Main and auxiliary platform distributed collaborative operation method and system for overhead distribution line drainage

Abstract

The invention belongs to the technical field of overhead distribution line drainage operation, and provides a main and auxiliary platform distributed collaborative operation method and system for overhead distribution line drainage. A main platform and an auxiliary platform are used for operation; the method comprises the following steps: firstly, obtaining a deployment scheme of a main platform and an auxiliary platform based on positioning data and environmental constraint conditions by taking optimal view division as a target according to a constructed feature map; then, blind area data of the main platform are collected through the auxiliary platform, and features of the data collected by the main platform and the auxiliary platform are extracted and fused; and finally, distributed collaborative operation control of the main platform and the auxiliary platform is carried out according to the fused features. Tasks which cannot be covered by the main platform are undertaken through the auxiliary platform, and the safety problem caused by blind areas in complex procedures can be solved through double-platform cooperation of the main platform and the auxiliary platform.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of overhead distribution line diversion operation, and particularly relates to a main and auxiliary platform distributed collaborative operation method and system for overhead distribution line diversion. BACKGROUND

[0002] With the power grid upgrading, large-scale distributed power (photovoltaic / energy storage) access to distribution network, overhead distribution line as the core carrier of power transmission, its diversion operation (such as branch line access to main line, fault line modification, load transfer) frequency increases year by year. The current distribution network operation requires non-stop operation to become the mainstream mode, and the diversion operation needs to be completed in the live state, such as insulation stripping, parallel groove clamp installation, branch line cutting, etc., and needs to cope with complex operation environment, and puts forward higher requirements for operation efficiency, safety and environmental adaptability. The emergence of operation platform (automatic robot) capable of stripping, wiring, parallel groove clamp installation and branch line cutting solves the problems of live operation and coping with complex operation environment.

[0003] The complex operation environment makes the operation point have obstacles, cross-over lines, vibration caused by platform action, and natural weather changes, etc. special conditions, which require high platform obstacle removal and precise operation; and the current single platform sensing unit has a blind area, which is difficult to ensure safety in complex processes, and there is a lack of environmental perception and operation coordination; the environment of overhead line is affected by external factors, and the cable is easy to swing, which brings challenges to the platform mounted on the cable and the data coordination of multiple platforms. SUMMARY

[0004] In order to solve the above problems, the present application provides a main and auxiliary platform distributed collaborative operation method and system for overhead distribution line diversion. According to the constructed feature map, based on positioning data and environmental constraint conditions, the deployment scheme of the main platform and the auxiliary platform is obtained with the optimal visual angle division as the target; then, the blind area data of the main platform is collected by the auxiliary platform, and the features of the data collected by the main platform and the auxiliary platform are extracted and fused; finally, the distributed collaborative operation control of the main platform and the auxiliary platform is carried out according to the fused features; the tasks that cannot be covered by the main platform are undertaken by the auxiliary platform, and the double platform cooperation of the main platform and the auxiliary platform can solve the safety problem caused by the blind area in complex processes.

[0005] In order to achieve the above purpose, the present application is realized by the following technical scheme: In the first aspect, the present application provides a main and auxiliary platform distributed collaborative operation method for overhead distribution line diversion, which uses a main platform and an auxiliary platform capable of completing walking, stripping, wiring, clamping, breaking and data collection; including: constructing a feature map of the operation area according to the scanning data; According to the constructed feature map, based on positioning data and environmental constraints, a deployment scheme of the main platform and the auxiliary platform is obtained with the optimal visual angle division as the goal; The blind area data of the main platform is collected by the auxiliary platform, and the features of the data collected by the main platform and the auxiliary platform are extracted and fused; According to the fused features, distributed collaborative operation control of the main platform and the auxiliary platform is performed.

[0006] Further, the feature map includes phase line trend, tower coordinate, and obstacle position feature point.

[0007] Further, the optimal visual angle division is: ; Wherein: Coverage is the wrapping range of the visual angle of the two platforms on the work point; Overlap is the proportion of the overlapping area of the visual angle; Distance is the moving distance of the auxiliary platform; 、 and are weights.

[0008] Further, the distance and sag height of the phase line where the main platform and the auxiliary platform are located are determined through the constructed feature map; based on positioning data and environmental constraints, a target function is solved by using a greedy algorithm, and the target coordinates are solved through a local optimal-global approximate optimal step-by-step optimization process, and finally the moving distance of the auxiliary platform along the cable is obtained.

