Air obstacle avoidance method, system, medium and device based on multi-machine vibration mutual inspection

Abstract

The present application relates to the technical field of unmanned aerial vehicle obstacle avoidance, and provides an aerial obstacle avoidance method, system, medium and equipment based on multi-aircraft vibration mutual detection, comprising: for the vibration data of each unmanned aerial vehicle, extracting time domain features, frequency domain features and time-frequency domain features, and after splicing, obtaining a merged feature vector; for each unmanned aerial vehicle, encapsulating the merged feature vector as a feature packet and periodically broadcasting, and after receiving the feature packet of other unmanned aerial vehicles, unpacking, splicing the merged feature vectors of all unmanned aerial vehicles in the same time window to obtain a multi-aircraft vibration feature set; based on the multi-aircraft vibration feature set, calculating the vibration difference between each pair of unmanned aerial vehicles to obtain a vibration mutual detection matrix; based on the vibration mutual detection matrix, identifying the air disturbance type and disturbance dominant direction, and generating an expected position and attitude adjustment amount to control the unmanned aerial vehicle to avoid obstacles. The anti-interference capability in a complex flight environment is significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) obstacle avoidance technology, and particularly relates to aerial obstacle avoidance methods, systems, media and equipment based on multi-aircraft vibration mutual detection. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Existing drone obstacle avoidance technologies almost entirely rely on sensing methods capable of perceiving "hard obstacles," such as vision, depth sensors, lidar, or millimeter-wave radar. However, for "soft obstacles"—those that cannot form clear reflections or present identifiable shapes in visual images—such as strong crosswinds, wake vortices, rising thermals, valley turbulence, low-density smoke disturbances, building airflow deflection, and strong wind shearing—traditional technologies completely lack effective detection capabilities. In these scenarios, drone flight stability depends on aerodynamic equilibrium, and airflow disturbances can directly trigger abrupt attitude changes, trajectory deviations, lift transients, and even cascading instability. However, existing technologies cannot identify such invisible airflow risks in advance, resulting in low safety and poor mission reliability for multi-drone swarm flights.

[0004] At the same time, although a single UAV can collect vibration and attitude changes through its own inertial measurement unit (IMU), due to factors such as self-vibration noise, propeller coupling, and structural resonance, the single-unit IMU cannot distinguish between "self-noise" and "external air disturbances", and cannot determine the direction, range, and propagation trend of the disturbances. Therefore, it cannot rely on single-unit inertial measurement to achieve effective soft obstacle identification.

[0005] In multi-drone scenarios, traditional collaborative methods only share macroscopic information such as position, velocity, and attitude, lacking a collaborative observation mechanism for "microscopic vibration behavior." This prevents multiple drones from mutually verifying airflow conditions and from inferring disturbance sources through collective information, thus hindering the construction of a spatial distribution model of environmental airflow. The lack of technology for sharing micro-vibration characteristics among multiple drones prevents the group from identifying air disturbances through "collective perception," representing a significant gap in existing technology.

[0006] In summary, existing UAV aerial obstacle avoidance technologies suffer from the following technical problems: the inability of UAVs to identify "soft obstacles" such as air disturbances; the inability of a single IMU to distinguish between airframe vibrations and external disturbances; the lack of a collaborative observation mechanism for airflow disturbances among multiple UAVs; the inability to predict the trends of disturbances such as wakes, crosswinds, and hot air masses in advance; and the high degree of dispersion and severe coupling of multi-UAV formations in complex airflows. Summary of the Invention

[0007] To address the technical problems mentioned above, this invention provides an aerial obstacle avoidance method, system, medium, and device based on multi-aircraft vibration mutual inspection. This method upgrades disturbance judgment from single-point observation to multi-point spatial distribution observation. By eliminating the influence of body structure vibration, propeller coupling, and other body noise through the vibration mutual inspection matrix, the identification results of air disturbances are more stable and reliable, significantly improving the anti-interference capability in complex flight environments.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, comprising: Obtain vibration data for each drone in the formation; For the vibration data of each drone, time-domain features, frequency-domain features, and time-frequency-domain features are extracted, and then concatenated to obtain a merged feature vector; For each UAV, its own merged feature vector is encapsulated into a feature message and broadcast periodically. After receiving feature messages from other UAVs, the message is unpacked. The merged feature vectors of all UAVs within the same time window are spliced ​​together to obtain a multi-UAV vibration feature set. Based on the multi-machine vibration feature set, the vibration difference between each pair of UAVs is calculated to obtain the vibration mutual inspection matrix; Based on the vibration mutual inspection matrix, after identifying the type and dominant direction of air disturbance, the desired position and attitude adjustment amount is generated to control the UAV for obstacle avoidance.

