Inspection robot path planning method and device based on neural network

Abstract

The invention provides an inspection robot path planning method and device based on a neural network, and relates to the technical field of path planning, and the method comprises the steps: searching nodes according to a starting point and a target point, and carrying out the node collision detection and node connection in the searching process, and obtaining an initial sampling path set; carrying out sampling probability calculation on nodes in the initial sampling path set, and carrying out iterative optimization through a neural network to obtain a sampling expansion scheme; performing node searching on the starting point and the target point based on the sampling expansion scheme, performing node collision detection in the searching process, when an obstacle is collided, constructing a sampling reconstruction model, performing asynchronous step length adjustment on the nodes obtained by searching according to the sampling expansion scheme according to substation layout information, and selecting a shortest path to obtain a feasible path; and carrying out smooth optimization on the feasible path to obtain a final inspection path. The problem of safety risks in operation and maintenance management of substation equipment is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular, to a path planning method and device for an inspection robot based on a neural network. Background Art

[0002] In the current technical field of path planning, for the regular inspection and detection work of substation equipment, although technology is advancing rapidly, the actual operation still mainly relies on manual methods. Inspectors need to carefully check each key equipment in the station one by one. This process is not only time-consuming and laborious, but also difficult to meet the high standards of modern power grid operation and maintenance management. Among them, the inspection cycle is relatively long, and it is difficult to achieve real-time supervision and management during manual inspection. In addition, there may be omissions and judgment errors during manual inspection, and abnormal situations that may occur during the operation of equipment are difficult to be discovered and processed in time, jointly leading to the problem of safety risks in the operation and maintenance management of substation equipment.

[0003] Therefore, there is an urgent need for a path planning method and device for an inspection robot based on a neural network to solve the problem of safety risks in the operation and maintenance management of substation equipment. Summary of the Invention

[0004] The purpose of the present invention is to provide a path planning method and device for an inspection robot based on a neural network to improve the above problems. To achieve the above purpose, the technical solutions adopted by the present invention are as follows: In the first aspect, the present application provides a path planning method for an inspection robot based on a neural network, including: Obtain substation layout information, where the substation layout information includes a starting point and a target point; Search for nodes according to the starting point and the target point, perform node collision detection during the search process, and connect the starting point, the target point, and the searched nodes to obtain an initial sampling path set; Calculate the sampling probability of the nodes in the initial sampling path set, combine the collision information and time cost information obtained from the collision detection, and iteratively optimize the nodes in the initial sampling path set through a preset neural network to obtain a sampling expansion scheme; Search for nodes for the starting point and the target point based on the sampling expansion scheme, perform node collision detection during the search process. Among them, when colliding with an obstacle, construct a sampling reconstruction model, adjust the nodes obtained by searching the sampling expansion scheme with different step sizes according to the substation layout information, and select the shortest path to obtain a feasible path; Perform smoothing optimization on the feasible path to obtain the final inspection path.

[0005] Second aspect, the present application also provides a path planning device for an inspection robot based on a neural network, including: An acquisition module, configured to acquire substation layout information, where the substation layout information includes a starting point and a target point; A first construction module, configured to search for nodes according to the starting point and the target point, perform node collision detection during the search process, and connect the starting point, the target point, and the searched nodes to obtain an initial sampling path set; A second construction module, configured to calculate sampling probabilities for the nodes in the initial sampling path set, combine the collision information and time cost information obtained from the collision detection, and iteratively optimize the nodes in the initial sampling path set through a preset neural network to obtain a sampling expansion scheme; A third construction module, configured to search for nodes based on the sampling expansion scheme for the starting point and the target point, and perform node collision detection during the search process. Wherein, when colliding with an obstacle, a sampling reconstruction model is constructed, and different step size adjustments are made to the nodes obtained by searching the sampling expansion scheme according to the substation layout information, and the shortest path is selected to obtain a feasible path; A fourth construction module, configured to perform smoothing optimization on the feasible path to obtain a final inspection path.

[0006] Third aspect, the present application also provides a path planning device for an inspection robot based on a neural network, including: A memory, configured to store a computer program; A processor, configured to implement the steps of the path planning method for the inspection robot based on the neural network when executing the computer program.

[0007] Fourth aspect, the present application also provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned path planning method for the inspection robot based on the neural network are implemented.

