A hummingbird-inspired bionic drone's fuselage spine folding control system and its working method

By using a hummingbird-inspired conical spine structure and multi-dimensional parameter acquisition, combined with a light-vibration adaptive noise reduction mechanism, a closed-loop control system was constructed. This solved the problems of fuselage height folding adaptability and calculation accuracy of UAVs during narrow space passage, and improved the stability and safety of UAVs during passage in narrow spaces.

CN123113605APending Publication Date: 2026-11-17HANGZHOU YIJIA BIONIC TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611477509.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing drones suffer from poor adaptability when traversing narrow spaces due to their highly folded fuselage, low accuracy in spatial parameter calculations, and lack of dynamic closed-loop optimization, resulting in poor safety and stability.

Method used

Adopting a hummingbird-inspired conical spine structure, combined with multi-dimensional spatial and fuselage parameter acquisition, and equipped with a light-vibration adaptive noise reduction mechanism, a closed-loop control system is constructed to achieve continuous adjustment of the fuselage height. Through full-process status monitoring and error iterative optimization, the adaptability to narrow space passage and calculation accuracy are improved.

Benefits of technology

It enables continuous adjustment of the drone's fuselage height, improves the structural adaptability and calculation accuracy for narrow space passage, reduces the risk of collisions in complex working conditions, and ensures flight stability and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113605A_ABST
    Figure CN123113605A_ABST
Patent Text Reader

Abstract

This invention discloses a hummingbird-inspired bionic drone's spinal folding control system and its operating method, belonging to the field of bionic flight technology for drones. The invention uses a binocular camera, lidar, and multiple onboard sensors to collect full-dimensional parameters of the drone's attitude within a confined space. Abnormal data is eliminated using a statistical thresholding method, and image noise is eliminated using adaptive dual filtering based on light and vibration. The actual accommodative height of the space is calculated by fusing tilt angle, obstacle protrusion, and channel dimensions, and the axial target retraction distance is calculated based on the geometric relationship of the conical spine. Front and rear micro-electric actuators work together to drive the nose and tail fins to retract synchronously, lowering the drone's height. After completion, a closed-loop verification of the drone's height is performed; if insufficient, a secondary retraction is performed, and error samples are used for parameter iteration and optimization. Throughout the flight, the folding amplitude is dynamically and adaptively adjusted based on flight speed, vibration, and channel dimensions. This invention features a biomimetic hummingbird-shaped conical spine axial telescopic structure, specifically designed for narrow passages with limited height. It constructs a fully closed-loop control system encompassing "data acquisition, noise reduction, calculation, folding, dynamic adjustment, and iterative optimization," effectively improving spatial calculation accuracy and stability in narrow spaces. This addresses the shortcomings of existing UAVs, such as poor spatial calculation accuracy, lack of adaptive folding, and susceptibility to collisions. It is suitable for narrow-space operation scenarios such as rubble search and rescue, power line inspection, and indoor detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicles (UAVs), and in particular to a hummingbird-inspired bionic UAV with a folding control system for its fuselage spine and its operating method. Background Technology

[0002] With the continuous maturation of drone technology, drones have been widely used in various fields such as power line inspection, indoor search and rescue, rubble detection, and building surveying. These application scenarios commonly involve narrow spaces with limited height, such as wall gaps, pipe joints, gaps in building beams, and narrow crevices in forests, placing extremely high demands on the drone's ability to navigate these confined spaces.

[0003] Most existing drones have fixed fuselage structures that cannot adapt to spatial dimensions and are prone to collisions with spatial boundaries or internal obstacles when traversing narrow spaces, resulting in fuselage damage, mission interruption, or even crashes. While some biomimetic drones have folding structures, these are mostly designed for folding wings or arms, which only reduces the lateral width. They lack an axial telescopic folding design that mimics the conical body shape of a hummingbird, making it impossible to achieve continuous and precise control of the fuselage height through axial retraction, and thus difficult to match narrow passages of different heights.

[0004] Meanwhile, existing drones lack deep integration between spatial perception and folding control. The spatial parameters of image analysis are easily affected by noise from ambient light and fuselage vibration. Conventional fixed parameter filtering schemes have large errors in spatial height calculation under complex conditions. Moreover, most schemes only execute a single folding command, lack a dynamic adjustment mechanism for the entire traversal, and also lack the ability to iteratively optimize folding errors in a closed loop, resulting in poor safety and stability when traversing narrow spaces.

[0005] Drawing inspiration from the axially extendable conical body design of hummingbirds, this invention proposes a fuselage spine folding system and method that combines precise image analysis, multi-parameter fusion calculation, adaptive folding control, and closed-loop verification. This is crucial for addressing the challenges of UAVs navigating narrow spaces. Therefore, to address the problems of poor fuselage height folding adaptability, low spatial parameter calculation accuracy, and lack of dynamic closed-loop optimization in existing technologies for UAVs navigating narrow spaces, this invention proposes a fuselage spine folding system and its operating method for a hummingbird-inspired UAV. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a hummingbird-inspired bionic drone's fuselage spine folding system and its operating method. Based on multi-dimensional spatial and fuselage parameter acquisition, and combined with a light-vibration adaptive dual noise reduction mechanism and a spatial size correction algorithm, it achieves reliable calculation of the accommodating height in height-restricted narrow spaces. Furthermore, based on the geometric relationship of the conical spine, it controls the axial contraction of the fuselage and the synchronous retraction of the head and tail, ensuring the fuselage height matches the dimensions of the narrow space. Simultaneously, it achieves dynamic adjustment of folding parameters through continuous status monitoring and utilizes error samples to perform iterative parameter optimization, constructing a fully closed-loop control system. This aims to improve the adaptability, calculation accuracy, and flight stability for navigating narrow spaces.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows.

