Path planning method and system for operation of transformer substation oil sample collection robot

Abstract

The invention relates to the field of path planning, in particular to a path planning method and system for operation of a transformer substation oil sample collection robot. The method comprises the following steps: the invention relates to the path planning method and system of the transformer substation oil sample collection robot. The method comprises the following steps: acquiring position information of a starting point and a target point, and constructing an elliptical state subset space; uniformly sampling in the space, generating random points and mapping the random points to an elliptical region; searching a nearest node and generating a new node; determining a neighborhood node set, selecting a father node with the minimum cost, and updating a path; constructing two random trees based on a reverse expansion strategy, and detecting an intersection condition to generate an initial feasible path; and dynamically adjusting ellipse parameters, and iteratively optimizing the path until a convergence condition is met. According to the method, the efficiency of path planning and the path quality are remarkably improved through the improved Inform-RRT algorithm in combination with the elliptic state subset sampling and bidirectional expansion strategy.

Description

Technical Field

[0001] The present invention relates to the field of path planning, and in particular to a path planning method and system for an oil sample collection robot operating in a transformer substation. Background Art

[0002] With the increasing intelligence of power systems, path planning technology for substation oil sampling robots has become a key component in ensuring efficient transformer operation and maintenance. Traditional path planning methods mainly include rail-mounted guidance, RRT (Rapidly Expanding Random Tree) algorithm, and RRT algorithm. Rail-mounted guidance requires the pre-laying of fixed rails. Although it can achieve basic navigation, it is costly and inflexible. It cannot adapt to the dynamic changes in substation equipment layout, especially in scenarios with limited space. The RRT algorithm explores paths through random sampling. Although it can adapt to complex environments, its blind sampling nature leads to inefficient path search. The generated paths are often redundant and uneven, making it difficult to meet the accuracy requirements of oil sampling. The RRT algorithm introduces a progressive optimization mechanism based on RRT, approximating the optimal path through multiple iterations. However, the sampling space is large, the number of nodes is large, and the convergence speed is slow, making it difficult to provide an efficient and feasible path for the robot within a limited time.

[0003] For example, patent CN117724496A discloses a path planning method for a power inspection robot based on RRT*. Although this method improves the efficiency of path planning to a certain extent, it still has the following shortcomings:

[0004] Slow path convergence: Although the RRT* algorithm approaches the optimal path through multiple iterations, its convergence speed is still slow in complex environments and it is difficult to meet real-time requirements.

[0005] Low sampling efficiency: The traditional RRT* algorithm performs random sampling in the entire space, resulting in a large number of invalid samples, especially near the target point, resulting in low sampling efficiency.

[0006] In response to the above problems, the present invention proposes a path planning method and system for an oil sample collection robot in a substation. Summary of the Invention

[0007] The present invention provides a path planning method and system for an oil sampling robot in a transformer substation, aiming to solve the problems of large blindness and slow convergence of traditional RRT algorithms.

[0008] To achieve the above objectives, the following technical solutions are adopted.

[0009] A path planning method for an oil sampling robot in a substation comprises the following steps:

[0010] S1: Obtain the position information of the robot's starting point and target point, and construct an elliptical state subset space based on the starting point and target point;

[0011] S2: Perform uniform sampling in the elliptical state subset space to generate a random point x rand , and maps the unit circle sampling points to the ellipse state subset space through the transformation matrix;

[0012] S3: Search for the same value as x in the random tree T rand The nearest node x nearest , and x nearest As the base point along x rand Direction generates a new node x new ;

[0013] S4: Determine x new The neighborhood node set X near and by traversing X near The node in selects the parent node x with the minimum cost min , change x new Insert into the random tree T and update the path connection relationship;

[0014] S5: Based on the reverse expansion strategy, two random trees are constructed with the starting point and the target point as the starting point of the bidirectional search. The node intersection condition is detected during the alternating expansion of the two trees. If there is an intersection, an initial feasible path is generated;

[0015] S6: Based on the cost C of the initial feasible path max The major axis parameters of the ellipse state subset are dynamically adjusted, and the optimization path is iteratively optimized until the preset convergence conditions are met.

