Inspection robot path planning and defect identification method for digital platform

Abstract

The application discloses a path planning and defect identification method for a digital platform-oriented inspection robot, relates to the field of inspection analysis management and control, and is mainly aimed at the case that an inspection technology only relies on visual defect identification, leading to diversified hidden fault missed inspection and a very high false detection rate. The method is characterized in that: through regional point position analysis, path layering screening and conflict point position avoidance, the method performs accurate screening and hierarchical definition of optimal inspection paths, distinguishes different types of path overlap and obstacle avoidance areas, and then builds a linkage temporary inspection task for diversified inspection subjects that cannot be identified by vision, so as to solve the problem of lagging behind in equipment correlation anomaly management and control. Finally, the method innovatively adopts a running mode in which static map navigation and real-time visual navigation logic are alternately operated, combines a three-level classification identification mechanism of an occlusion, and differentially processes rigid, flexible and penetrable occlusion obstacles, thereby abandoning the traditional one-size-fits-all obstacle avoidance logic.

Description

Technical Field

[0001] This invention relates to the field of inspection analysis and control, specifically to a method for path planning and defect identification of inspection robots for digital platforms. Background Technology

[0002] As the digital and intelligent transformation of industry continues to deepen, the equipment density in various industrial sites, power plant rooms, and intelligent computer rooms is constantly increasing. Manual inspection mode can no longer meet the needs of high-frequency, high-precision, and all-weather inspection operations. Inspection robots have become the core carrier for equipment operation and maintenance, defect detection, and risk prevention and control of digital platforms.

[0003] However, the current mainstream inspection robot technology system still has obvious technical shortcomings, making it difficult to adapt to the precision operation requirements of complex and intensive industrial inspection scenarios: At the path planning level, most of them adopt fixed preset inspection paths without combining real-time inspection area scenarios, robot start and stop positions, multiple inspection path overlaps and conflicts, and movement distance outside the area for dynamic adaptation analysis. Path selection relies only on a single distance dimension and does not distinguish between different types of overlapping areas and avoidance difficulties, resulting in chaotic priority division of inspection paths. Problems such as excessively long inspection detour distances, conflicts between multiple robot inspection points, and ineffective avoidance often occur, resulting in low overall inspection efficiency.

[0004] For example, in terms of defect identification and anomaly handling, routine inspections rely solely on a single visual image recognition method. This lacks the ability to identify hidden faults that cannot be captured by vision, such as lag, abnormal noise, and temperature drift. Furthermore, it does not consider the linkage and constraints between inspection equipment. After an anomaly is detected on a single device, it is impossible to simultaneously prevent and control the derivative risks of related and cooperating equipment. There is a time lag between the discovery of anomalies during inspections and the subsequent management and control, which can easily lead to secondary equipment failures.

[0005] In terms of navigation and obstacle avoidance, pure static map navigation or full visual navigation mode are commonly used. Pure static map navigation cannot adapt to non-rigid occlusion obstacles that are temporarily added on site, and is prone to inspection delays and missed points. Full visual navigation requires continuous high-load computing, which places high demands on the computing power of robot hardware, and has a large amount of recognition redundancy and serious battery drain.

[0006] To address the aforementioned technical shortcomings, this paper proposes a path planning and defect identification method for inspection robots on digital platforms. This method addresses industry pain points such as poor adaptability of traditional digital inspection robots, limited obstacle identification, delayed response to abnormal inspections, and missed defects due to multi-device linkage. The aim is to build a fully intelligent inspection system that integrates inspection area execution decision-making, fixed path temporary task setup, and dynamic and static alternating navigation and obstacle avoidance. Summary of the Invention

[0007] The purpose of this invention is to solve the problems mentioned above by proposing a path planning and defect identification method for inspection robots on digital platforms.

[0008] The objective of this invention can be achieved through the following technical solution: a method for path planning and defect identification of inspection robots for digital platforms, the method steps of which are as follows: Step 1: Inspection Area Execution Decision. The current inspection area and the inspection path matched by the inspection robot are used for adaptability analysis. The inspection execution decision is made based on the results of the area and path adaptability analysis to ensure inspection efficiency. Step 2: Establish temporary tasks based on fixed paths. The inspection paths are fixed according to the decision-making process. The main entities are identified based on the fixed inspection paths. Temporary tasks are established based on real-time inspection results to reduce the lag time of inspection and control. Step 3: Alternate route navigation and obstacle avoidance logic. During the inspection phase, navigation is alternated between static map of the inspection area and real-time visual navigation.

