Navigation matching correction method based on inspection robot

Abstract

The invention discloses a navigation matching correction method based on an inspection robot, and the method comprises the steps: building a feature database through the deployment of a physical calibration object and the recognition of a natural feature object, providing a reliable positioning reference for a robot, achieving the coarse positioning through the fusion of visual and laser radar data in a positioning process, and introducing a dynamic credibility evaluation mechanism. The positioning reliability is quantified in real time through an exponential decay model, when the credibility is lower than a threshold value, the system automatically triggers a compensation behavior to re-search features, in the aspect of multi-robot cooperation, secondary positioning correction is achieved through track matching and data fusion, high-confidence-coefficient reference data are screened through a clustering algorithm, the group positioning precision is improved, and the positioning accuracy is improved. Aiming at a key inspection area, multi-angle image matching is adopted to realize fine positioning, a positioning error is dynamically corrected through a sliding window, the continuity and accuracy of robot navigation in a complex environment are remarkably improved through closed-loop correction and self-adaptive optimization, and the method is suitable for intelligent inspection requirements of railway trains.

Description

Technical Field

[0001] The present invention relates to the technical field of train inspection, and more particularly to a navigation matching and correction method based on an inspection robot. Background Art

[0002] With the increasing demand for railway train inspections, the use of inspection robots for self-service inspections is becoming increasingly common. However, existing inspection robot navigation technology still has shortcomings in terms of path planning accuracy, environmental adaptability, and navigation matching and correction capabilities, resulting in inspection efficiency and accuracy that cannot meet actual needs.

[0003] After searching, a patrol robot navigation method, device, patrol robot and storage medium with publication number CN112729302B was disclosed, and the publication date was March 29, 2024. This patent determines the current position of the patrol robot and navigates by obtaining the patrol area map and patrol path, combining the matching degree between the detection area reference map of the estimated area and the actual detection area map. However, in this technical solution, the accuracy of navigation matching depends on the division of the estimated area and the construction of the reference map. If the estimated area is too large or too small, it may lead to increased matching errors, thereby affecting the accuracy of navigation. In addition, this solution does not take into account the real-time correction mechanism in a dynamic environment, and positioning deviations may occur due to environmental changes in complex scenarios.

[0004] After searching, a navigation method and device for an inspection robot with the publication number CN115979249B was disclosed, and the publication date is June 20, 2023. This patent uses pre-built visual SLAM technology to generate a global map, and selects the inspection route with the best energy consumption based on the optimization function and energy loss calculation model, while combining the yaw angle to complete the navigation control. However, in this technical solution, the construction of the global map has high requirements on the accuracy of the sensor and the richness of the environmental features. If the environmental features are sparse or the sensor data noise is large, it may lead to inaccurate map construction, thereby affecting the planning accuracy of the navigation path. In addition, the solution lacks a real-time response mechanism to dynamic obstacles, and path deviation or navigation failure may occur in complex dynamic environments.

[0005] The above problems show that the existing inspection robot navigation technology still has certain shortcomings in terms of navigation matching accuracy, adaptability to dynamic environments, and real-time correction capabilities. Therefore, the present invention provides a navigation matching and correction method based on an inspection robot, which aims to improve the accuracy of navigation matching, enhance the real-time correction capability in a dynamic environment, and optimize the flexibility and accuracy of path planning, thereby meeting the needs of the railway train inspection field for efficient and precise navigation. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the purpose of this invention is to overcome the existing inspection robot navigation methods' over-reliance on static environment models and the lack of dynamic quantization management and smooth correction mechanisms when dealing with persistent pose uncertainty caused by factors such as sensor drift and odometry cumulative error. Therefore, the present invention provides a navigation matching and correction method based on an inspection robot. This method aims to establish a closed-loop control system capable of dynamically quantifying, modeling, tracking, and correcting the confidence and accuracy of the robot's own position information. This system introduces a dual time decay model to manage the lifecycle of the confidence in the coarse positioning phase and the accuracy in the fine positioning phase. This system, combined with high-confidence information fusion and adaptive decay parameter correction within a multi-robot collaborative framework, fundamentally addresses the problem of navigation failure caused by the continuous accumulation of uncertainty, significantly improving the continuity, accuracy, and reliability of the inspection robot's autonomous navigation in complex industrial environments.

