An unmanned substation inspection robot and its autonomous inspection method

By analyzing the point cloud reflection intensity and distribution differences in the substation inspection area, the driving speed is dynamically adjusted and the optimal travel trajectory is planned, solving the problems of wheel slippage and posture divergence of unattended inspection robots in substation inspection, and realizing safe autonomous inspection of substations.

CN123113889APending Publication Date: 2026-11-17SHAANXI HUAHETONG ELECTRONIC TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611482896.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Unmanned inspection robots are prone to interruption of inspection tasks during substation inspections due to wheel slippage and position divergence, especially on longitudinally equidistant pillars in long corridors and on sloping concrete roads, where there are problems with lidar measurement errors and wheeled odometer calculation failures.

Method used

By acquiring the point cloud reflection intensity of the substation inspection area where the robot is currently located, analyzing the intensity distribution difference in local areas, determining the boundary distance between cement and gravel, constructing the vehicle's lateral offset distance and slip penalty value, dynamically adjusting the driving speed to avoid wheel slippage, and planning the optimal travel trajectory.

Benefits of technology

It effectively avoids wheel slippage and attitude divergence, ensures the continuity of inspection tasks, prevents wheel odometer trajectory calculation divergence and chassis attitude loss of control, and realizes safe and autonomous inspection of substations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113889A_ABST
    Figure CN123113889A_ABST
Patent Text Reader

Abstract

This invention relates to the field of robot inspection control technology, specifically to an unmanned substation inspection robot and its autonomous inspection method. The method includes: acquiring the point cloud reflection intensity in different local areas of the substation where the robot is currently located; determining the intensity distribution difference degree of each local area based on the difference between the point cloud reflection intensity in each local area and the baseline reflection intensity distribution of the cement road surface; determining the boundary distance between cement and gravel based on the intensity distribution difference degree of different local areas; determining the optimal travel trajectory based on the position distribution and boundary distance of path points on each candidate trajectory; determining the upper limit of travel speed using the wheel slip penalty value of the path points on the optimal travel trajectory; and the robot inspecting the substation inspection area according to the optimal travel trajectory and the upper limit of travel speed. This invention enables the robot to adaptively adjust its travel speed according to the actual slip risk of the road surface ahead, preventing tire slippage and wheel spin from the power source.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of robot inspection and control technology, specifically to an unmanned substation inspection robot and its autonomous inspection method. Background Technology

[0002] The outdoor switchyard corridor of a substation is typically equipped with equidistantly arranged metal insulated supports, a typical operating environment for unmanned inspection robots. During actual inspections, the following technical challenges arise: First, the equidistant longitudinal arrangement of supports in the corridor prevents LiDAR from detecting structural changes in that direction, making the point cloud registration algorithm prone to longitudinal degradation and positioning slippage. In this case, the system usually relies on the chassis's wheeled odometer for trajectory estimation, but accurate odometer calculations require the wheels to maintain a smooth, rolling motion without slippage. Second, the center of the corridor is usually a flat, hardened concrete surface, with loose gravel areas for insulation on both sides. The concrete surface generally has a 1% to 3% drainage slope. This slope introduces a height error component into the lateral distance calculated by conventional two-dimensional planar radar, reducing the accuracy of road edge recognition. Third, existing local obstacle avoidance algorithms typically rely solely on spatial geometric distance to plan obstacle avoidance trajectories. When the inspection robot approaches the edge of the corridor to avoid temporary obstacles, the obstacle avoidance trajectory can easily lead the outer wheels into the loose gravel area. Due to the significant abrupt change in the coefficient of friction between the cement surface and the gravel, the electromagnetic torque output by the bottom motor during start-up or acceleration can easily exceed the maximum static friction of the gravel road surface, causing the left and right wheels to spin and slip differentially. The spinning of the wheels directly undermines the calculation premise of the wheel odometer, causing the chassis posture to completely diverge and triggering a yaw collision, forcing the inspection task to be unexpectedly interrupted. Summary of the Invention

[0003] To address the problems encountered by existing inspection robots during autonomous substation inspections, such as wheel slippage, pose divergence, and inspection task interruption, this invention aims to provide an unmanned substation inspection robot and its autonomous inspection method. The specific technical solution adopted is as follows: In a first aspect, the present invention provides an unmanned substation inspection robot and its autonomous inspection method, the method comprising the following steps: Obtain the point cloud reflection intensity in different local sections of the substation inspection area where the robot is currently located; Based on the difference between the point cloud reflection intensity and the reference reflection intensity distribution of the cement pavement in each local interval, the intensity distribution difference degree of each local interval is determined; based on the intensity distribution difference degree of different local intervals, the boundary distance between cement and crushed stone is determined. Based on the positional distribution of path points on each candidate trajectory of the robot, the lateral offset distance of the vehicle corresponding to each path point is determined; based on the lateral offset distance of the vehicle and the boundary distance, the wheel slip penalty value of each path point on each candidate trajectory is obtained. The optimal travel trajectory is determined by comparing the cumulative wheel slip penalty values ​​of path points on different alternative trajectories; the upper limit of travel speed is determined using the wheel slip penalty values ​​of path points on the optimal travel trajectory; the robot then inspects the substation inspection area based on the optimal travel trajectory and the upper limit of travel speed.