[0009] Furthermore, the decision-making process for the inspection area in step one is as follows: Based on the entities required for inspection in the inspection area, determine the inspection points, generate inspection routes based on the distribution of the points, and filter the generated inspection routes to retain the optimal routes. Based on the cabin position set by the inspection robot, the inspection start point is marked. Based on the generated inspection route, the starting point and ending point of the corresponding route are determined. Based on the inspection start point, the distance that the inspection robot needs to move when it is at the starting point and ending point outside the inspection area is obtained and marked as the outside movement distance. A distance set is established based on the inspection route.

[0010] Furthermore, based on the movement routes outside the non-inspection area corresponding to each subset, the overlapping points of the inspection routes of non-current inspection robots corresponding to the movement routes outside the non-inspection area are obtained, as well as the avoidance frequency of the corresponding overlapping points. The avoidance frequency is obtained by estimating the time and overlapping points of the inspection tasks of non-current inspection robots. If the number of overlapping points of the non-inspection area movement route corresponding to the inspection route of the non-current inspection robot is less than the set point number threshold, and the frequency of avoiding the corresponding overlapping points is less than the avoidance frequency threshold, then the corresponding subset will be retained, that is, the corresponding inspection route will be retained. If the number of overlapping points of the non-inspection area movement route corresponding to the inspection route of the non-current inspection robot is not lower than the set point number threshold, or the frequency of avoiding the corresponding overlapping points is not lower than the avoidance frequency threshold, then the corresponding subset will not be retained, that is, the corresponding inspection route will be deleted.

[0011] Furthermore, based on the inspection routes corresponding to the remaining subsets within the set, the inspection robot enters each inspection point within the inspection area to complete the inspection, obtains the overlapping movement distance corresponding to the alternating inspection of the inspection points, and identifies whether all inspection routes are overlapping areas based on the area where the current overlapping movement distance is located. If the overlapping movement distance of the inspection points exceeds the overlapping movement distance threshold during alternating inspections, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as a long-distance unavoidable area. If the overlapping movement distance of the inspection points does not exceed the overlapping movement distance threshold when they are inspected alternately, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as an area that cannot be avoided in the short distance. If the overlapping movement distance of the inspection points exceeds the overlapping movement distance threshold during alternating inspections, and the area where the current overlapping movement distance is located is not the overlapping area of ​​all inspection routes, then the inspection route will be removed. If the overlapping movement distance of the inspection points does not exceed the overlapping movement distance threshold during alternating inspections, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area is marked as a feasible area that has not been avoided.

[0012] Furthermore, based on the remaining inspection routes, the inspection robot's planned inspection route trajectory is collected, and adjustments are made to each area type. Priority is given to adjustments to feasible areas that have not been avoided. The adjusted inspection trajectory is then used to identify long-distance and short-distance areas that cannot be avoided. If the long-distance unavoidable area is below the set long-distance threshold, and the consecutive frequency of the short-distance unavoidable area is below the set short-distance frequency threshold, then the corresponding adjusted inspection route will be set as the preferred inspection route for the current inspection area. When the inspection route is not unique, the priority will be divided according to the actual inspection robot movement distance of the inspection route from shortest to longest, that is, the shorter the movement distance, the higher the priority.

[0013] Furthermore, if the number of consecutive occurrences of long-distance unavoidable areas is not lower than the set long-distance threshold, or the number of consecutive occurrences of short-distance unavoidable areas is not lower than the set short-distance frequency threshold, then the corresponding adjusted inspection route will be set as the alternative inspection route for the current inspection area. When the preferred inspection route conflicts with other inspection tasks, the alternative inspection route is used as a reference to partially occupy the route. The occupied part and the intersection node are determined based on the actual alternative inspection route. If the current preferred inspection route is occupied by the main inspection body, and the real-time route cannot be determined, the alternative inspection route is used as a reference to occupy the route. After the occupation is completed, the route returns to the real-time preferred inspection route.

[0014] Furthermore, the process of establishing a temporary task for the solidified path in step two is as follows: Based on the inspection subject, obtain the corresponding fault type of the inspection subject. If the fault type cannot be identified by visual detection, mark the corresponding subject as a diversified inspection subject and set the corresponding detection and identification parameters as the inspection parameter type based on the fault type. Based on the operational process of the inspection area, the inspection entities that cooperate in the execution are determined. At the same time, the inspection parameter types of the cooperating inspection entities are extracted. If the inspection parameter type is abnormal, the operating parameters of the inspection entity are changed, and the current operating parameters interfere with the cooperating inspection entities. Then, the corresponding cooperating inspection entities are marked as restrictive entities. When inspecting multiple inspection subjects, the fluctuation range of the parameter values ​​corresponding to the inspection parameter types is identified. If the parameter fluctuation range exceeds the set fluctuation span, the system is reported that the current inspection subject is abnormal. At the same time, a temporary task is created, and the inspection subject of the temporary task is set as the subject that cooperates in the inspection. If the parameter fluctuation range does not exceed the set fluctuation span, the inspection task continues to be executed.