[0007] To achieve the above object, the present invention provides the following technical solutions: A navigation matching correction method based on an inspection robot comprises the following steps: Inspection scene feature calibration step: deploy physical calibration objects at preset locations within the inspection area and identify inherent natural features in the environment. 3D spatial coordinates and posture parameters of the physical calibration objects and natural features in the world coordinate system are obtained through 3D scanning. The geometric type, 3D coordinates, and posture parameters of the features are then combined into feature records and stored in the inspection robot's memory to form a feature and physical coordinate database. The robot coarse positioning step synchronously collects the image data stream and the three-dimensional point cloud data stream of the environment through the visual sensor and the lidar sensor. When at least one of the features is collaboratively identified, the perspective n-point algorithm is used to solve the three-dimensional translation vector and rotation matrix of the current camera coordinate system relative to the identified feature. The relative position in the camera coordinate system is converted into the absolute position of the inspection robot in the world coordinate system through the coordinate system homogeneous transformation by combining the feature with the coordinates and posture of the feature in the physical coordinate database; A quantitative calculation step of the coarse positioning confidence level, after each successful execution of the robot coarse positioning step, calculates an initial confidence level used to characterize the reliability of the coarse positioning result; The time decay step of the coarse positioning trust, after obtaining the initial trust, makes the initial trust change dynamically over time according to a preset exponential decay model to obtain a real-time trust that evolves over time, which is used to dynamically quantify and track the uncertainty of the positioning information of the inspection robot.

[0008] Furthermore, the quantitative calculation step of the coarse positioning confidence includes a basic score calculation sub-step, a basic score adjustment sub-step, a confidence correction sub-step and a sensor verification sub-step; The basic score calculation sub-step determines a basic score positively correlated with the number of SIFT feature matching point pairs used for the perspective n-point algorithm in the robot rough positioning step; The basic score adjustment sub-step determines the feature integrity weight according to the feature object outline integrity or occlusion degree to adjust the basic score by multiplication; The trustworthiness correction sub-step determines an environmental adaptability correction factor based on real-time illumination intensity or image clarity, and performs product correction again on the basic score after product adjustment; The sensor verification sub-step compares the positions of the feature objects calculated by the visual sensor and the lidar sensor, and adds or subtracts the confidence value based on whether the three-dimensional distance deviation between the two is within a preset threshold to obtain an initial confidence.

[0009] Furthermore, the mathematical expression of the exponential decay model is: ,in is an attenuation coefficient, and a linear mapping relationship is established between the attenuation coefficient and the moving speed of the inspection robot obtained by the wheel speed odometer. In addition, the attenuation coefficient is also affected by the feature density of the area where the inspection robot is located. When the number of calibrated features within the preset radius of the current position of the inspection robot is greater than a preset number threshold, the attenuation coefficient is multiplied by a first compensation coefficient less than 1; when the number of calibrated features within the preset radius is less than another preset number threshold, the attenuation coefficient is multiplied by a second compensation coefficient greater than 1.

[0010] Furthermore, it also includes real-time trust monitoring and compensation triggering steps and trust updating and attenuation resetting steps under multi-feature recognition; The real-time confidence monitoring and compensation triggering step monitors the real-time confidence value in real time. When the real-time confidence value decays below a preset compensation trigger threshold, an active coarse positioning compensation behavior is automatically triggered, and the inspection robot is controlled to scan to search for new available features. The trust update and attenuation reset steps under the multi-feature recognition are as follows: when the inspection robot continuously recognizes multiple features, if the newly calculated initial trust is higher than the current real-time trust after attenuation, the current trust is directly updated to the new initial trust, and the attenuation time is reset.

[0011] Furthermore, it also includes a secondary positioning fusion correction step for multi-robot interaction, in which each patrol robot transmits a data packet containing its own ID, timestamp, coarse positioning posture, real-time trust and historical trajectory point sequence to other patrol robots according to a preset period. When the patrol robot receives the data packet from the other machine, it extracts the historical trajectory feature sequence of this machine and the other machine, and uses the dynamic time warping DTW algorithm to calculate the similarity between the two trajectory feature sequences; when the similarity meets the preset conditions, it is determined that the path of the other machine and the local machine overlaps, and the positioning data points transmitted by the other machine in the overlapping segment are used as reference data to be processed.