[0008] The hummingbird bionic drone of this invention adopts a conical spine-type main load-bearing frame. The front and rear axial electric push rods drive the head movable frame and the tail movable frame to retract synchronously towards the middle along the axial direction. By utilizing the geometric relationship of the conical sidewalls, the axial retraction is converted into a reduction in the body height, realizing continuous adjustment of the body height. It is equipped with a closed-loop control system consisting of a sensing unit, a processor unit, an execution drive unit, and a position feedback unit. Combined with highly transparent deformable bionic skin, it still maintains environmental perception capabilities in the retracted state.

[0009] Furthermore, the frame, as the main load-bearing structure of the spine, adopts a titanium alloy honeycomb sandwich conical design, with the sidewalls forming an angle of 30°~45° with the fuselage axis. Batteries are installed at the bottom of the frame, with the front axial electric actuators located on the inner side of the front end of the frame and the rear axial electric actuators symmetrically located on the inner side of the rear end of the frame. The total retraction distance is borne by the front and rear electric actuators in a 1:1 ratio. Position sensor two is located at the front end of the top of the frame, and position sensor three is located at the rear end of the top of the frame. Speed ​​sensors, tilt sensors, and vibration sensors are integrated in the middle of the frame to monitor changes in flight attitude in real time. The processor is located in the middle of the frame, forming a distributed sensing network with the various sensors.

[0010] Furthermore, the bionic skin fully covers the surface of the frame, with the front end fixed to the outer edge of the head movable frame and the rear end fixed to the outer edge of the tail movable frame; a rigid light-transmitting window is embedded in the corresponding area of ​​the camera to avoid shrinkage and wrinkles obstructing the image.

[0011] Furthermore, the wing is hummingbird-shaped, with its root embedded in the maximum contour of the middle of both sides of the frame, and is driven to flap by the internal drive mechanism of the frame; the maximum outer edge of the flapping wing is always within the height envelope of the maximum cross section of the frame, so there is no need to store the wing separately after the fuselage is folded and stored.

[0012] Furthermore, the tail movable frame is located at the rear end of the frame, and the tail fin is fixedly installed at the end of the tail movable frame; the root of the tail movable frame is connected to the output end of the rear axial electric push rod, which can be driven to slide forward along the frame axis and be stored therein; four position sensors are embedded in the root of the tail movable frame to provide real-time feedback on the tail fin position.

[0013] Furthermore, the movable frame of the machine head is set at the front end of the frame, and two sets of panoramic cameras and LiDAR are set at the top. A position sensor is arranged at the top of the movable frame. When the movable frame is retracted, the panoramic cameras can continuously collect environmental data through the light-transmitting window.

[0014] Furthermore, the sensors, lidar, and cameras transmit the collected data to the processor, which processes the data and then sends commands to the front and rear axial electric actuators. The two actuators work together to achieve adaptive adjustment of the fuselage height, enabling the aircraft to navigate through narrow spaces.

[0015] The present invention also provides a corresponding control method, the core process of which is as follows.

[0016] The first step is to collect parameters in all dimensions, simultaneously acquiring environmental parameters in the confined space and the drone's own status parameters.

[0017] The second step is data preprocessing, which involves using statistical thresholds combined with physical correlation constraints to remove outlier data and generate a high-quality, effective parameter set.

[0018] The third step involves adaptive dual noise reduction and spatial calculation. The filtering parameters are dynamically adjusted based on light and vibration to eliminate image noise. Multiple parameters are fused to correct the actual accommodative height in space, and the shrinkage distance of the spinal target is calculated.

[0019] The fourth step involves precise folding control, which drives the front and rear electric push rods to retract synchronously. The height of the retracted body is verified based on the measured stroke. If it is insufficient, a second retraction is performed. At the same time, the folding accuracy error is calculated for iterative optimization.

[0020] The fifth step involves dynamic adjustment throughout the entire process. During the crossing, the folding range is dynamically adjusted according to real-time working conditions. Long and narrow passages are locked in place, and a safe retreat is triggered in case of abnormal working conditions to ensure the stability of the crossing.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention is based on a hummingbird-inspired conical spine folding structure. By axially contracting, it reduces the fuselage height, achieving continuous adjustment of the fuselage height. Unlike traditional wing folding solutions, it is specifically adapted to narrow space scenarios with limited height, improving the structural adaptability of UAVs for narrow space passage. 2. This invention adopts a light-vibration adaptive dual noise reduction mechanism, which dynamically adjusts the Gaussian filter intensity according to the ambient light brightness and the median filter window according to the body vibration frequency, specifically eliminating Gaussian noise and salt-and-pepper noise, thereby improving the reliability of spatial height calculation under complex working conditions and providing a stable data foundation for folding control. 3. This invention integrates tilt attitude, obstacle protrusion amount, channel axial length, fuselage projection envelope, and other multi-dimensional parameters to correct the space-accommodating height, and dynamically adjusts safety redundancy based on flight speed, making the folding amount calculation more consistent with actual flight conditions and reducing the collision risk during tilted flight. 4. This invention constructs a complete closed-loop control system of "acquisition-calculation-folding-verification-iterative optimization". The folding effect is verified based on the measured stroke. If it is insufficient, a secondary contraction is performed. At the same time, the core coefficients are continuously iterated through error samples. The folding control accuracy can be gradually optimized with use. 5. This invention sets up a dynamic adjustment mechanism and failure degradation strategy throughout the entire process. Long and narrow channels are locked in a folded state to avoid frequent adjustments. Safe retreat can be performed under abnormal working conditions. It has strong versatility and can be adapted to height-restricted narrow spaces of different sizes and with different obstacle distributions. It is suitable for a variety of drone operation scenarios. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the appearance of a hummingbird-inspired bionic drone provided in an embodiment of the present invention.

