ​​the multiple working condition features, ensuring the accuracy of the working condition area.

[0042] At this time, previous working images of multiple sub-working areas are collected. Based on the recognition of previous working images of multiple sub-working areas, multiple working condition features are determined. Feature extraction is performed by image recognition of previous working images of multiple sub-working areas. Feature types: structural features: equipment location, workpiece distribution, fixed obstacles; dynamic features: personnel location, mobile equipment status, workpiece flow; environmental features: lighting conditions, temperature, humidity; task features: hook-off point location, workpiece type, operation difficulty.

[0043] The working area is determined based on the location, shape, and movement path of the unhooking robot relative to the sub-working area. The principles for dividing the working area are: areas with similar features are grouped into the same working area; the density of the robot's historical movement paths is considered; and areas are classified according to the level of collision risk. The division method involves dividing the workspace into grid cells, using algorithms such as K-means and DBSCAN for region clustering, generating a heatmap based on the robot's historical paths, and identifying high-frequency areas.

[0044] Furthermore, multiple working condition data of the unhooking robot are determined based on the detection of the working area. At the same time, the morphology and service life of the unhooking robot are collected. Based on the multiple working condition data, the morphology and service life of the unhooking robot, corresponding multimodal data are constructed, which is compatible with the overall consideration of the detection of the working area and ensures the accuracy of the multiple working condition data of the unhooking robot.

[0045] At this time, data is collected in real time by various sensors installed on the unhooking robot and in the working environment, including: environmental sensors: temperature, humidity, light intensity, dust concentration, etc.; force sensors: torque, tension, pressure, etc. during the unhooking process; position sensors: precise coordinates of the robot, joint angles, etc.; speed sensors: movement speed, angular velocity, etc.; vision sensors: target recognition, obstacle detection, etc.; at the same time, the raw data is filtered, calibrated, and standardized to eliminate noise and outliers; key features such as maximum value, minimum value, average value, and rate of change are extracted from the processed data.

[0046] Robot morphology data acquisition includes: physical morphology (including robot arm length, number of joints, end effector type, etc.); working morphology (including current posture, extension state, load status, etc.); space occupation (a 3D volume model of the robot in the workspace); and lifespan data acquisition, including working time (cumulative operating hours), work cycle count (total number of times the unhooking action is completed), maintenance records (history of regular maintenance and component replacement), and performance degradation (wear level and accuracy reduction of key components). Simultaneously, multimodal data is constructed based on multiple working condition data, the robot's morphology, and its lifespan. Data from different types and sources is integrated to form a unified data representation, ensuring consistency in time and space between different modalities. Relationships between different modalities are established, such as the relationship between working conditions and robot performance. The fused data is organized into a structured format suitable for subsequent analysis and processing.

[0047] By constructing such detailed multimodal data, we can comprehensively reflect the working status, morphological characteristics, and service life of the unhooking robot under specific working conditions, providing a rich data foundation for subsequent complex working condition identification and collision avoidance strategy formulation; this multi-dimensional, multimodal data fusion method can more accurately assess the robot's working status and potential risks.

[0048] Therefore, multi-level working condition combinations are determined based on the detection of multi-modal data, and the corresponding complex working condition types are determined based on the identification of multi-level working condition combinations. This approach takes into account the overall consideration of multi-level working condition combination identification and ensures the accuracy of the corresponding complex working condition types.

[0049] At this point, the previously constructed multimodal data (including operating condition data, robot morphology data, and lifespan data) are integrated to form a unified data structure. These data come from different sensors and systems and need to be processed uniformly using data fusion technology. Key features are extracted from the multimodal data, such as: operating condition features: temperature change rate, torque fluctuation, position deviation, etc.; morphology features: joint angle combinations, end effector posture, etc.; lifespan features: cumulative working time, load cycle count, performance degradation index, etc. Feature selection algorithms (such as principal component analysis (PCA), recursive feature elimination (RFE), etc.) are used to select the most representative features.

[0050] Features are layered and combined according to different dimensions: First layer: Basic operating condition layer (environmental conditions, basic operating parameters); Second layer: Interactive operating condition layer (human-computer interaction, multi-device collaboration); Third layer: Abnormal operating condition layer (equipment failure, emergency); Features are combined within each layer, and relationships are established between layers; Operating condition combination representation: Multi-layer operating condition combinations are represented using vector, matrix, or graph structures.

[0051] Machine learning algorithms (such as SVM, random forest, neural networks, etc.) are used to classify working condition combinations and establish a working condition pattern library containing feature patterns of various typical working conditions. Based on the characteristics of working condition combinations, complex working conditions are divided into the following categories: High-Precision Cooperative Working Condition (HPC); Multi-Interference Cooperative Working Condition (MIC); Emergency Obstacle Avoidance Working Condition (EAO); Equipment Degradation Working Condition (EDD); and Environmental Anomaly Working Condition (EAC). Simultaneously, an evaluation index system is established, including: Complexity Index: comprehensively assessing the complexity of the working condition; Risk Level: assessing potential safety risks; Efficiency Impact: assessing the degree of impact on work efficiency. Changes in working condition combinations are monitored in real time, and the working condition type identification results are dynamically updated. A working condition transformation model is established to predict the trend of working condition changes.

[0052] refer to Figure 4 In step S13, the specific steps are as follows:

[0053] S131: Collect the working area, determine multiple working condition nodes based on the working area and the corresponding complex working condition types, and mark the complex working condition type of the working condition node. Determine the first unhooking path of the working area according to the node positions of multiple working condition nodes and the movement path of the unhooking robot relative to the working area.

[0054] S132: Determine the second unhooking path of the work area based on the types of complex working conditions corresponding to multiple working condition nodes and the movement path of the unhooking robot relative to the work area; construct multiple unhooking positions in the work area based on the first and second unhooking paths;

[0055] S133: Collect current images corresponding to multiple unhooking positions, and determine the shape of the corresponding unhooked part based on the recognition of the current image. Determine the unhooking behavior corresponding to the unhooking position based on the shape of the unhooked part, the unhooking position, and the working mode of the unhooking robot, so as to mark the unhooking behavior of the unhooking robot at each unhooking position.

[0056] In the embodiments of this application, a working area is collected, multiple working condition nodes are determined based on the working area and the corresponding complex working condition types, and the complex working condition types of the working condition nodes are marked. The first unhooking path of the working area is determined according to the node positions of the multiple working condition nodes and the movement path of the unhooking robot relative to the working area. This takes into account the overall consideration of the node positions of the multiple working condition nodes and the movement path of the unhooking robot relative to the working area, ensuring the accuracy of the first unhooking path of the working area.

[0057] At this point, complete spatial information of the current work area is obtained from the robot control system or environmental modeling system, including but not limited to: area boundary coordinates (such as polygon vertices, grid maps, CAD models, etc.); internal structure of the area (such as equipment locations, passages, obstacles, etc.); area attributes (such as lighting, temperature, ground material, etc.); the collected area data is converted into a format that the robot can recognize (such as two-dimensional / three-dimensional grid maps, topology maps, point clouds, etc.) to establish a digital twin model of the work area for subsequent path planning and node placement.

[0058] Nodes are arranged according to the area structure (such as corners, passageways, and near equipment) and complex operating conditions (such as MIC, EDD, and EAO). Node types include: path control nodes (used for path turning or speed control); decision nodes (used for operating condition switching or obstacle avoidance judgment); and target approach nodes (located near critical operation points of the unhooked component). Each node is labeled with a complex operating condition type, such as: "type":"MIC" (high-density multi-interference collaborative operating condition); "type":"EDD" (equipment performance degradation operating condition); and "type":"EAO" (extreme environment operating condition). Node attributes include: node ID, coordinates (x, y, z), operating condition type, node function description, and path constraints.

