A control method for remote obstacle avoidance of a power inspection robot

Abstract

The present application provides a control method for remote obstacle avoidance of a power inspection robot, which relates to the technical field of power inspection and includes the following steps: After receiving the start inspection signal of the inspection robot, the road image is obtained in real time, the target section is determined, and the corresponding local image is obtained. If there is an obstacle to be avoided in the local image, it is predicted whether the inspection robot can determine the type of the obstacle to be avoided; if it is predicted that the inspection robot cannot determine the type of the obstacle to be avoided, a dynamic obstacle avoidance strategy is obtained according to the obstacle to be avoided and the local image; The inspection robot is controlled to avoid obstacles with a dynamic obstacle avoidance strategy when approaching the obstacle to be avoided, and continue to travel along the preset inspection path after avoiding the obstacle to be avoided. The present application overcomes the limitations of the automatic obstacle avoidance system of the inspection robot for special scenarios, ensures the continuity of the inspection process, improves the reliability of the inspection robot, and guarantees the safety of the inspection robot and surrounding power equipment.

Description

Technical Field

[0001] This application relates to the technical field of power inspection, and particularly to a control method for remote obstacle avoidance of a power inspection robot. Background Art

[0002] In the current field of power inspection, inspection robots are widely used to improve the inspection efficiency and accuracy. Most of the existing inspection robots rely on an automatic obstacle avoidance system composed of their own sensors to judge faults in the traveling path and perform adaptive obstacle avoidance. However, this method has certain limitations.

[0003] In the prior art, power inspection robots usually handle obstacles only when they encounter them, lacking the ability to pre-judge obstacles. This causes the robot to spend time identifying and making decisions when encountering obstacles, thus reducing the inspection efficiency. Summary of the Invention

[0004] In view of the above defects or deficiencies in the prior art, this application aims to provide a control method for remote obstacle avoidance of a power inspection robot, including the following steps:

[0005] After receiving the start inspection signal of the inspection robot, obtain the road image in real time, where the road image is an image composed of all inspection roads available for the inspection robot to travel in the inspection site;

[0006] Determine the target section and obtain the local image corresponding to the target section, where the target section is a continuous section of road in the road image that the inspection robot is about to travel along the preset inspection path;

[0007] If there are obstacles to be avoided in the local image, then pre-judge whether the inspection robot can determine the type of the obstacle to be avoided;

[0008] If it is pre-judged that the inspection robot cannot determine the type of the obstacle to be avoided, then obtain a dynamic obstacle avoidance strategy according to the obstacle to be avoided and the local image;

[0009] Control the inspection robot to avoid obstacles with the dynamic obstacle avoidance strategy when it travels to the obstacle to be avoided, and continue to travel along the preset inspection path after avoiding the obstacle to be avoided.

[0010] According to the technical solution provided by the embodiment of this application, the determination of the target section specifically includes the following steps:

[0011] Set a first preset duration according to the initial traveling speed of the inspection robot and the data processing period of the remote obstacle avoidance system; the data processing period is the required duration from the time when the remote obstacle avoidance system receives the road image until the dynamic obstacle avoidance strategy is generated;

[0012] Taking the current coordinates of the inspection robot as the starting point, the road that will be traveled at the initial traveling speed within the next first preset time period is determined as the target section.

[0013] According to the technical solution provided by the embodiment of the present application, the step of predicting the type of the obstacle to be avoided for which the inspection robot can determine the obstacle to be avoided specifically includes the following steps:

[0014] Determine whether the inspection robot can recognize all the feature categories of the obstacle to be avoided, and the feature categories at least include color feature, shape feature, and size feature;

[0015] If the inspection robot can recognize all the feature categories of the obstacle to be avoided, then retrieve and traverse the automatic obstacle avoidance database of the inspection robot; the automatic obstacle avoidance database includes multiple preset obstacle types, and a preset feature set corresponding to each preset obstacle type, and the preset feature set includes multiple preset feature categories;

[0016] If a preset obstacle type corresponding to all the feature categories of the obstacle to be avoided is matched, it is determined that the inspection robot can determine the type of the obstacle to be avoided.

[0017] According to the technical solution provided by the embodiment of the present application, after determining whether the inspection robot can recognize all the feature categories of the obstacle to be avoided, the following steps are further included:

[0018] If the inspection robot cannot recognize at least one of the feature categories of the obstacle to be avoided, or, the inspection robot can recognize all the feature categories of the obstacle to be avoided and no preset obstacle type corresponding to all the feature categories of the obstacle to be avoided is matched, it is determined that the inspection robot cannot determine the type of the obstacle to be avoided.

[0019] According to the technical solution provided by the embodiment of the present application, the step of determining whether the inspection robot can recognize all the feature categories of the obstacle to be avoided specifically includes the following steps:

[0020] Obtain the first feature factor corresponding to each feature category of the obstacle to be avoided, and retrieve the recognition threshold of each feature category of the inspection robot;

[0021] If the first feature factor corresponding to each feature category is within the recognition threshold of the feature category, it is determined that the inspection robot can recognize all the feature categories of the obstacle to be avoided.