[0023] Figure 2 This is an extended schematic diagram of a hummingbird drone provided as an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of the retraction of a hummingbird drone provided in an embodiment of the present invention.

[0025] Figure 4 This is a schematic diagram of the electronic structure of a hummingbird drone provided in an embodiment of the present invention.

[0026] Figure 5 This is a schematic diagram of the system structure provided in an embodiment of the present invention.

[0027] Figure 6 This is a flowchart provided for an embodiment of the present invention.

[0028] Explanation of reference numerals in the attached figures: 10. Frame; 101. Battery; 102. Front electric actuator; 103. Rear electric actuator; 104. Position sensor two; 105. Position sensor three; 106. Speed ​​sensor; 107. Tilt sensor; 108. Vibration sensor; 109. Processor; 11. Bionic skin; 12. Wing; 13. Tail fin; 131. Position sensor four; 14. Movable frame; 141. Panoramic camera; 142. LiDAR; 143. Position sensor one. Detailed Implementation

[0029] To make the above-mentioned objectives, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments; thereby, the process of how the present application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0030] In this embodiment, all length measurements are expressed in millimeters (mm), flight speed in m / s, and vibration frequency in Hz. The terminology used throughout is consistent with "forward axial actuator," "rear axial actuator," "nose movable mount," and "tail movable mount." The tail fin is fixedly mounted to the end of the tail movable mount. This embodiment defines the fuselage coordinate system as follows: the direction from the nose to the tail is the X-axis (fuselage axis), the vertical upward direction is the Y-axis (height direction), and the horizontal direction is the Z-axis (width direction).

[0031] Example 1

[0032] This embodiment discloses a hummingbird bionic drone, including a frame, bionic skin, wings, tail, nose movable frame, and tail movable frame.

[0033] The frame, serving as the main load-bearing structure of the spine, adopts a titanium alloy honeycomb sandwich conical design. The angle α between the sidewall and the X-axis (body axis) is 30°~45°, preferably 38.66°, corresponding to a spine contraction height conversion coefficient k=tanα≈0.8, exhibiting high strength and lightweight characteristics. A battery is installed at the bottom of the frame to power the entire machine. The front axial electric actuator is fixedly installed on the inner side of the front end of the frame, arranged along the X-axis direction; the rear axial electric actuator is symmetrically fixed on the inner side of the rear end of the frame, also arranged along the X-axis direction. The total contraction distance is shared by the two electric actuators in a 1:1 ratio, but the distribution ratio can be adjusted according to actual structural requirements.

[0034] Conical spine stiffness check: The axial compressive stress formula of thin-walled conical shell is used for check. Under contraction, the axial compressive stress satisfies σ=F / A≤σs / 1.5, where F is the total axial push rod thrust, A is the equivalent bearing area of ​​the frame section, σs is the yield strength of titanium alloy material, and the safety factor is not less than 1.5 to ensure that the frame has no plastic deformation and structural instability during contraction.

[0035] Position sensor two is located at the front end of the top of the frame, and position sensor three is located at the rear end of the top of the frame, used to detect the extreme positions of the nose and tail movable frames, respectively. Speed, tilt, and vibration sensors are integrated in the middle of the frame to monitor attitude parameters such as flight speed, pitch angle, and fuselage vibration frequency in real time. The processor is located in the middle of the frame and is electrically connected to all sensors and actuators, forming a distributed sensing and control network.

[0036] The output end of the front axial electric actuator is fixedly connected to the lower end of the machine head movable frame by bolts. A slider is provided on the outer side of the machine head movable frame, and an axial slide rail that cooperates with the slider is provided on the inner side of the front end of the frame. When the front axial electric actuator extends or retracts, it drives the machine head movable frame to slide backward along the slide rail in the X-axis direction, realizing the extension and retraction of the machine head.

[0037] The rear axial electric actuators are symmetrically arranged on both sides of the rear end of the frame, and the output end of each rear axial electric actuator is fixedly connected to the root of the tail movable frame. A sliding sleeve is provided at the root of the tail movable frame, which is sleeved on the outside of the tail end of the frame. When the rear axial electric actuators extend or retract, they drive the sliding sleeve to slide forward along the X-axis, thereby causing the tail movable frame to extend or retract; the tail fin is fixedly installed at the end of the tail movable frame and moves synchronously with the fuselage.

[0038] Both the front and rear axial electric actuators are miniature DC linear actuators, with a maximum stroke of 30mm and a minimum stroke of 0mm per actuator. The movement speed is 2mm / s, and the stroke control accuracy is ±0.1mm, enabling high-precision telescopic control. The synchronization error tolerance for the two actuators is ±0.2mm. When this tolerance is exceeded, the system triggers a synchronization correction command, sending a compensation pulse to the side with the smaller stroke. The compensation step size is 0.1mm per pulse until the deviation returns to the tolerance range.