[0008] The beneficial effects of the present invention are: The present invention adopts a sampling expansion scheme, a sampling reconstruction model, and a step size adjustment strategy. Specifically, an initial sampling path set is obtained according to the substation layout, the sampling probability of the initial sampling path set is calculated, and the initial sampling path set is iteratively optimized through a neural network to obtain a sampling expansion scheme. The iterative optimization through the neural network is used to identify and reduce duplicate sampling areas, effectively reducing the deviation and error in the data collection process and improving the overall sampling efficiency and data quality. The sampling reconstruction model is constructed based on sampling collision detection and is used to plan the optimal path that can safely bypass these obstacles, not only enhancing the safety of the inspection path and ensuring the smooth passage of the robot. The step size adjustment strategy is used to select an appropriate step size according to the obstacles, making its movement step size more flexible and effectively avoiding obstacles. The present invention plans the inspection path by deploying a robot for substation inspection, integrating the sampling expansion scheme, the sampling reconstruction model, and the step size adjustment strategy and performing smoothing processing, thus solving the problem of safety risks in the operation and maintenance management of substation equipment.

[0009] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will become obvious from the specification or can be understood by implementing the embodiments of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written specification, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0011] Figure 1 It is a schematic flowchart of the inspection robot path planning method based on a neural network described in the embodiments of the present invention; Figure 2 It is a schematic structural diagram of the inspection robot path planning device based on a neural network described in the embodiments of the present invention.

[0012] Reference numerals in the figure: 800, inspection robot path planning device based on a neural network; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0014] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive distinction and cannot be construed as indicating or implying relative importance.

[0015] Embodiment 1: This embodiment provides a path planning method for an inspection robot based on a neural network.

[0016] See Figure 1 , which shows that this method includes steps S1 to S5, including: S1: Obtain substation layout information, where the substation layout information includes a starting point and a target point; S2: Search for nodes according to the starting point and the target point. During the search process, perform node collision detection, and connect the starting point, the target point, and the searched nodes to obtain an initial sampling path set; In this step, take the starting point as the root node, search for nodes through a probabilistic target biasing strategy during the iteration process, and detect whether the searched nodes collide with obstacles through the Collision function during the search process. If there is no collision, connect the starting point, the target point, and the searched nodes. If there is a collision, discard the currently searched node and continue to search for the next node until the target point is expanded to obtain an initial sampling path set; Among them, the Collision function is to judge whether the line connecting the searched node and the previous searched node intersects with an obstacle. This step is used to quickly eliminate those obviously unreasonable or inefficient paths from many candidate paths, reduce the amount of data for subsequent processing, effectively narrow the selection range, and provide a basis for subsequent more refined optimization.

[0017] S3: Calculate the sampling probabilities of the nodes in the initial sampling path set. Combining the collision information and time cost information obtained from the collision detection, iteratively optimize the nodes in the initial sampling path set through a preset neural network to obtain a sampling expansion plan. To clarify the specific method for obtaining the sampling expansion plan, steps S3 includes S31 to S36, specifically: S31: Obtain historical trajectory data. S32: Perform path planning processing on the initial sampling path set based on a preset adaptive positioning algorithm, and optimize the initial sampling path set through a preset path evaluation criterion to obtain a local path. To clarify the specific method for obtaining the local path, steps S32 includes S321 to S323, specifically: S321: Analyze and calculate the collision information and sampling information obtained from the collision detection based on a preset adaptive positioning algorithm to obtain the global sampling point position. In this step, the adaptive positioning algorithm uses the collision information to identify and avoid potential obstacles, ensuring that the robot can move without touching any obstacles. The adaptive positioning algorithm uses the sampling information to identify feature points in the environment, determine its position in the environment, and calculate based on the analyzed information to obtain the global sampling point position.

[0018] S322: Perform global path planning processing on the global sampling point position, and optimize the initial sampling path set through the path evaluation criterion to obtain a global path. In this step, plan from the global sampling point position to the target point, and optimize the initial sampling path set through the constraint of the path evaluation criterion to obtain a collision-free global path.

[0019] S323: Perform local path planning processing on the global path according to a preset evaluation function to obtain a local path.

[0020] In this step, the evaluation function is: (1); In the above formula (1), represents the total score of the evaluation function, , , and all represent weight coefficients, represents the path cost evaluation index; represents the distance evaluation index from the starting point to the global sampling point position; represents the difference evaluation index from the global path; Represents the time cost evaluation index; Among them, the total score of the evaluation function Is used to measure the quality of the local path.