[0016] Optionally, constructing the elliptical state subset space in step S1 includes:

[0017] Based on the theoretical shortest path cost C between the starting point and the target point min , calculate the major axis C of the ellipse best , and generate the transformation matrix L through Cholesky decomposition so that the elliptic state subset satisfies the equation:

[0018] (xx center ) T S(xx center )=1

[0019] Among them, S is the best and C min Determine the diagonal matrix, x center is the center of the ellipse, T is the transpose of the matrix, and x is the x-axis coordinate of the node.

[0020] Optionally, the mapping of the unit circle sampling points in step S2 includes:

[0021] For the unit circle n- ball Uniform sampling points x within ball Perform linear transformation using the formula:

[0022]

[0023] Where C is the rotation matrix solved by the Wahba problem, L is the translation matrix, X center is the center of the ellipse.

[0024] Optionally, the reverse expansion strategy in step S5 includes:

[0025] Sub-step S51: using the starting point and the target point as the root nodes of two random trees respectively, and selecting one of the trees for expansion in each iteration;

[0026] Sub-step S52: Select the nearest node from another random tree as the expansion direction, generate a new node and detect collision;

[0027] Sub-step S53: When the distance between the nodes in the two trees is less than a preset threshold, the paths are merged and an initial feasible path is output.

[0028] Optionally, the dynamic adjustment of the major axis parameters of the ellipse in step S6 includes: max Update the major axis C of the ellipse best , through the formula:

[0029] C best =α·r·C min

[0030] Among them, α is the expansion coefficient, and r is the dynamic adjustment factor based on the number of random tree nodes n.

[0031] Optionally, in step S4, the parent node x is selected min include:

[0032] Calculate the distance from the starting point to X near Midpoint xi to x new The path cost is calculated, and the node with the smallest total cost is selected as the parent node, and the original parent node connection is disconnected to optimize the path.

[0033] Optionally, the method further includes: a dynamic obstacle adaptation strategy for updating the elliptical state subset space according to real-time changes in the substation equipment layout; the dynamic obstacle adaptation strategy performs the following operations:

[0034] Acquire the location information of new or moving obstacles in the substation in real time and calibrate the three-dimensional coordinates of the obstacles through sensor fusion algorithms;

[0035] Based on the calibrated obstacle position, dynamically adjust the ellipse major axis parameter C best and the neighborhood radius r, so that the elliptical state subset space shrinks or expands adaptively with the obstacle distribution, where the major axis adjustment formula is:

[0036]

[0037] Among them, α is the dynamic attenuation coefficient, d obs is the minimum distance between the obstacle and the current path, d init is the initial path length;

[0038] Regenerate sampling points x based on the updated ellipse parameters rand The new path is then re-planned and the safety distance between it and the dynamic obstacle is verified by the collision detection unit. If a path conflict is detected, local path re-planning is triggered, and only the nodes in the conflict area are iteratively optimized, while the path structure in the non-conflict area is retained.

[0039] The local path replanning includes: deleting conflicting nodes and their sub-nodes, constructing temporary random tree branches in the conflicting area, and quickly generating a local optimized path through a bidirectional expansion strategy.

[0040] Optionally, the update condition of the ellipse state subset is:

[0041] When a newly generated node causes the path cost to decrease by more than a preset threshold, the ellipse parameters are recalculated and the sampling space is reduced.

[0042] A substation oil sample collection robot path planning system, comprising:

[0043] A sampling module, used to generate uniform sampling points in the elliptical state subset space;

[0044] The path generation module is used to build a random tree and generate new nodes, searching for the initial feasible path through the reverse expansion strategy;

[0045] Optimization module, used to dynamically adjust ellipse parameters and iteratively optimize path cost;

[0046] The control module is used to send the optimized path to the robot actuator to complete the oil sample collection task.

[0047] Optionally, the path generation module includes:

[0048] A bidirectional search unit for synchronously expanding the random tree from the starting point and the target point;

[0049] A collision detection unit, used to verify whether the new node collides with environmental obstacles;

[0050] The path merging unit is used to generate the final path when two random trees intersect.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention significantly improves the efficiency and quality of path planning by introducing an improved Informed-RRT algorithm, combining elliptical state subset sampling with a bidirectional expansion strategy. First, an elliptical state subset space is constructed based on the starting point and the target point, and a transformation matrix is ​​generated through Cholesky decomposition. The unit circle sampling points are mapped to the elliptical area, which reduces the invalid sampling range and solves the problems of large blindness and slow convergence of the traditional RRT algorithm. It enables the robot to quickly locate the approximate optimal path within a limited time. At the same time, a bidirectional expansion strategy is adopted to synchronously construct a random tree from the starting point and the end point. By alternating expansion and node intersection detection, the path search time is greatly shortened, which is particularly suitable for dynamic path planning in complex obstacle environments of substations. In addition, the major axis parameters of the ellipse are dynamically adjusted, and the search space is iteratively optimized in combination with path cost feedback, which further improves the adaptability of the algorithm to changes in the layout of substation equipment.

