Cable tunnel patrolling method, system, device and medium based on quadruped robot

Abstract

The application discloses a kind of cable tunnel patrol method, system, equipment and medium based on four-legged robot.The method of the application comprises: real-time acquisition of multi-source data;Multi-source data is time-synchronized and abnormally handled, extracts environmental feature points and parameters required by particle filtering, constructs structured fusion dataset supporting positioning and obstacle avoidance algorithm;Based on structured fusion dataset, combined with tunnel environment benchmark data, estimate robot real-time pose using fusion positioning algorithm, and generate the coordinate sequence of future patrol path;Based on the motion state data of the obstacle, predict the future trajectory of the obstacle, and perform spatio-temporal overlap analysis with the coordinate sequence of the robot's future patrol path, identify potential collision risks, adjust the robot's movement speed according to the risk level and trigger the corresponding obstacle avoidance response.The application improves positioning accuracy through multi-source data fusion and trajectory prediction, realizes active obstacle avoidance, reduces collision risk and adapts to complex tunnel environment.

Description

Technical Field

[0001] This invention relates to the field of intelligent robot control technology, specifically to a method, system, equipment, and medium for cable tunnel inspection based on a quadruped robot. Background Technology

[0002] Cable tunnel inspection is a crucial step in ensuring the safe operation of power systems. Currently, using intelligent robots to replace manual labor for automated inspections has become the mainstream trend.

[0003] However, the long, narrow, repetitive, and feature-sparse environment of cable tunnels poses a significant challenge to the long-term stable operation of robots. This is mainly reflected in two aspects: First, in terms of localization, localization methods relying on environmental feature matching are prone to accumulating errors in the repetitive and feature-sparse environment of tunnels, leading to drift in localization results and compromising long-term reliability. Simultaneously, probabilistic localization methods often suffer from particle degradation or divergence in monotonous environments, resulting in localization failures. Second, in terms of obstacle avoidance, existing methods are mostly "reactive obstacle avoidance," meaning they only brake or detour after detecting an obstacle. This approach cannot effectively handle suddenly appearing dynamic obstacles, exhibiting reaction lag and collision risks, and lacking sufficient intelligence. Current technologies struggle to simultaneously achieve high-precision localization and active safety obstacle avoidance in cable tunnels, limiting the reliability and widespread application of inspection robots.

[0004] Therefore, developing a robotic system capable of overcoming the challenges of the tunnel environment and achieving precise positioning, forward prediction, and active obstacle avoidance is of great practical significance and application value for improving the level of intelligent inspection of cable tunnels and ensuring operational safety. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, device and medium for cable tunnel inspection based on a quadruped robot, so as to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution.

[0007] In a first aspect, the present invention provides a cable tunnel inspection method based on a quadruped robot, comprising:

[0008] Step S1: During the inspection of the cable tunnel, the robot collects multi-source data in real time. The multi-source data includes: geometric feature data of obstacles in the tunnel, motion state data, robot's own path deviation data, and tunnel environment baseline data.

[0009] Step S2: Perform time synchronization and anomaly processing on the multi-source data, extract environmental feature points and parameters required for particle filtering, and construct a structured fusion dataset that supports localization and obstacle avoidance algorithms;

[0010] Step S3: Based on the structured fusion dataset and combined with the tunnel environment reference data, a fusion positioning algorithm is used to estimate the robot's real-time pose and generate a coordinate sequence for the future inspection path.

[0011] Step S4: Based on the motion state data of the obstacle, predict the future trajectory of the obstacle, perform spatiotemporal overlap analysis with the coordinate sequence of the robot's future inspection path, identify potential collision risks, adjust the robot's movement speed according to the risk level, and trigger the corresponding obstacle avoidance response.

[0012] Further, step S1 includes:

[0013] Step S1-1: Establish a dual coordinate system. A global coordinate system is established with the preset reference mark as the origin, and the direction extending along the tunnel axis is denoted as... The axis, perpendicular to the horizontal direction of the tunnel axis, is denoted as... An axis, vertically upward, is denoted as Axis; establish a local coordinate system with the robot's center as the origin, and denote the direction of robot movement as . The axis, the horizontal axis perpendicular to the robot's direction of movement, is denoted as... An axis, vertically upward, is denoted as axis;

[0014] Step S1-2: Generate 3D point cloud data in the local coordinate system by scanning the tunnel environment in real time using LiDAR. The coordinates are transformed and mapped to the global coordinate system; visual images are acquired simultaneously, obstacle depth information is calculated and converted into three-dimensional coordinates in the global coordinate system. Extract the outline dimensions (L, W, H) and depth information d of the obstacle; where L is the length of the obstacle along the tunnel axis, W is the width perpendicular to the axis, and H is the height;

[0015] Steps S1-3: Obtain the robot's angular velocity in real time. Linear acceleration a and attitude angle ,in For roll angle, For pitch angle, For heading angle; simultaneously acquire the instantaneous velocity v, direction angle α, and acceleration of dynamic obstacles. ;

[0016] Step S1-4: Based on the preset baseline path Real-time acquisition of the robot's current position P Lateral deviation from the preset baseline path and longitudinal deviation ,in For P to vertical distance, , Accumulate mileage along the preset baseline path from the current location. This is the mileage corresponding to the point closest to P on the baseline path;

[0017] Steps S1-5: When the robot moves to the detection range of the preset reference mark, acquire the mark detection data, including: the absolute coordinates of the mark in the global coordinate system. Identity identification The reliability parameters of this test are also included, and the preset benchmark marks are set at intervals along the tunnel path.

[0018] Further, step S2 includes:

[0019] Step S2-1: Use the robot's local clock as the base time. Obtain the collection timestamps of each source data. By calculating the time deviation between each timestamp and the reference time Based on this deviation, linear interpolation is used to map data from different sampling times to a unified time series. , where k=1,2,...,n, to synchronize multi-source data in the time dimension;

[0020] Step S2-2: For the synchronized obstacle geometric feature data, calculate the feature deviation between adjacent frames. Where F is the feature vector, containing the three-dimensional coordinates of the obstacle in the global coordinate system. Contour dimensions (L, W, H) and depth information d; when δ exceeds a preset threshold The data is judged as abnormal at that time; for motion state data, outliers are identified through reasonableness checks.

[0021] Step S2-3: For detected abnormal data, if it is an isolated point, the neighborhood mean is used as a substitute; if it is a continuous anomaly, complementary data from the backup data source is used to fill it in, and the data confidence level c is marked.