[0009] Furthermore, it also includes: before extracting time-domain features, frequency-domain features, and time-frequency-domain features, preprocessing the vibration data of each UAV, the preprocessing including DC removal, filtering by a digital bandpass filter, filtering by a notch filter, smoothing, and data normalization.

[0010] Furthermore, the UAV vibration data includes three-axis acceleration and three-axis angular velocity; The time-domain features include: root mean square value, mean, variance, and peak value; The frequency domain features include: spectrum, spectral power, high-frequency energy ratio, and spectral centroid; The time-frequency domain feature is multi-scale wavelet energy.

[0011] Furthermore, the feature message includes its own merged feature vector, local unique identifier, and local timestamp.

[0012] Furthermore, the vibration difference between the two drones is ;in, Let α and β be the 2-norm, and α and β be the weighting coefficients. The i-th and j-th UAVs at time... The merged feature vectors are respectively and .

[0013] Furthermore, the steps for identifying the type of air disturbance and the dominant direction of the disturbance include: Calculate the average difference value of each row in the vibration cross-check matrix. Calculate the overall average difference Where N is the total number of drones in the formation. The vibration difference between two drones; When there are fewer than the set number of drones Satisfy the average difference value Significantly greater than And drones When located upstream or in front of the formation, it is identified as a drone. Located in a region with a strong wake; when Spatial Index When the gradient is monotonically increasing or decreasing and is stable as positive or negative in a certain direction, it is determined to be a unilateral wind or wind shear phenomenon, and the wind direction is determined by the gradient sign. Calculate the variance of the differences in the vibration cross-check matrix. Among them, the mean of the difference values ,when and When the vibration mutual inspection matrix is ​​increased and there are no gradient structures or local isolated high values, the formation as a whole is determined to be in the turbulent region. Constructing a vertical energy index using vertical vibration characteristics When all If the value is higher than the historical baseline and the difference value is evenly distributed in the vibration mutual inspection matrix, the current disturbance type is determined to be a thermal rising air mass.

[0014] Furthermore, in the process of generating the desired position and attitude adjustment amounts, a control cost function is introduced to constrain the obstacle avoidance offset and energy consumption.

[0015] A second aspect of the present invention provides an aerial obstacle avoidance system based on multi-aircraft vibration mutual detection, comprising: The data acquisition module is configured to acquire vibration data of each drone in the formation. The feature extraction module is configured to extract time-domain features, frequency-domain features, and time-frequency-domain features from the vibration data of each UAV, and then concatenate them to obtain a merged feature vector. The interaction module is configured to: for each UAV, encapsulate its own merged feature vector into a feature message and broadcast it periodically, and receive feature messages from other UAVs and unpack them, and then concatenate the merged feature vectors of all UAVs within the same time window to obtain a multi-UAV vibration feature set. The vibration mutual inspection module is configured to: calculate the vibration difference between each pair of UAVs based on the multi-aircraft vibration feature set, and obtain the vibration mutual inspection matrix. The obstacle avoidance control module is configured to: based on the vibration mutual detection matrix, identify the type of air disturbance and the dominant direction of the disturbance, generate the desired position and attitude adjustment amount, and perform obstacle avoidance control on the UAV.

[0016] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for aerial obstacle avoidance based on multi-aircraft vibration mutual detection.

[0017] A fourth aspect of the present invention provides a computer device including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method for aerial obstacle avoidance based on multi-aircraft vibration mutual detection.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention solves the problem that a single IMU cannot distinguish between its own noise and external disturbances through collaborative analysis of vibration characteristics from multiple aircraft. It elevates disturbance identification from single-point observation to multi-point spatially distributed observation, and eliminates the influence of body noise such as structural vibration and propeller coupling through a vibration cross-checking matrix. This makes the identification results of air disturbances more stable and reliable, significantly improving the anti-interference capability in complex flight environments.