[0012] Furthermore, the multi-robot interaction secondary positioning fusion correction step also includes a valid cluster screening sub-step and a secondary positioning correction sub-step. The effective cluster screening sub-step processes all reference data to be processed using the density-based spatial clustering DBSCAN algorithm, sets the neighborhood radius and the minimum number of samples, divides the spatially closely spaced positioning data into clusters, and removes isolated data points determined by the algorithm to be noise, retaining only the largest cluster with a data volume exceeding a preset ratio as the effective reference data; The secondary positioning correction sub-step calculates the secondary positioning correction result of the aircraft through a weighted fusion algorithm. The final fusion coordinates of the aircraft are determined by its own coarse positioning result and all valid reference data. The calculation formula is: ,in is the coarse positioning coordinate of this machine, is the valid reference data coordinate, and are their respective weights.

[0013] Furthermore, the secondary positioning correction sub-step also includes dividing the real-time trust into multiple levels, making the weight and Dynamically associated with their respective real-time trust levels, before weighted average calculation, abnormal reference data points are checked and identified, and the weights of reference data points that deviate from the cluster mean by more than a preset multiple of the standard deviation are directly set to zero.

[0014] Furthermore, it also includes a fine positioning step. When the inspection robot confirms through positioning that it has entered the preset key component inspection area, the fine positioning step is activated, and the camera carried by the inspection robot is used to capture images of the pre-examined fine positioning feature objects from at least three different spatial positions and angles, and extract the two-dimensional key points of the fine positioning feature objects from them, and match the two-dimensional key points with the three-dimensional model of the feature object pre-stored in the feature and physical coordinate database, and solve the relative pose transformation matrix between the current camera coordinate system and the fine positioning feature object's own coordinate system, and combine the coordinates of the feature object in the world coordinate system in the database to obtain the current fine positioning coordinates of the inspection robot.

[0015] Furthermore, the method also includes a dynamic attenuation and correction step, in which the initial accuracy is calibrated according to the image matching error of the two-dimensional key points, and the initial accuracy is dynamically adjusted according to the vibration acceleration amplitude collected by the inspection robot. The initial accuracy deteriorates over time according to the linear attenuation rule. The attenuation expression of the accuracy value is: ,in for The precision of the moment, is the precision reduction factor, is the initial accuracy.

[0016] Furthermore, the dynamic attenuation and correction step also includes storing several recent fine positioning results through a sliding window. Whenever a new fine positioning result is obtained, the difference between the result and the historical result in the sliding window is calculated to obtain a difference sequence, the average value of the difference sequence is calculated, and the precision attenuation factor is increased or decreased by analyzing the changing trend of the average value over multiple consecutive calculation cycles.

[0017] Beneficial effects of the present invention: 1. By introducing a dual time decay model of coarse positioning trust and fine positioning accuracy, dynamic quantification and life cycle management of the reliability of robot positioning information are achieved. Combining real-time monitoring and compensation triggering mechanisms, a closed-loop control system is formed, which effectively solves the problem of continuous expansion of positioning deviations caused by sensor drift or cumulative errors in traditional methods, and significantly improves the continuity and stability of navigation. In addition, by fusing multimodal data of visual sensors and lidars, and dynamically adjusting trust in combination with real-time parameters such as ambient lighting and feature occlusion, the positioning robustness of the robot in complex or dynamic environments is enhanced. Even if some features are blocked or lighting conditions change, the system can still maintain a high positioning reliability through weighted fusion and attenuation compensation; 2. By sharing data and matching trajectories among multiple robots, and then achieving secondary positioning correction through clustering screening and weighted fusion, the positioning failure of a single robot due to the lack of local features is avoided. The group collaboration mechanism not only improves the positioning accuracy, but also enhances the fault tolerance of the system in sparse feature environments. In addition, the trust attenuation coefficient and the precision attenuation factor are dynamically associated with the robot's motion state, environmental feature density, etc., and can be adaptively adjusted according to the actual scenario; at the same time, the difference trend of historical positioning results is analyzed through a sliding window, and the attenuation parameters are corrected in real time, which further optimizes the long-term positioning accuracy and triggers high-precision image matching positioning in key inspection areas to ensure accurate detection of important components; and when features are missing or communication is interrupted, it automatically switches to the inertial navigation emergency mode and slows down the trust attenuation, ensuring the basic navigation function in extreme cases and improving the overall reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is the overall flow chart of the present invention; Figure 2 This is a flow chart of the sub-steps of quantitative calculation of coarse positioning confidence in the present invention; Figure 3 This is a flowchart of the secondary positioning fusion correction of multi-robot interaction in the present invention. DETAILED DESCRIPTION

[0019] The present invention will be described in further detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom," "top," "inner," and "outer" refer to directions toward or away from the geometric center of a particular component, respectively.