[0022] Step S2-4: According to the uniform time series After processing, each data record is denoted as... Where k is the time point index, with values ​​1, 2, ..., n. Represents a complete data record at the k-th time point; includes the field: obstacle feature set. Robot motion parameters Path deviation Tunnel environmental benchmark information And data confidence level c, forming a structured fusion dataset with spatiotemporal correlation. .

[0023] Furthermore, in step S2-2, outliers in the motion state data are identified through a rationality check. Specifically, this includes checking whether the absolute values ​​of the robot's linear acceleration 'a' and angular velocity 'ω' meet the following requirements. and ,in and These are the linear acceleration threshold and the angular velocity threshold, respectively; acceleration of dynamic obstacles. Check whether its absolute value satisfies , This is the acceleration threshold for dynamic obstacles.

[0024] Further, step S3 includes:

[0025] Step S3-1: Based on the robot motion parameters in the structured fusion dataset D and path deviation The initial pose of the robot is calculated using a kinematic model. ,in The initial 3D coordinates in the global coordinate system. The initial attitude angle is... The initial roll angle, The initial pitch angle, This is the initial heading angle;

[0026] Step S3-2: Combine the absolute coordinates marked in the global coordinate system from the tunnel environment reference data. When the robot detects a reference marker, it uses the absolute coordinates of the marker as the reference marker constraint and performs a weighted fusion algorithm to determine the initial pose. Initial three-dimensional coordinates and initial attitude angle The correction is performed, where the weights of the initial pose are determined based on the reliability of the motion parameter data. The weights of the baseline marker constraints are determined based on the reliability of the marker detection data. The corrected real-time pose is determined, and the sum of the two weights is 1. ;

[0027] Step S3-3: Based on real-time pose And the preset global inspection path, within the future time window T, in order to three-dimensional position Starting from the global preset baseline path The extension direction is selected sequentially at preset spatial intervals Δs, and the range of selected points is limited to a future time window T, generating coordinate points. The coordinates are in the global coordinate system, where t is the robot path time step index; all points must satisfy the deviation constraint from the global path and the distance between adjacent points must not exceed Δs, ultimately forming the coordinate sequence of the future inspection path. .

[0028] Further, step S4 includes:

[0029] Step S4-1: Based on obstacle motion state data in structured fusion dataset D A kinematic prediction model is used to generate a sequence of trajectory coordinates of obstacles within a future time window T. ,in This is an obstacle trajectory time step index, which corresponds one-to-one with the robot path time step index; where For obstacles at any time The three-dimensional coordinates of the trajectory points are consistent with the robot path coordinate sequence;

[0030] Step S4-2: Sequence of future trajectory coordinates of obstacles With the robot's future inspection path coordinate sequence Perform spatiotemporal matching to calculate the spatial distance at corresponding times. and time difference ,in For the robot to arrive At that moment, For the obstacle to reach The moment; Step S4-3: When and ≤ At that time, it was determined that there was a potential collision risk, among which As a space safety threshold, As a time safety threshold; and according to and Calculating risk level based on actual value ,in , The weighting coefficients and ;

[0031] Step S4-4: Execute the corresponding obstacle avoidance strategy according to the risk level R: when When, it triggers emergency braking and generates an alternative route; when When, adjust the robot's speed and direction to increase the safe distance; when At the same time, maintain the current path and continuously monitor the movement status of obstacles; among which Indicates a high-risk threshold. This indicates a low-risk threshold.

[0032] Furthermore, the method for adjusting the robot's speed and direction to increase the safe distance in step S4-4 includes:

[0033] Step S4-4-1: Based on the future trajectory coordinate sequence of obstacles With the robot's future path coordinate sequence Calculate the lateral offset direction of the obstacle relative to the robot, the lateral offset direction being along the global coordinate system. The positive and negative directions of the axis; and the minimum safe distance difference. ;

[0034] Step S4-4-2: Dynamically reduce the robot's cruising speed according to the risk level R, and adjust the speed accordingly. ,in Original cruising speed;

[0035] Step S4-4-3: Generate a deflection angle in the lateral offset direction away from the obstacle. ,in , This is a scaling factor that makes the robot path... The axis is offset, and the minimum distance between the offset path and the future trajectory of the obstacle is greater than or equal to... ;

[0036] Step S4-4-4: Real-time verification of whether the distance between the adjusted path and the tunnel wall is greater than or equal to the preset boundary threshold. If not, prioritize following the path. Adjust the path in the opposite direction of the axis.

[0037] In a second aspect, the present invention provides a cable tunnel inspection system based on a quadruped robot for performing the above-mentioned cable tunnel inspection method based on a quadruped robot, which includes: a data acquisition module, a data processing module, a pose estimation and path generation module, and an obstacle avoidance decision module.

[0038] The data acquisition module is used to collect multi-source data in real time during the robot's inspection of cable tunnels. The multi-source data includes: geometric feature data of obstacles in the tunnel, motion state data, robot path deviation data, and tunnel environmental baseline data.

[0039] The data processing module is used to perform time synchronization and anomaly processing on the multi-source data, extract environmental feature points and parameters required for particle filtering, and construct a structured fusion dataset that supports localization and obstacle avoidance algorithms.

[0040] The pose estimation and path generation module is used to estimate the robot's real-time pose based on the structured fusion dataset and combined with tunnel environment reference data, using a fusion positioning algorithm, and to generate a coordinate sequence for the future inspection path.

[0041] The obstacle avoidance decision module is used to predict the future trajectory of obstacles based on their motion state data, perform spatiotemporal overlap analysis with the coordinate sequence of the robot's future patrol path, identify potential collision risks, adjust the robot's movement speed according to the risk level, and trigger the corresponding obstacle avoidance response.

[0042] Thirdly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described cable tunnel inspection method based on a quadruped robot.

[0043] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described cable tunnel inspection method based on a quadruped robot.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] 1. This invention employs a fusion localization scheme combining an improved SLAM algorithm and an improved AMCL algorithm. The improved SLAM optimizes feature point matching by incorporating prior information about the tunnel structure, reducing map drift caused by environmental noise. The improved AMCL dynamically adjusts particle filter parameters and combines them with reference marker constraints to enhance global localization stability. After confidence-weighted fusion, the localization error can be controlled within ±5mm, completely solving the localization drift problem of traditional single algorithms in tunnel environments and providing a precise pose foundation for path planning and obstacle avoidance.

