Split type robot and robot for common box bus inspection and application thereof

Abstract

The application belongs to the technical field of inspection robots, and discloses a split robot and a robot for common-box bus inspection and application thereof. The split robot comprises a body and four leg structures arranged on the body. The body comprises a front body, a rear body and an intermediate connecting joint. The intermediate connecting joint comprises a first connecting rod, a second connecting rod and a central hinge pin. The rear end of the front body is in sliding connection with the front end of the first connecting rod, the rear end of the first connecting rod is in rotary connection with the front end of the second connecting rod, and the rear end of the second connecting rod is in sliding connection with the front end of the rear body. After the split robot and the common-box bus satisfy a specific relationship in size, the split robot becomes the robot for common-box bus inspection. The application of the robot is to control the robot to travel in the common-box bus for inspection. The split robot can reliably realize a right-angle turn in a narrow common-box bus without sacrificing the structural function, and effectively solves the problem of travel failure or operation interruption caused by a large turning radius in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of inspection robot technology, and relates to a split-type robot and a robot for inspection of common busbars and their applications. Background Technology

[0002] With the development of industrial robots, and to adapt to different application scenarios, existing robots are mainly wheeled / tracked robots and quadrupedal bionic robots. However, due to limitations in structural design and motion mechanisms, these two types of robots generally suffer from excessively large turning radii, as specifically manifested below:

[0003] Wheeled / tracked robots employ a rigid, integral chassis structure, with a minimum turning radius typically no less than 500mm, and some larger models even exceeding 800mm. In narrow, winding, enclosed environments, these robots cannot turn effectively and are prone to physical jamming.

[0004] While quadrupedal bionic robots possess the ability to rotate in place and can adjust their orientation with "zero radius" while stationary, they still need to maintain their overall body shape to navigate curves when making dynamic turns while traveling through narrow passages. The effective turning radius is generally between 400-800mm. Therefore, in narrow, winding, and enclosed environments, these robots cannot enter or pass through right-angle bends.

[0005] Therefore, it is necessary to develop a robot that can turn effectively in narrow, winding, enclosed environments.

[0006] Common busbars are typically narrow, winding, and enclosed environments. Existing technologies using robots for common busbar inspections encounter several challenges, including the inability to effectively navigate turns and the inability to perform blind-spot-free inspection of cylindrical insulators. Current inspection robots generally employ a "single camera + pan-tilt-zoom (PTZ)" scanning mode. For cylindrical insulators, a single viewpoint can only cover approximately one-third of the surface at most, requiring complex mechanical movements to compensate for the limited field of view. For example, patent application CN210270618U uses a "rotating device" and a "pan-tilt-zoom" with a "visible light camera" for inspection, while patent CN103640020B uses a hook arm to move a single-point sensor along the insulator surface. However, in both of these approaches, during continuous robot movement, the single camera cannot simultaneously cover the back and sides of the insulator, easily leading to missed cracks.

[0007] Therefore, it is necessary to develop a new type of busbar inspection robot that can simultaneously or at least partially solve the two major problems mentioned above. Summary of the Invention

[0008] One of the objectives of this invention is to provide a robot that can turn effectively in narrow, winding, enclosed environments.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0010] A split-type robot is a quadruped robot with straight-line and turning functions, including a body with four leg structures. The body includes a front body, a rear body and a middle connecting joint.

[0011] The intermediate connecting joint includes a first connecting rod, a second connecting rod, and a central hinge pin;

[0012] The rear end of the front body is slidably connected to the front end of the first connecting rod, the rear end of the first connecting rod is rotatably connected to the front end of the second connecting rod with the axis of rotation parallel to the vertical direction, and the rear end of the second connecting rod is slidably connected to the front end of the rear body. The sliding direction of all sliding connections is parallel to the width direction of the body.

[0013] The modular robot of this invention comprises multiple sub-modules connected by hinges or sliding joints, and its posture and position can be independently adjusted under a control system. Thanks to this design, during turning, each sub-module can pass through the corner area sequentially without needing to move synchronously around the same center, thus significantly reducing the radius of the smallest circle its body can draw during turning. Actual measurements have verified that the actual turning radius of the robot described in this invention can be controlled to ≤400mm, far smaller than the typical value of existing technologies (400-800mm). Therefore, this invention can reliably achieve right-angle turns in narrow spaces without sacrificing structural functionality, effectively solving the problem of passage failure or work interruption caused by excessively large turning radii in existing technologies.

[0014] As a preferred technical solution:

[0015] As described above, in a split-type robot, the rear end of the front body is slidably connected to the front end of the first connecting rod via a pair of sliding grooves and a slider. The sliding grooves are located at the rear end of the front body, and the sliders are fixed to the front end of the first connecting rod.

[0016] The rear end of the second connecting rod is slidably connected to the front end of the rear body through a pair of sliding grooves and sliders. The sliding grooves are opened at the front end of the rear body, and the sliders are fixed to the rear end of the second connecting rod.

[0017] The rear end of the first connecting rod and the front end of the second connecting rod are rotatably connected by a central hinge pin that runs vertically through both of them.

[0018] As described above, the leg structure of a split-type robot includes a hip joint connecting frame, a hip abduction and adduction motor, a transition motor, a hip flexion and extension motor, a thigh link, a knee flexion and extension motor, a lower leg link, a hub motor, and wheels.

[0019] The hip joint connecting frame is fixedly connected to the side of the front or rear body;

[0020] The housing of the hip abduction and adduction motor is fixedly connected to the hip joint connecting frame. The output shaft of the hip abduction and adduction motor is parallel to the length direction of the machine body and is fixedly connected to the housing of the transition motor. The output shaft of the transition motor is parallel to the width direction of the machine body and is fixedly connected to the housing of the hip flexion and extension motor. The output shaft of the hip flexion and extension motor is parallel to the width direction of the machine body and is fixedly connected to the upper end of the thigh connecting rod.

[0021] The outer casing of the knee flexion and extension motor is fixedly connected to the lower end of the thigh linkage, and the output shaft of the knee flexion and extension motor is parallel to the width direction of the machine body and fixedly connected to the upper end of the lower leg linkage.