[0009] Further, based on the main platform work phase line coordinates of the feature map, the point cloud data is screened, and only the point cloud data within the main platform range is retained; the original environmental data collected by each platform is processed, and through filtering and noise reduction processing: for laser radar point cloud, isolated voxels with less than a preset number of points and abnormal points deviating from the mean value by a preset multiple of standard deviation are removed; through curvature calculation, the corner points of the conductor and the platform mechanism contour are extracted; for depth camera data, the depth map is smoothed by bilateral filtering to eliminate depth jump noise.

[0010] Further, the feature extraction includes conductor contour extraction, platform contour extraction, and camera feature extraction: For the natural arc formed by the gravity and tension of the conductor, a segmented B-spline curve fitting is adopted; first, the point cloud is divided into multiple fitting segments according to the set distance along the conductor trend, and a B-spline curve is fitted for each segment of point cloud by using the least square method; The main platform body is segmented from the auxiliary platform radar point cloud data based on a region growing algorithm of a normal vector; a point with a reflection intensity not less than a preset intensity and a curvature less than a preset curvature is selected as a seed point in a suspected point cloud cluster; an angle between a normal vector of a neighboring point and the seed point is not greater than a preset angle, a distance between the neighboring point and the seed point is not greater than a preset distance, and if a grown point cloud cluster meets a preset shape, the main platform body is determined, and the body point cloud is output; For the main platform camera, the conductor insulation layer texture is extracted, the wire stripping position is marked by ORB feature point detection, and the three-dimensional coordinates of the wire stripping position are obtained in combination with the depth data; for the auxiliary platform camera, the conductor side arc texture is extracted, the conductor texture in the blind area of the main platform camera is supplemented, and the depth data is corrected in combination with the sag curve fitted by the radar.

[0011] Further, when the features are fused, the weight distribution of the conductor arc feature and the main platform module feature is dynamically adjusted: all features are converted into feature maps; the weight of the radar feature channel is increased, and the weight of the camera feature channel is reduced; the spatial attention weight of the key regions of the conductor sag highest point, the main platform wiring module clamping groove and the mechanical arm end is increased, and the feature is strengthened.

[0012] Further, the weighted feature maps are fused by a convolution layer to obtain an environmental feature map; the environmental feature map includes three layers of key information: a bottom layer, a middle layer and a top layer; wherein the bottom layer information is a three-dimensional profile of the conductor with an arc; the middle layer information is a three-dimensional bounding box and component coordinates of each module of the main platform; and the top layer information is a conductor texture and a texture annotation of a working state.

[0013] Further, after the main platform receives the target work point coordinates, the walking module is started to travel at a set speed, and the real-time position is fed back by using the self-positioning system; at the same time, the auxiliary platform moves to a preset position beside the main platform, the laser radar is started to scan the surrounding environment, the blind area data of the main platform is collected, the positioning data of the main platform and the laser radar point cloud of the auxiliary platform are fused and processed until the positioning data coincides with the target coordinates; The main platform starts the wire stripping module according to the preset parameters, and the real-time wire stripping torque is fed back by the built-in torque sensor; the auxiliary platform synchronously starts the infrared thermal imager and the industrial camera to collect the conductor temperature, the exposed conductor length and the insulation layer residue in the wire stripping area; the conductor temperature, the exposed conductor length and the visual information of the auxiliary platform are fused and analyzed; if the insulation layer residue exceeds the threshold, rework is performed; The end effector of the main platform robotic arm grasps the guide wire and guides it into the wiring module. The auxiliary platform's LiDAR tracks the robotic arm's movement in real time, while a depth camera captures the arm's posture. Based on the auxiliary platform's point cloud data and image data, motion deviations are analyzed in real time, and a PID control algorithm calculates the correction amount, which is then used to correct the trajectory of the main platform robotic arm. When the camera detects that the wire clamp and conductor adhesion rate has reached a preset value, a tightening command is issued. The auxiliary platform collects data at the drainage point where the line needs to be cut. It obtains the world coordinate system coordinates of the parallel trench clamp through radar point cloud data and coordinates. It selects the cutting position from the point cloud data of the drainage line and sends the coordinates of the operation position to the main platform for robotic arm path planning and end-effector attitude control.

[0014] Secondly, this invention also provides a distributed collaborative operation system for main and auxiliary platforms for overhead power distribution line diversion, which uses a main platform and an auxiliary platform capable of performing tasks such as walking, stripping, connecting, clamping, cutting, and data acquisition; including: Construct a feature map of the work area based on the scanned data; Based on the constructed feature map, and taking into account the location data and environmental constraints, the deployment schemes for the main platform and auxiliary platform are obtained with the goal of optimal division of labor from the perspective. Data from the blind spots of the main platform is collected by the auxiliary platform, and the features of the data collected by the main platform and the auxiliary platform are extracted and fused. Based on the integrated characteristics, distributed collaborative operation control is implemented between the main platform and auxiliary platforms.