[0046] 2. By generating the future trajectory of obstacles in advance through a kinematic prediction model and combining it with the robot's future path for spatiotemporal matching, the obstacle avoidance logic is upgraded from "real-time response" to "advance prediction". Based on the risk classification mechanism of spatial distance and time difference, high, medium and low collision risks can be accurately identified, and corresponding differentiated strategies such as emergency braking, speed adjustment and directional deviation can be executed. The accuracy of collision risk identification is improved and the incidence of collision accidents is reduced.

[0047] 3. A full-process data processing mechanism of "time synchronization - anomaly detection - data repair - confidence labeling" is constructed. Linear interpolation is used to solve the asynchronous problem of multi-source data, and neighborhood mean and backup data are used to fill in and repair outliers. Data quality is quantified using a confidence parameter. This mechanism effectively filters noise interference from sensors such as LiDAR and vision, improves the effective data rate of structured fusion datasets, and provides high-quality input for localization and obstacle avoidance algorithms. Attached Figure Description

[0048] Figure 1 This is a flowchart of a cable tunnel inspection method based on a quadruped robot according to the present invention;

[0049] Figure 2 This is a diagram illustrating the composition of a cable tunnel inspection system based on a quadruped robot according to the present invention.

[0050] Figure 3 This is a diagram showing the composition of the obstacle avoidance strategy execution unit in the cable tunnel inspection system of the present invention;

[0051] Figure 4 This is a schematic diagram of the logical structure of a computer device provided in Embodiment 4 of the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1

[0054] like Figure 1 As shown in the figure, this embodiment provides a cable tunnel inspection method based on a quadruped robot, the steps of which are as follows:

[0055] Step S1: During the inspection of the cable tunnel, the robot collects multi-source data in real time. The multi-source data includes: geometric feature data of obstacles in the tunnel, motion state data, robot's own path deviation data, and tunnel environment baseline data.

[0056] Step S2: Perform time synchronization and anomaly processing on the multi-source data, extract environmental feature points and parameters required for particle filtering, and construct a structured fusion dataset that supports localization and obstacle avoidance algorithms;

[0057] Step S3: Based on the structured fusion dataset and combined with the tunnel environment reference data, the fusion positioning algorithm is used to estimate the robot's real-time pose and generate a coordinate sequence for the future inspection path;

[0058] Step S4: Based on the motion state data of the obstacle, predict the future trajectory of the obstacle, perform spatiotemporal overlap analysis with the coordinate sequence of the robot's future patrol path, identify potential collision risks, adjust the robot's movement speed according to the risk level, and trigger the corresponding obstacle avoidance response.

[0059] Specifically, step S1 includes:

[0060] Step S1-1: Establish a global coordinate system with the first preset reference mark at the tunnel entrance as the origin. The reference mark is marked with a QR code and the positioning accuracy is ±2mm. The axis extends along the tunnel axis, with the positive direction being the direction of inspection and advancement. The axis is perpendicular to the horizontal direction of the tunnel axis, with the positive direction being the right side of the tunnel; The axis points vertically upwards; a local coordinate system is established with the geometric center of the robot body as the origin. The axis is along the robot's direction of travel. The axis is perpendicular to the horizontal direction of the robot's movement. The axis is vertically upward; the two coordinate systems are mapped through a preset transformation matrix, and the transformation error is controlled within 0.5mm;

[0061] Steps S1-2: The robot is equipped with a 16-line LiDAR, with a scanning frequency of 10Hz and a ranging range of 0.1-100m, to scan the tunnel environment in real time and generate 3D point cloud data in a local coordinate system. The transformation matrix established in step S1-1 is mapped to the global coordinate system; a high-definition industrial camera with a resolution of 1920×1080 and a frame rate of 20fps is simultaneously started to acquire visual images. The depth information of obstacles is calculated using a binocular vision ranging algorithm, and the LiDAR data is cross-validated and converted into three-dimensional coordinates in the global coordinate system. Extract the obstacle's contour dimensions (L, W, H) and depth information d from the point cloud data, where L is the obstacle's distance along the path. The length along the axial direction, W is the length along the axis. Width along the axial direction, H is the width along the axis. The height along the axis, d is the straight-line distance between the obstacle and the robot's current position; when maintenance personnel are detected in the tunnel, their outline dimensions are L=0.6m, W=0.4m, H=1.7m, depth information d=3.2m, and global coordinates are (15.2m, 0.3m, 0.9m);

[0062] Steps S1-3: Using the robot's built-in IMU sensor, with a sampling frequency of 100Hz, an acceleration measurement range of ±16g, and an angular velocity measurement range of ±2000° / s, the robot's angular velocity ω, linear acceleration a, and attitude angle are acquired in real time. φ is the roll angle, ranging from -180° to 180°; θ is the pitch angle, ranging from -90° to 90°; ψ is the heading angle, ranging from 0° to 360°; at a certain moment, the robot's angular velocity ω = 0.5° / s, linear acceleration a = 0.2 m / s², and attitude angles φ = 0.3°, θ = 0.1°, and ψ = 1.2° were collected; based on the inter-frame matching algorithm of LiDAR point cloud, the motion state data of dynamic obstacles were extracted synchronously: instantaneous velocity v, motion direction angle α, and acceleration. The instantaneous velocity of the aforementioned maintenance personnel is v = 0.5 m / s, the direction angle of motion is α = 185°, and the acceleration is... =0.1m / s²;

[0063] Step S1-4: Preset the baseline path To create a smooth curve in a global coordinate system based on tunnel CAD drawings, the robot motion controller acquires the current position P in real time. and The deviation parameter. Lateral deviation. For P to vertical distance, along Axial direction; longitudinal deviation , For the current position along The cumulative mileage is the distance from the starting point of the path to the projection point of P on the path; for The mileage corresponding to the nearest point to P; the global coordinates of the robot's current position P are (12.5m, 0.1m, 0.8m), and the mileage corresponding to the nearest point on the baseline path. =12.4m, current location, cumulative mileage along the baseline path =12.6m, then the lateral deviation is =0.1m, longitudinal deviation =|12.6-12.4|=0.2m;

[0064] Steps S1-5: Place a 20cm x 20cm QR code reference marker every 5m along the tunnel path. The code includes the marker ID and absolute coordinates. The robot is equipped with a QR code recognition camera with a recognition distance of 0.5-3m and a recognition accuracy of ±1mm. When the robot moves into the marker detection range, it automatically triggers recognition and acquires marker detection data. The marker detection data includes the absolute coordinates of the marker in the global coordinate system. Identity identification The reliability parameter for this detection is as follows: When the reference marker with ID M10 is detected, its absolute coordinates are (20.0m, 0.0m, 0.0m), and the detection reliability parameter is 0.98.

