A method and system for directional protection and dynamic speed-dependent obstacle avoidance of a cable crane hook

Abstract

The application relates to the technical field of intelligent hoisting equipment safety, and provides a cable crane hook directional protection and dynamic speed-following obstacle avoidance method and system, which comprises the following steps: receiving mechanism operation state data, a detection area enabling instruction and a dynamic safety distance issued by a ground host; generating and enabling a corresponding follow-up detection area in a follow-up coordinate system fixed to the hook; acquiring obstacle point cloud data in the enabled area, calculating a horizontal distance, a vertical distance and a shortest distance in combination with a hook circumscribed envelope; and reporting multi-dimensional distance data to the ground host to execute a hierarchical response strategy. The application realizes active perception of unknown obstacles in the near field of the hook and dynamic speed-following hierarchical protection, and solves the problems of a fixed no-entry area failing to adapt to a dynamic boundary and low matching precision of a safety distance.

Description

Technical Field

[0001] This application relates to the field of safety technology for intelligent lifting equipment, specifically to a method and system for directional protection and dynamic speed-following obstacle avoidance of a cable crane hook. Background Technology

[0002] Currently, in complex working conditions such as dam construction in water conservancy and hydropower projects, obstacle avoidance protection for cable cranes mainly adopts a combination of Global Positioning System (GPS) and fixed restricted zones. This scheme establishes static restricted zones by pre-mapping the dam structure, edges, and boundaries of structures, and compares the location of dynamic equipment equipped with GPS terminals with the restricted zones to trigger alarms or restrict movement. However, this obstacle avoidance method based on comparison of known target locations has significant shortcomings in practical applications: Firstly, its coverage is limited to objects with installed positioning terminals, and it cannot detect temporary equipment, personnel, material stockpiles, and sudden intrusions without terminals, resulting in blind spots. Secondly, the positions of the dam opening and edges change dynamically with the construction progress during dam construction, and fixing dynamic boundaries as static restricted zones can easily lead to delayed boundary updates, resulting in false alarms for old boundaries or missed alarms for new obstacles. In addition, the existing scheme has a coarse protection granularity, making it difficult to perform fine detection of the near-field three-dimensional space of the hook, and the alarm threshold is usually fixed, making it impossible to dynamically match the safety distance according to the operating speed and braking capacity of the mechanism, resulting in overly conservative protection at low speeds or insufficient protection at high speeds. Summary of the Invention

[0003] Purpose of the invention: The purpose of this invention is to solve the technical problems in the prior art and provide a method and system for directional protection and dynamic speed-based obstacle avoidance of cable crane hooks.

[0004] Technical solution:

[0005] In a first aspect, this application provides a method for directional protection and dynamic speed-following obstacle avoidance of a cable crane hook, applied to hook end equipment, including:

[0006] S100, Receive mechanism operation status data sent by the ground host, the mechanism operation status data includes the operation data of the traction mechanism and the operation data of the hoisting mechanism;

[0007] S200, receiving a detection area activation command and a dynamic safety distance generated by the ground host based on the mechanism's operating status data. The detection area activation command is used to indicate the activation of the traction follow-up detection area corresponding to the operating data of the traction mechanism and / or the hoisting follow-up detection area corresponding to the operating data of the hoisting mechanism. The dynamic safety distance is used to determine the size of the traction follow-up detection area and / or the hoisting follow-up detection area.

[0008] S300, according to the detection area activation command and the dynamic safety distance, generate and activate the corresponding follow-up detection area in the follow-up coordinate system, the follow-up coordinate system being fixed to the hook;

[0009] S400, acquire obstacle point cloud data located within the activated follow-up detection area, and calculate multi-dimensional distance data based on the obstacle point cloud data and the pre-recorded hook outer envelope. The multi-dimensional distance data includes horizontal distance, vertical distance and shortest distance.

[0010] S500, the multidimensional distance data is reported to the ground host so that the ground host can execute a graded response strategy based on the multidimensional distance data and the dynamic safety distance.

[0011] Preferably, before acquiring obstacle point cloud data located within the enabled follow-up detection area, the method further includes:

[0012] The point cloud data is acquired synchronously from multiple solid-state lidars, and the point cloud data is transformed to the servo coordinate system according to preset relative extrinsic parameters to obtain combined point cloud data; statistical outlier filtering is performed on the combined point cloud data to remove isolated points whose number of points in the neighborhood is less than a preset minimum neighbor threshold.

[0013] The point cloud data after statistical outlier filtering is subjected to wind speed-correlated dynamic background difference processing to obtain candidate obstacle point cloud data, including: for each point in the point cloud data after statistical outlier filtering, the corresponding spatial position is found in the background reference model and the relative displacement is calculated.

[0014] According to the formula Calculate the decision threshold;

[0015] Where dbg is the judgment threshold, d0 is the basic threshold, k is the wind speed adjustment coefficient, and w is the real-time wind speed;

[0016] When the relative displacement is greater than the determination threshold, the corresponding point is included in the candidate obstacle point cloud data; the candidate obstacle point cloud data is downsampled using voxels, the space is divided into voxels of a preset side length, and the average coordinate of the point falling in each non-empty voxel is used as the representative point to generate downsampled point cloud data.

[0017] Preferably, the step of acquiring obstacle point cloud data within the activated follow-up detection area, and calculating multidimensional distance data based on the obstacle point cloud data and the pre-recorded hook circumferential envelope, includes:

[0018] The obstacle point cloud data is obtained by extracting point clouds located within the activated servo detection area from the downsampled point cloud data; the obstacle point cloud data is then clustered to obtain obstacle point cloud clusters.

[0019] Based on the hook's outer envelope and the obstacle point cloud cluster, determine the global nearest obstacle point pair, and calculate the horizontal distance, the vertical distance, and the shortest distance based on the global nearest obstacle point pair.

[0020] Preferably, the step of calculating the horizontal distance, the vertical distance, and the shortest distance based on the global nearest obstacle pair includes:

[0021] The vector formed by the global nearest obstacle point pair is decomposed in the following coordinate system to obtain the component norm in the horizontal plane as the horizontal distance, the absolute value of the component in the vertical direction as the vertical distance, and the Euclidean distance of the global nearest obstacle point pair as the shortest distance.

[0022] Preferably, the step of generating and activating the corresponding follow-up detection area in the follow-up coordinate system according to the detection area activation command and the dynamic safety distance includes:

[0023] When the activation command of the detection area indicates the activation of the traction follow-up detection area, a forward-looking cuboid area is generated along the running direction of the traction mechanism with the origin of the follow-up coordinate system as the anchor point. The length of the forward-looking cuboid area is determined according to the dynamic safety distance.

[0024] When the activation command of the detection area indicates the activation of the lifting follow-up detection area, a downward-viewing cuboid area is generated along the running direction of the lifting mechanism with the origin of the follow-up coordinate system as the anchor point. The height of the downward-viewing cuboid area is determined according to the dynamic safety distance.

[0025] When the traction mechanism and the lifting mechanism are running simultaneously, the forward-looking cuboid region and the downward-looking cuboid region are activated at the same time, and the union of the forward-looking cuboid region and the downward-looking cuboid region is taken as the activated follow-up detection region.

[0026] Preferably, the operating data of the traction mechanism includes the operating speed of the traction mechanism, and the operating data of the hoisting mechanism includes the operating speed of the hoisting mechanism; the detection area activation command is generated by the ground host when it determines that the absolute value of the operating speed of the traction mechanism or the absolute value of the operating speed of the hoisting mechanism is greater than or equal to a preset speed threshold; the width and height of the forward-looking cuboid region are preset fixed values, and the length of the forward-looking cuboid region is equal to the dynamic safety distance corresponding to the traction mechanism; the horizontal cross-sectional side length of the downward-looking cuboid region is a preset fixed value, the height of the downward-looking cuboid region is equal to the dynamic safety distance corresponding to the hoisting mechanism, and the downward-looking cuboid region extends only in the vertically downward direction.

[0027] Preferably, acquiring obstacle point cloud data located within the activated follow-up detection area includes:

[0028] Point cloud data is collected synchronously by four solid-state lidars rigidly installed around the horizontal mounting beam in the middle of the hook. The four solid-state lidars are arranged circumferentially at 90-degree intervals in the horizontal direction. The optical axes of two solid-state lidars face the left and right banks to cover the traction three-dimensional corridor, and the optical axes of the other two solid-state lidars face the vertical downward direction to cover the lifting downward area.

