An Autonomous Navigation and Deception Defense System and Method for Unmanned Aerial Vehicles in GPS Denial Environments

By employing a tightly coupled navigation mode that combines visual/laser SLAM and geomagnetic vector field matching, the autonomous navigation problem of UAVs when GPS signals fail or are spoofed is solved, achieving high-precision, long-duration autonomous flight and spoofing defense capabilities.

CN122085308APending Publication Date: 2026-05-26BEIJING YIFANG INTELLIGENT TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610418399.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-31
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Existing drone navigation systems cannot achieve long-term, highly reliable autonomous navigation when GPS signals fail or are spoofed, and are easily affected by multi-sensor information contamination, leading to positioning errors.

Method used

The navigation mode employs a tightly coupled approach of visual/laser SLAM, geomagnetic vector field matching, and high-precision IMU. Through a deception detection and isolation module, multi-dimensional consistency checks and error state Kalman filters are used, combined with geomagnetic deviation terms, to fuse navigation data, achieving seamless switching and continuous navigation.

Benefits of technology

It provides strong anti-spoofing capabilities, high-precision continuous navigation, expands applicable scenarios, and ensures the system's self-consistency and security in GPS-denied environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122085308A_ABST
    Figure CN122085308A_ABST
Patent Text Reader

Abstract

This application discloses an autonomous navigation and deception defense system and method for unmanned aerial vehicles (UAVs) in GPS-denied environments. The system includes a multi-source sensor array, and parallel deception detection and isolation threads and a tightly coupled multi-source fusion navigation thread. The deception detection and isolation thread performs multi-dimensional consistency checks on received navigation data. When a dual comparison of geomagnetic fingerprints indicates a poor match between GPS position and geomagnetic readings, but a good match between visual position and geomagnetic readings, the GPS data is isolated from the fusion module. The tightly coupled fusion navigation module employs an error state Kalman filter with an IMU as its core. A geomagnetic deviation term is introduced into the state vector, and the original observation geometric information is directly fed into the filter for tight coupling. The difference between the geomagnetic vector obtained from a geomagnetic vector map query and the measured value from the magnetometer is used for fusion and updating, achieving high-precision continuous navigation in GPS-denied environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of autonomous navigation and flight safety technology for unmanned aerial vehicles (UAVs), and in particular to an autonomous navigation and deception defense system and method for UAVs in GPS denial environments. Background Technology

[0002] Global satellite navigation systems (such as GPS and BeiDou) are the core for drones to achieve high-precision positioning and navigation. However, in canyons, between urban buildings, indoors, or near critical facilities, satellite signals are easily blocked (denied) or subjected to malicious deception attacks, leading to drone positioning errors, loss of control, or even hijacking.

[0003] Existing solutions have significant limitations, as follows:

[0004] For pure inertial navigation systems: dead reckoning is performed using inertial measurement units (IMUs) after GPS failure. The drawbacks include: IMU errors accumulate rapidly over time (drift), resulting in significant positioning deviations within a short period, making them unsuitable for long-duration flights alone.

[0005] For visual / laser SLAM: simultaneous localization and map building are performed by sensing environmental features through cameras or lidar. Drawbacks include: susceptibility to failure in environments with sparse or highly repetitive features (such as deserts, open water, and long straight tunnels); significant interference from dynamic objects; high computational resource consumption; and challenges in scale ambiguity and loop closure detection.

[0006] Using the Earth's magnetic field as a reference has several drawbacks: the geomagnetic field is weak and easily disturbed by ferromagnetic materials in the environment (such as building steel bars and vehicles), causing severe anomalies; traditional single-point geomagnetic matching positioning completely fails in areas of interference.

[0007] For loosely coupled multi-sensor systems, information from GPS, IMU, and vision is simply superimposed. However, this approach has several drawbacks: in GPS spoofing scenarios, incorrect location information can "contaminate" the entire fusion system, leading to system misinterpretation; and each sensor operates independently, failing to form a robust framework for mutual verification.