[0022] According to the technical solution provided by the embodiment of the present application, after determining the target road section and obtaining the local image corresponding to the target road section, the following steps are further included:

[0023] Judge whether there is an obstacle on the target road section;

[0024] If there is an obstacle on the target road section, judge the attribute category of the obstacle on the target road section, and the attribute category includes general obstacles and special obstacles;

[0025] If the attribute category of the obstacle on the target road section is the special obstacle, it is determined that there is an obstacle to be avoided in the local image.

[0026] According to the technical solution provided by the embodiment of the present application, after receiving the start inspection signal of the inspection robot, the following steps are further included:

[0027] Judge whether the inspection robot is inspecting the inspection site for the first time;

[0028] The step of judging the attribute category of the obstacle on the target road section specifically includes the following steps:

[0029] If the inspection robot is inspecting the inspection site for the first time, retrieve the first inspection log of the inspection site, and the first inspection log records the first inspection information of other robots inspecting the inspection site; the first inspection information at least includes the first obstacle coordinates corresponding to the previous obstacles encountered by other robots during the inspection, and the obstacle avoidance results of other robots for the previous obstacles, and the obstacle avoidance results include successful obstacle avoidance and failed obstacle avoidance;

[0030] If the first obstacle coordinates identical to the coordinates of the obstacle on the target road section are matched in the first inspection log and the obstacle avoidance success rate of other robots for the previous obstacles matching the obstacle on the target road section is greater than the first preset threshold, the attribute category of the obstacle on the target road section is determined to be the general obstacle.

[0031] According to the technical solution provided by the embodiment of the present application, the step of judging the attribute category of the obstacle on the target road section further includes the following steps:

[0032] If the inspection robot is not inspecting the inspection site for the first time, retrieve the second inspection log of the inspection robot, and the second inspection log records the second inspection information of the inspection robot inspecting the inspection site; the second inspection information at least includes the second obstacle coordinates corresponding to the previous obstacles encountered by the inspection robot during the inspection, and the obstacle avoidance results of the inspection robot for the previous obstacles;

[0033] If the second obstacle coordinates that match the coordinates of the obstacle on the target section are found in the second inspection log and the obstacle avoidance success rate of the inspection robot for the previous obstacle that matches the obstacle on the target section is greater than the first preset threshold, the attribute category of the obstacle on the target section is determined as the general obstacle.

[0034] According to the technical solution provided by the embodiment of the present application, before obtaining the first feature factors corresponding to the respective feature categories of the obstacle to be avoided, the following steps are further included:

[0035] Obtain the light intensity of the surrounding environment where the obstacle to be avoided is located;

[0036] The obtaining of the first feature factors corresponding to the respective feature categories of the obstacle to be avoided specifically includes the following steps:

[0037] If the light intensity is less than the strong light threshold and greater than the dim light threshold, obtain the first feature factors corresponding to the respective feature categories of the obstacle to be avoided.

[0038] According to the technical solution provided by the embodiment of the present application, the automatic obstacle avoidance database further includes a preset obstacle avoidance strategy corresponding to each preset obstacle type; after obtaining the dynamic obstacle avoidance strategy based on the obstacle to be avoided and the local image, the following steps are further included:

[0039] Use the type of the obstacle to be avoided as a preset obstacle type, use all the feature categories of the obstacle to be avoided as the preset feature set corresponding to this preset obstacle type, and use the dynamic obstacle avoidance strategy of the type of the obstacle to be avoided as the preset obstacle avoidance strategy corresponding to this preset obstacle type, and update the automatic obstacle avoidance database.

[0040] Compared with the prior art, the beneficial effects of the present application are as follows: The present application actively collects real-time road images through a remote obstacle avoidance system, determines in advance the possible obstacles in the target section, pre-judges whether the automatic obstacle avoidance system of the inspection robot has the ability to identify or match the type of the obstacle, and when it is predicted that the obstacle avoidance ability of the inspection robot for the upcoming obstacle is limited, actively plan in advance to obtain a dynamic obstacle avoidance strategy. This pre-judgment ability allows the robot to make preparations in advance and improve the inspection efficiency. At the same time, when it is predicted that the inspection robot cannot determine the type of the obstacle, the present application can also generate a dynamic obstacle avoidance strategy based on the obstacle and the local image. This flexible processing method allows the inspection robot to quickly respond when encountering an obstacle, avoids unnecessary time waste, and improves the inspection efficiency, as well as the adaptability and reliability of the robot. Description of the Drawings

[0041] Figure 1This is the flowchart of the steps of the control method for remote obstacle avoidance of the power inspection robot provided by the embodiments of the present application. Detailed implementation manners

[0042] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention, rather than limiting the invention. In addition, it should be noted that for the convenience of description, only the parts related to the invention are shown in the drawings.

