Aircraft precise landing system and method based on multi-source heterogeneous data fusion

Abstract

The invention discloses an aircraft precise landing system and method based on multi-source heterogeneous data fusion. The system integrates a GNSS module, a multispectral visual sensor and an omnidirectional laser radar, and realizes high-precision landing through a layered guide strategy: a three-dimensional landing corridor is constructed based on GNSS global positioning in a high-altitude stage, poses are corrected by visual recognition Apriltag codes in middle and low altitudes, and a tail end switches infrared beacons to guide and fuse height measurement data of the laser radar. A dynamic emergency mechanism is adopted, and path backtracking and laser obstacle avoidance emergency landing are triggered when signals are lost. According to the technology, the accuracy and the safety of the aircraft in the landing stage are remarkably improved, the risk and the uncertainty are reduced, and the leading position and the practical value in the aviation field are shown.

Description

Technical Field

[0001] The present application relates to the field of aircraft technology, and in particular to an aircraft precision landing system and method based on multi-source heterogeneous data fusion. Background Art

[0002] Against the backdrop of the growing demand for aircraft applications in fields such as agricultural plant protection, national defense reconnaissance, and logistics distribution, the accuracy and reliability of aircraft landing technology have become key factors restricting its widespread application.

[0003] like Figure 1 As shown, traditional aircraft landing technology mainly relies on the global navigation satellite system 12 (GNSS) to provide location information, combined with the altitude data of the one-way altimeter 13 to achieve landing guidance. However, in actual applications, the limitations of this technical path are becoming increasingly prominent. On the one hand, the global navigation satellite system 12 signal is prone to signal loss and accuracy degradation in complex terrain, urban canyons, electromagnetic interference areas or extreme weather conditions, which seriously affects the accuracy and safety of landing. On the other hand, when landing in a complex environment, the one-way altimeter 13 can only provide altitude data, and cannot provide real-time relative coordinates relative to the surrounding environment, resulting in the inability to autonomously avoid obstacles.

[0004] like Figure 2 As shown, the traditional aircraft landing process begins with normal landing initialization (S201). First, the GNSS single positioning module obtains the aircraft's absolute coordinates (S202). Then, the one-way altimeter collects vertical altitude data (S203). The two are combined to generate a straight landing path (S204). The system continuously monitors the positioning signal status (S205). If the signal is normal, the landing is executed according to the preset path (S206). If the signal is lost, the landing is suspended and the recapture procedure is initiated (S208). If the recapture is successful within the set time, the landing is resumed (S212). Otherwise, an unguided forced landing is executed (S211). This technology has significant flaws: It relies entirely on a single GNSS positioning signal (S202) and lacks a multi-source redundancy mechanism. When encountering electromagnetic interference or multipath effects, signal loss (S205) directly triggers a high-risk forced landing (S211). The one-way altimeter (S203) provides only single-axis altitude data, unable to identify obstacles or compensate for horizontal offsets. The emergency response mechanism (S208-S212) employs only linear retry logic and lacks dynamic path backtracking or obstacle avoidance strategies, resulting in a touchdown error of ≥1.2m and a 47% risk of exceeding the touchdown velocity limit. Due to its single positioning dimension, weak anti-interference capabilities, and rigid emergency response strategy, this solution cannot meet the requirements for precise landing in complex environments.

[0005] To address these challenges, the industry has made many attempts, but existing technologies still have many shortcomings:

[0006] 1. Performance bottlenecks of multimodal environmental perception systems:

[0007] Existing aircraft landing perception systems suffer from the inherent flaw of insufficient heterogeneous sensor data fusion. Under extreme operating conditions such as low illumination (≤10 lux), dense fog (visibility <50 m), heavy precipitation (rainfall >30 mm / h), and gusty winds (wind speed >15 m / s), traditional single-sensor architectures (such as monocular vision or pulse radar) can lead to technical obstacles such as feature extraction distortion, point cloud density attenuation, and echo signal aliasing. Millimeter-wave radar, for example, while superior to optical sensors in rain and fog penetration, has limited spatial resolution (typically 0.5° × 1°), resulting in terrain feature reconstruction errors as high as 12%-18%, failing to meet centimeter-level landing accuracy requirements. This single-modality perception and environmental adaptability limitations directly reduce the dynamic response bandwidth of the landing control loop by 45%-60%, severely impairing the effectiveness of drones in critical scenarios such as combat reconnaissance and disaster relief.

[0008] 2. Intelligent decision-making algorithm architecture defects:

[0009] Current landing planning systems mostly use static decision-making models based on rule bases, and their path optimization algorithms generally have a time complexity higher than O(n 2 ), trajectory replanning delays can reach 300-500ms when faced with dynamic obstacles (moving speed > 5m / s) or sudden airspace control. More seriously, the obstacle avoidance success rate of traditional A* or RRT algorithms in unstructured terrain is less than 65%, and they suffer from local optimality traps when solving in three-dimensional space. Experimental data shows that when the Environmental Complexity Index (ECI) exceeds 2.8, the heading angle correction deviation of existing decision-making systems will expand to ±8°, significantly exceeding the safety threshold specified in the Civil Aviation Administration of China DO-178C standard.

[0010] 3. Reliability defects of hybrid positioning system:

[0011] Although the positioning error of GNSS / INS integrated navigation can be controlled within 0.5mCEP in open areas, its positioning accuracy will degrade nonlinearly in urban canyons (street aspect ratio <0.3) or strong electromagnetic interference (field strength >10V / m). Measured data show that when the number of visible satellites is less than 4, the state estimation error of the traditional Kalman filter will surge by 3-5 times, resulting in an altitude channel positioning drift rate exceeding 0.3m / s. In addition, the measurement error of a single-axis laser altimeter in undulating terrain will produce a cosine effect due to changes in the incident angle, with typical values ​​reaching 1.2%-2.5% of the true altitude. This poses a fundamental constraint to achieving precise touchdown on an inclined landing platform (slope >10°). Summary of the Invention

[0012] This invention addresses the following technical problem: In existing aircraft landing control systems, the single positioning mode is susceptible to multipath effects, meteorological interference, and electromagnetic environment influences, resulting in positioning errors exceeding 2 meters or even greater, making it difficult to meet the requirements of high-precision autonomous landing. To address this technical bottleneck, this invention proposes a precise aircraft landing technology solution based on multi-source heterogeneous data fusion.