[0029] Attitude data is acquired by an inertial measurement unit installed at the geometric center of the transverse mounting beam, with the inertial measurement unit serving as the origin of the servo coordinate system;

[0030] The four solid-state lidars are synchronized at the hardware level using a precise time protocol. Point cloud frames with timestamp differences falling within the allowable threshold are merged to obtain synchronized point cloud data.

[0031] Preferably, the method further includes:

[0032] Heartbeat frames are sent to the ground host at a preset period, and the heartbeat frames are sent together with the multidimensional distance data;

[0033] When the ground host loses the heartbeat frames consecutively for a preset number of times, it triggers the ground host to perform speed limiting protection and switch to the protection mode based on the ground-side three-dimensional environment model.

[0034] Preferably, the method further includes:

[0035] When the operating data of the traction mechanism and the operating data of the hoisting mechanism both meet the idle criterion and remain so for a preset time, the hook is controlled to drive multiple solid-state lidars to perform multi-station scanning in the construction area to obtain simplified point cloud data.

[0036] The simplified point cloud data is reported to the ground host, so that the ground host can combine the hook operation position information to construct the ground-side three-dimensional environment model;

[0037] In the protection mode, the ground host executes speed limiting constraints based on the ground-side three-dimensional environment model.

[0038] Secondly, this application provides a cable crane hook directional protection and dynamic speed-following obstacle avoidance system, including hook end equipment and ground host;

[0039] The hook end device is used for:

[0040] Receive the mechanism operation status data sent by the ground host, the mechanism operation status data including the operation data of the traction mechanism and the operation data of the hoisting mechanism;

[0041] The system receives a detection area activation command and a dynamic safety distance generated by the ground host based on the mechanism's operating status data. The detection area activation command is used to indicate the activation of the traction follow-up detection area corresponding to the operating data of the traction mechanism and / or the hoisting follow-up detection area corresponding to the operating data of the hoisting mechanism. The dynamic safety distance is used to determine the size of the traction follow-up detection area and / or the hoisting follow-up detection area.

[0042] According to the activation command of the detection area and the dynamic safety distance, a corresponding follow-up detection area is generated and activated in the follow-up coordinate system, which is fixed to the hook;

[0043] Obtain obstacle point cloud data within the activated follow-up detection area; calculate multidimensional distance data based on the obstacle point cloud data and the pre-recorded hook circumferential envelope; the multidimensional distance data includes horizontal distance, vertical distance, and shortest distance.

[0044] The multidimensional distance data is reported to the ground host;

[0045] The ground host is used for:

[0046] The dynamic safety distance is calculated based on the operational status data of the mechanism, and the activation command for the detection area is generated; a graded response strategy is executed based on the multi-dimensional distance data and the dynamic safety distance.

[0047] Beneficial effects:

[0048] 1. It achieves active three-dimensional perception without relying on pre-installed positioning terminals or complete loading of obstacles. By setting a follow-up orientation detection unit at the hook end, it directly detects unknown or temporary obstacles on the path, filling the blind spot of traditional comparison-based obstacle avoidance schemes in non-cooperative target perception and improving operational safety under complex working conditions.

[0049] 2. Achieving precise matching between the protection range and the dynamic characteristics of the mechanism's operation. By binding the size of the servo detection area to a dynamic safety distance calculated based on real-time speed, the detection area can adaptively expand and contract with the mechanism's operating speed and braking capability, avoiding the problems of low-speed hypersensitivity or high-speed lag caused by fixed thresholds, and improving the rationality of the obstacle avoidance strategy.

[0050] 3. Improved adaptability and operational continuity under dynamic boundary conditions. By adopting a strategy of independently activating the follow-up detection area according to the direction of mechanism movement, it can adapt to the dynamic changes of the pouring boundary without frequent re-measurement and resetting at the hopper or edge; at the same time, through the wind speed-correlated dynamic background difference mechanism, it effectively suppresses false detections caused by plateau winds and sand and cable sway, ensuring operational continuity.

[0051] 4. Enhanced system fault tolerance and reliability. By establishing rate-limiting protection under communication anomalies and a fallback mechanism based on a ground-side 3D environmental model, basic boundary constraints and obstacle warning capabilities can still be maintained even during short-term wireless link interruptions, forming a multi-layered security protection system. Attached Figure Description

[0052] Figure 1 This is a schematic diagram illustrating the direction of cable car movement and the definition of the axis system according to an embodiment of this application;

[0053] Figure 2 This is a top-view layout and field-of-view orientation diagram of the radar assembly according to an embodiment of this application;

[0054] Figure 3 This is a schematic diagram of the hook assembly and radar installation structure according to an embodiment of this application;

[0055] Figure 4 This is a schematic diagram of the radar coverage area according to an embodiment of this application;

[0056] Figure 5 This is a schematic diagram of the follower cuboid detection area of ​​the traction mechanism in an embodiment of this application;

[0057] Figure 6 This is a schematic diagram of the follow-up cuboid detection area of ​​the lifting mechanism according to an embodiment of this application;

[0058] Figure 7 This is a schematic diagram of dynamic safety distance and graded alarm in an embodiment of this application;

[0059] Figure 8 This is a flowchart illustrating the independent control of the traction and lifting detection area according to an embodiment of this application.

[0060] Figure 9 This is a schematic diagram of the system composition and wireless interaction according to an embodiment of this application;

[0061] Figure 10 This is a flowchart illustrating the three-dimensional environment model and communication anomaly switching process in an embodiment of this application;

[0062] Figure 11 This is a schematic diagram of the three modes of the main software flow in an embodiment of this application;

[0063] Figure 12 This is a flowchart of a method according to an embodiment of this application. Detailed Implementation

[0064] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art.

[0066] Example 1

[0067] like Figure 1 , Figure 9 , Figure 12 As shown, this embodiment provides a method for directional protection and dynamic speed-based obstacle avoidance of a cable crane hook. This method is applied to the hook-end equipment and achieves proactive perception and safety warning of the hook's near-field environment through an end-cloud collaborative architecture. In this embodiment, the hook-end equipment, as the perception execution node, is mainly responsible for environmental data collection, processing, and distance calculation, while the decision-making and dissemination of safety strategies are completed by the ground host. This division of labor ensures the real-time nature of perception and the globality of decision-making.

[0068] S100: Receives operational status data from the ground host, including operational data for the traction mechanism and the hoisting mechanism. Specifically, the hook-end device passively receives real-time status information from the ground host via a wireless communication link. The traction mechanism's operational data typically includes traction speed and direction of travel, while the hoisting mechanism's operational data includes hoisting speed and direction of movement. This data is directly obtained from the cable crane control system by the ground host, ensuring higher authority and accuracy. The hook-end device itself does not directly collect encoder data from the cable crane mechanism but relies on data from the ground host. This simplifies the hardware interface design of the hook end and avoids misjudgments caused by sensor installation errors or drift, ensuring that subsequent obstacle avoidance logic is based on a unified system-level status benchmark.

[0069] S200: The system receives a detection area activation command and a dynamic safety distance generated by the ground host based on the mechanism's operating status data. The detection area activation command instructs the activation of the traction follow-up detection area corresponding to the traction mechanism's operating data and / or the hoisting follow-up detection area corresponding to the hoisting mechanism's operating data. The dynamic safety distance determines the size of the traction follow-up detection area and / or the hoisting follow-up detection area. Specifically, the ground host calculates the appropriate safety protection range based on the current operating speed and braking performance, and converts it into specific control commands, which are then sent to the hook end. The detection area activation command is essentially a function switch, informing the hook end which direction of potential risk should be monitored. For example, when only traction operations are performed, only the traction direction detection area needs to be activated; when performing combined operations, areas in multiple directions may be activated simultaneously. The dynamic safety distance is a length-dimensional parameter that directly determines the extension depth of the detection area in the direction of movement. In this way, the detection range of the hook end is no longer a fixed static envelope, but an elastic protective cover that can be dynamically adjusted with the operation of the cable crane, avoiding oversensitivity at low speeds and insufficient protection at high speeds.