[0043] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and embodiments.

[0044] Embodiment 1

[0045] As mentioned in the background art, in view of the problems in the prior art, the present application proposes a control method for remote obstacle avoidance of a power inspection robot. The control method takes the remote obstacle avoidance system as the main body. The remote obstacle avoidance system relies on a high-performance server or cloud computing platform, and can remotely monitor and control the inspection robot to realize remote operation and troubleshooting of the robot. For example, when the automatic obstacle avoidance system fails or cannot handle complex obstacles, the inspection robot will send a help request to the remote obstacle avoidance system. The remote obstacle avoidance system can intervene and take over the control of the robot to ensure the smooth progress of the inspection task. However, if the response is made after waiting for the inspection robot to send a help request, the response is often relatively lagged. Therefore, this solution is proposed. The control method is described below, including the following steps:

[0046] S1. After receiving the start inspection signal of the inspection robot, the road image is obtained in real time. The road image is an image composed of all the inspection roads in the inspection site where the inspection robot can travel.

[0047] Specifically, before the inspection robot starts the inspection, the remote obstacle avoidance system performs an initialization operation to ensure a stable communication connection with various sensors in the inspection site and is ready to receive the start inspection signal from the inspection robot. At the same time, the remote obstacle avoidance system loads the required algorithm library and data model so that data analysis and decision-making can be quickly performed in the subsequent processing process.

[0048] S2. Determine the target section and obtain the local image corresponding to the target section. The target section is a continuous section of the road in the road image that the inspection robot is about to travel along the preset inspection path.

[0049] Specifically, the preset inspection path can be a series of preset coordinate points or specific road signs. By analyzing the road image, a continuous section of the road that the inspection robot is about to travel is determined as the target section, and the local image corresponding to the target section is extracted.

[0050] In a preferred embodiment, the determination of the target section specifically includes the following steps:

[0051] According to the initial traveling speed of the inspection robot and the data processing cycle of the remote obstacle avoidance system, a first preset duration is set; the data processing cycle is the required duration from the time when the remote obstacle avoidance system receives the road image until the dynamic obstacle avoidance strategy is generated;

[0052] Taking the current coordinate of the inspection robot as the starting point, the road that will be traveled at the initial traveling speed within the next first preset duration is determined as the target section.

[0053] Specifically, the initial traveling speed of the inspection robot is a preset value before the start of the inspection task. The data processing cycle is the time required for a series of operations such as analyzing the image, predicting the type of obstacle, and generating the dynamic obstacle avoidance strategy after the remote obstacle avoidance system receives the road image. The average data processing cycle can be statistically obtained through multiple tests on the system.

[0054] Exemplarily, it can be calculated by the formula: First preset duration = data processing cycle × initial traveling speed coefficient (a coefficient determined according to the actual situation to ensure that the robot does not travel too far beyond the control range within the data processing cycle). Exemplarily, the initial traveling speed coefficient is inversely proportional to the initial traveling speed. The initial traveling speed coefficient can be obtained in the following way: retrieve and traverse the coefficient database, which includes several initial traveling speed intervals and the corresponding reference coefficients for each interval; take the reference coefficient corresponding to the initial traveling speed interval that the initial traveling speed satisfies as the initial traveling speed coefficient. Among them, the larger the initial traveling speed interval, the smaller the corresponding reference coefficient. Exemplarily, a reference coefficient corresponding to an initial traveling speed interval [10m / s, 13m / s] is 0.5, and a reference coefficient corresponding to an initial traveling speed interval [2m / s, 5m / s] is 1.5. Taking the current coordinate of the inspection robot as the starting point, according to the initial traveling speed and the first preset duration, calculate the distance that the robot will travel at the initial traveling speed within the next first preset duration, and combine the preset inspection path and the road layout of the current site to determine the road within this distance range as the target section.

[0055] In a preferred embodiment, after determining the target section and obtaining the local image corresponding to the target section, the following steps are further included:

[0056] Determine whether there is an obstacle on the target section;

[0057] Specifically, perform image analysis and processing on the local image corresponding to the target section, and use image recognition technology to detect whether there are objects in the local image that may affect the movement of the inspection robot. If an object is detected, it is preliminarily determined that there is an obstacle on the target section.

[0058] If there is the obstacle on the target section, determine the attribute category of the obstacle on the target section, and the attribute category includes general obstacles and special obstacles;

[0059] Specifically, general obstacles can be understood as obstacles that are common in the inspection site and can be successfully identified and avoided by the automatic obstacle avoidance system of most inspection robots, while special obstacles are obstacles that are not common in the inspection site and may not be recognized by the automatic obstacle avoidance system of most inspection robots or cannot be avoided successfully even if recognized.

[0060] If the attribute category of the obstacle on the target section is the special obstacle, it is determined that there is an obstacle to be avoided in the local image.