[0015] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the main and auxiliary platform distributed collaborative operation method for overhead power distribution line diversion described in the first aspect.

[0016] Fourthly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the program to implement the steps of the main and auxiliary platform distributed collaborative operation method for overhead power distribution line diversion described in the first aspect.

[0017] Fifthly, the present invention also provides a computer program product, the computer program product comprising a computer program, which, when executed by a processor, implements the steps of the main and auxiliary platform distributed collaborative operation method for overhead power distribution line diversion described in the first aspect.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes two platforms: a main platform and an auxiliary platform. First, based on the constructed feature map, location data, and environmental constraints, and with the goal of optimal perspective division of labor, a deployment scheme for the main platform and auxiliary platform is obtained. Then, blind spot data of the main platform is collected through the auxiliary platform, and features of the data collected by the main platform and auxiliary platform are extracted and fused. Finally, distributed collaborative operation control of the main platform and auxiliary platform is performed based on the fused features. By having the auxiliary platform undertake tasks that the main platform cannot cover, the dual-platform cooperation of the main platform and auxiliary platform can solve the safety problems caused by blind spots in complex processes. Attached Figure Description

[0019] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0020] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the greedy algorithm in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the business platform structure in Embodiment 1 of the present invention; The components include: 1. Main support unit; 2. Wire stripping module; 3. UAV hoisting mechanism; 4. Camera; 5. Wiring module; 6. Wire cutting clamping module; and 7. Robotic arm. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0023] Main platform: Equipment responsible for core processes such as wire stripping, splicing, installation of parallel cable clamps, and branch line cutting during diversion operations.

[0024] Auxiliary platform: Equipped with additional tools (used to replace the end effector of the robotic arm on the main platform to complete power distribution tasks such as insulation restoration) and sensing units to assist the main platform in its work.

[0025] Distributed task allocation: Based on the job planning of the main platform, the auxiliary platform realizes real-time data interaction through 5G / edge computing and undertakes tasks that the main platform cannot cover.

[0026] Collaborative operation safety control: Design a space avoidance protocol for the main and auxiliary platforms—share the positions of both platforms in real time through positioning to ensure a distance of ≥1.5 meters (avoiding collisions).

[0027] Data interaction: Sensor data from auxiliary platforms and drones helps the main platform build environmental models and perform risk monitoring and other tasks.

[0028] Example 1: This embodiment provides a distributed collaborative operation method for main and auxiliary platforms for overhead power distribution line diversion. In order to realize the system corresponding to the method, the system clarifies the functional division of each main body, constructs a low-latency data interaction network, and designs a dynamic task scheduling and safety control mechanism to achieve the parallelization, intelligence and high safety of overhead power distribution line diversion operation.

[0029] The system in this embodiment includes an A-phase operation platform, a B-phase operation platform, and edge computing nodes. Both the A-phase and B-phase operation platforms have complete sensing and execution functions and can dynamically switch between primary and secondary roles according to operational requirements. The specific configuration is shown in Table 1. Table 1. Operating Platform Configuration

[0030] Collaborative multi-view 3D reconstruction: Based on the theory of non-overlapping perspective complementarity and heterogeneous sensing collaboration, this method utilizes the non-overlapping perspective formed by the spatial arrangement of phase lines on the main and auxiliary platforms. Combined with the distributed computing power and decision-making capabilities of edge computing, a closed loop of synchronous acquisition, perspective optimization, data fusion, and coordinate unification is established to achieve 3D environment modeling with both local details and global vision, thus solving the problem of blind spots in single-platform perspectives.

[0031] The primary and secondary platform sensing units have identical configurations and can switch roles at any time. Both collect data simultaneously and upload it to edge computing, improving the completeness of the environmental model through data overlay and feature complementarity.

[0032] Edge computing is based on the spatial distribution of phase lines and the dynamic calculation of the coordinates of the work points to determine the optimal perspective division of labor, driving the auxiliary platform to move to the expected position, minimizing the overlap of perspectives between the two platforms and reducing the perception blind spot as much as possible.

[0033] To avoid crosstalk between lidars, the main platform does not operate its own lidar during operation, but instead uses the lidar of the auxiliary platform to obtain more comprehensive environmental point cloud data.

[0034] Edge computing handles perspective decision-making and scheduling, multi-source data fusion, and coordinate system unification, solving key issues of multi-sensor data collaboration and multi-view stitching, and avoiding insufficient local computing power or data asynchrony on the main and auxiliary platforms.

