Vibration driving micro-robot based on machine vision and used for narrow environment detection

Abstract

The invention discloses a vibration driving micro-robot for narrow environment detection based on machine vision, and belongs to the field of narrow area detection. Aiming at the problem that the narrow area is difficult to detect, the novel device is designed, the structure is simple, meanwhile, a self-powered micro power supply is carried, and the condition of the narrow area is uploaded to the cloud in real time through visual navigation flaw detection and visual defect analysis. The invention discloses a vibration-driven micro-robot for narrow environment detection based on machine vision. The vibration-driven micro-robot comprises a camera assembly 1, a robot main body assembly 2 and a robot leg assembly 3, by exciting a first-order bending vibration mode, the tail end of the leg forms an elliptical motion track, the amplitude, frequency and motor phase difference are adjusted through PWM waves, multiple motion modes are achieved, a battery is arranged in the main body module, an external power source is not needed, and a camera is carried to construct a whole-process closed-loop crack analysis system. And the method is of great significance in improving the application of the micro-robot in narrow area detection.

Description

Technical Field

[0001] This invention relates to the field of narrow area detection, and more specifically to a vibration-driven microrobot based on machine vision for narrow environment detection. Background Technology

[0002] Detection in confined spaces currently faces numerous challenges. Traditional detection equipment is bulky and difficult to maneuver in narrow spaces. Even miniature probes may lack the flexibility to move and maneuver, failing to fully cover the detection area and leaving numerous blind spots. In dangerous confined areas, such as inside petrochemical pipelines or disaster ruins, manual detection poses significant safety risks, including threats from explosions, collapses, and toxic gases. Data acquired through traditional detection methods is mostly recorded on paper, and subsequent manual entry and processing are time-consuming, labor-intensive, and prone to errors. Even with digital acquisition methods, real-time processing and analysis of large amounts of complex data remains difficult, failing to meet the demands of rapid detection. Furthermore, the inability to connect to an external power source in confined spaces limits the equipment's operating time.

[0003] To address the challenge of detecting problems in narrow areas, a novel device has been designed that is structurally simple yet incorporates a self-powered micro-power source. This device utilizes visual navigation for flaw detection and visual defect analysis to upload real-time data about the narrow area to the cloud. This design is significant for enhancing the application of microrobots in narrow area detection. Summary of the Invention

[0004] This invention addresses the practical needs of narrow area detection by proposing a vibration-driven microrobot based on machine vision for narrow environment detection, aiming to achieve miniaturization, high efficiency, and automation.

[0005] The technical solution adopted by the present invention to achieve the above requirements is as follows:

[0006] A vibration-driven microrobot based on machine vision for detection in confined environments comprises three components: a camera assembly 1, a robot body assembly 2, and robot leg assemblies 3. The camera assembly 1 is mounted at the front end of the robot body assembly 2 to acquire images of the pipe. The robot body assembly 2 is connected to the robot leg assemblies 3 to integrate control and power supply modules. The robot leg assemblies 3 are symmetrically arranged on both sides of the body to generate elliptical motion trajectories to drive the robot's movement.

[0007] Furthermore, the camera assembly 1, as the main information capture component, consists of a camera 101 and a bolt 102. The camera mainly adopts a wide-angle camera to enable a larger capture range. It is fixed to the front end of the robot shell 201 by the bolt 102 and is used to collect images inside the pipeline in real time and transmit them via WiFi.

[0008] Furthermore, the robot main body component 2 is a core component of this device, including: robot shell 201, mainly responsible for housing internal components; battery 202 provides 3.7V power, eliminating the need for an external power source; 203 pin plate 1 fixes the battery 202; 204 pin 1 serves as a fixing component, fixing the motor drive board 205; the motor drive board 205 mainly generates PWM control signals for controlling the robot's straight-line movement and turning; embedded processor 206 serves as the robot's control center, responsible for processing sensor data and image transmission; 207 pin 2 serves as a fixing component, fixing the step-down module 209; 208 pin plate 2 fixes the step-down module 209, which reduces the voltage to 3.3V, supplying power to the embedded processor 206; nylon wheels 210 serve as balancing components, assisting in supporting the robot body; and robot cover 211 is used to seal the shell.