[0022] The stator (housing) of the hub motor is fixedly connected to the lower end of the lower leg connecting rod, and the rotor of the hub motor is parallel to the width direction of the body and fixedly connected to the wheel (usually an omnidirectional wheel or a Mecanum wheel).

[0023] As described above, a split-type robot has a forward-facing RGB camera installed at the very front of the front body and a lidar installed at the bottom.

[0024] The second objective of this invention is to provide a robot that can effectively turn within a common busbar.

[0025] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0026] A robot A for inspecting common busbars is a split-type robot as described above that satisfies a specific relationship with the dimensions of the common busbars.

[0027] The specific relationship is: the split-type robot reaches its maximum deflection angle at the intermediate connecting joint. Dynamic bounding rectangle width The internal clear width of the common busbar is smaller than that of the common busbar. Furthermore, the total height of the split-type robot at the preset cruising altitude The internal clear height of the common busbar is less than ;

[0028] The purpose of controlling the dimensions of the split robot and the common busbar to meet a specific relationship is to ensure that the split robot can move within the common busbar and perform straight or turning operations.

[0029] The third objective of this invention is to provide a robot that can effectively turn in a common busbar and perform blind-spot-free detection of columnar insulators.

[0030] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0031] A robot B for inspecting common busbars is a split-type robot as described above that satisfies a specific relationship with the dimensions of the common busbars.

[0032] The specific relationship is: the split-type robot reaches its maximum deflection angle at the intermediate connecting joint. Dynamic bounding rectangle width The internal clear width of the common busbar is smaller than that of the common busbar. Furthermore, the total height of the split-type robot at the preset cruising altitude The internal clear height of the common busbar is less than ;

[0033] The number of insulators in the same row of the common busbar is n, where n>1;

[0034] The bottom of the split-type robot is equipped with n insulator vision inspection systems, which are used to capture images of the entire circumferential surface of n insulators in the same row of the common busbar.

[0035] As a preferred technical solution:

[0036] As described above, a robot B for inspecting common busbars includes a left front micro-shell, a right front micro-shell, a left rear micro-shell, and a right rear micro-shell for each insulator visual inspection system.

[0037] The left front micro-shell bends towards the right front micro-shell, and the left rear micro-shell bends towards the right rear micro-shell;

[0038] The left front micro-shell, right front micro-shell, left rear micro-shell, and right rear micro-shell of each insulator visual inspection system form a ring-shaped inspection area;

[0039] Each miniature housing integrates a miniature industrial camera and a miniature LED fill light.

[0040] To demonstrate that the insulator visual inspection system described in this invention can cover the insulator surface, a mathematical model of the camera's field of view (FOV) coverage is established using 12 cameras as an example for verification. The steps are as follows:

[0041] Assume the insulator to be tested is a standard cylinder with a radius of... The radius of the annular inspection area formed by the four miniature shells of a single insulator visual inspection system is [missing information]. (R > r). The camera array consists of 12 cameras evenly distributed along the circumference, with the optical axis of each camera pointing towards the center of the circle.

[0042] (1) Circumferential Coverage Condition: To achieve seamless coverage along the circumference, the fields of view of two adjacent cameras must overlap on the surface of the insulator. Let the horizontal field of view of a single camera be... The angle between the optical axes of adjacent cameras is According to geometric relationships, the included angle Determined by the number of cameras:

[0043]

[0044] when hour, .

[0045] This leads to the mandatory requirements for camera selection in this invention: the horizontal field of view of a single camera. It must be strictly greater than 30° (i.e.) In practical applications, this invention selects a field of view angle. Industrial lenses, thus creating an overlap angle between adjacent cameras. :

[0046]

[0047] Mechanism conclusion: The above calculations prove that, relying on the fixed 15° redundant overlap area of ​​the physical structure, the detection blind zone in the circumferential direction of the insulator is completely eliminated from a geometrical perspective.

[0048] (2) Axial coverage and depth of field calculation: The insulator surface is not a smooth plane and has a skirt structure. Let the depth of field range of the camera be... The working distance L between the camera lens and the insulator surface is calculated as follows:

[0049]

[0050] in Allowance for mounting the miniature housing protruding from the front of the camera lens.

[0051] Mechanism conclusion: This invention ensures that the working distance L always falls within the camera's depth of field range by adjusting the diameter R of the annular detection area. This ensures that both the insulator's main body and the edges of its skirts can be clearly imaged.

[0052] Furthermore, while traditional robots require a "stop-take photo-go" process, this invention achieves "taking photos while moving," the core of which lies in solving the motion blur problem. The steps are as follows:

[0053] (1) Motion fuzz suppression model: When the robot moves at a speed of When moving linearly along the pipe axis, if the exposure time is The distance of the motion blur produced by the imaging for:

[0054]

[0055] in, To improve the speed of robot inspection, The effective exposure time of the camera. This refers to the distance of the image trail.

[0056] To ensure image sharpness, the motion blur distance must be less than the physical size of one pixel in the imaging system. (That is, the blur is within one pixel and imperceptible to the human eye). In other words, the following conditions must be met:

[0057]

[0058] Among them, pixel physical size It is determined by the camera resolution and magnification.

[0059] Mechanism Implementation: This invention utilizes high-frequency flash technology integrating a miniature LED fill light (273) to extend the effective exposure time. Compressed to the microsecond level (e.g.) Substitute typical inspection speed. The resulting motion blur is only It is much smaller than the regular pixel size, thus "freezing" the moment of motion at the physical level.

[0060] (2) Position synchronization triggering logic: The system uses LiDAR (260) to build a local map of the pipeline in real time and calculates the distance of the robot relative to the center plane of the insulator in real time. The trigger logic is set as follows: when... At the instant the plane of the ring detection area coincides with the central cross-section of the insulator, the central controller simultaneously sends a hard trigger signal to all 12 cameras. This mechanism ensures that the image is captured from the optimal viewing position, avoiding positional deviations caused by communication delays.