[0020] Since the existing inspection robot navigation technology still has certain deficiencies in terms of navigation matching accuracy, dynamic environment adaptability and real-time correction capability, the present invention provides a navigation matching correction method based on an inspection robot, which is applied to at least one inspection robot, wherein the inspection robot is equipped with a central processing unit, a memory, at least one set of visual sensors, a set of laser radar sensors, a set of inertial measurement units and a wireless communication module; Figure 1 As shown, the method includes the following steps: First, the inspection scene’s characteristic shapes are calibrated and three-dimensional space modeled. This step aims to provide a stable and recognizable geometric reference for the subsequent positioning process. Specifically, physical calibration objects with unique geometric shapes and high optical contrast are deployed at preset key locations within the inspection area. Components inherent in the inspection environment and with stable geometric features are identified as natural features. The physical calibration objects are red equilateral triangle calibration plates with a side length of 8 cm. The natural features include hexagonal bolt groups and rectangular inspection hatch covers on the train body. All physical calibration objects and natural features are scanned by a 3D laser scanner to obtain their precise 3D spatial coordinates in the pre-established world coordinate system ( ) and attitude parameters (yaw angle, pitch angle, roll angle), the geometric type, three-dimensional coordinates and attitude parameters of the feature object constitute a feature record, and all feature records are stored in the local memory of the inspection robot to form a feature and physical coordinate database.

[0021] At the same time, an industrial camera is used to collect a group of image samples of various types of feature objects under different lighting conditions and observation angles. The image samples are input into a pre-built regional convolutional neural network target detection model for training. In addition, for each type of feature object, its scale-invariant feature transform (SIFT) key points are extracted, and a standard SIFT feature descriptor template is generated and stored in the feature-physical coordinate database.

[0022] Secondly, the inspection robot is roughly positioned based on feature recognition. This step aims to provide the robot with an initial or periodic pose estimation with a controlled error range. The inspection robot is equipped with an industrial camera and a lidar. During the movement of the inspection robot, the industrial camera and the lidar synchronously and continuously collect the image data stream and the three-dimensional point cloud data stream of the environment; the central processing unit performs Gaussian filtering denoising and histogram equalization on the collected image frames in real time, and then calls the pre-trained Faster The R-CNN model detects whether there are calibrated features in the processed image frames. At the same time, the Euclidean clustering segmentation algorithm is applied to the collected point cloud data stream to extract point cloud clusters with preset geometric dimensions that match the calibrated features. When the visual detection and point cloud segmentation collaboratively identify at least one feature, the coarse positioning calculation is immediately triggered. Specifically, the camera intrinsic parameter matrix and distortion coefficient are used to solve the three-dimensional translation vector and rotation matrix of the current camera coordinate system relative to the identified feature through the perspective n-point algorithm. Subsequently, the coordinates and posture of the feature in the world coordinate system are retrieved from the feature and physical coordinate database. The relative posture in the camera coordinate system is converted to the absolute posture of the robot in the world coordinate system through the multiplication operation of the coordinate system homogeneous transformation matrix. .

[0023] Further, if Figure 2 As shown in the figure, a quantitative calculation and time decay mechanism for coarse positioning trust is introduced. This mechanism aims to quantitatively evaluate the reliability of each coarse positioning result and simulate its time decay over time. After each successful coarse positioning, an initial trust C0 is immediately calculated. The calculation of the initial trust integrates evaluation indicators from multiple dimensions: First, a base score based on the number of feature matching points is set. When the number of SIFT feature matching point pairs used for the perspective n-point algorithm is greater than or equal to 10, the base score is 0.9; when the number is between 5 and 9, the base score is 0.6; when the number is less than 5, the base score is 0.3; Secondly, a feature completeness weight is introduced. If the completeness of the identified feature outline is higher than 90%, 0.1 is added to the basic score; if there is less than 30% occlusion, 0.1 is deducted from the basic score. Finally, an environmental adaptability correction factor is introduced. The robot's onboard ambient light and humidity sensors acquire real-time environmental parameters. When the light intensity is greater than N1 or less than N2, the current confidence value is multiplied by a correction factor of 0.8. When the image clarity drops below a preset threshold, the value is multiplied by a correction factor of 0.7. Finally, a sensor consistency check is performed to compare the positions of feature objects calculated using the visual n-point algorithm and the LiDAR point cloud centroid matching. If the three-dimensional spatial distance deviation between the two is less than or equal to 5 cm, the confidence level is increased by 0.05; if the deviation is greater than 10 cm, the confidence level is reduced by 0.15. The final value calculated from all the above factors is used as the initial confidence level C0 for this coarse positioning.