[0019] This invention enables advance perception of air disturbances, providing predictive obstacle avoidance capabilities. By analyzing the temporal and spatial evolution characteristics of the vibration cross-check matrix, predictive identification can be completed before the disturbance causes a sudden change in attitude, providing the UAV with an advance response time of approximately 0.3 to 1.0 seconds. This transforms obstacle avoidance control from passive correction to active avoidance, significantly reducing the risk of flight instability.

[0020] This invention establishes an air disturbance type discrimination mechanism based on a vibration mutual inspection matrix structure. By comprehensively analyzing the matrix mean distribution, gradient change, and variance of differences, it can distinguish different disturbance types such as wake, crosswind, wind shear, overall turbulence, and rising thermal air mass, and determine the dominant direction of the disturbance, providing a clear basis for subsequent control decisions.

[0021] This invention proposes an optimal obstacle avoidance control strategy based on risk distribution. By mapping the vibration mutual inspection matrix to a spatial risk distribution and introducing a cost function to constrain obstacle avoidance offset and energy consumption, the minimum necessary control actions are generated under the premise of ensuring safety. This allows the UAV to escape the disturbance area with minimal attitude adjustment and trajectory deviation, avoiding excessive interference to the mission route and formation structure.

[0022] This invention achieves adaptive aerodynamic reconfiguration of multi-UAV formation structures. Based on the disturbance strength distribution reflected by the vibration cross-check matrix, the spacing and altitude distribution between UAVs can be automatically adjusted, transforming the formation from a highly coupled structure to a disturbance-resistant structure. This reduces the risk of wake superposition and aerodynamic resonance, significantly improving the overall stability of multi-UAV formations in complex wind fields.

[0023] This invention requires no additional sensors, offering advantages in terms of low cost and low power consumption. It utilizes only existing IMU vibration data from the drone to achieve soft obstacle perception and avoidance control, without relying on hardware such as vision, lidar, or millimeter-wave radar. It has low computational load and communication overhead, and can be directly deployed on existing drone platforms, demonstrating excellent engineering feasibility and scalability.

[0024] This invention maintains stable and reliable obstacle avoidance capabilities even under visual perception failure conditions. In scenarios where visual perception is compromised, such as at night, in backlight, dense fog, smoke, or against textureless backgrounds, the device can still accurately identify air disturbances and execute obstacle avoidance control through a vibration mutual detection mechanism, significantly improving the all-weather flight safety of UAVs in complex environments. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 This is a flowchart of an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0028] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] Example 1 This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection.

[0030] This embodiment provides an aerial obstacle avoidance method based on multi-drone vibration mutual detection. It utilizes the vibration mutual detection of multiple UAVs to realize air disturbance perception and autonomous obstacle avoidance, so as to solve the technical problems of unidentifiable soft obstacles, undeterminable disturbance direction, and easy instability of formation in wind fields.

[0031] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which solves the technical defect that UAVs cannot identify "soft obstacles" such as air disturbances, enabling UAVs to still have reliable obstacle avoidance capabilities in visually ineffective scenarios such as dense fog, smoke, night, and backlight.

[0032] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which solves the problem that a single IMU cannot distinguish between airframe vibration and external disturbances, so that the IMU vibration signal can be used for airflow state judgment, rather than just for attitude control.

[0033] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which solves the problem of the lack of a collaborative observation mechanism for airflow disturbances among multiple aircraft. By using a vibration mutual detection matrix, multiple aircraft can share micro-vibration changes, thereby achieving "collective intelligent perception".

[0034] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which solves the problem of being unable to predict the trends of disturbances such as wake turbulence, crosswinds, and hot air masses in advance, enabling UAVs to complete predictive obstacle avoidance 0.3–1.0 seconds before the disturbance causes attitude instability.

[0035] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which solves the problem that multi-aircraft formations are highly prone to dispersion and severe coupling in complex airflow. By coordinating vibration information, the formation can automatically reconstruct its formation, adjust its spacing, and form a windproof structure.

[0036] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which enables UAVs to accurately identify invisible air disturbances, predict flight risks, and perform intelligent obstacle avoidance without additional hardware sensors, significantly improving the operational safety of UAVs in real complex environments.

[0037] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, such as... Figure 1 As shown, it includes the following steps: Step 1: Acquisition and preprocessing of UAV vibration data.

[0038] Step 101: Install an inertial measurement unit (IMU) on each drone participating in the cooperative flight. The IMU is sampled at a fixed frequency. Collect vibration-related data of the machine body, including triaxial acceleration. , , With triaxial angular velocity , , Where t is the discrete-time index, and the sampling frequency satisfies ≥200.