[0004] Preferably, obtaining the point cloud reflection intensity in different local intervals within the substation inspection area where the robot is currently located includes: Using the center projection point at the bottom of the robot chassis as the origin of the coordinate system and the opposite direction of the vertical vector of gravity as the plane normal, an orthogonal projection plane of gravity is constructed. The lateral feature abundance vector is projected onto the gravity orthogonal projection plane to obtain the lateral projection coordinate axis; A local vertical sampling window is set along the horizontal projection coordinate axis to obtain the current sampling area; Based on the location distribution of laser points within the current sampling area, the current sampling area is divided into multiple local intervals; Distance compensation and normalization were performed on the initial reflection intensity of the point cloud in each local interval to obtain the reflection intensity of the point cloud in each local interval.

[0005] Preferably, obtaining the lateral feature abundance vector includes: A nonlinear least squares registration algorithm is used to iteratively solve the current frame of lidar data, and the registration convergence confidence of the algorithm output is extracted. The registration optimization information matrix is ​​obtained only if the registration convergence confidence is greater than the preset registration success threshold. The information matrix is ​​subjected to eigenvalue decomposition, and the eigenvector corresponding to the eigenvalue with the largest value is extracted from the decomposition result as the lateral feature abundance vector.

[0006] Preferably, determining the intensity distribution difference degree of each local interval based on the difference between the point cloud reflection intensity and the reference reflection intensity distribution of the cement pavement within each local interval includes: The intensity distribution difference of candidate local intervals is obtained by using the cumulative distribution function of the reflection intensity distribution map of the candidate local intervals and the cumulative distribution function of the reflection intensity benchmark distribution map of the smooth cement pavement; wherein, the reflection intensity distribution map of the candidate local intervals is generated based on all point cloud positions and point cloud reflection intensities within the candidate local intervals. The candidate local interval is any local interval.

[0007] Preferably, determining the boundary distance between cement and crushed stone based on the difference in strength distribution in different local areas includes: Determine the index number of each local interval, and compare the intensity distribution difference of each local interval with the difference threshold in ascending order of index number. Multiply the index number of the first local interval where the intensity distribution difference is greater than the difference threshold by the length of the horizontal sampling segment to obtain the boundary distance between cement and crushed stone. If the intensity distribution difference in all local intervals is less than or equal to the difference threshold, then the boundary distance between cement and crushed stone is set to a preset distance.

[0008] Preferably, determining the lateral offset distance of the vehicle corresponding to each path point based on the positional distribution of path points on each candidate trajectory of the robot includes: For any of the alternative trajectories: The three-dimensional spatial coordinates of each path point on any candidate trajectory are projected onto the lateral projection coordinate axis to obtain the absolute coordinate values ​​of each path point on the lateral projection coordinate axis. The absolute coordinate values ​​are used as the lateral offset distance of the vehicle corresponding to each path point.

[0009] Preferably, obtaining the wheel slip penalty value for each path point on each candidate trajectory based on the vehicle's lateral offset distance and the boundary distance includes: Calculate the first difference between the boundary distance and the safety tire half-width constant; When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is less than or equal to the first difference, the wheel slip penalty value of the path point to be analyzed is set as the basic penalty constant. When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is greater than the first difference and less than the boundary distance, the wheel slip penalty value of the path point to be analyzed is determined based on the difference between the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed and the first difference, and the pavement clearance ratio of the local interval where the path point to be analyzed is located. When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is greater than or equal to the boundary distance, the wheel slip penalty value of the path point to be analyzed is set to the truncation penalty constant. The base penalty constant is less than the truncation penalty constant.

[0010] The path point to be analyzed is any path point on any candidate trajectory.

[0011] Preferably, determining the optimal travel trajectory by comparing the cumulative results of wheel slip penalty values ​​at path points on different candidate trajectories includes: For any candidate trajectory: the sum of the wheel slip penalty values ​​of all path points on the candidate trajectory is taken as the total slip penalty value of the candidate trajectory; The candidate trajectory corresponding to the minimum total slip penalty value is taken as the optimal travel trajectory.

[0012] Preferably, determining the upper limit of driving speed using the wheel slip penalty value of the path point on the optimal travel trajectory includes: The speed reduction caused by the current road surface roughness is obtained by multiplying the wheel slip penalty value of the path point on the optimal travel trajectory with the speed reduction coefficient. The maximum value between the difference between the base cruising speed and the speed decay, and the absolute lower limit speed, is taken as the upper limit of driving speed.