[0024] After obtaining the initial trust C0, the trust is not static, but changes dynamically with time t according to an exponential decay model. Its mathematical expression is: , where the attenuation coefficient is a dynamic parameter whose value is related to the state of the robot. Specifically, The value establishes a linear mapping relationship with the robot moving speed v; In addition, the attenuation coefficient It is also affected by the feature density of the area where the robot is located. If there are more than or equal to 3 calibrated features within a 5-meter radius of the robot's current position, the currently calculated The value is multiplied by a compensation factor of 0.8; if there is less than 1 feature within a 10-meter radius, The value is multiplied by a compensation factor of 1.2.

[0025] When the value of C(t) decays to below the preset compensation trigger threshold of 0.3, the robot control system automatically triggers an active coarse positioning compensation behavior, controls the robot gimbal to perform a 360-degree scan to search for new available features, and plots the decay curve of C(t) over time in real time on the monitoring interface of the robot control system, and marks the warning line (C=0.5) and compensation trigger line (C=0.3). When the robot continuously recognizes multiple features, the maximum value priority principle is used to update the current trust. That is, if the newly calculated C0 is higher than the current attenuated C(t), the current trust of the system is directly updated to the new C0, and the decay time t is reset.

[0026] Further, if Figure 3 As shown, it also includes a secondary positioning fusion correction method for multi-robot interaction, which is applied to a cluster composed of multiple inspection robots. Each robot in the cluster builds a low-latency, high-bandwidth real-time communication network through its onboard communication module or Mesh self-organizing network module. Each robot transmits a message containing its own ID, current timestamp, and the position calculated by coarse positioning to other robots in the network in a period of 0.5 seconds. , the current trust C(t) and the data packet of the historical trajectory point sequence in the last 60 seconds.

[0027] When a robot (hereinafter referred to as "this robot") receives a positioning data packet from another robot (hereinafter referred to as "the other robot"), it performs a secondary positioning fusion calculation. The calculation first performs path matching and reference data screening. The local robot extracts its own historical trajectory feature sequence for the last 60 seconds. The sequence consists of a feature point containing the steering angle and speed recorded every 20 cm. The local robot uses the dynamic time warping (DTW) algorithm to calculate the similarity between its own trajectory feature sequence and the received trajectory feature sequence of the other robot. When the distance calculated by DTW is less than the preset threshold (indicating that the two paths are highly similar in shape, with a similarity greater than 80%), it is determined that the path of the other robot overlaps with that of the local robot in the corresponding time period; then, the local robot is broadcast to all other robots with overlapping paths within the overlapping segment. The positioning data points are processed using the density-based spatial clustering (DBSCAN) algorithm. The neighborhood radius eps of DBSCAN is set to 2 meters, and the minimum number of samples minPts is set to 3. The positioning data with close spatial positions are divided into a cluster, and isolated data points determined by the algorithm as noise are eliminated. Only the largest cluster with a data volume exceeding 70% of the total number of data points in the overlapping segment is retained. The mean 3D coordinate of all positioning points in the largest cluster is used as the effective reference data benchmark for this fusion calculation. Before fusion, all filtered data from other devices are time-synchronized and calibrated. For data with timestamp deviation within 100 milliseconds, linear interpolation is used to correct its posture to a position aligned with the current timestamp of the local device based on the velocity vector of the other device.

[0028] After obtaining the filtered and calibrated valid reference data, a weighted fusion algorithm based on the idea of federated filtering is used to calculate the secondary positioning correction result of the host, and the final fused coordinates of the host are determined by its own rough positioning result and all valid reference data jointly, and its calculation formula is: , where the and are not fixed values, but are dynamically associated with their respective confidence levels C(t). Specifically, the weights and are positively correlated with the confidence level C.