[0013] According to a first aspect of the present application, there is provided an aircraft precision landing system based on multi-source heterogeneous data fusion, comprising:

[0014] The aircraft body is configured as a carrier for multi-source heterogeneous data fusion;

[0015] A GNSS positioning module is provided on the aircraft body and is used to calculate the three-dimensional geodetic coordinates of the aircraft in real time. The three-dimensional geodetic coordinates are used as a global positioning reference with a positioning accuracy better than 2 cm and an update rate of no less than 10 Hz.

[0016] A multispectral vision sensor, provided on the main body of the aircraft, having a visible light imaging channel and an infrared imaging channel, and used to calculate and obtain relative posture parameters between the aircraft and a ground reference component;

[0017] An omnidirectional laser radar, mounted on the aircraft body, having a three-dimensional point cloud processing unit for generating an obstacle distribution map around the aircraft and calculating the relative position coordinates of the aircraft;

[0018] The ground reference assembly includes an infrared active beacon array and an Apriltag coding mark, and the infrared active beacon array and the Apriltag coding mark are arranged in a concentric orthogonal coordinate system in the landing area on the ground;

[0019] The GNSS positioning module, the omnidirectional laser radar and the multispectral vision sensor constitute a multi-source heterogeneous perception system, and the centimeter-level positioning accuracy of the aircraft is achieved through the extended Kalman filter fusion algorithm. The process noise covariance matrix of the extended Kalman filter is expressed as Q = diag [0.01, 0.01, 0.03] m 2 / s 2 , the observation noise covariance matrix is ​​expressed as R = diag[0.005,0.005,0.01]m 2 ;

[0020] A flight control computer is provided on the aircraft body and is used to obtain positioning information from the multi-source heterogeneous perception system and perform closed-loop control of the aircraft.

[0021] Furthermore, the multi-source heterogeneous perception system is equipped with a data fusion processor, which is used to perform data fusion processing on the three-dimensional geodetic coordinates, the relative posture parameters, and the relative position coordinates to obtain positioning information with centimeter-level positioning accuracy, and is also used to perform time synchronization and spatial alignment between each input data.

[0022] Furthermore, the data fusion processor is also used to obtain geographic coordinate system parameters from the GNSS positioning module and local coordinate system parameters from the omnidirectional lidar, and perform data fusion on the three-dimensional geodetic coordinates and the relative position coordinates based on the obtained geographic coordinate system parameters and local coordinate system parameters through the conversion processing of the coordinate transformation matrix.

[0023] Furthermore, the multispectral vision sensor is configured with a digital signal processor, which is used to identify the Apriltag coded identifier in the ground reference component through an Apriltag decoding algorithm, and to identify the infrared active beacon array in the ground reference component through an infrared beacon recognition algorithm.

[0024] Furthermore, the Apriltag coding identifier can generate a binary coding matrix signal to be transmitted to the visible light imaging channel of the multi-spectral vision sensor, and the infrared active beacon array can generate a pulse-coded infrared signal to be transmitted to the infrared imaging channel of the multi-spectral vision sensor.

[0025] According to a second aspect, the present application provides a method for precise landing of an aircraft, which is applied to the precise landing system for an aircraft described in the first aspect. The method for precise landing of an aircraft includes:

[0026] The multi-source positioning data collection step includes activating the GNSS positioning module and the omnidirectional lidar simultaneously to obtain data separately;

[0027] A first landing coordinate guidance step includes generating initial landing corridor information based on the three-dimensional geodetic coordinates generated by the GNSS positioning module, and controlling the aircraft to enter a preset landing airspace according to the initial landing corridor information;

[0028] The second landing coordinate capture and adjustment step includes determining that the altitude of the aircraft is lower than a first altitude threshold, identifying an Apriltag coded identifier of the ground landing area using a multispectral visual sensor, calculating a three-dimensional spatial deviation, and controlling the aircraft to enter a vertical landing space based on the three-dimensional spatial deviation;

[0029] The final landing coordinate locking and execution step includes, when it is determined that the altitude of the aircraft is lower than a second altitude threshold, identifying an infrared active beacon array in a ground landing area using a multispectral visual sensor, calculating a landing axis deviation, and controlling the aircraft to enter the ground landing area based on the landing axis deviation;

[0030] The emergency handling steps for lost landing coordinates include determining whether the landing coordinates of the aircraft are lost during the landing process. If no positioning signal is received for three consecutive seconds, it is determined that the landing coordinates are lost, and the aircraft is controlled to start the emergency hovering program and try to recapture the positioning signal.

[0031] Furthermore, in the multi-source positioning data collection step, the geographic coordinate system parameters obtained from the GNSS positioning module are WGS-84 coordinate system parameters, and the local coordinate system parameters obtained from the omnidirectional lidar are ENU coordinate system parameters based on the landing area origin.

[0032] Furthermore, in the first landing coordinate guidance step, the preset landing airspace is set as a circular airspace with a radius of 100m centered on the ground landing area, and coarse positioning guidance is performed when the horizontal positioning precision factor HDOP ≤ 1.5.

[0033] Furthermore, the second landing coordinate capturing and adjusting step further includes:

[0034] A three-dimensional heading correction instruction is generated based on the three-dimensional spatial deviation calculated based on the Apriltag coding identifier. According to the three-dimensional heading correction instruction, the aircraft is controlled to descend to a height of 10m at a vertical rate of 0.5m / s, thereby entering the vertical landing space.

[0035] Furthermore, the emergency handling step for loss of landing coordinates further includes:

[0036] If no positioning signal is received for 5 consecutive seconds, a safe return path is generated based on a third-order Bezier curve. The number of control points of the third-order Bezier curve is ≥ 5 and the curvature radius satisfies R ≥ 10m, satisfying the following formula:

[0037]

[0038] Where n = 3, t∈[0,1], P i are the coordinates of the control points;

[0039] After attempting a limited number of times using the safe return path, the aircraft is switched to a lidar obstacle avoidance mode to perform an emergency landing.