[0008] Therefore, there is an urgent need for an autonomous navigation system that can automatically shield the influence of unreliable GPS information, and utilize other sensors to build a mutually redundant and mutually verified system, while also possessing the ability to actively detect and resist deception attacks. Summary of the Invention

[0009] Based on this, this application provides an autonomous navigation and deception defense system and method for UAVs in GPS denial environments. When GPS is available, this application uses it for calibration; when the GPS signal quality deteriorates or is deceiving, it can automatically trigger a deception detection and isolation mechanism and seamlessly switch to a tightly coupled fusion navigation mode with "visual / laser SLAM + geomagnetic vector field matching + high-precision IMU" as the core, so as to achieve long-term, highly reliable autonomous flight.

[0010] Firstly, a system for autonomous navigation and deception defense in GPS-denied environments for unmanned aerial vehicles (UAVs) is provided. This system includes a deception detection and isolation module operating in parallel and a tightly coupled multi-source fusion navigation module, wherein:

[0011] The deception detection and isolation module is used to perform multi-dimensional consistency checks on navigation data. When the geomagnetic fingerprint anomaly detection determines that the GPS position-geomagnetic reading match is poor while the visual position-geomagnetic reading match is good, the GPS data is isolated from the fusion unit. The multi-dimensional consistency checks include receiver autonomous integrity monitoring, residual checks between IMU / visual predicted position and GPS position, and geomagnetic fingerprint anomaly detection.

[0012] The tightly coupled multi-source fusion navigation module employs an error state Kalman filter centered on an IMU to introduce a geomagnetic bias term into the state vector. The pixel coordinates of visual feature points, the point-to-surface distance of the laser point cloud, and geomagnetic vector observation data are directly fed into the filter as raw observation geometric information. The filter is then fused and updated by combining the difference between the geomagnetic vector obtained from the geomagnetic vector map query and the actual measurement value of the magnetometer. Based on the sensor reliability decision, the dominant navigation source is determined, enabling seamless switching and continuous navigation in GPS-denied environments.

[0013] Optionally, the system further includes a multi-source sensor array to acquire navigation data, wherein:

[0014] The multi-source sensor array specifically includes a multi-band, multi-system satellite receiver, a high-precision tactical-grade IMU, a solid-state lidar, a binocular vision module, and a three-axis high-precision magnetometer, which are used to collect raw satellite navigation observations, inertial measurement data, laser point clouds, visual image sequences, and geomagnetic vector data, respectively.

[0015] Optionally, the geomagnetic fingerprint anomaly detection in the deception detection and isolation module specifically includes:

[0016] The real-time magnetometer readings are compared with the geomagnetic reference values ​​obtained by querying the geomagnetic vector map based on the GPS location, and the real-time magnetometer readings are also compared with the geomagnetic reference values ​​obtained by querying the geomagnetic vector map based on the visual / SLAM estimated location.

[0017] If the geomagnetic reference value corresponding to the GPS location does not match well with the real-time magnetometer reading, but the geomagnetic reference value corresponding to the visual / SLAM estimated location matches well with the real-time magnetometer reading, then the GPS is determined to be spoofed.

[0018] Optionally, the raw observation geometric information in the tightly coupled multi-source fusion navigation module is directly fed into the filter, and then fused and updated by combining the difference between the geomagnetic vector obtained from the geomagnetic vector map query and the actual measurement value of the magnetometer. Specifically, this includes:

[0019] The pixel coordinates of feature points extracted by binocular vision are directly fed into the error state Kalman filter as the original observation geometric information, and geometrically fused with the feature positions predicted based on IMU.

[0020] The point-to-surface distance of the laser point cloud plane acquired by the solid-state lidar is directly fed into the error state Kalman filter as the original observation geometric information, and then geometrically fused with the point position predicted by the IMU.

[0021] Optionally, the geomagnetic tight coupling in the tightly coupled multi-source fusion navigation module is specifically used for:

[0022] The ground magnetic bias term is added to the state vector of the error state Kalman filter. The current pose is calculated by the IMU and the geomagnetic vector B_pred is predicted in the current pose by combining the geomagnetic vector map. The difference between the actual measured value B_meas of the magnetometer and the geomagnetic vector B_pred is used as the observation. The filter state, including position, attitude and geomagnetic bias, is updated to provide absolute position observation and heading angle observation to suppress IMU heading drift.