[0021] S33: Make a decision on the local path according to the historical trajectory data to obtain a walking path segment; In this step, make a decision on the local path according to the trajectory time and trajectory points in the historical trajectory data to obtain a walking path segment.

[0022] S34: Calculate the normalized probability of the walking path segment to obtain a sampling probability expression; In this step, the sampling probability expression is: (2); In the above formula (2), Represents the probability of selecting a feasible path segment in the section from the starting point to the target point, Represents probability, Represents a feasible path segment, Represents the target point, Represents the starting point, And Both represent parameters affecting path selection, Represents the sum of the exponential scores of all possible path segments, Represents the exponential function, Represents the score calculated by the evaluation function for the feasible path segment, Represents the scaling factor; Among them, the sum of the exponential scores of all possible path segments is used for normalization processing.

[0023] S35: Calculate the nodes in the initial sampling path set based on the sampling probability expression, and integrate the nodes in the initial sampling path set through the collision information and the time cost information to obtain an optimized sampling path set; S36: Iteratively optimize the nodes in the optimized sampling path set according to the neural network to obtain a sampling expansion plan.

[0024] In this step, the neural network extracts deep features for each node in the optimized sampling path set to obtain the deep features of the nodes, and iteratively adjusts each node in the optimized sampling path set based on the deep features of the nodes to obtain a sampling expansion plan.

[0025] During the iterative adjustment process, the overall performance of the optimized sampling path set is evaluated. Based on the evaluation results, the nodes with duplicates in the optimized sampling path set are taken as the optimization objects, and the neural network is used to predict better positions or alternative paths. Among them, the overall performance of the current path set includes multiple key indicators such as the length, smoothness, safety (i.e., the ability to avoid collisions and potential risks), and time cost of the path. The neural network is used to identify and reduce duplicate sampling areas, effectively reducing the bias and error in the data collection process, and improving the overall sampling efficiency and data quality.

[0026] S4: Search for nodes for the starting point and the target point based on the sampling expansion scheme, and perform node collision detection during the search process. Among them, when colliding with an obstacle, a sampling reconstruction model is constructed, and different step size adjustments are made to the nodes obtained by searching the sampling expansion scheme according to the substation layout information, and the shortest path is selected to obtain a feasible path. To clarify the specific method for obtaining the feasible path, steps S4 includes S41 to S44, specifically: S41: Search for nodes for the starting point and the target point based on the sampling expansion scheme to obtain expansion points. The expansion points are the nodes obtained by searching the sampling expansion scheme, and there are no less than two expansion points. In this step, based on the sampling expansion scheme, during the iteration process, the starting point and the target point are searched for nodes through the probability target biasing strategy to obtain expansion points.

[0027] S42: Perform collision detection on each of the expansion points during the search process. When an expansion point collides with an obstacle, the initial sampling path set needs to be reconstructed. In this step, during the search process, the Collision function is used to detect whether the searched node collides with an obstacle. When an expansion point collides with an obstacle, the initial sampling path set needs to be reconstructed.

[0028] S43: Random sampling construction is performed based on the starting point and the target point to obtain a sampling reconstruction model. To clarify the specific method for obtaining the sampling reconstruction model, steps S43 includes S431 to S433, specifically: S431: Perform multiple random samplings and construction based on the distance between the starting point and the target point to obtain a reconstructed trajectory. To clarify the specific method for obtaining the reconstructed trajectory, steps S431 includes S4311 to S43114, specifically: S4311: Randomly sample the distance between the starting point and the target point multiple times based on a preset path planning algorithm to obtain a first random sampling point and a second random sampling point; In this step, the first random sampling point is , and the second random sampling point is .

[0029] S4312: Calculate the probability for the first random sampling point and the second random sampling point to obtain a path selection probability; In this step, calculate the utility values for the first random sampling point and the second random sampling point to obtain a first utility value and a second utility value, and perform an evaluation calculation based on the first utility value and the second utility value to obtain a path selection probability.

[0030] In this step, the path selection probability is: (3); In the above formula (3), represents the path selection probability, represents the utility value of the lower path, represents the utility value of the lower path, represents all possible paths; Among them, the utility value of the lower path represents the first utility value, and the utility value of the lower path represents the second utility value.

[0031] S4313: Perform path scoring based on the path selection probability, select the path with the highest score to obtain a matching path, and the path scoring is for the shortest path planning, comprehensiveness of the inspection area, and real-time nature of environmental changes; S4314: Perform priority selection on the initial sampling path set based on the matching path and the path selection probability to obtain a reconstructed trajectory.