[0039] Step 102: Preprocess the raw IMU signal directly to eliminate interference components unrelated to air disturbances.

[0040] The preprocessing steps include: First, a DC removal operation is performed on each axial signal. Let the original signal be... The signal mean is Where T is the total number of sampling points within the time window, which can be taken as 50, to calculate the signal after DC removal. This is used to eliminate the effects of gravity components and static bias. Secondly, for Filtering is performed using a digital bandpass filter, with the bandpass cutoff frequency set to [value missing]. ,satisfy This suppresses slow low-frequency attitude changes and broadband noise interference above the target frequency band. Secondly, for motors and propellers with inherent frequencies In this case, based on bandpass filtering and the natural frequencies of the motor and propeller... Construct notch filters using several integer multiples thereof to explicitly suppress the energy in this frequency band, thus obtaining the net vibration signal. ; Subsequently, the net vibration signal Smoothing is performed using a sliding time window of fixed length L. The output signal within the window is defined as follows: This smoothing operation reduces transient random noise. Finally, data normalization is performed.

[0041] After the above processing, a net vibration sequence is obtained for subsequent feature extraction. , , , , , This sequence retains the dynamic changes caused by air disturbances over time, while significantly reducing the interference components from the body structure and drive components.

[0042] Step 2: Extraction and construction of multi-scale vibration feature vectors.

[0043] Step 201: Temporal Feature Extraction. Multi-scale feature extraction is performed on the net vibration sequence of each UAV locally to construct a vibration feature vector characterizing the current airflow disturbance state.

[0044] For the vertical acceleration within the time window [t, t+L-1] , Calculate the root mean square value: ; And calculate the variance within that time window: mean ; And kurtosis: ; Used to characterize vibrational energy and impulse characteristics.

[0045] For other axial acceleration and angular velocity signals, the mean, variance, RMS, kurtosis and other time-domain statistical characteristics are calculated in the same way.

[0046] As another implementation method, empirical mode decomposition (EMD) or variational mode decomposition (VMD) is used to separate different intrinsic modes of the vibration signal, and the instantaneous energy, instantaneous frequency and statistics of each mode are used as vibration characteristics.

[0047] Step 202, Regarding frequency domain characteristics: Performing a Fast Fourier Transform of length L on the net vibration sequence within a window yields the spectrum: , j represents the imaginary unit. This indicates the output signal within the nth time window; Calculate spectral power ; And define the high-frequency energy ratio ;in, To define the high-frequency threshold, the proportion of high-frequency perturbation components is characterized by HER. k This represents the actual frequency value corresponding to the k-th frequency component; Simultaneously, calculate the spectral centroid: ; Used to reflect the frequency band where the main vibration energy is concentrated.

[0048] As another implementation, time-frequency features are constructed using the Hilbert-Huang spectrum, and the Hilbert amplitude envelope is directly used as the perturbation-sensitive feature.

[0049] Step 203, Regarding time-frequency characteristics: The net vibration signal is analyzed using discrete wavelet transform. Decompose the wavelet decomposition into detail coefficients and approximation coefficients at several scales. Let the number of wavelet decomposition levels be J. For each level of detail coefficients... Calculate wavelet energy: ; This yields the multi-scale energy vector, i.e., the wavelet multi-scale eigenvector. It is used to distinguish the energy distribution patterns of different types of disturbances on the frequency scale.

[0050] As another implementation method, the energy from multiple frequency bands is accumulated to construct a disturbance energy vector: ,in, It is the integral energy of a fixed frequency band, used to replace the continuous spectrum characteristics.

[0051] Step 204: By combining the time-domain, frequency-domain, and time-frequency-domain features described above, the features are concatenated in a fixed order to form a vibration feature vector of dimension D, i.e., the feature vectors are merged. Where i is the drone number, D is the preset fixed dimension, and x d Let d be the d-th vibration feature component. To adapt to the wireless communication bandwidth, quantization encoding is performed on the feature vector, mapping the floating-point feature to a fixed-point representation. Then, through linear transformation and the selected encoding scheme, the feature vector is converted into a fixed-length binary data frame, which serves as the basic data structure for subsequent multi-machine sharing.

[0052] Step 3: Sharing and time synchronization of vibration characteristics of multiple UAVs.