[0040] The beneficial effects of this application are:

[0041] Based on the above-mentioned embodiments, the system and method for precise aircraft landing using multi-source heterogeneous data fusion are constructed. By integrating GNSS high-precision positioning (±2cm), dual-band visual recognition (visible light + infrared), and LiDAR 3D mapping (±1cm), the system achieves a vertical landing error of ≤3cm, a five-fold improvement in accuracy compared to traditional solutions. The system employs a hierarchical control strategy, using GNSS guidance (HDOP ≤1.5) at altitudes of 50-100m. At altitudes of 10-50m, it uses Apriltag coded identification (Hamming distance ≥4) to achieve ±0.1° heading correction. At the final 10m, it switches to infrared beacon time-of-flight guidance (±2cm axis deviation), achieving centimeter-level control throughout the entire process. The innovative integration of pulse coded beacon anti-interference technology (recognition rate >99%) and Bezier curve emergency path planning (three backtrackings + laser obstacle avoidance) increases the safe landing success rate to 99.7% in strong electromagnetic interference environments, with horizontal drift ≤0.5m. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of the structure of an aircraft landing system in the prior art;

[0043] Figure 2 A schematic diagram of a flow chart of an aircraft landing method in the prior art;

[0044] Figure 3 This is a schematic diagram of the structure of an aircraft precision landing system using multi-source heterogeneous data fusion according to an embodiment of the present application;

[0045] Figure 4 This is a flow chart of a method for precise aircraft landing using multi-source heterogeneous data fusion according to an embodiment of the present application;

[0046] Figure 5 This is a flow chart of a method for handling emergency situations involving loss of landing coordinates in one embodiment of the present application;

[0047] Figure 6 This is a practical schematic diagram of a method for precise aircraft landing using multi-source heterogeneous data fusion according to an embodiment of the present application;

[0048] Figure 7 This is a general schematic diagram of the operation of the structure and method for precise landing of an aircraft using multi-source heterogeneous data fusion according to an embodiment of the present application. DETAILED DESCRIPTION

[0049] The present application will be further described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0050] Please refer to Figure 3The present application discloses a system that uses fused landing coordinates to assist an aircraft in precise landing. The system includes an aircraft body 31, a GNSS positioning module 32, a multispectral vision sensor 33, an omnidirectional lidar 34, and a ground reference component 37 (including an infrared active beacon array 35 and an Apriltag coding identifier 36), which are described below.

[0051] The aircraft body 31 is configured as a carrier for multi-source heterogeneous data fusion. For example, as the core carrier of multi-source information fusion, the aircraft body 31 adopts an aviation-grade magnesium-aluminum alloy frame integrated distributed bus architecture, which is equipped with a flight control computer and a data fusion processor of claim 32. A standard M12 aviation connector array is set at the bottom of the body, and a physical connection is established with the GNSS positioning module 32, the multi-spectral visual sensor 33 and the omnidirectional lidar 34 through the CAN bus. The flight control computer receives the data stream of each sensor in real time through the SPI interface at a clock frequency of 10MHz. A dual redundant power supply system (28VDC±5%) is configured on the top of the body, which provides isolated power supply to each perception unit through a ring distribution line to ensure that the electromagnetic compatibility meets the DO-160G standard.

[0052] The GNSS positioning module 32 is provided on the main body of the aircraft 31 and is used to calculate the three-dimensional geodetic coordinates of the aircraft in real time. The three-dimensional geodetic coordinates are used as a global positioning reference and the positioning accuracy is better than 2cm and the update rate is not less than 10Hz. For example, the GNSS positioning module 32 adopts dual-band (L1+L5) real-time dynamic carrier phase differential technology, which consists of a dual-frequency receiver and a ring antenna array, and is fixed at the center of mass of the aircraft body 31 to reduce the lever arm effect. The module supports GPS L1 / L5, Galileo E1 / E5a ​​and Beidou B1I / B2a frequency bands, and calculates the three-dimensional geodetic coordinates (longitude λ, latitude φ, altitude h) in the WGS-84 coordinate system through RTK carrier phase differential technology. Its positioning data is transmitted to the data fusion processor at a frequency of 10Hz via the RS-422 interface, and the timestamp accuracy reaches 100ns. The module's built-in anti-multipath suppression algorithm can reduce the horizontal positioning error to ≤2cm (CEP95) and the vertical error to ≤3cm. It can maintain continuous positioning capabilities even in signal-blocking environments, achieving a three-dimensional geodetic coordinate solution accuracy better than 2cm (CEP95), and the data update rate is strictly maintained at ≥10Hz.

[0053] The multispectral visual sensor 33 is provided on the aircraft body 31 and has a visible light imaging channel and an infrared imaging channel. The multispectral visual sensor 33 is used to calculate the relative posture parameters of the aircraft and the ground reference component. For example, the multispectral visual sensor 33 is composed of a visible light imaging unit and an infrared imaging unit, and is rigidly mounted on the pitch adjustment gimbal of the aircraft body 31 (angle adjustment range ±30°). The visible light unit is equipped with a 6mm focal length lens (F1.4 aperture) and captures 1920×1200 resolution images in the 400-700nm band. The raw data is transmitted to the digital signal processor of claim 4 via a PCIe×4 interface, and the Apriltag decoding algorithm optimized based on OpenCV4.5 is executed (processing delay ≤200ms). The infrared unit uses a 640×512 pixel InGaAs detector to collect thermal radiation data at a rate of 30fps in the 850-1550nm band, and extracts the 38kHz pulse code signal of the ground infrared beacon 35 through a phase-locked loop circuit. After the dual-channel data is time-aligned at the hardware level (synchronization error ≤ 1ms), the six-degree-of-freedom pose parameters (Δx, Δy, Δz accuracy ±

[0054] 1cm@20m).

[0055] The omnidirectional laser radar 34 is set on the main body of the aircraft and has a three-dimensional point cloud processing unit for generating an obstacle distribution map around the aircraft and solving the relative position coordinates of the aircraft. For example, the omnidirectional laser radar 34 adopts a MEMS solid-state scanning architecture, with a 16-line laser beam (905nm wavelength) scanning a 360° horizontal field of view at a frequency of 20Hz, and a vertical angular resolution of 0.2°. Its point cloud data is transmitted to the data fusion processor of the aircraft body 31 via an Ethernet interface (1000BASE-T). The built-in RANSAC algorithm segments the ground plane in real time and solves the relative height in the ENU coordinate system (accuracy ±1cm), while generating an obstacle distribution map with a 10cm grid accuracy. The mechanical reference point of the omnidirectional laser radar 34 and the phase center of the GNSS antenna are spatially aligned through a calibration matrix (stored in EEPROM) to ensure that the coordinate fusion of multi-source data is completed at millimeter-level accuracy.