[0039] The bionic skin is made of transparent silicone or polydimethylsiloxane (PDMS) elastomer material, which has a light transmittance of not less than 90% in the visible light band and an elongation at break of not less than 300%, and can recover its original shape without permanent deformation after repeated stretching and contraction. The thickness of the bionic skin is 0.5 mm to 1.0 mm. When the moving frame of the machine head is fully retracted, the bionic skin produces axial compression wrinkles but does not break; when fully extended, the bionic skin stretches axially without tearing.

[0040] Fatigue life of bionic skin: Based on the maximum stretching strain under working conditions ≤30% (far lower than the material's elongation at break), the estimated fatigue cycle life is no less than 10,000 stretching cycles; it is recommended to replace the bionic skin when the cumulative stretching reaches 8,000 cycles or after 12 months of continuous use to avoid aging and cracking that affects light transmission and sealing performance.

[0041] The front end of the bionic skin is fixedly connected to the outer edge of the head frame via adhesive bonding, and the rear end is fixedly connected to the outer edge of the tail frame via adhesive bonding, allowing the bionic skin to extend and retract synchronously with the sliding of the head and tail frames. The bionic skin area corresponding to the panoramic camera is fitted with a rigid PC light-transmitting window, which is rigidly fixed to the head frame and extends and retracts synchronously with the head. The skin around the window is fixed to the outer edge of the window with pressing edges, so that when contracted, wrinkles only occur in the body area behind the window and do not obstruct the camera's field of view.

[0042] The wing is hummingbird-shaped, with its root embedded in the maximum diameter profile of the middle section of both sides of the frame. It can be driven by a flapping wing drive mechanism inside the frame to achieve biomimetic flapping flight. The highest and lowest points of the wing flapping do not exceed the original maximum height profile of the frame: since the fuselage folds only along the axial direction, the height of the maximum cross-section in the middle of the frame remains unchanged. The wing is installed at this maximum cross-section. Therefore, when the fuselage spine folds and retracts, the wing maintains its original installation position without additional storage. After folding, the overall height of the fuselage is determined only by the amount of spine contraction, and the outer edge of the wing flapping is always within the fuselage height envelope.

[0043] Two sets of panoramic cameras are installed at the upper end of the nose movable frame, a lidar is installed at the top of the movable frame, and a position sensor is installed at the top of the movable frame to provide real-time feedback on the extension and retraction position of the nose movable frame. A position sensor is embedded at the base of the tail movable frame to provide real-time feedback on the extension and retraction position of the tail movable frame.

[0044] The battery, various position sensors, speed sensors, tilt sensors, vibration sensors, panoramic cameras, and lidar transmit the collected data to the processor. After processing by the processor, control commands are sent to the front and rear axial electric actuators, driving them to extend and retract in tandem, achieving adaptive adjustment of the fuselage height, thus safely passing through narrow spaces.

[0045] Example 2

[0046] This embodiment provides a folding control system for the fuselage spine of a hummingbird bionic drone. When the system program is executed by the processor, it can realize the folding control of the hummingbird bionic drone. The system includes a sensing unit, a processor unit, an execution drive unit, and a position feedback unit, forming a closed-loop control link of "sensing-computation-execution-feedback-iteration".

[0047] The perception unit includes a binocular panoramic camera, a LiDAR, a speed sensor, a tilt sensor, a vibration sensor, and multiple sets of position sensors. It is used to collect environmental parameters in confined spaces and the UAV's own state parameters, outputting raw parameter sets to the processor unit. Specifically, it interfaces with the binocular panoramic camera to collect the horizontal distance between the left and right boundaries of the space, the axial extension length of the confined space, and the original vertical distance between the upper and lower boundaries of the space; it interfaces with the LiDAR to collect spatial point cloud data, identifying the equivalent protrusion height of obstacles and the ambient light intensity value; and it interfaces with the UAV's internal sensors to collect the original fuselage height, nose extension length, tail extension length, flight speed, fuselage pitch angle, and effective fuselage vibration frequency.

[0048] The processor unit integrates a data preprocessing module, an image analysis and data processing module, a spine folding control module, and a status monitoring and dynamic adjustment module.

[0049] Data preprocessing module: Receives the original parameter set, removes invalid data based on single-parameter statistical features and multi-parameter correlation constraints, and generates a valid parameter set.

[0050] Image analysis and data processing module: Performs dual noise reduction processing on the acquired images, corrects the space-accommodating height, and calculates the difference in body height reduction and the shrinkage distance of the spine target.

[0051] The spinal folding control module generates coordinated control commands for the front and rear push rods, receives measured data from the position feedback unit to complete the height verification after folding, performs secondary folding as needed, and calculates the error for parameter iteration.

[0052] Status monitoring and dynamic adjustment module: continuously monitors the environment and fuselage status throughout the narrow space crossing process, outputs dynamic adjustment commands, and triggers safety degradation strategies under abnormal operating conditions.

[0053] The actuator includes a front axial electric actuator and a rear axial electric actuator. It receives control commands from the processor unit and drives the head movable frame and the tail movable frame to slide along the fuselage axis, thereby realizing the retraction and extension of the fuselage spine.

[0054] The position feedback unit includes position sensor 1 distributed at the head movable frame, position sensor 2 at the front end of the frame, position sensor 3 at the rear end of the frame, and position sensor 4 at the root of the tail movable frame. It is used to collect the measured travel of the head and tail extension in real time and transmit it back to the processor unit to provide measured data support for closed-loop verification.