[0015] Furthermore, temporary tasks are established by using the cooperating inspection entities corresponding to the current inspection entity as the detection objects, and repeating this process to construct a temporary task monitoring trajectory. The inspection parameter type of the current inspection entity is used as the source of abnormal parameters, and the operating parameters corresponding to the adjacent cooperating inspection entities are used as temporary task parameters. All types of temporary task parameters are identified, and when abnormal fluctuations in parameters are identified, they are reported synchronously.

[0016] Furthermore, the alternation process of the route navigation obstacle avoidance logic in step three is as follows: The frequency of non-adjacent non-rigid obstacles appearing in long-distance unavoidable areas of the inspection area is obtained, as well as the obstacle avoidance inspection frequency of the inspection robot when non-rigid obstacles appear. Here, non-rigid means temporary obstacles set up in the area where the inspection body is located. If the frequency of non-adjacent non-rigid occlusions in long-distance unavoidable areas of the inspection zone corresponds one-to-one with the obstacle avoidance inspection frequency of the inspection robot when non-rigid occlusions appear, then visual navigation is set and the features of the occluded objects are extracted based on visual detection technology; obstacle avoidance decisions are made based on the visual detection results. If the frequency of non-adjacent non-rigid obstructions in long-distance unavoidable areas of the inspection zone does not correspond one-to-one with the obstacle avoidance inspection frequency of the inspection robot when non-rigid obstructions occur, then the inspection will continue to ensure the completion rate of the inspection points.

[0017] Compared with the prior art, the beneficial effects of the present invention are: By implementing a full-process decision-making logic that includes route generation from inspection points, statistics on movement distance outside the area, hierarchical division of distance subsets, prediction of overlapping points and avoidance frequency, identification of overlapping area types, and hierarchical screening of path priorities, the system achieves intelligent and refined adaptation and screening of inspection paths, completely solving the problems of strong blindness and poor adaptability in traditional path planning. Innovatively, it distinguishes between three types of overlapping areas: long-distance unavoidable, short-distance unavoidable, and feasible but not avoided. It then performs targeted trajectory optimization and adjustment, quantifies the difficulty of path avoidance, and avoids excessive detours and inspection failures. In addition, by setting long and short distance thresholds and frequency thresholds, the optimized route is divided into preferred and alternative routes, with priority determined based on travel distance. When the preferred route is occupied or there is a task conflict, the alternative route can be quickly called to complete the partial replacement without interrupting the overall inspection task. This eliminates the problem of missing inspection subjects and significantly improves the adaptability, stability, and fault tolerance of inspection paths in complex industrial scenarios.

[0018] Breaking through the technical limitations of traditional single visual recognition in inspections, we construct a defect identification and anomaly handling system that integrates "visual recognition + multi-source parameter acquisition + equipment linkage prevention and control + dynamic temporary tasks," thereby enhancing the comprehensiveness of digital inspections and their ability to prevent risks in advance. It has the ability to dynamically generate temporary tasks. Based on real-time abnormal parameter fluctuations, it can automatically set up temporary inspection tasks with related and cooperating equipment as the monitoring objects, continuously adding monitoring trajectories, and continuously tracking and monitoring the source of abnormal parameters and linked equipment. This changes the shortcomings of traditional fixed inspections that only detect once and have no follow-up, and can promptly capture potential risks from minor parameter fluctuations. It enables early detection, early warning, and early handling of equipment failures, completely solving the industry problem of abnormality detection and lagging control during inspections, and comprehensively improving the risk prevention and control level of the digital inspection platform.