[0053] A dedicated wireless communication link is established within the multi-drone formation, and all drones perform periodic broadcasting and receiving of vibration characteristics through a unified protocol.

[0054] (1) Packaging feature frames: After each UAV completes the vibration feature extraction within the current time window, it packages its own feature vector. Together with the local unique identifier IDi and the local timestamp Ti, they are encapsulated into a feature message. The fixed fields in the message format include the ID field, the timestamp field, and the feature data field.

[0055] (2) Periodically broadcast and receive feature frames from other UAVs: The UAV sends feature messages at a fixed period ΔT. Other UAVs unpack the messages at the receiving end, write the features into the corresponding buffer slot according to IDi, and align them with the current time of the local machine according to the timestamp Ti.

[0056] (3) Timestamp alignment: A periodic time synchronization mechanism is adopted to adjust the local clock of each UAV by broadcasting synchronization messages within a preset time interval, so that the time error of different UAVs is kept within a predetermined threshold, thereby ensuring the matching degree of the feature sampling window on the time axis.

[0057] (4) Update the feature set: After the above process, at any time Each drone holds a set of features from all drones within the same time window. Where N is the total number of drones in the formation; this set provides complete input for constructing the vibration mutual inspection matrix.

[0058] Step 4: Construction of the vibration mutual inspection matrix for multiple UAVs.

[0059] Based on the multi-machine vibration feature set obtained in step 3 A multi-UAV vibration cross-check matrix is ​​constructed to quantify the relative differences in airflow disturbance environments in which different UAVs operate.

[0060] Step 401: Let the i-th and j-th UAVs be at time... The vibration characteristics are respectively and Define the vibration difference function: ; in, The L2 norm is used, and α and β are weighting coefficients that act on the Euclidean distance term and the cosine distance term, respectively. This combined metric simultaneously characterizes the differences in the magnitude and directional angle of feature differences within the same framework.

[0061] Step 402: For all i,j∈{1,2,…,N}, calculate the corresponding dissimilarity. And construct a vibration mutual inspection matrix: ; Among them, diagonal elements Defined as 0.

[0062] matrix This spatially reflects the similarities and differences in vibration modes among various UAVs. Over time... The advancement of vibration mutual inspection matrix sequence Describe the propagation and evolution characteristics of air disturbances in formation.

[0063] Step 5: Identification of air disturbance type and direction based on vibration cross-check matrix.

[0064] By analyzing the vibration mutual inspection matrix Structural analysis identifies the types of air disturbances in the current environment and the direction of their effects.

[0065] For ease of explanation, assume that the order of the drones in physical space corresponds one-to-one with the indices 1, ..., N.

[0066] Step 501: Calculate the average difference value of each row in the vibration cross-check matrix: And calculate the overall average difference. .

[0067] Step 502: When one or a few indexes exist satisfy Significantly greater than And the drone When the drone is located upstream or in front of the formation, this phenomenon is identified as the appearance of the core of the wake disturbance in front, indicating that the drone... It is located in a region with a strong wake.

[0068] Step 503, when Spatial Index Monotonically increasing or decreasing, and with a gradient When the gradient is consistently positive or negative in a certain direction, the pattern is identified as a unilateral wind or wind shear phenomenon, and the wind direction is determined by the gradient sign. Indicates the first A drone at all times The average vibration difference value, This indicates the spatial distance between two adjacent drones along the main direction of the formation.

[0069] Step 504: For the overall turbulent flow case, analyze the variance of the difference values ​​in the vibration cross-check matrix: ; Among them, the mean of the difference values .

[0070] when and When the vibration mutual inspection matrix increases and there is no obvious gradient structure or local isolated high value, the formation as a whole is determined to be in the turbulent region.

[0071] The following equation indicates that there is no obvious spatial gradient structure in the vibration mutual detection matrix: In the formula Spatial gradient threshold; The following equation indicates that there are no locally isolated nodes with high dissimilarity values. In the formula Local peak threshold; Under typical experimental conditions of small multi-rotor UAV formations (formation spacing of 1–3 m, IMU sampling frequency of not less than 200 Hz), empirically, it is advisable to take... = =2.5, which effectively suppresses misjudgments caused by random turbulence while ensuring sensitivity to crosswind and wake disturbances.

[0072] Step 505: For thermally rising air masses, construct a vertical energy index using vertical vibration characteristics. Calculate its average value The calculation method is as follows: Let the first Drones in the time window The net vibration sequence of vertical acceleration within is as follows: (The result after DC removal, filtering, and smoothing) is then defined as follows: Single-unit vertical energy index ; Formation average vertical energy ; Where L is the length of the sliding time window, which is set manually.