[0032] S432: Modify the initial sampling path set according to the reconstructed trajectory, and perform constraints based on the obstacle positions obtained through the collision detection to obtain a reconstructed path set; In this step, according to the parameters and characteristics of the reconstructed trajectory, compare and analyze each path in the initial sampling path set one by one, identify the parts that do not match or conflict with the reconstructed trajectory, and perform further constraint processing on the modified path set based on the obstacle positions obtained through the collision detection to obtain a reconstructed path set.

[0033] S433: Iteratively optimize the reconstructed path set according to a preset maximum number of iterations, and obtain a sampling reconstruction model through statistical analysis of the error data in the initial sampling path set.

[0034] In this step, the sampling reconstruction model can remove sampling points that are far away globally and add globally better sampling points to the path, so as to plan an optimal path that can safely bypass these obstacles, which not only enhances the safety of the inspection path and ensures that the robot can pass smoothly.

[0035] S44: Adjust the step sizes of each expansion point differently according to the sampling reconstruction model and the substation layout information, and select the shortest path to obtain a feasible path.

[0036] To clarify the specific method for obtaining the feasible path, steps S441 to S446 are included in step S44, specifically: S441: Calculate according to the sampling reconstruction model, the different regional environments in the substation layout information, and the collision information to obtain the number of obstacles; In this step, considering the occurrence of dynamic obstacles on the roads in the substation layout information and the route planning in narrow spaces in the substation, it is applicable to the daily inspection of high-voltage transmission lines in the substation and provides the possibility for multi-robot inspection at the same time.

[0037] S442: Construct according to the number of obstacles and different range constraints of the preset number of obstacles to obtain a dynamic step size expression; In this step, the dynamic step size expression is: (4); In the above formula (4), represents the dynamic step size, represents the maximum value of the step size, represents that the step size increases when the current number of obstacles is small, represents that the step size is moderate when the number of obstacles reaches a certain range, represents that the step size decreases when the number of obstacles is large, represents the minimum value of the step size, represents the current number of obstacles, represents the preset number of obstacles; S443: Judge each expansion point according to the dynamic step size expression to obtain a node judgment result; In this step, each expansion point is judged according to the current number of obstacles; Among them, the current number of obstacles is ; When When there is no obstacle, the step size is x; When is greater than 0 but less than or equal to the number of obstacles is small, and the step size is ; When is greater than but less than or equal to the number of obstacles is moderate, and the step size is , being the initial step size; When is greater than but less than the number of obstacles is large, and the step size is ; When there is a complete obstacle or the number of obstacles is excessive within the range, the step size is .

[0038] S444: Based on the node judgment result, select different step sizes to adjust each of the expansion points to obtain the final expansion points; S445: Connect the starting point, the target point, and the final expansion points to obtain the final path set; S446: Select the shortest path for each cumulative distance in the final path set to obtain the feasible path.

[0039] In this step, the step size adjustment strategy selects an appropriate step size according to the obstacles, making the movement step size more flexible and effectively avoiding obstacles.

[0040] S5: Smoothly optimize the feasible path to obtain the final inspection path.

[0041] To clarify the specific acquisition method of the final inspection path, step S5 includes S51 to S53, specifically: S51: Divide the feasible path into multiple sub-regions to obtain path blocks, and the number of path blocks is not less than two; S52: Construct a function for each path block according to the preset cubic spline interpolation method to obtain the piecewise interpolation function expression; In this step, the piecewise interpolation function expression is: (5); In the above formula (5), represents the th piecewise interpolation function, , , and Coefficients representing each sub - interval, representing the independent variable the difference between and the nodes of the piece - wise interpolation function representing the independent variable the difference between and the nodes of the piece - wise interpolation function representing the independent variable the difference between and the nodes of the piece - wise interpolation function representing the interpolation points of the \(i\) - th interval.

[0042] Among them, the interpolation conditions are satisfied, that is: For all \(i = 0,1,\cdots,n\); the interpolation curve is smooth, that is and its derivative function and are continuous.

[0043] S53: Based on the piece - wise interpolation function expression, construct the starting point and the target point, analyze the safety distance of obstacles and the path rotation limit, and perform smoothing optimization on the feasible path to obtain the final inspection path.