[0035] like Figure 1 As shown, the method in this embodiment includes: S1, Map Building: The collaborative UAV first performs a global scan of the operation area and uses the SfM (Structure of Motion) algorithm to construct a sparse feature map of the operation area, which includes 3D feature points such as phase line direction, tower coordinates, and fixed obstacle positions. The map data is then bound to the GNSS (Global Navigation Satellite System) differential positioning reference to lay the foundation for subsequent work.

[0036] S2, Optimal Viewpoint Calculation: Edge computing is based on the objective function of maximizing information to calculate the optimal division of labor from different perspectives. ; Where: Coverage is the area of ​​the work point covered by the two platforms' perspectives (target ≥ 95%); Overlap is the percentage of overlapping areas (target ≤ 15%); Distance is the distance the auxiliary platform moves (target ≤ 3m). , and For weights.

[0037] like Figure 2 As shown, the distance between the phase lines of the main and auxiliary platforms is obtained through a pre-built map. Based on the positioning data and environmental constraints, a greedy algorithm is used to solve the objective function. The target coordinates are solved through a stepwise optimization process of local optimum and global approximate optimum, and finally the distance the auxiliary platform moves along the cable is obtained.

[0038] S3, Main and Auxiliary Platform Deployment: Edge computing issues deployment commands to the operation platform based on coordinate data and optimal viewpoint calculations. After the platform moves along the cable to the target location, it verifies the deployment effect by calculating the proportion of key features.

[0039] S4. Multi-source data preprocessing: S4.1 Coordinate Constraints: Based on the coordinates of the main platform's operational phase line using a pre-built map, the point cloud data is filtered, retaining only the point cloud data within the main platform's scope.

[0040] S4.2, Noise Reduction: The raw environmental data collected from each platform is processed in the edge computing nodes. Through filtering and noise reduction, key feature data is preserved while reducing the computational load.

[0041] For lidar point clouds, a combination of voxel filtering and statistical filtering is used for noise reduction to remove isolated voxels with fewer than three points and outliers deviating from the mean by three times the standard deviation. Curvature calculations are used to extract corner points of the guide wires and platform mechanism contours, preserving key geometric features and reducing subsequent computational load.

[0042] For depth camera data, the depth map is smoothed using bilateral filtering to eliminate depth jump noise. The texture map uses the Retinex algorithm to enhance dark details and avoid texture feature loss due to uneven lighting.

[0043] S4.3 Time Synchronization: By processing environmental data from different sensor units through time synchronization, the system can know the sensor data from different perspectives at the same time, thereby better guiding the operation process.

[0044] Edge nodes generate unified timestamps based on GNSS differential positioning and simultaneously transmit them to the main and auxiliary platforms via 5G. Timestamp alignment is performed on the data from each sensor: if the timestamp of the main platform's LiDAR data is T1 and the timestamp of the auxiliary platform's camera data is T2, edge computing corrects T2 to T1 ± 0.5ms using linear interpolation to avoid reconstruction misalignment caused by spatiotemporal deviations.

[0045] S4.4 Spatial Synchronization (Coordinate Unification): S4.4.1, Matching within the auxiliary platform: Because the LiDAR and depth camera are fixed to the same gimbal on the auxiliary platform, taking advantage of their fixed relative positions, the ICP (Iterative Closest Point) algorithm is used to fuse the LiDAR point cloud and camera data to quickly obtain comprehensive and high-precision environmental information. The ICP algorithm iteratively optimizes a rigid transformation (rotation matrix R and translation vector t) to ensure that the source point cloud (depth camera point cloud) overlaps as much as possible with the target point cloud (LiDAR point cloud) after the transformation. The ICP algorithm includes: Nearest Neighbor Search: For each point in the transformed source point cloud P, find the nearest corresponding point in the target point cloud Q. Euclidean distance is used as the nearest neighbor criterion, and a KD-Tree is employed to accelerate the nearest neighbor search. By constructing a spatial binary tree structure, the KD-Tree reduces the average search complexity from O(N) to O(logN), significantly improving efficiency.

[0046] Removing erroneous point pairs: Not all initially matched point pairs are correct. Here, point pairs whose distance exceeds a preset threshold are discarded to remove erroneous matches, thus improving the robustness and accuracy of the algorithm.

[0047] Calculate the optimal transformation: Based on the current valid corresponding point pairs, find an optimal rigid transformation that minimizes the error between corresponding points. This is a least squares problem, solved using Singular Value Decomposition (SVD) to find a closed-form solution. First, calculate the centroids of the valid source and target point sets respectively, obtaining the covariance matrix H of the decentralized point set. Then, perform Singular Value Decomposition on matrix H, and calculate the rotation matrix R and translation vector t based on the SVD results. Apply the calculated R and t to the original source point cloud P to obtain a new source point cloud P', which is used for the next iteration.