[0009] Furthermore, the robot's leg assembly 3, serving as the power core, mainly includes: a honeycomb vibration isolation module 301, which is connected to the motor box 303 via bolts 302. It divides vibration energy through multiple units, and the hollow structure utilizes air damping and wave scattering to attenuate energy, avoiding resonance and adapting to the characteristics of the vibration source, ultimately achieving efficient vibration isolation and suppressing vibration transmission to the main body; a motor box 303, which houses an eccentric vibration motor 304; the eccentric vibration motor 304, as the main power-providing component, generates periodic centrifugal force. When the two polarization motors begin to rotate at the same speed, they generate the same excitation force; a drive foot 305, fixed to the motor output shaft, converting vibration into an elliptical trajectory; and a silicone rubber composite material plate 306, which enhances foot friction and ground adaptability, preventing the robot from lifting up or vibrating in place.

[0010] Furthermore, considering the requirements for detecting narrow pipes, and to achieve miniaturized, highly maneuverable autonomous navigation, a detection method for vibration-driven microrobots used for narrow environment detection based on machine vision is proposed, comprising:

[0011] A machine vision-based motion control method for vibration-driven microrobots used for narrow environment detection, comprising the following steps:

[0012] The S100 eccentric vibration motor generates periodic centrifugal force when the eccentric rotor inside rotates, outputting vibration energy.

[0013] The S200 and ESP32 main controller generate PWM signals to drive the vibration motor, and the amplitude can be adjusted by the duty cycle.

[0014] The S300, by adjusting the motor frequency to match the natural frequency of the cantilever beam, excites the first-order bending vibration mode, causing the end of the leg to form an elliptical motion trajectory.

[0015] The S400 has two motors that receive PWM signals of the same frequency and phase. The direction of the elliptical trajectory is consistent with the direction of the speed magnitude, which propels the robot to move forward in a straight line.

[0016] The S500 adjusts the PWM frequency or phase difference of a single-sided motor to make the bipedal elliptical trajectory asymmetrical, generating a differential effect to change the direction and speed of motion and achieve steering.

[0017] A machine vision-based method for automatic navigation control of vibration-driven microrobots for narrow environment detection, comprising the following steps:

[0018] D100, Pipeline Environment Sensing and Positioning: Upon startup, the onboard miniature camera 1 continuously collects environmental images as it travels within the pipeline. A flexible vibration isolation module 301 is installed between the main body and the drive foot 3 to effectively suppress the interference of vibrations generated by the vibration motor during operation on the camera imaging, ensuring image clarity. The onboard embedded processor transmits the compressed image data to an external control terminal in real time via a WiFi module.

[0019] D200, Dynamic Target Locking and Pose Adjustment: Camera 1 acquires images at a fixed frequency. The control terminal performs real-time analysis of the received images.

[0020] D210, Target Detection: Determines whether a suspected crack target is detected in the current image. If no crack is detected, the pipeline axis continues to move and explore.

[0021] D220, Target Locking and Mode Selection: Once a suspected crack target is detected, the control terminal generates control commands based on the target's position (left, center, right) in the image.

[0022] If the target is located on the left side of the image, a left turn command is sent to control the drive foot 3 to adjust its movement direction. The left drive foot decelerates, while the right drive foot accelerates, allowing the camera 1 to more accurately align with the target area.

[0023] If the target is located on the right side of the image, send a right turn command to control the drive foot 3 to adjust the direction of movement and make an adjustment in the opposite direction.

[0024] If the target is basically centered or centered after adjustment, a return-to-center command is sent to maintain the current state and enter the fine detection mode, triggering the crack identification process of Example 3.

[0025] D300, closed-loop navigation control: During pose adjustment, camera 1 continuously reports the target position. When the target remains within the center 5% area of ​​the image for three consecutive frames, a successful lock is determined. If the adjustment times out, the current position is recorded and the investigation continues.

[0026] A machine vision-based method for crack classification alarm of vibration-driven microrobots for narrow environment detection, comprising the following steps:

[0027] C100: Measurement of key crack parameters: Calculation of geometric parameters for the crack profile extracted in Example 3:

[0028] C110, Calculate the minimum bounding rectangle: Calculate the minimum bounding rectangle for each crack profile.

[0029] C120, crack length L: The length of the longest side of the smallest bounding rectangle is taken as the main direction length of the crack, which is also the crack length.

[0030] C130, crack width W: The length of the shorter side of the smallest bounding rectangle is used as the crack width characterization. To more accurately reflect the crack width variation, the maximum value among all width measurement points on the profile is usually taken as the final crack width W.

[0031] C200 Crack Classification and Alarm Decision: Cracks are classified based on the measured crack length L and width W, and an alarm is triggered accordingly.