[0059] Algorithms such as RRT (Rapid Random Tree) and DWA (Dynamic Window Method) are used to generate an initial path based on the node positions. The path generation considers: reachability between nodes (no collisions); path smoothness (reducing sharp turns); speed planning (different speeds under different working conditions); the path consists of a series of path points, each point including: coordinates (x, y, z); orientation angle (θ); speed limit (v); working condition type label.

[0060] Specifically, the work area is a decoupling and uncoupling operation area of ​​a railway freight marshalling yard, about 30 meters long and 10 meters wide, containing tracks, carriages, temporary storage materials and personnel activity areas; the complex working conditions include both high-density multi-interference cooperative working conditions (MIC) and equipment performance degradation working conditions (EDD).

[0061] Work area boundary coordinates: top left corner (0,10), top right corner (30,10), bottom right corner (30,0), bottom left corner (0,0); Internal structure: tracks are distributed longitudinally along y=5; carriages are located between x=10 and x=20; temporary storage area is located between x=22 and x=25; personnel activity area is located between x=5 and x=8; data acquisition working condition node diagram table, which is shown in Table 1:

[0062] Table 1. Schematic Diagram of Working Conditions and Nodes

[0063] Node ID Coordinates (x, y) Operating conditions Function Description N01 (2,8) MIC Starting point, edge of the personnel activity area N02 (7,5) MIC orbital approach point N03 (15,5) EDD Carriage operation point N04 (23,5) MIC Storage area proximity point N05 (28,5) EDD End point, equipment performance degradation zone

[0064] The first de-hooking path is generated using a method such as RRT (Rapid Random Tree), and a schematic table of the first de-hooking path is collected, as shown in Table 2:

[0065] Table 2: Schematic diagram of the first unhooking path

[0066]

[0067]

[0068] The path starts from the starting point P01 (edge ​​of the personnel activity area), passes through the track approach point P03, enters the carriage operation area P05, then approaches the stacking area P07, and finally reaches the destination P09. In the MIC area (such as P01-P03, P06-P07), the speed is lower, emphasizing obstacle avoidance. In the EDD area (such as P04-P05, P08-P09), the speed is further reduced to protect the equipment. By accurately collecting the working area, rationally arranging the working condition nodes, and generating the first unhooking path based on the node positions and complex working condition types, the foundation is laid for subsequent path optimization and unhooking behavior decisions. This process not only considers the spatial structure but also incorporates dynamic working condition information, significantly improving the adaptability and safety of the unhooking robot in complex environments.

[0069] Furthermore, a second unhooking path for the work area is determined based on the types of complex working conditions corresponding to multiple working nodes and the movement path of the unhooking robot relative to the work area. Multiple unhooking positions in the work area are constructed based on the first and second unhooking paths, which takes into account the overall consideration of the types of complex working conditions corresponding to multiple working nodes and the movement path of the unhooking robot relative to the work area, ensuring the accuracy of the second unhooking path in the work area.

[0070] At this point, a second unhooking path for the work area is determined based on the types of complex working conditions corresponding to multiple working condition nodes and the movement path of the unhooking robot relative to the work area. The second unhooking path is an optimized version of the first unhooking path, with key considerations including: types of complex working conditions (such as MIC, EDD, EAO): different working conditions have different requirements for path safety, efficiency, and obstacle avoidance strategies; robot movement characteristics: such as turning radius, acceleration and deceleration capabilities, and robotic arm extension range; and dynamic environmental changes: such as personnel movement, temporary obstacles, and changes in lighting.

[0071] Optimization strategies include: Smoothness optimization: reducing sharp turns and improving motion continuity; Safety enhancement: adding buffer zones in the MIC area and reducing speed in the EDD area; Time efficiency optimization: shortening path length or reducing waiting time while ensuring safety; Multi-objective trade-offs: balancing multiple dimensions such as time, energy consumption, and equipment wear and tear; Path representation: typically using a sequence of waypoints, each point containing: coordinates (x, y, z); heading; speed limit (maxspeed); and operating condition labels (such as MIC, EDD).

[0072] Multiple unhooking positions are constructed in the working area based on the first and second unhooking paths. The unhooking positions are key points for the robot to perform unhooking operations and usually have the following characteristics: good field of vision (the unhooked object can be clearly identified); reachable by the robotic arm (within the kinematic range); moderate distance from the unhooked object (too close affects recognition, too far reduces accuracy); and minimal environmental interference (such as no obstruction and low vibration).

[0073] The logic for constructing unhooking locations is as follows: Identify "overlapping sections," "optimized sections," and "alternative sections" in the first and second paths; extract suitable nodes or sections for unhooking from the two paths; score locations based on indicators such as unhooking success rate, time efficiency, and safety, and retain the optimal location; output a list of unhooking locations: each location includes: coordinates and orientation; applicable working condition type; recommended unhooking behavior; and risk level.

[0074] Specifically, the work area is a railway freight yard, including the carriage area, the material stacking area, and the personnel passage area; the complex working conditions are: MIC: areas with frequent personnel activity; EDD: areas with aging equipment and limited accuracy; the unhooking robot model is RB003, which is equipped with a robotic arm and a vision recognition module; the first unhooking path (generated by S131) ​​already contains 9 nodes, and the total path length is about 28 meters.

[0075] First unhooking path node sequence (partial):

[0076] P01(2,8),P02(5,7),P03(7,5),P04(12,5),P05(15,5),P06(20,5),P07(23,5),P08(26,5),P09(28,5);

[0077] Operating condition labels for each node: P01-P03: MIC; P04-P05: EDD; P06-P07: MIC;

[0078] P08-P09: EDD;

[0079] Optimization Process: Smoothing: Add an intermediate point P02a(6,6) in the P02-P03 segment to make the turn smoother; Safety Enhancement: Shift the path outward by 0.5 meters in the MIC area (e.g., P06-P07), away from the pedestrian walkway; Speed ​​Adjustment: Reduce the target speed to 0.3 m / s in the EDD area (P04-P05, P08-P09); Path Shortening: Replace the original broken line with a straight path in the end section (P08-P09); Output Second Unhooking Path Node Sequence:

[0080] P01(2,8),P02(5,7),P02a(6,6),P03(7,5),P04(12,5),P05(15,5),P06(20,5),P07(23,5),P08(26,5),P09(28,5).

[0081] The two paths were compared: overlapping sections: P01-P02, P03-P04, P05-P06, P07-P08; optimized sections: P02-P03 (with the addition of P02a), P08-P09 (straight path); key candidate unhooking points: P03, P05, P07, P09. By combining the types of complex working conditions and the robot's movement characteristics, the first unhooking path was optimized to generate a safer, more efficient, and more adaptable second unhooking path. Based on the comparative analysis of the two paths, multiple unhooking positions were constructed. These unhooking positions not only considered spatial accessibility but also incorporated multi-dimensional factors such as environmental interference and working condition risks, providing solid data support for subsequent unhooking behavior decisions.

[0082] Therefore, multiple images corresponding to the unhooking positions are acquired, and the shape of the corresponding unhooked item is determined based on the recognition of the current image. The unhooking behavior corresponding to the unhooking position is determined based on the shape of the unhooked item, the unhooking position, and the working mode of the unhooking robot. This marks the unhooking behavior of the unhooking robot at each unhooking position, taking into account the overall consideration of the recognition of the current image and ensuring the accuracy of the shape of the corresponding unhooked item. At the same time, complex working conditions are introduced to further control the complex working conditions, taking into account the overall consideration of the current images corresponding to multiple unhooking positions and the working mode of the unhooking robot, thereby improving the accuracy of the unhooking behavior of the unhooking robot at each unhooking position.

[0083] At this time, the vision system (such as RGB camera, depth camera, LiDAR, etc.) on the unhooking robot collects environmental images of the unhooking position in real time; multiple angle images are collected at each unhooking position to cover different perspectives of the unhooked part; the image acquisition time is usually 1 to 2 seconds before the robot approaches the unhooking position to ensure that the image is consistent with the actual environment.