[0061] In a preferred embodiment, after receiving the start inspection signal of the inspection robot, the following steps are further included:

[0062] Determine whether the inspection robot is inspecting the inspection site for the first time;

[0063] The determination of the attribute category of the obstacle on the target section specifically includes the following steps:

[0064] If the inspection robot is inspecting the inspection site for the first time, retrieve the first inspection log of the inspection site, and the first inspection log records the first inspection information of other robots inspecting the inspection site; the first inspection information at least includes the first obstacle coordinates corresponding to the previous obstacles encountered by other robots during the inspection, and the obstacle avoidance results of other robots for the previous obstacles, and the obstacle avoidance results include successful obstacle avoidance and failed obstacle avoidance;

[0065] If the first obstacle coordinates identical to the coordinates of the obstacle on the target section are matched in the first inspection log and the obstacle avoidance success rate of other robots for the previous obstacles matching the obstacle on the target section is greater than the first preset threshold, the attribute category of the obstacle on the target section is determined to be the general obstacle.

[0066] Specifically, if the first obstacle coordinates identical to those of the obstacle on the target section are not found in the first patrol log, or if the first obstacle coordinates identical to those of the obstacle on the target section are found in the first patrol log and the obstacle avoidance success rate of other robots for the previous obstacle matching the obstacle on the target section is less than or equal to the first preset threshold, then the attribute category of the obstacle on the target section is determined as the special obstacle, that is, it is determined that there is an obstacle to be avoided in the local image, and the remote obstacle avoidance system needs to provide assistance subsequently.

[0067] Specifically, if the coordinates of the obstacle on the target section are the same as the first obstacle coordinates, it is very likely that this obstacle on the target section is the previous obstacle that also appeared in previous patrols. Then, by checking the obstacle avoidance results of other robots (referring to other patrol robots except the patrol robot used in this solution currently), it is determined whether this obstacle on the target section often appears and is easily avoided successfully. If the first obstacle coordinates identical to those of the obstacle on the target section are found in the first patrol log and the obstacle avoidance success rate of other robots for the previous obstacle matching the obstacle on the target section is greater than the first preset threshold (which can be set to 80%), it indicates that the obstacle on the target section often appears and is easily avoided successfully, and it is then identified as an ordinary obstacle. The remote obstacle avoidance system does not provide assistance, and only the patrol robot relies on its own automatic obstacle avoidance system to successfully avoid it.

[0068] In this embodiment, it is considered that if the patrol robot enters a patrol site for the first time and conducts a patrol on this patrol site, then the previous experience of other patrol robots can be combined to judge the situation of the obstacles that appear. If the experience of other robots shows that a certain type of obstacle has caused difficulties in recognition or obstacle avoidance failure many times, then when the remote obstacle avoidance system discovers a similar obstacle during the current patrol, it can be aware in advance of the recognition and avoidance risks that the patrol robot may face, and thus conduct monitoring and intervention more targeted.

[0069] In a preferred embodiment, the judging of the attribute category of the obstacle on the target section further includes the following steps:

[0070] If the patrol robot is not patrolling the patrol site for the first time, then retrieve the second patrol log of the patrol robot. The second patrol log records the second patrol information of the patrol robot patrolling the patrol site; the second patrol information at least includes the second obstacle coordinates corresponding to the previous obstacles encountered by the patrol robot during the patrol, and the obstacle avoidance results of the patrol robot for the previous obstacles.

[0071] If the second obstacle coordinates identical to the coordinates of the obstacles on the target section are matched in the second patrol log and the obstacle avoidance success rate of the patrol robot for the previous obstacles matched with the obstacles on the target section is greater than the first preset threshold, then the attribute category of the obstacles on the target section is determined to be the general obstacles.

[0072] Specifically, when the patrol robot is not entering the patrol site for the first time, the previous patrol log of itself is retrieved to judge the current obstacle situation. If it is found that the robot has successfully avoided a certain type of obstacles with specific sizes and colors many times in the past, then when similar obstacles are encountered again, the remote obstacle avoidance system marks them as ordinary obstacles. Otherwise, they are marked as special obstacles, and a conclusion that there are obstacles to be avoided in the local image is drawn to execute the following steps.

[0073] S3. If there are obstacles to be avoided in the local image, then it is predicted whether the patrol robot can determine the type of the obstacles to be avoided.

[0074] In a preferred embodiment, the prediction of whether the patrol robot can determine the type of the obstacles to be avoided specifically includes the following steps:

[0075] Judge whether the patrol robot can recognize all the characteristic categories of the obstacles to be avoided, and the characteristic categories at least include color characteristics, shape characteristics, and size characteristics.

[0076] Further, the judgment of whether the patrol robot can recognize all the characteristic categories of the obstacles to be avoided specifically includes the following steps:

[0077] Obtain the first characteristic factors corresponding to each of the characteristic categories of the obstacles to be avoided, and retrieve the recognition thresholds of the patrol robot for each of the characteristic categories.

[0078] If the first characteristic factor corresponding to each of the characteristic categories is within the recognition threshold of the corresponding characteristic category, it is determined that the patrol robot can recognize all the characteristic categories of the obstacles to be avoided.