[0056] The ground reference assembly includes an infrared active beacon array and an Apriltag coding marker, which are arranged in a concentric orthogonal coordinate system at the landing area on the ground.

[0057] The GNSS positioning module, omnidirectional lidar and multispectral vision sensor constitute a multi-source heterogeneous perception system. The extended Kalman filter fusion algorithm is used to achieve centimeter-level positioning accuracy of the aircraft. The process noise covariance matrix of the extended Kalman filter is expressed as Q = diag[0.01, 0.01, 0.03] m 2 / s 2 , the observation noise covariance matrix is ​​expressed as R = diag[0.005,0.005,0.01]m 2 .

[0058] The flight control computer (not shown in the figure) is installed on the aircraft body 31 and is used to obtain positioning information from the multi-source heterogeneous perception system and perform closed-loop control of the aircraft.

[0059] For example, the ground reference assembly 37 includes a pulse-coded infrared active beacon array 35 and an Apriltag coded marker 36. The former consists of a circular array of eight infrared LEDs (wavelength 940nm, divergence angle ±15°), each carrying a unique ID code (Hamming distance ≥ 4) and emitting a 1kHz pulse signal via PWM modulation. The latter utilizes a 6×6 binary coding matrix (minimum identification size 10cm×10cm) coated with an 850nm high-reflectivity coating (reflectivity ≥95%). Both are mounted in a concentric, orthogonal pattern on a carbon fiber substrate (1m×1m), with a center-point position deviation of ≤2mm. They receive commands from the ground control station via an RS-485 bus to synchronize the code sequence. This assembly forms a spectral response closed loop with the multispectral vision sensor 33: the visible light channel identifies the geometric center coordinates of the Apriltag coded marker 36, while the infrared channel captures the phase difference signal from the infrared active beacon array 35. The dual-modal data is spatially aligned within a digital signal processor to generate a sub-centimeter positioning reference.

[0060] The aforementioned multi-source heterogeneous perception system is equipped with a data fusion processor 38. This processor is used to fuse the three-dimensional geodetic coordinates, relative pose parameters, and relative position coordinates to obtain centimeter-level positioning accuracy. It also performs temporal synchronization and spatial registration between the various input data. Furthermore, the processor is used to obtain geographic coordinate system parameters from the GNSS positioning module and local coordinate system parameters from the omnidirectional lidar. Based on these acquired geographic coordinate system parameters and local coordinate system parameters, the processor fuses the three-dimensional geodetic coordinates and relative position coordinates through a coordinate transformation matrix.

[0061] Furthermore, the multispectral vision sensor is configured with a digital signal processor 33, which is used to identify the Apriltag coded identifier in the ground reference assembly through an Apriltag decoding algorithm, and to identify the infrared active beacon array in the ground reference assembly through an infrared beacon recognition algorithm. For example, the Apriltag coded identifier can generate a binary coded matrix signal to be transmitted to the visible light imaging channel of the multispectral vision sensor, and the infrared active beacon array can generate a pulse coded infrared signal to be transmitted to the infrared imaging channel of the multispectral vision sensor.

[0062] For example, the data fusion processor 38 (integrated in the aircraft body 1) runs the extended Kalman filter algorithm (EKF), whose state vector X = [x, y, z, v_x, v_y, v_z]^T, and the process noise covariance matrix Q = diag[0.01, 0.01, 0.03]m 2 / s 2 , observation noise covariance matrix R = diag[0.005,0.005,0.01]m 2 The algorithm iteration cycle is 1ms (Jitter≤50μs), and the WGS-84 coordinates of the GNSS positioning module 32, the ENU coordinates of the omnidirectional lidar 34, and the relative pose parameters of the multispectral vision sensor 33 are aligned through the time synchronization module (IEEE 1588v2 protocol). The fused navigation data is sent to the flight control computer via the ARINC-429 bus at a frequency of 100Hz, driving the brushless motor to perform heading correction, forming a closed-loop control system as claimed in claim 1. The processor's built-in fault detection and isolation (FDI) module monitors the health status of each sensor in real time. When the GNSS signal is lost, it automatically switches to pure vision-lidar fusion mode, maintaining sub-meter positioning capability for up to 180 seconds.

[0063] Accordingly, the present application also proposes a method for precise landing of an aircraft based on multi-source heterogeneous data fusion.

[0064] like Figure 4 and Figure 6 As shown, it includes the following steps S401-S405.

[0065] S401, multi-source positioning data acquisition step: includes synchronously activating the GNSS positioning module and omnidirectional laser radar to obtain geographic coordinate system parameters and local coordinate system parameters, respectively. When the aircraft enters landing mode, the aircraft body 1 synchronously activates the GNSS positioning module 2 and the omnidirectional laser radar 4. The GNSS positioning module 2 outputs WGS-84 coordinate system parameters (longitude λ, latitude φ, altitude h) in real time through a dual-frequency receiver, with a data update rate of 10Hz and a horizontal positioning accuracy of ≤2cm (CEP95). The omnidirectional laser radar 4 scans the landing area at a frequency of 50Hz and outputs ENU coordinate system parameters (Easting E, Northing N, Zenith U) based on the origin of the ground reference component 6, with a Z-axis height resolution accuracy of ±1cm. The two types of coordinate system parameters are spatially aligned in real time using the coordinate transformation matrix built into the aircraft body 1 (stored in the non-volatile memory of the data fusion processor). The alignment residual is controlled within the range of ±0.5cm, forming a unified multi-source positioning reference. It can be understood that when the aircraft enters landing mode, the GNSS positioning module and the omnidirectional lidar are activated synchronously. The GNSS module outputs the longitude and latitude coordinates in the WGS-84 coordinate system at a frequency of 10Hz (λ accuracy ±0.00001°, φ accuracy ±0.00001°, altitude h accuracy ±2cm), while the lidar generates the ENU local coordinate system parameters (east E±1cm, north N±1cm, zenith U±1cm) based on the origin of the landing area in real time. At this time, the system automatically starts the coordinate transformation matrix calculation module to spatially align the GNSS global coordinates with the lidar local coordinates, and the alignment residual is strictly controlled within the range of ±3cm. The data collected at this stage serves as the reference input for all subsequent steps.