[0053] Through the aforementioned technical means, the present invention not only inherits the asymptotic optimality of the RRT* algorithm, but also optimizes path curvature by removing redundant nodes through path smoothing, making the robot's motion more efficient and stable. The system's modular design separates sampling, path generation, and optimization functions, combining collision detection with real-time parameter adjustment to ensure high accuracy and reliability in oil sampling tasks. These improvements enable the robot to flexibly avoid obstacles in the confined environment of substations while reducing hardware deployment costs, providing efficient technical support for intelligent power system operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The figure is a control flow diagram of an embodiment of a path planning method for an oil sampling robot in a substation according to the present invention.

[0055] Figure 2 This is a schematic diagram of a path planning system module for an oil sampling robot in a substation according to the present invention. DETAILED DESCRIPTION

[0056] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0057] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise indicated, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0058] Example 1

[0059] like Figure 1 As shown, the present invention provides a path planning method for a substation oil sampling robot. Its core lies in combining an improved Informed-RRT algorithm with a bidirectional expansion strategy. By dynamically adjusting the elliptical sampling space and optimizing the path convergence speed, this method addresses the problems of blind sampling, slow convergence, and uneven paths encountered by traditional RRT algorithms in complex substation environments. The following details the algorithm's core process, mathematical implementation details, dynamic parameter adjustment strategies, and system modular design.

[0060] Construction and Sampling Implementation of Elliptic State Subset Space

[0061] When the robot is initialized, it is necessary to obtain the position coordinates of the starting point and the target point (for example, determined by a laser radar or the internal coordinate system of the substation). Based on the theoretical shortest path cost C between the two points min (i.e. Euclidean distance), construct the ellipse state subset space. Ellipse major axis C best The initial value is the current optimal path cost C max , the minor axis is given by the formula calculate.

[0062] Compared with random sampling of the entire state space, based on the elliptical state subset x ellipse ~U(X ellipse ) can be sampled uniformly through the unit circle n- ball , x ball ~U(X ball ) are uniformly distributed samples through matrix transformation, rotation and translation.

[0063] x ellipse =Lx ball +x center

[0064] Among them, x ball represents the uniform sampling points in the unit circle, L represents the transformation matrix, is the center of the ellipse, and its two foci are xf1 and xf2.

[0065] Ellipse center x centerThe midpoint coordinates of the starting point and the target point are generated by Cholesky decomposition to generate the transformation matrix L. The transformation matrix is ​​calculated using the Cholesky decomposition hyperelliosoid matrix idea. T ≡S,S∈R n×n .

[0066] So that the ellipse equation (xx center ) T S(xx center )=1, S is the diagonal matrix determined by Cbest and Cmin, x center is the center of the ellipse, T is the transpose of the matrix, and x is the x-axis coordinate of the node.

[0067] Decompose to get the hyperelliosoid matrix:

[0068] diag{·} represents a diagonal matrix

[0069] Matrix decomposition is done through linear algebra libraries (such as Eigen). Therefore, the unit circle n- ball , x ball ~U(X ball ) are processed by the transformation matrix, rotation matrix, and translation to obtain uniform sampling points in the elliptical state subset:

[0070] Xf=CLx ball +x center .

[0071] Here, C is the rotation matrix C=U diag{1,…,1,det(U)det(V)}VT calculated by the Wahba problem, which is used to align the ellipse direction with the line connecting the starting point and the target point.

[0072]

[0073] This process limits the sampling range to the ellipse area, avoiding the inefficiency of traditional RRT* random sampling in the entire space. For example, in the scene with the starting point A(0,0) and the target point B(10,0), the initial major axis of the ellipse is 10 meters and the minor axis is 6 meters (assuming C max =10, C min =8), the sampling points are only distributed within the elliptical area, which significantly reduces invalid exploration.