[0048] The resulting fused data combines the geometric accuracy of the LiDAR (±0.5mm) with the texture information of the depth camera, and is used to cover blind spots of the main platform (such as obstacles below and to the side of the work point).

[0049] S4.4.2 Unified coordinates between primary and secondary platforms: Edge computing uses the GNSS differential positioning data carried by the main and auxiliary platforms to perform coordinate mapping of the sensor point cloud based on the world coordinate system of the pre-built map.

[0050] Specifically, based on the platform's own GNSS coordinates and the relative coordinates of the feature points obtained by the sensing unit, the feature points are mapped to the world coordinate system.

[0051] S5, Feature layer extraction: S5.1, Conductor Profile Extraction: To address the natural curvature of conductors caused by gravity and tension (the sag of 10kV distribution lines is typically 2-5m), piecewise B-spline curve fitting is used instead of traditional RANSAC linear fitting.

[0052] First, the point cloud is divided into fitting segments at predetermined distances along the conductor's direction (set as the X-axis). A B-spline curve is then fitted to each segment using the least squares method. The fitted curve is then corrected using the sag formula for power distribution lines to ensure it conforms to physical and mechanical properties. This ultimately forms a curved 3D conductor profile, providing a physically realistic geometric skeleton for subsequent modeling.

[0053] S5.2 Platform contour extraction: A region growing algorithm based on normal vectors is used to segment the main platform body from the auxiliary platform's radar point cloud data. From the suspected point cloud clusters, areas with a reflection intensity ≥80dB and curvature <0.1m are selected. - The point marked with "¹" is used as the seed point. The growth rule is as follows: The angle between the normal vectors of adjacent points and the seed point should be ≤5° (ensuring they belong to the same plane, such as the outer shell of the main body); the distance between adjacent points and the seed point should be ≤5mm (ensuring spatial continuity). If the grown point cloud cluster satisfies the condition that "its volume and shape are close to a cuboid", it is determined to be the main body of the main platform, and the main body point cloud is output.

[0054] Based on the main point cloud, morphological templates are used to distinguish the various functional modules of the platform. The main point cloud is clustered according to the distance between adjacent points ≤3mm, resulting in multiple sub-point cloud clusters. According to the preset module size, clusters with mismatched sizes are eliminated, retaining 3-4 potential module clusters. A template matching algorithm is used to calculate the chamfer distance between the potential clusters and the template, thereby matching the corresponding modules.

[0055] S5.3, Camera Feature Extraction: For the main platform camera (facing the work area), extract the texture of the wire insulation layer (such as stripes and markings on the surface of the insulation layer), use ORB feature points to detect and mark the stripping location, and combine the depth data to obtain the three-dimensional coordinates of the stripping location.

[0056] For the auxiliary platform camera (side-viewing main platform), extract the lateral curvature texture of the guide wire to supplement the guide wire texture in the blind spot of the main platform camera's field of view, and combine it with the sag curve fitted by the radar to correct the depth data.

[0057] The auxiliary platform camera plays a primary role in data acquisition, capturing images of the platform module's operational status and wiring clamp status. Template matching is used to confirm the module's operational status and wiring effectiveness. It also captures the end effector posture of the main platform's robotic arm, extracting the robotic arm joint contours through edge detection and calculating the relative distance between the end effector and the wiring module using depth data, thus supplementing the radar contour with detailed information. The main platform camera is responsible for acquiring the position and posture information of the robotic arm's end effector from this perspective.

[0058] S6. Feature Fusion: Feature Complementary Fusion The edge computing loading space-channel dual attention CNN model dynamically adjusts the weight allocation based on the corrected wire curvature features and the main platform module features.

[0059] S6.1 Feature Input: The feature input layer inputs radar and camera features. Radar features include control points of the curved wire B-spline curve (100 points / span), 3D bounding boxes of each module of the main platform, and coordinates of key components. Camera features include ORB feature points of the wire insulation layer texture (500 points / m), template matching results of the main platform modules, and depth data of the robotic arm's end effector. All data is formatted uniformly, converting all features into 256×256×64 feature maps (radar features occupy 32 channels, and camera features occupy 32 channels).

[0060] S6.2 Attention Weight Allocation: S6.3, Channel Attention: The radar feature channel weight has been increased to 0.75 (from 0.7), as the conductor curvature and the main platform module outline are the geometric basis for 3D modeling, and the accuracy has been improved to ±0.3mm after correction. The camera feature channel weight has been adjusted to 0.25 (from 0.3), mainly responsible for supplementing texture details (such as insulation layer damage identification) and confirming module status (such as whether the wire clamp is open).