[0029] Introduce a hierarchical incentive mechanism: divide the confidence level into three levels: high (C≥0.7), medium (0.3<C<0.7), and low (C≤0.3). When calculating the weights, if the confidence level of a data source (whether it is the host or another machine) belongs to the high level, its weight is multiplied by an incentive coefficient of 1.2 based on its confidence level value; if it belongs to the low level, it is multiplied by an inhibitory coefficient of 0.5. Before performing the weighted average calculation, an outlier rejection based on Grubbs' test is also performed, and the weights of the reference data points whose deviation from the mean of the clustering cluster exceeds 3 times the standard deviation are directly set to zero. After the fusion calculation, not only the corrected coordinates are output, but also its 95% confidence interval is calculated. If the radius of this interval (i.e., the standard deviation σ) is greater than 50 cm, the system will trigger a request to increase the amount of reference data or shorten the data acquisition period for re-fusion.

[0030] Furthermore, the present invention also includes a fine positioning module based on high-precision image calibration. When the robot confirms through secondary positioning that it has entered a preset key component inspection area (for example, less than 1 meter away from the axle end of the target wheel set), this module is activated; the inspection robot is equipped with an industrial camera, and the robot controls the robotic arm to collect images of the preset fine positioning feature (for example, a laser-etched circular mark with a diameter of 10 mm at the center of the axle end of the wheel set) from at least three different spatial positions and angles. At the same time, the lidar performs high-density scanning on this feature to generate a local three-dimensional point cloud model; For each frame of high-precision image collected, the image processing unit extracts the 2D key points of the feature (for example, the exact center pixel coordinates and the edge point set of the circular mark); matches these 2D key points with the three-dimensional model of this feature pre-stored in the feature and physical coordinate database, and applies a variant of the perspective-n-point algorithm - the EPnP algorithm to accurately calculate the relative pose transformation matrix between the current camera coordinate system and the coordinate system of the feature itself. This matrix contains the translation vector and rotation matrix. Combined with the precise coordinates of the feature in the database in the world coordinate system, the robot's current fine positioning coordinates in the world coordinate system are calculated through coordinate transformation. If multiple fine positioning features are identified simultaneously, the multiple relative poses calculated by EPnP are put into a least squares optimization framework to minimize the sum of all feature reprojection errors, thereby solving for an optimal, multi-feature-integrated fine robot pose.

[0031] Corresponding to the coarse positioning trust attenuation model, the present invention also establishes a dynamic attenuation and correction model for the accuracy of fine positioning. The model aims to quantify the accuracy degradation of fine positioning results over time caused by factors such as small vibrations and IMU drift. After each fine positioning is completed, the system calibrates an initial accuracy S0 based on the quality of feature matching. For example, for highly reflective, clear-edge laser-etched marks, if the matching error of their key points in the image is less than 0.1 pixels, S0 is calibrated to ±2 mm; for natural weld features on the train surface, if the matching error is between 0.1 and 0.3 pixels, S0 is calibrated to ±5 mm. The initial accuracy S0 is also dynamically adjusted according to environmental parameters. For example, if the vibration acceleration amplitude detected by the IMU is greater than 0.5g, a compensation amount of ±2 mm is added to the calibrated S0.

[0032] After obtaining the initial accuracy S0, the accuracy value will deteriorate over time t according to a linear decay rule, the mathematical expression of which is: ,in, is the precision reduction factor, and its initial value is set to 0.005 / second. It is not static, but is adaptively modified based on the robot's operating status and historical positioning performance. Specifically, The setting is associated with the scene: When the robot moves smoothly, 0.005 / s; when the IMU detects that the robot vibration acceleration is greater than 0.5g, Automatically increase to 0.01 / second, accelerating accuracy attenuation; when the robot is in a stationary state performing fixed-point detection tasks, Reduced to 0.002 / second, slowing down the decay.

[0033] To achieve the attenuation factor The closed-loop correction of the present invention adopts an algorithm based on difference feedback. The system uses a sliding window with a size of 5 and a step length of 1 second to store the fine positioning results of the last 5 times (i.e. within 5 seconds). Whenever a new fine positioning result is obtained, the difference between the result and the historical results in the sliding window is calculated to obtain a difference sequence, and the average value of the difference sequence is calculated. , by analyzing the continuous calculation cycles The changing trend of Make corrections if It shows a monotonically increasing trend, indicating that the current If the value is set too small, the actual accuracy deteriorates faster than the model prediction, then a preset linear interpolation formula is used to To be adjusted upwards: New , on the contrary, if A decreasing trend indicates If the setting is too large, Adjusted downwards: New ,like If the fluctuation is within the stable range of ±1 mm, then Remain unchanged. Will be applied immediately Calculation, when three consecutive corrections When the value changes are greater than 20%, the system will automatically trigger the sensor self-test process to check for hardware failures.