[0061] To ensure sufficient image fidelity and accuracy, taking n=3 as an example, this invention first physically acquires 12 overlapping local images (because the four cameras in each insulator vision inspection system adopt a layout of facing curves and are all equipped with large field-of-view industrial lenses, the fields of view of the four cameras corresponding to a single insulator will form a redundant overlap area of ​​15° on the circumferential surface of the insulator during shooting. Therefore, each local image of the insulator captured by the miniature industrial camera has the feature of field-of-view overlap. All 12 images from the three sets of insulator vision inspection systems have this overlap attribute, which are the 12 overlapping local images described in this invention). Then, an algorithm is used to "stitch" them together. "To obtain a complete unfolded image of an insulator (the specific process is as follows: during actual data acquisition, the center plane positions of three insulators are first precisely located in real time using LiDAR. The moment the robot travels to the annular detection area and its plane completely overlaps with the center cross-section of the three insulators, the central controller simultaneously sends hard-trigger shooting signals to these 12 miniature industrial cameras. Combined with the high-frequency flash microsecond-level exposure technology of the miniature LED supplementary lights, synchronous shooting without motion blur is achieved, thus acquiring these 12 overlapping partial images of the insulator. Subsequently, the four overlapping partial images corresponding to each individual insulator are stitched together to form a complete 360° circumferential unfolded image of that insulator). This process is not a simple image stitching, but rather a reverse cylindrical projection based on a precise geometric model."

[0062] (1) Coordinate Mapping Model: To eliminate perspective distortion caused by camera tilting, this invention establishes a mapping relationship from a "two-dimensional image plane" to a "three-dimensional cylindrical surface." Let the coordinates of any pixel on the insulator unfolded diagram be... Its corresponding three-dimensional cylindrical coordinates for:

[0063]

[0064] in This represents the total width of the unfolded diagram.

[0065] Based on the pre-calibrated camera intrinsic parameter matrix K and the first... Extrinsic parameter matrix of a camera (Guaranteed by the machining accuracy of the annular detection area), this spatial point is projected onto the first... Pixel coordinates on the camera's imaging plane Satisfy the following projection equations:

[0066]

[0067] Mechanism conclusion: As long as the insulator radius is known... and mechanical structure parameters For each point on the unfolded image, its position on the original image can be accurately found using the above formula.

[0068] (2) Weighted fusion algorithm: In the overlapping area of ​​the two cameras (i.e., within the aforementioned 15° overlap angle), the pixel values ​​are not simply averaged, but rather a fade-in / fade-out weighted fusion is used:

[0069]

[0070] Among them, weight The distance from the point to the image boundary is proportional. This eliminates stitching gaps caused by uneven lighting or minor mechanical vibrations.

[0071] In summary, this invention employs a geometrically overlapping design (ensuring no blind spots in space), a temporally short-exposure synchronization (ensuring no blurring in dynamics), and a mathematically inverse projection (ensuring distortion-free reconstruction). These three interconnected elements constitute a complete omnidirectional dynamic detection mechanism for insulators. This mechanism, combined with the aforementioned "robot movement mechanism," enables efficient and fully automated inspection in confined spaces. To demonstrate the effectiveness of the 12-camera array described in this invention (such as...), Figure 4 (As shown) To cover the surface of the insulator, a mathematical model of the camera's field of view (FOV) coverage needs to be established.

[0072] As described above, a robot B for inspecting common busbars has a split-type robot body consisting of (n+1) uprights at the front.

[0073] (n+1) uprights are arranged at intervals along the width of the front body to form n gaps;

[0074] The visual inspection systems for n insulators are located in n gaps respectively;

[0075] In each insulator visual inspection system, the left front micro shell and the left rear micro shell are fixedly connected to the bottom of one pole, and the right front micro shell and the right rear micro shell are fixedly connected to the bottom of another pole.

[0076] Accordingly, the present invention also provides a method for inspecting a common busbar, controlling a robot B, as described above, for inspecting a common busbar to move within the common busbar, comprising the following steps:

[0077] (1) Environmental perception and map building;

[0078] The robot uses its onboard sensors (including a two-dimensional high-frequency LiDAR for scanning the internal cross-section of the common busbar and building an environmental map, and a forward-looking camera to assist in locating the insulator and monitoring the distance between the pipe wall) to acquire real-time three-dimensional environmental information inside the common busbar in order to build or update its navigation map.

[0079] (2) Global and local path planning;

[0080] Based on the constructed environmental map and inspection task objectives, the robot plans the optimal or feasible path from the starting point to the end point and handles local problems such as turning and obstacle avoidance in real time.

[0081] (3) Motor control and gait execution;

[0082] The key step in translating the planned path into actual limb movements is for the robot to coordinate its multiple legs to complete actions such as walking in a straight line, turning, and crossing with a stable gait.

[0083] (4) Task execution and data collection;

[0084] While moving along the planned path, the robot uses its onboard insulator vision inspection system to perform its tasks.

[0085] The fourth objective of this invention is to provide a robot that, although unable to effectively turn within a common busbar, can perform blind-spot-free detection of columnar insulators.

[0086] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0087] A robot C for inspecting common busbars is basically the same as robot B for inspecting common busbars, except that robot C has a rigid integrated body and does not use a split structure of front body, rear body and intermediate connecting joints. It retains the circumferential array of insulator vision inspection system (including miniature industrial cameras and supplementary lights), which can be used for the inspection of straight sections of common busbars, and can achieve 360° blind-spot-free inspection of columnar insulators through the array camera.

[0088] Beneficial effects:

[0089] (1) This invention solves the problem of limited movement of traditional robots in extremely narrow and winding environments such as common busbars by using a split wheel-foot composite mechanism and a passive adaptive design of intermediate connecting joints, and realizes flexible turning and stable movement in confined spaces.

[0090] (2) This invention uses multiple insulator visual inspection systems and adopts dynamic scanning imaging to complete the acquisition of 360° panoramic images of insulators in one go, fundamentally eliminating the blind spot problem of single-camera inspection and greatly improving inspection efficiency and coverage.

[0091] (3) This invention combines the robot’s mobility with an intelligent recognition system, and achieves automatic identification and location of defects through advanced machine vision algorithms, reducing human intervention and improving the objectivity and accuracy of the detection results.