[0061] S6.4 Spatial Attention: For key areas such as the highest point of the conductor sag, the main platform wiring module slot, and the end effector of the robotic arm, the spatial attention weight is set to 0.9 (0.1 for the background area) to ensure that the features of the core working area are enhanced. For example, the feature point at the main platform wire stripping location (X=100.2m, Y=50.1m) has the highest weight in the spatial attention map and is given priority in the fusion calculation.

[0062] S6.5 Feature Fusion Output: The weighted feature maps are fused using a CNN convolutional layer (3×3 kernel, ReLU activation function) to output a 1024×1024×3 environmental feature map; the map includes key information from three layers: bottom, middle, and top. The bottom layer information contains the 3D contour of the conductor with curvature (B-spline curve fitting result); the middle layer information contains the 3D bounding boxes of each module of the main platform and the coordinates of key components (such as the center point of the stripper and the wiring slot); the top layer information contains the texture of the conductor and the texture annotation of the module's working status (such as the rotation of the stripper and the opening and closing of the robotic arm).

[0063] The positioning accuracy of key areas (such as wiring modules) in the fused feature map is ≤ ±0.5mm, providing geometrically accurate and state-clear input data for subsequent 3D modeling.

[0064] S7. 3D Model Generation and Update: Edge computing is based on the fused feature map and constructs a dense point cloud model through the MVS (Multi-View Stereo Matching) algorithm; it uses the Poisson reconstruction algorithm to generate a mesh model and combines it with the texture data of the main platform camera for mapping. The model update frequency is 10Hz, providing millimeter-level and blind-zone-free environmental support for processes such as wire stripping and splicing.

[0065] S8, Main and Auxiliary Platform System Operations: like Figure 3 As shown, both the main and auxiliary platforms include a main support unit 1, and a wire stripping module 2, a drone hoisting mechanism 3, a camera 4, a wiring module 5, a wire cutting clamping module 6, and a robotic arm 7, all mounted on the main support unit 1.

[0066] The platform has processes such as walking, stripping, connecting, clamping, and cutting. Walking and stripping are achieved through specific modules integrated into the platform. Connecting is achieved by clamping the lead wire with the end effector of the robotic arm, guiding the end of the lead wire to extend into the parallel groove clamp of the connecting module, and the tightening mechanism of the connecting module tightens the parallel groove clamp to connect the lead wire. Cutting and clearing obstacles are achieved by the end effector of the robotic arm, which has a clamping device and a cutting device.

[0067] Based on the division of labor of main execution, auxiliary perception and edge collaboration, the auxiliary platform provides blind spot data and effect verification in each process, while edge computing is responsible for data fusion, parameter decision and process scheduling to ensure the accuracy and safety of core processes.

[0068] S8.1 Walking and Positioning (Auxiliary Platform Blind Spot Filling + Edge Computing Fusion): After receiving the target work point coordinates from the edge node, the main platform activates its walking module to move at a set speed and uses its own GNSS positioning system to report its real-time position. Simultaneously, the auxiliary platform moves to a position 2.5 meters to the side of the main platform, activates its lidar to scan the surrounding environment, collects blind spot data including that of the main platform, and transmits it to the edge node via the 5G network. The edge computing node fuses the GNSS positioning data from the main platform with the lidar point cloud from the auxiliary platform, assesses the walking risk in real time, and continues until the positioning data coincides with the target coordinates. Then, a clamping command is issued, controlling the clamping mechanism to grip the overhead cable and secure the platform to it.

[0069] S8.2, Stripping (Auxiliary Platform Evaluation + Edge Computing Decision): The main platform activates the wire stripping module based on preset parameters and provides real-time stripping torque feedback via a built-in torque sensor. Simultaneously, the auxiliary platform uses an infrared thermal imager and an industrial camera to collect data on the wire temperature, exposed wire length, and insulation residue in the stripping area. Edge nodes fuse and analyze the temperature, length, and visual information uploaded from the auxiliary platform to generate a wire stripping performance evaluation report. If the result is satisfactory, a "Stop stripping, switch wiring module" command is sent to the main platform; if insulation residue exceeds a threshold, rework is initiated, adjusting the stripping blade speed and feed distance, and performing additional stripping.