[0055] Example 3.

[0056] This embodiment provides a method for controlling the folding of the fuselage spine of a hummingbird-inspired bionic drone, and the specific implementation steps are as follows.

[0057] S1: Parameter Acquisition Phase

[0058] The drone collects environmental parameters in the confined space in front of it using a binocular panoramic camera and LiDAR, including the original vertical distance between the upper and lower boundaries of the space, the horizontal distance between the spaces, the axial extension length of the confined space, the distribution of obstacles, and the ambient light intensity. At the same time, the drone collects its own status parameters through built-in sensors, including the original fuselage height, the length of the nose extension, the length of the tail extension, the flight speed, the fuselage pitch angle, and the fuselage vibration frequency, thus completing the full-dimensional data collection.

[0059] S2: Data Preprocessing Stage

[0060] The system receives the original parameter set, performs outlier screening based on single-parameter statistical features and multi-parameter correlation constraint mechanisms, and generates a time-series aligned effective parameter set, with the specific rules as follows.

[0061] The sampling window is set to collect data for 10 consecutive frames. The mean and standard deviation of each parameter sequence are calculated, and outliers exceeding the mean ± 2 times the standard deviation are removed.

[0062] A transient protection mechanism is set up: when the distance to an obstacle detected by the lidar decreases by more than 20% within one frame, it is determined to be a sudden obstacle transient event. The outlier removal of this parameter is suspended, and the original data is retained for obstacle avoidance decision-making.

[0063] Verify the physical correlation constraints of multiple parameters and eliminate abnormal samples that do not conform to physical laws. This includes three rules: ① The vertical height and horizontal spacing are positively correlated, and abnormal narrow slit data with height much greater than width are excluded; ② The range of fuselage pitch angle is limited to ±30°, and samples exceeding this range are judged as sensor abnormalities; ③ The push rod stroke and fuselage height must satisfy the geometric mapping relationship ΔH=ΔL×k, and samples with a deviation of more than 10% are judged as invalid samples.

[0064] The minimum effective sample threshold is set to 60% of the sampling window. When the number of effective samples is lower than the threshold, the effective parameter set of the previous frame is used as the backoff value, and the data confidence level is marked.

[0065] S3: Image Denoising and Spatial Parameter Calculation Stage

[0066] A dual noise reduction mechanism combining Gaussian filtering and median filtering is used to eliminate Gaussian noise in the image and salt-and-pepper noise from sensor vibration. The intensity of the Gaussian filter is adaptively adjusted according to the ambient light brightness, and the median filter window is adaptively adjusted according to the vibration frequency of the machine body. Then, the space-accommodating height and shrinkage distance are calculated.

[0067] Adaptive dual noise reduction calculation The formula for Gaussian filtering is: G(x,y)=1 / (2πσ_G^2)·e^(-(x^2+y^2) / (2σ_G^2))*I(x,y) Where * represents the convolution operation, I(x,y) is the pixel value of the original image; the standard deviation of the filter σ_G = 0.8 + 0.4 × I_light, I_light is the normalized ambient light brightness value, with a value range of [0,1], and σ_G has a value range of [0.8,1.2].

[0068] The median filtering formula is: M(i,j)=Med{I(i+k,j+l)|k,l∈[-r,r]} Where Med is the median operation, the filter window radius r = round(1 + 2 × f_vib^*); f_vib^* is the normalized effective vibration frequency of the fuselage, which is obtained by dividing the measured vibration frequency by the reference vibration frequency of 10Hz, and the value range is [0,1]; r is rounded to the range of [1,3].

[0069] Calculation of the actual height that the space can accommodate Integrating multiple parameters such as the original vertical height of the space, the fuselage pitch angle, the equivalent protrusion height of obstacles, the axial length of the narrow space, and the total axial length of the fuselage, the actual accommodative height of the space corresponding to the fuselage projection envelope under tilted conditions is corrected using the following formula: H_space=H_raw×cosθ-H_obstacle-k_L×L_space-L_body×|sinθ| Where H_raw is the original vertical distance between the upper and lower boundaries of the space, θ is the pitch angle of the fuselage, H_obstacle is the equivalent protrusion height of the obstacle in the space, which is obtained by identifying the maximum vertical protrusion of the obstacle from the lidar point cloud; L_space is the axial extension length of the narrow space; k_L is the axial length safety redundancy coefficient, which is 10mm / m, representing a 10mm height safety margin reserved for every 1m of axial extension; L_body is the total axial length of the UAV fuselage, and L_body×|sinθ| is the vertical projection increment of the fuselage when flying at an angle, which is used to correct the occupancy of the fuselage envelope on the passage height.

[0070] Target contraction distance calculation First, define the target safe passage height: H_target=H_space-ΔH_safe×(1+0.05V) Where ΔH_safe is the basic safety redundancy height, with a value of 5mm, and V is the UAV flight speed in m / s; the higher the flight speed, the greater the safety redundancy, and the more stringent the corresponding space safety passage threshold.

[0071] The formula for calculating the required reduction in fuselage height is: ΔH=max(0,H_original-H_target) H_original represents the original height of the drone. When the original height already meets the passage requirements, the height difference reduction is 0, and there is no need to perform folding.

[0072] The formula for calculating the total contraction distance of the spinal target is: ΔL_total=ΔH / k Where k is the spinal contraction height conversion coefficient, k=tanα, and α is the angle between the frame sidewall and the fuselage axis; the total contraction distance is shared by the front axial electric actuator and the rear axial electric actuator in a 1:1 ratio.