[0023] Optionally, the tightly coupled multi-source fusion navigation module further includes an adaptive fusion unit, which dynamically adjusts the noise covariance of the observation quantities corresponding to each original observation geometric information based on the visual feature tracking quality, laser point cloud density, and the degree of local disturbance of the geomagnetic field, thereby achieving adaptive weighted fusion.

[0024] Optionally, the system also includes a geomagnetic vector map management module, which is used to pre-store geomagnetic vector maps obtained through prior measurements before flight, or to construct geomagnetic vector maps online in real time using SLAM technology. The geomagnetic vector map includes the total magnetic field strength, components and their spatial gradient, and provides geomagnetic reference value query services for the deception detection and isolation module and the tightly coupled multi-source fusion navigation module.

[0025] Optionally, the decision-making process based on sensor confidence specifically includes: when the confidence level of spoofing detection exceeds a threshold, immediately reducing the GPS data weight to zero and seamlessly switching to a tightly coupled fusion navigation mode centered on visual / laser SLAM, geomagnetic vector matching, and high-precision IMU.

[0026] Optionally, after flying out of the deception risk zone, if the GPS signal is confirmed to be reliable through a consistency check, it can be gradually reintroduced into the fusion filter to restore full-state navigation.

[0027] Secondly, a method for autonomous navigation and deception defense in GPS-denied environments for unmanned aerial vehicles (UAVs) is provided, implemented in any of the systems described in the first aspect above. This method includes parallel execution of a deception detection and isolation thread and a tightly coupled multi-source fusion navigation thread, wherein:

[0028] In the deception detection and isolation thread, the navigation data is subjected to multi-dimensional consistency checks. These multi-dimensional consistency checks include receiver autonomous integrity monitoring, residual checks between IMU / visual predicted position and GPS position, and geomagnetic fingerprint anomaly detection. When the geomagnetic fingerprint anomaly detection determines that the GPS position-geomagnetic reading match is poor while the visual position-geomagnetic reading match is good, the GPS data is isolated from the fusion unit.

[0029] In the tightly coupled multi-source fusion navigation thread, an error state Kalman filter with IMU as the core is adopted. A geomagnetic deviation term is introduced into the state vector. The pixel coordinates of visual feature points and the point-to-surface distance of laser point cloud are directly fed into the filter as the original observation geometric information for tight coupling. At the same time, the difference between the actual measurement value of the magnetometer and the geomagnetic vector obtained by querying the geomagnetic vector map is used as the geomagnetic vector observation data for tight coupling. The navigation source is determined based on the reliability of the sensor, so as to achieve seamless switching and continuous navigation in GPS denied environment.

[0030] The beneficial effects of the technical solutions provided in this application include at least the following:

[0031] (1) Strong anti-spoofing capability: By introducing the geomagnetic vector as an independent reference source, an "arbitration" mechanism between GPS and visual / SLAM is created, which can efficiently and reliably detect and isolate satellite spoofing signals.

[0032] (2) High-precision continuous navigation: After GPS rejection, the IMU is provided with rich absolute and relative observations through the tight coupling and fusion of visual / lidar and geomagnetic vectors, which reduces the growth rate of navigation error by an order of magnitude and enables precise autonomous flight over distances of kilometers.

[0033] (3) Stronger environmental adaptability: Geomagnetic information remains effective in environments lacking visual / laser features (such as the sky and water surface); visual / laser can dominate positioning in areas with strong geomagnetic interference (such as steel structures). The two complement each other, greatly expanding the applicable scenarios.

[0034] (4) System self-consistency and safety: All sensors work tightly coupled under a unified mathematical model. If any sensor fails or is interfered with, the system can maintain basic navigation capabilities through the remaining sensors and provide a reliability assessment, thus ensuring high safety. Attached Figure Description

[0035] To more clearly illustrate the embodiments of this application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0036] Figure 1 A hardware composition and information flow diagram of an autonomous navigation and deception defense system for unmanned aerial vehicles (UAVs) in GPS denial environments is provided for embodiments of this application.