[0019] It adopts an operation mechanism that alternates between static map navigation and real-time visual navigation logic, combined with three-level classification and identification technology for obstructions, to completely solve the core defects of traditional single navigation mode and one-size-fits-all obstacle avoidance logic, and achieve multiple balances in inspection efficiency, identification accuracy and operational stability. Attached Figure Description

[0020] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0021] Figure 1 This is a schematic block diagram of the process method of the present invention; Figure 2 This is a flowchart illustrating the specific process of the method described in this invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] Please see Figures 1-2 As shown, this invention proposes a path planning and defect identification method for inspection robots oriented towards digital platforms. The method steps are as follows: Step 1: Inspection Area Execution Decision. The current inspection area and the inspection path matched by the inspection robot are used for adaptability analysis. The inspection execution decision is made based on the results of the area and path adaptability analysis to ensure inspection efficiency. Step 2: Establish temporary tasks based on the fixed path. The inspection path is fixed according to the decision and the main body association is identified based on the fixed inspection path. Temporary tasks are established based on the real-time inspection results to quickly respond to inspection anomalies and avoid the problem of timely inspection but lagging control. Step 3: Alternating Route Navigation and Obstacle Avoidance Logic. During the inspection phase, navigation based on the static map of the inspection area and real-time visual navigation alternate. This avoids interference from actual obstacles in the static map navigation and reduces the visual processing intensity of the visual navigation, ensuring the stability of the inspection execution. By statistically analyzing the matching relationship between the frequency of non-rigid obstructions and the frequency of obstacle avoidance inspection, the navigation mode is dynamically switched. In scenarios suitable for static navigation, global static map navigation is used first, which significantly reduces the real-time computation intensity of the visual algorithm, reduces the consumption of robot hardware computing power and battery life, and ensures the completion of inspection points. In scenarios with inefficient routes, visual navigation is activated to achieve dynamic obstacle adaptation, balancing navigation accuracy and computational efficiency. Secondly, this step uses the combined identification of laser point cloud morphology features and image texture features to classify obstructions into three categories: rigid permanent obstructions, flexible temporary obstructions, and penetrable semi-obstructions. This enables differentiated obstacle avoidance decisions, abandons the rigid logic of uniformly classifying all obstructions as high-risk, cancels ineffective obstacle avoidance for temporary lightweight obstructions that have no collision risk and do not affect observation, eliminates blind spots in inspection, and avoids unnecessary interruption of inspection tasks and point shielding issues. Meanwhile, the alternating operation of static maps and visual navigation can simultaneously update static map data in real time, allowing the fixed map to adapt to changes in dynamic on-site scenarios, continuously improving the accuracy and stability of navigation and obstacle avoidance, and ensuring the continuity and reliability of inspection operations in complex dynamic scenarios. The decision-making process for the inspection area in Step 1 is as follows: Based on the entities required for inspection in the inspection area, determine the inspection points, generate inspection routes based on the distribution of the points, and filter the generated inspection routes to retain the optimal routes. Based on the cabin position set by the inspection robot, mark it as the inspection start point. Based on the generated inspection route, determine the start point and end point of the corresponding route. Based on the inspection start point, obtain the distance that the inspection robot needs to move when it is at the start point and end point outside the inspection area, and mark it as the outside movement distance. Establish a distance set based on the inspection route. Based on the movement routes outside the non-inspection area corresponding to each subset, the overlapping points of the inspection routes of non-current inspection robots corresponding to the movement routes outside the non-inspection area are obtained, as well as the avoidance frequency of the corresponding overlapping points. The avoidance frequency is obtained by estimating the time and overlapping points of the inspection tasks of non-current inspection robots. If the number of overlapping points of the non-inspection area movement route corresponding to the inspection route of the non-current inspection robot is less than the set point number threshold, and the frequency of avoiding the corresponding overlapping points is less than the avoidance frequency threshold, then