[0032] C210, width grading:

[0033] If so, it is defined as a narrow crack, i.e., b=2.

[0034] If b=1, then it is defined as a medium crack.

[0035] If so, it is defined as a wide crack, i.e., b=0.

[0036] C220, length classification:

[0037] If so, it is defined as a short crack, i.e., a=2.

[0038] If so, it is defined as a medium crack, i.e., a=1.

[0039] If so, it is defined as a long crack, i.e., a=0.

[0040] C230, generating a grading vector: Combining the width and length grading results, a crack state vector F(a,b) is formed:

[0041] b: Width safety indicator. b=2 indicates width safety (narrow crack), b=0,1 indicates width danger (wide crack, medium crack).

[0042] a: Length safety indicator. a=2 indicates length safety (short crack), a=0,1 indicates length danger (long crack, medium crack).

[0043] C240, Alarm Condition Determination: Query the preset alarm condition table based on the value of vector F(a,b):

[0044] condition Handling method a=0, b=0 Emergency alarm a=0, b=1 High-priority alarms a=1, b=0 Medium priority alarm a=1, b=1 Medium priority alarm Other combinations neglect

[0045] C300, Alarm Execution and Data Logging:

[0046] C310, alarm trigger: If an alarm is required (i.e. (a,b) is (0,0), (0,1), (1,0), or (1,1)), the control terminal sends a command via the wireless link:

[0047] Turn on the LED lights to provide on-site visual alarm.

[0048] At the same time, the control terminal generates alarm information of the corresponding level (emergency / high priority / medium priority) and transmits it to the upper monitoring center or mobile terminal through wireless network (such as 4G / 5G) or data antenna.

[0049] C320, Data Recording and Transmission: Regardless of whether an alarm is triggered, detected crack images, key parameters (L, W), grading results (wide / medium / narrow, long / medium / short), and information such as time and location are all recorded. This data is continuously transmitted back to the control terminal via a WiFi module or data antenna for storage and analysis, used for pipeline health status assessment and report generation.

[0050] The advantages of this invention compared to existing technologies, due to the adoption of the above-mentioned techniques, are as follows:

[0051] (1) The present invention adopts a “compact integration of three components” (camera + main body + legs), and the main body module has a built-in battery, control and power supply unit (no external power supply required), realizing the miniaturization of the whole machine, which can directly penetrate into narrow environments to complete the detection, filling the gap of existing technology in efficient detection in narrow spaces.

[0052] (2) This invention uses an eccentric vibration motor to drive the vibration, exciting a first-order bending vibration mode, causing the end of the leg to form an elliptical motion trajectory. Combined with the bottom silicone rubber composite plate, it enhances friction to prevent slippage and reduces dependence on ground flatness, making it adaptable to various complex terrains. Simultaneously, by adjusting the amplitude, frequency, and motor phase difference through PWM waves, it achieves multiple motion modes such as straight-line movement, turning, and obstacle avoidance, offering far greater flexibility than traditional drive methods.

[0053] (3) The present invention effectively suppresses the transmission of motor vibration to the main body (and camera) through the honeycomb vibration isolation module (multi-unit segmented vibration energy + air damping attenuation), ensuring that the camera can still collect clear images during robot movement, providing a high-quality data foundation for subsequent crack detection and target locking, and solving the contradiction of "vibration-imaging" in the prior art.

[0054] (4) This invention constructs a “full-process closed-loop crack analysis system” to realize the functional upgrade from “detection” to “analysis-early warning-source tracing”. For suspected cracks detected, the crack outline is automatically extracted and key geometric parameters are calculated. The long side of the smallest bounding rectangle of the crack is the “crack length L” and the maximum value of the short side is the “crack width W”. Objective data replaces traditional subjective judgment to ensure the accuracy of crack status assessment. A grading standard is established based on the dual dimensions of “length (short / medium / long) + width (narrow / medium / wide)”. On-site visual warning is achieved through LED lights, and emergency / high / medium priority alarm information is sent to the monitoring center remotely through 4G / 5G and other networks to achieve “hazard matching early warning” and avoid excessive alarm or delayed early warning.