[0037] Figure 2 System software architecture and core algorithm thread block diagram provided for embodiments of this application;

[0038] Figure 3 This is a schematic diagram illustrating the principle of geomagnetic vector anomaly detection provided in an embodiment of this application;

[0039] Figure 4 The principle and data flow diagram of the tightly coupled multi-source fusion navigation filter provided in the embodiments of this application;

[0040] Figure 5 This is a schematic diagram illustrating the workflow of a typical GPS denial and spoofing scenario provided in the embodiments of this application. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0042] In the description of this application, the terms "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or apparatuses, or steps or units added based on further optimizations conceived in this application.

[0043] This application belongs to the field of autonomous navigation and flight safety technology for unmanned aerial vehicles (UAVs). Specifically, it relates to a system and method that enables UAVs to achieve continuous, reliable, and spoof-resistant autonomous navigation, positioning, and trajectory tracking without relying on external satellite signals in complex electromagnetic environments where satellite navigation signals fail (GPS denial) or are subjected to malicious spoofing interference. This is achieved by fusing visual, lidar, geomagnetic field, and inertial measurement data. This application aims to overcome the shortcomings of existing technologies and provide a UAV autonomous navigation and spoofing defense system and method based on tightly coupled fusion of multi-source information. When GPS is available, the system uses it for calibration; when GPS signal quality deteriorates or is spoofed, it can automatically trigger a spoofing detection and isolation mechanism and seamlessly switch to a tightly coupled fusion navigation mode centered on "visual / laser SLAM + geomagnetic vector field matching + high-precision IMU," achieving long-duration, highly reliable autonomous flight.

[0044] The core of this application lies in a navigation architecture of "adaptive sensor reliability assessment and tight coupling fusion". This architecture includes a centralized fusion filter (such as an error state Kalman filter) and innovatively introduces a "geomagnetic vector anomaly detection and auxiliary correction" module.

[0045] Please refer to Figure 1 The document illustrates a flowchart of an autonomous navigation and deception defense system for unmanned aerial vehicles (UAVs) in GPS denial environments, as provided in an embodiment of this application. The system includes a navigation sensor array comprising a multi-band, multi-system satellite receiver (outputting raw observations and signal quality), a high-precision tactical-grade IMU (outputting angular velocity and specific force), a solid-state lidar and binocular vision module (outputting feature points and point clouds), and a three-axis high-precision magnetometer (outputting geomagnetic vectors). Data from all sensors is input to an onboard high-performance computing unit. The computing unit operates a centralized, tightly coupled fusion filter and a security decision-making and switching module. The filter functions as a tightly coupled multi-source filter after GPS isolation. The core of fusion navigation is to tightly couple and fuse the pixel coordinates of visual feature points, the point-to-surface distance of laser point clouds, and geomagnetic vector observation data as the original observation geometric information. The safety decision and switching module selects the dominant navigation source based on the judgment results of deception detection and isolation—that is, to confirm whether GPS has been deceived through multi-dimensional consistency checks (including geomagnetic fingerprint anomaly detection). When GPS is trusted, it is the dominant source; when GPS is deceived, its weight is reduced to zero and the system switches to the tightly coupled fusion navigation mode. The system's bottom layer ultimately outputs reliable pose, velocity, and navigation integrity alarms, achieving seamless switching and continuous navigation in GPS denial environments.

[0046] Specifically, the system includes a deception detection and isolation module that operates in parallel and a tightly coupled multi-source fusion navigation module, wherein:

[0047] The deception detection and isolation module is used to perform multi-dimensional consistency checks on navigation data. When the geomagnetic fingerprint anomaly detection determines that the GPS position-geomagnetic reading match is poor while the visual position-geomagnetic reading match is good, the GPS data is isolated from the fusion unit. The multi-dimensional consistency checks include receiver autonomous integrity monitoring, residual checks between IMU / visual predicted position and GPS position, and geomagnetic fingerprint anomaly detection.

