A UAV obstacle avoidance method and system based on inertial navigation

Abstract

The present invention provides a method and system for obstacle avoidance of unmanned aerial vehicles (UAVs) based on inertial navigation, which belongs to the technical field of obstacle avoidance systems for UAVs. It aims to solve the problem of insufficient obstacle avoidance capabilities of existing UAVs in environments such as weak GPS signals, limited vision, or sensor detection blind spots. Collect broadband high-frequency dynamic response data output in real time by the UAV's inertial navigation system; establish or obtain a baseline high-frequency dynamic response model of the UAV in a specific flight state and power system working mode without the influence of external obstacles; based on the high-frequency dynamic response signal; identify whether there are specific air medium disturbance echo characteristics formed by the disturbance generated by the UAV's own movement through external obstacles; and finally generate and execute obstacle avoidance flight instructions based on the estimated information. The method of the present invention does not rely on traditional external environmental sensors for initial detection, but uses INS to perceive physical interactions, thereby improving the autonomous obstacle avoidance capability and safety of the UAV in complex and challenging environments.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) obstacle avoidance systems, and in particular to an UAV obstacle avoidance method and system based on inertial navigation. Background Art

[0002] Currently, drone obstacle avoidance relies primarily on sensors such as GPS, vision, LiDAR, ultrasound, or infrared to perceive the environment. However, these technologies have significant limitations in specific scenarios: GPS fails indoors, in canyons, or when exposed to interference; vision is easily affected by lighting, weather, and environmental texture, and requires a high level of computation; while LiDAR offers high accuracy, its performance degrades in adverse weather conditions such as rain, fog, and dust, and it is costly and has difficulty detecting certain materials; and ultrasound and infrared sensors have short ranges and are susceptible to environmental interference.

[0003] Currently, wireless drones rely heavily on GPS and other signals for positioning and obstacle avoidance. Tethered drones (using optical fiber or cables) face significant challenges with their accessibility. When operating in complex, sensor-restricted environments, such as indoors, dense vegetation, tunnels, inclement weather, or with strong interference, the reliability of these traditional obstacle avoidance methods can be significantly reduced or even completely ineffective. This technical bottleneck severely restricts the application and operational safety of drones. Summary of the Invention

[0004] Technical problems solved

[0005] In view of the shortcomings of the existing technology, the present invention provides a method and system for avoiding obstacles in a UAV based on inertial navigation, which solves the problems of the existing technology.

[0006] Technical Solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method and system for avoiding obstacles in a UAV based on inertial navigation, comprising the following steps:

[0008] Sp1. Driving the UAV power system to generate motion and synchronously collecting broadband high-frequency dynamic response data output by the inertial navigation system, wherein the broadband high-frequency dynamic response data includes a three-axis high-frequency linear acceleration signal;

[0009] Sp2. Establish or obtain a benchmark high-frequency dynamic response model of the UAV under specific flight conditions and power system operating modes, considering only its own motion and interaction with the free space air medium;

[0010] Sp3, based on the current UAV flight state and power system operating mode parameters, using the benchmark high-frequency dynamic response model to predict the current expected high-frequency dynamic response signal of the UAV in real time;

[0011] Sp4, comparing the real-time high-frequency dynamic response data collected in step Sp1 with the expected high-frequency dynamic response signal predicted in step Sp3, and extracting the residual high-frequency signal representing the abnormal disturbance of the external environment;

[0012] Sp5. Perform time domain, frequency domain, or time-frequency domain analysis on the residual high-frequency signal to identify whether there is a specific air medium disturbance echo feature formed by the disturbance generated by the drone's own motion after being reflected, diffracted, or interfered by external obstacles;

[0013] If the specific air medium disturbance echo feature is identified and the energy or correlation index of the feature exceeds a preset determination threshold, it is confirmed that an obstacle is detected;

[0014] Sp7. estimating the relative position, size or shape of the obstacle based on the time delay information, frequency components, multi-axis signal correlation or spatial distribution pattern of the specific air medium disturbance echo characteristics;

[0015] Sp8. Generate and execute obstacle avoidance flight instructions based on the estimated obstacle information.

[0016] Preferably, the reference high-frequency dynamic response model in Sp2 is established by combining computational fluid dynamics simulation with structural dynamics analysis, or is driven by data acquired through calibration flight in a known large obstacle-free space.

[0017] Preferably, in the Sp5, the identification of the specific air medium disturbance echo characteristics includes: detecting the modulation signal related to the passing frequency of the drone rotor blade or its harmonics, the Doppler frequency shift characteristics, or the abnormal energy enhancement of a specific frequency band.

[0018] Preferably, the drone obstacle avoidance method further includes:

[0019] In Sp1, a small disturbance modulation signal with specific spectral characteristics is actively applied to the UAV power system;

[0020] In Sp5, the analysis is focused on the relevant components in the residual high-frequency signal that are synchronized with the disturbance modulation signal or have a specific delay, so as to enhance the detection sensitivity and anti-interference capability of the specific air medium disturbance echo characteristics.