[0034] Finally, the present invention forms a dynamic compensation and optimization closed loop for the entire process. During the inspection process, if the robot's visual system recognizes a new feature that is not pre-registered in the database but meets the preset geometric rules (such as circle, rectangle), the system will trigger a temporary coarse positioning compensation; it will calculate the approximate world coordinates of the new feature based on the coordinates of the nearest known feature in the surrounding area, and use it to perform an instant positioning to compensate for the accumulated error of the odometer. In a multi-robot collaborative scenario, if the coarse positioning trust of a robot is If the value is lower than 0.3, it will actively broadcast a positioning calibration request to the network. A robot with a high confidence level higher than 0.7 will send its current fine positioning results and related local point cloud map fragments to the requester. After receiving the data, the requester uses the iterative closest point (ICP) algorithm to align the path point cloud calculated by the odometer with the received high-precision local map, thereby correcting its own accumulated positioning deviation in one go and in a large scale. In addition, the system also has an emergency processing mechanism. If the robot fails to identify any known features for 5 consecutive seconds and does not receive any valid positioning data from other machines, the system will switch to a pure inertial navigation emergency mode, relying entirely on the wheel speed odometer and IMU for track calculation, and at the same time, the attenuation coefficient of the coarse positioning confidence will be reduced. It is temporarily set to a minimum value of 0.001 to maximize the "shelf life" of the last valid positioning result until the robot re-enters the feature area or restores network communication.

[0035] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.

Claims

1. A navigation matching and correction method based on an inspection robot, characterized by: The steps include: Inspection scene feature calibration step: deploy physical calibration objects at preset locations within the inspection area and identify inherent natural features in the environment. 3D spatial coordinates and posture parameters of the physical calibration objects and natural features in the world coordinate system are obtained through 3D scanning. The geometric type, 3D coordinates, and posture parameters of the features are then combined into feature records and stored in the inspection robot's memory to form a feature and physical coordinate database. The robot coarse positioning step synchronously collects the image data stream and the three-dimensional point cloud data stream of the environment through the visual sensor and the lidar sensor. When at least one of the features is collaboratively identified, the perspective n-point algorithm is used to solve the three-dimensional translation vector and rotation matrix of the current camera coordinate system relative to the identified feature. The relative position in the camera coordinate system is converted into the absolute position of the inspection robot in the world coordinate system through the coordinate system homogeneous transformation by combining the feature with the coordinates and posture of the feature in the physical coordinate database; A quantitative calculation step of the coarse positioning confidence level, after each successful execution of the robot coarse positioning step, calculates an initial confidence level used to characterize the reliability of the coarse positioning result; The time decay step of the coarse positioning trust, after obtaining the initial trust, makes the initial trust change dynamically over time according to a preset exponential decay model to obtain a real-time trust that evolves over time, which is used to dynamically quantify and track the uncertainty of the positioning information of the inspection robot.

2. The navigation matching correction method based on the inspection robot according to claim 1, characterized in that: The quantitative calculation step of the coarse positioning confidence includes a basic score calculation sub-step, a basic score adjustment sub-step, a confidence correction sub-step and a sensor verification sub-step; The basic score calculation sub-step determines a basic score positively correlated with the number of SIFT feature matching point pairs used for the perspective n-point algorithm in the robot rough positioning step; The basic score adjustment sub-step determines the feature integrity weight according to the feature object outline integrity or occlusion degree to adjust the basic score by multiplication; The trustworthiness correction sub-step determines an environmental adaptability correction factor based on real-time illumination intensity or image clarity, and performs product correction again on the basic score after product adjustment; The sensor verification sub-step compares the positions of the feature objects calculated by the visual sensor and the lidar sensor, and adds or subtracts the confidence value based on whether the three-dimensional distance deviation between the two is within a preset threshold to obtain an initial confidence.

3. A navigation matching correction method based on an inspection robot according to claim 1 or 2, characterized in that: The mathematical expression of the exponential decay model is: ,in is an attenuation coefficient, and a linear mapping relationship is established between the attenuation coefficient and the moving speed of the inspection robot obtained by the wheel speed odometer. In addition, the attenuation coefficient is also affected by the feature density of the area where the inspection robot is located. When the number of calibrated features within the preset radius of the current position of the inspection robot is greater than a preset number threshold, the attenuation coefficient is multiplied by a first compensation coefficient less than 1; when the number of calibrated features within the preset radius is less than another preset number threshold, the attenuation coefficient is multiplied by a second compensation coefficient greater than 1.