[0070] S300, based on the detection area activation command and the dynamic safety distance, a corresponding follow-up detection area is generated and activated in the follow-up coordinate system, the follow-up coordinate system being fixed to the hook; combined with Figure 1 As shown, the cable crane mainly consists of a left bank tower, a right bank tower, a load-bearing cable erected between the two towers, a traction trolley that travels back and forth along the load-bearing cable, and a hook suspended below the traction trolley. The traction trolley moves horizontally between the left and right banks along the load-bearing cable to achieve traction motion, while the hook achieves vertical lifting motion via a lifting wire rope. The entire cable crane can also travel upstream and downstream. Based on the above structure, this application defines three axes of the follow-up coordinate system: the X-axis is parallel to the left and right bank directions (i.e., the traction direction), the Y-axis is parallel to the upstream and downstream directions (i.e., the travel direction), and the Z-axis is vertically upward (i.e., the lifting direction). Since this coordinate system is fixed to the hook, it moves synchronously with the translation and attitude changes of the hook. This means that no matter where the hook is located on the dam pouring surface, the generated follow-up detection area always maintains a fixed orientation relative to the hook body. For example, the traction follow-up detection area always points forward of the traction motion, and the lifting follow-up detection area always covers the space directly below the hook. This homing characteristic is the core foundation for achieving "directional protection," enabling sensing resources to accurately focus on risk areas under the current movement trend, without having to perform indiscriminate scanning and processing of the entire space. This significantly reduces the computational load and improves the response speed to obstacles in key directions.

[0071] S400: Obtain obstacle point cloud data within the activated follow-up detection area. Based on the obstacle point cloud data and the pre-recorded hook circumferential envelope, calculate multi-dimensional distance data, including horizontal distance, vertical distance, and shortest distance. Specifically, the hook-end device first extracts the valid point cloud data falling within the currently activated follow-up detection area from the original point cloud, removing irrelevant background interference. Subsequently, spatially compare these environmental point clouds with the pre-stored hook circumferential envelope. Here, the hook circumferential envelope refers to the maximum geometric contour including the hook body and any potentially mounted canisters, serving as the reference boundary for the "local entity," rather than simply treating the hook as a point mass. The multi-dimensional distances calculated based on this have clear physical orientation: the horizontal distance reflects the proximity of obstacles on the horizontal plane, primarily used to assess traction collision risk; the vertical distance reflects the distance between obstacles in the vertical direction, primarily used to assess lifting compression or impact risk; and the shortest distance is the Euclidean distance between the two, representing the most direct physical contact risk. These three distance components are derived from the same source and together constitute a three-dimensional description of near-field risk, providing sufficient data support for subsequent hierarchical decision-making.

[0072] In step S500, the multi-dimensional distance data is reported to the ground host, enabling the ground host to execute a tiered response strategy based on the multi-dimensional distance data and the dynamic safety distance. Specifically, after completing the distance calculation, the hook-end device transmits data packets containing horizontal distance, vertical distance, and shortest distance back to the ground host in real time via a wireless link. The hook end itself does not perform safety control actions such as deceleration or stopping; instead, it submits the "perception results" to the ground host, which has global control authority, for "decision judgment." After receiving the multi-dimensional distance data, the ground host compares it with the dynamic safety distance calculated in step S200. Figure 7As shown, the dynamic safety distance adopts a two-level nested threshold structure: the outer dashed box represents the warning zone D_warn, and the inner solid box represents the braking zone D_brake, with D_warn > D_brake. The warning zone D_warn corresponds to the trigger boundary of audible and visual alarms and speed-limited operation, while the braking zone D_brake corresponds to the trigger boundary of emergency braking. The lengths of both threshold levels are dynamically adjusted with the operating speed of the mechanism: when the operating speed increases, the braking distance increases, and D_warn and D_brake expand synchronously; when the operating speed decreases, both threshold levels contract synchronously. This speed-dependent dynamic adjustment mechanism ensures that the safety protection distance always matches the current braking capacity under different speed conditions. When the multi-dimensional distance data falls into the warning zone, an audible and visual alarm is triggered and speed-limited operation is initiated; when the multi-dimensional distance data falls into the braking zone, emergency braking is triggered. This hierarchical response strategy, with its "sensing on the hook and decision-making on the ground" architecture, fully utilizes the abundant computing resources and complete control interfaces on the ground side, while ensuring the lightweight and real-time nature of the hook-end sensing, forming a complete safety closed loop. It should be understood that although this embodiment describes a specific data flow, in other embodiments, as long as the functions of hook-end sensing and ground-end decision-making are decoupled and coordinated, they all fall within the protection scope of this application.

[0073] Example 2

[0074] This embodiment, based on Embodiment 1, further details the specific implementation process of acquiring obstacle point cloud data and calculating multidimensional distance data in step S400. In this embodiment, to improve the accuracy and robustness of the sensing data, a preprocessing procedure for the original point cloud is included before acquiring obstacle point cloud data located within the activated servo detection area. Specifically, this preprocessing procedure is strictly executed in the order of statistical outlier filtering, wind speed-correlated dynamic background subtraction, and voxel downsampling. This specific timing is crucial to ensuring the reliability of subsequent distance calculations.

[0075] First, point cloud data from multiple solid-state lidar sensors are acquired synchronously. The point cloud data is then transformed to a moving coordinate system based on preset relative extrinsic parameters, resulting in combined point cloud data. This step achieves spatiotemporal unification of multi-sensor data, enabling point clouds from different installation locations to be fused within the same reference system with the hook as the origin. Subsequently, statistical outlier filtering is applied to the combined point cloud data to remove isolated points whose neighboring points are less than a preset minimum neighbor threshold. Specifically, for each point in the combined point cloud, the number of points N in the neighborhood is counted with that point as the center and a preset neighborhood radius R as the range. In a preferred example of this application, the neighborhood radius R is set to 0.1m. If the number of neighboring points N for a point is less than the preset minimum neighbor threshold N_min, the point is determined to be a floating noise point or random interference and is removed. This step effectively removes discrete false points caused by electromagnetic interference, rain / fog scattering, or sensor noise, preventing these isolated noise points from being misidentified as obstacles in subsequent processing.

[0076] After denoising, the point cloud data after statistical outlier filtering is subjected to wind speed-correlated dynamic background difference processing to obtain candidate obstacle point cloud data. This is a key technical means for addressing complex working conditions in plateau and mountainous areas. The specific processing includes: for each point in the point cloud data after statistical outlier filtering, finding its spatial corresponding position in the background reference model and calculating its relative displacement. Here, the background reference model is a local environment model stored locally on the hook-end equipment, used to characterize the steady-state background structure in the near field of the hook, independent of the global three-dimensional environment model on the ground side. Then, according to the formula... Calculate the decision threshold, where, To determine the threshold, Based on the threshold, This is the wind speed adjustment coefficient. This refers to real-time wind speed. The decision is made when the calculated relative displacement exceeds a threshold value. When the relative displacement is less than or equal to the threshold, the corresponding point is included in the candidate obstacle point cloud data; otherwise, if the relative displacement is less than or equal to the threshold, the point is considered to be a background disturbance or wind and dust and is suppressed.

[0077] Introducing wind speed variable Dynamically adjusting the judgment threshold has clear physical significance and technical benefits. In high-altitude scenarios such as dam construction, when wind speeds are high, increasing the wind speed adjustment coefficient can be effective. With real-time wind speed The product term automatically raises the judgment threshold. This effectively avoids false alarms caused by slight cable swaying or sandstorms; when the wind speed is low, the judgment threshold returns to the basic threshold. Nearby, maintain high sensitivity to detect real obstacles, thus balancing safety and operational efficiency.

[0078] After obtaining the candidate obstacle point cloud data, voxel downsampling is performed, dividing the space into voxels of preset side lengths. The average coordinates of the points falling within each non-empty voxel are used as representative points to generate downsampled point cloud data. In a preferred example of this application, the voxel side length... The value is set to 0.02m. It is crucial to emphasize that voxel downsampling must be performed after wind speed-related dynamic background differencing. If downsampling is performed before denoising, the coordinates of isolated noise points may be averaged into the representative points of the voxels, causing noise points that should have been removed to "contaminate" the valid data, thus affecting the accuracy of subsequent distance calculations. By performing downsampling after differencing, the purity of the input data is ensured, and the computational load of subsequent clustering and distance calculations is significantly reduced by homogenizing the point cloud density, achieving a balance between real-time performance and accuracy.