[0073] S4: Spinal Folding Control Stage

[0074] Based on the calculated total retraction distance of the spinal target, the front and rear axial electric actuators are controlled to move in tandem, driving the head and tail movable frames to retract synchronously towards the center along the fuselage axis, thereby achieving spinal axial retraction of the fuselage and reducing the overall height of the fuselage to fit the narrow space.

[0075] After retraction is complete, the measured strokes of the front push rod L_f_measured and the rear push rod L_r_measured are collected by the position sensor to calculate the actual height of the fuselage after retraction.

[0076] H_body_measured^'=H_original-(L_f_measured+L_r_measured)×k The actual height of the fuselage is compared with the target's safe passage height. The condition for passing through narrow spaces is: H_body_measured^'≤H_target.

[0077] If the conditions are met, the folding is deemed qualified, and the accuracy error of this folding is calculated.

[0078] If the conditions are not met, perform a second shrinkage correction, and follow the termination rules below: The maximum number of corrections is N_max=3. The contraction will terminate if the passage condition is still not met after 3 consecutive secondary contractions. Stroke saturation determination: When the total stroke of the front and rear push rods reaches 60mm of maximum stroke, it is determined that the stroke is saturated and the retraction will stop. Infeasible Retreat Branch: When the travel is saturated and the passage conditions are still not met, or the space height is less than the minimum folding height of the fuselage, the system triggers the safe retreat logic: maintain the current folding state, decelerate and hover, then reverse to exit the narrow space, and at the same time report the space as impassable.

[0079] The formula for calculating folding accuracy error is: E_fold=|H_body_measured^'-H_target|×(1+0.1×f_vib+0.1×I_light) The folding accuracy error is stored in the parameter sample library according to the corresponding working conditions. The safety redundancy coefficient and conversion coefficient are periodically corrected by fitting the least squares method to achieve closed-loop parameter iterative optimization.

[0080] S5: Status Monitoring and Dynamic Adjustment Phase

[0081] The drone updates environmental and flight status parameters in real time throughout its passage through narrow spaces. A segmented state machine mechanism dynamically adjusts the folding and unfolding of its spine, including altitude margin hysteresis determination and channel unlocking conditions. The specific adjustment formula is as follows: Among them, ΔL_adjust is positive for spinal extension and negative for further spinal contraction; ΔH_hyst is the height hysteresis, with a value of 3mm, to avoid frequent state switching; L_exit is the axial length threshold for exit determination, with a value of 0.2m.

[0082] The adjustment factors are defined as follows: The speed adjustment factor K_V = 1 / (1+0.2V), and the faster the flight speed, the smaller the adjustment range. The vibration adjustment factor K_f = 1 / (1+0.3f_vib), the higher the vibration frequency, the smaller the adjustment range; The horizontal spacing adjustment factor K_D = D_horizontal / D_0, where D_horizontal is the spatial horizontal spacing and D_0 is the baseline horizontal spacing. The value is the width of the drone's own body. The narrower the space, the smaller the adjustment range. The axial length adjustment factor K_L = L_0 / L_space, where L_0 is the baseline axial length, with a value of 1m. The longer the axial length, the smaller the adjustment range.

[0083] When the axial length of a confined space exceeds 2 meters and the horizontal spacing is less than 1.2 times the fuselage width, the system enters a locked state, prohibiting active deployment to avoid frequent adjustments that could lead to attitude instability. After passing through the exit and detecting an open external space with an axial length less than the exit threshold, the system gradually unlocks and deploys the fuselage, resuming normal flight.

[0084] Folding failure safety degradation strategy: If a fault such as stick jamming or no feedback from the position sensor is detected during flight, the system immediately enters safety degradation mode: locks the current stick position, disables all folding dynamic adjustment functions, decelerates and hovers, and triggers a return-to-home command to avoid forced passage and collision.