the corresponding subset will be retained, that is, the corresponding inspection route will be retained. If the number of overlapping points of the non-inspection area movement route corresponding to the inspection route of the non-current inspection robot is not lower than the set point number threshold, or the frequency of avoiding the corresponding overlapping points is not lower than the avoidance frequency threshold, then the corresponding subset will not be retained, that is, the corresponding inspection route will be deleted. Based on the inspection routes corresponding to the remaining subsets within the set, the inspection robot enters each inspection point in the inspection area to complete the inspection, obtains the corresponding overlapping movement distance when the inspection points are inspected alternately, and identifies whether all inspection routes are overlapping areas based on the area where the current overlapping movement distance is located. If the overlapping movement distance of the inspection points exceeds the overlapping movement distance threshold during alternating inspections, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as a long-distance unavoidable area. If the overlapping movement distance of the inspection points does not exceed the overlapping movement distance threshold when they are inspected alternately, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as an area that cannot be avoided in the short distance. If the overlapping movement distance of the inspection points exceeds the overlapping movement distance threshold during alternating inspections, and the area where the current overlapping movement distance is located is not the overlapping area of ​​all inspection routes, then the inspection route will be removed. If the overlapping movement distance of the inspection points does not exceed the overlapping movement distance threshold when they are inspected alternately, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area is marked as a feasible area that has not been avoided. Based on the remaining inspection route, the inspection robot's planned inspection route trajectory is collected. Based on the inspection trajectory, long-distance unavoidable areas, short-distance unavoidable areas, and feasible but not implemented avoidance areas are identified. Adjustments are made to each area type, prioritizing adjustments to feasible but not implemented avoidance areas. The adjusted inspection trajectory is then used to identify long-distance unavoidable areas and short-distance unavoidable areas. If the long-distance inescapable area is lower than the set long-distance threshold, and the consecutive frequency of the short-distance inescapable area is lower than the set short-distance frequency threshold, then the corresponding adjusted inspection route will be set as the preferred inspection route for the current inspection area. When the inspection route is not unique, the priority will be divided according to the actual inspection robot movement distance of the inspection route from short to long, that is, the shorter the movement distance, the higher the priority. If the number of consecutive occurrences of long-distance unavoidable areas is not lower than the set long-distance threshold, or the number of consecutive occurrences of short-distance unavoidable areas is not lower than the set short-distance frequency threshold, then the corresponding adjusted inspection route will be set as the alternative inspection route for the current inspection area. When the preferred inspection route conflicts with other inspection tasks, the alternative inspection route is used as a reference to partially occupy the route. The occupied part and the intersection node are determined based on the actual alternative inspection route. If the current preferred inspection route is occupied by an inspection subject, and the real-time route cannot be determined, the alternative inspection route is used as a reference to occupy the route. After the occupation is completed, the route returns to the real-time preferred inspection route. This ensures that all inspection subjects complete their inspections and also ensures that no inspection subject is missed when changing routes, thus improving the feasibility of route changes. It needs to be explained that the threshold for the number of points and the threshold for avoidance frequency are scenario-adaptive experience thresholds. These thresholds are obtained by collecting historical data of more than 30 days of routine multi-robot collaborative inspections of the target inspection area, statistically analyzing the distribution range of the number of overlapping points and the frequency of daily avoidance for different inspection routes, and removing extreme abnormal data. The industry's conventional safety inspection threshold is then used as the benchmark threshold. At the same time, the threshold can be dynamically adjusted according to the density of equipment and the number of robots in the inspection area. The denser the equipment and the more inspection robots there are, the lower the threshold should be to ensure that route conflicts are minimized. In scenarios with sparse equipment, the threshold can be raised appropriately to expand the range of selectable routes.