[0073] When all If the value is significantly higher than the historical baseline and the difference values ​​in the cross-check matrix are relatively evenly distributed, the current disturbance type is determined to be a thermal rising air mass.

[0074] The method for determining whether the difference values ​​in the mutual inspection matrix are relatively evenly distributed is as follows: First calculate the mean with standard deviation : ; ; Subsequently, the uniformity index was calculated. : ; like If the difference values ​​are relatively evenly distributed, then the threshold value is considered to be the most uniform. Take 0.2.

[0075] Based on the above decision-making logic, at each moment, a disturbance type label (such as wake, crosswind, turbulence, or hot air mass) and the dominant direction of the disturbance are output.

[0076] Step 6: Generation of obstacle avoidance and attitude stabilization control strategies.

[0077] After obtaining information on the type and direction of air disturbances, a multi-UAV obstacle avoidance and stability control strategy is generated based on the vibration mutual inspection matrix and the spatial arrangement of UAVs.

[0078] Step 601, let the first... The drone's current location is The current danger direction unit vector is: (For example, wind direction or wake propagation direction), for the configuration position correction of the drone in the high-risk area (i.e., obstacle avoidance offset): ; in, In accordance with A positive scalar coefficient, calibrated with the disturbance intensity, is used to control the avoidance distance.

[0079] Step 602: For overall disturbances such as turbulence or hot air masses, select a uniform height adjustment strategy based on the disturbance type and construct a uniform vertical correction amount. ,at this time, .

[0080] For systemic disturbances such as turbulence or hot air masses, a uniform height adjustment strategy is selected based on the disturbance type to construct a uniform vertical correction. Specifically, the formation's average vertical energy is calculated based on the aforementioned single-unit vertical energy index. and compared it with the baseline value under historical stable flight conditions. When comparing, At that time, it was considered that the overall vertical disturbance was relatively weak, and therefore... ;when At that time, it is determined based on the proportion that it exceeds the baseline value. Size: ; Among them, the larger the excess ratio, the larger the uniform vertical correction amount. The larger it is, and the greater it is, subject to the safe flight altitude range of drones. For thermally rising air masses, The direction is used to counteract the upward or downward trend; for overall turbulence, The direction is determined by comparing the strength of disturbances in the vertical direction, and the side with weaker disturbance is selected first for altitude synchronization adjustment.

[0081] Step 7: Obstacle avoidance control command execution and closed-loop update.

[0082] During the control execution phase, the desired position and attitude adjustment amounts generated in step 6 are... (That is, the result obtained in step 601 or step 602) As the control target, it is input into the flight control system of each UAV.

[0083] The flight control system based on the desired position and desired attitude angle The calculation formula is as follows: the thrust of each motor or the control command of the control surface is calculated through the attitude loop and the position loop, and the roll angle, pitch angle, yaw angle and altitude are adjusted to achieve obstacle avoidance and attitude stability in physical space.

[0084] Specifically, first calculate the desired acceleration within the control period: , ; In the formula, , These represent the target displacements that need to be achieved in the x and y directions, respectively. and These are the expected accelerations in the x and y directions, respectively. This refers to the single control update cycle of the UAV flight control system.

[0085] Then, the roll angle is calculated. and pitch angle : ; ; In the formula, For gravitational acceleration, yaw angle The default setting is to maintain the original heading.

[0086] After the control command is executed, the UAV continues to collect the updated IMU vibration data according to step 1, and re-executes the preprocessing, feature extraction, feature sharing, mutual inspection matrix construction and disturbance identification steps.

[0087] This embodiment uses a fixed control cycle to perform the above-mentioned cycle, realizing a real-time closed loop of air disturbance perception-decision-control. Through continuous closed-loop updates, it maintains the tracking of the environmental state and the dynamic adjustment of obstacle avoidance strategies in the case of rapid changes in the disturbance field, thereby maintaining the flight safety and stability of multi-UAV formations in complex soft obstacle environments such as strong crosswinds, wakes, turbulence and hot air masses.

