A robotic arm path planning and maintenance method for uninterrupted power distribution network operations

Abstract

This invention provides a method for robotic arm path planning and maintenance for live-line work in power distribution networks, belonging to the field of robotic arm path planning and maintenance technology. It plans a maintenance path for the robotic arm performing live-line work that meets geometric collision-free, electrical insulation safety, and compliance with live-line operation procedures, establishing a four-dimensional composite configuration space including the amplitude of the electric field intensity in the work area. Rigid prohibition points are marked according to the maximum allowable electric field intensity threshold, and a risk potential field layer is constructed by combining the electric field amplitude and distance risk potential field function. The task type is analyzed, and electrical safety operation timing logic rules are retrieved and converted into path search cost penalty terms. A fusion path search is performed based on an improved fast random expansion tree algorithm, and geometric collision, electric field rigid prohibition, cumulative risk potential field, and timing logic consistency checks are performed on the generated new nodes. This invention embeds electromagnetic situational awareness into the planning kernel, achieving dual geometric and electrical obstacle avoidance, significantly improving the safety of live-line work.

Description

Technical Field

[0001] This invention provides a robotic arm path planning and maintenance method for uninterrupted power distribution network operations, belonging to the field of robotic arm path planning and maintenance technology. Background Technology

[0002] With the increasing demands for power supply reliability from the national economy, live-line work on distribution networks has become a key means of ensuring electricity supply for people's livelihoods and production. Traditional manual live-line work requires workers to wear heavy full-body shielding suits and directly contact high-voltage live conductors at heights of several meters to more than ten meters in strong electric fields. This not only involves extremely high work intensity but also poses a serious risk of electric shock. Therefore, using dual-arm or multi-arm robots to replace manual labor for tasks such as connecting live wires, disconnecting leads, and replacing insulators has become an inevitable trend in the industry.

[0003] In the control technology of live-line working robots for power distribution networks, path planning for the robotic arm is one of the core challenges. Existing path planning methods, such as the fast randomized expanded tree algorithm, probabilistic path graph method, or A* algorithm, are mainly applied in the industrial manufacturing field. When these control methods (algorithms) are directly transferred to the live-line working scenario of power distribution networks, a mismatch between function and effect will occur, mainly manifested in the following ways:

[0004] The problem of singular planning dimensions: Existing technologies typically simplify the distribution network operation environment to a geometric space composed of poles, conductors, insulators, etc., and the planning goal is only to find a collision-free geometric path from the starting point to the end point. However, the core characteristic of live distribution network operation is high voltage and liveness, and the operation space is filled with highly irregular power frequency electromagnetic fields. Existing algorithms completely ignore the physical effects of electromagnetic fields on the robotic arm body. As a result, although the planned path is geometrically collision-free, the upper arm or joints of the robotic arm may get too close to the live conductors during the movement, which can easily cause gap breakdown discharge, or cause the end effector to generate induced voltage due to electromagnetic induction, damaging the precision sensors.

[0005] The problem of the separation between perception and decision-making: Although some improved algorithms have introduced visual obstacle avoidance, they usually adopt a serial architecture of perception first and planning later. That is, environmental modeling is performed before planning begins to fix the obstacle information. This architecture cannot cope with the dynamic changes in electric field distribution caused by the movement of the robotic arm itself during the operation. For example, when a metal robotic arm approaches a wire, it will distort the spatial electric field, making the originally safe area dangerous.

[0006] The problem of missing special operation logic: Distribution network maintenance operations require strict adherence to electrical operation sequences, such as connecting the live wire first and then disconnecting it, or disconnecting the lead wire first and then disconnecting the main line. Existing planning methods only focus on kinematic constraints and cannot embed such electrical safety operation procedures as hard constraints into the path search process. This will cause the robotic arm to follow a trajectory that conforms to geometric kinematics but violates electrical safety rules.

[0007] Therefore, at present, how to plan a maintenance path for the robotic arm in a dynamically changing strong electromagnetic environment that satisfies geometric collision-free operation, electrical insulation safety, and compliance with live-line operation procedures is a technical problem that urgently needs to be solved. Summary of the Invention

[0008] To address the technical problems existing in the background art, the present invention provides a method for robotic arm path planning and maintenance for uninterrupted power distribution network operations, comprising the following steps:

[0009] Step 1: Establish a composite configuration space that includes the electrical properties of the workspace. The composite configuration space is a four-dimensional space formed by calculating the electric field intensity amplitude at discrete spatial points and attaching the electric field intensity amplitude as the fourth dimension data to the three-dimensional coordinate points based on the three-dimensional geometric model.

[0010] Step 2: In the composite configuration space, rigid prohibited access points are marked according to the maximum allowable electric field strength threshold. For non-prohibited access points, the risk potential field value of each non-prohibited access point is calculated by combining the electric field strength amplitude with the risk potential field function of the distance to the nearest charged body surface, and a risk potential field layer is constructed.

[0011] Step 3: Analyze the task type of the maintenance task, retrieve the corresponding standard electrical safety operation timing logic rules according to the task type, and convert the timing logic rules into a cost function penalty term in the path search process;

[0012] Step 4: Perform fusion path search based on the improved fast random expansion tree algorithm. When generating new nodes, perform safety checks on the new nodes, including at least geometric collision detection, electric field rigidity prohibition detection, cumulative risk potential field value detection, and temporal logic consistency detection. Only when all checks are passed will the new node be added to the search tree.

[0013] Step 5: Smooth and post-process the initial planned path generated in Step 4 to generate the final executable maintenance path;

[0014] Step 6: Control the robotic arm to move according to the final executable maintenance path generated in Step 5. During the movement, monitor the status of the robotic arm and the environment in real time. If it is determined that there is a sudden change in the environment or electric field distribution, control the robotic arm to pause the movement and return to Step 1 to trigger the dynamic replanning of the path.

[0015] The specific method for step 1 is as follows:

[0016] Step 1.1: Collect point cloud data and two-dimensional image data of the current working scene using a lidar and binocular vision camera installed on the robotic arm base or bucket truck;

[0017] Step 1.2: Match the collected point cloud data and 2D image data with the pre-stored prior knowledge base of tower type, conductor type, and insulator string length to identify charged and non-charged bodies in the scene, and mark the voltage level of charged bodies;

[0018] Step 1.3: Based on the marking results, construct a preliminary geometric model of the workspace that includes location coordinates, geometric shape, and voltage level information;

[0019] Step 1.4: Based on the preliminary geometric model, calculate the electric field distribution in the work space using the finite element method or boundary element method, and solve for the electric field intensity amplitude at discrete spatial points with the robot arm base coordinate system as the reference.

[0020] Step 1.5: Use the electric field intensity amplitude at all discrete points as the fourth dimension data and attach it to the three-dimensional coordinate points of the preliminary geometric model to form a four-dimensional composite configuration space C_space_E containing the spatial coordinate axis X, spatial coordinate axis Y, spatial coordinate axis Z and the electric field intensity amplitude E;

[0021] In the composite configuration space C_space_E, each discrete point can indicate whether the location is occupied by a geometric entity and the scalar value of the electric field intensity at that location.