[0079] Based on the downsampled point cloud data obtained through the above preprocessing, this embodiment further refines the logic for acquiring obstacle point cloud data and calculating multidimensional distances in step S400. Specifically, obstacle point cloud data located within the activated servo detection area is acquired. Based on the obstacle point cloud data and the pre-recorded hook circumscribed envelope, multidimensional distance data is calculated, including: extracting point clouds located within the activated servo detection area from the downsampled point cloud data to obtain obstacle point cloud data; clustering the obstacle point cloud data to obtain obstacle point cloud clusters; determining the globally nearest obstacle point pair based on the hook circumscribed envelope and obstacle point cloud clusters, and calculating the horizontal distance, vertical distance, and shortest distance based on the globally nearest obstacle point pair.

[0080] like Figure 4As shown, the spatial relationship between the radar coverage area and the servo detection area is illustrated under two typical operating conditions. The left-hand condition corresponds to the traction motion state, where radars 2 and 4 form a symmetrical wide fan-shaped detection area along the X-axis, covering the traction-oriented three-dimensional corridor. The right-hand condition corresponds to the lifting motion state, where radars 1 and 3 form a longitudinal detection area along the Y-axis, covering the area below the lifting direction. Although the physical field of view of the radar may cover a wide range (e.g., +40° to -80°), this application focuses only on the point cloud within the activated servo detection area (the cuboid area shown in the figure). Through geometric truncation, the processing range is strictly limited to the risk space related to the current motion direction, eliminating interference from irrelevant areas. Euclidean clustering is performed on the truncated point cloud, grouping spatially adjacent points into the same obstacle point cloud cluster. For example, a clustering distance threshold of 0.1m can be set, meaning that two points with a distance less than this value are considered the same object. At the same time, clusters containing too few points (e.g., less than 5 points) are removed to further filter out residual minor interference. Based on this, the system iterates through the set of points circumscribed in the hook's envelope and all obstacle point clouds to find the pair of points with the smallest Euclidean distance, defining this as the globally nearest obstacle pair. This pair represents the most dangerous contact position between the hook and environmental obstacles at the current moment.

[0081] After determining the global nearest obstacle pair, the specific methods for calculating the horizontal distance, vertical distance, and shortest distance based on the pair include: decomposing the vector formed by the global nearest obstacle pair in the moving coordinate system, obtaining the component norm in the horizontal plane as the horizontal distance, obtaining the absolute value of the component in the vertical direction as the vertical distance, and using the Euclidean distance of the global nearest obstacle pair as the shortest distance.

[0082] This multi-dimensional distance decomposition calculation method directly serves the ground host's hierarchical response strategy. Horizontal distance primarily reflects the proximity of the hook to obstacles in the traction direction and is the primary criterion for obstacle avoidance decisions by the traction mechanism; vertical distance primarily reflects the distance between the hook and obstacles below in the lifting direction and is the primary criterion for obstacle avoidance decisions by the lifting mechanism; while the shortest distance serves as a fallback indicator for safety verification under complex attitudes such as oblique approach. By decoupling a single spatial proximity into components with clear motion semantics, this application enables the ground host to implement differentiated and precise control based on the motion characteristics of different mechanisms, avoiding the problem of poor adaptability of traditional single distance thresholds under different working conditions.

[0083] Example 3

[0084] This embodiment, based on Embodiment 1, provides a more detailed explanation of the generation rules and dynamic activation mechanism of the follow-up detection area in step S300. According to the detection area activation command and the dynamic safety distance, the corresponding follow-up detection area is generated and activated in the follow-up coordinate system, including: when the detection area activation command indicates the activation of the traction follow-up detection area, a forward-looking cuboid area is generated along the running direction of the traction mechanism, with the origin of the follow-up coordinate system as the anchor point, and the length of the forward-looking cuboid area is determined according to the dynamic safety distance; when the detection area activation command indicates the activation of the hoisting follow-up detection area, a downward-looking cuboid area is generated along the running direction of the hoisting mechanism, with the origin of the follow-up coordinate system as the anchor point, and the height of the downward-looking cuboid area is determined according to the dynamic safety distance; when the traction mechanism and the hoisting mechanism are running simultaneously, both the forward-looking cuboid area and the downward-looking cuboid area are activated simultaneously, and the union of the forward-looking cuboid area and the downward-looking cuboid area is taken as the activated follow-up detection area.

[0085] Specifically, the servo detection region is not a fixed, static envelope, but a geometric space that dynamically adjusts with the operation of the mechanism. For example... Figure 5 and Figure 6 As shown, the generation of this region strictly relies on the origin of the servo coordinate system, namely the inertial measurement unit (IMU) installed at the geometric center of the hook beam 6 of the hook assembly. This generation rule, anchored by the IMU, ensures that the detection area always maintains a fixed relative position with the hook body. Regardless of the hook's attitude or position, the detection area can accurately cover the motion risk zone relative to the hook. For different motion mechanisms, the detection area exhibits significant directional differences. For the traction mechanism, the generated forward-looking cuboid region extends forward along the traction direction (X-axis), its core purpose being to detect potential obstacles on the hook's translation path in advance; therefore, its length dimension is directly tied to the dynamic safety distance reflecting braking capability. For the hoisting mechanism, the generated downward-looking cuboid region extends downward along the vertical direction (negative Z-axis), focusing on monitoring the risk of compression or impact below the hook; its height dimension is also determined by the dynamic safety distance. This directional differentiation design allows the sensing resources to accurately focus on the main risk direction under the current motion trend, avoiding ineffective scanning of irrelevant spaces.

[0086] Furthermore, the operating data of the traction mechanism includes the operating speed of the traction mechanism, and the operating data of the hoisting mechanism includes the operating speed of the hoisting mechanism;

[0087] The activation command for the detection zone is generated by the ground host when it determines that the absolute value of the traction mechanism's operating speed or the hoisting mechanism's operating speed is greater than or equal to a preset speed threshold. In a preferred example of this application, the preset speed threshold is set to 0.05 m / s. This speed criterion constitutes the logical threshold for activating the servo detection zone. When the mechanism's operating speed is below this threshold, the system determines that the hook is in a quasi-stationary or slightly moving state. At this time, the risk of collision is extremely low, or the existing low-speed protection measures are sufficient. Therefore, the active detection zone is not activated, which saves computing resources and avoids false alarms caused by sensor noise or slight shaking. Only when the speed exceeds this threshold, indicating that the mechanism has entered a substantial movement phase, does the system activate the corresponding servo detection zone and begin executing the active obstacle avoidance logic.

[0088] Regarding the specific dimensions of the follow-up detection area, the width and height of the forward-looking cuboid area are preset fixed values, and the length of the forward-looking cuboid area is equal to the dynamic safety distance corresponding to the traction mechanism; the horizontal cross-sectional side length of the downward-looking cuboid area is preset fixed values, the height of the downward-looking cuboid area is equal to the dynamic safety distance corresponding to the hoisting mechanism, and the downward-looking cuboid area extends only in the vertically downward direction. For example... Figure 5 As shown, in a specific implementation, the width (Y direction) of the traction-guided detection area (forward-looking cuboid) can be set to 30m, the height (Z direction) can be set to 50m, and its length (X direction) is equal in real time to the dynamic safety distance D_warn_trac on the traction side issued by the ground host. Figure 6 As shown, the horizontal cross-section (XY plane) of the hoisting follow-up detection area (downward-view cuboid) can be set as a 30m × 30m square, and its height (negative Z-direction) is equal to the dynamic safety distance D_warn_hoist on the hoisting side in real time. Setting some side lengths to fixed values ​​is based on engineering considerations of the cable crane structure and operating environment: for example, a width of 30m is sufficient to cover the typical range of dam pouring opening and adjacent operations on both sides, while a height of 50m takes into account the maximum possible swing amplitude of the hook and the sensing redundancy in the vertical direction. Binding the side length in the direction of movement to the dynamic safety distance achieves adaptive matching between the detection range and braking capacity. When the operating speed increases, the dynamic safety distance increases accordingly, and the detection area automatically extends in the direction of movement, thereby providing an earlier warning time for high-speed movement; conversely, when the speed decreases, the detection area automatically shrinks, reducing the detection of irrelevant targets at a distance and reducing the false alarm rate. This "speed-dependent expansion and contraction" mechanism fundamentally solves the problem of poor adaptability of traditional fixed threshold schemes under variable speed conditions.