4. The navigation matching and correction method based on the inspection robot according to claim 3, characterized in that: It also includes real-time trust monitoring and compensation triggering steps and trust updating and attenuation resetting steps under multi-feature recognition; The real-time confidence monitoring and compensation triggering step monitors the real-time confidence value in real time. When the real-time confidence value decays below a preset compensation trigger threshold, an active coarse positioning compensation behavior is automatically triggered, and the inspection robot is controlled to scan to search for new available features. The trust update and attenuation reset steps under the multi-feature recognition are as follows: when the inspection robot continuously recognizes multiple features, if the newly calculated initial trust is higher than the current real-time trust after attenuation, the current trust is directly updated to the new initial trust, and the attenuation time is reset.

5. The navigation matching correction method based on the inspection robot according to claim 1, characterized in that: It also includes a secondary positioning fusion correction step for multi-robot interaction, in which each inspection robot transmits a data packet containing its own ID, timestamp, coarse positioning posture, real-time trust and historical trajectory point sequence to other inspection robots according to a preset period. When the inspection robot receives the data packet from the other robot, it extracts the historical trajectory feature sequence of this robot and the other robot, and uses the dynamic time warping DTW algorithm to calculate the similarity between the two trajectory feature sequences; when the similarity meets the preset conditions, it is determined that the path of the other robot and the local robot overlaps, and the positioning data points transmitted by the other robot in the overlapping segment are used as reference data to be processed.

6. The navigation matching correction method based on the inspection robot according to claim 5, characterized in that: The multi-robot interaction secondary positioning fusion correction step also includes a valid cluster screening sub-step and a secondary positioning correction sub-step. The effective cluster screening sub-step processes all reference data to be processed using the density-based spatial clustering DBSCAN algorithm, sets the neighborhood radius and the minimum number of samples, divides the spatially closely spaced positioning data into clusters, and removes isolated data points determined by the algorithm to be noise, retaining only the largest cluster with a data volume exceeding a preset ratio as the effective reference data; The secondary positioning correction sub-step calculates the secondary positioning correction result of the aircraft through a weighted fusion algorithm. The final fusion coordinates of the aircraft are determined by its own coarse positioning result and all valid reference data. The calculation formula is: ,in is the coarse positioning coordinate of this machine, is the valid reference data coordinate, and are their respective weights.

7. The navigation matching and correction method based on the inspection robot according to claim 6, characterized in that: The secondary positioning correction sub-step also includes dividing the real-time trust into multiple levels, so that the weight and Dynamically associated with their respective real-time trust levels, before weighted average calculation, abnormal reference data points are checked and identified, and the weights of reference data points that deviate from the cluster mean by more than a preset multiple of the standard deviation are directly set to zero.

8. The navigation matching and correction method based on the inspection robot according to claim 1, characterized in that: It also includes a fine positioning step. When the inspection robot confirms through positioning that it has entered the preset key component inspection area, the fine positioning step is activated. The camera carried by the inspection robot is used to capture images of the pre-examined fine positioning feature objects from at least three different spatial positions and angles, and the two-dimensional key points of the fine positioning feature objects are extracted from them. The two-dimensional key points are matched with the three-dimensional model of the feature object pre-stored in the feature and physical coordinate database, and the relative posture transformation matrix between the current camera coordinate system and the fine positioning feature object's own coordinate system is solved. Combined with the coordinates of the feature object in the world coordinate system in the database, the current fine positioning coordinates of the inspection robot are obtained.

9. The navigation matching and correction method based on the inspection robot according to claim 8, characterized in that: The method also includes a dynamic attenuation and correction step. The initial accuracy is calibrated according to the image matching error of the two-dimensional key points, and the initial accuracy is dynamically adjusted according to the vibration acceleration amplitude collected by the inspection robot. The initial accuracy deteriorates over time according to the linear attenuation rule. The attenuation expression of the accuracy value is: ,in for The precision of the moment, is the precision reduction factor, is the initial accuracy.

10. The navigation matching and correction method based on the inspection robot according to claim 9, characterized in that: The dynamic attenuation and correction step also includes storing several recent precise fine positioning results through a sliding window. Whenever a new fine positioning result is obtained, the difference between the result and the historical results in the sliding window is calculated to obtain a difference sequence, the average value of the difference sequence is calculated, and the precision reduction factor is adjusted up or down by analyzing the changing trend of the average value over multiple consecutive calculation cycles.

Images