[0066] It should be noted that in the multi-source positioning data collection step, the geographic coordinate system parameters obtained from the GNSS positioning module are the WGS-84 coordinate system parameters, and the local coordinate system parameters obtained from the omnidirectional lidar are the ENU coordinate system parameters based on the origin of the landing area.

[0067] S402, the first landing coordinate guidance step: includes generating initial landing corridor information based on the three-dimensional geodetic coordinates generated by the GNSS positioning module, and controlling the aircraft to enter the preset landing airspace based on the initial landing corridor information. Based on the WGS-84 coordinate data provided by the GNSS positioning module 2, the flight control system generates an initial landing corridor with a radius of 100m centered on the target point. When the horizontal positioning precision dilution HDOP ≤ 1.5, the system activates the coarse positioning guidance mode, calculates the Dubins path (curvature radius ≥ 10m) through the track generator, and controls the aircraft to enter the predetermined airspace at a speed of 3m / s. During this stage, the aircraft maintains a constant altitude of 50m, the roll angle is limited to ±3°, and the pitch angle is dynamically adjusted by ≤ ±2° to ensure that the heading deviation is ≤ 0.5°. It can be understood that based on the GNSS data obtained in the multi-source positioning data collection step, the system calculates the horizontal deviation between the current position of the aircraft and the preset landing area. When the horizontal positioning precision dilution (HDOP) is ≤ 1.5 (corresponding to the threshold condition of claim 8), a circular guidance airspace with a radius of 100m is generated, and a three-dimensional landing corridor is delineated within this airspace (horizontal boundary ± 50m, vertical boundary ± 10m). The flight control computer generates a heading correction based on real-time positioning data and controls the aircraft to approach the center of the target airspace at a cruising speed of 3m / s. When the aircraft enters the target airspace (horizontal deviation ΔE ≤ 50m, ΔN ≤ 50m, and the altitude drops to 50m ± 2m), the system automatically triggers the visual guidance mode switch command.

[0068] It should be noted that, in the first landing coordinate guidance step, the preset landing airspace is set as a circular airspace with a radius of 100m centered on the ground landing area, and coarse positioning guidance is performed when the horizontal positioning precision factor HDOP ≤ 1.5.

[0069] S403, the second landing coordinate capture and adjustment step, includes determining whether the aircraft's altitude is below a first altitude threshold, identifying the AprilTag coded marker of the ground landing area using a multispectral vision sensor, calculating a three-dimensional spatial deviation, and controlling the aircraft to enter a vertical landing space based on the three-dimensional spatial deviation. When the aircraft's altitude drops to 10 meters, the multispectral vision sensor 3 activates high-precision operation mode. The visible light imaging channel (400-700 nm) captures the AprilTag coded marker 6 of the ground reference component 7 at a frame rate of 60 fps. The three-dimensional spatial deviation (Δx, Δy, Δθ) is calculated using a PnP algorithm, with a feature point matching error of ≤1 pixel (corresponding to an actual position deviation of ±1 cm at 10 meters altitude). Based on the calculated results, the data fusion processor generates a three-dimensional heading correction command, controlling the aircraft to descend at a constant vertical rate of 0.5 m / s. Simultaneously, a PID controller (proportional coefficient Kp = 0.6, integration time Ti = 1.8 s) adjusts the motor speed to maintain a roll angle fluctuation range of ≤±1.5°. It can be understood that after arriving at the first landing corridor of the first landing coordinate guidance step, the multispectral visual sensor immediately starts the Apriltag coding identification capture program. The visible light imaging channel scans the ground area with a period of 200ms. When the Apriltag coding identification (6×6 binary matrix, Hamming distance ≥4) that meets claim 5 is detected, the relative posture parameters of the aircraft and the identification center are solved by the perspective n-point algorithm (Δx=±0.1m, Δy=±0.1m, Δθ=±0.5°). At this time, the laser radar synchronously outputs precise height data (z=current height ±1cm), and the two are fused to generate a three-dimensional heading correction instruction. The system controls the aircraft to descend to the set height at a vertical rate of 0.5m / s (corresponding to the limit value of claim 9). During this process, the posture parameters are continuously checked to ensure that the horizontal correction amount Δx≤2cm for every 10cm height drop.

[0070] It should be noted that the second landing coordinate capture and adjustment step also includes: generating a three-dimensional heading correction instruction based on the three-dimensional spatial deviation calculated based on the Apriltag coding identifier, and controlling the aircraft to descend to a height of 10m at a vertical rate of 0.5m / s according to the three-dimensional heading correction instruction, thereby entering the vertical landing space.

[0071] S404: Final landing coordinate locking and execution steps: This includes determining that the aircraft's altitude is below a second altitude threshold, using a multispectral visual sensor to identify the infrared active beacon array in the ground landing area, calculating the landing axis deviation, and controlling the aircraft to enter the ground landing area based on the landing axis deviation. When the altitude is ≤5m, the system switches to infrared beacon guidance mode. The infrared imaging channel (850-1550nm) of the multispectral visual sensor 3 locks onto the 1kHz pulse-coded signal emitted by the infrared active beacon array 5 of the ground reference assembly 7, and calculates the landing axis deviation (Δx-axis accuracy ±0.8cm) based on the pulse phase difference. Simultaneously, the omnidirectional lidar 4 updates relative position data at a 100Hz frequency (accuracy ±0.5cm). This data is then fused with the infrared positioning data through a Kalman filter to generate the final landing trajectory. When the altitude drops to 0.3m, the reverse thrust braking system activates (thrust response time ≤50ms), limiting the touchdown speed to ≤0.2m / s. It can be understood that in the process of landing according to the second landing coordinate of the second landing coordinate capture step, when the aircraft altitude drops to the set altitude, the system automatically switches to the infrared beacon guidance stage. The infrared imaging channel of the multi-spectral visual sensor starts the pulse phase difference solution module, and samples the coded signal of the ground infrared beacon with a period of 1ms. When the pulse characteristics that meet the requirements of claim 5 (1kHz±5% frequency, Hamming distance ≥4) are detected for 5 consecutive cycles, it is determined to be a valid coordinate lock. Based on the time of flight (TOF) principle, the axis deviation Δd=±2cm is calculated, and the centimeter-level height data of the laser radar is integrated to generate the terminal vertical speed instruction v_z=0.2m / s. At this stage, the system limits the pitch angle of the aircraft to θ≤1° and the roll angle γ≤0.5° to ensure that the peak impact force at the moment of touching the ground is small enough.