[0084] Image recognition processing flow: Preprocessing: noise reduction, contrast enhancement, normalization, etc.; Target detection: using algorithms such as YOLO and Faster R-CNN to identify the removed hooks; Morphology classification: classifying the removed hooks into different morphologies based on the recognition results, for example: Morphology A: hooks hang vertically and are stable; Morphology B: hooks are tilted and sway slightly; Morphology C: hooks are obstructed or entangled with other objects; Morphology D: hooks are severely deformed or damaged; Morphology judgment criteria: Geometric features: hook angle, length, and opening direction; Dynamic features: swaying frequency and amplitude; Environmental features: lighting conditions and background interference.

[0085] The unhooking behavior corresponding to the unhooking position is determined based on the shape of the unhooked part, the unhooking position, and the working mode of the unhooking robot. The shape of the unhooked part is as described above (A / B / C / D categories); the unhooking position includes position coordinates, accessibility, and working condition type (e.g., MIC, EDD); the working mode is the robot's currently set operating mode, such as: standard mode (balancing efficiency and safety); high-efficiency mode (prioritizing speed, suitable for low-risk environments); and safety mode (prioritizing obstacle avoidance and protection, suitable for high-risk environments).

[0086] Unhooking behavior decision logic: The behavior decision is made using a rule engine and a machine learning model. Common unhooking behaviors include: Direct unhooking: Applicable to form A, where the position is stable, and the robot directly performs the unhooking action; Unhooking after adjusting posture: Applicable to form B, where the robot adjusts the angle of the hook before performing the action; Unhooking after obstacle avoidance: Applicable to form C, where there are obstacles and the path needs to be replanned; Alarm shutdown: Applicable to form D, where the hook and ring are abnormal and cannot be safely unhooked, triggering an alarm.

[0087] In the robot control system, a behavior label is attached to each unhooking position; the label content includes: behavior type; execution conditions (e.g., if form = A and working condition = MIC); execution parameters (e.g., speed, force, angle adjustment range); usually stored in JSON or XML structure for easy robot parsing.

[0088] Specifically, assuming the unhooking robot is located in a railway freight station and is performing the task of unhooking the carriage, the specific scenario is as follows: Unhooking position H01: coordinates: [10,5,0]; working condition type: MIC (medium interference complex working condition); working mode: standard mode; the acquired image shows that the hook is suspended vertically without swinging (morphology A); decision behavior: unhooking directly, speed 0.8m / s, force limit 50N; Unhooking position H02: coordinates: [15,5,0]; working condition type: EDD (equipment-dense area); working mode: safe mode; the acquired image shows that the hook is slightly tilted, accompanied by a small swing (morphology B); decision behavior: unhooking after adjusting the posture, the speed is reduced to 0.4m / s, and the hook angle is first adjusted to 15°.

[0089] Unhooking position H03: Coordinates: [23,5,0]; Working condition type: EAO (Abnormal environment area); Working mode: High efficiency mode; Acquired image shows that the hook and loop are partially obscured (shape C); Decision behavior: Unhook after obstacle avoidance and replanning the path, the detour distance increases by 1.2m; Unhooking position H04: Coordinates: [30,5,0]; Working condition type: MIC; Working mode: Standard mode; Acquired image shows that the hook and loop are severely deformed (shape D); Decision behavior: Alarm shutdown, notify manual intervention; By recognizing the shape of the unhooked part in real-time image recognition, combined with the unhooking position and working mode, the robot intelligently decides and marks the best unhooking behavior for each unhooking position; This process not only improves the robot's environmental adaptability, but also significantly enhances the operational safety and task success rate under complex working conditions.

[0090] refer to Figure 5 In step S14, the specific steps are as follows:

[0091] S141: Collect each unhooking behavior, determine the corresponding unhooking path based on the recognition of each unhooking behavior, match the corresponding unhooking path at each unhooking position, and predict the collision area of ​​the unhooking robot based on the matching of the corresponding unhooking path at each unhooking position, the posture of the unhooking robot, and the current image corresponding to the unhooking position.

[0092] S142: Based on the identification of the collision region, determine multiple sub-collision regions, determine multiple collision nodes according to the current image corresponding to the region location, region shape and hooking position of the multiple sub-collision regions, and determine the corresponding collision avoidance space based on the multiple collision nodes and the corresponding hooking behavior;

[0093] S143: Based on the collision avoidance space, the posture of the unhooking robot and the corresponding working mode, the optimized unhooking behavior is determined to gradually avoid multiple predicted collision nodes and eliminate multiple collision nodes, thus ensuring the collision-free operation of the unhooking robot.

[0094] In the embodiments of this application, each unhooking behavior is collected, the corresponding unhooking path is determined based on the identification of each unhooking behavior, and the corresponding unhooking path is matched at each unhooking position. Based on the matching of the corresponding unhooking path at each unhooking position, the posture of the unhooking robot, and the current image corresponding to the unhooking position, the collision area of ​​the unhooking robot is predicted. This takes into account the overall consideration of the identification of each unhooking behavior and ensures the accuracy of the corresponding unhooking path.

[0095] At this point, the unhooking behavior of the unhooking robot at each unhooking position is obtained from step S133 (e.g., "direct unhooking", "unhooking after adjusting posture", "unhooking after obstacle avoidance", etc.). Each behavior is not only an action label, but also contains a series of execution parameters, such as: action type (grabbing, rotating, lifting, translating, etc.); speed range (e.g., 0.2 to 0.5 m / s); force control (e.g., 10 to 50 N); angle range of each joint of the robotic arm; and execution time limit.

[0096] Based on the unhooking behavior, the system matches or generates corresponding unhooking paths from the path library in real time. Each path consists of a series of key points, each containing three-dimensional coordinates (x, y, z) and pose (roll, pitch, yaw). Path types include: straight (suitable for obstacle-free areas), curved (suitable for navigating obstacles), and segmented (suitable for complex multi-obstacle scenarios). The system matches one or more candidate paths for each unhooking location. Matching criteria include: the spatial coordinates of the unhooking location; the distribution of obstacles identified in the current environment image; and the execution requirements of the unhooking behavior (such as accuracy and speed).

[0097] Based on the matching of the corresponding unhooking path at each unhooking position, the posture of the unhooking robot, and the current image corresponding to the unhooking position, the collision area of ​​the unhooking robot is predicted. Input elements are integrated: Unhooking path: a spatial trajectory composed of a sequence of key points.

[0098] Robot posture: including base position, joint angles, and end effector posture; Current image: environmental image acquired through a vision system, identifying static and dynamic obstacles; Collision region prediction method: modeling the robot's motion trajectory using bounding boxes or voxel grids; dividing the space occupied by the robot during movement into multiple small regions (voxels), each voxel labeled as "occupied" or "free"; identifying obstacles in the environment using image recognition technologies (such as YOLO, Mask R-CNN); mapping the position and size of obstacles to the same coordinate system as the robot; static obstacles (such as equipment, pipes) are directly marked as permanently occupied areas; dynamic obstacles (such as people, mobile devices) are predicted to have future positions based on their motion trajectories.

[0099] Check if each voxel of the robot's motion trajectory overlaps with a voxel occupied by an obstacle; if they overlap, mark the area as a "collision region"; the collision region is not just a point, but a three-dimensional spatial range, including: a core region with a high probability of collision; and edge regions where collisions occur; output: a list of collision regions, each region containing: region coordinates (x, y, z); region dimensions (length, width, height); collision probability (based on obstacle motion prediction); and associated hook-off path segment.

[0100] Specifically, assuming the unhooking robot is performing a task, there are currently three unhooking positions (H01, H02, H03), and the unhooking behavior and environment are different at each position.

[0101] Unhooking position H01: Unhooking behavior: Direct unhooking; Current image: Shows a fixed support (static obstacle) in front of the unhooking position; Unhooking path: Straight path, from the starting point [10,5,1] to the ending point [10,5,0.5]; Robot posture: The robotic arm is vertically downward, and the end effector is facing the hook; Collision area prediction: The path overlaps with the bounding box of the support at [10,5,0.7]; Collision area coordinates: [9.8,4.8,0.6] to [10.2,5.2,0.8]; Collision probability: 100% (static obstacle); Output: Collision area C01.