[0048] Specifically, the geomagnetic fingerprint anomaly detection in the deception detection and isolation module includes:

[0049] The real-time magnetometer readings are compared with the geomagnetic reference values ​​obtained from the geomagnetic vector map based on the GPS location. Simultaneously, the real-time magnetometer readings are compared with the geomagnetic reference values ​​obtained from the geomagnetic vector map based on the visual / SLAM estimated location. If the geomagnetic reference value corresponding to the GPS location does not match well with the real-time magnetometer reading, but the geomagnetic reference value corresponding to the visual / SLAM estimated location matches well with the real-time magnetometer reading, then the GPS is determined to be spoofed.

[0050] The tightly coupled multi-source fusion navigation module employs an error state Kalman filter centered on an IMU to introduce a geomagnetic bias term into the state vector. The pixel coordinates of visual feature points, the point-to-surface distance of the laser point cloud, and geomagnetic vector observation data are directly fed into the filter as raw observation geometric information. The filter is then fused and updated by combining the difference between the geomagnetic vector obtained from the geomagnetic vector map query and the actual measurement value of the magnetometer. Based on the sensor reliability decision, the dominant navigation source is determined, enabling seamless switching and continuous navigation in GPS-denied environments.

[0051] Specifically, the raw observation geometric information in the tightly coupled multi-source fusion navigation module is directly fed into the filter, and the difference between the geomagnetic vector obtained from the geomagnetic vector map query and the actual measurement value of the magnetometer is fused and updated. This includes: directly feeding the pixel coordinates of the feature points extracted by binocular vision into the error state Kalman filter as raw observation geometric information, and performing geometric constraint fusion with the feature positions predicted based on IMU; and directly feeding the point-to-surface distance of the laser point cloud plane obtained by solid-state lidar into the error state Kalman filter as raw observation geometric information, and performing geometric constraint fusion with the point positions predicted based on IMU.

[0052] The geomagnetic tight coupling in the tightly coupled multi-source fusion navigation module is specifically used to: expand the geomagnetic deviation term into the state vector of the error state Kalman filter; calculate the current pose through the IMU and predict the geomagnetic vector B_pred under the current pose by combining the geomagnetic vector map; use the difference between the actual measured value B_meas of the magnetometer and the geomagnetic vector B_pred as the observation; update the filter state, including position, attitude and geomagnetic deviation, to provide absolute position observation and heading angle observation to suppress IMU heading drift.

[0053] Specifically, the navigation source decision based on sensor reliability includes: when the spoofing detection confidence exceeds a threshold, immediately reducing the GPS data weight to zero and seamlessly switching to a tightly coupled fusion navigation mode centered on visual / laser SLAM, geomagnetic vector matching, and a high-precision IMU. After flying out of the spoofing risk zone, a consistency check confirms that the GPS signal has recovered its reliability, and it is gradually reintroduced into the fusion filter to restore full-state navigation.

[0054] In this embodiment of the application, the system also includes a multi-source sensor array to acquire navigation data. Specifically, the multi-source sensor array includes a multi-band multi-system satellite receiver, a high-precision tactical-grade IMU, a solid-state lidar, a binocular vision module, and a three-axis high-precision magnetometer, which are used to collect raw satellite navigation observations, inertial measurement data, laser point clouds, visual image sequences, and geomagnetic vector data, respectively.

[0055] The tightly coupled multi-source fusion navigation module also includes an adaptive fusion unit, which dynamically adjusts the noise covariance of the observation quantities corresponding to each original observation geometric information based on the tracking quality of visual features, the density of laser point clouds, and the degree of local disturbance of the geomagnetic field, so as to achieve adaptive weighted fusion.

[0056] The system also includes a geomagnetic vector map management module, which is used to pre-store geomagnetic vector maps obtained through prior measurements before flight, or to build geomagnetic vector maps online in real time using SLAM technology. The geomagnetic vector map contains the total magnetic field strength, components and their spatial gradient, and provides geomagnetic reference value query services for the deception detection and isolation module and the tightly coupled multi-source fusion navigation module.