[0092] (4) Through the synergistic innovation of mechanical structure and insulator visual inspection system, this invention has successfully solved the key technical problem in the inspection of common busbar insulators, and provides a safe, efficient and intelligent solution for power equipment maintenance. It has important industry promotion value and economic benefits. Attached Figure Description

[0093] Figure 1 This is an overall schematic diagram of robot B used for common busbar inspection according to the present invention;

[0094] Figure 2 This is a schematic diagram of the leg structure of the present invention;

[0095] Figure 3 This is a schematic diagram of the intermediate connecting joint of the present invention; in the figure, a is an overall schematic diagram of the intermediate connecting joint, b is a schematic diagram of the connection between the first connecting rod and the second connecting rod, and c is a schematic diagram of the connection between the second connecting rod and the rear body.

[0096] Figure 4 This is a partial schematic diagram of the front body of the present invention;

[0097] Figures 5-12 This is a schematic diagram of the action sequence of the robot B used for common busbar inspection in this invention when passing through a right-angle bend; in each figure, a is the action diagram of the robot B used for common busbar inspection when passing through a right-angle bend, and b is the form diagram of the robot B used for common busbar inspection when passing through a right-angle bend.

[0098] Figure 13 The robot B used for common busbar inspection in this invention achieves its maximum deflection angle at the intermediate connecting joint. Dynamic circumscribed rectangle width The internal net width of the common busbar A schematic diagram;

[0099] Figure 14 The total height of robot B, used for inspection of shared busbars according to the present invention, at a preset cruising altitude. Internal clearance of the common busbar A schematic diagram;

[0100] Figure 15 This is an overall schematic diagram of the robot C used for common busbar inspection according to the present invention; in the figure, a is a motion diagram of the robot C for common busbar inspection passing through a straight line segment, and b is a schematic diagram of the robot C for common busbar inspection.

[0101] In the diagram, 200 is the front body, 221 is the thigh link, 222 is the knee flexion / extension motor, 223 is the lower leg link, 231 is the hip joint connecting frame, 232 is the hip abduction / adduction motor, 233 is the hip flexion / extension motor, 234 is the transition motor, 240 is the hub motor, 250 is the forward-facing RGB camera, 260 is the LiDAR, 272 is the miniature industrial camera, 273 is the miniature LED fill light, 280 is the upright, 291 is the left front miniature shell, 292 is the right front miniature shell, 293 is the left rear miniature shell, 294 is the right rear miniature shell, 300 is the rear body, 400 is the intermediate connecting joint, 410 is the first connecting rod, 420 is the second connecting rod, and 430 is the central hinge pin. Detailed Implementation

[0102] The present invention will be further described below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0103] Example 1

[0104] A split-type robot, a quadruped robot with straight-line and turning functions, includes a body with four leg structures on the body;

[0105] The fuselage consists of a front fuselage, a rear fuselage, and a central connecting joint;

[0106] A forward-facing RGB camera is installed at the very front of the front unit, and a lidar is installed at the bottom;

[0107] The intermediate connecting joint includes a first connecting rod, a second connecting rod, and a central hinge pin;

[0108] The rear end of the front body is slidably connected to the front end of the first connecting rod through a pair of sliding grooves and a slider. The sliding grooves are opened at the rear end of the front body, and the sliders are fixed at the front end of the first connecting rod.

[0109] The rear end of the first connecting rod and the front end of the second connecting rod are rotatably connected by a central hinge pin that runs vertically through them, and the axis of rotation is parallel to the vertical direction.

[0110] The rear end of the second connecting rod is slidably connected to the front end of the rear body through a pair of sliding grooves and sliders. The sliding grooves are opened at the front end of the rear body, and the sliders are fixed to the rear end of the second connecting rod.

[0111] The sliding direction of all sliding connections is parallel to the width direction of the fuselage;

[0112] The leg structure includes a hip joint connecting frame, a hip abduction and adduction motor, a transition motor, a hip flexion and extension motor, a thigh link, a knee flexion and extension motor, a lower leg link, a hub motor, and a wheel;

[0113] The hip joint connecting frame is fixedly connected to the side of the front or rear body;

[0114] The housing of the hip abduction and adduction motor is fixedly connected to the hip joint connecting frame. The output shaft of the hip abduction and adduction motor is parallel to the length direction of the machine body and is fixedly connected to the housing of the transition motor. The output shaft of the transition motor is parallel to the width direction of the machine body and is fixedly connected to the housing of the hip flexion and extension motor. The output shaft of the hip flexion and extension motor is parallel to the width direction of the machine body and is fixedly connected to the upper end of the thigh connecting rod.

[0115] The outer casing of the knee flexion and extension motor is fixedly connected to the lower end of the thigh linkage, and the output shaft of the knee flexion and extension motor is parallel to the width direction of the machine body and fixedly connected to the upper end of the lower leg linkage.

[0116] The stator of the hub motor is fixedly connected to the lower end of the lower leg connecting rod, and the rotor of the hub motor is parallel to the width direction of the machine body and fixedly connected to the wheel.

[0117] Example 2

[0118] A robot B for inspecting common busbars, such as Figures 1-4 As shown, this is a split-type robot whose dimensions satisfy a specific relationship with the dimensions of the shared busbar;

[0119] The number of insulators in the same row of the common busbar is n, where n>1;

[0120] A split-type robot is a quadruped robot with straight-line and turning functions, including a body and four leg structures set on the body;

[0121] like Figure 1 , Figure 3 As shown, the fuselage includes a front fuselage 200, a rear fuselage 300, and an intermediate connecting joint 400;

[0122] The intermediate connecting joint 400 includes a first connecting rod 410, a second connecting rod 420, and a central hinge pin 430;

[0123] like Figure 1 , Figure 4 As shown, the front body 200 includes (n+1) uprights 280; the (n+1) uprights 280 are arranged at intervals along the width direction of the front body 200 to form n gaps;

[0124] n insulator visual inspection systems are located in n gaps respectively. The n insulator visual inspection systems are used to capture images of the entire circumferential surface of n insulators in the same row of the common busbar.