[0021] Preferably, in the above-mentioned Sp7, the relative position of the obstacle is estimated by analyzing the phase difference or amplitude ratio of the echo characteristics of the specific air medium disturbance on different axial sensors or different position sensors of the inertial navigation system.

[0022] Preferably, in the above-mentioned Sp7, the relative distance of the obstacle is estimated by analyzing the time delay of the specific air medium disturbance echo characteristics relative to the initial disturbance generated by the drone itself, and calculating it in combination with the disturbance propagation speed model in the air medium.

[0023] Preferably, the preset determination threshold is dynamically adjusted according to the current environmental background noise level, the UAV flight speed or the estimated value of the atmospheric turbulence intensity.

[0024] Preferably, the drone obstacle avoidance method further includes:

[0025] fusing sensor data, wherein the fused sensor data includes at least one type of sensor data;

[0026] The fused sensor data is used to confirm the identified echo features in step Sp6, or to calibrate the estimated obstacle information in step Sp7, or to assist in optimizing the obstacle avoidance flight instructions in step Sp8, but the initial detection of the obstacle mainly depends on the analysis process based on the inertial navigation system defined in steps Sp1 to Sp6.

[0027] Preferably, the drone obstacle avoidance method further includes:

[0028] Continuously record the detected echo characteristics, estimated obstacle information and obstacle avoidance execution results;

[0029] By using these recorded data, the reference model in step Sp2, the feature recognition algorithm in step Sp5 or the obstacle information estimation algorithm in step Sp7 is iteratively optimized through an online or offline learning algorithm.

[0030] Preferably, the drone system includes:

[0031] An inertial navigation system for broadband high frequency dynamic response data, a processor, and a memory storing computer executable instructions configured to be executed by the processor.

[0032] Beneficial effects

[0033] The present invention provides a method and system for avoiding obstacles in a UAV based on inertial navigation. It has the following beneficial effects:

[0034] This invention uses an INS as a high-frequency "vibration pickup" to detect air medium disturbance wave information, which is generated by the drone's own motion and modulated by external obstacles. It no longer relies on the macroscopic forces or torques exerted by obstacles on the drone, but instead captures the more subtle and long-range medium disturbance effects. This provides a potential alternative or supplementary obstacle avoidance approach based on fundamental physical principles in scenarios where traditional sensors are limited or ineffective, and represents a groundbreaking new use case for the original inertial navigation system. This breakthrough technology application is achieved without increasing hardware costs, or even at a lower cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a system architecture diagram of the present invention;

[0036] Figure 2 A cloud diagram of the system composition of the present invention;

[0037] Figure 3 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Specific embodiment one:

[0040] An inertial navigation system (INS) is a navigation system that uses an inertial measurement unit (IMU) to independently calculate the position, attitude, and velocity of vehicles such as drones, aircraft, missiles, and submarines, independent of external information such as GPS satellite signals or ground base stations. Components include accelerometers, which measure the linear acceleration of the vehicle along three mutually perpendicular axes. Acceleration is integrated once to obtain velocity, and secondarily integrated to obtain displacement.

[0041] Gyroscope: Used to measure the angular velocity of the vehicle around three mutually perpendicular axes. By integrating the angular velocity, the vehicle's attitude angle, pitch angle, roll angle, yaw angle, etc. are obtained.

[0042] The INS reads accelerometer and gyroscope data at high frequency, performs integration operations, and performs coordinate system transformations to continuously infer the position, velocity, and attitude of the drone relative to its initial state. As previously discussed, the INS provides high-frequency, high-precision, real-time data on the drone's motion, including acceleration and angular velocity. Its high-frequency response is used to capture subtle vibrations or dynamic response characteristics caused by the external environment (such as air disturbances caused by obstacles) acting on the drone's body, thereby indirectly detecting obstacles.

[0043] like Figure 1-Figure 3 As shown, a method and system for avoiding obstacles in a UAV based on inertial navigation includes the following steps:

[0044] Sp1. Drive the drone's propulsion system to generate motion and simultaneously collect broadband, high-frequency dynamic response data from the inertial navigation system. This broadband, high-frequency dynamic response data includes triaxial, high-frequency linear acceleration signals. This step is the foundation of the entire method. The drone's propulsion system, including motors and propellers, drives the drone in motion according to flight control commands. In this process, the propulsion system not only provides thrust but also continuously radiates energy into the surrounding environment in the form of sound waves, body vibrations, and air pressure fluctuations. The inertial navigation system installed on the aircraft, particularly the triaxial accelerometer within it, must possess wideband characteristics. Its effective measurement frequency range must cover the main frequency components and their harmonics generated by the power system's operation, as well as the characteristic frequencies expected to appear in obstacle echoes. Typically, its operating bandwidth must be at least 1 kHz, and may even exceed several thousand Hz. Furthermore, to capture weak echo signals, the accelerometer's noise density must be sufficiently low, less than a few hundred micrograms per root-of-hertz. The system synchronously acquires the three-axis, high-frequency linear acceleration signals output by the inertial navigation system at an extremely high sampling rate of 2 kHz to 10 kHz or higher, generating a raw, broadband, high-frequency dynamic response data stream. Crucially, this data acquisition must be precisely time-synchronized with the operating state parameters of the UAV's powertrain, such as the real-time RPM or PWM command values ​​of each motor, as well as essential flight state parameters such as airspeed, attitude angle, and angular rate. Synchronization accuracy must reach microseconds, achieved through hardware triggering or precision time protocols such as PTP, ensuring that subsequent signal processing accurately correlates the response signal with the source state that generated it.