[0022] The specific method for step 2 is as follows:

[0023] Step 2.1: Introduce the maximum permissible electric field strength threshold specified in the safety regulations for live-line working. ;

[0024] Step 2.2: Traverse all discrete points in the composite configuration space C_space_E, and compare the electric field intensity amplitude E at each discrete point with the maximum allowable electric field intensity threshold. Compare;

[0025] Step 2.3: If the electric field intensity amplitude E at a certain discrete point is greater than or equal to the maximum allowable electric field intensity threshold... If so, then mark the point as a rigidly forbidden access point;

[0026] Step 2.4: If the electric field intensity amplitude E at a certain discrete point is less than the maximum allowable electric field intensity threshold... Then, based on the distance d between the point and the nearest charged body surface and the electric field intensity amplitude E at the point, the risk potential field value R at that point is calculated using the risk potential field function. The formula for calculating the risk potential field value R is as follows:

[0027] ;

[0028] in, This is the electric field strength weighting coefficient. This is the distance weighting coefficient. To prevent singular value constants where the denominator is zero;

[0029] Step 2.5: Through the above calculations, assign a risk potential field value R between 0 and 1 to each non-forbidden access point, thereby constructing a risk potential field layer in the composite configuration space C_space_E that superimposes electric field intensity risk and geometric proximity risk.

[0030] The specific method for step 3 is as follows:

[0031] Step 3.1: Analyze the task type of the current maintenance task. The task type includes live-line connection work, live-line disconnection work, or live-line insulator replacement work.

[0032] Step 3.2: Based on the task type, retrieve the corresponding standard electrical safety operation sequence logic rules from the database;

[0033] Step 3.3: Formalize the timing logic rules into sequence constraints on the state of the robotic arm end effector. The sequence constraints specifically include: specifying the grounded or floating state that the robotic arm end effector must maintain in different task stages, and specifying the order in which the robotic arm end effector contacts different potential bodies.

[0034] Step 3.4: Convert the above sequence constraints into a penalty term in the cost function of the path search process. That is, when the path search algorithm attempts to generate a continuous path, if the state sequence of the end effector corresponding to the path violates the temporal logic rules, a very large penalty value is added to the cost function of the path, so that the path is excluded in the subsequent optimization process.

[0035] The specific method for step 4 is as follows:

[0036] Step 4.1: Initialize the search tree T, and set the configuration points corresponding to the current pose of the robotic arm. As the root node of the search tree, the configuration point contains the angle information of each joint of the robotic arm and the coordinate information in the composite configuration space C_space_E;

[0037] Step 4.2: Randomly generate a target bias sampling point in the composite configuration space C_space_E. When generating, the target point of the maintenance task is directly set with a preset probability p. as sampling points The remaining probabilities are randomly sampled throughout the space;

[0038] Step 4.3: Traverse the search tree T to find the distance from the sampling point. Recent existing nodes ;

[0039] Step 4.4: From the nearest node To sampling point The direction is expanded with a fixed step size to generate a new node. During the expansion process, the forward kinematics model of the robotic arm is called to calculate the volume occupied by each link of the robotic arm in space;

[0040] Step 4.5: For the newly generated node Conduct safety testing;

[0041] Step 4.6: If the new node If it passes all security checks, it is added to the search tree T, and its origin is recorded. arrive The cost of moving and from arrive The cumulative risk potential value;

[0042] Step 4.7: Determine the new node Has the target point been reached? Within its neighborhood:

[0043] If so, the path search is successful. Backtrack from the target point to the root node to generate an initial planned path.

[0044] If not, return to step 4.2 and continue iterating.

[0045] The security checks performed in step 4.5 specifically include:

[0046] First, perform geometric collision detection:

[0047] Position the robotic arm in the configuration The volume occupied by each link is compared with the geometric obstacles in the composite configuration space C_space_E. If geometric penetration occurs, the detection is deemed unsuccessful and the new node is discarded. ;

[0048] Secondly, perform electric field rigidity prohibition detection:

[0049] Query new node The electric field intensity amplitude E at each point on the surface of the robotic arm in the composite configuration space C_space_E is defined as follows: if the electric field intensity amplitude E at any point on the robotic arm is greater than or equal to the maximum allowable electric field intensity threshold... If the test fails, the new node is discarded. ;

[0050] Next, perform cumulative risk potential field value detection:

[0051] Calculate from the root node To the current new node The cumulative sum of the risk potential values ​​R of all nodes along the entire path; if this cumulative sum exceeds the preset total path risk threshold... If the path is deemed too risky, the new node is discarded. ;

[0052] Finally, perform a timing logic consistency check:

[0053] New node After adding a node to the current path, update the state sequence of the robotic arm's end effector on the path. Compare this sequence with the standard electrical safety operation timing logic rules. If the state sequence violates the specified contact sequence or state transition rules, the detection is deemed failed, and the new node is discarded. .

[0054] In the cumulative risk potential value detection in step 4.5, starting from the root node To the current new node The method for calculating the cumulative sum of the risk potential values ​​R of all nodes along the entire path is as follows:

[0055] First query the new node The risk potential field values ​​at multiple key points on the robotic arm are used to determine the maximum value of these values, which is then taken as the new node. Node risk value Then From the root node To the nearest node Cumulative risk value Add them together to get a new cumulative risk value. The total risk threshold of the path It is a fixed value preset based on the complexity of the task and the insulation safety level.

[0056] The specific method for step 5 is as follows:

[0057] An interpolation algorithm is used to insert intermediate points between adjacent nodes of the initial planned path generated in step 4, so that the motion trajectory of each joint of the robotic arm is continuous. The interpolation algorithm is a fifth-order polynomial interpolation algorithm.

[0058] Using the conjugate gradient optimization algorithm, under the premise of satisfying all safety constraints in step 4, the node positions on the path are locally fine-tuned with the objectives of minimizing the total path length and minimizing the cumulative risk potential field value of the path, so as to generate the final executable maintenance path.

[0059] The specific method for triggering dynamic replanning of the path in step 6 is as follows:

[0060] During the movement of the robotic arm along the final executable maintenance path, the contact force between the robotic arm and the environment and the discharge pulse current are monitored in real time by the six-dimensional force sensor and current sensor installed on the wrist of the robotic arm.

[0061] If a sudden change in contact force or a discharge pulse signal in a characteristic frequency band is detected, it is determined that a sudden change has occurred in the current environment or electric field distribution. The movement of the robotic arm is immediately paused, and the process returns to step 1 to re-collect environmental data and trigger a replanning process until the end effector of the robotic arm safely reaches the target work point and completes the maintenance operation.

[0062] The specific method for detecting the discharge pulse signal in the characteristic frequency band in step 6 is as follows:

[0063] This is achieved by using a high-frequency current transformer connected to the grounding wire of the robotic arm. The high-frequency current transformer is used to collect discharge pulse current in the 30 MHz to 100 MHz frequency band. When the amplitude of the monitored discharge pulse current exceeds the preset threshold of 5 mA, it is determined that a partial discharge has occurred, triggering dynamic replanning of the path.