[0125] Each insulator visual inspection system includes a left front miniature housing 291, a right front miniature housing 292, a left rear miniature housing 293, and a right rear miniature housing 294; the left front miniature housing 291 and the right front miniature housing 292 are bent towards each other, and the left rear miniature housing 293 and the right rear miniature housing 294 are bent towards each other; each miniature housing integrates a miniature industrial camera 272 and a miniature LED supplementary light 273;

[0126] In each insulator visual inspection system, the left front miniature shell 291 and the left rear miniature shell 293 are fixedly connected to the bottom of a pole 280, and the right front miniature shell 292 and the right rear miniature shell 294 are fixedly connected to the bottom of another pole 280.

[0127] A front-view RGB camera 250 is installed at the very front of the front unit 200, and a lidar 260 is installed at the bottom;

[0128] The rear end of the front body 200 is slidably connected to the front end of the first connecting rod 410 through a pair of sliding grooves and a slider. The sliding grooves are opened at the rear end of the front body 200, and the sliders are fixed at the front end of the first connecting rod 410.

[0129] The rear end of the first connecting rod 410 and the front end of the second connecting rod 420 are rotatably connected by a central hinge pin 430 that passes vertically through them, and the axis of rotation is parallel to the vertical direction.

[0130] The rear end of the second connecting rod 420 is slidably connected to the front end of the rear body 300 through a pair of sliding grooves and sliders. The sliding grooves are opened at the front end of the rear body 300, and the sliders are fixed at the rear end of the second connecting rod 420. The sliding direction of all sliding connections is parallel to the width direction of the body.

[0131] like Figure 1 , Figure 2 As shown, the leg structure includes a hip joint connecting frame 231, a hip abduction and adduction motor 232, a transition motor 234, a hip flexion and extension motor 233, a thigh link 221, a knee flexion and extension motor 222, a lower leg link 223, a hub motor 240, and a wheel;

[0132] The hip joint connecting frame 231 is fixedly connected to the side of the front body 200 or the rear body 300;

[0133] The housing of the hip abduction and adduction motor 232 is fixedly connected to the hip joint connecting frame 231. The output shaft of the hip abduction and adduction motor 232 is parallel to the length direction of the machine body and is fixedly connected to the housing of the transition motor 234. The output shaft of the transition motor 234 is parallel to the width direction of the machine body and is fixedly connected to the housing of the hip flexion and extension motor 233. The output shaft of the hip flexion and extension motor 233 is parallel to the width direction of the machine body and is fixedly connected to the upper end of the thigh connecting rod 221.

[0134] The outer casing of the knee flexion and extension motor 222 is fixedly connected to the lower end of the thigh link 221, and the output shaft of the knee flexion and extension motor 222 is parallel to the width direction of the machine body and fixedly connected to the upper end of the lower leg link 223.

[0135] The stator of the hub motor 240 is fixedly connected to the lower end of the lower leg connecting rod 223, and the rotor of the hub motor 240 is parallel to the width direction of the machine body and fixedly connected to the wheel.

[0136] Specific relationships are: such as Figure 13 , Figure 14 As shown, the split-type robot reaches its maximum deflection angle at the intermediate connecting joint 400. Dynamic bounding rectangle width The internal clear width of the common busbar is smaller than that of the common busbar. Furthermore, the total height of the split-type robot at the preset cruising altitude The internal clear height of the common busbar is less than .

[0137] A method for inspecting a common busbar, comprising controlling a robot B, as described above, to move within the common busbar, includes the following steps:

[0138] (1) Environmental perception and map building;

[0139] The robot uses its onboard sensors to acquire real-time three-dimensional environmental information inside the common busbar in order to build or update its navigation map.

[0140] (2) Global and local path planning;

[0141] Based on the constructed environmental map and inspection task objectives, the robot plans the optimal or feasible path from the starting point to the end point and handles local problems such as turning and obstacle avoidance in real time.

[0142] (3) Motor control and gait execution;

[0143] The key step in translating the planned path into actual limb movements is for the robot to coordinate its multiple legs to complete actions such as walking in a straight line, turning, and crossing with a stable gait.

[0144] (4) Task execution and data collection;

[0145] While moving along the planned path, the robot uses its onboard insulator vision inspection system to perform its tasks.

[0146] When using the above-mentioned inspection method for the common busbar, after the robot B used for the common busbar inspection is powered on and started, the two-dimensional high-frequency laser radar located at the bottom immediately starts working at a scanning frequency of 10Hz. It performs real-time scanning of the environmental cross-section inside the common busbar, which is 500mm high and 1000mm wide, and constructs a two-dimensional occupancy grid map with a resolution of centimeters to identify the environmental contours and obstacle features. At the same time, in order to distinguish different leg structures, the four leg structures are respectively labeled as the left leg structure of the front body, the right leg structure of the front body, the left leg structure of the rear body, and the right leg structure of the rear body.

[0147] When the lidar point cloud data shows a vertical step with a height abruptly changing to 50mm at a distance of 200mm ahead, the control algorithm identifies it as the starting point of an obstacle platform measuring 880mm in length and 105mm in width. It then triggers a deceleration command and stops moving forward 100mm from the edge of the step. Simultaneously, it switches the left and right leg structures of the front fuselage to foot-based mode (when obstacle crossing is required, the control system instructs the knee flexion / extension motor of a specific leg to actuate, raising the lower leg and foot wheel to simulate walking to cross the obstacle; in this mode, the hub motor can be braked). (Locking, turning the wheels into fixed fulcrums) Preparing for climbing; At the start of the climbing action, the right leg structure of the front unit remains locked and supported, while the hip flexion-extension motor of the left leg structure of the front unit swings forward 18° while the knee flexion-extension motor bends inward 28° (the specific angles are calculated by real-time inverse kinematics). This combined action raises the bottom surface of the 100mm diameter foot wheel from the ground to a height of 72mm (i.e., crossing a 50mm step with a 22mm safety margin), then extends forward and falls to the surface of a 50mm high obstacle, landing on the ground. The front unit then... The right leg structure repeats the same leg-lifting and stepping motion; after the left and right leg structures of the front unit have landed steadily, the hub motors of the left and right leg structures of the rear unit drive the rear unit to crawl forward until the left and right leg structures of the rear unit approach the step. The left and right leg structures of the rear unit then execute the same foot-climbing procedure to lift the entire body onto the obstacle platform; after the system confirms that all four legs are on the obstacle platform through joint torque feedback, it switches back to wheel mode (i.e., when the hub motors of all legs continue to drive and the knee joint angle remains constant). When locked, the robot moves efficiently and smoothly in the form of a four-wheeled vehicle. At this time, in order to prevent the robot body from getting too close to the top of the busbar due to the 50mm rise of the base surface, the control system will automatically adjust the angle of the leg joints to reduce the height of the robot body relative to the center of the wheel axle by about 40mm. Then, the robot drives the wheels to travel along the 880mm long obstacle surface and performs 360° circumferential dynamic detection without blind spots on the three insulators with a diameter of 130mm and a height of 130mm distributed on it through the insulator vision inspection system. Simultaneously, it collects local overlapping images of each insulator and completes the preliminary defect identification.