[0025] Overlapping movement distance threshold: Based on the normal moving speed of the inspection robot and the standard inspection time of a single point, combined with the setting of the robot's maximum effective detour tolerance distance, through multiple field tests, the maximum critical movement distance that does not affect the overall inspection sequence and does not generate invalid energy consumption during the robot's overlapping movement is statistically determined. This critical value is set as the overlapping movement distance threshold to distinguish between long and short distance overlapping avoidance areas.

[0026] Long-distance threshold and short-distance frequency threshold: The long-distance threshold is set based on the average detour distance of the regional inspection route. It is based on the average overlapping avoidance distance of all effective inspection routes in the region. Areas exceeding the average are judged as long-distance unavoidable areas, and areas below the average are judged as short-distance manageable areas. The short-distance continuous frequency threshold is determined by statistically analyzing the fault association frequency and inspection delay frequency of short-distance avoidance areas in historical inspections. The maximum continuous frequency without inspection faults and task delays is taken as the critical threshold to ensure that short-distance avoidance does not accumulate and affect inspection efficiency.

[0027] The process of establishing a temporary task for the solidified path in step two is as follows: After determining the inspection route, the inspection robot performs the inspection task according to the set inspection route. It identifies the inspection subject within the route according to the determined inspection route, and obtains the fault type of the corresponding inspection subject. If the fault type cannot be identified by visual detection, the corresponding subject is marked as a diversified inspection subject. The detection and identification parameters corresponding to the fault type are set as the inspection parameter type, such as temperature, noise value, etc. Based on the operational flow of the inspection area, the cooperating inspection entities are determined, and the inspection parameter types of the cooperating inspection entities are extracted. If the inspection parameter type is abnormal, the operating parameters of the inspection entity are changed, and the current operating parameters interfere with the cooperating inspection entity. Then, the corresponding cooperating inspection entity is marked as a restrictive entity. For example, if the inspection parameter of the inspection entity is fan jamming, the noise is very loud and cannot be detected visually. At this time, the temperature of the inspection entity will remain high for a long time, and there is an overload risk. In this case, the cooperating inspection entity cannot cooperate effectively, and there is a risk of short circuit or open circuit. When inspecting multiple inspection subjects, multi-source data is collected according to the corresponding inspection parameter type, and the fluctuation range of the parameter value corresponding to the inspection parameter type is identified by the sensor. If the parameter fluctuation range exceeds the set fluctuation span, the system is reported that there is an anomaly in the current inspection subject. At the same time, a temporary task is created, and the inspection subject of the temporary task is set as the subject to cooperate in the inspection. If the parameter fluctuation range does not exceed the set fluctuation span, the inspection task continues to be executed. Temporary tasks are established by using the cooperating inspection entities corresponding to the current inspection entity as the detection objects, and repeating this process to construct a temporary task monitoring trajectory. The inspection parameter type of the current inspection entity is used as the source of abnormal parameters, and the operating parameters corresponding to the adjacent cooperating inspection entities are used as temporary task parameters. All types of temporary task parameters are identified, and when abnormal parameter fluctuations are detected, the corresponding fault type of the abnormal parameters of the inspection entity is reported synchronously. After receiving the report, the system performs abnormal maintenance on the inspection entity and abnormal prevention and control on the cooperating inspection entities. It's important to explain the parameter fluctuation range threshold: This threshold is specific to the equipment's operating conditions and is determined based on the factory-rated operating parameters and industry standard operating parameter ranges for different types of inspection equipment, combined with the parameter fluctuation tolerance range during normal equipment operation. For various inspection parameters such as temperature, noise, operating current, and voltage, multiple sets of parameter data are collected under normal steady-state operation. The standard fluctuation range is calculated, and the range exceeding the maximum steady-state fluctuation value is set as the abnormal fluctuation range threshold. Furthermore, differentiated adaptation calibration is performed for industrial equipment of different power and models to avoid false alarms and missed alarms caused by a uniform threshold, ensuring the accuracy of parameter anomaly identification.

[0028] The alternation process of the route navigation obstacle avoidance logic in step three is as follows: When the inspection robot performs inspection tasks according to the inspection route, it moves based on the inspection route and the static map of the inspection area. It should be explained that navigation and obstacle avoidance are completed based on the global static environment map. This study aims to obtain the frequency of non-adjacent non-rigid obstacles in long-distance, unavoidable areas within the inspection zone, as well as the obstacle avoidance frequency of the inspection robot when non-rigid obstacles appear. Non-rigid obstacles refer to temporary obstacles within the inspection area, such as equipment dust covers, temporary cable arrangements, lightweight shielding panels, and stacked maintenance consumables. These obstacles do not affect equipment inspection and observation and pose no collision risk. Current technology uniformly classifies all obstacles as high-risk obstacles, forcing detours or stopping inspections, resulting in the blocking of many normally inspectable points, creating blind spots and task interruptions. If the frequency of non-adjacent non-rigid obstructions in long-distance, unavoidable areas of the inspection zone corresponds one-to-one with the obstacle avoidance inspection frequency of the inspection robot when non-rigid obstructions occur, it indicates that the inspection robot's route movement is inefficient. Visual navigation is set up, and visual detection technology is used to extract differentiated scene features such as the material penetration, obstruction persistence, and collision risk level of the obstructing objects. Through joint identification using laser point cloud morphology features and image texture features, obstructions in the inspection scene are classified into three levels: Level 1 rigid permanent obstructions (walls, cabinets, metal structures), Level 2 flexible temporary obstructions (dust covers, temporary cables, lightweight baffles), and Level 3 penetrable semi-obstructions (mesh obstructions, thin consumables). Obstruction avoidance decisions are made based on the visual detection results to avoid unnecessary obstacle avoidance. If the frequency of non-adjacent non-rigid obstructions in long-distance unavoidable areas of the inspection zone does not correspond one-to-one with the obstacle avoidance inspection frequency of the inspection robot when non-rigid obstructions occur, it indicates that the inspection zone is suitable for the current static map navigation, and the inspection completion rate of the inspection points will continue to be ensured. The visual algorithm intensity will be reduced, and the visual recognition intensity will be reduced without affecting the inspection efficiency.

[0029] Non-rigid obstruction frequency and obstacle avoidance inspection frequency matching threshold: obtained through on-site routine inspection test calibration, continuously collect data on the occurrence frequency, duration and corresponding obstacle avoidance operation frequency of various non-rigid temporary obstructions in the inspection area, and establish a frequency matching database. When the frequency of obstruction and the frequency of obstacle avoidance are completely one-to-one, it proves that the static map cannot adapt to the dynamic occlusion scenario and is judged as an inefficient route movement scenario. When the frequencies of the two do not match, it proves that the existing static map navigation can meet the inspection requirements and there is no need to start visual navigation. At the same time, the frequency matching rules are dynamically calibrated by combining the existence attributes of the obstruction and the dynamic change frequency of the scene to adapt to inspection scenarios of different complexity.