[0102] Unhooking position H02: Unhooking behavior: Unhooking after adjusting posture; Current image: Shows a worker moving near the path (dynamic obstacle); Unhooking path: Curved path, from the starting point [20,10,1] to the ending point [20,10,0.5]; Robot posture: The robotic arm is tilted at 30 degrees, and the end effector is deflected at 15 degrees; Collision area prediction: The worker's activity area overlaps with the path at [20,10,0.7]; Collision area coordinates: [19.5,9.5,0.6] to [20.5,10.5,0.8]; Collision probability: 60% (based on worker's movement speed prediction); Output: Collision area C02.

[0103] Unhooking position H03: Unhooking behavior: Unhooking after obstacle avoidance; Current image: Shows multiple intersecting pipes, forming a complex obstacle; Unhooking path: Segmented path, from the starting point [30,15,1] through the intermediate point [30,15.5,1.2] to the ending point [30,15,0.5]; Robot posture: Multi-joint linkage of the robotic arm, with dynamic adjustment of the end effector; Collision area prediction: The path overlaps with pipes at multiple points; Collision area coordinates: C03: [29.8,14.8,1.0] to [30.2,15.2,1]. .4];C04: [29.8,15.3,0.8] to [30.2,15.7,1.0];Collision probability: 80% (fixed pipes but dense paths);Output: Collision areas C03, C04;Through accurate hook removal behavior recognition, path matching and multi-dimensional collision prediction, key technical support is provided for the safe operation of the hook removal robot in complex working conditions. This method not only considers the robot's own motion characteristics, but also effectively copes with static and dynamic obstacles in the environment, significantly improving the safety and reliability of the system.

[0104] Furthermore, multiple sub-collision regions are determined based on the identification of the collision region. Multiple collision nodes are determined based on the current image corresponding to the regional location, regional shape, and unhooking position of the multiple sub-collision regions. The corresponding collision avoidance space is determined based on the multiple collision nodes and the corresponding unhooking behavior. This approach takes into account the overall consideration of multiple collision nodes and the corresponding unhooking behavior, ensuring the accuracy of the corresponding collision avoidance space.

[0105] At this point, the purpose of sub-collision region division is to further subdivide the initially predicted collision regions (such as C01, C02, C03, etc. output by S141) into smaller units for more precise avoidance planning. The division criteria include: spatial density: areas with dense obstacle distribution are divided into finer units; risk level: high-risk areas (such as personnel activity areas, precision equipment areas) are divided into finer units; path characteristics: areas with large path curvature and frequent turns are divided into finer units. Division methods include: grid division: dividing the collision region into uniform cubic grids (such as 0.1m × 0.1m × 0.1m); adaptive division: dynamically adjusting the size of sub-regions based on obstacle distribution; boundary expansion: expanding the safety buffer zone outside the obstacle boundary to form multiple sub-regions. Each sub-region includes: spatial coordinates (x, y, z); region shape (cube, cylinder, irregular polyhedron, etc.); risk level (high, medium, low); obstacle type (static, dynamic, semi-dynamic).

[0106] Multiple collision nodes are determined based on the current image corresponding to the regional location, regional shape, and unhooking position of multiple sub-collision regions. The collision node is defined as the key point in the sub-collision region with the highest risk and where a collision will occur. The node selection criteria are: closest to the robot path; highest obstacle density; and the point through which dynamic obstacles pass.

[0107] Node determination methods: Nearest neighbor method: select the center point of the sub-region closest to the path; Density peak method: select the sub-region with the highest obstacle density; Dynamic prediction method: predict the trajectory of dynamic obstacles and select the intersecting points; Node attributes: Each collision node includes: three-dimensional coordinates (x, y, z); risk level; obstacle type; predicted collision time (for dynamic obstacles).

[0108] Based on multiple collision nodes and the corresponding unhooking behavior, the corresponding collision avoidance space is determined. The avoidance space is defined as a temporary safe area planned by the robot to avoid collision nodes. The avoidance space can be: point-like: temporary stopping point; line-like: detour path; surface-like: temporary working plane; volume-like: three-dimensional safe space.

[0109] Avoidance space generation methods: Geometric offset method: generates an offset region around the collision node; Trajectory replanning method: replans the robot path to bypass the collision node; Time window method: plans time windows to pass through dynamic obstacles in staggered time windows; Avoidance space attributes: Each avoidance space includes: spatial range (start point, end point, boundary); time range (start time, end time); applicable unhooking behavior type; safety margin (minimum safe distance from the obstacle).

[0110] Specifically, the unhooking robot performs the unhooking task; the work area includes multiple static devices (such as conveyor belts and control cabinets) and dynamic obstacles (such as moving assembly vehicles and workers); the unhooking positions H01, H02, and H03 are located at different workstations.

[0111] Collision Zone C01 (near H01): Divided into 3 sub-zones: SA01: [10.0, 5.0, 0.8] to [10.2, 5.2, 1.0] (near the conveyor belt); SA02: [10.2, 5.0, 0.8] to [10.4, 5.2, 1.0] (near the control cabinet); SA03: [10.0, 5.2, 0.8] to [10.2, 5.4, 1.0] (worker activity area); Collision Zone C02 (near H02): Divided into 2 sub-zones: SB01: [20.3, 10.3, 0.7] to [20.5, 10.5, 0.9] (assembly vehicle aisle); SB02: [20.5, 10.3, 0.7] to [20.7, 10.5, 0.9] (equipment edge).

[0112] Collision Node Determination: H01 Unhook Position: Sub-region SA01: Shortest Path, Select Node N01 ([10.1,5.1,0.9]); Sub-region SA03: Dense Dynamic Obstacles, Select Node N02 ([10.1,5.3,0.9]); H02 Unhook Position: Sub-region SB01: Frequent Passage of Assembly Vehicles, Select Node N03 ([20.4,10.4,0.8]); Sub-region SB02: Static Equipment, Select Node N04 ([20.6,10.4,0.8]); Collision Avoidance Space Generation: H01 Unhook Position: Node N01: Avoidance Space: Linear Detour Path, from [10.0,5.0,0.8] through [10.0 [4.8,0.9] to [10.2,5.2,1.0]; Applicable behavior: Unhook after adjusting attitude; Safety margin: 0.2m; Node N02: Avoidance space: Time window, pass through after a 2-second delay; Applicable behavior: Unhook after obstacle avoidance; Safety margin: 0.5m; H02 Unhook position: Node N03: Avoidance space: Point-like stopping point [20.3,10.2,0.8], waiting for the assembly vehicle to pass; Applicable behavior: Unhook directly; Safety margin: 0.3m; Node N04: Avoidance space: Planar safety plane, from [20.5,10.2,0.7] to [20.7,10.6,0.9]; Applicable behavior: Unhook after obstacle avoidance; Safety margin: 0.4m.

[0113] By refining the sub-collision region division, collision node identification, and avoidance space generation, this method provides key technical support for the safe operation of the unhooking robot under complex working conditions. It not only considers the spatial dimension but also incorporates the temporal dimension, effectively dealing with static and dynamic obstacles and significantly improving the system's safety and adaptability.

[0114] Therefore, based on the collision avoidance space, the posture of the unhooking robot, and the corresponding working mode, the optimized unhooking behavior is determined to gradually avoid multiple predicted collision nodes and eliminate multiple collision nodes, ensuring the collision-free operation of the unhooking robot. This approach takes into account the overall considerations of the collision avoidance space, the posture of the unhooking robot, and the corresponding working mode, ensuring the accuracy of the optimized unhooking behavior.