[0148] When the radar detects the approaching edge of the step at the end of the obstacle, the front legs switch back to foot mode. The knee extension motors on the left side of the front fuselage extend the legs downwards to reach the ground across a 50mm height difference. At the same time, the knee extension motors on the right side of the front fuselage bend and retract to actively lower the center of gravity of the front fuselage, assisting the left side of the front fuselage to land smoothly. The right side of the front fuselage then repeats the downward movement. After the front fuselage completes the step-down movement, the rear fuselage moves forward and completes the step-down movement in the same way. Finally, the entire aircraft returns to the initial wheeled cruising posture and continues to travel to the next obstacle area.

[0149] When the robot, in wheel mode, detects a 90° left-turn right-angle corner 500mm ahead by the vehicle-mounted 2D LiDAR, the control system immediately instructs the robot to decelerate to a crawling speed of 0.1m / s and switch its four legs to foot mode in preparation for executing the split-type cornering mechanism.

[0150] like Figure 5 , Figure 6 As shown, at the start of the turning maneuver, in order to generate sufficient yaw torque and guide direction within the confined space, the hip abduction and adduction motor of the right leg structure of the front fuselage rotates outward by 25°, while its knee flexion and extension motor extends slightly to increase leg length and maintain reliable grounding over a large turning radius. Meanwhile, the hip abduction and adduction motor of the left leg structure of the front fuselage rotates inward by 20° with a slight knee flexion to serve as an inner fulcrum. Subsequently, the front fuselage is driven to rotate 22.5° to the left, achieving an initial adjustment of the direction of motion. Afterward, the front fuselage continues to rotate 22.5° to the left (accumulating to 45°). Meanwhile, the lidar and forward-looking camera continuously monitor the relative position of the robot to the side wall of the pipe, and the lidar and forward-looking camera enter a high alert state to prepare for possible obstacle avoidance; as the front body begins to deflect to the left under the driving force, the passive connecting joint is forced to bend through the central hinge pin. The key is that at this time, the sliders at both ends of the first and second connecting rods inside the middle connecting joint are forced to slide passively inward by 35mm in the longitudinal groove. This physical displacement compensation effectively shortens the equivalent wheelbase when turning and prevents the robot's edges from getting stuck on the pipe walls on both sides.

[0151] When the front body axis deflects at an angle of 45° relative to the initial path, the system accurately identifies the row of insulators to be inspected at the corner based on radar positioning data and performs a pause during travel. It immediately triggers the insulator visual inspection system to complete high-definition fixed-point capture of the three insulators at the corner, supplements the acquisition of circumferential detail overlapping images, improves the detection data, and completes the fixed-point capture task.

[0152] like Figure 7As shown, after the detection is completed, the front body rotates 22.5° to the left (to a cumulative 67.5°). At this point, if the lidar and forward-facing camera detect that the two leg structures of the front body are about to contact the inner wall of the pipe or that there is a block, the control system will immediately instruct the knee flexion and extension motor on that side to activate, driving the lower leg to lift up, thus achieving "leg lifting" to overcome the obstacle, successfully avoiding structural interference and ensuring the continuity of movement. The intermediate connecting joints continue to adapt to the movement requirements;

[0153] like Figure 8 As shown, when the front unit completes its final 22.5° rotation (reaching a cumulative 90°), it is fully inserted into the target pipe, with its axis parallel to the pipe's direction. Subsequently, the hub motors of the two leg structures of the front unit stop driving or apply reverse braking, at which point the deformation of the intermediate connecting joint reaches its maximum.

[0154] like Figure 9 As shown, the two leg structures of the front body lock the current posture, the two leg structures of the rear body begin to execute the turning sequence and rotate 22.5° to the left, and as the rear legs move, the middle connecting joint begins to deform and recover.

[0155] like Figure 10 As shown, the two leg structures of the rear body continue to rotate 22.5° to the left (to a cumulative total of 45°), the deformation of the connecting joint in the middle continues to recover, and the relative positions of the front and rear bodies are gradually adjusted.

[0156] like Figure 11 As shown, the two leg structures of the rear fuselage continue to rotate 22.5° to the left (to a cumulative 67.5°), the deformation of the connecting joint in the middle is further restored, the angle between the axes of the front and rear fuselage continues to decrease, the legs and the pipe wall always maintain non-interference contact, and the overall attitude of the fuselage gradually transitions to a straight-line driving state.

[0157] like Figure 12 As shown, the two leg structures of the rear body continue to rotate to the left until the final 22.5° rotation is completed (accumulating to 90°), so that its axis is completely parallel to the direction of the target pipe, the deformation of the middle connecting joint is completely restored, the axes of the front and rear bodies are realigned, and after the system confirms the posture alignment, it switches back to the wheel mode, adjusts the body height back to the standard value of 365mm, and accelerates to continue the inspection along the new path.

[0158] Example 3

[0159] A robot A for inspecting common busbars is basically the same as in Embodiment 2, except that the front body does not include the pole and the visual inspection system for n insulators.