[0030] The three-level classification threshold for occlusions is determined based on three core indicators: laser point cloud morphological feature parameters, image texture penetration rate, and collision risk coefficient. Through extensive sample training and field experiments, thresholds are set for material penetration, occlusion duration, and collision risk level. Permanent rigid occlusions have a duration approaching infinity, a penetration rate of 0%, and a collision risk coefficient of 100%. Secondary flexible temporary occlusions have a short duration, low penetration rate, and no collision risk. Tertiary penetrable occlusions have a penetration rate higher than the set penetration threshold and no collision risk. This achieves accurate classification of the three types of occlusions.

[0031] In summary, through the coordinated efforts of three core steps, a fully automated, high-precision, and highly stable robotic inspection system adapted to digital platforms is constructed, comprehensively breaking through the bottlenecks of traditional inspection technologies from three core dimensions: path planning, defect identification, and navigation and obstacle avoidance.

[0032] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for path planning and defect identification of inspection robots for digital platforms, characterized in that, The steps are as follows: Step 1: Inspection Area Execution Decision. The current inspection area and the inspection path matched by the inspection robot are used for adaptability analysis. The inspection execution decision is made based on the results of the area and path adaptability analysis to ensure inspection efficiency. Step 2: Establish temporary tasks based on fixed paths. The inspection paths are fixed according to the decision-making process. The main entities are identified based on the fixed inspection paths. Temporary tasks are established based on real-time inspection results to reduce the lag time of inspection and control. Step 3: Alternate route navigation and obstacle avoidance logic. During the inspection phase, navigation is alternated between static map of the inspection area and real-time visual navigation.

2. The method for path planning and defect identification of inspection robots for digital platforms according to claim 1, characterized in that, The decision-making process for the inspection area in Step 1 is as follows: Based on the entities required for inspection in the inspection area, determine the inspection points, generate inspection routes based on the distribution of the points, and filter the generated inspection routes to retain the optimal routes. Based on the cabin position set by the inspection robot, the inspection start point is marked. Based on the generated inspection route, the starting point and ending point of the corresponding route are determined. Based on the inspection start point, the distance that the inspection robot needs to move when it is at the starting point and ending point outside the inspection area is obtained and marked as the outside movement distance. A distance set is established based on the inspection route.

3. The method for path planning and defect identification of inspection robots for digital platforms according to claim 2, characterized in that, Based on the movement routes outside the non-inspection area corresponding to each subset, the overlapping points of the inspection routes of non-current inspection robots corresponding to the movement routes outside the non-inspection area are obtained, as well as the avoidance frequency of the corresponding overlapping points. The avoidance frequency is obtained by estimating the time and overlapping points of the inspection tasks of non-current inspection robots. If the number of overlapping points of the non-inspection area movement route corresponding to the inspection route of the non-current inspection robot is less than the set point number threshold, and the frequency of avoiding the corresponding overlapping points is less than the avoidance frequency threshold, then the corresponding subset will be retained, that is, the corresponding inspection route will be retained. If the number of overlapping points of the non-inspection area movement route corresponding to the inspection route of the non-current inspection robot is not lower than the set point number threshold, or the frequency of avoiding the corresponding overlapping points is not lower than the avoidance frequency threshold, then the corresponding subset will not be retained, that is, the corresponding inspection route will be deleted.

4. The method for path planning and defect identification of inspection robots for digital platforms according to claim 3, characterized in that, Based on the inspection routes corresponding to the remaining subsets within the set, the inspection robot enters each inspection point in the inspection area to complete the inspection, obtains the corresponding overlapping movement distance when the inspection points are inspected alternately, and identifies whether all inspection routes are overlapping areas based on the area where the current overlapping movement distance is located. If the overlapping movement distance of the inspection points exceeds the overlapping movement distance threshold during alternating inspections, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as a long-distance unavoidable area. If the overlapping movement distance of the inspection points does not exceed the overlapping movement distance threshold when they are inspected alternately, and the current overlapping movement distance is located in the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as an area that cannot be avoided in the short distance. If the overlapping movement distance of the inspection points exceeds the overlapping movement distance threshold during alternating inspections, and the area where the current overlapping movement distance is located is not the overlapping area of ​​all inspection routes, then the inspection route will be removed; if the overlapping movement distance of the inspection points does not exceed the overlapping movement distance threshold during alternating inspections, and the area where the current overlapping movement distance is located is the overlapping area of ​​all inspection routes, then the current overlapping area will be marked as a feasible area that has not been avoided.