[0013] Secondly, the present invention provides an unmanned substation inspection robot, including a memory and a processor, wherein the processor executes a computer program stored in the memory to implement the method described in the first aspect above.

[0014] The present invention has at least the following beneficial effects: This invention acquires the point cloud reflection intensity in different local sections of the substation inspection area where the robot is currently located, and analyzes the differences between the reflection intensity distribution of each local section and the baseline reflection intensity distribution of the cement road surface. It quantifies the intensity distribution difference of each local section, reflecting the changes in the physical properties of the road surface material. This provides a reliable basis for subsequently identifying the boundary distance between cement and gravel, avoiding the inaccurate road edge identification caused by height errors when relying solely on geometric distance or two-dimensional radar on inclined roads. Then, based on the intensity distribution difference of different local sections, the boundary distance between cement and gravel is determined. This boundary distance dynamically reflects the actual physical boundary between safe road surfaces and high-risk gravel areas in the current inspection area. Using the positional distribution of path points on each of the robot's alternative trajectories, the lateral offset distance of the vehicle at each path point is determined. This approach links the trajectory planning problem with the lateral position risk of the wheels. Based on the vehicle's lateral offset distance and boundary distance, a wheel slip penalty value is constructed for each path point on each candidate trajectory. This penalty value quantifies the risk of the robot's wheels slipping when traveling along that path point. By comparing the cumulative results of the wheel slip penalty values ​​for path points on different candidate trajectories, the optimal travel trajectory is selected. This avoids problems such as differential wheel spin and slippage caused by the outer wheel running into a loose gravel area, as well as the resulting divergence in wheel odometer trajectory calculation, chassis pose loss of control, and yaw collisions. The wheel slip penalty value for the path points on the optimal travel trajectory is used to determine the upper limit of the travel speed, enabling the robot's travel speed to be dynamically and adaptively adjusted according to the actual slippage risk on the road ahead, preventing tire slippage and spin from the power source. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 The flowchart illustrates an autonomous inspection method for an unattended substation inspection robot provided in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description, in conjunction with the accompanying drawings and preferred embodiments, describes an unmanned substation inspection robot and its autonomous inspection method proposed according to the present invention.

[0018] The following description, in conjunction with the accompanying drawings, details the specific scheme of the unmanned substation inspection robot and its autonomous inspection method provided by this invention.

[0019] An embodiment of an autonomous inspection method for an unmanned substation inspection robot: This embodiment proposes an autonomous inspection method for an unmanned substation inspection robot, such as... Figure 1 As shown, the autonomous inspection method of an unmanned substation inspection robot in this embodiment includes the following steps: Step S1: Obtain the point cloud reflection intensity in different local sections of the substation inspection area where the robot is currently located.

[0020] First, the inspection robot is controlled to travel in a straight line on a smooth, hardened cement road surface. During this straight-line movement, a 3D LiDAR is used to collect continuous scanning point clouds and extract the reflection intensity values ​​of discrete points hitting the ground. Based on the extracted reflection intensity values, probability mass statistics are performed to generate a baseline distribution map of the cement road surface reflection intensity. Simultaneously, the robot's inherent chassis dimension parameters were read, and the safety tire half-width constant was extracted from them. The safety tire half-width constant is the lateral physical distance from the geometric center of the robot chassis to the outermost edge of the tire.

[0021] During the actual inspection process of the robot, the current frame of LiDAR data sampled at the current timestamp is obtained, and the single frame of laser scanning point cloud under the current timestamp is extracted. Then, every discrete point inside the point cloud is traversed to obtain the three-dimensional spatial coordinates and initial reflection intensity of each discrete point. Each discrete point is equivalent to a discrete laser point.

[0022] During the actual inspection of a substation by a robot, the metal insulated pillars arranged at equal intervals on both sides of the corridor in the substation inspection area cause the lidar to lack structural feature changes in the extension direction parallel to the corridor. This lack of geometric features will induce the laser odometry to produce erroneous slippage when calculating the relative pose. In order to solve this problem, this embodiment will separate the degradation direction lacking geometric constraints from the reference direction rich in features, and use them as a three-dimensional reference skeleton for evaluating the lateral offset of the vehicle.

[0023] Specifically, a nonlinear least squares registration algorithm is used to iteratively solve the current frame of LiDAR data. This algorithm can be the Iterative Closest Point (ICP) algorithm or the Normal Distributions Transform (NDT) algorithm, among others. As a specific implementation, the NDT algorithm is chosen, as it is an existing algorithm and will not be elaborated upon further here. During the iterative solution process, the registration convergence confidence score (the rate of change of iterative error in the NDT algorithm) is extracted. The registration optimization information matrix generated at the algorithm's underlying layer is obtained only if the registration convergence confidence score is greater than a preset registration success threshold. The preset registration success threshold can be 0.85, which can be set by the implementer according to specific circumstances.