[0055] (5) Through a fully wireless design, the present invention eliminates the need for cable connections, enabling the robot to operate autonomously in narrow or high-risk environments (such as deeply buried pipes or small cavities) that are inaccessible to humans. This not only expands the detection range but also avoids the risk of personnel directly contacting dangerous environments, achieving truly unmanned detection. Attached Figure Description

[0056] Figure 1 Overall structural diagram of the invention

[0057] Figure 2 Schematic diagram of the internal device of the robot body of the present invention

[0058] Figure 3 Schematic diagram of the internal structure of the drive leg and the vibration isolation module of the present invention

[0059] Figure 4 Schematic diagram of the internal device of the robot body of the present invention

[0060] Figure 5 Flowchart of Automatic Navigation Control for Crack Identification in this Invention

[0061] Figure 6 Flowchart of crack identification and error reporting in this invention

[0062] Figure 7 Flowchart of crack classification in this invention

[0063] Figure 8 Crack identification pattern diagram of the present invention

[0064] Figure 9 Schematic diagram of the motion principle of this invention

[0065] Figure 10 Schematic diagram of the driving signal of the present invention

[0066] In the attached diagram: 1. Camera assembly; 2. Robot body assembly; 3. Robot leg assembly; 101. Camera; 102. Bolt 2; 201. Robot shell; 202. Battery; 203. Pin plate 1; 204. Pin 1; 205. Motor drive board; 206. Embedded processor; 207. Pin 2; 208. Pin plate 2; 209. Step-down module; 210. Nylon wheel; 211. Robot cover; 301. Honeycomb vibration isolation module; 302. Bolt 1; 303. Motor box; 304. Motor; 305. Drive leg; 306. Silicone rubber composite material plate. Detailed Implementation

[0067] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0068] Please refer to Figures 1 to 10 As shown, a vibration-driven microrobot for narrow environment detection based on machine vision is illustrated, comprising: a camera assembly 1, a robot body assembly 2, and a robot leg assembly 3; the camera assembly 1 is mounted on the robot body assembly 2 using 102 bolts 2, and the robot leg assembly 3 is fixed to the robot body assembly 2 using 302 bolts 1.

[0069] Furthermore, in a preferred embodiment, the internal devices of the robot body refer to Figure 2 As shown, the fixed motor drive board 205 is fixed to the robot shell 201 by pin 204, the battery 202 is placed on pin 203 and fixed to the rear of the robot shell 201, pin 207 fixes the embedded processor 206, the step-down module 209 is placed on pin 208 and fixed to the front of the robot shell 201, and the nylon wheel 210 is installed at the lower end of the robot shell 201 to help support the body.

[0070] Furthermore, in a preferred embodiment, the drive leg interior and vibration isolation module 3, refer to... Figure 3 As shown, the system includes a 301 honeycomb vibration isolation module, 302 bolt 1, 303 motor housing, 304 motor, 305 drive foot, and 306 silicone rubber composite material plate. The eccentric vibration motor 304 is placed in the motor housing 303 and assembled at a specific position on the drive foot 305. The honeycomb vibration isolation module 301 is installed at one end of the drive foot 305 to suppress vibration transmission to the main body. Bolt 302 securely connects the vibration isolation module to the robot shell 201. Figure 4 As shown, the silicone rubber composite plate 306 is attached to the lower end of the drive foot 305 to increase the friction between the drive foot and the contact surface and prevent the robot from tilting upwards.

[0071] Furthermore, in a preferred embodiment, the present invention provides a machine vision-based vibration-driven microrobot driving method for narrow environment detection, combined with... Figure 10 As shown, taking one rotation cycle as an example, the path of one rotation is divided into four quadrants, each quadrant representing one-quarter of the rotation, or 90°. Assuming the polarizing rotor starts in the first quadrant, when it enters the first quadrant, the horizontal component of the force gradually decreases, causing the driving foot's speed to decrease. When it enters the second quadrant, the horizontal component of the force gradually increases along the negative direction of motion, causing the driving foot's speed to change to the negative y-axis direction. When it enters the third quadrant, the horizontal component of the force gradually decreases along the negative direction of motion, causing the driving foot to eventually stop moving due to continuous friction from the ground. When it enters the fourth quadrant, the horizontal component of the force continuously propels the driving foot to move along the positive y-axis. Therefore, the polarizing rotor reciprocates within one rotation cycle. In summary, when the driving foot moves in the positive direction, the vertically downward force gradually decreases, and the friction decreases. Conversely, when the driving foot moves in the negative direction, the vertically downward force gradually increases, and the friction increases. Therefore, this kind of friction-driven propulsion foot needs to have a greater displacement in the positive direction than in the negative direction in order to achieve forward movement of the robot.