[0064] The present invention has the following advantages over the prior art:

[0065] I. This invention breaks through the limitations of traditional path planning that only considers geometric obstacle avoidance. By establishing a four-dimensional composite configuration space that includes the amplitude of electric field intensity, the high-voltage electromagnetic field is directly embedded into the planning kernel as an insurmountable hard constraint. When searching for a path, the robotic arm simultaneously evaluates the electric field risk at each step, ensuring that the planned path is not only geometrically collision-free, but also meets the electrical insulation safety requirements. This eliminates the risk of gap discharge caused by the robotic arm entering a high field strength region from the algorithmic root, achieving a leapfrog improvement in safety.

[0066] Second, this invention transforms the abstract electric field distribution into a quantifiable path cost by constructing a risk potential field function based on the electric field strength amplitude and distance. At the same time, it transforms the electrical safety operation sequence logic rules of live-line work into the cost function penalty term in the path search. This design enables the planning algorithm to autonomously weigh path length and electromagnetic risk, and strictly adhere to operating procedures such as equipotential before contact, thus solving the core problems of the separation of perception and decision-making and the disconnect between planning and electrical logic in traditional methods.

[0067] Third, this invention introduces a dynamic replanning mechanism based on discharge pulse current monitoring during path tracking. By setting a high-frequency current transformer on the grounding wire of the robotic arm, characteristic discharge signals from 30 MHz to 100 MHz are captured in real time. When environmental changes (such as wire swing) cause electric field distortion and trigger discharge, the system can immediately sense, pause and trigger replanning. This closed-loop architecture enables the robotic arm to have the ability to adapt to complex dynamic electromagnetic environments, greatly improving the reliability and fault tolerance of live-line operations. Attached Figure Description

[0068] The present invention will be further described below with reference to the accompanying drawings:

[0069] Fig. 1 This is a flowchart of the steps of the robotic arm path planning and maintenance method for uninterrupted power supply operations in power distribution networks according to the present invention.

[0070] Fig. 2 This is a flowchart of the step 3 of the present invention, which introduces timing logic constraints for live-line work.

[0071] Fig. 3 This is a flowchart of the path smoothing and post-processing steps in step 5 of the present invention. Detailed Implementation

[0072] like Figs. 1 to 3 As shown, this invention provides a robotic arm path planning and maintenance method for live-line work, mainly including: establishing a four-dimensional composite configuration space containing the electric field intensity amplitude of the work space; marking rigid prohibition points according to the maximum allowable electric field intensity threshold, and constructing a risk potential field layer by combining the electric field amplitude and distance risk potential field function; parsing the task type and retrieving the electrical safety operation timing logic rules, converting them into path search cost penalty terms; performing fusion path search based on an improved fast random expansion tree algorithm, performing four checks on the generated new nodes: geometric collision, electric field rigid prohibition, accumulated risk potential field, and timing logic consistency, adding nodes only when all checks pass; smoothly optimizing the initial path to generate the final path; and monitoring the discharge pulse signal in real time during path tracking, triggering replanning if a sudden change occurs. This invention embeds electromagnetic situational awareness into the planning kernel, achieving dual obstacle avoidance of geometry and electricity, significantly improving the safety of live-line work.

[0073] Furthermore, in order to realize the path planning of the robotic arm for live-line operation of power distribution network, the present invention specifically selects a dual-arm live-line operation robot as the path planning target. The robot is installed in the working bucket of the insulated bucket truck. Both robotic arms are 6-degree-of-freedom redundant robotic arms, and the ends are equipped with wire grippers and hydraulic crimping pliers, respectively. The robot control system includes a sensing unit, an electromagnetic field calculation unit, a path planning unit, and a safety monitoring unit.

[0074] Based on the above-mentioned live-line working robot, the method for path planning and maintenance of the robotic arm includes:

[0075] Step 1: Establish a composite configuration space that includes the electrical properties of the workspace. The composite configuration space is a four-dimensional space formed by calculating the electric field intensity amplitude at discrete spatial points and attaching the electric field intensity amplitude as a fourth dimension to the three-dimensional coordinate points based on the three-dimensional geometric model.

[0076] Step 2: In the composite configuration space, rigid restricted access points are marked according to the maximum allowable electric field strength threshold. For non-restricted access points, the risk potential field value of each non-restricted access point is calculated by combining the electric field strength amplitude with the risk potential field function of the distance to the nearest charged body surface, and a risk potential field layer is constructed.

[0077] Step 3: Analyze the task type of the maintenance task, retrieve the corresponding standard electrical safety operation timing logic rules according to the task type, and convert the timing logic rules into a cost function penalty term in the path search process.

[0078] Step 4: Perform fusion path search based on the improved fast random expansion tree algorithm. When generating new nodes, perform at least four safety checks on the new nodes, including geometric collision detection, electric field rigidity prohibition detection, cumulative risk potential field value detection, and temporal logic consistency detection. Only when all four checks are passed will the new node be added to the search tree.

[0079] Step 5: Smooth and post-process the initial planned path generated in Step 4 to generate the final executable maintenance path.

[0080] Step 6: During the movement of the robotic arm according to the final executable maintenance path, monitor the status of the robotic arm and the environment in real time. If it is determined that there is a sudden change in the environment or electric field distribution, pause the movement of the robotic arm and return to step 1 to trigger replanning.

[0081] The specific method for establishing the composite configuration space including the electrical attributes of the workspace in step 1 is as follows:

[0082] Step 1.1: Environmental Sensing and Data Acquisition

[0083] After the boom lift is launched and stabilized, the robot control system activates the perception system. The perception system includes a 16-line LiDAR mounted above the center of the two robotic arm bases, and two industrial-grade binocular vision cameras mounted on the elbows of the left and right robotic arms, respectively. The LiDAR scans the work area at a frequency of 10 Hz to acquire raw point cloud data of the surrounding environment, with a point cloud density better than 5 millimeters at a distance of 10 meters. Simultaneously, the two binocular vision cameras synchronously acquire high-resolution images with a resolution of 1920×1080 and a frame rate of 30 frames per second.

[0084] Step 1.2: Identify charged and uncharged bodies:

[0085] The control system fuses and registers the collected point cloud data with 2D image data. First, a pre-trained deep learning model is used to detect targets in the images, identifying key components such as conductors, insulators, and tower crossarms. Then, the image recognition results are registered with the point cloud data, and semantic labels are assigned to the corresponding point clouds. Next, the control system retrieves prior information stored in its local database, such as typical tower models (e.g., 15-meter poles), conductor models (LGJ-120 / 20), and insulator string lengths (e.g., 6 standard insulators, approximately 0.8 meters in total length), and matches and verifies this information with the perceived data. Ultimately, this accurately identifies live components (including the three-phase main line and the lead wire to be connected) and non-live components (towers, crossarms, and insulators) in the work scenario. In this example, the line voltage level is confirmed to be 10 kV.

[0086] Step 1.3: Constructing the preliminary geometric model:

[0087] Based on the above identification results, the control system establishes a three-dimensional geometric model with the center of the boom truck base as the origin O. The model is stored using an octree data structure, and each voxel is 2 cm × 2 cm × 2 cm in size. Each voxel in the model contains geometric attributes (whether it is occupied by a geometric entity) and preliminary electrical attributes (whether it is located on the surface of a charged body; if so, the voltage level is marked as 10 kV). This model is denoted as the preliminary geometric model.