[0089] like Figure 8As shown, the activation control process of the detection area is as follows: First, the motion states of the traction speed Vtrac and the lifting speed Vhoist are acquired, followed by the speed state judgment stage. The system checks three mutually exclusive conditions in sequence: if only Vtrac is greater than zero, the traction direction cube detection area ROI is enabled and the lifting ROI is disabled; if only Vhoist is greater than zero, the lifting direction ROI is enabled and the traction ROI is disabled; if both Vtrac and Vhoist are greater than zero, the ROIs in the two directions are enabled independently and in parallel. When the traction mechanism and the lifting mechanism are running simultaneously, the system does not simply switch or select one area, but simultaneously enables the forward-looking cuboid area and the downward-looking cuboid area, and uses the union of the two as the final effective detection range. This union interception strategy is the key advantage of this application that distinguishes it from the traditional omnidirectional symmetric envelope scheme. While using an omnidirectional spherical or cubic envelope covers all directions, it also includes a large number of "safe zones" where collisions are impossible in the current motion state (such as the rear area during forward traction and the upper area during ascent and descent). This not only wastes valuable point cloud processing computing power but also increases the probability of misclassifying irrelevant backgrounds as obstacles. By independently enabling and combining the union of data, this application constructs an asymmetric composite detection space that closely follows the actual motion trend. While ensuring full coverage of key risk directions, it eliminates redundant data to the greatest extent, improving the system's real-time performance and robustness.

[0090] Example 4

[0091] This embodiment, based on embodiments 1 to 3, further details the hardware implementation for acquiring obstacle point cloud data within the activated servo detection area. To achieve high-precision stereo perception in the near field of the hook and acquire obstacle point cloud data within the activated servo detection area, the method includes: synchronously acquiring point cloud data using four solid-state lidars rigidly mounted around the hook beam 6 in the middle of the hook assembly. The four solid-state lidars are arranged circumferentially at 90-degree intervals in the horizontal direction. Two of the solid-state lidars have their optical axes oriented towards the left and right banks to cover the traction-oriented three-dimensional corridor, while the optical axes of the other two are oriented approximately vertically downwards to cover the lifting-down area. Specifically, as... Figure 3 , Figure 2 and Figure 1 As shown, the hook assembly includes a top hook pulley 5, a middle hook beam 6, and a bottom hook 7. The middle section of the hook assembly features a hook beam 6, which not only provides structural support but also serves as a reference platform for sensor integration. Figure 2As shown, four solid-state lidar units are arranged in a cross shape around the IMU, numbered sequentially in a clockwise direction as Radar 1, Radar 2, Radar 3, and Radar 4. The X-axis is parallel to the left and right bank directions (traction direction), the Y-axis is parallel to the upstream and downstream directions (travel direction), and the Z-axis is vertically upward (lifting direction). Figure 4 As shown, each radar has a vertical field of view of -18° to 7° and a horizontal field of view of 120° (+40° to 80°). Radars numbered 2 and 4 form a wider fan-shaped detection area in the X-axis direction to cover the traction-oriented three-dimensional corridor, while radars numbered 1 and 3 form a longitudinal detection area in the Y-axis direction to cover the area below the lift.

[0092] Four solid-state lidar units are rigidly fixed to the perimeter of the hook beam 6 via radar brackets 1#, 2#, 3#, and 4# (marked as 1, 2, 3, and 4). This rigid connection ensures the long-term stability of the relative position between the radar and the hook body, preventing extrinsic parameter drift caused by operational vibrations. In terms of layout, the four radars are evenly distributed at 90-degree intervals in the horizontal plane, forming a 360-degree circumferential sensing capability. However, this circumferential arrangement is not a simple omnidirectional symmetrical scan, but rather a directional division of labor based on the operational characteristics of the cable crane. Radars numbered 2 and 4 have their optical axes facing the right and left banks respectively (i.e., ±X directions), specifically responsible for covering the three-dimensional corridor in the direction of the traction mechanism's operation, with their field of view configuration focusing on long-range detection in the horizontal direction. Radars numbered 1 and 3 have their optical axes facing an approximately vertically downward direction (i.e., -Z direction), located on the upstream and downstream sides respectively, specifically responsible for covering the lower area in the direction of the hoisting mechanism's operation, with their field of view configuration focusing on short-range blind spot monitoring in the vertical direction. This directional layout strategy based on the motion axis enables sensor resources to be precisely matched to the protection requirements of the two core mechanisms of cable crane traction and hoisting. Compared with the traditional omnidirectional uniform layout, it achieves higher point cloud density and detection reliability in key risk directions.

[0093] In the aforementioned hardware layout, establishing a spatial reference base is crucial. Attitude data is acquired through an inertial measurement unit (IMU) installed at the geometric center of the hook beam 6, with the IMU serving as the origin of the servo coordinate system. Specifically, the IMU is precisely installed at the geometric center of the hook beam 6, which is also the symmetry center of the spatial distribution of the four radars. Setting the IMU as the origin of the servo coordinate system has dual technical significance: firstly, the IMU can sense the hook's attitude changes (pitch, roll, yaw) in three-dimensional space in real time, providing a dynamic reference for point cloud attitude compensation; secondly, establishing a coordinate system with its geometric center as the origin makes the relative position vectors of the four radars more numerically regular, which simplifies subsequent coordinate transformation calculations. During operation, regardless of the hook's motion state or attitude, all sensor data will be uniformly converted to this servo coordinate system with the IMU as the origin for processing, thus ensuring the consistency between the sensed data and the hook's own motion.

[0094] Since multiple radars work collaboratively, consistency in the temporal dimension is a prerequisite for spatial fusion. Hardware-level time synchronization of the four solid-state lidars is achieved through a Precision Time Protocol (PTP), merging point cloud frames whose timestamp differences fall within an allowable threshold to obtain synchronized point cloud data. Specifically, this application employs a Precision Time Protocol (PTP) rather than software-level time alignment. This is because the cable crane hook is in continuous motion during operation; if there are significant differences in the acquisition times of each radar, the same obstacle will appear in different spatial locations in the point clouds of different radars, resulting in ghosting, misalignment, or false gaps in the fused point cloud, severely affecting the accuracy of obstacle detection. The PTP protocol, through a hardware stamping mechanism, can control the time synchronization accuracy of the four radars to the microsecond level, far superior to the radar frame period. During the data fusion stage, the system only merges point cloud frames whose timestamp differences are less than a preset allowable threshold. For example, this allowable threshold can be set to half the radar frame period to ensure that the data in each frame participating in the fusion reflects a snapshot of the environment at the same instant. This strict temporal constraint mechanism fundamentally eliminates the motion distortion problem caused by asynchronous sensor acquisition, providing a high-quality data source with spatiotemporal consistency for subsequent wind speed correlation background difference and multidimensional distance calculation.

[0095] After completing the hardware synchronization and data acquisition, it is also necessary to unify the local data of each radar into the servo coordinate system. Although this embodiment does not describe the calibration process itself in detail, during operation, the system will use pre-calibrated and stored relative extrinsic parameters to transform the point cloud of each radar in real time. Specifically, for each frame of the original point cloud acquired, the system will perform a real-time transformation based on a preset rotation matrix. Translation vector , will the Points in the radar coordinate system Points transformed to the following coordinate system The transformation formula is: This transformation process is performed in real time during each frame of data processing, making the four physically independent radars logically equivalent to a single servo detection unit.

[0096] Example 5

[0097] This embodiment, based on embodiments 1 to 4, further details the fault-tolerant protection mechanism of the system under abnormal wireless communication conditions and the construction and application of the ground-side three-dimensional environment model. As a supplementary safety mechanism to the method flow, this application establishes a dual guarantee of heartbeat monitoring and model fallback to ensure that even if the wireless link between the hook end and the ground host experiences a short-term interruption or severe interference, the cable crane can still maintain basic safety protection capabilities, preventing the risk of loss of control due to missing sensing data.

[0098] Heartbeat frames are sent to the ground host at preset intervals, along with multidimensional distance data. Specifically, when performing normal obstacle avoidance sensing and reporting tasks, the hook-end device includes a heartbeat identifier in each data transmission cycle. In a preferred example of this application, the heartbeat frame transmission cycle is set to 50ms, which is consistent with the reporting frequency of multidimensional distance data, ensuring real-time status monitoring while avoiding unnecessary consumption of valuable wireless bandwidth resources. The heartbeat frame not only confirms link connectivity but also implicitly contains the hook-end device's own health status information.