[0115] At this point, the optimized unhooking behavior is determined based on the collision avoidance space, the robot's posture, and the corresponding working mode. Input conditions: Collision avoidance space: the avoidance space generated by S142 (such as linear path, time window, point stopping point, planar safety plane, etc.); Unhooking robot posture: including current position, joint angle, end effector state, etc.; Working mode: such as "standard mode", "high precision mode", "fast mode", etc., which affects path planning and action execution strategy; Decision logic for optimizing unhooking behavior: Avoidance strategy selection: if the avoidance space is a linear path, a path tracking strategy is adopted; if the avoidance space is a time window, a delay waiting strategy is adopted; if the avoidance space is a point stopping point, a pause observation strategy is adopted; if the avoidance space is a planar safety plane, a free movement strategy within the plane is adopted.

[0116] Attitude adjustment strategy: Adjust the robot's attitude according to the avoidance space shape (e.g., adjust the joint angle of the robotic arm, change the orientation of the end effector); ensure stability and operability during avoidance; Working mode adaptation: Standard mode: balance speed and safety; High precision mode: prioritize motion accuracy and appropriately reduce speed; Fast mode: maximize efficiency under the premise of safety; Optimized hook unhooking line includes: motion sequence (e.g., "forward → pause → adjust attitude → unhook"); speed curve (e.g., acceleration → constant speed → deceleration); force control parameters (e.g., gripping force, rotation force); safety monitoring parameters (e.g., collision detection threshold, emergency stop conditions).

[0117] Multiple predicted collision nodes are gradually avoided and eliminated. The gradual avoidance strategy is as follows: Prioritization: Collision nodes are sorted according to factors such as risk level, spatial location, and time window; high-risk nodes are dealt with first, and dynamic nodes are given priority over static nodes; Phased avoidance: Phase 1: Avoid high-risk nodes (such as areas with personnel activity and areas with precision equipment); Phase 2: Avoid medium-risk nodes (such as fixed obstacles and slow-moving mobile equipment); Phase 3: Avoid low-risk nodes (such as temporary storage and non-critical equipment).

[0118] Collision Node Elimination Mechanism: Spatial Elimination: The robot trajectory completely avoids the space where the node is located through path planning; Verification Method: Spatial interference detection algorithm (such as OBB bounding box detection); Temporal Elimination: The robot's passage time through the node is staggered from the obstacle's activity time through delay or acceleration; Verification Method: Time window matching algorithm; Dynamic Elimination: Real-time monitoring of obstacle positions and dynamic path adjustment; Verification Method: Dynamic path replanning algorithm; Collision-Free Operation Guarantee: Real-time Monitoring: Real-time detection of environmental changes through sensors (such as LiDAR, vision system); If a new collision risk occurs, replanning is immediately triggered; Emergency Mechanism: Setting multiple safety thresholds (such as distance threshold, speed threshold); Automatic emergency stop or obstacle avoidance action is triggered when the threshold is exceeded; Through intelligent avoidance strategies, phased node elimination, and strict safety monitoring, collision-free operation of the unhooking robot under complex working conditions is achieved. This method not only considers the spatial and temporal dimensions but also adapts to dynamically changing environments, providing a reliable guarantee for safe operation in industrial automation.

[0119] refer to Figure 6 In step S15, the specific steps are as follows:

[0120] S151: Divide each unhooking behavior into multiple optimized unhooking behaviors and the original unhooking behavior. Determine the first anti-collision coefficient based on the optimized unhooking behavior and the unhooking robot's movement path. Determine the second anti-collision coefficient based on the original unhooking behavior and the unhooking robot's movement path.

[0121] S152: Determine the anti-collision processing system of the unhooking robot based on the first anti-collision coefficient, the second anti-collision coefficient, and the anti-collision mapping relationship; In the anti-collision processing system, the working process of the unhooking robot is dynamically monitored, and the safe distance between the posture of the unhooking robot during the working process and the unhooking space of the unhooking position is marked, and the safe distance is dynamically maintained to be greater than the preset safe distance threshold.

[0122] S153: With the assistance of the anti-collision processing system, the unhooking robot sequentially performs the corresponding unhooking behavior at each unhooking position and marks the unhooking efficiency of the current unhooking behavior. Based on the unhooking efficiency of the current unhooking behavior, the current image of the remaining unhooking positions, and the shape of the unhooked part, the work efficiency system is determined, and the unhooking efficiency of the unhooking robot at the remaining unhooking positions is gradually improved along this work efficiency system.

[0123] In the embodiments of this application, each unhooking behavior is divided into multiple optimized unhooking behaviors and the original unhooking behavior. A first anti-collision coefficient is determined based on the optimized unhooking behavior and the unhooking robot's movement path, and a second anti-collision coefficient is determined based on the original unhooking behavior and the unhooking robot's movement path. This approach takes into account the overall consideration of the original unhooking behavior and the unhooking robot's movement path, ensuring the accuracy of the second anti-collision coefficient.

[0124] At this point, the optimized unhooking behavior has been adjusted through step S143 and has obstacle avoidance capability, path optimization and attitude adjustment features, making it suitable for complex or high-risk environments; the original unhooking behavior is the initially set unhooking behavior, which has not been optimized for obstacle avoidance and is suitable for barrier-free or low-risk environments.

[0125] Classification criteria: Path characteristics: Optimized behavior: The path bypasses obstacles, is longer but safer; Original behavior: The path is a straight line or a simple curve, is shorter but has a risk of collision.

[0126] Posture characteristics: Optimized behavior: Frequent joint angle adjustments, smooth posture changes; Original behavior: Less joint angle change, fixed posture;

[0127] Action characteristics: Optimized behavior: includes obstacle avoidance actions (such as waiting, detouring, and slowing down); Original behavior: only includes basic hook removal actions (such as grabbing, rotating, and lifting);

[0128] Classification process: Traverse all unhooking behaviors; check if the behavior contains obstacle avoidance actions or path optimization; mark it as an optimized behavior or an existing behavior; generate a behavior classification list.

[0129] The first anti-collision coefficient (C1) is determined based on the optimized unhooking behavior and the unhooking robot's movement path. The first anti-collision coefficient (C1) is defined as follows: it is used to evaluate the anti-collision capability of the optimized unhooking behavior; its value range is 0 to 1, with a larger value indicating stronger anti-collision capability; the calculation method is as follows: Path safety score (S1): calculates the minimum distance (d_min) between the path and obstacles; if d_min > safety threshold (e.g., 0.5m), S11; if d_min < safety threshold, S1d_min / safety threshold; Posture stability score (S2): checks whether the joint angles are within the safe range; if all relevant... If all joint angles are within the safe range, S21; otherwise, S2: number of joints within the safe range / total number of joints; motion smoothness score (S3): calculate the rate of change of velocity (Δv); if Δv < smoothing threshold (e.g., 0.1m / s2), S31; otherwise, S3: smoothing threshold / Δv; comprehensive calculation: C1 = (S1 × 0.5) + (S2 × 0.3) + (S3 × 0.2); calculation process: extract the movement path of the optimized behavior; calculate the path safety score; extract robot posture data; calculate posture stability score; extract motion velocity data; calculate motion smoothness score; comprehensive calculation C1.

[0130] The second anti-collision coefficient (C2) is determined based on the original unhooking behavior and the unhooking robot's movement path. The second anti-collision coefficient (C2) is defined as: used to evaluate the anti-collision capability of the original unhooking behavior; its value range is 0 to 1, with a larger value indicating stronger anti-collision capability; the calculation method is similar to C1, but the path safety score is usually lower (because it is not optimized); the path safety score (S1) is: the original behavior path is usually a straight line or a simple curve, close to obstacles; the calculation method for S1 is the same as C1, but d_min is usually smaller; the posture stability score (S2) is: the original behavior posture is fixed, and S2 is usually 1; the motion smoothness score (S3) is: the original behavior motion is simple, and S3 is usually 1; the overall calculation is: C2 = (S1 × 0.5) + (S2 × 0.3) + (S3 × 0.2); the calculation process is: extract the movement path of the original behavior; calculate the path safety score; extract robot posture data; calculate the posture stability score; extract motion speed data; calculate the motion smoothness score; and calculate C2.