[0074] Random tree construction and node optimization strategy

[0075] Generate a random point x within the ellipse rand Finally, the algorithm uses KD-Tree to accelerate the nearest neighbor search and find the nearest node x from the random tree T. nearest. New node x new Along x rand Direction generation, step length is subject to robot kinematic constraints (such as maximum steering angle 30°, step length upper limit 1 meter). new If there is no collision with environmental obstacles (verified by ray detection or bounding box collision detection), then its neighborhood node set X is determined near , the neighborhood radius r is dynamically adjusted to (N is the current number of nodes, d is the spatial dimension, and γ=1.5 is the adjustment coefficient). min Calculate the distance from the starting point to x i ∈X near to x new The total path cost (including turn penalty and path length) is selected, and the node with the minimum cost is disconnected through the Rewire operation. For example, if x new to x i The path cost of x is 20% lower than that of the original parent node, then update x i The parent node is x new , and delete the connecting edge of the original parent node.

[0076] Bidirectional expansion strategy and path fusion mechanism

[0077] To improve the search efficiency, the algorithm constructs two random trees T from the starting point and the target point simultaneously. start With T goal At each iteration, a tree is selected for expansion: x is generated in the direction of the nearest node of another tree. new , and detect whether the distance between the two tree nodes is less than the threshold = 0.5 meters. If the condition is met, the paths are merged by linear interpolation to generate the initial feasible path. For example, when T start Node n1 and T goal When the distance between nodes n2 is 0.3 meters, n1 and n2 are directly connected to form a complete path. Collision detection is accelerated using a bounding volume hierarchy (BVH) algorithm. Obstacles such as transformers and cable trenches are modeled as 3D geometric objects, ensuring a minimum safe distance between the path and obstacles of ≥ 0.3 meters. After the initial path is generated, redundant nodes (such as intermediate points between adjacent nodes that can be directly connected) are deleted and Bezier curve interpolation is used to optimize the path curvature, ensuring a smooth trajectory that complies with the maximum centrifugal force constraint.

[0078] Dynamic parameter adjustment and iterative optimization process

[0079] After the initial path is generated, according to its cost C max Dynamically adjust the major axis C of the ellipse best .

[0080] The formula is C best =λ·r

[0081]

[0082] Where λ is the expansion coefficient of the elliptical state subset. γ is a constant and n is the number of random tree nodes. max As the path is optimized, it is gradually reduced. This process is achieved through cyclic iteration until the path cost change rate is less than 0.1% or the maximum number of iterations (such as 1000) is reached. In order to accelerate convergence, a simulated annealing strategy is introduced: with probability P = e -ΔCost / T Accept temporary cost increases (T is a temperature parameter that decays with the number of iterations) to avoid falling into local optimality.

[0083] Example 2

[0084] The specific implementation of the dynamic obstacle adaptation strategy in the present invention includes the following steps:

[0085] The three-dimensional coordinate data of obstacles in the substation are collected in real time through LiDAR, visual SLAM (Simultaneous Localization and Mapping), and ultrasonic sensors. The ICP (Iterative Closest Point) algorithm is used to fuse and calibrate the multi-sensor data, including:

[0086] The LiDAR obtains obstacle point cloud data, removes noise points through the RANSAC (Random Sampling Consensus) algorithm, and generates the initial three-dimensional coordinates of the obstacle;

[0087] Visual SLAM locates obstacles through feature matching (such as ORB feature points) and extracts obstacle contour information in key frames;

[0088] The coordinate data of the lidar and visual SLAM are fused through the Kalman filter to output the calibrated obstacle center coordinates, and the positioning error is controlled within ±5cm.

[0089] Based on the calibrated obstacle position, dynamically update the ellipse major axis parameter C best And the neighborhood radius r, the specific formula is:

[0090]

[0091] Among them, d obs is the minimum Euclidean distance between the obstacle and the current path, d init is the initial path length, α is the dynamic attenuation coefficient, and its value range is [0.5, 1.5] [0.5, 1.5]. When d obs When d is less than 1m, α = 0.8 to reduce the elliptical sampling space to 4.8% of the original area; when d obs When the distance is greater than 3m, α=1.5 to expand the elliptical space to 120% of the original area. The neighborhood radius r is dynamically adjusted according to the number of nodes n, and the formula is:

[0092]

[0093] The adjusted elliptical state subset space is decomposed by Cholesky to update the transformation matrix L.