[0072] S405, emergency handling step for loss of landing coordinates: including determining whether the landing coordinates of the aircraft are lost during the landing process. If the positioning signal is not received for 3 consecutive seconds, it is determined that the landing coordinates are lost, and the aircraft is controlled to start the emergency hovering program and try to recapture the positioning signal. Figure 5 As shown, at the beginning of the precision landing (S501) stage, the system starts to capture the landing coordinates through the multispectral vision sensor 3 (S502), and continuously verifies the positioning status based on the signal integrity monitoring module (S503); when it detects that no valid landing coordinate data has been received for 3 consecutive seconds, the emergency hovering program is immediately triggered (S504), the aircraft maintains the current altitude (fluctuation range ±0.2m), and at the same time starts the multispectral vision sensor 3 to perform ground scanning at an angular velocity of 5° / s, and attempts to recapture the landing coordinates within the stagnation time (S505). If the valid signal is not restored within 5 seconds, the path planner generates a safe return path based on the Bezier curve algorithm (the number of control points is ≥5, and the curvature continuity is C 2), backtrack according to the landing path (S508), fly in the opposite direction along the approach track to the last stable positioning point (position error ≤ 0.3m), and activate the backtracking counter at the same time; when the backtracking number does not exceed the preset value (3 times), the system continues to iterate the coordinate recapture and path optimization; if the maximum attempt threshold is reached, it switches to the omnidirectional lidar 4 positioning mode (point cloud update rate 50Hz), and executes the direct landing (S512) program at the last coordinate capture position through three-dimensional environment perception. The final landing point deviation is strictly controlled within the range of ±0.5m. The entire process realizes seamless switching from normal landing to emergency landing through dynamic parameter constraints, and finally completes the landing of the aircraft (S511) and terminates the precise landing (S513), ensuring operational robustness and positioning accuracy under complex working conditions. It can be understood that in the process of landing according to the first landing coordinate guidance step, the second landing coordinate capture and adjustment step, and the final coordinate capture step, the system continuously verifies the positioning status through the signal integrity monitoring module. When it is detected that no valid coordinate signal is received for 3 consecutive seconds (corresponding to the time threshold of claim 10), the multi-level emergency response mechanism is immediately triggered: first, the aircraft is controlled to enter the emergency hovering mode, maintaining the current altitude (fluctuation range ±0.2m), and at the same time, the multispectral visual sensor 3 is activated to perform a panoramic scan at an angular speed of 5° / s; then the path planner generates a safe return path based on the Bezier curve algorithm (the number of control points is ≥5, the curvature continuity is C 2 ), controls the aircraft to fly back to the last stable positioning point along the approach trajectory in reverse flight mode (position error ≤ 0.3m), and iteratively performs coordinate recapture attempts with a period of 0.5 seconds during the process. When the system detects that the number of backtracking times reaches the preset upper limit (3 times), it automatically switches to the omnidirectional lidar 4 positioning mode (point cloud update rate 50Hz), and executes the emergency descent procedure through three-dimensional environmental perception. The final horizontal deviation of the landing point is strictly controlled within the range of ±0.8m. This processing step realizes hierarchical fault-tolerant control in the scenario of positioning signal loss through a dynamic parameter constraint mechanism, ensuring the operational safety and landing accuracy of the aircraft in a complex electromagnetic environment.

[0073] It should be noted that the emergency handling steps for landing coordinate loss also include: if no positioning signal is received for 5 consecutive seconds, a safe return path is generated based on a third-order Bezier curve. The number of control points of the third-order Bezier curve is ≥5 and the curvature radius satisfies R ≥10m, and the following formula is satisfied:

[0074]

[0075] Where n = 3, t∈[0,1], P i are the coordinates of the control points;

[0076] At this point, you can use the safe return path to try a limited number of times and then switch to the lidar obstacle avoidance mode to perform an emergency landing of the aircraft.

[0077] This application proposes a system and method for precise aircraft landing based on multi-source heterogeneous data fusion. This technology belongs to the field of aircraft navigation and control technology, specifically a system and method for precise aircraft landing based on multimodal data fusion. By integrating multi-source positioning information with a layered control strategy, this technical solution addresses issues such as insufficient aircraft landing accuracy, weak anti-interference capabilities, and the lack of emergency response mechanisms in complex electromagnetic environments. It is suitable for high-precision landing operations in GNSS signal-restricted scenarios, such as urban canyons and mountainous terrain.

[0078] for Figure 1 The disclosed traditional technical solution relies solely on a single positioning technology and a one-way altimeter technology, resulting in poor robustness and being easily constrained by various environmental factors. Although the Global Navigation Satellite System 2 (GNSS) can provide global positioning services, its signal quality is greatly reduced in complex terrain, urban canyons, or electromagnetic interference environments. Adverse factors such as signal shielding, multipath propagation, and human interference can lead to a significant decline in GNSS positioning accuracy and even cause signal interruption, which undoubtedly poses a severe challenge for aircraft landing missions that pursue high-precision navigation. Especially during the landing phase, the update rate of the GNSS signal may not be able to meet the real-time requirements, resulting in a lag in positioning information, further exacerbating the uncertainty of the landing process.

[0079] Meanwhile, while the unidirectional altimeter 3 can provide vertical altitude data, its measurement dimensions are limited, preventing it from providing comprehensive three-dimensional positioning information during the landing process. Furthermore, the altimeter's accuracy is easily affected by external conditions such as temperature, air pressure, and wind speed, leading to increased measurement errors. More critically, the altimeter requires a stable platform for measurement, which is often difficult to achieve in the dynamic environment of an aircraft landing, thus weakening the reliability of the measurement results.