[0079] Specifically, for the color feature category of the obstacle to be avoided, the color information of the obstacle to be avoided in the local image is extracted through image analysis technology. For example, a color space model (such as RGB, HSV, etc.) can be used to represent the color. The extracted color information is converted into specific color values, which are used as the first feature factor of the color feature category. For the shape feature category, a shape recognition algorithm is used to determine the shape of the obstacle to be avoided, such as circular, square, irregular shape, etc., and the shape information is converted into a corresponding numerical representation of the shape, which is used as the first feature factor of the shape feature category. For the size feature category, the pixel size of the obstacle to be avoided in the image is measured, and combined with the known scale information, the actual size is calculated, and the size value is used as the first feature factor of the size feature category.

[0080] Specifically, during the design and training process of the inspection robot, the recognition thresholds for different feature categories are determined. These recognition thresholds are ranges set according to the performance of the robot and the actual application scenarios. For the color feature category, the recognition threshold can be a specific color range. For example, in the RGB color space, the recognition threshold for a certain color feature can be a numerical range such as (R: 100 - 200, G: 50 - 150, B: 0 - 100).

[0081] For the shape feature category, the recognition threshold can be a series of allowable shape parameter ranges, such as the radius range of a circle, the side length range of a square, etc. For the size feature category, the recognition threshold can be the upper and lower limit ranges of dimensions such as length, width, and height.

[0082] If the inspection robot can recognize all the feature categories of the obstacle to be avoided, then the automatic obstacle avoidance database of the inspection robot is retrieved and traversed; the automatic obstacle avoidance database includes multiple preset obstacle types, and a preset feature set corresponding to each preset obstacle type, and the preset feature set includes multiple preset feature categories;

[0083] Specifically, the preset obstacle types are the obstacle types that the inspection robot can autonomously recognize, obtained according to the automatic obstacle avoidance system carried by the inspection robot. Exemplarily, the preset obstacle types can be classified according to the shape of the obstacle into box - type (such as a lost toolbox), column - type (such as a telegraph pole), and spherical - type (such as the spherical protective cover on a fallen power device).

[0084] Specifically, if the first feature factor of the color feature category is within the color recognition threshold range, the first feature factor of the shape feature category is within the shape recognition threshold range, and the first feature factor of the size feature category is within the size recognition threshold range, it is determined that the inspection robot can recognize all the feature categories of the obstacle to be avoided.

[0085] Exemplarily, it is assumed that the inspection robot is inspecting an inspection site. A red rectangular box, which is an obstacle to be avoided, appears in the local image. The color information of the box extracted through image analysis is RGB (200, 50, 30), which is used as the first feature factor of the color feature category.

[0086] The recognition threshold of the inspection robot for color features is in the range of the red color system, such as RGB (150 - 255, 0 - 100, 0 - 100). The shape recognition algorithm determines that the box is in the shape of a rectangular cuboid, and the parameter representation of the rectangular cuboid shape is used as the first feature factor of the shape feature category. The recognition threshold of the inspection robot for shape features includes the parameter range of the rectangular cuboid shape, such as the proportional relationship of length, width, and height. The pixel size of the box in the image is measured, and combined with the scale information of the inspection site, the actual size of the box is calculated to be 80 cm in length, 60 cm in width, and 40 cm in height, and these size information are used as the first feature factor of the size feature category. The recognition threshold of the inspection robot for size features is in the range of 50 - 100 cm in length, 40 - 80 cm in width, and 30 - 60 cm in height. Since the color feature factor of the box is within the color recognition threshold, the shape feature factor is within the shape recognition threshold, and the size feature factor is within the size recognition threshold, it is determined that the inspection robot can recognize all the feature categories of this obstacle to be avoided.

[0087] If a preset obstacle type corresponding to all the feature categories of the obstacle to be avoided is matched, it is determined that the inspection robot can determine the type of the obstacle to be avoided.

[0088] In a preferred embodiment, after determining whether the inspection robot can recognize all the feature categories of the obstacle to be avoided, the following steps are further included:

[0089] If the inspection robot cannot recognize at least one of the feature categories of the obstacle to be avoided, or, the inspection robot can recognize all the feature categories of the obstacle to be avoided and no preset obstacle type corresponding to all the feature categories of the obstacle to be avoided is matched, it is determined that the inspection robot cannot determine the type of the obstacle to be avoided.

[0090] Specifically, once a preset obstacle type is matched in the automatic obstacle avoidance database, the inspection robot can adopt corresponding obstacle avoidance strategies according to the obstacle type, indicating that subsequent remote obstacle avoidance system assistance is not required. In addition, since the automatic obstacle avoidance database of the inspection robot is constructed based on preset obstacle types and their corresponding feature sets. The obstacle avoidance strategies of the robot are usually associated with these known preset obstacle types. Only when a specific preset obstacle type is matched can the corresponding pre-trained or preset obstacle avoidance strategies be retrieved. If no preset obstacle type matching the feature category of the obstacle to be avoided is found after traversing the entire automatic obstacle avoidance database, it is determined that the inspection robot cannot determine the obstacle type to be avoided, indicating that the inspection robot cannot accurately obtain a suitable obstacle avoidance strategy from the automatic obstacle avoidance database, and subsequent remote obstacle avoidance system assistance is required to ensure the continuity of the inspection process.