[0044] Embodiment 2: This embodiment provides a path planning device for an inspection robot based on a neural network. The device includes: An acquisition module, configured to acquire substation layout information, where the substation layout information includes a starting point and a target point; A first construction module, configured to search for nodes according to the starting point and the target point, perform node collision detection during the search process, and connect the starting point, the target point, and the searched nodes to obtain an initial sampling path set; A second construction module, configured to calculate the sampling probability of the nodes in the initial sampling path set, combine the collision information and time - cost information obtained from the collision detection, and perform iterative optimization on the nodes in the initial sampling path set through a preset neural network to obtain a sampling expansion scheme; To clarify the specific acquisition method of the second construction module, specifically: A first acquisition unit, configured to acquire historical trajectory data; A first processing unit, configured to perform path planning on the initial sampling path set based on a preset adaptive positioning algorithm, and optimize the initial sampling path set through a preset path evaluation criterion to obtain a local path; A second processing unit, configured to perform decision processing on the local path according to the historical trajectory data to obtain a walking path segment; A third processing unit, configured to perform normalized probability calculation on the walking path segment to obtain a sampling probability expression; A fourth processing unit, configured to calculate the nodes in the initial sampling path set based on the sampling probability expression, and integrate the nodes in the initial sampling path set through the collision information and the time cost information to obtain an optimized sampling path set; A fifth processing unit, configured to perform iterative optimization on the nodes in the optimized sampling path set according to the neural network to obtain a sampling expansion scheme.

[0045] A third construction module, configured to search for nodes for the starting point and the target point based on the sampling expansion scheme, and perform node collision detection during the search process. Wherein, when colliding with an obstacle, a sampling reconstruction model is constructed, and different step size adjustments are performed on the nodes obtained by searching the sampling expansion scheme according to the substation layout information and the shortest path is selected to obtain a feasible path; To clarify the specific acquisition method of the third construction module, specifically: A fifth processing unit, configured to search for nodes for the starting point and the target point based on the sampling expansion scheme to obtain expansion points, where the expansion points are the nodes obtained by searching the sampling expansion scheme, and there are no less than two expansion points; A sixth processing unit, configured to perform collision detection on each expansion point during the search process. When the expansion point collides with an obstacle, the initial sampling path set needs to be reconstructed; A seventh processing unit, configured to perform random sampling construction between the starting point and the target point to obtain a sampling reconstruction model; An eighth processing unit, configured to perform different step size adjustments on each expansion point according to the sampling reconstruction model and the substation layout information and select the shortest path to obtain a feasible path.

[0046] A fourth construction module, configured to perform smoothing optimization on the feasible path to obtain a final inspection path.

[0047] To clarify the specific acquisition method of the fourth construction module, specifically: A first optimization unit, configured to divide the feasible path into multiple sub-regions to obtain path blocks, where there are no less than two path blocks; A second optimization unit, configured to construct a function for each path block according to the preset cubic spline interpolation method to obtain a piecewise interpolation function expression; A third optimization unit is configured to construct the starting point and the target point based on the piecewise interpolation function expression, analyze the safety distance of the obstacle and the path rotation limit, perform a smoothing optimization process on the feasible path, and obtain a final inspection path.

[0048] It should be noted that for the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0049] Embodiment 3: Corresponding to the above method embodiment, a path planning device for an inspection robot based on a neural network is also provided in this embodiment. The path planning device for an inspection robot based on a neural network described below can be correspondingly referred to the path planning method for an inspection robot based on a neural network described above.

[0050] Figure 2 is a block diagram of a path planning device 800 for an inspection robot based on a neural network shown according to an exemplary embodiment. As Figure 2 shown, the path planning device 800 for an inspection robot based on a neural network may include: a processor 801, a memory 802. The path planning device 800 for an inspection robot based on a neural network may further include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0051] Among them, the processor 801 is used to control the overall operation of the inspection robot path planning device 800 based on a neural network to complete all or part of the steps in the above-mentioned inspection robot path planning method based on a neural network. The memory 802 is used to store various types of data to support the operation of the inspection robot path planning device 800 based on a neural network. These data may include, for example, instructions for any application or method operating on the inspection robot path planning device 800 based on a neural network, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, and so on. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk. The multimedia component 803 may include a screen and an audio component. Among them, the screen can be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone, and the microphone is used to receive external audio signals. The received audio signals can be further stored in the memory 802 or sent through the communication component 805. The audio component further includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be a keyboard, a mouse, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the inspection robot path planning device 800 based on a neural network and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination of one or more of them. Therefore, the corresponding communication component 805 may include: a Wi-Fi module, a Bluetooth module, and an NFC module.