[0131] Specifically, the unhooking robot needs to perform the unhooking task at three unhooking locations (P01, P02, P03); P01: no obstacles, using the original unhooking behavior; P02: with static obstacles (pipes), using the optimized unhooking behavior; P03: with dynamic obstacles (moving vehicles), using the optimized unhooking behavior.

[0132] Behavior classification: P01: Original unhooking behavior (straight path, fixed posture); P02: Optimized unhooking behavior (detour path, posture adjustment); P03: Optimized unhooking behavior (waiting + detour path, posture adjustment).

[0133] Calculate C1 (optimized behavior): P02: Path safety score: d_min = 0.6m > 0.5m; S11: Posture stability score: All joints are within the safe range; S21: Movement smoothness score: Δv = 0.08m / s 2 <0.1m / s 2 S31; C1=(1×0.5)+(1×0.3)+(1×0.2)=1; P03: Path safety score: d_min=0.4m<0.5m, S1=0.4 / 0.50.8; Posture stability score: All joints are within the safe range, S2=1; Movement smoothness score: Δv=0.12m / s 2 >0.1m / s 2 , S3=0.1 / 0.12=0.83; C1=(0.8×0.5)+(1×0.3)+(0.83×0.2)=0.4+0.3+0.166=0.866.

[0134] Calculate C2 (existing behavior):

[0135] P01: Path safety score: d_min=0.3m<0.5m, S1=0.3 / 0.5=0.6; Posture stability score: All joints are within the safe range, S21; Movement smoothness score: Δv=0.05m / s2<0.1m / s2, S3=1; C2=(0.6×0.5)+(1×0.3)+(1×0.2)=0.3+0.3+0.2=0.8.

[0136] Output results: Behavior classification: P01: Original behavior; P02: Optimized behavior; P03: Optimized behavior; Collision avoidance coefficient: P01: C2 = 0.8; P02: C1 = 1; P03: C1 = 0.866.

[0137] Furthermore, an anti-collision system for the unhooking robot is determined based on the first anti-collision coefficient, the second anti-collision coefficient, and the anti-collision mapping relationship. In the anti-collision system, the working process of the unhooking robot is dynamically monitored, and the safe distance between the robot's posture during the working process and the unhooking space at the unhooking position is marked. This safe distance is dynamically maintained to be greater than a preset safe distance threshold. This system takes into account the overall consideration of the first anti-collision coefficient, the second anti-collision coefficient, and the anti-collision mapping relationship, ensuring the accuracy of the anti-collision system for the unhooking robot.

[0138] At this time, a collision avoidance processing system for the hook unhooking robot is determined based on the first collision avoidance coefficient, the second collision avoidance coefficient, and the collision avoidance mapping relationship; the first collision avoidance coefficient (C1): for the optimized hook unhooking behavior, reflecting the obstacle avoidance ability; the second collision avoidance coefficient (C2): for the original hook unhooking behavior, reflecting the basic collision avoidance ability; the collision avoidance mapping relationship: a predefined rule or model for mapping the collision avoidance coefficient into a specific processing strategy.

[0139] Construction of the collision avoidance processing system: Collect the comprehensive collision avoidance index I: Determine the processing strategy according to the mapping relationship: I>0.9: Advanced obstacle avoidance mode (real-time path replanning, dynamic obstacle avoidance); 0.7≤I≤0.9: Standard obstacle avoidance mode (fixed obstacle avoidance path, attitude adjustment); I<0.7: Basic obstacle avoidance mode (deceleration, alarm); Generate a processing system document, which includes strategies, trigger conditions, execution processes, monitoring indicators, etc.

[0140] In the collision avoidance processing system, dynamically monitor the working process of the hook unhooking robot. Monitoring targets: The attitude of the hook unhooking robot (joint angles, end position); The hook space at the hook unhooking position (obstacle distribution, safety area); The actual safety distance (the minimum distance between the robot and the obstacle); Monitoring methods: Sensor data collection: LiDAR: Measure the distance to the obstacle; IMU: Monitor the robot's attitude; Vision system: Identify the hook space.

[0141] Calculate the safety distance d in real time; Update the attitude matrix T; Generate a monitoring report; Monitoring process: Initialize monitoring parameters (sampling frequency, alarm threshold); Continuously collect sensor data in a loop; Calculate the safety distance and attitude deviation; Determine whether to trigger an alarm; Record the monitoring data.

[0142] Mark the safety distance between the attitude of the hook unhooking robot during the working process and the hook space at the hook unhooking position, and dynamically maintain this safety distance to be greater than the preset safety distance threshold. Definition of the safety distance: The preset safety distance threshold dth: Set according to the working conditions, such as 0.5m; Can be dynamically adjusted (such as increasing the threshold at high speed); Safety distance marking: In the monitoring interface: Mark the safety status with different colors: Green: d>dth; Yellow: dthX0.8<d≤dth; Red: d≤dthX0.8; In the data record: Store the timestamp, position, and safety distance value; Dynamic maintenance method: When d≤dth: Trigger obstacle avoidance actions: Decelerate; Adjust the attitude; Plan a new path; Update the processing strategy: Switch to a more advanced obstacle avoidance mode; Increase the monitoring frequency; Verify the adjustment effect: Recalculate the safety distance; Continue working after confirming d>dth.

[0143] Specifically, the unhooking robot performs steel ladle unhooking operations; unhooking positions: P01, P02, P03; obstacles: pipes, supports, mobile equipment; preset safety distance threshold dth 0.5m; specific execution process: calculate collision avoidance coefficients: P01: C2 = 0.8 (original behavior); P02: C1 = 1 (optimized behavior); P03: C1 = 0.866 (optimized behavior); comprehensive index I = (0.5 x 1) + (0.5 x 0.8) = 0.9; determine the collision avoidance system: mapping relationship: I > 0.9 → advanced obstacle avoidance mode.

[0144] Processing strategy: Real-time path replanning; dynamic obstacle avoidance; high-frequency monitoring (10Hz); Dynamic monitoring execution: At position P01: Data acquisition: LiDAR measured d = 0.6m; IMU showed normal attitude; vision system identified no change in the hook-off space; marking status: green (safe); At position P02: Data acquisition: LiDAR measured d = 0.4m; IMU showed attitude deviation of 5°; vision system identified a new obstacle; marking status: yellow (warning); Obstacle avoidance triggered: decelerate to 0.05m / s; adjust attitude; replan path; Verify adjustment: d under the new path =0.55m; Marking status: Green (safe); Dynamically maintain safe distance: At position P03: Initial d = 0.45m; Trigger obstacle avoidance: Switch to advanced obstacle avoidance mode; Adjust path to bypass obstacles; Final d = 0.6m; Marking status: Green (safe); Output result: Collision avoidance system: Advanced obstacle avoidance mode; Monitoring record: P01: d = 0.6m (green); P02: d = 0.4m → 0.55m (yellow → green); P03: d = 0.45m → 0.6m (yellow → green); Safe distance maintenance: d > 0.5m throughout.

[0145] Therefore, with the assistance of the anti-collision processing system, the unhooking robot sequentially performs corresponding unhooking actions at each unhooking position and marks the unhooking efficiency of the current unhooking action. Based on the unhooking efficiency of the current unhooking action, the current image of the remaining unhooking positions, and the shape of the unhooked item, a work efficiency system is determined. The unhooking efficiency of the unhooking robot at the remaining unhooking positions is gradually improved along this work efficiency system. This system takes into account the overall consideration of the unhooking efficiency of the current unhooking action, the current image of the remaining unhooking positions, and the shape of the unhooked item, ensuring the accuracy of the work efficiency system. At the same time, multiple collision nodes are introduced to further ensure the accuracy of the optimized unhooking actions. This system realizes the overall consideration of multiple optimized unhooking actions, the original unhooking actions, and the unhooking robot's movement path, improving the accuracy of the anti-collision processing system of the unhooking robot and progressively improving the unhooking efficiency of the unhooking robot.