[0024] The information matrix obtained above is subjected to eigenvalue decomposition to obtain a set of eigenvalues ​​and a set of eigenvectors corresponding to each eigenvalue; the eigenvector corresponding to the eigenvalue with the largest value is extracted from the decomposition results as the lateral feature abundance vector. This lateral feature abundance vector Pointing to the side perpendicular to the insulating pillar, it has rich structural texture.

[0025] Simultaneously, a read command is sent to the inertial measurement unit (IMU) inside the chassis to intercept the vertical gravity vector output by the IMU, pointing absolutely downwards from the Earth's center. Using the center projection point of the robot chassis bottom at the current moment as the origin of the coordinate system and the opposite direction of the vertical vector of gravity as the plane normal, construct an orthogonal projection surface of gravity.

[0026] The lateral feature abundance vector is projected onto the gravity orthogonal projection plane to obtain the lateral projection coordinate axis.

[0027] A local longitudinal sampling window is defined with the current geometric center of the robot's chassis as the origin and its longitudinal length limited to a preset effective ranging range. The obtained area within this window is recorded as the current sampling area. For example, this window can be set as a rectangular area between 1.0 meter and 3.0 meters in front of the chassis. After setting this window, all discrete points falling outside this window are directly discarded, thereby ensuring that the incident angle variation range of the laser point and radar participating in subsequent calculations remains convergent, avoiding large-scale signal distortion.

[0028] After obtaining a point cloud set with a true physical representation, the terrain slope interference is eliminated using a gravity benchmark, and the point cloud is divided into various local intervals.

[0029] The above steps yielded the lateral projection coordinate axes. Next, set the horizontal sampling segment length. Along this horizontal projection coordinate axis The current sampling area is divided into equidistant segments on both sides of the road to determine the maximum number of sampling intervals allowed by the current environment, and this number is set as the maximum index number. The length of the lateral sampling segment can be set to 0.2 meters.

[0030] Subsequently, iterate through all discrete point clouds within the candidate local interval, and for any discrete point cloud, calculate its projection coordinates on the horizontal axis. The absolute distance component is set as the vehicle's lateral offset distance. To convert continuous lateral vehicle offset distances into array-addressable index numbers, this embodiment uses the lateral vehicle offset distance... Divide by the length of the lateral sampling segment The initial index number is obtained by rounding down the ratio. To prevent array out-of-bounds errors from causing memory overflow, the initial index number is set to... With the maximum index number The two values ​​are compared, and the smaller value is selected as the final local interval index number of the corresponding point cloud. .

[0031] Using the above methods, we can obtain the index number of each point cloud, group all point clouds with the same index number into a local interval, divide all point clouds in the current sampling area into multiple local intervals, and use the index number of the point cloud in the local interval as the index number of the local interval.

[0032] After determining the index number of each local interval, for the k-th local interval, the reflection intensity of the point cloud is written into memory. In the data set corresponding to each array unit. After completing the allocation and writing of all reserved laser points in the current frame, in each corresponding local interval index number. Probability distribution statistics are performed within the corresponding set to generate the horizontal projection coordinate axis. Reflection intensity distribution map of each local area .

[0033] The system sequentially acquires the three-dimensional spatial coordinates of all discrete laser points within each local interval and calculates the actual spatial ranging distance of each laser point relative to the laser radar hardware emission center. After calculating the actual ranging distance, the system calls the laser radar factory calibration curve (i.e., the distance-intensity attenuation response model) preset in memory. Based on this model, distance compensation is performed on the initial reflection intensity values ​​of each extracted laser point, uniformly mapping and normalizing them to an equivalent intensity value at a standard reference distance (e.g., uniformly equivalent to 2.0 meters in front). After completing the distance compensation calculation, a normalized reflection intensity that eliminates spatial attenuation noise is generated, which is used as the true physical characterization for determining the road surface material. It should be noted that the point cloud reflection intensity mentioned below refers to the point cloud reflection intensity obtained after the above distance compensation and normalization processing.

[0034] Step S2: Determine the intensity distribution difference of each local interval based on the difference between the point cloud reflection intensity and the reference reflection intensity distribution of the cement pavement; determine the boundary distance between cement and crushed stone based on the intensity distribution difference of different local intervals.

[0035] The following explanation uses any local interval as an example. Other local intervals can be processed using the method provided in this embodiment.

[0036] Specifically, any local interval is designated as a candidate local interval. The cumulative distribution function of the reflection intensity distribution map for the candidate local interval and the cumulative distribution function of the baseline reflection intensity distribution map for the smooth cement pavement are calculated respectively. Based on the cumulative distribution functions of the reflection intensity distribution maps within the candidate local intervals and the baseline reflection intensity distribution map for the smooth cement pavement, the intensity distribution difference between the candidate local intervals is obtained.