[0091] Specifically, the hardware resources and energy supply of the inspection robot itself are usually limited. It needs to integrate various functional components in a compact body, such as a power system, a sensor system, a communication system, etc. Loading the complex remote obstacle avoidance system algorithm into the robot's own obstacle avoidance system will exceed the robot's own processor and memory, and running complex algorithms will consume more energy, which will shorten the robot's battery life. For robots that need to perform long-term and large-scale inspections, the effective utilization of energy is crucial. If additional computing tasks are loaded onto the robot, it may lead to frequent charging and affect the inspection efficiency. Based on this, this application generates obstacle avoidance strategies based on the remote obstacle avoidance system through known information. For a large number of robots distributed in different inspection sites, the remote obstacle avoidance system can also centrally manage and update the algorithms to ensure that all robots can obtain the latest obstacle avoidance strategy generation technology in a timely manner.

[0092] This embodiment takes into account that predicting that the inspection robot can determine the obstacle type to be avoided requires both steps to succeed. The first step is that the inspection robot can completely identify all the first feature factors of the feature categories of the obstacle to be avoided, that is, it has the ability to identify the characteristics of the obstacle. The second step is that there is a preset obstacle in the automatic obstacle avoidance database stored in the automatic obstacle avoidance system of the inspection robot that matches all the feature categories of the obstacle to be avoided, that is, the database is rich.

[0093] In a preferred embodiment, before obtaining the first feature factors corresponding to the respective feature categories of the obstacle to be avoided, the following steps are further included:

[0094] Obtain the light intensity of the surrounding environment where the obstacle to be avoided is located;

[0095] The obtaining of the first feature factors corresponding to the respective feature categories of the obstacle to be avoided specifically includes the following steps:

[0096] When the light intensity is less than the strong light threshold and greater than the dim threshold, obtain the first feature factors corresponding to the respective feature categories of the obstacle to be avoided.

[0097] Optionally, a light sensor that can rotate 360 degrees and can be telescoped up and down is mounted at the top of the inspection robot (a better position for the field of view). When the remote obstacle avoidance system determines that there is an obstacle to be avoided in the local image, it sends a signal to the inspection robot. This signal includes the direction of the obstacle to be avoided relative to the current position of the inspection robot. The inspection robot can control the direction of the light sensor to face the obstacle to be avoided, so as to collect the light intensity at the position of the obstacle to be avoided. The remote obstacle avoidance system can directly obtain the light intensity by reading the data of the light sensor. If there is no light sensor, the light intensity can be estimated by analyzing the features such as the brightness and contrast of the local image. For example, calculate the average brightness value of the pixels in the image and convert it into a relative value of the light intensity according to a certain standard.

[0098] Specifically, the strong light threshold and the dim threshold can be set according to the performance of the inspection robot and the actual application scenario. The strong light threshold is a relatively high light intensity value, representing that the ambient light is very bright, which may affect the recognition of obstacle features. For example, by testing the inspection robot in different strong light environments, a light intensity value can be determined as the strong light threshold. When the light intensity exceeds this value, it is considered a strong light environment. The dim threshold can be a relatively low light intensity value, representing that the ambient light is very dark, which may also affect the recognition of obstacle features. Similarly, the dim threshold can be determined through testing. When the light intensity is lower than this value, it is considered a dim environment.

[0099] Specifically, compare the obtained light intensity with the strong light threshold and the dim threshold. If the light intensity is within this range, it means that the ambient light is neither too strong nor too dim, and is in a state suitable for the inspection robot to recognize obstacle features. On the contrary, if the light is too strong or too dim (the light intensity is less than the strong light threshold or greater than the dim threshold), it is directly determined that the inspection robot cannot recognize at least one feature category of the obstacle to be avoided.

[0100] S4. If it is predicted that the inspection robot cannot determine the type of the obstacle to be avoided, then obtain a dynamic obstacle avoidance strategy according to the obstacle to be avoided and the local image;