[0065] Specifically, step S2 includes:

[0066] Step S2-1: Use the robot's local clock as the base time. With an accuracy of 1ms, it reads the acquisition timestamps from various data sources such as LiDAR, cameras, and IMU. Calculate time deviation For data with different sampling frequencies (100Hz for IMU and 20Hz for camera), linear interpolation is used to map all data to a unified time series. k=1,2,...,n, with a time interval Δt=0.05s to ensure that multi-source data are synchronized in the time dimension; the data collected by the lidar at t=10.02s and the data collected by the camera at t=10.00s are interpolated and mapped to the same time points such as t=10.00s and 10.05s.

[0067] Step S2-2: Construct a feature vector F from the synchronized obstacle geometric feature data, containing... ,L,W,H,d, calculate the feature deviation between adjacent frames Preset threshold =0.5m, when The time was determined to be abnormal; the feature vectors of the obstacles in two adjacent frames were respectively =(15.2m,0.6m,0.9m,0.6m,0.4m,1.7m,3.2m)、

[0068] =(15.3m, 0.7m, 0.9m, 0.6m, 0.4m, 1.7m, 3.1m), calculated to... ≈0.17m< =0.5m, judged as normal data; a certain frame was affected by point cloud noise. = (15.2m, 1.0m, 0.9m, 0.6m, 0.4m, 1.7m, 3.2m), then δ = 0.6m> The data was deemed abnormal; the motion state data was validated for reasonableness: robot linear acceleration threshold. =2 angular velocity threshold =100° / s; Dynamic obstacle acceleration threshold =1.5 When the detected value exceeds the corresponding threshold, it is judged as abnormal; the robot's linear acceleration a = 2.5 was collected. Exceeding =2 This is considered abnormal.

[0069] Step S2-3: For detected abnormal data, if it is an isolated point (i.e., a single-frame anomaly), the neighborhood mean of the three adjacent frames is used as the replacement; if it is a continuous anomaly (i.e., anomalies in three or more frames), complementary data from a backup data source is used to fill the gap; the linear acceleration a = 2.5. This is an isolated anomaly; the linear accelerations of the three adjacent frames are 0.2. 0.3 0.2 Using the mean ≈0.23 Replace the outlier; assign a confidence level c to the repaired data, ranging from 0 to 1; the confidence level of the motion parameter data is denoted as... The confidence level of the labeled detection data is denoted as The more reliable the data, i.e., the fewer repair attempts and the higher the accuracy of the source data, the greater the confidence value; the confidence value of unrepaired robot angular velocity data. =0.95, confidence level of linear acceleration data after neighborhood mean replacement. =0.75, the confidence level of the benchmark marker detection data with ID M10 above. =0.98;

[0070] Step S2-4: By time series The processed data, each data record Includes the following fields: obstacle feature set Robot motion parameters Path deviation Environmental benchmark information and data confidence level c; data confidence level includes , To form a structured dataset with spatiotemporal correlation. The data is stored in the robot's local database and updated in real time. Data record at 10.00s The obstacle feature set is: {(15.2m, 0.6m, 0.9m, 0.6m, 0.4m, 1.7m, 3.2m)}.

[0071] Robot motion parameters (0.5° / s, 0.23m / s², 0.3°, 0.1°, 1.2°), path deviation (0.1m, 0.2m), environmental reference marker information {(20.0m, 0.0m, 0.0m, M10)}, data confidence level ( =0.85, =0.98).

[0072] Specifically, step S3 includes:

[0073] Step S3-1: Based on the robot motion parameters in the fusion dataset D and path deviation The initial pose of the robot is calculated using a kinematic model:

[0074] , The initial three-dimensional position in the global coordinate system is calculated by integrating the pose, linear acceleration, and angular velocity from the previous moment. The initial attitude angle, To bypass The initial roll angle of the axis rotation, To bypass The initial pitch angle of the axis rotation. To bypass The initial heading angle of the axis rotation; based on the previous pose (12.3m, 0.1m, 0.8m, 0.2°, 0.1°, 1.1°), with a time interval Δt = 0.05s, combined with the current motion parameters ω = 0.5° / s and a = 0.23. The displacement increment was calculated. ≈0.00029m, heading angle increment =0.5×0.05=0.025°, thus obtaining the initial pose. =(12.5m,0.1m,0.8m,0.3°,0.1°,1.2°);

[0075] Step S3-2: Combine the marked coordinates from the tunnel environmental reference data When the robot detects a reference marker, it uses the absolute coordinates of the marker as the reference marker constraint and performs a weighted fusion algorithm to determine the initial pose. Initial three-dimensional coordinates and initial attitude angle The correction is performed, where the weights of the initial pose are determined based on the reliability of the motion parameter data. The weights of the baseline marker constraints are determined based on the reliability of the marker detection data. It is determined that the sum of the two weights is 1; the weight calculation formula is: the initial pose weight is The baseline marker constraint weight is The fusion localization algorithm is built upon improved SLAM and AMCL algorithms. The improved SLAM algorithm optimizes feature point matching by introducing prior information about the tunnel structure, reducing environmental noise interference with map construction and improving local positioning accuracy, keeping local positioning errors within ±3mm. The improved AMCL algorithm dynamically adjusts the number of particles in the particle filter and the resampling strategy, combining this with the absolute coordinates of the reference marker to achieve global positioning correction. The number of particles is dynamically adjusted between 500 and 2000 based on environmental complexity, and the error can be reduced to ±1mm after global positioning correction. The reliability of the current motion parameter data is also assessed. =0.85, reliability of benchmark marker detection data =0.98, the initial pose weights are calculated using weighting. ≈0.46, the baseline mark constraint weight is The initial pose X-coordinate is 12.502m, the reference marker X-coordinate is 12.500m, and the corrected X-coordinate is 12.502×0.46+12.500×0.54=5.75092+6.75=12.50092m≈12.5m. The optimized real-time pose is thus obtained. =(12.5m,0.1m,0.8m,0.3°,0.1°,1.2°), with positioning error controlled within ±5mm.