[0037] As a specific implementation method, a specific formula for calculating the intensity distribution difference is given. The intensity distribution difference of the k-th local interval can be expressed as: in, This represents the degree of difference in intensity distribution within the k-th local interval. Indicates the reflection intensity value. This represents the maximum value of the reflected intensity. This indicates that the reflection intensity value within the k-th local interval is less than or equal to the reflection intensity value. The sum of probabilities, This indicates that the reflection intensity value in the baseline distribution map of reflection intensity of smooth cement pavement is less than or equal to the reflection intensity value. The sum of probabilities.

[0038] Indicates the reflection intensity value Above, the absolute probability difference in the strength distribution of the current local pavement compared to the reference pure cement pavement. Integration operation. It represents the minimum overall workload required to transport and reshape the current actual probability distribution into the ideal baseline distribution.

[0039] When radar detects enlarging cracks and coarser materials in the ground (e.g., in areas with loose gravel), more laser photons undergo diffuse reflection and attenuation within the pores, causing a collapse-like shift in the overall distribution towards lower intensity zones. At this point, the cumulative transported area calculated by integration increases rapidly, leading to a sharp rise in the intensity distribution difference. Therefore, the magnitude of the intensity distribution difference can directly and objectively quantify the degree of porosity degradation in the current area of ​​the road surface.

[0040] The intensity distribution difference of each local interval can be obtained by using the above methods.

[0041] By comparing the intensity distribution difference and the difference threshold of the corresponding local intervals in ascending order of index number, the index number of the first local region with an intensity distribution difference greater than the difference threshold is multiplied by the length of the horizontal sampling segment to obtain the boundary distance between cement and crushed stone at the current timestamp. The difference threshold is set as a multiple of the standard deviation of the baseline distribution. Specifically, if the difference in intensity distribution of all local intervals is less than or equal to the difference threshold, the boundary distance between cement and gravel is set to a preset distance. The preset distance is a large distance value, ensuring that in such open scenarios without obstacles or gravel threats, all lateral distances within the current time period are forcibly determined to be within the safe road surface range. In this embodiment, the preset distance is 99,999 meters.

[0042] Step S3: Based on the positional distribution of path points on each candidate trajectory of the robot, determine the lateral offset distance of the vehicle corresponding to each path point; based on the lateral offset distance of the vehicle and the boundary distance, obtain the wheel slip penalty value of each path point on each candidate trajectory.

[0043] Because once the outer tires of the substation inspection robot's chassis cross the boundary of a flat cement road surface and press into a loose gravel area, the friction coefficient between the left and right wheels and the ground will change abruptly, which can easily cause the wheels to lose traction and cause the odometer track calculation to diverge. Therefore, it is necessary to dynamically transform the static boundary distance and local porosity (i.e. roughness) into a mathematical model, with the vehicle's lateral offset distance as a continuous independent variable, to construct a piecewise penalty mathematical model that exhibits an exponential increase when approaching the dangerous boundary. This serves as the quantitative basis for limiting the transient torque of the wheels by the underlying drive.

[0044] Before inspecting the substation inspection area, the robot pre-plans multiple alternative trajectories. Conventional local path planners, when guiding the robot around obstacles left in corridors, are prone to generating dangerous curves that lead the outer tires of the vehicle into gravel areas. In order to cut off such infeasible paths during the geometric planning stage, it is necessary to quantify and filter the slippage of each alternative trajectory to select the optimal passage.

[0045] The following explanation uses one alternative trajectory as an example. The method provided in this embodiment can be used to explain other alternative trajectories.

[0046] For any candidate trajectory: Project the three-dimensional spatial coordinates of each path point on the candidate trajectory onto the lateral projection coordinate axis to obtain the absolute coordinate values ​​of each path point on the lateral projection coordinate axis, and use the absolute coordinate values ​​as the lateral offset distance of the vehicle corresponding to each path point.

[0047] Record any path point on the candidate trajectory as the path point to be analyzed.

[0048] The difference between the distance between the cement and gravel boundary and the safety tire half-width constant is recorded as the first difference. When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is less than or equal to the first difference, it indicates that even considering the physical width of the tire, the contact surfaces of the robot's wheels on both sides are completely within the safe and flat cement hardened road surface. In this case, the wheel slippage penalty value of the path point to be analyzed is set as the basic penalty constant. When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is greater than or equal to the distance between the cement and gravel boundary, it indicates that the wheel has completely crossed the cement boundary and completely pressed into the gravel area with a very low coefficient of friction. In this case, the wheel slippage penalty value of the path point to be analyzed is set as the truncation penalty constant. The basic penalty constant is less than the truncation penalty constant. The basic penalty constant is a preset, extremely small positive number greater than 0. In this embodiment, the basic penalty constant is 1, and the truncation penalty constant is 255.