[0101] Specifically, the strategy generation module for the remote obstacle avoidance system in this application is implemented through existing technologies. Here is an optional method: Analyze in detail the position, size, shape, etc. of the obstacle to be avoided in the local image. Image recognition algorithms can be used to determine the contour and approximate size of the obstacle, and its relative position can be determined by comparing with the surrounding environment. Then, based on multiple consecutive images, judge the mobility of the obstacle and analyze the surrounding environment in the local image, including terrain, other possible obstacles, and passable areas. Determine whether there is enough space nearby for the inspection robot to detour or adjust its traveling direction. If the obstacle is small and there is enough space around, a detour strategy can be generated. Determine a suitable detour direction and distance to ensure that the inspection robot can safely avoid the obstacle. For example, calculate a path that bypasses the obstacle at the minimum distance based on the position and shape of the obstacle. If the obstacle is large or the surrounding space is limited, a pause and wait strategy can be generated. Wait for a period of time to observe whether the obstacle will move or whether other feasible paths will appear. At the same time, continuously monitor the local image to adjust the strategy in a timely manner. If the nature of the obstacle is uncertain, a tentative approach strategy can be adopted. Slowly approach the obstacle while closely observing the reaction of the obstacle and the changes in the surrounding environment to adjust the traveling direction or speed in a timely manner.

[0102] S5. Control the inspection robot to avoid obstacles with the dynamic obstacle avoidance strategy when it travels to the obstacle to be avoided, and continue to travel along the preset inspection path after avoiding the obstacle to be avoided.

[0103] Specifically, when the inspection robot successfully avoids the obstacle to be avoided, the remote obstacle avoidance system determines the current position and the next traveling direction of the robot according to the information of the preset inspection path, and guides the inspection robot to gradually return to the preset inspection path to continue the inspection task. The traveling direction and speed of the robot can be adjusted by sending navigation instructions to make it move towards the next target point on the preset inspection path.

[0104] In a preferred embodiment, the automatic obstacle avoidance database further includes a preset obstacle avoidance strategy corresponding to each preset obstacle type; after obtaining the dynamic obstacle avoidance strategy according to the obstacle to be avoided and the local image, the following steps are further included:

[0105] Update the automatic obstacle avoidance database by taking the type of the obstacle to be avoided as a preset obstacle type, taking all the feature categories of the obstacle to be avoided as the preset feature set corresponding to the preset obstacle type, and taking the dynamic obstacle avoidance strategy of the type of the obstacle to be avoided as the preset obstacle avoidance strategy corresponding to the preset obstacle type.

[0106] Specifically, as new types of obstacles, their characteristics, and obstacle avoidance strategies are added to the automatic obstacle avoidance database, when the inspection robot encounters similar obstacles in subsequent inspection tasks, and on the premise that the inspection robot can identify all feature categories of the obstacles to be avoided, the inspection robot can match the corresponding preset obstacle avoidance strategy and no longer requires the assistance of the remote obstacle avoidance system. This greatly improves the adaptability of the robot to different inspection environments and reduces stagnation or incorrect decisions caused by unknown obstacles.

[0107] At the same time, after encountering a new obstacle to be avoided and generating a dynamic obstacle avoidance strategy each time, the database is updated, enabling the inspection robot to continuously learn and accumulate experience. Over time, the database will become increasingly rich, and the obstacle avoidance ability of the robot will also be continuously optimized, improving the efficiency and reliability of the inspection.

[0108] In this article, specific examples are used to elaborate on the principles and implementation methods of this application. The descriptions of the above embodiments are only used to help understand the method and its core idea of this application. The above are only the preferred implementation methods of this application. It should be noted that due to the limitations of literal expression and the objectively infinite specific structures, for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements, refinements, or changes can be made, or the above technical features can be combined in an appropriate manner; these improvements, refinements, changes, or combinations, or directly applying the concept and technical solution of the invention to other occasions without improvement, should all be regarded as the protection scope of this application.

Claims

1. A control method for remote obstacle avoidance of a power inspection robot, characterized in that, Including the following steps: After receiving the start inspection signal of the inspection robot, obtain road images in real time. The road images are images composed of all inspection roads in the inspection site where the inspection robot can travel. Determine the target section and obtain the local image corresponding to the target section. The target section is a continuous section of road in the road image that the inspection robot is about to travel along the preset inspection path. If there are obstacles to be avoided in the local image, then predict whether the inspection robot can determine the type of the obstacle to be avoided. If it is predicted that the inspection robot cannot determine the type of the obstacle to be avoided, then obtain a dynamic obstacle avoidance strategy based on the obstacle to be avoided and the local image. Control the inspection robot to avoid obstacles with the dynamic obstacle avoidance strategy when it travels to the obstacle to be avoided, and continue to travel along the preset inspection path after avoiding the obstacle to be avoided. After determining the target section and obtaining the local image corresponding to the target section, the following steps are further included: Judge whether there are target section obstacles in the target section. If there are target section obstacles, then judge the attribute category of the target section obstacles. The attribute category includes general obstacles and special obstacles. If the attribute category of the target section obstacle is the special obstacle, then it is determined that there is an obstacle to be avoided in the local image. After receiving the start inspection signal of the inspection robot, the following steps are further included: Judge whether the inspection robot is inspecting the inspection site for the first time. The judging of the attribute category of the target section obstacle specifically includes the following steps: If the inspection robot is inspecting the inspection site for the first time, then retrieve the first inspection log of the inspection site. The first inspection log records the first inspection information of other robots inspecting the inspection site. The first inspection information at least includes the first obstacle coordinates corresponding to the previous obstacles encountered by other robots during the inspection, and the obstacle avoidance results of other robots for the previous obstacles. The obstacle avoidance results include successful obstacle avoidance and failed obstacle avoidance. If the first obstacle coordinates identical to the coordinates of the target section obstacle are matched in the first inspection log and the obstacle avoidance success rate of other robots for the previous obstacle matching the target section obstacle is greater than the first preset threshold, then the attribute category of the target section obstacle is determined to be the general obstacle.