[0045] Sp2. Establish or obtain a baseline high-frequency dynamic response model for the drone under specific flight conditions and propulsion system operating modes, considering only its own motion and interaction with the free air medium. The goal of this step is to accurately predict or characterize the high-frequency dynamic response signals generated by the drone's own operating state under ideal conditions without obstacle interference. This model forms the basis for subsequent extraction of abnormal signals. The following two approaches are preferred for model establishment:

[0046] First, physical simulation drives modeling. This method first uses computational fluid dynamics (CFD) software to simulate the complex unsteady flow field generated by propeller rotation and flow around the fuselage under different flight conditions (speed, angle of attack, sideslip angle) and power system operating modes (rotational speed), accurately calculating the detailed pressure distribution acting on the fuselage surface. This time-varying pressure distribution is then applied as an external load to a detailed UAV structural dynamics model established using the finite element method (FEM). By solving the structural dynamics equations, the three-axis high-frequency vibration acceleration response at the inertial navigation system installation point is obtained. This process is computationally intensive and is usually completed offline. The results are stored in the form of parameterized functions, high-dimensional lookup tables, or reduced-order models for online real-time access. The model must consider the structural material properties, component connection methods, and the dynamic characteristics of sensor installation.

[0047] Second, data-driven modeling. This approach requires comprehensive calibration flights in a large, known, safe space free of reflective obstacles, such as a large anechoic chamber or an open area hundreds of meters above sea level. During flight, the drone must traverse a variety of typical flight state combinations within its operating envelope, including varying speeds, altitudes, attitudes, and power output levels. Simultaneously, the complete flight parameters for each state and simultaneously acquired high-frequency inertial navigation system data are recorded with high precision. Leveraging this extensive data, advanced system identification techniques are employed to construct a model. Possible models include the nonlinear autoregressive exogenous model (NARX), recurrent neural networks with long-term memory capabilities such as long short-term memory (LSTM), Gaussian process regression (GPR) with uncertainty estimation, or deep neural networks (DNN) with end-to-end learning. The training goal is to establish a precise mapping from the input real-time flight and power state parameter vectors to the output expected three-axis high-frequency acceleration signal time series. Data-driven models must prioritize the completeness and representativeness of the training data and incorporate online adaptive update mechanisms to address environmental changes or drift in vehicle characteristics.

[0048] Regardless of the method used, the final baseline model Able to input state matrix in real time , They are motor speed, flight speed, and flight mode parameters. The input state matrix includes but is not limited to the above parameters to predict the benchmark high-frequency dynamic response signal .

[0049] Sp3, based on the current UAV flight state and power system working mode parameters, use the benchmark high-frequency dynamic response model to predict the UAV's current expected high-frequency dynamic response signal in real time; this step is run in real time during the UAV flight. The system obtains the current UAV flight state parameters (including speed, attitude, angular rate, altitude) and power system working mode parameters (including each motor RPM). These real-time parameters are used as input and substituted into the benchmark high-frequency dynamic response model established or obtained in step Sp2. The model then outputs the expected high-frequency dynamic response signal for the current moment or the next very short time window. This prediction process requires efficient computing to ensure that its output rate can match the data acquisition rate of the inertial navigation system and meet real-time requirements.

[0050] Sp4, compare the real-time high-frequency dynamic response data collected in step Sp1 with the expected high-frequency dynamic response signal predicted in step Sp3, and extract the residual high-frequency signal that represents the abnormal disturbance of the external environment; this step aims to remove the known part generated by the drone's own operation from the real-time measurement signal containing all information, thereby highlighting the abnormal disturbance caused by external unknown factors, especially the obstacle echo. The specific operation is to compare the high-frequency dynamic response data collected in real time in step Sp1 with the expected high-frequency dynamic response signal predicted in step Sp3, and extract the residual high-frequency signal that represents the abnormal disturbance of the external environment; this step aims to remove the known part generated by the drone's own operation from the real-time measurement signal containing all information, thereby highlighting the abnormal disturbance caused by external unknown factors, especially the obstacle echo. Expected high-frequency dynamic response signal predicted in real time with step Sp3 After precise alignment in time, perform vector subtraction: . This is the residual high-frequency signal. Ideally, if the model is perfect and there are no external obstacles, this residual signal should be close to zero-mean white noise. In practice, it will contain model errors, unmodeled internal noise sources, atmospheric turbulence, and the desired obstacle return signals. Therefore, subsequent steps require further target feature extraction from this residual signal.