[0049] When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is greater than the first difference and less than the boundary distance between cement and gravel, the wheel slip penalty value of the path point to be analyzed is determined based on the difference between the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed and the first difference, and the pavement porosity of the local interval where the path point to be analyzed is located. As a specific implementation method, the specific calculation formula of the wheel slip penalty value of the path point to be analyzed in this case is given: in, This represents the wheel slip penalty value at the path point to be analyzed. Represents the basic penalty constant. This represents the cost amplification factor used to adjust the sensitivity of punishment. This represents the ratio of the intensity distribution difference within the local interval where the path point to be analyzed is located to the maximum reflection intensity. This represents the lateral offset distance of the vehicle corresponding to the path point to be analyzed. Indicates the absolute value sign. Indicates the distance between the cement and the crushed stone. This represents the safety tire half-width constant. Represents the natural constant.

[0050] The maximum reflection intensity is the upper limit of the reflection intensity range set by the lidar hardware.

[0051] The cost amplification factor ranges from [1.5, 3.0]. In this embodiment, it is set to 2.0. In specific applications, the implementer can adjust it according to the robot's weight and tire tread type. The base penalty constant is used to ensure that the entire exponential mapping curve starts smoothly from a non-zero safety base, preventing logical deadlocks where the multiplication of formulas results in zero.

[0052] The intensity distribution difference of the local interval where the path point to be analyzed is located represents the pavement porosity of the current transverse interval. This value plays the role of the gain slope in the power of the exponential function: the rougher the road surface in the local interval, the higher the intensity distribution difference, resulting in a more drastic increase in the penalty value. The penetration depth is used to characterize the absolute physical depth to which the outer edge of a wheel crosses the safety limit and intrudes into a hazardous material area.

[0053] The closer the lateral offset distance of the vehicle corresponding to the path point to be analyzed is to the boundary between cement and gravel, the deeper the penetration depth. At the same time, if the intensity distribution difference in the local area where the path point to be analyzed is located is greater, the product of the two will cause the exponential term to expand rapidly. The wheel slip penalty value of the path point to be analyzed shows a non-linear (exponential) steep increase as it approaches the danger zone. This violent penalty response can quantify the sharp increase in the risk of slippage with extreme sensitivity.

[0054] By using the above method, the wheel slip penalty value of each path point on each candidate trajectory can be obtained. The larger the wheel slip penalty value, the higher the porosity of the road surface at the corresponding path point, the smaller the static friction force that can be provided to the tire, and the higher the probability of the wheel spinning and slipping when it reaches that point.

[0055] Step S4: By comparing the cumulative results of wheel slip penalty values ​​of path points on different alternative trajectories, the optimal travel trajectory is determined; the upper limit of travel speed is determined using the wheel slip penalty values ​​of path points on the optimal travel trajectory; the robot inspects the substation inspection area based on the optimal travel trajectory and the upper limit of travel speed.

[0056] In step S3 of this embodiment, the wheel slip penalty value of each path point on each candidate trajectory is determined. Next, the comprehensive penalty value of each candidate path is determined based on the wheel slip penalty values ​​of all path points on each candidate path, and then the optimal travel path is selected from all candidate paths.

[0057] For any candidate trajectory: the sum of the wheel slip penalty values ​​of all path points on the candidate trajectory is taken as the total slip penalty value of the candidate trajectory. Using this method, the total slip penalty value for each candidate trajectory can be obtained. The candidate trajectory corresponding to the minimum total slip penalty value is taken as the optimal travel trajectory, and the robot will subsequently perform inspections along the optimal travel trajectory.

[0058] During the robot's inspection along the optimal path, when approaching the boundary between cement and gravel, if the underlying drive motor maintains excessive output torque during start-up or acceleration, the tire surface is highly susceptible to exceeding the static friction limit of the gravel, causing it to slip and spin instantly. Therefore, based on the real-time penalty magnitude of the robot's current position, the maximum allowable speed and acceleration of the motor are dynamically reduced to suppress slippage from the power source.

[0059] Specifically, the maximum value of the wheel slip penalty for all path points within the look-ahead window on the optimal travel trajectory is obtained. The product of this maximum value and the speed attenuation coefficient is used as the speed attenuation caused by the current road surface roughness. The speed attenuation coefficient is obtained by calculating the difference between the robot's base cruising speed and the absolute lower limit speed, and then using the ratio of this difference to the truncation penalty constant as the speed attenuation coefficient. The absolute lower limit speed is a speed set to prevent motor stalling and locking; in this embodiment, the absolute lower limit speed is 0.2 m / s. The look-ahead window extends forward from the robot's current chassis center along the optimal travel trajectory to a preset look-ahead distance, which can be set to 3.0 meters in front of the chassis. The difference between the base cruising speed and the speed decay is calculated. This difference represents the theoretically maximum safe speed allowed after deducting the decay from the base cruising speed. The maximum value between this difference and the absolute lower limit speed is used as the upper limit of the driving speed. The larger the wheel slip penalty value on the road ahead, the larger the calculated speed decay reduction, thus significantly lowering the final allowable speed limit threshold. Simultaneously, the outer boundary clamping function ensures that even if the penalty value reaches the maximum truncation constant, the speed limit threshold will only smoothly decrease to the absolute lower limit speed, thus avoiding chassis lock-up caused by the speed limit returning to zero. In this embodiment, the robot's base cruising speed is 1.0 m / s.