[0099] When the ground host loses heartbeat frames consecutively for a preset number of times, it triggers rate limiting protection and switches to a protection mode based on the ground-side 3D environment model. Specifically, the ground host maintains a heartbeat counter. The counter is reset to zero each time a valid heartbeat frame is received, and incremented by one if no heartbeat is received within a transmission cycle. In a preferred example of this application, the preset number of times is set to 3, meaning that if no heartbeat is received for 150ms (3 × 50ms) consecutively, the ground host determines that the current wireless communication is in an abnormal state. This threshold design balances anti-interference capability and safety response speed: if the threshold is too small (e.g., 1 time), it is easy to misjudge due to instantaneous jitter or packet loss in the wireless channel, causing unnecessary downtime; if the threshold is too large (e.g., 10 times), it will lead to excessively long fault detection delays, increasing safety risks. Once a communication anomaly is detected, the ground control unit immediately takes over safety control. First, it implements speed-limiting protection for the relevant mechanisms, restricting the operating speed of the traction and hoisting mechanisms to a preset safe low-speed range (e.g., 10% of the rated speed or an absolute value of 0.1 m / s), fundamentally reducing the kinetic energy of a potential collision. Simultaneously, the system automatically switches from a "normal mode" relying on real-time perception to a "protection mode" relying on pre-stored maps. Figure 10As shown, the complete process of 3D environment model construction and communication anomaly switching is as follows: First, during idle periods, the hook-carrying lidar scans multiple locations, stitching together a local 3D environment model; then, it enters normal mode, performing real-time distance detection and continuously updating the 3D environment model; the system continuously judges the communication status, and if communication is normal, it maintains the current mode; if a communication anomaly is detected (such as consecutive loss of heartbeat frames reaching a preset number), it switches to 3D model protection mode and applies speed limits to relevant mechanisms; in protection mode, it cyclically checks whether communication has recovered, and once communication is restored, it switches back to real-time detection mode and uses the latest data to correct and update the 3D environment model. This mode switching is completed on the ground host side, and the hook-end equipment does not need to sense changes in communication status, thus simplifying the end-side logic and improving the robustness of the system.

[0100] To support the effective operation of the aforementioned protection mode, this application also designs a ground-side 3D environment model construction mechanism based on idle periods. When the operating data of the traction mechanism and the hoisting mechanism both meet the idle criterion and remain so for a preset time, the control hook drives multiple solid-state lidars to perform multi-station scanning in the construction area to acquire simplified point cloud data. Specifically, the idle criterion refers to the mechanism being in a quasi-stationary state, defined in this embodiment as the absolute values ​​of both the traction speed and the hoisting speed being less than 0.05 m / s. The preset time T can be configured to 30 seconds or other values ​​suitable for the rhythm of on-site operations. The purpose of setting this criterion is to automatically collect environmental data during work breaks (such as waiting for unloading, tank changing, and other non-production periods) without interfering with normal, efficient operations. After the idle condition is met, the hook-end equipment automatically starts the scanning program, controlling the radar to cover the construction area within the current visible range in a multi-station scanning manner. The "simplified point cloud data" acquired here is lightweight data after downsampling or feature extraction, and its data volume is much smaller than the full point cloud used for real-time obstacle avoidance, aiming to adapt to the needs of limited communication bandwidth or storage efficiency.

[0101] The simplified point cloud data is reported to the ground host, which then combines it with the hook's operating position information to construct a ground-side 3D environment model. After receiving multiple frames of simplified point cloud data, the ground host uses the cable crane's own encoder or global position information provided by GNSS to transform each frame of point cloud data to a geodetic coordinate system or dam coordinate system for stitching and fusion. This process gradually constructs and updates a 3D environment model reflecting the actual geometric structure of the construction site (usually stored in voxel map format). It is important to emphasize that the "ground-side 3D environment model" in this embodiment and the "background reference model" mentioned in Embodiment 2 are two completely independent concepts. The background reference model is stored locally on the hook-end equipment and dynamically updated with the hook's movement. It is only used for wind speed-related background difference analysis to suppress false detections of wind and sand and does not contain global geographic information. The ground-side 3D environment model, on the other hand, is stored on the ground host, constructed based on a global coordinate system, and represents the macroscopic geometric structure of the static or semi-static environment, such as the dam's pouring surface, edges, and structures. It is specifically used to provide spatial boundary constraints during communication interruptions. The two have no overlap in storage location, update timing, data content, or purpose. This dual physical and logical isolation ensures the purity and reliability of their respective functions.

[0102] In protection mode, the ground control unit executes speed-limiting constraints based on a ground-side 3D environmental model. Specifically, when a communication anomaly triggers protection mode, since the real-time multi-dimensional distance data reported by the hook cannot be obtained, the ground control unit uses a pre-built ground-side 3D environmental model as an alternative sensing source. Based on the hook's current position fed back by the cable crane control system (this position information typically comes from the ground-side encoder or positioning system and does not rely on wireless reporting from the hook), the ground control unit queries the 3D environmental model for the distribution of obstacles and boundary information within a certain range around the hook. If the hook is detected approaching an edge, hopper edge, or other known obstacle recorded in the model, the ground control unit will further impose constraints on top of the existing speed limit, and may even trigger emergency braking when approaching a dangerous boundary. Although this protection based on a pre-stored model has lower accuracy than real-time sensing, it provides a last line of defense in extreme situations where communication is interrupted, preventing the hook from blindly entering unknown or dangerous areas.

[0103] Furthermore, the fault tolerance mechanism of this application also includes self-healing logic after communication is restored. When the ground host receives a valid continuous heartbeat frame again, it determines that communication has returned to normal, and then exits the protection mode, resuming the hierarchical response strategy based on real-time multidimensional distance data. At the same time, the latest real-time point cloud data acquired during the communication restoration period will also be used to correct and update the ground-side 3D environment model, eliminating model lag caused by environmental changes (such as adding material or adjusting templates), and ensuring the timeliness of map data when entering the protection mode again.

[0104] This complete state machine closed loop of "normal → abnormal → recovery → correction" enables the system to have comprehensive adaptability to cope with complex electromagnetic environments and dynamic construction scenarios.

[0105] like Figure 11 As shown, from the perspective of the overall software operation logic, the collaborative work between the hook-end equipment and the ground host can be divided into three operating states: operation mode, idle mode, and communication failure mode. In operation mode, the software sequentially executes radar scanning to acquire point clouds, point cloud preprocessing (filtering, noise reduction, etc.), target distance calculation to complete obstacle detection, and sends the distance results to the PLC, forming the main process of real-time obstacle avoidance. In idle mode, the software controls the hook movement to drive radar rotation scanning, then performs multi-radar point cloud fusion to construct a 3D point cloud, ultimately completing 3D modeling and generating a 3D environmental model. In communication failure mode, the software sequentially executes communication failure detection, enables speed-limited operation, performs protection control based on the 3D model, and performs communication recovery detection until normal operation is restored. These three modes are interconnected through operating state criteria, together forming the complete main software process.

[0106] Example 6

[0107] like Figure 9 As shown, this embodiment provides a directional protection and dynamic speed-following obstacle avoidance system for cable crane hooks. This system constructs an end-to-cloud collaborative security architecture through clear physical boundary division and functional allocation. The system includes hook-end devices and a ground host. The hook-end devices include four lidars (1 to 4), which collect point cloud data and aggregate it to an embedded IPC for local processing. The hook-end devices and the ground host PLC are connected via a dual-band 2.4GHz and 5GHz wireless link. The downlink channel carries low-frequency, high-reliability data such as lifting speed, traction speed, wind speed, and status commands, while the uplink channel transmits high-frequency, real-time data such as horizontal distance, vertical distance, shortest distance, heartbeat indicators, fault codes, and point cloud data. Unlike the aforementioned method embodiments that focus on algorithmic flow, this embodiment focuses on describing the physical devices implementing the above method and their interaction relationships, providing a basis for product-level implementation and protection.

[0108] The hook-end equipment is used to receive the mechanism's operational status data from the ground host, including the operational data of the traction mechanism and the hoisting mechanism. The hook-end equipment typically integrates an embedded industrial control computer, sensor components, and a wireless communication module mounted on the hook assembly. It also integrates a high-performance computing unit for point cloud processing and distance calculation tasks.