[0076] Step S3-3: Based on real-time pose and the preset global inspection path, which is related to Consistent, within the future time window T=3s, with Location Starting from (12.5m, 0.1m, 0.8m), along... The extension direction is determined by taking points sequentially at preset spatial intervals Δs = 0.1m. The range of points is limited to T, corresponding to a theoretical driving distance L = T × =0.5m / s, calculated L=3×0.5=1.5m, that is, the total length of the point is 1.5m; the generated coordinate points The coordinates are the three-dimensional position coordinates in the global coordinate system, where t is the robot path time step index; all points must meet the deviation constraints from the global path, with a lateral deviation less than or equal to 0.2m, a longitudinal deviation less than or equal to 0.1m, and the distance between adjacent points not greater than Δs = 0.1m; the first point =(12.6m, 0.1m, 0.8m), the second point =(12.7m, 0.1m, 0.8m), and so on, to ultimately form the coordinate sequence of the future inspection path. The sequence contains 15 coordinate points, covering the planned path within the next 3 seconds, for subsequent collision risk analysis.

[0077] Specifically, step S4 includes:

[0078] Step S4-1: Based on obstacle motion state data in fusion dataset D Using a kinematic prediction model, i.e., a uniformly accelerated linear motion model, a sequence of trajectory coordinates of the obstacle is generated within a future time window T=3s. First, the motion parameters in the local coordinate system are transformed to the global coordinate system to obtain the velocity components. and acceleration components The formula for calculating the velocity component is: =v×cosα, =v×sinα; the formula for calculating the acceleration components is: = ×cosα, = ×sinα; where For obstacles at any time The three-dimensional coordinates of the trajectory points are consistent with the robot's path coordinate sequence (0.2s); for the aforementioned maintenance personnel, their motion state data are v=0.5m / s, α=185°, =0.1 The calculations yield cos185°≈-0.9962, sin185°≈-0.0872, therefore =0.5×(-0.9962)≈-0.4981m / s, =0.5×(-0.0872)≈-0.0436m / s, =0.1×(-0.9962)≈-0.099 , =0.1×(-0.0872)≈-0.0087 Starting from the current global coordinates of the obstacle (15.2m, 0.6m, 0.9m), and with a time step of 0.2s consistent with the time interval of the robot's path coordinate sequence, the coordinates at future moments are calculated. =(15.039m, 0.591m, 0.9m); and so on, the trajectory coordinates from t=1 to t=15 are calculated to form a sequence. ;

[0079] Step S4-2: Sequence of future trajectory coordinates of obstacles With the coordinate sequence of the robot's future inspection path Perform spatiotemporal matching, targeting the corresponding time. , =1,2,…,15, calculate spatial distance That is, Euclidean distance and time difference. ; For the robot to arrive At that moment, For the obstacle to reach At that moment, because their time steps are consistent, = =t×0.2s, =10.1s + t × 0.2s, therefore =0.1s; when t=10, When =10, robot path coordinates =(13.5m, 0.1m, 0.8m), obstacle trajectory coordinates =(14.0m, 0.5m, 0.9m), calculate the spatial distance. ≈0.648m;

[0080] Step S4-3: Preset space safety threshold =0.5m, time safety threshold =0.3s, when and ≤ At that time, a potential collision risk was determined; the above t=10, When =10, =0.648m>0.5m, If the time interval is 0.1s ≤ 0.3s, a potential collision risk is still identified; a weighted average risk level is used to calculate the risk level. ,in =0.6、 =0.4, weighted sum is 1; substituting the data, we get R=0.6×(0.5 / 0.648)+0.4×(0.3 / 0.1)=0.6×0.772+0.4×3≈0.463+1.2=1.663; preset high-risk threshold =1.8, low-risk threshold =1.2, current =1.2 <R=1.663< =1.8, classified as medium risk;

[0081] Step S4-4: When In high-risk situations, emergency braking is triggered when the braking acceleration is greater than or equal to 2. Simultaneously, based on the improved SLAM algorithm, a local environment map is constructed to generate a detour path. The minimum distance between the detour path and obstacles is greater than or equal to 0.8m. In the aforementioned high-risk scenario, the robot immediately initiates emergency braking, decelerating from a speed of 0.5m / s to a stop with a braking distance of approximately 0.06m. At the same time, a detour path coordinate sequence is generated and directed towards... The axis is offset by 0.4m in the negative direction, and the starting point of the detour path is (12.5m, -0.3m, 0.8m);

[0082] when In cases of medium risk, adjust speed and direction to increase safe distance using the following methods:

[0083] Step S4-4-1: Calculate the lateral offset direction of the obstacle relative to the robot and the minimum safe distance difference Δd. The lateral offset direction is along the global coordinate system. The positive and negative directions of the axes are used to calculate the lateral offset direction of the obstacle relative to the robot and the minimum safe distance difference Δd; the lateral offset direction is along the global coordinate system. The positive and negative directions of the axis are determined by comparing the difference between the Y-coordinate of the obstacle and the robot; the obstacle's Y-coordinate is 0.5m, and the robot's Y-coordinate is 0.1m. The lateral offset direction is Positive axis direction; minimum safe distance difference =0.5-0.648=-0.148m, a negative value indicates that the current distance is slightly greater than the safety threshold, and the distance needs to be maintained and increased;

[0084] Step S4-4-2: Dynamically reduce cruise speed, adjusted speed v0 = 0.5 m / s, R = 1.663 =1.2, =1.8, =0.5×0.228≈0.114m / s;

[0085] Step S4-4-3: Generate a deflection angle in the lateral offset direction away from the obstacle. , , =0.8rad / m, making the path... Offset along the axis, ensuring that the minimum distance between the offset and the obstacle's trajectory is greater than or equal to... ; =0.8 × 0.148 ≈ 0.118 rad ≈ 6.76°, path direction The robot's Y-coordinate shifts in the negative direction, resulting in a value of 0.1 - 0.148 × 0.5 = 0.1 - 0.074 = 0.026m. The shift is set to 50% of Δd to avoid excessive shifting. The distance between the robot and the obstacle at Y-coordinate of 0.5m reaches 0.474m, which is close to the safety threshold. The distance will be further increased by adjusting the speed.

[0086] Step S4-4-4: Verify whether the distance between the adjusted path and the tunnel wall is greater than or equal to 0.3m, i.e., the boundary threshold. After adjustment, the robot's Y-coordinate is 0.026m, and the Y-coordinate of the left side wall of the tunnel is 0.0m. The distance is 0.026m < 0.3m, so it needs to be adjusted accordingly. The positive axis was adjusted by 0.274m. After the adjustment, the Y coordinate was 0.026 + 0.274 = 0.3m, and the distance from the tunnel wall was 0.3m, which meets the requirements.