2. The control method for remote obstacle avoidance of the power inspection robot according to claim 1, characterized in that: The determination of the target section specifically includes the following steps: Set a first preset duration according to the initial traveling speed of the inspection robot and the data processing period of the remote obstacle avoidance system. The data processing period is the required duration from the time when the remote obstacle avoidance system receives the road image until the dynamic obstacle avoidance strategy is generated. Taking the current coordinate of the inspection robot as the starting point, determine the road that will be traveled at the initial traveling speed within the next first preset duration as the target section.

3. The control method for remote obstacle avoidance of the power inspection robot according to claim 1, characterized in that: The prediction of whether the inspection robot can determine the type of the obstacle to be avoided specifically includes the following steps: Determine whether the inspection robot can recognize all the feature categories of the obstacle to be avoided, where the feature categories at least include color features, shape features, and size features; If the inspection robot can recognize all the feature categories of the obstacle to be avoided, then retrieve and traverse the automatic obstacle avoidance database of the inspection robot; the automatic obstacle avoidance database includes multiple preset obstacle types, and a preset feature set corresponding to each preset obstacle type, and the preset feature set includes multiple preset feature categories; If a preset obstacle type corresponding to all the feature categories of the obstacle to be avoided is matched, it is determined that the inspection robot can determine the type of the obstacle to be avoided.

4. The control method for remote obstacle avoidance of the power inspection robot according to claim 3, characterized in that: After determining whether the inspection robot can recognize all the feature categories of the obstacle to be avoided, the following steps are further included: If the inspection robot cannot recognize at least one of the feature categories of the obstacle to be avoided, or, the inspection robot can recognize all the feature categories of the obstacle to be avoided and no preset obstacle type corresponding to all the feature categories of the obstacle to be avoided is matched, it is determined that the inspection robot cannot determine the type of the obstacle to be avoided.

5. The control method for remote obstacle avoidance of the power inspection robot according to claim 3, wherein: Determining whether the inspection robot can recognize all the feature categories of the obstacle to be avoided specifically includes the following steps: Obtain the first feature factor corresponding to each feature category of the obstacle to be avoided, and retrieve the recognition threshold of each feature category by the inspection robot; If the first feature factor corresponding to each feature category is within the recognition threshold of the feature category, it is determined that the inspection robot can recognize all the feature categories of the obstacle to be avoided.

6. The control method for remote obstacle avoidance of the power inspection robot according to claim 1, characterized in that: Judging the attribute category of the obstacle on the target section further includes the following steps: If the inspection robot is not inspecting the inspection site for the first time, retrieve the second inspection log of the inspection robot, and the second inspection log records the second inspection information of the inspection robot inspecting the inspection site; the second inspection information at least includes the second obstacle coordinates corresponding to the previous obstacles encountered by the inspection robot during the inspection, and the obstacle avoidance result of the inspection robot for the previous obstacles; If the second obstacle coordinates identical to the coordinates of the obstacle on the target section are matched in the second inspection log and the obstacle avoidance success rate of the inspection robot for the previous obstacle matching the obstacle on the target section is greater than the first preset threshold, the attribute category of the obstacle on the target section is determined to be the general obstacle.

7. The control method for remote obstacle avoidance of the power inspection robot according to claim 5, characterized in that: Before obtaining the first feature factor corresponding to each feature category of the obstacle to be avoided, the following steps are further included: Obtain the light intensity of the surrounding environment where the obstacle to be avoided is located; Obtaining the first feature factor corresponding to each feature category of the obstacle to be avoided specifically includes the following steps: If the light intensity is less than the strong light threshold and greater than the dim threshold, obtain the first feature factor corresponding to each feature category of the obstacle to be avoided.

8. The control method for remote obstacle avoidance of the power inspection robot according to claim 3, characterized in that: The automatic obstacle avoidance database further includes a preset obstacle avoidance strategy corresponding to each of the preset obstacle types; after obtaining the dynamic obstacle avoidance strategy based on the obstacle to be avoided and the local image, the following steps are further included: Taking the type of the obstacle to be avoided as a preset obstacle type, taking all the feature categories of the obstacle to be avoided as a preset feature set corresponding to the preset obstacle type, and taking the dynamic obstacle avoidance strategy of the type of the obstacle to be avoided as the preset obstacle avoidance strategy corresponding to the preset obstacle type, to update the automatic obstacle avoidance database.

Images