[0146] At this time, with the assistance of the anti-collision processing system, the unhooking robot performs the corresponding unhooking behavior at each unhooking position in sequence, and marks the unhooking efficiency of the current unhooking behavior. The unhooking robot visits each unhooking position in sequence according to the strategy determined by the anti-collision processing system (S152); each position executes the corresponding unhooking behavior (optimized behavior or original behavior).

[0147]

[0148] At this point, T: Time required to complete unhooking (seconds); D: Path length (meters); A: Number of attitude adjustments; α, β: Weighting coefficients, typically α = 0.1, β = 0.2; Efficiency rating: High efficiency: E > 0.03; Medium efficiency: 0.02 ≤ E ≤ 0.03; Low efficiency: E < 0.02; Records for each unhooking position: Unhooking behavior type; Execution time; Path length; Number of attitude adjustments; Calculated efficiency value E.

[0149] The efficiency system is determined based on the current hook removal efficiency, the current image of the remaining hook removal positions, and the shape of the hooked item. Input data: Efficiency of the current hook removal action; current image of the remaining hook removal positions (from the vision system); shape of the hooked item (e.g., hook type, size, material, etc.); Efficiency system construction: Image analysis: using image processing algorithms (e.g., YOLO, OpenCV) to identify: obstacle location; shape of the hooked item; passable area; Shape classification: hooked item shape classification: simple hook type: such as standard round hook; complex hook type: such as irregular hook, multi-hook combination; dynamic hook type: such as swing hook.

[0150] Efficiency system rules: Based on the current efficiency (E) and shape type, formulate efficiency improvement strategies: Simple hook type: increase speed and reduce posture adjustment; Complex hook type: increase accuracy and appropriately reduce speed; Dynamic hook type: real-time tracking and dynamic path adjustment; Generate efficiency system document: including: current efficiency (E); image analysis results of the remaining positions; shape of the hook being removed; efficiency improvement strategies.

[0151] The unhooking efficiency of the robot at the remaining unhooking positions is gradually improved according to this efficiency system. Based on the efficiency system document, the unhooking behavior at the remaining positions is adjusted as follows: Path optimization: shorten the path length D; Speed ​​adjustment: optimize the time T; Posture optimization: reduce the number of adjustments A; Dynamic adjustment: real-time monitoring: LiDAR, IMU, and vision system provide real-time data; dynamically update the path and posture; Iterative optimization: recalculate E after each unhooking position is completed; update the efficiency system document; adjust subsequent strategies; Output result: the efficiency of the unhooking behavior at the remaining positions gradually improves; ultimately maximizing overall efficiency.

[0152] Specifically, the unhooking robot needs to complete the unhooking task at three locations (P01, P02, P03); Collision avoidance system: advanced obstacle avoidance mode; Preset safe distance threshold: 0.5m; Shape of the unhooked part: P01: simple hook type; P02: complex hook type; P03: dynamic hook type; P01 position: Unhooking behavior: original behavior (straight path); Time T = 10s; Path length D = 2m; Number of posture adjustments A = 1; E = 0.096; Mark: efficient (E>0.03), Behavior type: original behavior;

[0153] P02 Position: Current Image Analysis: Complex hook shape identified; Obstacle detected; Efficiency System Construction: Current E = 0.096; Shape: Complex hook shape; Strategy: Increase accuracy, reduce speed; Execute hook removal behavior: Optimized behavior (detour path); T = 15s; Path length D = 3m; Number of attitude adjustments A = 2; E = 0.064, Mark: High efficiency (E > 0.03), Behavior type: Optimized behavior;

[0154] P03 Position: Current Image Analysis: Identify dynamic hook shape; Detect swinging obstacle; Efficiency System Construction: E = 0.064; Shape: Dynamic hook shape; Strategy: Real-time tracking, dynamic adjustment; Execute hook removal behavior: Optimized behavior (dynamic tracking path); Time T = 12s; Path length D = 2.5m; Number of attitude adjustments A = 1; E = 0.080; Mark: High efficiency (E > 0.03); Behavior type: Optimized behavior; Output results: Hook removal behavior efficiency: P01: 0.096; P02: 0.064; P03: 0.080; Efficiency improvement: P02 and P03, through optimized behavior, maintain high efficiency despite the increased path length, through dynamic adjustment and real-time tracking; Overall efficiency: Average E ≈ 0.080, higher than the initial setting of 0.03.

[0155] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of the anti-collision system for the unhooking robot under complex working conditions in an embodiment of the present invention; the anti-collision system for the unhooking robot under complex working conditions includes:

[0156] Previous working image module 21 is used to determine multiple sub-working areas based on the detection of the working area of ​​the unhooking robot, and to mark the previous working images of the multiple sub-working areas;

[0157] The multimodal data module 22 is used to determine multiple working condition data of the unhooking robot based on previous working images of multiple sub-working areas, and to determine the corresponding complex working condition type based on multiple working condition data, the shape of the unhooking robot and the service life of the unhooking robot.

[0158] The unhooking behavior module 23 is used to determine multiple unhooking positions in the work area based on the work area, the type of complex working conditions, and the unhooking robot, and to determine the unhooking behavior of the unhooking robot at each unhooking position according to the current image corresponding to the multiple unhooking positions and the working mode of the unhooking robot.

[0159] The optimization module 24 is used to predict the collision area of ​​the unhooking robot based on the current image corresponding to each unhooking behavior and unhooking position, determine multiple collision nodes based on the identification of the collision area, and determine the optimized unhooking behavior based on the multiple collision nodes, the corresponding unhooking behavior and the posture of the unhooking robot to eliminate multiple collision nodes.

[0160] The anti-collision processing system module 25 is used to determine the anti-collision processing system of the unhooking robot based on multiple optimized unhooking behaviors, the original unhooking behaviors and the unhooking robot's movement path, and to improve the unhooking efficiency of the unhooking robot step by step.

[0161] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A complex working condition unhooking robot anti-collision processing method, characterized in that, The method comprises the following steps: According to the detection of the working area of the hooking robot, a plurality of sub-working areas are determined, and the past working images of the plurality of sub-working areas are marked. According to the past working images of the plurality of sub-working areas, a plurality of working condition data of the hooking robot are determined, according to the plurality of working condition data, the shape of the hooking robot and the service life of the hooking robot, a corresponding complex working condition type is determined, comprising: collecting the past working images of the plurality of sub-working areas, determining a plurality of working condition characteristics according to the recognition of the past working images of the plurality of sub-working areas, and determining a working condition area according to the position, feature shape and moving path of the hooking robot relative to the sub-working area; according to the detection of the working condition area, a plurality of working condition data of the hooking robot are determined, at the same time, the shape of the hooking robot and the service life of the hooking robot are collected, and a corresponding multi-modal data is constructed according to the plurality of working condition data, the shape of the hooking robot and the service life of the hooking robot; according to the detection of the multi-modal data, a plurality of working condition combinations are determined, and according to the recognition of the plurality of working condition combinations, a corresponding complex working condition type is determined; Based on the working area, the complex working condition type and the hooking robot, a plurality of hooking positions of the working area are determined, and according to the current image corresponding to the plurality of hooking positions and the working mode of the hooking robot, the hooking behavior of the hooking robot at each hooking position is determined; Based on each hooking behavior and the current image corresponding to the hooking position, the collision area of the hooking robot is predicted, and according to the recognition of the collision area, a plurality of collision nodes are determined; according to the plurality of collision nodes, the corresponding hooking behavior and the posture of the hooking robot, an optimized hooking behavior is determined to eliminate the plurality of collision nodes; According to the plurality of optimized hooking behaviors, the original hooking behavior and the moving path of the hooking robot, a collision avoidance processing system of the hooking robot is determined, and the hooking working efficiency of the hooking robot is improved step by step.