[0051] Sp5, perform time domain, frequency domain or time-frequency domain analysis on the residual high-frequency signal to identify whether there are specific air medium disturbance echo characteristics formed by the disturbance generated by the drone's own movement after being reflected, diffracted or interfered by external obstacles; in the signal processing stage, the goal is to detect and identify weak obstacle echo characteristics from background noise and interference. The residual high-frequency signal obtained in step Sp4 needs to be analyzed. Perform in-depth analysis and use Fast Fourier Transform (FT) or improved spectrum estimation algorithms to calculate the power spectrum density of the residual signal. Focus on the characteristic frequencies related to the UAV power system, especially the rotor blade pass frequency (BPF) and its harmonic frequencies. The presence of obstacles causes abnormal energy enhancement peaks near these frequency points, or causes slight frequency shifts due to the Doppler effect. BPF can be Calculated, where is the number of blades. Time-frequency domain analysis: For non-stationary echo signals or scenarios where their occurrence time needs to be located, time-frequency analysis tools such as short-time Fourier transform (STFT), wavelet transforms (such as continuous wavelet transform (CWT) or discrete wavelet transform (DWT), or Hilbert-Huang transform (HHT) are used. These tools can display the distribution of signal energy over time and frequency, helping to identify transient echo characteristics or the temporal variation patterns of specific frequency components.

[0052] Active modulation signal analysis: If an active disturbance modulation signal with a specific spectrum is applied to the power system in step Sp1, for example, a frequency of Sinusoidal RPM fluctuations. In this step, the focus should be on analyzing the residual signal The modulated signal and its interaction frequency with the natural frequency of BPF (such as ) synchronization component. This can be achieved digitally using the principles of a lock-in amplifier or by calculating the cross-correlation function between the residual signal and a known modulation signal template (taking into account propagation delay). The presence of a high correlation peak or a specific frequency-locked component greatly enhances detection confidence and effectively suppresses uncorrelated noise.

[0053] Pattern Recognition and Machine Learning: Design or train pattern recognition algorithms, detectors based on template matching, or use machine learning models including convolutional neural networks (CNNs) or recurrent neural networks (RNNs) to learn and recognize complex echo signature patterns directly from the raw fragments of the residual signal or its time-spectrogram.

[0054] The output of this step is a decision on whether an echo feature exists, as well as the extracted feature parameters, including the echo signal's center frequency, bandwidth, energy intensity, signal-to-noise ratio (SNR), occurrence time, duration, and correlation with the active modulation signal.

[0055] Sp6. If a specific air medium disturbance echo feature is identified, and the energy or correlation index of the feature exceeds the preset judgment threshold, it is confirmed that the obstacle is detected; this step is based on the feature identified in step Sp5 to make the final confirmation of the existence of the obstacle. Compare the key quantitative indicators of the extracted echo features, such as the SNR of a specific frequency peak, the height of the cross-correlation peak, the output strength of the matched filter, or the confidence score of the machine learning classifier, with a preset judgment threshold. Only when the indicator significantly exceeds the threshold is it confirmed that the obstacle has been detected to control the false alarm rate. The preset judgment threshold should not be fixed, and the preferred solution is to use a dynamic adaptive threshold. The basis for adjusting the threshold includes: real-time background noise level: by analyzing the energy level of the residual signal in the unexpected echo frequency area, the background noise power is estimated in real time. , threshold setting ,in is the basic threshold, Is the proportional coefficient. UAV flight speed: When flying at high speed, the aerodynamic noise and structural vibration generated by itself are usually stronger, and the threshold needs to be appropriately increased. Atmospheric turbulence intensity: Turbulence intensity is estimated using low-frequency inertial data or atmospheric sensors. Strong turbulence environments introduce more interference, and the threshold needs to be increased. Sensor fusion confirmation: When step Sp5 preliminarily identifies suspicious features, query other low-cost, short-range sensors such as ultrasonic or infrared sensors. If these sensors also report potential objects in the corresponding direction, even if the information is vague, the confirmation threshold should be appropriately lowered. , improve detection sensitivity; conversely, if other sensors clearly indicate that there is no object in that direction, the threshold is raised to reduce false alarms. The confirmation logic is designed as "continuous confirmation", which requires the feature to stably exceed the threshold in multiple consecutive analysis windows, or "multi-feature confirmation", which requires multiple different types of echo feature indicators to meet the standards simultaneously.

[0056] Sp7: Estimate the relative position, size, or shape of the obstacle based on the time delay, frequency content, multi-axis signal correlation, or spatial distribution pattern of the echo characteristics of a specific air disturbance. After confirming the detection of an obstacle, this step uses the echo characteristics extracted in step Sp5 to estimate relevant information about the obstacle, providing a basis for obstacle avoidance. Single INS three-axis analysis: Analyze the distribution ratio of the confirmed echo signal energy along the three measurement axes of the inertial navigation system (X, Y, and Z). If the Z-axis energy is significantly greater than the X and Y axes, it indicates that the obstacle is primarily above or below. This only provides rough directional information. If the drone is equipped with multiple (at least two) inertial measurement units (IMUs) with precisely calibrated relative position and attitude, the echo's direction of arrival (i.e., the obstacle's relative azimuth and elevation) can be more accurately determined by calculating the time difference (TDOA) or phase difference (PDoA) of the same echo signal arriving at different units. TDOA is calculated using the peak position of the cross-correlation function, or PDoA is obtained by calculating the cross-power spectrum of the signals between sensors. This requires extremely high temporal synchronization between the sensors. For distributed INS, the amplitude differences of the same echo signal measured on different sensors are also analyzed, and the direction is inferred by combining the signal attenuation model.