[0088] Step 1.4: Calculate the electric field distribution:

[0089] The control system initiates the electromagnetic field solving module. Based on the preliminary geometric model, the boundary element method is used to calculate the electric field distribution in the workspace. The calculation boundary conditions are set as follows: a potential is applied to the surfaces of all charged bodies marked with 10 kV. (Relative to ground voltage), the tower, the boom arm of the boom truck, and the ground are set to zero potential. The solution domain is a spherical space with a radius of 3.5 meters centered at the working center of the robotic arm, and the discretized mesh resolution is 2 cm × 2 cm × 2 cm. The solver traverses all mesh points and outputs the electric field strength amplitude E at each discrete point, in volts per meter (V / m). For example, at a mesh point 0.1 meters from the surface of a charged conductor, the calculated electric field strength amplitude is... .

[0090] Step 1.5: Forming a four-dimensional composite configuration space:

[0091] The control system uses the electric field amplitude as a fourth dimension, appending it to the three-dimensional coordinate points to form a four-dimensional array. The array has a dimension of 350×350×350 grid points. For grid points occupied by geometric entities, the electric field strength amplitude E is set to infinity to indicate that the location is unreachable. For free space grid points, E is the calculated value. At this point, the composite configuration space is complete. Each discrete point in this space indicates not only whether the location is occupied by a geometric entity, but also the scalar value of the electric field strength at that location.

[0092] The specific method for constructing the risk potential field based on the electric field state in step 2 is as follows:

[0093] Step 2.1: Set the electric field safety threshold:

[0094] Based on the requirements for the limit of electric field exposure for the human body in the "Management Specification for Live Working on 10kV Overhead Distribution Lines", and considering the insulation characteristics of the metal parts of the robotic arm and the air gap, this invention sets a strict maximum allowable electric field strength threshold. This threshold ensures that there is a safety margin of at least 5 centimeters between any metal part of the robotic arm and a charged body (corresponding to an air breakdown field strength of approximately 30 kilovolts per centimeter in a uniform electric field).

[0095] Step 2.2: Traverse the spatial points and calculate the risk potential field:

[0096] The control system traverses the complex configuration space All grid points. For each grid point, first determine whether its electric field amplitude E is greater than or equal to... If so, mark the grid point as a rigid no-access point, and no part of the robotic arm may enter this point during subsequent path planning.

[0097] like Then, the risk potential value R of the grid point is calculated. First, the Euclidean distance d from the grid point to the nearest charged body surface is calculated. The control system quickly calculates the minimum distance from each grid point to all charged body surfaces through a distance transformation algorithm.

[0098] Then, substitute the risk potential function:

[0099] ;

[0100] In this method, an electric field intensity weighting coefficient is set. Distance weighting coefficient Singularity prevention constant The first term of this function reflects the risk proportion of the electric field strength itself, and the second term reflects the risk of proximity to a charged body (the closer the distance, the higher the risk).

[0101] For example, a grid point is located near a wire. , Then the calculation yields:

[0102] ;

[0103] If another grid point is far from the conductor but has a high field distortion, a high R value will also be obtained. All calculated R values ​​constitute a risk potential field layer with the same dimension as the composite configuration space.

[0104] The specific method for introducing live-line working timing logic constraints in step 3 is as follows:

[0105] Step 3.1: Parse the task type:

[0106] The task type in this example is "live-line connection". The control system parses the specific type of the current maintenance task from the local task rule base. The task type includes live-line connection work, live-line disconnection work, or live-line insulator replacement work.

[0107] Step 3.2: Retrieve the timing logic rules:

[0108] Based on the task type, the control system retrieves the corresponding standard electrical safety operation timing logic rules from the database. Taking live wire connection as an example, this rule is described using a finite state machine:

[0109] Initial state: The end effectors of both robotic arms (the wire gripper and the hydraulic clamp) are reliably grounded through the grounding wire of the boom truck, and the potential is zero.

[0110] Rule 1: Before the right arm hydraulic clamps strip and crimp the main cable, ensure that the jaws of the right arm hydraulic clamps are equipotentially connected to the lead wire clamp via a dedicated conductive slip ring. This process should be performed after the left arm cable gripper has gripped the lead wire and stabilized it.

[0111] Rule 2: During the entire crimping process, it is strictly forbidden for two robotic arms to simultaneously contact conductors with different potentials (for example, the left arm wire gripper contacts the lead wire (energized), and the right arm hydraulic clamp contacts the main wire (energized). Since both have the same potential, 10 kV, this is allowed; however, if the left arm contacts the grounding body and the right arm contacts the energized body, a step voltage path is formed, which is absolutely prohibited).

[0112] Rule 3: The left arm cable gripper must not release the drainage line before the drainage line is secured to the main line.

[0113] Step 3.3: Formalize into sequence constraints:

[0114] The control system formalizes the aforementioned timing logic rules into sequential constraints on the state of the robotic arm's end effector. Specifically, this includes: specifying whether the end effector must remain in a grounded or suspended state during different task phases, and specifying the order in which the end effector contacts different potential bodies.

[0115] Step 3.4: Convert to a cost function penalty term:

[0116] These sequence constraints are converted into a penalty term in the cost function of the path search process. During the path search, a virtual state variable is maintained to record whether the current end effector is in a "grounded" or "equipotential" state. When the path search algorithm attempts to generate a continuous path, if the state sequence of the end effector corresponding to that path violates the aforementioned timing logic rules, a very large penalty value (e.g., ...) is added to the cost function of that path. This path was excluded during subsequent optimization.

[0117] The specific method for performing the fusion path search based on the improved fast random expanded tree algorithm in step 4 is as follows:

[0118] Step 4.1: Initialize the search tree:

[0119] Taking the right arm (performing the pressing task) as an example, plan the path from the current initial pose to the target work point (mainline peeling position). This involves creating a vector from the angle values ​​of the current 6 joints of the right arm. As the root node of the search tree, the configuration point contains the angle information of each joint of the robotic arm and the information within the composite configuration space. The coordinate information in the middle (can be calculated through forward kinematics).

[0120] Step 4.2: Generate target bias sampling points:

[0121] Set the target bias probability p = 0.1. Generate a random number in each iteration. If rand < 0.1, then directly set the target configuration. (The joint angle at the point of skinning, calculated using inverse kinematics) is used as the sampling point. Otherwise, a 6-dimensional vector is randomly generated within the joint space as... .

[0122] Step 4.3: Find the nearest node:

[0123] In the search tree, calculate the relationship between all existing nodes and... Find the node with the smallest Euclidean distance in joint space. .

[0124] Step 4.4: Expand and generate new nodes:

[0125] from Towards Directional interpolation with a step size of 0.02 radians generates a new joint space vector. Using the forward kinematics model of the robotic arm, the kinematics of the joint angle is calculated. At that time, the approximate outer envelope of the right upper arm, forearm, wrist, and distal gripper in Cartesian space (simplified using a sphere or capsule).