[0088] This embodiment provides an in-flight obstacle avoidance method based on multi-aircraft vibration mutual detection. By collaboratively analyzing the vibration characteristics of multiple aircraft, it solves the problem that a single-aircraft IMU cannot distinguish between its own noise and external disturbances. It elevates disturbance judgment from single-point observation to multi-point spatially distributed observation, and eliminates the influence of body noise such as airframe structural vibration and propeller coupling through a vibration mutual detection matrix. This makes the identification results of air disturbances more stable and reliable, significantly improving the anti-interference capability in complex flight environments.

[0089] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which realizes advance perception of air disturbances and has predictive obstacle avoidance capabilities. By analyzing the evolution characteristics of the vibration mutual detection matrix in time and space, predictive identification can be completed before the disturbance causes a sudden change in attitude, providing the UAV with an advance response time of about 0.3 to 1.0 seconds, enabling obstacle avoidance control to change from passive correction to active avoidance, and significantly reducing the risk of flight instability.

[0090] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, establishing an air disturbance type discrimination mechanism based on a vibration mutual detection matrix structure. By comprehensively analyzing the matrix mean distribution, gradient change, and variance of difference, it can distinguish different disturbance types such as wake, crosswind, wind shear, overall turbulence, and thermal rising air mass, and determine the dominant direction of the disturbance, providing a clear basis for subsequent control decisions.

[0091] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, proposing an optimal obstacle avoidance control strategy based on risk distribution. By mapping the vibration mutual detection matrix to a spatial risk distribution and introducing a cost function to constrain obstacle avoidance offset and energy consumption, the minimum necessary control actions are generated under the premise of ensuring safety. This allows the UAV to escape the disturbance area with minimal attitude adjustment and trajectory deviation, avoiding excessive interference to the mission route and formation structure.

[0092] This embodiment provides an aerial obstacle avoidance method based on multi-drone vibration mutual detection, which realizes adaptive aerodynamic reconfiguration of multi-drone formation structures. According to the disturbance strength distribution reflected by the vibration mutual detection matrix, the spacing and altitude distribution between drones can be automatically adjusted, transforming the formation from a highly coupled structure to a disturbance-resistant structure, thereby reducing the risk of wake superposition and aerodynamic resonance, and significantly improving the overall stability of multi-drone formations in complex wind fields.

[0093] This embodiment provides an aerial obstacle avoidance method based on multi-drone vibration mutual detection, which requires no additional sensors and has the engineering advantages of low cost and low power consumption. It only uses the existing IMU vibration data of the UAV to complete soft obstacle perception and obstacle avoidance control, without relying on hardware such as vision, lidar or millimeter-wave radar. It has low computational load and low communication load, and can be directly deployed on existing UAV platforms, with good engineering feasibility and large-scale promotion value.

[0094] This embodiment provides an aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, which still possesses stable and reliable obstacle avoidance capabilities under visual perception failure conditions. In scenarios where visual perception fails, such as nighttime, backlighting, dense fog, smoke, and textureless backgrounds, the vibration mutual detection mechanism can still accurately identify air disturbances and execute obstacle avoidance control, significantly improving the all-weather flight safety of UAVs in complex environments.

[0095] Example 2 This embodiment provides an aerial obstacle avoidance system based on multi-aircraft vibration mutual detection, comprising: The data acquisition module is configured to acquire vibration data of each drone in the formation. The feature extraction module is configured to extract time-domain features, frequency-domain features, and time-frequency-domain features from the vibration data of each UAV, and then concatenate them to obtain a merged feature vector. The interaction module is configured to: for each UAV, encapsulate its own merged feature vector into a feature message and broadcast it periodically, and receive feature messages from other UAVs and unpack them, and then concatenate the merged feature vectors of all UAVs within the same time window to obtain a multi-UAV vibration feature set. The vibration mutual inspection module is configured to: calculate the vibration difference between each pair of UAVs based on the multi-aircraft vibration feature set, and obtain the vibration mutual inspection matrix. The obstacle avoidance control module is configured to: based on the vibration mutual detection matrix, identify the type of air disturbance and the dominant direction of the disturbance, generate the desired position and attitude adjustment amount, and perform obstacle avoidance control on the UAV.

[0096] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0097] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in Embodiment 1 above.

[0098] Example 4 This embodiment provides a computer device, such as... Figure 2 As shown, the system includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001. The processor 1001, communication interface 1002, and computer-readable storage medium 1003 can be connected via a bus or other means. The communication interface 1002 is used to receive and transmit data. When the processor 1001 executes the program, it implements the steps in the aerial obstacle avoidance method based on multi-aircraft vibration mutual detection described in Embodiment 1 above.