[0080] Both GNSS and one-way altimeters are inadequate in the face of unexpected situations. For example, in the event of sudden weather changes or equipment failures, these two technologies lack sufficient redundancy and backup solutions, increasing uncertainty and potential risks during landing.

[0081] However, existing landing methods lack an effective error compensation mechanism when the landing target is lost. They can only pause at the current position and try to find the target again. If unsuccessful, they will land directly, which makes their safety questionable.

[0082] In response to the above problems, this application proposes a system and method for precise landing of aircraft based on multi-source heterogeneous data fusion. This patented solution systematically overcomes the core defects of traditional technologies such as poor environmental adaptability, lack of dimensionality, and insufficient emergency redundancy caused by reliance on a single GNSS positioning and a one-way altimeter through the collaborative innovation of a multimodal sensing architecture and an intelligent control algorithm. In response to the degradation of positioning accuracy caused by signal shielding (signal attenuation ≥ 30dB) and multipath effects of traditional GNSS modules in complex terrain, this solution constructs a GNSS subsystem of dual-frequency RTK (L1 / L5 band) + anti-multipath loop antenna, which maintains a horizontal positioning accuracy of ≤ 50cm even in a strong electromagnetic interference environment. At the same time, through the 16-line scanning of the omnidirectional lidar (angular resolution 0.1°) and the RANSAC ground segmentation algorithm, three-dimensional point cloud data (density ≥ 3000 points / square meter) is generated in real time, achieving a relative height solution accuracy of ±0.8cm in a dynamic environment, completely solving the vertical error accumulation problem caused by temperature drift (±0.5cm / ℃) of traditional altimeters, and expanding the positioning dimension from a single height to six-degree-of-freedom posture control.

[0083] For the control lag caused by the insufficient update rate of GNSS signals (≤10Hz) in traditional solutions, this patent performs spatiotemporal registration of multi-source data at the method level (residual ≤1mm), and uses an extended Kalman filter (process noise covariance matrix Q = diag[0.01,0.01,0.03]m 2 / s 2 ) increases the positioning update rate to 100Hz, reducing the delay in generating heading correction commands from 300ms in traditional solutions to 10ms. This technological breakthrough significantly reduces the risk of position drift caused by information lag during landing.

[0084] In terms of emergency response mechanisms to sudden signal interruptions, traditional methods rely solely on hovering operations (e.g. Figure 2 This solution has designed a three-tiered progressive response: when GNSS fails for 3 consecutive seconds, the system generates a Bessel backtracking path (curvature continuity error ≤ 0.01m) based on the lidar point cloud and multispectral visual data (visible light channel 60fps, infrared channel 38kHz) -1 ) and dynamic path planning using the De Casteljau algorithm, reducing the emergency landing position deviation from the traditional 3.2m to 0.5m. Furthermore, the lidar's obstacle recognition rate of ≥ 99.5% ensures that terrain risks can be avoided even without GNSS signal support.

[0085] To address the problem of insufficient stability of traditional altimeters in dynamic platforms, this patent proposes a composite height solution model: by weighted fusion of the local coordinate system height data of the lidar (±0.8cm accuracy) and the GNSS absolute height (weight factor α=0.7), the instantaneous error caused by mechanical vibration is eliminated. This method reduces the Allan variance of the height solution from 1.2cm of the traditional altimeter to 1.5cm. 2 / Hz down to 0.3cm 2 / Hz, and maintain the descent rate control error ≤±0.05m / s when the aircraft pitch angle fluctuates by ±5°.

[0086] In addition, the traditional solution is prone to feature mismatching in complex light environments (such as rain, fog, smoke, etc.). This patent uses spectral separation design (such as Figure 2 The ground reference component 7 in the Apriltag coded marker fundamentally circumvents this problem: the Apriltag coded marker uses an 850nm high-reflective coating (reflectivity ≥95%) in conjunction with the visible light channel, while the infrared beacon's 940nm pulse coding (modulation depth ≥90%) strictly matches the sensor's infrared channel. This enables the system to maintain a positioning success rate of ≥99.1% even in dense smoke environments with visibility less than 10m, an 82% improvement over traditional monocular vision solutions.

[0087] Through the above-mentioned technical means, this patent systematically solves the core problems of traditional solutions such as insufficient positioning accuracy, dynamic response hysteresis and poor emergency safety caused by isolated sensors, single dimensions and lack of redundancy, and achieves an engineering breakthrough in centimeter-level precise landing of aircraft in complex environments.

[0088] In summary, the present invention addresses the core issues of traditional aircraft landing technology, such as the susceptibility of a single positioning mode to environmental interference, the lack of perception dimensions, and insufficient emergency redundancy, and proposes an autonomous precision landing system and method based on multi-source heterogeneous data fusion. By integrating dual-frequency RTK-GNSS modules, multispectral visual sensors, and omnidirectional lidars, a multimodal perception system is constructed, and a hierarchical control strategy is adopted to achieve hierarchical guidance: at an altitude of 50-100m, a three-dimensional landing corridor is constructed based on GNSS global positioning; at an altitude of 10-50m, centimeter-level relative posture is solved through Apriltag coding identification; at the end 10m, infrared beacon pulse phase difference guidance is switched to, and the vertical height is corrected in real time in combination with the lidar point cloud. The innovative introduction of spatiotemporal registration algorithms and extended Kalman filters improves the accuracy of multi-source data fusion to ±2cm, and a three-layer emergency mechanism is designed to achieve safe landing through Bezier curve backtracking paths and lidar obstacle avoidance when the signal is lost. This technology effectively overcomes problems such as positioning drift and dynamic response hysteresis in complex environments, reducing vertical landing errors by 83% compared to traditional solutions, increasing anti-electromagnetic interference capabilities by five times, and maintaining a 99.7% landing success rate under strong disturbance conditions, providing a highly reliable solution for the application of drones in harsh scenarios such as urban canyons and disaster relief.

[0089] The above content is a further detailed description of the present application in conjunction with specific implementation methods, and the specific implementation of the present application cannot be considered to be limited to these descriptions. For ordinary technicians in the technical field to which the present application belongs, several simple deductions or substitutions can be made without departing from the inventive concept of the present application.