[0060] After determining the upper limit of the travel speed, this upper limit is sent as a hard intervention command to each motor driver inside the chassis. In particular, the underlying motor driver is equipped with a speed and acceleration smoothing controller. After receiving the upper limit of the travel speed, the driver forcibly clamps the robot's travel speed below the upper limit, and at the same time, proportionally reduces the upper limit of the transient acceleration output of the motor according to the attenuation ratio of the upper limit of the travel speed.

[0061] By strictly limiting the upper limit of transient acceleration, the transient peak electromagnetic torque generated by the motor during start-up or rapid acceleration is suppressed. This physical intervention mechanism ensures that the tangential traction force applied by the wheel to the ground is always less than the maximum static friction force under the current gravel porosity, fundamentally preventing the tire from slipping and spinning due to exceeding the physical friction limit, and maintaining the pure rolling premise required for accurate calculation of wheel odometers.

[0062] Thus, the method described above has enabled the use of robots for autonomous inspection of the area where unattended substations are located.

[0063] This embodiment acquires the point cloud reflection intensity in different local sections of the substation inspection area where the robot is currently located, and analyzes the differences between the reflection intensity distribution of each local section and the reference reflection intensity distribution of the cement road surface. The intensity distribution difference of each local section is quantified, reflecting the change in the physical properties of the road surface material. This provides a reliable basis for subsequently identifying the boundary distance between cement and gravel, avoiding the inaccurate road edge identification caused by height errors when relying solely on geometric distance or two-dimensional radar on inclined roads. Then, based on the intensity distribution difference of different local sections, the boundary distance between cement and gravel is determined. This boundary distance dynamically reflects the actual physical boundary between safe road surfaces and high-risk gravel areas in the current inspection area. Using the positional distribution of path points on each of the robot's alternative trajectories, the lateral offset distance of the vehicle at each path point is determined. This approach links the trajectory planning problem with the lateral position risk of the wheels. Based on the vehicle's lateral offset distance and boundary distance, a wheel slip penalty value is constructed for each path point on each candidate trajectory. This penalty value quantifies the risk of the robot's wheels slipping when traveling along that path point. By comparing the cumulative results of the wheel slip penalty values ​​for path points on different candidate trajectories, the optimal travel trajectory is selected. This avoids problems such as differential wheel spin and slippage caused by the outer wheel running into a loose gravel area, as well as the resulting divergence in wheel odometer trajectory calculation, chassis pose loss of control, and yaw collisions. The wheel slip penalty value of the path point on the optimal travel trajectory is used to determine the upper limit of the travel speed, enabling the robot's travel speed to be dynamically and adaptively adjusted according to the actual slip risk on the road ahead, preventing tire slippage and spin from the power source.

[0064] An example of an unmanned substation inspection robot: An embodiment of the present invention provides an unmanned substation inspection robot, including a memory and a processor. The processor executes a computer program stored in the memory to implement the above-described autonomous inspection method of the unmanned substation inspection robot. The autonomous inspection method of the unmanned substation inspection robot has been described in detail in the above-described embodiment of the autonomous inspection method of the unmanned substation inspection robot, so it will not be described in detail here.

[0065] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An autonomous inspection method for an unmanned substation inspection robot, characterized in that, The method includes the following steps: Obtain the point cloud reflection intensity in different local sections of the substation inspection area where the robot is currently located; Based on the difference between the point cloud reflection intensity and the reference reflection intensity distribution of the cement pavement in each local interval, the intensity distribution difference degree of each local interval is determined; based on the intensity distribution difference degree of different local intervals, the boundary distance between cement and crushed stone is determined. Based on the positional distribution of path points on each candidate trajectory of the robot, the lateral offset distance of the vehicle corresponding to each path point is determined; based on the lateral offset distance of the vehicle and the boundary distance, the wheel slip penalty value of each path point on each candidate trajectory is obtained. The optimal travel trajectory is determined by comparing the cumulative wheel slip penalty values ​​of path points on different alternative trajectories; the upper limit of travel speed is determined using the wheel slip penalty values ​​of path points on the optimal travel trajectory; the robot then inspects the substation inspection area based on the optimal travel trajectory and the upper limit of travel speed.

2. The autonomous inspection method for an unmanned substation inspection robot according to claim 1, characterized in that, The acquisition of point cloud reflection intensity in different local sections of the substation inspection area where the robot is currently located includes: Using the center projection point at the bottom of the robot chassis as the origin of the coordinate system and the opposite direction of the vertical vector of gravity as the plane normal, an orthogonal projection plane of gravity is constructed. The lateral feature abundance vector is projected onto the gravity orthogonal projection plane to obtain the lateral projection coordinate axis; A local vertical sampling window is set along the horizontal projection coordinate axis to obtain the current sampling area; Based on the location distribution of laser points within the current sampling area, the current sampling area is divided into multiple local intervals; Distance compensation and normalization were performed on the initial reflection intensity of the point cloud in each local interval to obtain the reflection intensity of the point cloud in each local interval.