[0057] Relative distance estimation is based on echo time delay: This is the primary distance estimation method. It is necessary to identify the echo feature (a pulse or correlation peak in the residual signal) associated with a specific energy emission event of the drone itself (the moment when a specific blade passes a fixed reference point on the fuselage, or a specific phase point of the active modulation signal). The time difference from the emission event to the reception of the corresponding echo feature is measured. Using the known propagation speed of air medium disturbance (Speed ​​of sound, needs to be corrected according to environmental parameters such as temperature), obstacle distance It can be estimated as The accuracy of this method is highly dependent on the time delay Measurement accuracy and precise definition of the initial emission time are crucial. In the case of continuous wave or narrowband signals, the rate of change of distance is inferred rather than the absolute distance using phase delay information. Size or shape estimation is performed by analyzing characteristics such as the echo signal's spectral width (large objects reflect a wider frequency band), energy attenuation with frequency (scattering characteristics of different materials or shapes), and the echo signal's duration or Doppler spread, combined with a priori databases or complex scattering models for preliminary inference. This typically requires more complex signal processing and machine learning models. Sensor fusion calibration: If sensors that provide accurate 3D point clouds, such as LiDAR or structured light vision, are integrated, even if this data is sparse or discontinuous, it is used to calibrate or refine the position and range estimates derived from INS echo features, improving overall estimation accuracy and reliability. This step ultimately outputs a state vector containing information such as the confidence level of the obstacle's presence, the estimated relative position (e.g., horizontal angle, pitch angle), and the distance (or range level, closing rate).

[0058] Sp8. Generate and execute obstacle avoidance flight instructions based on the estimated obstacle information.

[0059] The baseline high-frequency dynamic response model in Sp2 is established through computational fluid dynamics simulation combined with structural dynamics analysis, or is driven by data obtained through calibration flight in a known large obstacle-free space. After receiving the obstacle information estimated in step Sp7, this step is responsible for making decisions and executing obstacle avoidance actions.

[0060] Input: obstacle state vector, current state of the drone (precise position, velocity, attitude), predetermined mission waypoints or trajectories, and flight performance constraints (maximum acceleration, angular rate, etc.).

[0061] The strategies are divided into the following cases:

[0062] 1. Reactive: Simple logic and fast response. If an obstacle is within a certain distance, the system will initiate an emergency brake or evade it upwards, downwards, left, or right. The evasion direction is opposite to the obstacle, and the evasion range is inversely proportional to the distance.

[0063] 2. Based on geometric programming: Use the speed barrier method or ORCA method to calculate the speed set to avoid collision in the speed space and select the optimal avoidance speed.

[0064] 3. Optimization-based: Using model predictive control (MPC), the trajectory of the drone and obstacles is predicted within a finite future horizon. An optimization problem is then solved to find a set of control inputs that maximizes the obstacle avoidance margin and minimizes path deviation while satisfying the constraints. The MPC cost function requires careful weighting to balance safety, efficiency, and comfort. This approach can be customized for specific application environments, including those for FPV drones, entertainment drones, or feature drones, allowing for the development of different threshold control schemes.

[0065] The output of the decision logic is converted into commands acceptable to the underlying flight controller, including target attitude angle / angular rate, target velocity vector, target acceleration, or target thrust / throttle commands. These commands are then sent to the flight controller for execution. During obstacle avoidance, steps 1-7 are continuously executed to monitor changes in obstacle status and dynamically adjust the avoidance strategy until the threat is determined to have been resolved.

[0066] Sensor fusion optimization: Obstacle avoidance path planning fully utilizes the environmental map information provided by the fused sensors to select a more globally optimal avoidance path, bypassing rather than simply stopping or turning.

[0067] In Sp5, the identification of specific air medium disturbance echo characteristics includes: detecting modulation signals related to the passing frequency of the UAV rotor blades or its harmonics, Doppler frequency shift characteristics, or abnormal energy enhancement in specific frequency bands.

[0068] Drone obstacle avoidance methods also include:

[0069] In Sp1, a small disturbance modulation signal with specific spectral characteristics is actively applied to the UAV power system;

[0070] In Sp5, the focus is on analyzing the relevant components in the residual high-frequency signal that are synchronized with the disturbance modulation signal or have a specific delay, so as to enhance the detection sensitivity and anti-interference ability of the specific air medium disturbance echo characteristics.

[0071] In Sp7, the relative position of the obstacle is estimated by analyzing the phase difference or amplitude ratio of the echo characteristics of the specific air medium disturbance on different axial sensors or different position sensors of the inertial navigation system.