[0057] like Figure 2 The system software architecture and core algorithm thread diagram are presented. Core thread one is the spoofing detection and isolation thread, which generates spoofing confidence through multi-dimensional consistency checks (including receiver autonomous integrity monitoring, IMU / visual prediction and GPS position comparison, and geomagnetic fingerprint anomaly detection dual verification). After sensor confidence assessment and decision, when the GPS confidence is determined to be low, the GPS data isolation mechanism is triggered. Core thread two is the tightly coupled multi-source fusion navigation thread, which adopts an error state Kalman filter with IMU as the core. The raw visual / laser observations (feature point pixel coordinates, point cloud plane distance) and geomagnetic vector observations (difference between magnetometer measured values ​​and geomagnetic map predicted values) are directly fed into the filter for tight coupling as raw observation geometric information. A geomagnetic deviation term is introduced into the state vector for online estimation. At the same time, adaptive noise adjustment is performed according to the noise weight of each sensor, and finally reliable navigation information is output to achieve continuous high-precision navigation in GPS denied environments.

[0058] In another optional embodiment of this application, the system includes:

[0059] a) Navigation sensor array:

[0060] Multi-band, multi-system satellite receiver: used to receive signals from GPS, BeiDou, etc., and output raw observations and signal quality indicators.

[0061] High-precision tactical-grade IMU: Provides angular velocity and specific force measurements.

[0062] Solid-state LiDAR and binocular vision module: used for environmental perception and feature extraction.

[0063] Triaxial high-precision magnetometer: measures the geomagnetic field vector in the carrier coordinate system.

[0064] b) Onboard high-performance computing unit: runs fusion algorithms, spoofing detection and SLAM calculations.

[0065] c) Safety decision and switching module: Based on sensor reliability, the decision-making guides the navigation source.

[0066] The software portion of this embodiment includes two parallel core threads: a deception detection and isolation thread and a tightly coupled multi-source fusion navigation thread.

[0067] S1: Deception Detection and Isolation Thread

[0068] Multidimensional consistency check:

[0069] Receiver autonomous integrity monitoring: checks satellite geometry and signal-to-noise ratio variations.

[0070] IMU / Visual Prediction vs. GPS Location Comparison: The location predicted by IMU and vision in the short term is compared with the location calculated by GPS to check the residuals. If the residuals continuously exceed the limits, it indicates that the GPS may be misled.

[0071] Geomagnetic fingerprint anomaly detection: The system pre-stores or generates online a geomagnetic vector map of the flight area (including the total magnetic field strength, components, and spatial gradient). Real-time magnetometer readings are compared with geomagnetic reference values ​​obtained from a table based on the current GPS location. Simultaneously, these readings are also compared with geomagnetic reference values ​​obtained from a table based on visual / SLAM estimated location.

[0072] If the "GPS location - geomagnetic reading" match well, but the "visual location - geomagnetic reading" match poorly, then visual drift may occur.

[0073] If the GPS location-geomagnetic readings match poorly, but the visual location-geomagnetic readings match well, then the GPS is very likely to be spoofed.

[0074] This double verification greatly improves the confidence level of spoofing detection.

[0075] Isolation Decision: When the confidence level of spoofing detection exceeds the threshold, the system immediately isolates the GPS data from the fusion unit (reducing the weight to zero) and issues an alarm.

[0076] S2: Tightly coupled multi-source fusion navigation thread (dominated after GPS is isolated)

[0077] A tightly coupled error state Kalman filter with an IMU as its core is used.

[0078] Tight coupling of vision / LiDAR: Unlike traditional methods that directly input pose, this approach directly feeds the raw observation geometry information (such as pixel coordinates and point-to-surface distance) of the extracted feature points (or LiDAR point cloud plane) into the filter as observations, fusing it with the feature locations predicted by the IMU. This is superior to loose coupling and can better handle dynamic scenes and short-term tracking loss.

[0079] Tight coupling of geomagnetic vectors:

[0080] The filter state vector is expanded with a local geomagnetic bias term.