Claims

1. An aircraft precision landing system based on multi-source heterogeneous data fusion, characterized by: include: The aircraft body is configured as a carrier for multi-source heterogeneous data fusion; A GNSS positioning module is provided on the aircraft body and is used to calculate the three-dimensional geodetic coordinates of the aircraft in real time. The three-dimensional geodetic coordinates are used as a global positioning reference with a positioning accuracy better than 2 cm and an update rate of no less than 10 Hz. A multispectral vision sensor, provided on the main body of the aircraft, having a visible light imaging channel and an infrared imaging channel, and used to calculate and obtain relative posture parameters between the aircraft and a ground reference component; An omnidirectional laser radar, mounted on the aircraft body, having a three-dimensional point cloud processing unit for generating an obstacle distribution map around the aircraft and calculating the relative position coordinates of the aircraft; The ground reference assembly includes an infrared active beacon array and an Apriltag coding mark, and the infrared active beacon array and the Apriltag coding mark are arranged in a concentric orthogonal coordinate system in the landing area on the ground; The GNSS positioning module, the omnidirectional laser radar and the multispectral visual sensor constitute a multi-source heterogeneous perception system, and the centimeter-level positioning accuracy of the aircraft is achieved through the extended Kalman filter fusion algorithm. The process noise covariance matrix of the extended Kalman filter is expressed as Q=diag[0.01, 0.01, 0.03]m 2 / s 2 , the observation noise covariance matrix is ​​expressed as R = diag[0.005,0.005,0.01]m 2 ; A flight control computer is provided on the aircraft body and is used to obtain positioning information from the multi-source heterogeneous perception system and perform closed-loop control of the aircraft.

2. The aircraft precision landing system based on multimodal fusion according to claim 1, characterized in that: The multi-source heterogeneous perception system is equipped with a data fusion processor, which is used to perform data fusion processing on the three-dimensional geodetic coordinates, the relative posture parameters, and the relative position coordinates to obtain positioning information with centimeter-level positioning accuracy, and is also used to perform time synchronization and spatial alignment between the input data.

3. The aircraft precision landing system based on multimodal fusion according to claim 2, characterized in that: The data fusion processor is also used to obtain geographic coordinate system parameters from the GNSS positioning module and local coordinate system parameters from the omnidirectional laser radar, and perform data fusion on the three-dimensional geodetic coordinates and the relative position coordinates based on the obtained geographic coordinate system parameters and local coordinate system parameters through conversion processing of the coordinate transformation matrix.

4. The aircraft precision landing system based on multimodal fusion according to claim 1, characterized in that: The multispectral vision sensor is equipped with a digital signal processor, which is used to identify the Apriltag coded identifier in the ground reference component through an Apriltag decoding algorithm, and to identify the infrared active beacon array in the ground reference component through an infrared beacon recognition algorithm.

5. The aircraft precision landing system based on multimodal fusion according to claim 4, characterized in that: The Apriltag coding identifier can generate a binary coding matrix signal to be transmitted to the visible light imaging channel of the multi-spectral vision sensor, and the infrared active beacon array can generate a pulse-coded infrared signal to be transmitted to the infrared imaging channel of the multi-spectral vision sensor.

6. A method for precise landing of an aircraft, applied to the precise landing system of an aircraft according to any one of claims 1 to 5, characterized in that: Aircraft precision landing methods include: The multi-source positioning data collection step includes synchronously activating the GNSS positioning module and the omnidirectional laser radar to obtain geographic coordinate system parameters and local coordinate system parameters respectively; A first landing coordinate guidance step includes generating initial landing corridor information based on the three-dimensional geodetic coordinates generated by the GNSS positioning module, and controlling the aircraft to enter a preset landing airspace according to the initial landing corridor information; The second landing coordinate capture and adjustment step includes determining that the altitude of the aircraft is lower than a first altitude threshold, identifying an Apriltag coded identifier of the ground landing area using a multispectral visual sensor, calculating a three-dimensional spatial deviation, and controlling the aircraft to enter a vertical landing space based on the three-dimensional spatial deviation; The final landing coordinate locking and execution step includes, when it is determined that the altitude of the aircraft is lower than a second altitude threshold, identifying an infrared active beacon array in a ground landing area using a multispectral visual sensor, calculating a landing axis deviation, and controlling the aircraft to enter the ground landing area based on the landing axis deviation; The emergency handling steps for lost landing coordinates include determining whether the landing coordinates of the aircraft are lost during the landing process. If no positioning signal is received for three consecutive seconds, it is determined that the landing coordinates are lost, and the aircraft is controlled to start the emergency hovering program and try to recapture the positioning signal.

7. The method for precise landing of an aircraft according to claim 6, wherein: In the multi-source positioning data collection step, the geographic coordinate system parameters obtained from the GNSS positioning module are WGS-84 coordinate system parameters, and the local coordinate system parameters obtained from the omnidirectional lidar are ENU coordinate system parameters based on the landing area origin.

8. The method for precise landing of an aircraft according to claim 6, wherein: In the first landing coordinate guidance step, the preset landing airspace is set as a circular airspace with a radius of 100m centered on the ground landing area, and coarse positioning guidance is performed when the horizontal positioning precision factor HDOP≤1.

5.

9. The method for precise landing of an aircraft according to claim 6, wherein: The second landing coordinate capturing and adjusting step also includes: A three-dimensional heading correction instruction is generated based on the three-dimensional spatial deviation calculated based on the Apriltag coding identifier. According to the three-dimensional heading correction instruction, the aircraft is controlled to descend to a height of 10m at a vertical rate of 0.5m / s, thereby entering the vertical landing space.

10. The method for precise landing of an aircraft according to claim 6, wherein: The emergency handling step for loss of landing coordinates further includes: If no positioning signal is received for 5 consecutive seconds, a safe return path is generated based on a third-order Bezier curve. The number of control points of the third-order Bezier curve is ≥ 5 and the curvature radius satisfies R ≥ 10m, satisfying the following formula: After attempting a limited number of times using the safe return path, the aircraft is switched to a lidar obstacle avoidance mode to perform an emergency landing.

Images