[0126] Step 4.5: Quadruple security checks:

[0127] For new nodes Perform the following four tests:

[0128] Geometric collision detection: Interference checks are performed between the outer envelope and geometric obstacles (towers, insulators, other arms, etc.) in the composite configuration space. Specifically, the outer envelope is discretized into several key points, and it is checked whether the voxels of these key points in the composite configuration space are occupied by geometric entities. If geometric penetration occurs (i.e., any key point is located within a voxel occupied by a geometric entity), the detection is deemed unsuccessful, and the new node is discarded. Then return to step 4.2.

[0129] Electric field rigidity prohibition detection: Take the point on the outer envelope closest to the charged body (determined based on the distance field d calculated in step 2), and query the electric field amplitude E of that point in the composite configuration space. If the electric field amplitude E at any point on the robotic arm is greater than or equal to the maximum allowable electric field intensity threshold... If the test fails, the new node is discarded. .

[0130] Cumulative risk potential value detection: First, calculate the risk value of the new node itself. Take the risk potential field values ​​of three key points on the outer envelope of the robotic arm (mid-arm, mid-arm, and end effector gripper). Take the maximum value as :

[0131] ;

[0132] Then get the root node To the nearest node Cumulative risk value (This value is recorded when a node is added). Calculate the cumulative risk value of the new path. Preset total risk threshold for the path. .like If the path is deemed too risky, the new node is discarded. .

[0133] Timing logic consistency check: New node After adding the node to the current path, update the state sequence of the robotic arm's end effector on the path. Compare this sequence with the standard electrical safety operation timing logic rules established in step 3. If the state sequence violates the prescribed contact sequence or state transition rules (e.g., attempting to contact a live conductor without completing equipotential bonding), the detection is deemed unsuccessful, and the new node is discarded. .

[0134] Step 4.6: Add a node:

[0135] If the new node If it passes all four security checks, it is added to the search tree T, and its origin is recorded. arrive The movement cost (joint-space Euclidean distance) and from arrive Cumulative risk value .

[0136] Step 4.7: Determine if the target has been reached:

[0137] Calculate new node With the target point The Euclidean distance in joint space. If it is less than a preset threshold of 0.1 radians, the path search is considered successful. From the target point... Backtrack to the root node This yields an initial planned path consisting of several configuration points. If the objective is not achieved, return to step 4.2 to continue iterating.

[0138] The specific method for path smoothing and post-processing in step 5 is as follows:

[0139] The initial planned path generated in step 4 is post-processed and optimized. First, a fifth-order polynomial interpolation algorithm is used to insert intermediate points between adjacent path points, ensuring that the motion trajectory of each joint is continuous in terms of position, velocity, and acceleration. For example, when joint 1 moves from 30 degrees to 35 degrees, three interpolation points are inserted to ensure a smooth transition in angular velocity and angular acceleration, guaranteeing stable movement of the robotic arm.

[0140] Then, the conjugate gradient optimization algorithm is used, with the nodes on the path as optimization variables, and local fine-tuning is performed with the goal of minimizing the comprehensive cost function. The comprehensive cost function is defined as:

[0141] ;

[0142] in The weights are the coefficients, and the total path length is the sum of the Euclidean distances in the joint space between all adjacent nodes. This represents the cumulative risk value for the entire path. The optimization process is conducted while satisfying all safety constraints in step 4, meaning that each new node after each iteration still needs to pass geometric collision detection, electric field rigidity prohibition detection, cumulative risk potential field value detection, and temporal logic consistency detection. After optimization, a smooth, short, and low-risk executable maintenance path is finally obtained.

[0143] The specific method for dynamic path replanning in step 6 is as follows:

[0144] As the robotic arm moves along the final executable maintenance path, the safety monitoring unit operates continuously. A six-dimensional force sensor is installed on the robotic arm's wrist to monitor the contact force between the robotic arm and the environment at a frequency of 1 kHz; simultaneously, a high-frequency current transformer is connected to the robotic arm's grounding wire to monitor the discharge pulse current in the 30 MHz to 100 MHz frequency band, with a threshold set at 5 mA.

[0145] If a sudden change in contact force is detected (e.g., the force value exceeds a preset rate of change of 5 Newtons) or a discharge pulse signal in a characteristic frequency band is detected (discharge pulse current amplitude exceeds 5 mA), it is determined that a sudden change has occurred in the current environment or electric field distribution. The robotic arm movement is immediately paused, and the process returns to step 1 to re-acquire environmental data and trigger a replanning process. During replanning, the perception system re-acquires the point cloud and image of the current scene and updates the composite configuration space. And the risk potential field layer, then use the current hover configuration as the new root node. Repeat steps 2 through 5 until the end effector of the robotic arm safely reaches the target work point and completes the maintenance operation.

[0146] Based on the above steps, this invention enables the planning and maintenance of robotic arm paths by deeply integrating electromagnetic situational awareness and physical space planning during live-line work in power distribution networks. This method ensures that the robotic arm not only avoids geometric obstacles when performing tasks, but also actively avoids high field strength areas and strictly follows electrical safety operating procedures, greatly improving the safety and reliability of live-line work.

[0147] In an embodiment of the present invention, a dual-arm live-line working robot is used. This robot is installed in the working bucket of an insulated bucket truck. Both arms are 6-DOF redundant arms, and their ends are equipped with a wire gripper and a hydraulic crimping clamp, respectively. The robot control system includes a sensing unit, an electromagnetic field calculation unit, a path planning unit, and a safety monitoring unit. The embodiment executes the following planned maintenance steps:

[0148] Step 1: After the boom truck is lifted into the air and stabilizes, the robot control system first executes the creation of a composite configuration space containing the electrical attributes of the workspace, including:

[0149] 1.1 Environmental Perception and Modeling:

[0150] The robot's perception system consists of a 16-line LiDAR mounted above the center of the two robotic arm bases, and two industrial-grade binocular vision cameras mounted on the elbows of the left and right robotic arms, respectively. The LiDAR scans the work area at a frequency of 10 Hz to acquire raw point cloud data of the surrounding environment, with a point cloud density better than 5 millimeters at a distance of 10 meters. Simultaneously, the two binocular vision cameras acquire high-resolution images (1920×1080 resolution, 30 frames per second).

[0151] The control system fuses point cloud data with image data. First, it performs target detection on the images, using a pre-trained deep learning model to identify key components such as conductors, insulators, and tower crossarms. Then, it registers the image recognition results with the point cloud data, assigning semantic labels to the corresponding point clouds. Next, the control system retrieves prior information stored in its local database, such as typical tower models (e.g., 15-meter poles), conductor models (LGJ-120 / 20), and insulator string lengths (e.g., 6 standard insulators, approximately 0.8 meters in total length), and matches and verifies this information with the perceived data. Ultimately, it accurately identifies live components (including the three-phase main line and the lead wire to be connected) and non-live components (towers, crossarms, and insulators) in the work scenario. In this example, the line voltage level is confirmed to be 10 kV.

[0152] Based on the above information, the control system establishes a three-dimensional geometric model with the center of the boom truck base as the origin O. The model is stored using an octree data structure, and each voxel is 2 cm × 2 cm × 2 cm in size. Each voxel in the model contains geometric attributes (whether it is occupied by a geometric entity) and electrical attributes (whether the point is located on the surface of a charged body; if so, the voltage level is marked as 10 kV).