[0094] Make sure that the ellipse equation (xx center ) T S(xx center )=1 is valid.

[0095] When the collision detection unit detects that the safe distance between the path and the obstacle is less than 0.5m, local path replanning is triggered:

[0096] Conflict node processing: Delete the conflicting node and its child nodes, and build a temporary Sub-RRT* tree branch within a radius of 2 meters centered on the conflict point.

[0097] Bidirectional expansion strategy: Synchronously expand two random trees from the conflict point to the starting point and the end point, manage the node expansion order through the priority queue, and select the node closest to the conflict point to generate a new node x in each iteration new The two-way intersection condition (node ​​spacing ∈ < 0.3m) is used as the termination condition.

[0098] Path fusion: merge the temporary tree branches with the main path, smooth the path through the B-spline curve, and ensure that the robot steering angle θ≤30° and the maximum curvature κ max ≤0.2m -1 .

[0099] To reduce the computational overhead of global iteration, only the nodes less than 5% of the conflict area are optimized. The local sampling points x are optimized by GPU parallel computing (CUDA acceleration). rand Generation and collision detection are accelerated, and the time consumption of a single local replanning is controlled within 50ms to ensure real-time performance.

[0100] During the local path replanning process, non-critical tasks (such as data collection) are suspended to prioritize path safety. If multiple dynamic obstacles are detected at the same time, the obstacles are prioritized according to their distance (d obs The redundant path backup mechanism is used to ensure that at least one feasible path is available.

[0101] Example 3

[0102] like Figure 2 As shown in the figure, a path planning system for a substation oil sampling robot consists of four modules:

[0103] Sampling module: Generate uniform sampling points in the elliptical space through multi-threaded parallel computing, use CUDA to accelerate GPU computing, and can process 105 sampling points.

[0104] Path Generation Module: This module includes a bidirectional search unit, a collision detection unit, and a path merging unit. The bidirectional search unit uses a priority queue to manage the node expansion order. The collision detection unit implements real-time 3D collision detection based on OpenGL. The path merging unit synchronizes the status of the two trees using timestamps.

[0105] Optimization module: monitors path costs in real time, dynamically adjusts ellipse parameters, and calls path smoothing algorithms (such as B-spline curve fitting).

[0106] Control module: Converts the final path into robot motion instructions (Path messages in ROS), adjusts the wheel speed and steering angle through the PID controller, and drives the actuator to complete oil sampling.

[0107] In terms of hardware deployment, an embedded platform (such as the NVIDIA Jetson AGX Xavier) runs the core algorithm, communicating with the robot chassis via the CAN bus. The robotic arm's end effector synchronizes path coordinates via the EtherCAT protocol. For example, after reaching the sampling point, the robotic arm uses force feedback sensors to adjust its grip to ensure a stable grip on the oil sample bottle.

[0108] It is understood from common technical knowledge that the present invention may be implemented by other embodiments that do not depart from its spirit or essential features. Therefore, the embodiments disclosed above are, in all respects, merely illustrative and not exclusive. All modifications within the scope of the present invention or equivalent to the scope of the present invention are intended to be encompassed by the present invention.

Claims

1. A path planning method for an oil sampling robot in a substation, characterized in that: The following steps are involved: S1: Obtain the position information of the robot's starting point and target point, and construct an elliptical state subset space based on the starting point and target point; S2: Perform uniform sampling in the elliptical state subset space to generate a random point x rand , and maps the unit circle sampling points to the ellipse state subset space through the transformation matrix; S3: Search for the same value as x in the random tree T rand The nearest node x nearest , and x nearest As the base point along x rand Direction generates a new node x new ; S4: Determine x new The neighborhood node set X near and by traversing X near The node in selects the parent node x with the minimum cost min , change x new Insert into the random tree T and update the path connection relationship; S5: Based on the reverse expansion strategy, two random trees are constructed with the starting point and the target point as the starting point of the bidirectional search. The node intersection condition is detected during the alternating expansion of the two trees. If there is an intersection, an initial feasible path is generated; S6: Based on the cost C of the initial feasible path max The major axis parameters of the ellipse state subset are dynamically adjusted, and the optimization path is iteratively optimized until the preset convergence conditions are met.