[0085] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for controlling the folding of the fuselage spine of a hummingbird-inspired bionic drone, characterized in that, Includes the following steps: S1: Parameter acquisition phase. The original vertical distance between the upper and lower boundaries of the narrow space in front of the drone, the horizontal distance between the upper and lower boundaries, the axial extension length of the narrow space, and the environmental parameters of the obstacle distribution are collected by binocular panoramic camera and lidar. The original fuselage height, nose extension length, tail extension length, flight speed, fuselage pitch angle, and fuselage vibration frequency are collected by the built-in sensors of the drone to complete the full-dimensional parameter acquisition. S2: In the data preprocessing stage, the average value and standard deviation of each parameter collection sequence are calculated within the set sampling window, and invalid outlier data exceeding the mean ± 2 times the standard deviation are removed; at the same time, the physical correlation constraints between parameters are verified, and abnormal samples that do not conform to physical laws are removed; when the effective sample size is lower than the minimum threshold, the effective parameter set of the previous frame is used as the backoff value to generate a time-aligned and reliable effective parameter set. S3: In the image denoising and spatial parameter calculation stage, a dual denoising mechanism combining Gaussian filtering and median filtering is used to eliminate Gaussian noise in the image and salt-and-pepper noise from sensor vibration. The intensity of the Gaussian filter is adaptively adjusted according to the ambient light brightness, and the median filter window is adaptively adjusted according to the fuselage vibration frequency. Multiple parameters such as the original vertical height of the space, the fuselage pitch angle, the equivalent protrusion height of obstacles, the axial length of the narrow space, and the total axial length of the fuselage are integrated to correct the actual accommodating height of the space corresponding to the fuselage projection envelope under tilted conditions. The required reduction height difference of the fuselage is calculated by combining the flight speed and the original fuselage height. Finally, based on the cone-shaped spine shrinkage height conversion coefficient, the total shrinkage distance of the fuselage spine target is calculated. S4: In the spinal folding control stage, based on the calculated total retraction distance of the spinal target, the front and rear axial electric actuators are controlled to move in tandem, driving the head and tail movable frames to retract synchronously towards the center along the fuselage axis, thereby achieving spinal axial retraction of the fuselage and reducing the overall height of the fuselage to fit the narrow space. After retraction, the actual height of the fuselage is calculated by measuring the head and tail strokes using position sensors and compared with the target safe passage threshold. If the passage requirements are not met and the maximum number of corrections has not been reached or the stroke saturation has not been reached, a second retraction is performed, and the folding accuracy error is calculated and stored in the parameter iteration queue. The core coefficients are periodically corrected by fitting using the least squares method to complete the closed-loop verification of the folding effect and parameter iteration optimization. S5: Status monitoring and dynamic adjustment phase. Throughout the entire process of the UAV traversing narrow spaces, environmental space parameters and fuselage flight status parameters are updated in real time. The folding and unfolding amplitude of the spine is dynamically adjusted in combination with flight speed, fuselage vibration frequency, horizontal spatial distance, and axial spatial length. When the flight speed is high and the fuselage vibration is severe, the folding adjustment amplitude is reduced. When the space is horizontally narrow and the axial length is long, the folding state is locked and arbitrary unfolding is prohibited. After passing through the exit, the unfolding state is gradually restored when the unlocking conditions are met, ensuring the stability of narrow space traversal.

2. The method for controlling the folding of the fuselage spine of a hummingbird-inspired unmanned aerial vehicle according to claim 1, characterized in that, In step S3, the Gaussian filter calculation formula is as follows: G(x,y)=1 / (2πσ_G^2)·e^(-(x^2+y^2) / (2σ_G^2))*I(x,y) Where * represents the convolution operation, I(x,y) is the original image pixel value, the filter standard deviation σ_G=0.8+0.4×I_light, I_light is the normalized ambient light brightness value, with a value range of [0,1], and σ_G has a value range of [0.8,1.2]. The median filtering formula is as follows: M(i,j)=Med{I(i+k,j+l)|k,l∈[-r,r]} Where Med is the median operation, the filter window radius r = round(1 + 2 × f_vib^*); f_vib^* is the normalized effective vibration frequency of the fuselage, which is obtained by dividing the measured vibration frequency by the reference vibration frequency of 10Hz, and the value range is [0,1]; r is rounded to the range of [1,3].

3. The method for controlling the folding of the fuselage spine of a hummingbird-inspired unmanned aerial vehicle according to claim 1, characterized in that, In step S3, the formula for calculating the actual accommodating height of the space is as follows: H_space=H_raw×cosθ-H_obstacle-k_L×L_space-L_body×|sinθ| Where H_raw is the original vertical distance between the upper and lower boundaries of the space, θ is the pitch angle of the fuselage, H_obstacle is the equivalent protrusion height of the obstacle in the space, which is obtained by identifying the maximum vertical protrusion of the obstacle from the lidar point cloud; L_space is the axial extension length of the narrow space; k_L is the axial length safety redundancy coefficient, which is 10mm / m, representing a 10mm height safety margin reserved for every 1m of axial extension; L_body is the total axial length of the UAV fuselage, and L_body×|sinθ| is the vertical projection increment of the fuselage when flying at an angle, which is used to correct the occupancy of the fuselage envelope on the passage height.

4. The method for controlling the folding of the fuselage spine of a hummingbird-inspired unmanned aerial vehicle according to claim 3, characterized in that, In step S3, first define the target safe passage height: H_target=H_space-ΔH_safe×(1+0.05V) Where ΔH_safe is the basic safety redundancy height, with a value of 5mm, and V is the UAV flight speed in m / s; The formula for calculating the required reduction in fuselage height is: ΔH=max(0,H_original-H_target) H_original represents the original height of the drone. When the original height already meets the passage requirements, the height difference reduction is 0, and folding is not required. The formula for calculating the total contraction distance of the spinal target is: ΔL_total=ΔH / k Where k is the spinal contraction height conversion coefficient, k=tanα, and α is the angle between the frame sidewall and the fuselage axis; the total contraction distance is shared by the front electric actuator and the rear electric actuator in a 1:1 ratio.

5. The method for controlling the folding of the fuselage spine of a hummingbird-inspired unmanned aerial vehicle according to claim 4, characterized in that, In step S4, after retraction, the front push rod travel L_f_measured and the rear push rod travel L_r_measured are measured using position sensors to calculate the actual height of the computer body. H_body_measured^'=H_original-(L_f_measured+L_r_measured)×k The criteria for determining passage through narrow spaces are: H_body_measured^'≤H_target The formula for calculating folding accuracy error is: E_fold=|H_body_measured^'-H_target|×(1+0.1×f_vib+0.1×I_light) The folding accuracy error is stored in the parameter sample library according to the corresponding working conditions. The safety redundancy coefficient and conversion coefficient are periodically corrected by fitting using the least squares method to achieve closed-loop parameter iterative optimization.