[0087] when In low-risk situations, maintain the current path and cruising speed, update obstacle movement data every 0.1 seconds, and continuously monitor changes in risk.

[0088] Example 2

[0089] This embodiment provides a cable tunnel inspection system based on a quadruped robot, used to execute the cable tunnel inspection method based on a quadruped robot described in Embodiment 1. It consists of a data acquisition module, a data processing module, a pose estimation and path generation module, and an obstacle avoidance decision module. Figure 2 As shown.

[0090] The data acquisition module is used to collect multi-source data in real time during the robot's inspection of cable tunnels. The multi-source data includes: geometric feature data of obstacles in the tunnel, motion state data, robot path deviation data, and tunnel environmental baseline data.

[0091] The data processing module is used to perform time synchronization and anomaly processing on the multi-source data, extract environmental feature points and parameters required for particle filtering, and construct a structured fusion dataset that supports localization and obstacle avoidance algorithms.

[0092] The pose estimation and path generation module is used to estimate the robot's real-time pose based on the structured fusion dataset and combined with tunnel environment reference data, using a fusion positioning algorithm, and to generate a coordinate sequence for the future inspection path.

[0093] The obstacle avoidance decision module is used to predict the future trajectory of obstacles based on their motion state data, perform spatiotemporal overlap analysis with the coordinate sequence of the robot's future patrol path, identify potential collision risks, adjust the robot's movement speed according to the risk level, and trigger the corresponding obstacle avoidance response.

[0094] Specifically, the data acquisition module includes:

[0095] The coordinate system establishment unit is used to establish a dual coordinate system. A global coordinate system is established with a preset reference mark as the origin, and the direction extending along the tunnel axis is denoted as... The axis, perpendicular to the horizontal direction of the tunnel axis, is denoted as... An axis, vertically upward, is denoted as Axis; establish a local coordinate system with the robot's center as the origin, and denote it as the axis along the robot's forward direction. The axis, the horizontal axis perpendicular to the direction of travel, is denoted as... An axis, vertically upward, is denoted as axis;

[0096] The environmental perception unit is used to generate three-dimensional point cloud data in a local coordinate system by scanning the tunnel environment in real time using lidar. The coordinates are transformed and mapped to the global coordinate system; visual images are acquired simultaneously, obstacle depth information is calculated and converted into three-dimensional coordinates in the global coordinate system. Extract the outline dimensions (L, W, H) and depth information d of the obstacle; where L is the length of the obstacle along the tunnel axis, W is the width perpendicular to the axis, and H is the height;

[0097] The motion state acquisition unit is used to acquire the robot's angular velocity in real time. Linear acceleration a and attitude angle ,in For roll angle, For pitch angle, For heading angle; simultaneously acquire the instantaneous velocity v, direction angle α, and acceleration of dynamic obstacles. ;

[0098] The path deviation detection unit is used to preset the baseline path. Real-time acquisition of the robot's current position P Lateral deviation from the preset baseline path and longitudinal deviation ,in For P to vertical distance, , Accumulate mileage along the preset baseline path from the current location. This is the mileage corresponding to the point closest to P on the baseline path;

[0099] The reference marker detection unit is used to acquire marker detection data, including the absolute coordinates of the marker in the global coordinate system, when the robot moves into the detection range of a preset reference marker. and identity markers The reliability parameters of this test are also included, and the preset benchmark marks are set at intervals along the tunnel path.

[0100] Specifically, the data processing module includes:

[0101] The time synchronization unit is used to base time on the robot's local clock. Obtain the collection timestamps of each source data. By calculating the time deviation between each timestamp and the reference time Based on this deviation, linear interpolation is used to map data from different sampling times to a unified time series. , where k=1,2,...,n, to synchronize multi-source data in the time dimension;

[0102] The anomaly detection unit is used to analyze the synchronized obstacle geometric feature data by calculating the feature deviation between adjacent frames. Where F is the feature vector, containing the three-dimensional coordinates of the obstacle in the global coordinate system. Contour dimensions (L, W, H) and depth information d; when δ exceeds a preset threshold Anomalies are identified when the motion state data is checked for abnormalities. Specifically, this includes verifying whether the absolute values ​​of the robot's linear acceleration *a* and angular velocity *ω* meet certain requirements. and ,in and These are the linear acceleration and angular velocity thresholds, respectively; acceleration relative to dynamic obstacles. Check whether its absolute value satisfies ;

[0103] The data repair unit is used to replace isolated points with the neighborhood mean for detected abnormal data, and to fill in the gaps with complementary data from a backup data source for continuous anomalies, and to label the data confidence level c, where the confidence level of motion parameter data is denoted as c. The confidence level of the labeled detection data is denoted as ;

[0104] Dataset construction unit, used according to the unified time series After processing, each data record is denoted as... Where k is the time point index, with values ​​1, 2, ..., n. Represents a complete data record at the k-th time point; includes the field: obstacle feature set. Robot motion parameters Path deviation Tunnel environmental benchmark information And data confidence level c, forming a structured dataset with spatiotemporal correlation. .

[0105] Specifically, the pose estimation and path generation module includes:

[0106] The initial pose calculation unit is used to calculate robot motion parameters based on the fused dataset D. and path deviation The initial pose of the robot is calculated using a kinematic model. ,in The initial 3D coordinates in the global coordinate system. The initial attitude angle is... This is represented as the initial roll angle. The initial pitch angle, This is the initial heading angle;

[0107] The pose correction unit is used to combine the absolute coordinates of the markers in the tunnel environment reference data in the global coordinate system. When the robot detects a reference marker, it uses the absolute coordinates of the marker as the reference marker constraint and performs a weighted fusion algorithm to determine the initial pose. Initial three-dimensional coordinates and initial attitude angle The correction is performed, where the weights of the initial pose are determined based on the reliability of the motion parameter data. The weights of the baseline marker constraints are determined based on the reliability of the marker detection data. The corrected real-time pose is determined, and the sum of the two weights is 1. ;

[0108] Future path generation unit, used for real-time pose generation And the preset global inspection path, within the future time window T, in order to Location Starting from the global preset baseline path The extension direction is selected sequentially at preset spatial intervals Δs, and the range of the selected points is limited to a future time window T, generating coordinate points. The coordinates are in the global coordinate system, where t is the robot path time step index; all points must satisfy the deviation constraint from the global path and the distance between adjacent points must not exceed Δs, ultimately forming the coordinate sequence of the future inspection path. .