[0072] In Sp7, the relative distance of obstacles is estimated by analyzing the time delay of the echo characteristics of a specific air medium disturbance relative to the initial disturbance generated by the drone itself, and combining it with the disturbance propagation speed model in the air medium.

[0073] The preset judgment threshold is dynamically adjusted according to the current environmental background noise level, drone flight speed or atmospheric turbulence intensity estimation.

[0074] Drone obstacle avoidance methods also include:

[0075] fusing sensor data, where the fused sensor data includes at least one type of sensor data;

[0076] The fused sensor data is used to confirm the identified echo features in step Sp6, or to calibrate the estimated obstacle information in step Sp7, or to assist in optimizing the obstacle avoidance flight instructions in step Sp8, but the initial detection of obstacles mainly depends on the analysis process based on the inertial navigation system defined in steps Sp1 to Sp6.

[0077] Drone obstacle avoidance methods also include:

[0078] Continuously record the detected echo characteristics, estimated obstacle information and obstacle avoidance execution results;

[0079] By using these recorded data, the benchmark model in step Sp2, the feature recognition algorithm in step Sp5 or the obstacle information estimation algorithm in step Sp7 is iteratively optimized through an online or offline learning algorithm. Specific embodiment two:

[0081] like Figure 1-Figure 3 As shown, the UAV system includes:

[0082] An inertial navigation system (INS) with broadband, high-frequency dynamic response data, a processor, and memory, with the memory storing computer-executable instructions configured for execution by the processor, requires a high-bandwidth, low-noise INS sensor. The processor must possess strong real-time computing capabilities, employing a multi-core CPU, GPU, or FPGA to handle complex signal processing, model prediction, and optimization computations. A high-speed data bus and ample memory are required. A real-time operating system (RTOS) ensures deterministic task scheduling. The software is modular, encompassing modules such as data acquisition and synchronization, baseline model prediction, signal processing and feature extraction, obstacle estimation, decision-making and control, sensor fusion, data logging, and human-computer interaction and monitoring.

[0083] The system should have data logging capabilities, recording detailed data for each detection event, including raw signals, processing results, decision-making processes, flight responses, and mission scenario information. This data can be used to train and optimize baseline models, feature extractors, estimators, and decision logic offline, and even employ reinforcement learning methods to optimize obstacle avoidance strategies. Online, real-time adjustments to noise models and judgment thresholds are also designed. For safety-critical obstacle avoidance systems, failure mode analysis must be considered, and necessary redundancy and backup mechanisms must be designed. If the INS detection method fails or confidence is low, the system can smoothly switch to an obstacle avoidance mode based on other sensors or execute a pre-set safe hover / return procedure. Specific embodiment three:

[0085] Based on the technical solutions of the specific embodiment 1 and the specific embodiment 2, further application descriptions in combination with the environment are given:

[0086] Drones need to fly autonomously in large warehouses with numerous shelves, and automatically count the inventory on high-level shelves using the scanning equipment they carry (such as RFID readers or barcode scanners).

[0087] GPS signals are completely unavailable, requiring reliance on internal positioning systems (such as SLAM or UWB), which can suffer from accumulated errors or drift. The warehouse layout is complex, with narrow aisles between shelves and numerous metal structures. This can cause texture loss or repetitive areas for visual SLAM and multipath interference for LiDAR. Inventory is irregularly stacked, sometimes with temporarily placed forklifts, stacked goods, or equipment that hasn't been returned to its original location, creating static or dynamic obstacles not marked on the map. Lighting conditions are also uneven.

[0088] A quadrotor drone is used, equipped with a high-precision positioning module, a mission payload, and the inertial navigation-based obstacle avoidance system described in this solution. The INS uses an industrial-grade MEMS inertial measurement unit (INS) with a sampling rate of at least 4 kHz and an accelerometer noise density below 150 micro-g / sqrt (Hz). A baseline high-frequency dynamic response model has been established through calibration flights in a safe area within a warehouse. The system uses an active modulation strategy to apply a weak sinusoidal perturbation to the motor RPM.

[0089] The drone flies slowly along the shelf aisle at a speed of 0.5 meters per second according to the predetermined path. The main navigation system provides positioning and tracking control.

[0090] Sp1-Sp3: The obstacle avoidance system continuously collects 4kHz INS acceleration data, synchronously recording the motor RPM and active modulation signal phase. Based on the current flight speed, attitude, and RPM, it uses a data-driven benchmark model in real time to predict the expected high-frequency acceleration signal when there are no obstacles.

[0091] Sp4: Subtract the predicted reference signal from the real-time INS data to obtain the residual high-frequency signal.

[0092] The drone is approaching a corner of a corridor with a specific obstacle not included in the pre-installed map. The drone's visual or LiDAR sensors cannot clearly identify the obstacle due to the angle or distance.

[0093] Sp5: The signal processing module analyzes the residual signal. Because active modulation is enabled, the system focuses specifically on components related to the modulation frequency and its interaction with the sideband frequencies generated by the blade pass frequency (BPF). Analysis revealed a significant and stable peak in the residual signal energy at one of the sideband frequencies, with a signal-to-noise ratio (SNR) reaching 8dB. Furthermore, cross-correlation calculations revealed a stable time delay between this frequency component and the transmitted modulated signal.