[0109] The hook-end device also receives detection area activation commands and dynamic safety distances generated by the ground host based on the mechanism's operating status data. The detection area activation command instructs the activation of the traction follow-up detection area corresponding to the traction mechanism's operating data and / or the hoisting follow-up detection area corresponding to the hoisting mechanism's operating data. The dynamic safety distance determines the dimensions of the traction follow-up detection area and / or the hoisting follow-up detection area. After parsing the configuration parameters sent by the ground host, the hook-end device dynamically constructs the corresponding geometric detection area in its local memory, allowing the sensing range to be reshaped in real time according to the operating conditions.

[0110] The hook-end device is also used to generate and activate the corresponding follow-up detection area in the follow-up coordinate system based on the detection area activation command and dynamic safety distance. The follow-up coordinate system is fixed to the hook. The hook-end device corrects the spatial orientation of the detection area in real time based on the data from the internally integrated inertial measurement unit (IMU) to ensure that it always maintains the correct orientation relative to the hook body.

[0111] The hook-end device is also used to acquire obstacle point cloud data within the activated servo detection area. Based on the obstacle point cloud data and the pre-recorded hook circumferential envelope, it calculates multi-dimensional distance data, including horizontal distance, vertical distance, and shortest distance. Utilizing edge computing capabilities, the hook-end device condenses the raw data from the four solid-state LiDARs into three values—horizontal distance, vertical distance, and shortest distance—after synchronous acquisition, coordinate transformation, filtering and noise reduction, and feature extraction. This significantly reduces the demand for wireless uplink bandwidth and ensures real-time obstacle avoidance response.

[0112] The hook-end device is also used to report multi-dimensional distance data to the ground host. The hook-end device sends data packets containing horizontal distance, vertical distance, shortest distance and heartbeat identifier to the ground host at preset intervals via wireless communication module. It only acts as an "information provider" and does not perform any deceleration or stopping operations.

[0113] The ground control unit is used to calculate the dynamic safety distance based on the mechanism's operating status data and generate activation commands for the detection area. The ground control unit is usually a programmable logic controller (PLC) or an industrial computer, which is directly connected to peripheral equipment such as the cable crane's drivers, encoders, and weather stations. Based on complete global status information, it accurately calculates the theoretical braking distance, system delay compensation distance, and safety margin required at the current speed, and then generates appropriate dynamic safety distance parameters.

[0114] The ground control unit is also used to execute tiered response strategies based on multi-dimensional distance data and dynamic safety distances. The ground control unit compares the received multi-dimensional distance data with internally calculated warning and braking thresholds in real time. Once a collision risk is determined, it issues speed-limiting or emergency stop commands to the cable car drive system and triggers audible and visual alarms. Because the control loop is completely closed on the ground side, the direct impact of wireless link jitter on safe execution is avoided, ensuring that the ultimate safety protection capability is in the hands of the most reliable control node.

[0115] A dual-band wireless link plays a crucial role as the communication channel between the hook-end equipment and the ground host. Specifically, this link typically employs a dual-band redundant design of 2.4GHz and 5GHz to cope with the complex electromagnetic interference environment at construction sites. The downlink channel carries low-frequency, high-reliability data such as mechanism status, control commands, and wind speed, while the uplink channel carries high-frequency, real-time data such as multi-dimensional distance and heartbeat. This bidirectional asynchronous data stream design ensures both timely issuance of control commands and continuous transmission of sensing data.

[0116] Through the aforementioned system architecture, this application achieves the synergistic advantage of "perception on the hook, decision-making on the ground." The hook-end equipment, utilizing its unique perspective close to the work surface, solves the problems of blind spots and lag in ground-side perception; the ground-based host, with its powerful computing and control capabilities, compensates for the limitations of limited computing power and difficulties in safe execution at the hook end. The two work closely together through standardized data interfaces and interaction protocols, forming a complete protection system that possesses both near-field fine-grained perception capabilities and global safety management capabilities. This modular and decoupled system design not only improves the feasibility and maintainability of the technical solution but also reserves ample interface space for subsequent functional expansion (such as adding visual sensors, connecting to a digital twin system, etc.).

[0117] Example 7

[0118] This embodiment provides a comprehensive application example in a dam construction scenario, aiming to verify the actual operational effectiveness of the cable crane hook directional protection and dynamic speed-following obstacle avoidance method and system described in the previous embodiment through typical working conditions. It should be understood that the following description is merely illustrative and not intended to limit the scope of protection of this application. In practical applications, the technical solution of this application can be widely applied to various complex operating environments such as water conservancy and hydropower projects, bridge construction, and the hoisting of large components.

[0119] The specific operational scenario is set near the concrete pouring sump of a dam, where a crane performs unloading operations using a hoisted container. This scenario exhibits typical dynamic boundary characteristics, meaning the location of the sump changes continuously as the pouring layer rises, and the site is often accompanied by uncertainties such as sudden strong winds at high altitudes and temporary material stockpiles. Under this scenario, the comprehensive operation process of the technical solution proposed in this application is as follows:

[0120] As the hook carries the container from the material collection point to the pouring silo opening, the ground control unit collects the real-time operating speeds of the traction and hoisting mechanisms when the hook approaches the silo opening for alignment. If the traction speed is 2.0 m / s and the hoisting speed is 0.8 m / s, the ground control unit calculates the dynamic safety distances D_warn_trac on the traction side and D_warn_hoist on the hoisting side based on preset braking parameters and sends them to the hook-end equipment. The hook-end equipment immediately generates and activates the corresponding traction and hoisting follow-up detection areas, and their lengths in the direction of movement automatically expand to the values ​​indicated by the corresponding dynamic safety distances. When the hook decelerates and makes minor adjustments for alignment, the detection areas shrink accordingly to avoid misjudging known background elements such as the silo wall as obstacles.

[0121] During operation, if a sudden strong wind, common in high-altitude mountainous areas, is encountered, the wind speed sensor detects a sudden increase in real-time wind speed w to 15 m / s. When the hook-end equipment performs wind speed-related dynamic background differential processing, it calculates the wind speed according to the formula... Automatically increase the judgment threshold Due to the existence of the wind speed adjustment coefficient k, the judgment threshold increases linearly with the increase of wind speed, which can effectively filter out false echoes caused by sandstorms or slight swaying of cables, prevent the system from frequently malfunctioning and even triggering emergency braking erroneously, and maintain the continuity and stability of operations even under severe weather conditions.

[0122] As the hook continued its descent towards the silo surface, radar detected a temporary material pile (such as a bundle of rebar or formwork components) at the edge of the silo opening that was not pre-entered into the system. The hook-end equipment, through active stereo sensing, promptly detected this unknown obstacle, clustered the downsampled point cloud data, and calculated the horizontal, vertical, and shortest distances, reporting this information to the ground control unit in real time. The ground control unit determined that the vertical distance was less than the warning threshold but greater than the braking threshold, and immediately triggered a tiered response strategy: issuing audible and visual warning signals and limiting the hoisting mechanism speed to a safe low range (e.g., 20% of the rated speed), prohibiting further acceleration but allowing continued controlled descent, effectively preventing a collision.

[0123] In the above operation process, if the wireless link between the hook-end equipment and the ground host is briefly interrupted due to electromagnetic interference or obstruction, the system will automatically enter fault-tolerant protection mode. When the ground host loses heartbeat frames consecutively for a preset number (e.g., 3 times), it immediately implements rate-limiting protection for the relevant mechanisms and switches to a protection mode based on the ground-side 3D environment model. In this mode, the ground host uses the voxel map previously built during idle periods, combined with the current hook position fed back by the cable crane encoder, to perform constraint detection. When wireless communication returns to normal, the ground host automatically exits the protection mode and switches back to the real-time detection mode, while simultaneously using the latest data to correct the ground-side 3D environment model.

[0124] In summary, the technical solution of this application has demonstrated multiple technical advantages through the above-mentioned typical scenarios, including active perception of unknown obstacles, adaptive adjustment of protection range according to speed, intelligent suppression of environmental interference, and fault tolerance for communication anomalies. It effectively solves the technical pain points mentioned in the background technology, such as large protection blind spots, poor dynamic adaptability, and low accuracy of safety distance matching in fixed restricted area solutions.