[0072] Furthermore, in a preferred embodiment, the generation of the drive signal and its obstacle avoidance method are as follows: Battery 202 provides power to drive motor drive board 205, which generates a PWM control signal to control the vibration of left and right eccentric vibration motors 304. The drive foot has two degrees of freedom, allowing it to turn freely to avoid obstacles during walking. Movement is achieved through vibration and ground friction, requiring less surface flatness and enabling it to handle various complex terrains. Multiple movement modes can also be achieved through PWM wave adjustment. Figure 10 As shown, when the eccentric vibration motor 304 vibrates, it causes the robot's driven feet 305 to generate elliptical motion, contacting the ground. The forward driving force generated by friction with the ground enables the robot's forward movement. Remote Bluetooth communication is achieved using the BLE communication module in the embedded processor 206. The host computer can send data commands to the embedded processor 206. When the command is interpreted as straight movement, the left and right eccentric vibration motors 304 move at the same speed, with the speed and direction of movement being the same. When the command is interpreted as clockwise movement, the PWM frequency of one motor is adjusted, causing the elliptical trajectory of the two feet to deviate, creating a differential speed effect. The speed on the outer side of the rotation direction is greater than the speed on the inner side, and the movement speed is in the robot's forward direction. When the command is interpreted as clockwise movement, the speed on the outer side of the rotation direction is greater than the speed on the inner side, and the movement speed is in the robot's backward direction.

[0073] Furthermore, in a preferred embodiment, the crack identification automatic navigation control process method of the present invention refers to... Figures 5 to 8 As shown:

[0074] Reference Figure 5 As shown in the flowchart of the automatic navigation control for crack identification in this invention, the robot first detects the pipeline image in real time using camera 101. When a crack is detected, its location is identified. When the crack is on the left, the amplitude of the right leg of the bipedal robot increases, while the amplitude of the left leg decreases. When the crack is on the right, the amplitude of the left leg increases, while the amplitude of the right leg decreases. When the crack is in the center, the robot decelerates and focuses, and the amplitude of both legs decreases. When the size of the crack exceeds a set threshold, the robot stops moving, activates LED lighting for image recording, and initiates wireless data transmission.

[0075] Reference Figure 6 Flowchart of crack identification and error reporting in this invention and Figure 8 As shown in Figure a, the crack identification method of this invention involves stabilizing the suspected crack area on the inner wall of the pipe, acquiring a high-definition RGB crack image of the pipe, and transmitting it to the embedded processor 206 for preprocessing: firstly, grayscale processing is performed, using a weighted average method to convert the color pipe crack RGB image into a grayscale image, significantly reducing the amount of data and focusing on the crack structure features; as shown in Figure a. Figure 8 The crack identification pattern of this invention is shown in Figure b. Gaussian filtering is then applied to smooth and denoise the grayscale image, effectively eliminating noise interference caused by particles adhering to the inner wall of the pipe and uneven lighting. Furthermore, adaptive binarization segmentation is implemented, dynamically calculating the threshold through local windows to enhance the contrast between the crack and the metal background, highlighting the crack features. Figure 8 The crack recognition pattern of this invention is shown in Figure c. Based on this, morphological optimization is performed: the opening operation uses an elliptical kernel to first erode and then expand the crack image, eliminating pseudo-crack connections formed by rust particles; the closing operation uses the same structural element to first expand and then erode, filling the microscopic pores inside the real crack. After the above processing, since cracks are generally irregular in shape, precise crack contour extraction is performed: the Canny edge detection algorithm is used to accurately extract all potential crack edge contours in the crack image. After four steps of Gaussian filtering pre-smoothing, gradient magnitude calculation, non-maximum suppression, and double threshold hysteresis processing, the edge points are accurately identified. After finding and obtaining all candidate contours, strict screening is performed based on the morphological characteristics of the pipe crack—micro-noise contours with an area of ​​less than 50 pixels are removed, and interference such as circular weld points with an aspect ratio ≤ 5 are excluded. Finally, crack contours with significant slender features are retained, establishing a reliable foundation for subsequent geometric parameter measurements.

[0076] Reference Figure 7 The crack classification flowchart of this invention shows that the crack classification vector is obtained based on the width and length classification, and the alarm condition is determined according to the alarm condition table.