[0094] Sp6: Real-time background noise estimation indicates the current ambient noise level is normal. The calculated SNR (8dB) exceeds the confirmation threshold, which is dynamically adjusted based on the current noise level (typically set to 6dB). The system confirms the detection of an obstacle.

[0095] Sp7: By analyzing the energy distribution of the echo signature along the three axes of the INS, we preliminarily determined that the obstacle was primarily located to the right of the front. Using the precisely measured time delay and the air speed model, we estimated the obstacle's distance to be approximately 2.5 meters.

[0096] Sp8: The obstacle avoidance module receives information about an obstacle (approximately 2.5 meters to the right). Since the drone is flying within a corridor with limited left and right space, the system generates a command: immediately stop forward and nudge 0.5 meters to the left. It also issues an alarm or pauses the mission. The drone executes this command smoothly.

[0097] The drone successfully avoided a collision with an unexpected ladder, ensuring equipment safety and the continuity of the inventory task. In complex indoor environments characterized by GPS failure, uneven lighting, and sensor blind spots, this INS-based approach, independent of external positioning and traditional vision / LiDAR perception, successfully detected unmodeled static obstacles by analyzing its own disturbance echoes, demonstrating its unique environmental adaptability. The active modulation strategy improves detection sensitivity and reliability. Specific embodiment four:

[0099] Based on the technical solutions of the specific embodiment 1 and the specific embodiment 2, further application descriptions in combination with the environment are given:

[0100] Drones are required to inspect concrete surface corrosion and cracks near the waterline of large sea-crossing bridge piers at close range. The massive concrete structures of these piers severely obstruct GPS signals and cause multipath interference, making positioning accuracy unreliable. Water reflections, waves, fog, and water vapor can interfere with the detection performance of visual and LiDAR sensors. Strong and erratic sea breezes increase flight control challenges and background noise. The very close inspection distance (less than 2 meters) requires precise sensing of the distance and relative attitude to the pier surface to avoid collisions.

[0101] A multi-rotor drone with adequate wind resistance was used, equipped with a high-definition zoom camera, a laser rangefinder (for auxiliary verification and distance calibration), and the proposed obstacle avoidance system. The INS was a high-performance MEMS or fiber optic gyroscope (FOG) with a 6kHz sampling rate and low accelerometer noise. The benchmark model combined physical simulation (accounting for the unique wind conditions near bridge piers) with calibration data from an actual open area on the bridge.

[0102] Under remote monitoring by the operator or based on feedback from a laser rangefinder, the drone slowly approaches the surface of the pier, attempting to maintain an inspection distance of 1.5 meters.

[0103] Sp1-Sp4: The system collects 6kHz INS data, predicts the reference signal (the model already takes into account the influence of ground effect and wall effect when approaching large structures), and extracts the residual signal.

[0104] A sudden sideways sea breeze pushed the drone closer to the bridge pier. Simultaneously, the baseline model's prediction accuracy temporarily decreased due to the drone's changing attitude and rapidly changing distance from the wall. The laser rangefinder's infrequent updates prevented it from promptly reflecting the most dangerous approach.

[0105] Sp5: The residual signal analysis module detected a sharp increase in energy levels across a wide frequency range (particularly those associated with the aircraft's structural natural frequencies and the BPF), far exceeding the fluctuations caused by normal wind disturbances. The signal characteristics exhibited asymmetry, indicating strong reflections from specific directions.

[0106] Sp6: The overall energy integral of the residual signal or the peaks of multiple key frequencies simultaneously exceed the dynamic threshold (which has been raised based on wind speed estimation). The system determines with high confidence that a close obstacle (i.e., the bridge pier surface) is rapidly approaching.

[0107] Sp7: Analyzing the energy distribution of the echo signal in the aircraft's coordinate system (the right sensor's axial signal is significantly stronger than the left), we confirm that the threat originates primarily from the right. Time delay analysis (if discernible) or simply a sharp increase in energy intensity indicates a rapidly decreasing distance and an extremely high risk level.

[0108] Sp8: The obstacle avoidance system triggers the highest priority avoidance command: immediately apply maximum lateral thrust to the left (away from the bridge pier) while slightly increasing the total thrust to resist the height loss caused by wind pressure and lateral displacement.

[0109] The drone was forcibly pushed away before colliding with the bridge pier and returned to a safe distance. Although the inspection mission was temporarily interrupted, expensive equipment damage was avoided. In situations where GPS is unreliable and other sensors are slow to respond or fail due to environmental factors (reflections, moisture, gusts of wind, and low update rates), the INS-based approach provides the ultimate, rapid safety guarantee by sensing abnormal air interactions (strong reflections at very close range) caused by rapid distance changes and strong wind disturbances. It is extremely sensitive to dynamically changing distances and relative attitude changes, making it suitable as a last-ditch defense for close-range operations.