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An aerial obstacle avoidance method based on multi-aircraft vibration mutual detection, characterized in that, include: Obtain vibration data for each drone in the formation; For the vibration data of each drone, time-domain features, frequency-domain features, and time-frequency-domain features are extracted, and then concatenated to obtain a merged feature vector; For each UAV, its own merged feature vector is encapsulated into a feature message and broadcast periodically. After receiving feature messages from other UAVs, the message is unpacked. The merged feature vectors of all UAVs within the same time window are spliced ​​together to obtain a multi-UAV vibration feature set. Based on the multi-machine vibration feature set, the vibration difference between each pair of UAVs is calculated to obtain the vibration mutual inspection matrix; Based on the vibration cross-check matrix, after identifying the type and dominant direction of air disturbance, the desired position and attitude adjustment amount is generated to control the UAV for obstacle avoidance.

2. The aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in claim 1, characterized in that, Also includes: Before extracting time-domain features, frequency-domain features, and time-frequency-domain features, the vibration data of each UAV is preprocessed. The preprocessing includes DC removal, filtering with a digital bandpass filter, filtering with a notch filter, smoothing, and data normalization.

3. The aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in claim 1, characterized in that, The drone vibration data includes triaxial acceleration and triaxial angular velocity; The time-domain features include: root mean square value, mean, variance, and peak value; The frequency domain features include: spectrum, spectral power, high-frequency energy ratio, and spectral centroid; The time-frequency domain feature is multi-scale wavelet energy.

4. The aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in claim 1, characterized in that, The feature message includes its own merged feature vector, local unique identifier, and local timestamp.

5. The aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in claim 1, characterized in that, The vibration difference between the two drones is ;in, Let α and β be the 2-norm, and α and β be the weighting coefficients. The i-th and j-th UAVs at time... The merged feature vectors are respectively and .

6. The aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in claim 1, characterized in that, The steps for identifying the type of air disturbance and the dominant direction of the disturbance include: Calculate the average difference value of each row in the vibration cross-check matrix. Calculate the overall average difference Where N is the total number of drones in the formation. The vibration difference between two drones; When there are fewer than the set number of drones Satisfy the average difference value Significantly greater than And drones When located upstream or in front of the formation, it is identified as a drone. Located in a region with a strong wake; when Spatial Index When the gradient is monotonically increasing or decreasing and is stable as positive or negative in a certain direction, it is determined to be a unilateral wind or wind shear phenomenon, and the wind direction is determined by the gradient sign. Calculate the variance of the differences in the vibration cross-check matrix. Among them, the mean of the difference values ,when and When the vibration mutual inspection matrix is ​​increased and there are no gradient structures or local isolated high values, the formation as a whole is determined to be in the turbulent region. Constructing a vertical energy index using vertical vibration characteristics When all If the value is higher than the historical baseline and the difference value is evenly distributed in the vibration mutual inspection matrix, the current disturbance type is determined to be a thermal rising air mass.

7. The aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in claim 1, characterized in that, In the process of generating the desired position and attitude adjustment, a control cost function is introduced to constrain the obstacle avoidance offset and energy consumption.

8. An aerial obstacle avoidance system based on multi-aircraft vibration mutual detection, characterized in that, include: The data acquisition module is configured to acquire vibration data of each drone in the formation. The feature extraction module is configured to extract time-domain features, frequency-domain features, and time-frequency-domain features from the vibration data of each UAV, and then concatenate them to obtain a merged feature vector. The interaction module is configured to: for each UAV, encapsulate its own merged feature vector into a feature message and broadcast it periodically, and receive feature messages from other UAVs and unpack them, and then concatenate the merged feature vectors of all UAVs within the same time window to obtain a multi-UAV vibration feature set. The vibration mutual inspection module is configured to: calculate the vibration difference between each pair of UAVs based on the multi-aircraft vibration feature set, and obtain the vibration mutual inspection matrix. The obstacle avoidance control module is configured to: based on the vibration mutual detection matrix, identify the type of air disturbance and the dominant direction of the disturbance, generate the desired position and attitude adjustment amount, and perform obstacle avoidance control on the UAV.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in any one of claims 1-7.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the aerial obstacle avoidance method based on multi-aircraft vibration mutual detection as described in any one of claims 1-7.

Images