[0070] S8.3 Wiring (Auxiliary Platform Trajectory Assistance + Edge Calculation Correction): The end effector of the main platform's robotic arm grasps the guide wire and guides it into the wiring module. During this process, the auxiliary platform's LiDAR tracks the robotic arm's movement in real time, while a depth camera captures and uploads images of the robotic arm's posture. The edge node analyzes motion deviations in real time based on the auxiliary platform's point cloud and image data, calculates corrections using a PID control algorithm, and sends these corrections to the main platform's robotic arm controller for trajectory correction. When the camera detects that the wire clamp and conductor have a 98% fit, the edge node further issues a tightening command. After wiring is completed, the edge node determines the wiring quality is acceptable based on the temperature and temperature rise data monitored at the wiring point by the auxiliary platform's infrared thermal imager.

[0071] S8.4 Disconnection (Auxiliary Platform Attitude Monitoring + Edge Computing Security Verification): The auxiliary platform collects data at the point where the wire needs to be cut. The edge computing node obtains the world coordinates of that point and the wire clamp using radar point cloud data and its own GNSS coordinates. The edge node selects a suitable cutting location from the point cloud data of the wire and sends the coordinates of this location to the main platform for robotic arm path planning and end-effector attitude control. During the operation, the edge computing node tracks the robotic arm path and end-effector attitude using radar and camera data from the auxiliary platform and makes real-time corrections. After confirming that the distance between the intersecting lines meets safety requirements, the edge node issues a cutting command. After the wire is cut, the auxiliary platform further reports the distance between the branch line and ground obstacles, and the edge node uses this information to guide the main platform in adjusting the branch line's lowering speed.

[0072] Example 2: This embodiment provides a distributed collaborative operation system for main and auxiliary platforms for overhead power distribution line diversion, using a main platform and an auxiliary platform capable of performing tasks such as walking, stripping, connecting, clamping, disconnecting, and data acquisition; including: Construct a feature map of the work area based on the scanned data; Based on the constructed feature map, and taking into account the location data and environmental constraints, the deployment schemes for the main platform and auxiliary platform are obtained with the goal of optimal division of labor from the perspective. Data from the blind spots of the main platform is collected by the auxiliary platform, and the features of the data collected by the main platform and the auxiliary platform are extracted and fused. Based on the integrated characteristics, distributed collaborative operation control is implemented between the main platform and auxiliary platforms.

[0073] The working method of the system is the same as the distributed collaborative operation method of the main and auxiliary platforms for overhead power distribution line diversion in Embodiment 1, and will not be repeated here.

[0074] Example 3: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the main and auxiliary platform distributed collaborative operation method for overhead power distribution line diversion described in Embodiment 1.

[0075] Example 4: This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the program, it implements the steps of the main and auxiliary platform distributed collaborative operation method for overhead power distribution line diversion described in Embodiment 1.

[0076] Example 5: This embodiment provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the main and auxiliary platform distributed collaborative operation method for overhead power distribution line diversion described in Embodiment 1.

[0077] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A distributed collaborative operation method for main and auxiliary platforms for overhead power distribution line diversion, which uses a main platform and an auxiliary platform capable of walking, stripping, connecting, clamping, cutting, and data acquisition; characterized in that, include: Construct a feature map of the work area based on the scanned data; Based on the constructed feature map, and taking into account the location data and environmental constraints, the deployment schemes for the main platform and auxiliary platform are obtained with the goal of optimal division of labor from the perspective. Data from the blind spots of the main platform is collected by the auxiliary platform, and the features of the data collected by the main platform and the auxiliary platform are extracted and fused. Based on the integrated characteristics, distributed collaborative operation control is implemented between the main platform and auxiliary platforms.

2. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 1, characterized in that, The feature map includes phase line direction, tower coordinates, and obstacle location feature points.

3. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 1, characterized in that, The optimal division of labor from the perspective is as follows: ； Where: Coverage is the area encompassed by the two platforms' perspectives over the work point; Overlap is the percentage of overlapping areas; Distance is the distance the auxiliary platform moves. , and For weights.

4. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 3, characterized in that, The distance between the phase lines of the main platform and the auxiliary platform is determined by constructing a feature map; based on the positioning data and environmental constraints, the objective function is solved using a greedy algorithm, and the target coordinates are solved through a stepwise optimization process of local optimum and global approximate optimum, and finally the distance the auxiliary platform moves along the cable is obtained.

5. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 1, characterized in that, Based on the phase line coordinates of the main platform operation using the feature map, the point cloud data is filtered, retaining only the point cloud data within the main platform area; the raw environmental data collected by each platform is processed through filtering and noise reduction: for LiDAR point clouds, isolated voxels with fewer than a preset number of points and outliers deviating from the mean by a preset number of standard deviations are removed; corner points of the guide wires and platform mechanism contours are extracted through curvature calculation; for depth camera data, the depth map is smoothed using bilateral filtering to eliminate depth jump noise.

6. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 1, characterized in that, Feature extraction includes conductor contour extraction, platform contour extraction, and camera feature extraction: To address the natural curvature of the conductor caused by gravity and tension, a piecewise B-spline curve fitting method is used. First, the point cloud is divided into multiple fitting segments along the conductor's direction at a set distance. For each segment of the point cloud, the least squares method is used to fit a B-spline curve. A region growing algorithm based on normal vectors is used to segment the main platform body from the radar point cloud data of the auxiliary platform; points with reflection intensity not less than a preset intensity and curvature less than a preset curvature are selected as seed points from the suspected point cloud clusters. The angle between the normal vectors of adjacent points and the seed point is no greater than a preset angle, and the distance between adjacent points and the seed point is no greater than a preset distance. If the grown point cloud cluster meets the preset shape, it is determined to be the main body of the main platform, and the main point cloud is output. For the main platform camera, the insulation layer texture of the conductor is extracted, and the stripping position is marked by ORB feature point detection. The three-dimensional coordinates of the stripping position are obtained by combining the depth data. For the auxiliary platform camera, the lateral curvature texture of the conductor is extracted to supplement the conductor texture in the blind spot of the main platform camera's view. The sag curve fitted by the radar is combined to correct the depth data.

7. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 6, characterized in that, During feature fusion, the weight allocation for conductor curvature features and main platform module features is dynamically adjusted: all features are converted into feature maps. Increase the weight of radar feature channels and decrease the weight of camera feature channels; Spatial attention weights are increased for the highest point of the conductor sag, the main platform wiring module slot, and the key area at the end of the robotic arm to enhance features.

8. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 7, characterized in that, The weighted feature maps are fused by convolutional layers to obtain an environmental feature map. The environmental feature map includes three layers of key information: bottom layer, middle layer and top layer. The bottom layer information is the three-dimensional contour of the curved wire; the middle layer information is the three-dimensional bounding box and component coordinates of each module of the main platform; and the top layer information is the texture of the wire and the texture annotation of the working status.

9. The distributed collaborative operation method of main and auxiliary platforms for overhead power distribution line diversion as described in claim 1, characterized in that, After receiving the coordinates of the target work point, the main platform starts the walking module and moves at a set speed, and uses its own positioning system to report the real-time position. At the same time, the auxiliary platform moves to a preset position to the side of the main platform, starts the lidar to scan the surrounding environment, collects the blind spot data of the main platform, and fuses the positioning data of the main platform with the lidar point cloud of the auxiliary platform until the positioning data coincides with the target coordinates. The main platform starts the wire stripping module according to preset parameters and feeds back the real-time wire stripping torque through the built-in torque sensor; the auxiliary platform simultaneously activates the infrared thermal imager and industrial camera to collect the wire temperature, exposed wire length and insulation layer residue in the stripping area; the wire temperature, exposed wire length and visual information of the auxiliary platform are fused and analyzed; if the insulation layer residue exceeds the threshold, rework is performed; The end effector of the main platform robotic arm grasps the guide wire and guides it into the wiring module. The auxiliary platform's LiDAR tracks the robotic arm's movement in real time, while a depth camera captures the arm's posture. Based on the auxiliary platform's point cloud data and image data, motion deviations are analyzed in real time, and a PID control algorithm calculates the correction amount, which is then used to correct the trajectory of the main platform robotic arm. When the camera detects that the wire clamp and conductor adhesion rate has reached a preset value, a tightening command is issued. The auxiliary platform collects data at the drainage point where the line needs to be cut. It obtains the world coordinate system coordinates of the parallel trench clamp through radar point cloud data and coordinates. It selects the cutting position from the point cloud data of the drainage line and sends the coordinates of the operation position to the main platform for robotic arm path planning and end-effector attitude control.

10. A distributed collaborative operation system for main and auxiliary platforms for overhead power distribution line diversion, which uses a main platform and an auxiliary platform capable of performing tasks such as walking, stripping, connecting, clamping, cutting, and data acquisition; characterized in that, include: Construct a feature map of the work area based on the scanned data; Based on the constructed feature map, and taking into account the location data and environmental constraints, the deployment schemes for the main platform and auxiliary platform are obtained with the goal of optimal division of labor from the perspective. Data from the blind spots of the main platform is collected by the auxiliary platform, and the features of the data collected by the main platform and the auxiliary platform are extracted and fused. Based on the integrated characteristics, distributed collaborative operation control is implemented between the main platform and auxiliary platforms.

Images