[0160] Example 4

[0161] A robot C for inspecting common busbars is basically the same as in Embodiment 2, except that: Figure 15 As shown, the robot C used for inspection of common busbars has a rigid integrated body, without adopting a split structure of front body, rear body and intermediate connecting joints. It retains the circumferential array of insulator vision inspection system (including miniature industrial camera and supplementary light), which can be used for inspection of straight sections of common busbars, and can achieve 360° blind spot detection of columnar insulators through the array camera.

[0162] The inspection method of the robot C used for common busbar inspection is basically the same as that in Example 2, except that the robot used is robot C used for common busbar inspection.

[0163] The steps for inspecting the straight section of the common busbar using the above-mentioned inspection method are as follows:

[0164] (1) Detection startup and image acquisition;

[0165] When the two-dimensional high-frequency LiDAR of robot C, used for common busbar inspection, scans the contour features of the insulator to be inspected within the straight section in real time, the system immediately activates the three sets of insulator vision inspection systems at the bottom of robot C to inspect the insulator. The inspection process is as follows:

[0166] First, four miniature industrial cameras in each system are evenly distributed circumferentially along the annular detection area, synchronously triggering exposure and acquisition. Adaptive illumination is achieved using miniature LED supplementary lights to obtain locally overlapping images of the insulator's circumference. The overlap angle of adjacent camera fields of view is controlled between 10° and 20° to ensure no blind spots. To suppress motion blur caused by high-speed linear travel (moving speed v ranging from 0.5 m / s to 2 m / s), an established motion blur suppression model is adopted.

[0167]

[0168] Where v is the robot's current inspection speed, and texp is the effective exposure time of the camera. The control system limits texp to the range of 10μs to 30μs to ensure the imaging motion blur distance. It meets the detection clarity requirements for minute defects on the surface of insulators (such as cracks with a width ≥ 0.1 mm);

[0169] To address inter-frame displacement during high-speed linear motion, the Lucas-Kanade optical flow method is further employed to compensate for motion distortion. The optical flow constraint equation is as follows:

[0170]

[0171] in, and For the image in and gradient of direction, For time gradient, The optical flow displacement of a pixel is given; the inter-frame displacement is obtained by minimizing the sum of squared errors within a window (window size 15×15~21×21). And perform affine transformation compensation on the image:

[0172]

[0173] After compensation, the image motion distortion rate is controlled to ≤1%, further improving the accuracy of defect localization.

[0174] (2) Image preprocessing and coordinate mapping;

[0175] The 12 acquired partial images of insulators were preprocessed sequentially as follows:

[0176] 1) Noise suppression: Gaussian filtering is used (convolution kernel size 3×3~5×5, (Values ​​range from 0.8 to 1.2) Removes environmental and electronic noise introduced during image acquisition;

[0177] 2) Contrast Enhancement: By using histogram equalization or adaptive histogram equalization (CLAHE), the contrast between the surface texture and defect areas of the insulator is improved, thereby enhancing the identification of defect features;

[0178] 3) Distortion Correction: Radial distortion correction is performed using camera calibration parameters (intrinsic parameter matrix, distortion coefficients). The image distortion rate is controlled within a certain range after correction. This ensures the accuracy of subsequent geometric calculations.

[0179] After preprocessing, based on the established coordinate mapping model, the pixel coordinates of the planar image are mapped to the three-dimensional cylindrical surface of the insulator:

[0180]

[0181] in, Let be the planar coordinates of any pixel on the insulator unfolded diagram. The total width of the insulator unfolded image is defined by the insulator circumference, ranging from 300mm to 500mm. θ represents the circumferential angle of the insulator corresponding to that pixel (ranging from 0 to 2π), and h represents the axial height coordinate (ranging from 0 to the total height of the insulator, typically 100mm to 150mm). This step aligns each local image to a unified three-dimensional geometric reference for the insulator, providing geometric constraints for subsequent image stitching and precise defect location.

[0182] (3) Defect feature extraction and identification;

[0183] Defect feature extraction and intelligent recognition are performed on four overlapping local images corresponding to a single insulator.

[0184] 1) Feature extraction: The edge contour features of defects such as cracks, damage, and dirt on the surface of the insulator are extracted by the Canny edge detection algorithm (high and low threshold range of 50~150). Combined with texture features (such as gray-level co-occurrence matrix GLCM), defects are distinguished from normal textures.

[0185] 2) Multi-view redundancy verification: multi-view feature vectors of a single defect Cosine similarity is used to calculate the consistency between features:

[0186]

[0187] when At that time, multiple perspective observations were used to determine the same defect; subsequently, the confidence level of each perspective detection was assessed. We perform weighted fusion to obtain the final defect confidence score:

[0188]

[0189] in, For the first The weights of each viewpoint (and reprojection error) (inversely proportional) To prevent division by zero of small constants (values) This fusion formula eliminates false detections caused by single-view occlusion, lighting changes, or noise.

[0190] 3) Intelligent Classification: The preprocessed local image patch (size 224×224~448×448 pixels) is input into the deep learning-based defect classification model, which outputs the defect type (crack / damage / fouling / normal) and the location of the defect in the three-dimensional coordinates of the insulator. With a confidence score (range 0~1), the model's accuracy in identifying different types of defects is controlled between 90% and 95%, meeting the reliability requirements of industrial inspection.

[0191] (4) Image stitching and result fusion output;

[0192] Four local images corresponding to a single insulator are stitched together, based on geometric constraints of coordinate mapping and feature matching of overlapping areas, to form a complete 360° circumferential unfolded image of the insulator.

[0193] The Normalized Cross-Correlation (NCC) algorithm is used for feature matching of overlapping regions. The matching similarity is defined as:

[0194]

[0195] in, These represent the pixel grayscale values ​​of two image blocks within the overlapping region. These are the grayscale mean values ​​of the two image patches, respectively; when When a match is deemed valid, the reprojection error of the stitched image is controlled within:

[0196] Pixel;

[0197] in, To match the number of feature points, The original feature point coordinates, To ensure that the coordinates are reprojected after splicing, the surface of the insulator is free from distortion and misalignment after splicing.

[0198] The multi-view defect identification results were then fused.