[0153] 1.2 Calculation of electric field distribution:

[0154] The control system initiates the electromagnetic field solving module. Based on the established geometric model, the boundary element method is used to calculate the electric field distribution in the workspace. The calculation boundary conditions are set as follows: a potential is applied to the surfaces of all charged bodies marked with 10 kV. (Relative to ground voltage), the tower, the boom arm of the boom truck, and the ground are set to zero potential. The solution domain is a spherical space with a radius of 3.5 meters centered at the working center of the robotic arm, and the discretized mesh resolution is 2 cm × 2 cm × 2 cm.

[0155] The solver iterates through all grid points and outputs the electric field amplitude E at each discrete point, in volts per meter (V / m). For example, at a grid point 0.1 meters from the surface of a charged conductor, the calculated electric field amplitude is... (i.e., 12 kV per centimeter).

[0156] 1.3 Formation of a four-dimensional composite configuration space:

[0157] The control system uses the electric field amplitude as a fourth dimension, appending it to the three-dimensional coordinate points to form a four-dimensional array. The array has a dimension of 350×350×350 grid points. For grid points occupied by geometric entities, the electric field strength amplitude E is set to infinity to represent inaccessibility. For free space grid points, E is the calculated value. At this point, the composite configuration space is complete, which describes not only the location of the object but also the strength of the electric field.

[0158] Step 2: Construct a risk potential field based on the electric field state, including:

[0159] 2.1 Set the electric field safety threshold:

[0160] Based on the requirements for the limit of electric field exposure for the human body in the "Management Specification for Live Working on 10kV Overhead Distribution Lines", and considering the insulation characteristics of the metal parts of the robotic arm and the air gap, this embodiment sets a strict maximum allowable electric field strength threshold. This threshold ensures that there is a safety margin of at least 5 centimeters between any metal part of the robotic arm and a charged body (corresponding to an air breakdown field strength of approximately 30 kilovolts per centimeter in a uniform electric field).

[0161] 2.2 Traverse spatial points and calculate the risk potential field:

[0162] The control system traverses the complex configuration space All grid points. For each grid point, first determine whether its electric field amplitude E is greater than or equal to... If so, mark the grid point as a "rigidly prohibited access point," and no part of the robotic arm may enter this point in subsequent path planning.

[0163] like Then, the risk potential value R of the grid point is calculated. First, the Euclidean distance d from the grid point to the nearest charged body surface is calculated. The control system quickly calculates the minimum distance from each grid point to all charged body surfaces through a distance transformation algorithm.

[0164] Then, substitute the risk potential function:

[0165] ;

[0166] In this embodiment, an electric field strength weighting coefficient is set. Distance weighting coefficient Singularity prevention constant The first term of this function reflects the risk proportion of the electric field strength itself, and the second term reflects the risk of proximity to a charged object (the closer the distance, the higher the risk). For example, if a grid point is located near a wire, , Then the calculation yields:

[0167] ;

[0168] If another grid point is far from the conductor but has a high field distortion (e.g., due to the robot arm's own induction), it will also obtain a high R value. All the calculated R values ​​constitute a risk potential field layer with the same dimension as the composite configuration space.

[0169] Step 3: Introduce live-line working timing logic constraints, including:

[0170] The task type in this example is live wire connection. The control system retrieves the standard electrical safety operation timing logic rule corresponding to this task from the local task rule base. This rule is described by a finite state machine:

[0171] Initial state: The end effectors of both robotic arms (the wire gripper and the hydraulic clamp) are reliably grounded through the grounding wire of the boom truck, and the potential is zero.

[0172] Rule 1: Before the right arm hydraulic clamps strip and crimp the main cable, ensure that the jaws of the right arm hydraulic clamps are equipotentially connected to the lead wire clamp via a dedicated conductive slip ring. This process should be performed after the left arm cable gripper has gripped the lead wire and stabilized it.

[0173] Rule 2: During the entire crimping process, it is strictly forbidden for two robotic arms to simultaneously contact conductors with different potentials (for example, the left arm wire gripper contacts the lead wire (energized), and the right arm hydraulic clamp contacts the main wire (energized), but both have the same potential, 10 kV, so this is allowed; however, if the left arm contacts the grounding body and the right arm contacts the energized body, a step voltage path is formed, which is absolutely prohibited).

[0174] Rule 3: The left arm cable gripper must not release the drainage line before the drainage line is secured to the main line.

[0175] The control system formalizes these rules as a sequence of constraints on the state of the robotic arm's end effector. Specifically, during path search, a virtual state variable is maintained to record whether the end effector is currently in a "grounded" or "equipotential" state. Any path sequence that leads to an illegal state transition (e.g., directly changing from "grounded" to "contact with a live conductor" without going through the "equipotential connection" step) will be subject to a significant penalty (e.g., ...). Thus, they are naturally eliminated in the search.

[0176] Step 4: Perform fusion path search based on the improved fast random expanded tree algorithm, including:

[0177] 4.1 Initialize the search tree:

[0178] Taking the right arm (performing the pressing task) as an example, plan the path from the current initial pose to the target work point (mainline peeling position). This involves creating a vector from the angle values ​​of the current 6 joints of the right arm. As the root node of the search tree.

[0179] 4.2 Generate target bias sampling points:

[0180] Set the target bias probability p = 0.1. Generate a random number in each iteration. If rand < 0.1, then directly set the target configuration. (The joint angle at the point of skinning, calculated using inverse kinematics) is used as the sampling point. Otherwise, a 6-dimensional vector is randomly generated within the joint space as... .

[0181] 4.3 Finding the nearest node:

[0182] In the search tree, calculate the relationship between all existing nodes and... Find the node with the smallest Euclidean distance in joint space. .

[0183] 4.4 Extend and generate new nodes:

[0184] from Towards Directional interpolation with a step size of 0.02 radians (approximately 1.15 degrees) generates a new joint space vector. Using the forward kinematics model of the robotic arm, the kinematics of the joint angle is calculated. At that time, the approximate outer envelope of the right upper arm, forearm, wrist, and distal gripper in Cartesian space (simplified using a sphere or capsule).

[0185] 4.5 Quadruple security testing:

[0186] For new nodes Perform the following four tests:

[0187] Geometric collision detection: This checks for interference between the outer envelope and geometric obstacles (towers, insulators, other arms, etc.) in the complex configuration space. For example, if the center of a sphere in the outer envelope lies within a voxel occupied by a geometric entity, a collision is considered to have occurred, the detection fails, and the object is discarded. Then return to step 4.2.

[0188] Electric field rigidity prohibition detection: Take the point on the outer envelope closest to the charged body (determined based on the distance field d calculated in step 2), and query the electric field amplitude E of that point in the composite configuration space. If the point... This indicates that a certain part of the robotic arm has entered a dangerous field strength zone under this configuration, the test fails, and the robot is discarded. .