2. A path planning method for a substation oil sampling robot according to claim 1, characterized in that: The step S1 of constructing the elliptical state subset space includes: Based on the theoretical shortest path cost C between the starting point and the target point min , calculate the major axis C of the ellipse best , and generate the transformation matrix L through Cholesky decomposition so that the elliptic state subset satisfies the equation: (x-x center ) T S(x-x center )=1 Among them, S is the best and C min Determine the diagonal matrix, x center is the center of the ellipse, T is the transpose of the matrix, and x is the x-axis coordinate of the node.

3. The path planning method for a substation oil sampling robot according to claim 1, characterized in that: The mapping of the unit circle sampling points in step S2 includes: For the unit circle n- ball Uniform sampling points x within ball Perform linear transformation using the formula: Where C is the rotation matrix solved by the Wahba problem, L is the translation matrix, and x center is the center of the ellipse.

4. The path planning method for a substation oil sampling robot according to claim 1, characterized in that: The reverse expansion strategy in step S5 includes: Sub-step S51: using the starting point and the target point as the root nodes of two random trees respectively, and selecting one of the trees for expansion in each iteration; Sub-step S52: Select the nearest node from another random tree as the expansion direction, generate a new node and detect collision; Sub-step S53: When the distance between the nodes in the two trees is less than a preset threshold, the paths are merged and an initial feasible path is output.

5. The path planning method for a substation oil sampling robot according to claim 1, characterized in that: The step S6 of dynamically adjusting the major axis parameters of the ellipse includes: max Update the major axis C of the ellipse best , through the formula: Cbest=λ·r Where λ is the expansion coefficient of the elliptical state subset, γ is a constant, n is the number of random tree nodes, and d is the spatial dimension.

6. The path planning method for a substation oil sampling robot according to claim 1, characterized in that: In step S4, the parent node x is selected min include: Calculate the distance from the starting point to X near Midpoint xi to x new The path cost is calculated, and the node with the smallest total cost is selected as the parent node, and the original parent node connection is disconnected to optimize the path.

7. The path planning method for a substation oil sampling robot according to claim 1, characterized in that: Also includes: Dynamic obstacle adaptation strategy, used to update the elliptical state subset space according to the real-time changes in substation equipment layout; The dynamic obstacle adaptation strategy performs the following operations: Acquire the location information of new or moving obstacles in the substation in real time and calibrate the three-dimensional coordinates of the obstacles through sensor fusion algorithms; Based on the calibrated obstacle positions, the ellipse major axis parameter Cbest and the neighborhood radius r are dynamically adjusted so that the ellipse state subset space can adaptively shrink or expand with the obstacle distribution. The major axis adjustment formula is: Among them, α is the dynamic attenuation coefficient, d obs is the minimum distance between the obstacle and the current path, d init is the initial path length; The sampling points xrand are regenerated based on the updated ellipse parameters, and the safe distance between the new path and dynamic obstacles is verified by the collision detection unit. If a path conflict is detected, local path replanning is triggered, and only the nodes in the conflicting area are iteratively optimized, while the path structure in the non-conflicting area is retained. The local path replanning includes: deleting conflicting nodes and their sub-nodes, constructing temporary random tree branches in the conflicting area, and quickly generating a local optimized path through a bidirectional expansion strategy.

8. The path planning method for a substation oil sampling robot according to claim 1, characterized in that: The update condition of the ellipse state subset is: When a newly generated node causes the path cost to decrease by more than a preset threshold, the ellipse parameters are recalculated and the sampling space is reduced.

9. A substation oil sample collection robot path planning system, based on a substation oil sample collection robot path planning method according to any one of claims 1 to 8, characterized in that: include, A sampling module, used to generate uniform sampling points in the elliptical state subset space; The path generation module is used to build a random tree and generate new nodes, searching for the initial feasible path through the reverse expansion strategy; Optimization module, used to dynamically adjust ellipse parameters and iteratively optimize path cost; The control module is used to send the optimized path to the robot actuator to complete the oil sample collection task.

10. A path planning system for oil sampling robot operation in a substation according to claim 9, characterized in that: The path generation module includes: A bidirectional search unit for synchronously expanding the random tree from the starting point and the target point; A collision detection unit, used to verify whether the new node collides with environmental obstacles; The path merging unit is used to generate the final path when two random trees intersect.

Images