[0199] 1) Duplicate detection elimination: based on defect location coordinates Combine feature similarity with multi-view annotations of the same defect to avoid duplicate statistics;

[0200] 2) False alarm filtering: Detection results with confidence scores below 0.7~0.8 are filtered to reduce the false alarm rate caused by environmental interference;

[0201] 3) Results Summary: The inspection results of the three insulators are summarized in parallel to generate a final inspection report. For the detected defects, the defect area ratio or defect length ratio is calculated based on their projection onto the 3D model of the insulator, serving as a basis for severity assessment.

[0202]

[0203] in, The projected area of ​​the defect. This represents the total surface area of ​​the insulator. The length of the crack / breakage. The circumferential circumference of the insulator; according to or Severity of scope classification:

[0204] slight:

[0205] generally:

[0206] serious:

[0207] The final inspection report includes the insulator number, defect location (circumferential angle θ, axial height h), defect type, severity (classified as minor / moderate / severe according to defect size / area), and corresponding original image clips. It is stored in the local inspection database and synchronized to the backend monitoring platform. If a severe defect (such as a through crack or large-area damage) is detected, the system immediately triggers an audible and visual alarm (alarm response time ≤ 50ms) to alert maintenance personnel for timely handling.

Claims

1. A split-type robot, a quadruped robot with straight-line and turning functions, comprising a body, wherein four leg structures are provided on the body, characterized in that, The fuselage includes a front fuselage (200), a rear fuselage (300), and an intermediate connecting joint (400). The intermediate connecting joint (400) includes a first connecting rod (410), a second connecting rod (420), and a central hinge pin (430). The rear end of the front body (200) and the front end of the first connecting rod (410) are slidably connected by a pair of sliding grooves and a slider. The sliding grooves are opened at the rear end of the front body (200), and the sliders are fixed at the front end of the first connecting rod (410). The rear end of the first connecting rod (410) and the front end of the second connecting rod (420) are rotatably connected by a central hinge pin (430) that runs vertically through them, and the axis of rotation is parallel to the vertical direction. The rear end of the second connecting rod (420) is slidably connected to the front end of the rear body (300) through a pair of sliding grooves and a slider. The sliding grooves are opened at the front end of the rear body (300), and the sliders are fixed at the rear end of the second connecting rod (420). The sliding direction of all sliding connections is parallel to the width direction of the body. During the process of the front body (200) starting to deflect under the driving force, the sliders at both ends of the first connecting rod (410) and the second connecting rod (420) inside the intermediate connecting joint (400) are forced to slide passively inward in the longitudinal groove. This physical displacement compensation effectively shortens the equivalent wheelbase when turning and prevents the body edges from getting stuck on the side tube walls.

2. The split-type robot according to claim 1, characterized in that, The leg structure includes a hip joint connecting frame (231), a hip abduction and adduction motor (232), a transition motor (234), a hip flexion and extension motor (233), a thigh link (221), a knee flexion and extension motor (222), a lower leg link (223), a hub motor (240), and a wheel; The hip joint connecting frame (231) is fixedly connected to the side of the front body (200) or the rear body (300); The housing of the hip abduction and adduction motor (232) is fixedly connected to the hip joint connecting frame (231). The output shaft of the hip abduction and adduction motor (232) is parallel to the length direction of the machine body and is fixedly connected to the housing of the transition motor (234). The output shaft of the transition motor (234) is parallel to the width direction of the machine body and is fixedly connected to the housing of the hip flexion and extension motor (233). The output shaft of the hip flexion and extension motor (233) is parallel to the width direction of the machine body and is fixedly connected to the upper end of the thigh link (221). The housing of the knee flexion and extension motor (222) is fixedly connected to the lower end of the thigh link (221), and the output shaft of the knee flexion and extension motor (222) is parallel to the width direction of the machine body and fixedly connected to the upper end of the calf link (223). The stator of the hub motor (240) is fixedly connected to the lower end of the lower leg connecting rod (223), and the rotor of the hub motor (240) is parallel to the width direction of the body and fixedly connected to the wheel.

3. A split-type robot according to claim 1, characterized in that, A forward-facing RGB camera (250) is installed at the very front of the front unit (200), and a lidar (260) is installed at the bottom.

4. A robot for inspecting common busbars, characterized in that, A split-type robot as described in any one of claims 1 to 3, wherein the dimensions of the shared busbar satisfy a specific relationship; The specific relationship is: the split robot reaches its maximum deflection angle at the intermediate connecting joint (400). Dynamic bounding rectangle width The internal clear width of the common busbar is smaller than that of the common busbar. Furthermore, the total height of the split-type robot at the preset cruising altitude The internal clear height of the common busbar is less than .

5. A robot for inspecting common busbars according to claim 4, characterized in that, The number of insulators in the same row of the common busbar is n, where n>1; The bottom of the split-type robot is equipped with n insulator vision inspection systems, which are used to capture images of the entire circumferential surface of n insulators in the same row of the common busbar.

6. A robot for inspecting common busbars according to claim 5, characterized in that, Each insulator visual inspection system includes a left front micro-shell, a right front micro-shell, a left rear micro-shell, and a right rear micro-shell; The left front micro-shell bends towards the right front micro-shell, and the left rear micro-shell bends towards the right rear micro-shell; Each miniature housing integrates a miniature industrial camera (272) and a miniature LED fill light (273).

7. A robot for inspecting common busbars according to claim 6, characterized in that, The front body (200) of the split robot includes n+1 uprights; n+1 uprights are arranged at intervals along the width direction of the front body (200) to form n gaps; The visual inspection systems for n insulators are located in n gaps respectively; In each insulator visual inspection system, the left front miniature shell and the left rear miniature shell are fixedly connected to the bottom of one pole, and the right front miniature shell and the right rear miniature shell are fixedly connected to the bottom of another pole.

8. A method for inspecting a common busbar, characterized in that, Controlling a robot for inspecting a common busbar as described in any one of claims 4 to 7 to move within the common busbar includes the following steps: (1) Environmental perception and map building; (2) Global and local path planning; (3) Motor control and gait execution; (4) Task execution and data collection.

Images