[0052] In an exemplary embodiment, the neural network-based inspection robot path planning device 800 can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components, and is used to execute the above-mentioned neural network-based inspection robot path planning method.

[0053] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When the program instructions are executed by a processor, the steps of the above-mentioned neural network-based inspection robot path planning method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the neural network-based inspection robot path planning device 800 to complete the above-mentioned neural network-based inspection robot path planning method.

[0054] Embodiment 4: Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. A readable storage medium described below can be correspondingly referred to the neural network-based inspection robot path planning method described above.

[0055] A readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the neural network-based inspection robot path planning method in the above method embodiment are implemented.

[0056] The readable storage medium can specifically be various readable storage media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0057] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0058] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A path planning method for an inspection robot based on a neural network, characterized in that: include: Acquire substation layout information, wherein the substation layout information includes a starting point and a target point; Searching for nodes according to the starting point and the target point, performing node collision detection during the search process, connecting the starting point, the target point and the searched nodes to obtain an initial sampling path set; Calculate the sampling probability of the nodes in the initial sampling path set, combine the collision information and time cost information obtained by the collision detection, iteratively optimize the nodes in the initial sampling path set through a preset neural network, and obtain a sampling expansion plan; Based on the sampling expansion scheme, nodes are searched for the starting point and the target point, and node collision detection is performed during the search process, wherein, when colliding with an obstacle, a sampling reconstruction model is constructed, and different step lengths are adjusted for the nodes searched by the sampling expansion scheme according to the substation layout information, and the shortest path is selected to obtain a feasible path; The feasible path is smoothly optimized to obtain a final inspection path.

2. The neural network-based inspection robot path planning method according to claim 1, characterized in that: The sampling probability of the nodes in the initial sampling path set is calculated, and the collision information and time cost information obtained by the collision detection are combined, and the nodes in the initial sampling path set are iteratively optimized through a preset neural network to obtain a sampling expansion scheme, including: Get historical trajectory data; Performing path planning processing on the initial sampling path set based on a preset adaptive positioning algorithm, optimizing the initial sampling path set according to a preset path evaluation criterion, and obtaining a local path; Perform decision processing on the local path according to the historical trajectory data to obtain a walking path segment; Performing normalized probability calculation on the walking path segment to obtain a sampling probability expression; Calculating the nodes in the initial sampling path set based on the sampling probability expression, integrating the nodes in the initial sampling path set through the collision information and the time cost information, and obtaining an optimized sampling path set; The nodes in the optimized sampling path set are iteratively optimized according to the neural network to obtain a sampling expansion solution.

3. The inspection robot path planning method based on neural network according to claim 1 is characterized in that: Based on the sampling expansion scheme, nodes are searched for the starting point and the target point, and node collision detection is performed during the search process. When an obstacle is collided with, a sampling reconstruction model is constructed, and different step lengths are adjusted for the nodes searched by the sampling expansion scheme according to the substation layout information, and the shortest path is selected to obtain a feasible path, including: Searching nodes for the starting point and the target point based on the sampling expansion scheme to obtain expansion points, where the expansion points are nodes obtained by searching the sampling expansion scheme, and there are no less than two expansion points; During the search process, collision detection is performed on each of the expansion points. When the expansion point collides with an obstacle, the initial sampling path set needs to be reconstructed; Based on random sampling between the starting point and the target point, a sampling reconstruction model is obtained; Different step lengths are adjusted for each of the expansion points according to the sampling reconstruction model and the substation layout information, and the shortest path is selected to obtain a feasible path.

4. The inspection robot path planning method based on neural network according to claim 3 is characterized in that: Based on random sampling between the starting point and the target point, a sampling reconstruction model is obtained, including: Performing multiple random sampling and construction based on the distance between the starting point and the target point to obtain a reconstructed trajectory; The initial sampling path set is modified according to the reconstructed trajectory, and the obstacle position obtained by the collision detection is constrained to obtain a reconstructed path set; The reconstructed path set is iteratively optimized according to a preset maximum number of iterations, and a sampling reconstruction model is obtained by statistically analyzing the error data in the initial sampling path set.