2. The complex case unhooking robot anti-collision processing method according to claim 1, characterized in that, According to the detection of the working area of the hooking robot, a plurality of sub-working areas are determined, and the past working images of the plurality of sub-working areas are marked, comprising: The model of the hooking robot is collected, the working area of the hooking robot is determined according to the matching of the signal of the hooking robot and the work schedule table, and a plurality of sub-working areas are determined based on the area position, area shape and work schedule table of the working area of the hooking robot; According to the detection of each sub-working area, a corresponding static node and a dynamic node are determined, and the node position corresponding to the static node and the node position corresponding to the dynamic node are collected, and the past working images of the plurality of sub-working areas are determined based on the tracing of the node position corresponding to the static node and the node position corresponding to the dynamic node.

3. The complex case unhooking robot anti-collision processing method according to claim 1, characterized in that, The plurality of hooking positions of the working area are determined based on the working area, the complex working condition type and the hooking robot, and the hooking behavior of the hooking robot at each hooking position is determined according to the current image corresponding to the plurality of hooking positions and the working mode of the hooking robot, comprising: The collection working area, based on the working area and the corresponding complex working condition category, determine a plurality of working condition nodes, and mark the complex working condition category of the working condition node, according to the node position of the plurality of working condition nodes and the moving path of the hooking robot relative to the working area, determine the first hooking path of the working area; According to the complex working condition category corresponding to the plurality of working condition nodes and the moving path of the hooking robot relative to the working area, determine the second hooking path of the working area; based on the first hooking path and the second hooking path, construct a plurality of hooking positions of the working area.

4. The complex case unhooking robot anti-collision processing method according to claim 3, characterized in that, The plurality of hooking positions of the working area are determined based on the working area, the complex working condition category and the hooking robot, and the hooking behavior of the hooking robot at each hooking position is determined according to the current image corresponding to the plurality of hooking positions and the working mode of the hooking robot, which further comprises: Collect the current image corresponding to the plurality of hooking positions, and determine the form of the corresponding hooked part according to the recognition of the current image, determine the hooking behavior corresponding to the hooking position according to the form of the hooked part, the hooking position and the working mode of the hooking robot, and mark the hooking behavior of the hooking robot at each hooking position.

5. The complex case unhooking robot anti-collision processing method according to claim 1, characterized in that, The collision area of the hooking robot is predicted based on the current image corresponding to each hooking behavior and hooking position, and a plurality of collision nodes are determined according to the recognition of the collision area; the optimized hooking behavior is determined according to the plurality of collision nodes, the corresponding hooking behavior and the posture of the hooking robot, so as to eliminate the plurality of collision nodes, comprising: Collect each hooking behavior, determine the corresponding hooking path according to the recognition of each hooking behavior, and match the corresponding hooking path at each hooking position, predict the collision area of the hooking robot based on the matching of the corresponding hooking path at each hooking position, the posture of the hooking robot and the current image corresponding to the hooking position.

6. The complex case unhooking robot anti-collision processing method according to claim 5, characterized in that, The collision area of the hooking robot is predicted based on the current image corresponding to each hooking behavior and hooking position, and a plurality of collision nodes are determined according to the recognition of the collision area; the optimized hooking behavior is determined according to the plurality of collision nodes, the corresponding hooking behavior and the posture of the hooking robot, so as to eliminate the plurality of collision nodes, comprising: Based on the recognition of the collision area, a plurality of sub-collision areas are determined, a plurality of collision nodes are determined according to the area position, the area form of the plurality of sub-collision areas and the current image corresponding to the hooking position, and the corresponding collision avoidance space is determined based on the plurality of collision nodes and the corresponding hooking behavior; According to the collision avoidance space, the posture of the hooking robot and the corresponding working mode, the optimized hooking behavior is determined, so as to gradually avoid the plurality of predicted collision nodes and eliminate the plurality of collision nodes, and the collision-free operation of the hooking robot is ensured.

7. The complex case unhooking robot anti-collision processing method according to claim 1, characterized in that, The anti-collision processing system of the hooking robot is determined according to the plurality of optimized hooking behaviors, the original hooking behavior and the moving path of the hooking robot, and the hooking work efficiency of the hooking robot is improved step by step, comprising: Each hooking behavior is divided into a plurality of optimized hooking behaviors and original hooking behaviors, a first anti-collision coefficient is determined according to the optimized hooking behavior and the moving path of the hooking robot, and a second anti-collision coefficient is determined according to the original hooking behavior and the moving path of the hooking robot; Determine the anti-collision processing system of the hooking robot based on the first anti-collision coefficient, the second anti-collision coefficient and the anti-collision mapping relationship; in the anti-collision processing system, dynamically monitor the working process of the hooking robot, mark the safety distance between the posture of the hooking robot in the working process and the hooking space of the hooking position, and dynamically maintain the safety distance greater than the preset safety distance threshold.

8. The complex case unhooking robot anti-collision processing method according to claim 7, characterized in that, The anti-collision processing system of the hooking robot is determined according to the plurality of optimized hooking behaviors, the original hooking behavior and the moving path of the hooking robot, and the hooking working efficiency of the hooking robot is gradually improved, and the anti-collision processing system of the hooking robot further comprises: The hooking robot sequentially performs corresponding hooking behaviors on each hooking position under the assistance of the anti-collision processing system, marks the hooking working efficiency of the current hooking behavior, determines a working efficiency system according to the hooking working efficiency of the current hooking behavior, the current image of the remaining hooking position and the form of the hooked part, and gradually improves the hooking working efficiency of the hooking behavior of the hooking robot at the remaining hooking position along the working efficiency system.

9. A complex working condition unhooking robot anti-collision processing system, characterized in that, The anti-collision processing system of the hooking robot under complex working conditions is applied to the anti-collision processing method of the hooking robot under complex working conditions as claimed in any one of claims 1-8, and the anti-collision processing system of the hooking robot under complex working conditions comprises: A past working image module is configured to determine a plurality of sub-working areas according to detection of a working area of the hooking robot, and mark past working images of the plurality of sub-working areas. A complex working condition type module is configured to determine a plurality of working condition data of the hooking robot according to the past working images of the plurality of sub-working areas, determine a corresponding complex working condition type according to the plurality of working condition data, the form of the hooking robot and the service life of the hooking robot, and comprise: collecting the past working images of the plurality of sub-working areas, determining a plurality of working condition features according to recognition of the past working images of the plurality of sub-working areas, determining a working condition area according to positions, feature forms of the plurality of working condition features and a moving path of the hooking robot relative to the sub-working area; determining a plurality of working condition data of the hooking robot according to detection of the working condition area, collecting the form of the hooking robot and the service life of the hooking robot at the same time, and constructing a corresponding multi-modal data according to the plurality of working condition data, the form of the hooking robot and the service life of the hooking robot; determining a multi-layer working condition combination based on detection of the multi-modal data, and determining a corresponding complex working condition type according to recognition of the multi-layer working condition combination. A hooking behavior module is configured to determine a plurality of hooking positions of the working area of the hooking robot based on the working area, the complex working condition type and the hooking robot, and determine hooking behaviors of the hooking robot at each hooking position according to current images corresponding to the plurality of hooking positions and a working mode of the hooking robot. An optimization module is configured to predict a collision area of the hooking robot based on each hooking behavior and the current image corresponding to the hooking position, determine a plurality of collision nodes according to recognition of the collision area, determine an optimized hooking behavior according to the plurality of collision nodes, the corresponding hooking behavior and the posture of the hooking robot, and eliminate the plurality of collision nodes. The collision avoidance processing system module is used for determining the collision avoidance processing system of the hooking robot according to the plurality of optimized hooking behaviors, the original hooking behavior and the hooking robot moving path, and gradually improving the hooking work efficiency of the hooking robot.

Images