[0109] Specifically, the obstacle avoidance decision module includes:

[0110] The obstacle trajectory prediction unit is used to predict obstacle motion state data based on the fused dataset D. A kinematic prediction model is used to generate a sequence of trajectory coordinates of obstacles within a future time window T. ,in This is an obstacle trajectory time step index, which corresponds one-to-one with the robot path time step index; where For obstacles at any time The three-dimensional coordinates of the trajectory points are consistent with the robot path coordinate sequence;

[0111] The spatiotemporal matching unit is used to generate a sequence of future trajectory coordinates of obstacles. With the coordinate sequence of the robot's future inspection path Perform spatiotemporal matching to calculate the spatial distance at corresponding times. and time difference ,in For the robot to arrive At that moment, For the obstacle to reach The moment;

[0112] Risk level calculation unit, used when and ≤ At that time, it was determined that there was a potential collision risk, among which As a space safety threshold, As a time safety threshold; and according to and Calculating risk level based on actual value ,in , The weighting coefficients and ;

[0113] The obstacle avoidance strategy execution unit is used to execute the corresponding obstacle avoidance strategy according to the risk level R: when When, it triggers emergency braking and generates an alternative route; when When, adjust the robot's speed and direction to increase the safe distance; when At the same time, maintain the current path and continuously monitor the movement status of obstacles; among which Indicates a high-risk threshold. This indicates a low-risk threshold.

[0114] In the obstacle avoidance strategy execution unit, such as Figure 3 As shown, it includes:

[0115] Offset calculation subunit, used for calculating the future trajectory coordinate sequence of obstacles. With the robot's future path coordinate sequence Calculate the lateral offset direction of the obstacle relative to the robot, the lateral offset direction being along the global coordinate system. The positive and negative directions of the axis; and the minimum safe distance difference. ;

[0116] The speed adjustment subunit is used to dynamically reduce the robot's cruising speed according to the risk level R. ,in Original cruising speed;

[0117] The direction adjustment subunit is used to generate a deflection angle in the lateral offset direction away from the obstacle. ,in , This is a scaling factor that makes the robot path... The axis is offset, and the minimum distance between the offset path and the future trajectory of the obstacle is greater than or equal to... ;

[0118] The boundary verification subunit is used to verify in real time whether the distance between the adjusted path and the tunnel wall is greater than or equal to a preset boundary threshold. If not, it prioritizes following the boundary threshold. Adjust the path in the opposite direction of the axis.

[0119] Example 3

[0120] This embodiment provides a computer device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform the method according to Embodiment 1 of the present invention.

[0121] Example 4

[0122] This embodiment provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform the method according to Embodiment 1 of the present invention.

[0123] refer to Figure 4 The present invention will now be described in the form of a structural block diagram of an electronic device 400 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0124] like Figure 4 As shown, the electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. The RAM 403 may also store various programs and data required for the operation of the electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0125] Multiple components in electronic device 400 are connected to I / O interface 405, including: input unit 406, output unit 407, storage unit 408, and communication unit 409. Input unit 406 can be any type of device capable of inputting information to electronic device 400. Input unit 406 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 407 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 408 may include, but is not limited to, disks and optical discs. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0126] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above. For example, in some embodiments, the aforementioned cable tunnel inspection method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. In some embodiments, the computing unit 401 can be configured to perform the aforementioned cable tunnel inspection method by any other suitable means (e.g., by means of firmware).

[0127] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0128] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0129] As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0130] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0131] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0132] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0133] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. It will be apparent to those skilled in the art that various modifications can be made to the above embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. A cable tunnel inspection method based on a quadruped robot, characterized in that, include: Step S1: During the inspection of the cable tunnel, the robot collects multi-source data in real time. The multi-source data includes: geometric feature data of obstacles in the tunnel, motion state data, robot's own path deviation data, and tunnel environment baseline data. Step S2: Perform time synchronization and anomaly processing on the multi-source data, extract environmental feature points and parameters required for particle filtering, and construct a structured fusion dataset that supports localization and obstacle avoidance algorithms; Step S3: Based on the structured fusion dataset and combined with the tunnel environment reference data, a fusion positioning algorithm is used to estimate the robot's real-time pose and generate a coordinate sequence for the future inspection path. Step S4: Based on the motion state data of the obstacle, predict the future trajectory of the obstacle, perform spatiotemporal overlap analysis with the coordinate sequence of the robot's future patrol path, identify potential collision risks, adjust the robot's movement speed according to the risk level and trigger the corresponding obstacle avoidance response. Step S3 includes: Step S3-1: Based on the robot motion parameters in the structured fusion dataset D and path deviation The initial pose of the robot is calculated using a kinematic model. ,in The initial 3D coordinates in the global coordinate system. The initial attitude angle is... The initial roll angle, The initial pitch angle, This is the initial heading angle; Step S3-2: Combine the absolute coordinates marked in the global coordinate system from the tunnel environment reference data. When the robot detects a reference marker, it uses the absolute coordinates of the marker as the reference marker constraint and performs a weighted fusion algorithm to determine the initial pose. Initial three-dimensional coordinates and initial attitude angle The correction is performed, where the weights of the initial pose are determined based on the reliability of the motion parameter data. The weights of the baseline marker constraints are determined based on the reliability of the marker detection data. The corrected real-time pose is determined, and the sum of the two weights is 1. ; Step S3-3: Based on real-time pose And the preset global inspection path, within the future time window T, in order to three-dimensional position Starting from the global preset baseline path The extension direction is selected sequentially at preset spatial intervals Δs, and the range of selected points is limited to a future time window T, generating coordinate points. The coordinates are in the global coordinate system, where t is the robot path time step index; all points must satisfy the deviation constraint from the global path and the distance between adjacent points must not exceed Δs, ultimately forming the coordinate sequence of the future inspection path. .