[0189] Cumulative risk potential value detection: First, calculate the risk value of the new node itself. Take the risk potential field values ​​of three key points on the outer envelope of the robotic arm (mid-arm, mid-arm, and end effector gripper). Take its maximum value as :

[0190] ;

[0191] Then calculate the distance from the root node to the nearest node. Cumulative risk value (The cumulative risk value of each node). The cumulative risk value of the new path. Preset total risk threshold for the path. .like This indicates that the cumulative risk of this path is too high, the test fails, and it is discarded. .

[0192] Timing logic consistency check: After adding to the current path, update the state sequence of the robotic arm's end effector on the path. For example, upon arrival... Previously, the end of the right arm was always in a grounded state; but The corresponding configuration requires the end gripper to contact the drain wire (potential rise). At this point, it's necessary to check if the state transition is valid: has the equipotential connection operation been completed (achieved through coordinated action of the left arm)? If the rule base specifies that equipotential connection must occur before contact, and the current state sequence indicates that the equipotential operation has not yet been performed, then it's considered a violation of timing logic, the check fails, and the device is discarded. .

[0193] 4.6 Adding nodes:

[0194] like If it passes all four of the above checks, it is added as a new node to the search tree T, and its existence is recorded from... arrive The movement cost (joint-space Euclidean distance) and from arrive Cumulative risk value .

[0195] 4.7 Determine if the target has been reached

[0196] calculate With the target point The Euclidean distance in joint space. If it is less than a preset threshold of 0.1 radians, the path search is considered successful. From Backtracking This yields an initial planning path consisting of several configuration points.

[0197] Step 5: Path smoothing and post-processing, including:

[0198] The initial path is post-processed. First, a fifth-order polynomial interpolation algorithm is used to insert intermediate points between adjacent path points to ensure that the motion trajectory of each joint is continuous in terms of position, velocity, and acceleration. For example, when joint 1 moves from 30 degrees to 35 degrees, three interpolation points are inserted to ensure a smooth transition in angular velocity and angular acceleration.

[0199] Then, the conjugate gradient optimization algorithm is used, with the nodes on the path as optimization variables, and local fine-tuning is performed with the goal of minimizing the comprehensive cost function. The comprehensive cost function is defined as:

[0200] ;

[0201] in The weighting coefficients are used. The optimization process is carried out under the premise of satisfying four-fold safety checks, meaning that new nodes after each iteration must still pass the above checks. After optimization, a smooth, short, and low-risk final executable maintenance path is obtained.

[0202] Step 6: Dynamic path replanning, including:

[0203] As the robotic arm moves along its final path, the safety monitoring unit operates continuously. A six-dimensional force sensor (sampling frequency 1 kHz) monitors the contact force, while a high-frequency current transformer is connected to the robotic arm's grounding wire to monitor the discharge pulse current in the 30 MHz to 100 MHz frequency band, with a threshold set at 5 mA.

[0204] As the robotic arm approaches the drain line, the distance between the drain line and the end of the robotic arm suddenly shortens due to the swaying of the drain line caused by the wind. The high-frequency current transformer detects a pulse current amplitude of 6 mA in the characteristic frequency band (around 50 MHz), exceeding the threshold of 5 mA. The control system determines that a partial discharge has occurred. At this moment, an emergency stop command is immediately issued, and the robotic arm comes to a stop.

[0205] Subsequently, the control system automatically returns to step 1, triggering a replanning process. The sensing system re-collects environmental data, updates the position of the guide wire, and recalculates the electric field distribution and risk potential field. Then, using the current hovering configuration as the new... Then, steps 2 through 5 are repeated to plan a new path that bypasses the dangerous discharge area. The robotic arm continues to move along the new path and eventually safely reaches the target work point to complete the main line stripping and crimping operations.

[0206] To verify the effectiveness of this invention, a path planning algorithm based solely on geometric collisions was used in the same 10 kV live-line connection operation scenario as a comparative example. The traditional RRT algorithm only performs geometric collision detection and does not include electric field risk fields or temporal logic constraints.

[0207] Each method was run 30 times, and the statistical results are shown in the table below:

[0208]

[0209] As shown in the table above, although the average planning time of the method of this invention is slightly longer than that of the traditional RRT method (due to the addition of complex electric field and timing detection), it has significant advantages in terms of safety: it eliminates electric field over-limit and timing logic errors, no discharge events occurred during actual execution, and the task success rate reached 100%. In contrast, the traditional RRT method, which only focuses on geometric obstacle avoidance, caused the robotic arm to repeatedly enter high-field areas during movement, triggering discharges and violating operating procedures, resulting in a task success rate of only 60%. This demonstrates that the present invention significantly improves the safety of live-line work on distribution networks by embedding electromagnetic situation and timing logic into the planning kernel.

[0210] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for path planning and maintenance of a robotic arm for uninterrupted power distribution network operations, characterized in that: Includes the following steps: Step 1: Establish a composite configuration space that includes the electrical properties of the workspace. The composite configuration space is a four-dimensional space formed by calculating the electric field intensity amplitude at discrete spatial points and attaching the electric field intensity amplitude as the fourth dimension data to the three-dimensional coordinate points based on the three-dimensional geometric model. Step 2: In the composite configuration space, rigid prohibited access points are marked according to the maximum allowable electric field strength threshold. For non-prohibited access points, the risk potential field value of each non-prohibited access point is calculated by combining the electric field strength amplitude with the risk potential field function of the distance to the nearest charged body surface, and a risk potential field layer is constructed. Step 3: Analyze the task type of the maintenance task, retrieve the corresponding standard electrical safety operation timing logic rules according to the task type, and convert the timing logic rules into a cost function penalty term in the path search process; Step 4: Perform fusion path search based on the improved fast random expansion tree algorithm. When generating new nodes, perform safety checks on the new nodes, including at least geometric collision detection, electric field rigidity prohibition detection, cumulative risk potential field value detection, and temporal logic consistency detection. Only when all checks are passed will the new node be added to the search tree. Step 5: Smooth and post-process the initial planned path generated in Step 4 to generate the final executable maintenance path; Step 6: Control the robotic arm to move according to the final executable maintenance path generated in Step 5. During the movement, monitor the status of the robotic arm and the environment in real time. If it is determined that there is a sudden change in the environment or electric field distribution, control the robotic arm to pause the movement and return to Step 1 to trigger the dynamic replanning of the path.

2. The method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 1, characterized in that: The specific method for step 1 is as follows: Step 1.1: Collect point cloud data and two-dimensional image data of the current working scene using a lidar and binocular vision camera installed on the robotic arm base or bucket truck; Step 1.2: Match the collected point cloud data and 2D image data with the pre-stored prior knowledge base of tower type, conductor type, and insulator string length to identify charged and non-charged bodies in the scene, and mark the voltage level of charged bodies; Step 1.3: Based on the marking results, construct a preliminary geometric model of the workspace that includes location coordinates, geometric shape, and voltage level information; Step 1.4: Based on the preliminary geometric model, calculate the electric field distribution in the work space using the finite element method or boundary element method, and solve for the electric field intensity amplitude at discrete spatial points with the robot arm base coordinate system as the reference. Step 1.5: Use the electric field intensity amplitude at all discrete points as the fourth dimension data and attach it to the three-dimensional coordinate points of the preliminary geometric model to form a four-dimensional composite configuration space C_space_E containing the spatial coordinate axis X, spatial coordinate axis Y, spatial coordinate axis Z and the electric field intensity amplitude E; In the composite configuration space C_space_E, each discrete point can indicate whether the location is occupied by a geometric entity and the scalar value of the electric field intensity at that location.