5. The method for path planning and defect identification of inspection robots for digital platforms according to claim 4, characterized in that, Based on the remaining inspection route, the inspection robot's planned inspection route trajectory is collected, and adjustments are made to each area type. Priority is given to adjustments to feasible areas that have not been avoided. The adjusted inspection trajectory is then used to identify areas that cannot be avoided over long distances and areas that cannot be avoided over short distances. If the long-distance unavoidable area is below the set long-distance threshold, and the consecutive frequency of the short-distance unavoidable area is below the set short-distance frequency threshold, then the corresponding adjusted inspection route will be set as the preferred inspection route for the current inspection area. When the inspection route is not unique, the priority will be divided according to the actual inspection robot movement distance of the inspection route from shortest to longest, that is, the shorter the movement distance, the higher the priority.

6. The method for path planning and defect identification of inspection robots for digital platforms according to claim 5, characterized in that, If the number of consecutive occurrences of long-distance unavoidable areas is not lower than the set long-distance threshold, or the number of consecutive occurrences of short-distance unavoidable areas is not lower than the set short-distance frequency threshold, then the corresponding adjusted inspection route will be set as the alternative inspection route for the current inspection area. When the preferred inspection route conflicts with other inspection tasks, the alternative inspection route is used as a reference to partially occupy the route. The occupied part and the intersection node are determined based on the actual alternative inspection route. If the real-time route cannot be determined, the alternative inspection route is used as a reference to occupy the route. After the occupation is completed, the route returns to the real-time preferred inspection route.

7. The method for path planning and defect identification of inspection robots for digital platforms according to claim 1, characterized in that, The process of establishing a temporary task for the solidified path in step two is as follows: Based on the inspection subject, obtain the corresponding fault type of the inspection subject. If the fault type cannot be identified by visual detection, mark the corresponding subject as a diversified inspection subject and set the corresponding detection and identification parameters as the inspection parameter type based on the fault type. Based on the operational process of the inspection area, the inspection entities that cooperate in the execution are determined. At the same time, the inspection parameter types of the cooperating inspection entities are extracted. If the inspection parameter type is abnormal, the operating parameters of the inspection entity are changed, and the current operating parameters interfere with the cooperating inspection entities. Then, the corresponding cooperating inspection entities are marked as restrictive entities. When inspecting multiple inspection subjects, identify the fluctuation range of the parameter values ​​corresponding to the inspection parameter types. If the parameter fluctuation range exceeds the set fluctuation span, report to the system that there is an anomaly in the current inspection subject, and at the same time create a temporary task and set the inspection subject of the temporary task as the subject that cooperates in the inspection. If the parameter fluctuation range does not exceed the set fluctuation span, the inspection task will continue.

8. The method for path planning and defect identification of inspection robots for digital platforms according to claim 7, characterized in that, Temporary tasks are established by using the cooperating inspection entities corresponding to the current inspection entity as the detection objects, and repeating this process to construct a temporary task monitoring trajectory. The inspection parameter type of the current inspection entity is used as the source of abnormal parameters, and the operating parameters corresponding to the adjacent cooperating inspection entities are used as temporary task parameters. All types of temporary task parameters are identified, and when abnormal parameter fluctuations are detected, they are reported synchronously.

9. The method for path planning and defect identification of inspection robots for digital platforms according to claim 1, characterized in that, The alternation process of the route navigation obstacle avoidance logic in step three is as follows: The frequency of non-adjacent non-rigid obstacles appearing in long-distance unavoidable areas of the inspection area is obtained, as well as the obstacle avoidance inspection frequency of the inspection robot when non-rigid obstacles appear. Here, non-rigid means temporary obstacles set up in the area where the inspection body is located. If the frequency of non-adjacent non-rigid occlusions in long-distance unavoidable areas of the inspection zone corresponds one-to-one with the obstacle avoidance inspection frequency of the inspection robot when non-rigid occlusions appear, then visual navigation is set and the features of the occluded objects are extracted according to visual detection technology, and obstacle avoidance decisions are made based on the visual detection results. If the frequency of non-adjacent non-rigid obstructions in long-distance unavoidable areas of the inspection zone does not correspond one-to-one with the obstacle avoidance inspection frequency of the inspection robot when non-rigid obstructions occur, then the inspection will continue to ensure the completion rate of the inspection points.

Images