6. The method for controlling the folding of the fuselage spine of a hummingbird-inspired unmanned aerial vehicle according to claim 5, characterized in that, In step S5, the formula for dynamically adjusting the spinal folding range is: Among them, ΔL_adjust indicates positive spinal extension and negative spinal contraction; ΔH_hyst is the height hysteresis, with a value of 3mm, to avoid frequent state switching. When the axial length of a narrow space is greater than 2m and the horizontal spacing is less than 1.2 times the fuselage width, the fuselage is locked in a folded state and cannot be actively unfolded; when passing through an exit and the axial length of the narrow space is less than L_exit=0.2m, the fuselage is gradually unlocked and unfolded. The speed adjustment factor K_V = 1 / (1+0.2V), and the faster the flight speed, the smaller the adjustment range. The vibration adjustment factor K_f = 1 / (1+0.3f_vib), the higher the vibration frequency, the smaller the adjustment range; The horizontal spacing adjustment factor K_D = D_horizontal / D_0, where D_horizontal is the spatial horizontal spacing and D_0 is the baseline horizontal spacing. The value is the width of the drone's own body. The narrower the space, the smaller the adjustment range. The axial length adjustment factor K_L = L_0 / L_space, where L_0 is the baseline axial length, with a value of 1m. The longer the axial length, the smaller the adjustment range.

7. A fuselage spine folding control system for a hummingbird-inspired bionic unmanned aerial vehicle, used to execute the fuselage spine folding control method according to any one of claims 1 to 6, characterized in that, It includes a sensing unit, a processor unit, an execution drive unit, and a position feedback unit; The sensing unit includes a binocular panoramic camera, a lidar, a speed sensor, a tilt sensor, a vibration sensor, and multiple sets of position sensors, used to collect environmental parameters in narrow spaces and the UAV's own state parameters, and output the raw parameter set to the processor unit. The processor unit has built-in data preprocessing module, image analysis and data processing module, spine folding control module, and status monitoring and dynamic adjustment module, which are used to complete abnormal data screening, image adaptive noise reduction, spatial accommodating height calculation, folding instruction generation and dynamic adjustment decision-making throughout the traversal. The execution drive unit includes a front axial electric actuator and a rear axial electric actuator. It receives folding control commands from the processor unit and drives the head movable frame and the tail movable frame to slide along the fuselage axis, thereby realizing the retraction and unfolding of the fuselage spine. The position feedback unit includes multiple sets of position sensors distributed on the head movable frame, the front and rear ends of the frame, and the tail movable frame. These sensors are used to collect the measured travel of the head and tail extension in real time and transmit it back to the processor unit, forming a closed-loop verification link for the folding effect.

8. A hummingbird-inspired unmanned aerial vehicle, characterized in that, The fuselage spine folding control system described in claim 7 includes a frame, bionic skin, wings, tail, nose movable frame, and tail movable frame; The frame is a titanium alloy honeycomb sandwich spine main load-bearing structure, which is conical in shape. The angle between the side wall and the axis of the body is 30°~45°, preferably 38.66°, corresponding to the spine contraction height conversion coefficient k=tanα≈0.

8. The battery is installed at the bottom of the frame. A front electric push rod is set on the inner side of the front end of the frame, and a rear electric push rod is symmetrically set on the inner side of the rear end. Position sensor 2 is integrated at the front end of the top of the frame, and position sensor 3 is integrated at the rear end of the top. A speed sensor, tilt sensor, vibration sensor and processor are integrated in the middle of the frame. The sensors and processor constitute a distributed sensing network. The movable frame of the machine head is assembled at the front end of the frame. The top of the movable frame is equipped with two sets of panoramic cameras, a lidar and a position sensor. The output end of the front electric push rod is fixedly connected to the lower end of the movable frame, which can drive the movable frame to slide backward along the axial slide rail of the frame. The tail movable frame is located at the rear end of the frame, and the tail fin is fixedly installed at the end of the tail movable frame; four position sensors are embedded in the root of the tail movable frame, and the output end of the rear electric push rod is fixedly connected to the root of the tail movable frame, which can drive the tail movable frame to slide forward along the frame axis and retract, thereby driving the tail fin to move synchronously. The bionic skin is made of a highly transparent elastomer, which fully covers the surface of the frame. Both ends are fixed to the outer edges of the head frame and the tail frame, respectively, and can deform synchronously with the body's extension and retraction. The bionic skin in the area corresponding to the camera is fitted with a rigid light-transmitting window, which is rigidly fixed to the head frame. When the camera is retracted, there are no wrinkles that obstruct the camera's data acquisition. The wing has a hummingbird-shaped structure, with its root embedded in the maximum outline of the middle of both sides of the frame, and is driven to flap by an internal drive mechanism of the frame; after the fuselage is folded and stored, the maximum outer edge of the flapping wing does not exceed the maximum height outline of the fuselage.

9. The hummingbird-inspired bionic drone according to claim 8, characterized in that, The front and rear electric actuators are miniature DC linear electric actuators with a maximum stroke of 30mm, a minimum stroke of 0mm, a movement speed of 2mm / s, and a stroke control accuracy of ±0.1mm. The bionic skin is made of transparent silicone or polydimethylsiloxane material, with a visible light transmittance ≥90%, an elongation at break ≥300%, and a thickness of 0.5mm~1.0mm. It can be repeatedly stretched and contracted without permanent deformation.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements all the steps of the fuselage spine folding control method of the hummingbird bionic UAV as described in any one of claims 1 to 6.