3. The autonomous inspection method for an unmanned substation inspection robot according to claim 2, characterized in that, The acquisition of the lateral feature abundance vector includes: A nonlinear least squares registration algorithm is used to iteratively solve the current frame of lidar data, and the registration convergence confidence of the algorithm output is extracted. The registration optimization information matrix is ​​obtained only if the registration convergence confidence is greater than the preset registration success threshold. The information matrix is ​​subjected to eigenvalue decomposition, and the eigenvector corresponding to the eigenvalue with the largest value is extracted from the decomposition result as the lateral feature abundance vector.

4. The autonomous inspection method for an unmanned substation inspection robot according to claim 1, characterized in that, The determination of the intensity distribution difference degree in each local interval based on the difference between the point cloud reflection intensity and the reference reflection intensity distribution of the cement pavement in each local interval includes: The intensity distribution difference of candidate local intervals is obtained by using the cumulative distribution function of the reflection intensity distribution map of the candidate local intervals and the cumulative distribution function of the reflection intensity benchmark distribution map of the smooth cement pavement; wherein, the reflection intensity distribution map of the candidate local intervals is generated based on all point cloud positions and point cloud reflection intensities within the candidate local intervals. The candidate local interval is any local interval.

5. The autonomous inspection method for an unmanned substation inspection robot according to claim 2, characterized in that, The determination of the boundary distance between cement and crushed stone based on the difference in strength distribution in different local areas includes: Determine the index number of each local interval, and compare the intensity distribution difference degree and the difference degree threshold of the corresponding local intervals in ascending order of index number; Multiply the index number of the first local interval where the intensity distribution difference is greater than the difference threshold by the length of the horizontal sampling segment to obtain the boundary distance between cement and crushed stone. If the intensity distribution difference in all local intervals is less than or equal to the difference threshold, then the boundary distance between cement and crushed stone is set to a preset distance.

6. The autonomous inspection method for an unmanned substation inspection robot according to claim 2, characterized in that, The determination of the lateral offset distance of the vehicle corresponding to each path point, based on the positional distribution of path points on each candidate trajectory of the robot, includes: For any alternative trajectory: The three-dimensional spatial coordinates of each path point on any candidate trajectory are projected onto the lateral projection coordinate axis to obtain the absolute coordinate values ​​of each path point on the lateral projection coordinate axis. The absolute coordinate values ​​are used as the lateral offset distance of the vehicle corresponding to each path point.

7. The autonomous inspection method for an unmanned substation inspection robot according to claim 1, characterized in that, The step of obtaining the wheel slip penalty value for each path point on each candidate trajectory based on the vehicle's lateral offset distance and the boundary distance includes: Calculate the first difference between the boundary distance and the safety tire half-width constant; When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is less than or equal to the first difference, the wheel slip penalty value of the path point to be analyzed is set as the basic penalty constant. When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is greater than the first difference and less than the boundary distance, the wheel slip penalty value of the path point to be analyzed is determined based on the difference between the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed and the first difference, and the pavement clearance ratio of the local interval where the path point to be analyzed is located. When the absolute value of the lateral offset distance of the vehicle corresponding to the path point to be analyzed is greater than or equal to the boundary distance, the wheel slip penalty value of the path point to be analyzed is set to the truncation penalty constant. The base penalty constant is less than the truncation penalty constant; The path point to be analyzed is any path point on any candidate trajectory.

8. The autonomous inspection method for an unmanned substation inspection robot according to claim 1, characterized in that, The process of determining the optimal travel trajectory by comparing the cumulative results of wheel slip penalty values ​​at path points on different candidate trajectories includes: For any candidate trajectory: the sum of the wheel slip penalty values ​​of all path points on the candidate trajectory is taken as the total slip penalty value of the candidate trajectory; The candidate trajectory corresponding to the minimum total slip penalty value is taken as the optimal travel trajectory.

9. The autonomous inspection method for an unmanned substation inspection robot according to claim 1, characterized in that, The method of determining the upper limit of driving speed using the wheel slip penalty value of the path point on the optimal travel trajectory includes: The speed reduction caused by the current road surface roughness is obtained by multiplying the wheel slip penalty value of the path point on the optimal travel trajectory with the speed reduction coefficient. The maximum value between the difference between the base cruising speed and the speed decay, and the absolute lower limit speed, is taken as the upper limit of driving speed.

10. An unmanned substation inspection robot, comprising a memory and a processor, characterized in that, The processor executes the computer program stored in the memory to implement the method of claim 1.