3. The method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 2, characterized in that: The specific method for step 2 is as follows: Step 2.1: Introduce the maximum permissible electric field strength threshold specified in the safety regulations for live-line working. ; Step 2.2: Traverse all discrete points in the composite configuration space C_space_E, and compare the electric field intensity amplitude E at each discrete point with the maximum allowable electric field intensity threshold. Compare; Step 2.3: If the electric field intensity amplitude E at a certain discrete point is greater than or equal to the maximum allowable electric field intensity threshold... If so, then mark the point as a rigidly forbidden access point; Step 2.4: If the electric field intensity amplitude E at a certain discrete point is less than the maximum allowable electric field intensity threshold... Then, based on the distance d between the point and the nearest charged body surface and the electric field intensity amplitude E at the point, the risk potential field value R at that point is calculated using the risk potential field function. The formula for calculating the risk potential field value R is as follows: ； in, This is the electric field strength weighting coefficient. This is the distance weighting coefficient. To prevent singular value constants where the denominator is zero; Step 2.5: Through the above calculations, assign a risk potential field value R between 0 and 1 to each non-forbidden access point, thereby constructing a risk potential field layer in the composite configuration space C_space_E that superimposes electric field intensity risk and geometric proximity risk.

4. The method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 3, characterized in that: The specific method for step 3 is as follows: Step 3.1: Analyze the task type of the current maintenance task. The task type includes live-line connection work, live-line disconnection work, or live-line insulator replacement work. Step 3.2: Based on the task type, retrieve the corresponding standard electrical safety operation sequence logic rules from the database; Step 3.3: Formalize the timing logic rules into sequence constraints on the state of the robotic arm end effector. The sequence constraints specifically include: specifying the grounded or floating state that the robotic arm end effector must maintain in different task stages, and specifying the order in which the robotic arm end effector contacts different potential bodies. Step 3.4: Convert the above sequence constraints into a penalty term in the cost function of the path search process. That is, when the path search algorithm attempts to generate a continuous path, if the state sequence of the end effector corresponding to the path violates the temporal logic rules, a very large penalty value is added to the cost function of the path, so that the path is excluded in the subsequent optimization process.

5. A method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 4, characterized in that: The specific method for step 4 is as follows: Step 4.1: Initialize the search tree T, and set the configuration points corresponding to the current pose of the robotic arm. As the root node of the search tree, the configuration point contains the angle information of each joint of the robotic arm and the coordinate information in the composite configuration space C_space_E; Step 4.2: Randomly generate a target bias sampling point in the composite configuration space C_space_E. When generating, the target point of the maintenance task is directly set with a preset probability p. as sampling points The remaining probabilities are randomly sampled throughout the space; Step 4.3: Traverse the search tree T to find the distance from the sampling point. Recent existing nodes ; Step 4.4: From the nearest node To sampling point The direction is expanded with a fixed step size to generate a new node. During the expansion process, the forward kinematics model of the robotic arm is called to calculate the volume occupied by each link of the robotic arm in space; Step 4.5: For the newly generated nodes Conduct safety testing; Step 4.6: If the new node If it passes all security checks, it is added to the search tree T, and its origin is recorded. arrive The cost of moving and from arrive The cumulative risk potential value; Step 4.7: Determine the new node Has the target point been reached? Within its neighborhood: If so, the path search is successful. Backtrack from the target point to the root node to generate an initial planned path. If not, return to step 4.2 and continue iterating.

6. The method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 5, characterized in that: The security checks performed in step 4.5 specifically include: First, perform geometric collision detection: Position the robotic arm in the configuration The volume occupied by each link is compared with the geometric obstacles in the composite configuration space C_space_E. If geometric penetration occurs, the detection is deemed unsuccessful and the new node is discarded. ; Secondly, perform electric field rigidity prohibition detection: Query new node The electric field intensity amplitude E at each point on the surface of the robotic arm in the composite configuration space C_space_E is defined as follows: if the electric field intensity amplitude E at any point on the robotic arm is greater than or equal to the maximum allowable electric field intensity threshold... If the test fails, the new node is discarded. ; Next, perform cumulative risk potential field value detection: Calculate from the root node To the current new node The cumulative sum of the risk potential values ​​R of all nodes along the entire path; if this cumulative sum exceeds the preset total path risk threshold... If the path is deemed too risky, the new node is discarded. ; Finally, perform a timing logic consistency check: New node After adding a node to the current path, update the state sequence of the robotic arm's end effector on the path. Compare this sequence with the standard electrical safety operation timing logic rules. If the state sequence violates the specified contact sequence or state transition rules, the detection is deemed failed, and the new node is discarded. .

7. A method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 6, characterized in that: In the cumulative risk potential value detection in step 4.5, starting from the root node To the current new node The method for calculating the cumulative sum of the risk potential values ​​R of all nodes along the entire path is as follows: First query the new node The risk potential field values ​​at multiple key points on the robotic arm are used to determine the maximum value of these values, which is then taken as the new node. Node risk value Then From the root node To the nearest node Cumulative risk value Add them together to get a new cumulative risk value. The total risk threshold of the path It is a fixed value preset based on the complexity of the task and the insulation safety level.

8. A method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 7, characterized in that: The specific method for step 5 is as follows: An interpolation algorithm is used to insert intermediate points between adjacent nodes of the initial planned path generated in step 4, so that the motion trajectory of each joint of the robotic arm is continuous. The interpolation algorithm is a fifth-order polynomial interpolation algorithm. Using the conjugate gradient optimization algorithm, under the premise of satisfying all safety constraints in step 4, the node positions on the path are locally fine-tuned with the objectives of minimizing the total path length and minimizing the cumulative risk potential field value of the path, so as to generate the final executable maintenance path.

9. A method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 8, characterized in that: The specific method for triggering dynamic replanning of the path in step 6 is as follows: During the movement of the robotic arm along the final executable maintenance path, the contact force between the robotic arm and the environment and the discharge pulse current are monitored in real time by the six-dimensional force sensor and current sensor installed on the wrist of the robotic arm. If a sudden change in contact force or a discharge pulse signal in a characteristic frequency band is detected, it is determined that a sudden change has occurred in the current environment or electric field distribution. The movement of the robotic arm is immediately paused, and the process returns to step 1 to re-collect environmental data and trigger a replanning process until the end effector of the robotic arm safely reaches the target work point and completes the maintenance operation.

10. A method for robotic arm path planning and maintenance for uninterrupted power distribution network operations according to claim 9, characterized in that: The specific method for detecting the discharge pulse signal in the characteristic frequency band in step 6 is as follows: This is achieved by using a high-frequency current transformer connected to the grounding wire of the robotic arm. The high-frequency current transformer is used to collect discharge pulse current in the 30 MHz to 100 MHz frequency band. When the amplitude of the monitored discharge pulse current exceeds the preset threshold of 5 mA, it is determined that a partial discharge has occurred, triggering dynamic replanning of the path.

Images