[0110] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further restrictions, an element defined by the statement "comprising a reference structure" does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0111] While embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations may be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for avoiding obstacles in a UAV based on inertial navigation, characterized in that: The following steps are involved: Sp1. Driving the UAV power system to generate motion and synchronously collecting broadband high-frequency dynamic response data output by the inertial navigation system, wherein the broadband high-frequency dynamic response data includes a three-axis high-frequency linear acceleration signal; Sp2. Establish or obtain a baseline high-frequency dynamic response model of the UAV under a specific flight state and power system operating mode, considering only its own motion and interaction with the free air medium. The specific flight state is the flight state of the UAV without obstacle interference. Sp3, based on the current UAV flight state and power system operating mode parameters, using the benchmark high-frequency dynamic response model to predict the current expected high-frequency dynamic response signal of the UAV in real time; Sp4, comparing the real-time high-frequency dynamic response data collected in step Sp1 with the expected high-frequency dynamic response signal predicted in step Sp3, and extracting the residual high-frequency signal representing the abnormal disturbance of the external environment; Sp5. Perform time domain, frequency domain, or time-frequency domain analysis on the residual high-frequency signal to identify whether there are specific air medium disturbance echo characteristics formed by the disturbance generated by the UAV's own motion after being reflected, diffracted, or interfered by external obstacles. The specific air medium disturbance echo characteristics are characteristic frequencies related to the UAV's power system, including the rotor blade pass frequency and subharmonic frequencies. Obstacles cause abnormal energy enhancement peaks near these frequency points, or slight frequency shifts caused by the Doppler effect; If the specific air medium disturbance echo feature is identified and the energy or correlation index of the feature exceeds a preset determination threshold, it is confirmed that an obstacle is detected; Sp7. estimating the relative position, size or shape of the obstacle based on the time delay information, frequency components, multi-axis signal correlation or spatial distribution pattern of the specific air medium disturbance echo characteristics; Sp8. Generate and execute obstacle avoidance flight instructions based on the estimated obstacle information.

2. The obstacle avoidance method for a UAV based on inertial navigation according to claim 1, characterized in that: The benchmark high-frequency dynamic response model in the Sp2 is established by combining computational fluid dynamics simulation with structural dynamics analysis, or is driven by data obtained through calibration flight in a known large obstacle-free space.

3. The obstacle avoidance method for a UAV based on inertial navigation according to claim 1, characterized in that: In the above-mentioned Sp5, the identification of the specific air medium disturbance echo characteristics includes: detecting the modulation signal related to the passing frequency of the UAV rotor blade or its harmonics, the Doppler frequency shift characteristics, or the abnormal energy enhancement of a specific frequency band.

4. The obstacle avoidance method for a UAV based on inertial navigation according to claim 1, characterized in that: The UAV obstacle avoidance method further includes: In Sp1, a small disturbance modulation signal with specific spectral characteristics is actively applied to the UAV power system; In Sp5, the analysis is focused on the relevant components in the residual high-frequency signal that are synchronized with the disturbance modulation signal or have a specific delay, so as to enhance the detection sensitivity and anti-interference capability of the specific air medium disturbance echo characteristics.

5. The obstacle avoidance method for a UAV based on inertial navigation according to claim 1, characterized in that: In the above-mentioned Sp7, the relative position of the obstacle is estimated by analyzing the phase difference or amplitude ratio of the echo characteristics of the specific air medium disturbance on different axial sensors or different position sensors of the inertial navigation system.

6. The method for avoiding obstacles in a UAV based on inertial navigation according to claim 1, characterized in that: In the above-mentioned Sp7, the relative distance of the obstacle is estimated by analyzing the time delay of the echo characteristics of the specific air medium disturbance relative to the initial disturbance generated by the drone itself, and calculating it in combination with the disturbance propagation speed model in the air medium.

7. The method for avoiding obstacles in a UAV based on inertial navigation according to claim 1, characterized in that: The preset judgment threshold is dynamically adjusted according to the current environmental background noise level, the UAV flight speed or the estimated value of the atmospheric turbulence intensity.

8. The obstacle avoidance method for a UAV based on inertial navigation according to claim 1, characterized in that: The UAV obstacle avoidance method further includes: fusing sensor data, wherein the fused sensor data includes at least one type of sensor data; The fused sensor data is used to confirm the identified echo features in Sp6, or to calibrate the estimated obstacle information in Sp7, or to assist in optimizing the obstacle avoidance flight instructions in Sp8.

9. The obstacle avoidance method for a UAV based on inertial navigation according to claim 1, characterized in that: The UAV obstacle avoidance method further includes: Continuously record the detected echo characteristics, estimated obstacle information and obstacle avoidance execution results; By using these recorded data, the benchmark model in Sp2, the feature recognition algorithm in Sp5 or the obstacle information estimation algorithm in Sp7 is iteratively optimized through an online or offline learning algorithm.

10. A UAV system corresponding to the UAV obstacle avoidance method based on inertial navigation according to any one of claims 1 to 9, characterized in that: The drone system includes: An inertial navigation system for broadband high frequency dynamic response data, a processor, and a memory storing computer executable instructions configured to be executed by the processor.

Images