2. The cable tunnel inspection method according to claim 1, characterized in that, Step S1 includes: Step S1-1: Establish a dual coordinate system. A global coordinate system is established with the preset reference mark as the origin, and the direction extending along the tunnel axis is denoted as... The axis, perpendicular to the horizontal direction of the tunnel axis, is denoted as... An axis, vertically upward, is denoted as Axis; establish a local coordinate system with the robot's center as the origin, and denote the direction of robot movement as . The axis, the horizontal axis perpendicular to the robot's direction of movement, is denoted as... An axis, vertically upward, is denoted as axis; Step S1-2: Generate 3D point cloud data in the local coordinate system by scanning the tunnel environment in real time using LiDAR. The coordinates are transformed and mapped to the global coordinate system; visual images are acquired simultaneously, obstacle depth information is calculated and converted into three-dimensional coordinates in the global coordinate system. Extract the outline dimensions (L, W, H) and depth information d of the obstacle; where L is the length of the obstacle along the tunnel axis, W is the width perpendicular to the axis, and H is the height; Steps S1-3: Obtain the robot's angular velocity in real time. Linear acceleration a and attitude angle ,in For roll angle, For pitch angle, For heading angle; simultaneously acquire the instantaneous velocity v, direction angle α, and acceleration of dynamic obstacles. ; Step S1-4: Based on the preset baseline path Real-time acquisition of the robot's current position P Lateral deviation from the preset baseline path and longitudinal deviation ,in For P to vertical distance, , Accumulate mileage along the preset baseline path from the current location. This is the mileage corresponding to the point closest to P on the baseline path; Steps S1-5: When the robot moves to the detection range of the preset reference mark, acquire the mark detection data, including: the absolute coordinates of the mark in the global coordinate system. Identity identification The reliability parameters of this test are also included, and the preset benchmark marks are set at intervals along the tunnel path.

3. The cable tunnel inspection method according to claim 2, characterized in that, Step S2 includes: Step S2-1: Use the robot's local clock as the base time. Obtain the collection timestamps of each source data. By calculating the time deviation between each timestamp and the reference time Based on this deviation, linear interpolation is used to map data from different sampling times to a unified time series. , where k=1,2,...,n, to synchronize multi-source data in the time dimension; Step S2-2: For the synchronized obstacle geometric feature data, calculate the feature deviation between adjacent frames. Where F is the feature vector, containing the three-dimensional coordinates of the obstacle in the global coordinate system. Contour dimensions (L, W, H) and depth information d; when δ exceeds a preset threshold The data is judged as abnormal at that time; for motion state data, outliers are identified through reasonableness checks. Step S2-3: For detected abnormal data, if it is an isolated point, the neighborhood mean is used as a substitute; if it is a continuous anomaly, complementary data from the backup data source is used to fill it in, and the data confidence level c is marked. Step S2-4: According to the uniform time series After processing, each data record is denoted as... Where k is the time point index, with values ​​1, 2, ..., n. Represents a complete data record at the k-th time point; includes the field: obstacle feature set. Robot motion parameters Path deviation Tunnel environmental benchmark information And data confidence level c, forming a structured fusion dataset with spatiotemporal correlation. .

4. The cable tunnel inspection method according to claim 3, characterized in that, In step S2-2, the motion state data is checked for outliers to identify them. Specifically, this includes checking whether the absolute values ​​of the robot's linear acceleration 'a' and angular velocity 'ω' meet the requirements. and ,in and These are the linear acceleration threshold and the angular velocity threshold, respectively; acceleration of dynamic obstacles. Check whether its absolute value satisfies , This is the acceleration threshold for dynamic obstacles.

5. The cable tunnel inspection method according to claim 4, characterized in that, Step S4 includes: Step S4-1: Based on obstacle motion state data in structured fusion dataset D A kinematic prediction model is used to generate a sequence of trajectory coordinates of obstacles within a future time window T. ,in This is a future time step index, corresponding one-to-one with the time step index of the robot's future path; where For obstacles at any time The three-dimensional coordinates of the trajectory points are consistent with the robot path coordinate sequence; Step S4-2: Sequence of future trajectory coordinates of obstacles With the coordinate sequence of the robot's future inspection path Perform spatiotemporal matching to calculate the spatial distance at corresponding times. and time difference ,in For the robot to arrive At that moment, For the obstacle to reach The moment; Step S4-3: When and ≤ At that time, it was determined that there was a potential collision risk, among which As a space safety threshold, As a time safety threshold; and according to and Calculating risk level based on actual value ,in , The weighting coefficients and ; Step S4-4: Execute the corresponding obstacle avoidance strategy according to the risk level R: when When, it triggers emergency braking and generates an alternative route; when When, adjust the robot's speed and direction to increase the safe distance; when At the same time, maintain the current path and continuously monitor the movement status of obstacles; among which Indicates a high-risk threshold. This indicates a low-risk threshold.

6. The cable tunnel inspection method according to claim 5, characterized in that, The method for adjusting the robot's speed and direction to increase the safe distance in step S4-4 includes: Step S4-4-1: Based on the future trajectory coordinate sequence of obstacles With the robot's future path coordinate sequence Calculate the lateral offset direction of the obstacle relative to the robot, the lateral offset direction being along the global coordinate system. The positive and negative directions of the axis; and the minimum safe distance difference. ; Step S4-4-2: Dynamically reduce the robot's cruising speed according to the risk level R, and adjust the speed accordingly. ,in Original cruising speed; Step S4-4-3: Generate a deflection angle in the lateral offset direction away from the obstacle. ,in , This is a scaling factor that makes the robot path... The axis is offset, and the minimum distance between the offset path and the future trajectory of the obstacle is greater than or equal to... ; Step S4-4-4: Real-time verification of whether the distance between the adjusted path and the tunnel wall is greater than or equal to the preset boundary threshold. If not, prioritize following the path. Adjust the path in the opposite direction of the axis.

7. A cable tunnel inspection system based on a quadruped robot, used to perform the cable tunnel inspection method based on a quadruped robot as described in any one of claims 1-6, characterized in that, include: The module includes a data acquisition module, a data processing module, a pose estimation and path generation module, and an obstacle avoidance decision-making module. The data acquisition module is used to collect multi-source data in real time during the robot's inspection of cable tunnels. The multi-source data includes: geometric feature data of obstacles in the tunnel, motion state data, robot path deviation data, and tunnel environmental baseline data. The data processing module is used to perform time synchronization and anomaly processing on the multi-source data, extract environmental feature points and parameters required for particle filtering, and construct a structured fusion dataset that supports localization and obstacle avoidance algorithms. The pose estimation and path generation module is used to estimate the robot's real-time pose based on the structured fusion dataset and combined with tunnel environment reference data, using a fusion positioning algorithm, and to generate a coordinate sequence for the future inspection path. The obstacle avoidance decision module is used to predict the future trajectory of obstacles based on their motion state data, perform spatiotemporal overlap analysis with the coordinate sequence of the robot's future patrol path, identify potential collision risks, adjust the robot's movement speed according to the risk level, and trigger the corresponding obstacle avoidance response.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images