[0125] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Those skilled in the art should understand that the specific parameter values ​​given in the above embodiments (such as neighborhood radius, voxel side length, velocity threshold, heartbeat cycle, fixed side length of the detection area, etc.), the number and layout of radars, the type of communication link, and the protection mode constraint strategy, etc., are all illustrative examples. In practical applications, they can be adaptively adjusted according to sensor performance, cable machine model, operating environment, and safety regulations. All changes or substitutions made within the technical scope disclosed in this application should be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for directional protection and dynamic speed-following obstacle avoidance of a cable crane hook, applied to hook end equipment, characterized in that, include: S100, Receive mechanism operation status data sent by the ground host, the mechanism operation status data includes the operation data of the traction mechanism and the operation data of the hoisting mechanism; S200, receiving a detection area activation command and a dynamic safety distance generated by the ground host based on the mechanism's operating status data. The detection area activation command is used to indicate the activation of the traction follow-up detection area corresponding to the operating data of the traction mechanism and / or the hoisting follow-up detection area corresponding to the operating data of the hoisting mechanism. The dynamic safety distance is used to determine the size of the traction follow-up detection area and / or the hoisting follow-up detection area. S300, according to the detection area activation command and the dynamic safety distance, generate and activate the corresponding follow-up detection area in the follow-up coordinate system, the follow-up coordinate system being fixed to the hook; S400, acquire obstacle point cloud data located within the activated follow-up detection area, and calculate multi-dimensional distance data based on the obstacle point cloud data and the pre-recorded hook outer envelope. The multi-dimensional distance data includes horizontal distance, vertical distance and shortest distance. S500, the multidimensional distance data is reported to the ground host so that the ground host can execute a graded response strategy based on the multidimensional distance data and the dynamic safety distance.

2. The method according to claim 1, characterized in that, Prior to acquiring obstacle point cloud data within the enabled servo detection area, the method further includes: The point cloud data is acquired synchronously from multiple solid-state lidars, and the point cloud data is transformed to the servo coordinate system according to preset relative extrinsic parameters to obtain combined point cloud data; statistical outlier filtering is performed on the combined point cloud data to remove isolated points whose number of points in the neighborhood is less than a preset minimum neighbor threshold. The point cloud data after statistical outlier filtering is subjected to wind speed-correlated dynamic background difference processing to obtain candidate obstacle point cloud data, including: for each point in the point cloud data after statistical outlier filtering, the corresponding spatial position is found in the background reference model and the relative displacement is calculated. According to the formula Calculate the decision threshold; in, The threshold value is the one used for determining the threshold. Here, k is the base threshold, k is the wind speed adjustment coefficient, and w is the real-time wind speed. When the relative displacement is greater than the determination threshold, the corresponding point is included in the candidate obstacle point cloud data; The candidate obstacle point cloud data is downsampled using voxels. The space is divided into voxels of a preset side length, and the average coordinate of the point falling in each non-empty voxel is used as a representative point to generate downsampled point cloud data.

3. The method according to claim 2, characterized in that, The process of acquiring obstacle point cloud data within the activated servo detection area, and calculating multidimensional distance data based on the obstacle point cloud data and the pre-recorded hook circumferential envelope, includes: The obstacle point cloud data is obtained by extracting point clouds located within the activated servo detection area from the downsampled point cloud data; the obstacle point cloud data is then clustered to obtain obstacle point cloud clusters. Based on the hook's outer envelope and the obstacle point cloud cluster, determine the global nearest obstacle point pair, and calculate the horizontal distance, the vertical distance, and the shortest distance based on the global nearest obstacle point pair.

4. The method according to claim 3, characterized in that, The step of calculating the horizontal distance, the vertical distance, and the shortest distance based on the global nearest obstacle pair includes: The vector formed by the global nearest obstacle point pair is decomposed in the following coordinate system to obtain the component norm in the horizontal plane as the horizontal distance, the absolute value of the component in the vertical direction as the vertical distance, and the Euclidean distance of the global nearest obstacle point pair as the shortest distance.

5. The method according to claim 1, characterized in that, The step of generating and activating a corresponding follow-up detection area in the follow-up coordinate system according to the detection area activation command and the dynamic safety distance includes: When the activation command of the detection area indicates the activation of the traction follow-up detection area, a forward-looking cuboid area is generated along the running direction of the traction mechanism with the origin of the follow-up coordinate system as the anchor point. The length of the forward-looking cuboid area is determined according to the dynamic safety distance. When the activation command of the detection area indicates the activation of the lifting follow-up detection area, a downward-viewing cuboid area is generated along the running direction of the lifting mechanism with the origin of the follow-up coordinate system as the anchor point. The height of the downward-viewing cuboid area is determined according to the dynamic safety distance. When the traction mechanism and the lifting mechanism are running simultaneously, the forward-looking cuboid region and the downward-looking cuboid region are activated at the same time, and the union of the forward-looking cuboid region and the downward-looking cuboid region is taken as the activated follow-up detection region.

6. The method according to claim 5, characterized in that, The operating data of the traction mechanism includes the operating speed of the traction mechanism, and the operating data of the hoisting mechanism includes the operating speed of the hoisting mechanism; The activation command for the detection area is generated by the ground host when it determines that the absolute value of the operating speed of the traction mechanism or the absolute value of the operating speed of the hoisting mechanism is greater than or equal to a preset speed threshold. The width and height of the forward-looking cuboid region are preset fixed values, and the length of the forward-looking cuboid region is equal to the dynamic safety distance corresponding to the traction mechanism. The horizontal cross-sectional side length of the lower-view cuboid region is a preset fixed value, the height of the lower-view cuboid region is equal to the dynamic safety distance corresponding to the lifting mechanism, and the lower-view cuboid region extends only in the vertically downward direction.

7. The method according to claim 1, characterized in that, The acquisition of obstacle point cloud data within the activated follow-up detection area includes: Point cloud data is collected synchronously by four solid-state lidars rigidly installed around the horizontal mounting beam in the middle of the hook. The four solid-state lidars are arranged circumferentially at 90-degree intervals in the horizontal direction. The optical axes of two solid-state lidars face the left and right banks to cover the traction three-dimensional corridor, and the optical axes of the other two solid-state lidars face the vertical downward direction to cover the lifting downward area. Attitude data is acquired by an inertial measurement unit installed at the geometric center of the transverse mounting beam, with the inertial measurement unit serving as the origin of the servo coordinate system; The four solid-state lidars are synchronized at the hardware level using a precise time protocol. Point cloud frames with timestamp differences falling within the allowable threshold are merged to obtain synchronized point cloud data.

8. The method according to claim 1, characterized in that, The method further includes: Heartbeat frames are sent to the ground host at a preset period, and the heartbeat frames are sent together with the multidimensional distance data; When the ground host loses the heartbeat frames consecutively for a preset number of times, it triggers the ground host to perform speed limiting protection and switch to the protection mode based on the ground-side three-dimensional environment model.

9. The method according to claim 8, characterized in that, The method further includes: When the operating data of the traction mechanism and the operating data of the hoisting mechanism both meet the idle criterion and remain so for a preset time, the hook is controlled to drive multiple solid-state lidars to perform multi-station scanning in the construction area to obtain simplified point cloud data. The simplified point cloud data is reported to the ground host, so that the ground host can combine the hook operation position information to construct the ground-side three-dimensional environment model; In the protection mode, the ground host executes speed limiting constraints based on the ground-side three-dimensional environment model.

10. A cable crane hook directional protection and dynamic speed-following obstacle avoidance system, characterized in that, Includes hook-end equipment and ground-based main unit; The hook end device is used for: Receive the mechanism operation status data sent by the ground host, the mechanism operation status data including the operation data of the traction mechanism and the operation data of the hoisting mechanism; The system receives a detection area activation command and a dynamic safety distance generated by the ground host based on the mechanism's operating status data. The detection area activation command is used to indicate the activation of the traction follow-up detection area corresponding to the operating data of the traction mechanism and / or the hoisting follow-up detection area corresponding to the operating data of the hoisting mechanism. The dynamic safety distance is used to determine the size of the traction follow-up detection area and / or the hoisting follow-up detection area. According to the activation command of the detection area and the dynamic safety distance, a corresponding follow-up detection area is generated and activated in the follow-up coordinate system, which is fixed to the hook; Obtain obstacle point cloud data within the activated follow-up detection area; calculate multidimensional distance data based on the obstacle point cloud data and the pre-recorded hook circumferential envelope; the multidimensional distance data includes horizontal distance, vertical distance, and shortest distance. The multidimensional distance data is reported to the ground host; The ground host is used for: The dynamic safety distance is calculated based on the operational status data of the mechanism, and an activation command for the detection area is generated; A tiered response strategy is executed based on the multidimensional distance data and the dynamic safety distance.

Images