[0077] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A vibration-driven microrobot for narrow environment detection based on machine vision, characterized in that, include: The robot consists of a camera assembly (1), a robot body assembly (2), and a robot leg assembly (3); the camera assembly (1) is mounted on top of the robot body assembly (2) using a (102) bolt 2, and the robot leg assembly (3) is fixed to the robot body assembly (2) using a (302) bolt 1; The camera assembly (1) is installed at the front end of the robot body assembly (2) to collect images of the pipeline; the robot body assembly (2) is connected to the robot leg assembly (3) to integrate the control and power supply modules; the robot leg assembly (3) is symmetrically arranged on both sides of the body to generate elliptical motion trajectories to drive the robot to move.

2. The vibration-driven microrobot for narrow environment detection based on machine vision according to claim 1, characterized in that, The camera assembly (1) is the main information capture component, consisting of a camera (101) and a bolt (102). The camera is mainly a wide-angle camera, which makes the camera capture range larger. It is fixed to the upper end of the robot body assembly (2) by bolts (102) for real-time acquisition of images inside the pipeline and transmission via WiFi.

3. The vibration-driven microrobot for narrow environment detection based on machine vision according to claim 2, characterized in that, Featuring a wide-angle camera design, its core advantage is a significantly expanded image capture range, adapting to the information collection needs of complex environments within pipelines. Equipped with WiFi transmission capabilities, it can quickly transmit real-time pipeline image data to the backend, supporting real-time monitoring and analysis.

4. The vibration-driven microrobot for narrow environment detection based on machine vision according to claim 2, characterized in that, The main robot component (2) includes: a robot shell (201), which is mainly responsible for housing internal components; a battery (202) that provides 3.7V power, requiring no external power supply; (203) a pin plate 1 that fixes the battery (202); (204) a pin 1 that serves as a fixing component to fix the motor drive board (205); the motor drive board (205) mainly generates PWM control signals to control the robot's straight-line movement and turning; an embedded processor (206) that serves as the robot's control center and is responsible for processing sensor data and image transmission; (207) a pin 2 that serves as a fixing component to fix the step-down module (209); (208) a pin plate 2 that fixes the step-down module (209) to reduce the voltage to 3.3V, which is responsible for supplying the embedded processor (206); nylon wheels (210) that serve as balancing components to assist in supporting the robot body; and a robot cover (211) that is used to seal the shell.

5. The vibration-driven microrobot for narrow environment detection based on machine vision according to claim 4, characterized in that, The core motion device of the robot includes: a honeycomb vibration isolation module (301) connected to the motor box (303) by bolts (302), which divides the vibration energy through multiple units, and the hollow structure uses air damping and wave scattering to attenuate the energy, avoid resonance and adapt to the characteristics of the vibration source, and finally achieve efficient vibration isolation and suppress the transmission of vibration to the main body; the motor box (303) is used to house the eccentric vibration motor (304); the eccentric vibration motor (304) is the main power-providing component, which generates periodic centrifugal force. When the two polarization motors start to rotate at the same speed, they will generate the same excitation force; the driving foot (305) is fixed to the motor output shaft and converts the vibration into an elliptical trajectory; the silicone rubber composite plate (306) can enhance the foot friction and ground adaptability, and prevent the robot from lifting up or vibrating in place.

6. The vibration-driven microrobot for narrow environment detection based on machine vision according to claim 5, characterized in that, After startup, the onboard miniature camera (1) continuously collects environmental images as it moves through the pipe. A flexible vibration isolation module (301) is provided between the robot's main body component (2) and the robot's leg component (3) to effectively suppress the interference of vibration generated by the vibration motor during operation on the camera imaging, ensuring image clarity. The onboard embedded processor transmits the compressed image data to the external control terminal in real time via the WiFi module. Once a suspected crack target is detected, the embedded processor (206) generates control commands based on the target's position (left, center, right) in the image. When the eccentric rotor inside the eccentric vibration motor (304) rotates, it generates periodic centrifugal force and outputs vibration energy. The embedded processor (206) and the main control motor drive board (205) generate PWM control signals and adjust the amplitude by adjusting the duty cycle. By adjusting the motor frequency to match the natural frequency of the cantilever beam, the first-order bending vibration mode is excited, causing the end of the leg to form an elliptical motion trajectory. The two motors receive the same frequency and phase PWM signals, and the direction of the elliptical trajectory is consistent with the direction of the speed magnitude, propelling the robot forward in a straight line. Adjusting the PWM frequency or phase difference of a single motor makes the bipedal elliptical trajectory asymmetrical, generating a differential effect to change the direction and speed of motion and achieve steering.

Images