5. The inspection robot path planning method based on neural network according to claim 3 is characterized in that: According to the sampling reconstruction model and the substation layout information, different step lengths are adjusted for each of the expansion points and the shortest path is selected to obtain a feasible path, including: Calculating according to the sampling reconstruction model, different regional environments in the substation layout information and the collision information to obtain the number of obstacles; According to different range constraints of the number of obstacles and the preset number of obstacles, a dynamic step length expression is constructed; Judging each of the expansion points according to the dynamic step length expression to obtain a node judgment result; Selecting different step lengths based on the node judgment result to adjust each of the expansion points to obtain a final expansion point; Connecting the starting point, the target point and the final extension point to obtain a final path set; The shortest path is selected for each cumulative distance in the final path set to obtain a feasible path.

6. The neural network-based inspection robot path planning method according to claim 1, characterized in that: The feasible path is smoothly optimized to obtain a final inspection path, including: Dividing the feasible path into multiple sub-regions to obtain path blocks, wherein the path blocks are no less than two; Constructing a function for each path block according to a preset cubic spline interpolation method to obtain a piecewise interpolation function expression; The starting point and the target point are constructed based on the piecewise interpolation function expression, and the safe distance of obstacles and the path rotation limit are analyzed to perform smooth optimization processing on the feasible path to obtain the final inspection path.

7. A path planning device for an inspection robot based on a neural network, characterized in that: include: An acquisition module, used to acquire substation layout information, wherein the substation layout information includes a starting point and a target point; A first construction module is used to search for nodes according to the starting point and the target point, perform node collision detection during the search process, and connect the starting point, the target point and the searched nodes to obtain an initial sampling path set; The second construction module is used to calculate the sampling probability of the nodes in the initial sampling path set, combine the collision information and time cost information obtained by the collision detection, iteratively optimize the nodes in the initial sampling path set through a preset neural network, and obtain a sampling expansion plan; The third construction module is used to search for nodes for the starting point and the target point based on the sampling expansion scheme, and perform node collision detection during the search process, wherein, when colliding with an obstacle, a sampling reconstruction model is constructed, and different step lengths are adjusted for the nodes searched by the sampling expansion scheme according to the substation layout information, and the shortest path is selected to obtain a feasible path; The fourth construction module is used to smoothly optimize the feasible path to obtain a final inspection path.

8. The neural network-based inspection robot path planning device according to claim 7, characterized in that: The second building block includes: A first acquisition unit, used to acquire historical trajectory data; A first processing unit is used to perform path planning processing on the initial sampling path set based on a preset adaptive positioning algorithm, and optimize the initial sampling path set according to a preset path evaluation criterion to obtain a local path; A second processing unit is used to perform decision processing on the local path according to the historical trajectory data to obtain a walking path segment; A third processing unit is used to perform normalized probability calculation on the walking path segment to obtain a sampling probability expression; a fourth processing unit, configured to calculate the nodes in the initial sampling path set based on the sampling probability expression, and integrate the nodes in the initial sampling path set through the collision information and the time cost information to obtain an optimized sampling path set; The fifth processing unit is used to iteratively optimize the nodes in the optimized sampling path set according to the neural network to obtain a sampling expansion solution.

9. The neural network-based inspection robot path planning device according to claim 7, characterized in that: The third building block includes: A fifth processing unit is used to search nodes for the starting point and the target point based on the sampling and expansion scheme to obtain expansion points, where the expansion points are nodes obtained by searching the sampling and expansion scheme, and there are no less than two expansion points; A sixth processing unit, configured to perform collision detection on each of the expansion points during the search process, and when the expansion point collides with an obstacle, the initial sampling path set needs to be reconstructed; A seventh processing unit, configured to perform random sampling construction between the starting point and the target point to obtain a sampling reconstruction model; An eighth processing unit is used to adjust the step length of each of the expansion points according to the sampling reconstruction model and the substation layout information and select the shortest path to obtain a feasible path.

10. The neural network-based inspection robot path planning device according to claim 7, characterized in that: The fourth building block includes: A first optimization unit is used to divide the feasible path into multiple sub-regions to obtain path blocks, and the path blocks are no less than two; A second optimization unit is used to construct a function for each of the path blocks according to a preset cubic spline interpolation method to obtain a piecewise interpolation function expression; The third optimization unit is used to construct the starting point and the target point based on the piecewise interpolation function expression, analyze the safety distance of the obstacle and the path rotation limit, perform smooth optimization processing on the feasible path, and obtain a final inspection path.

Images