[0081] Prediction steps: The IMU calculates the current pose and, combined with the pre-stored geomagnetic vector map, predicts the geomagnetic vector B_pred that the magnetometer "should" read under the current pose.

[0082] Update steps: Use the difference between the actual measured value B_meas and B_pred of the magnetometer as the observation to update the filter state (including position, attitude, and geomagnetic deviation).

[0083] Geomagnetic matching not only provides absolute position observations, but also crucial heading angle observations, effectively suppressing heading drift of the IMU.

[0084] Figure 3This is a schematic diagram illustrating the principle of geomagnetic vector anomaly detection. In scenario one (normal GPS), the predicted value B_pred_gps obtained by querying the pre-stored geomagnetic vector map based on the GPS estimated location matches well with the measured value B_meas from the magnetometer, indicating that the GPS is reliable. However, the predicted value B_pred_vis obtained by querying the location based on vision / SLAM has a moderate match with B_meas, suggesting possible visual drift. In scenario two (GPS spoofed), the B_pred_gps corresponding to the spoofed GPS estimated location has a poor match with B_meas (marked as a key anomaly), while the B_pred_vis corresponding to the location based on vision / SLAM has a good match with B_meas (marked as key evidence). Based on this, the system comprehensively determines that the GPS has been spoofed, thereby achieving high-confidence spoofing detection and isolation decisions through dual comparison of geomagnetic fingerprints.

[0085] Adaptive fusion: The filter dynamically adjusts the noise covariance of each observation based on the visual feature tracking quality, point cloud density, and the degree of local disturbance of the geomagnetic field, to achieve adaptive weighted fusion.

[0086] The following are specific embodiments of this application, including:

[0087] The working process of this invention is explained in detail by taking an example of a drone performing long-distance pipeline inspection flying over a substation area known to have GPS spoofing risks.

[0088] 1. Preliminary preparations:

[0089] Before the flight, a high-precision geomagnetic vector map of the mission area has been generated through prior measurements or offline calculations and loaded into the airborne system.

[0090] 2. Normal flight phase:

[0091] In open terrain, drones rely on GPS / IMU / visual fusion navigation. Geomagnetic observations are continuously used to calibrate magnetometer bias and assist heading.

[0092] 3. Entering the deception risk zone and detection:

[0093] As the substation approaches, ground-based deception devices emit false GPS signals.

[0094] The position calculated by the drone's GPS receiver began to slowly shift.

[0095] 4. Deception detection thread starts:

[0096] There are slight residuals between the IMU / visual short-term predicted location and the GPS location.

[0097] Key detection: The system queries the geomagnetic map and finds that the difference between the geomagnetic prediction value B_pred_gps corresponding to the current GPS location and the actual measured value B_meas by the magnetometer is getting larger and larger; while at the same time, the geomagnetic prediction value B_pred_vis corresponding to the visual / SLAM estimated location remains well matched with B_meas.

[0098] The system determined that the reliability of the GPS signal had dropped sharply, and the confidence level for deception had exceeded the threshold.

[0099] 5. Seamless switching and continuous navigation:

[0100] The security decision module immediately resets the GPS data rights to zero.

[0101] Tightly coupled fusion navigation threads automatically become dominant. Within the filter:

[0102] IMU performs state prediction.

[0103] The visual front end provides stable observation of pipeline feature points.

[0104] LiDAR provides planar constrained observation of the pipe surface and the ground.

[0105] The magnetometer provides B_meas, which is fused with B_pred_vis, obtained from a lookup table based on visual / laser-estimated pose, to continuously correct position, attitude (especially heading), and local magnetic deflection.

[0106] The drone relies entirely on this fusion system to fly stably along the pipeline trajectory, unaffected by any deceptive signals.

[0107] 6. Flying out of the risk zone and recovery:

[0108] After flying out of the substation's interference range, the spoofing signal disappeared. The system confirmed the reliable recovery of the GPS signal through a consistency check, and gradually reintroduced it into the fusion filter, restoring full-state navigation.

[0109] Figure 4 This paper presents the principle and data flow diagram of a tightly coupled multi-source fusion navigation filter. It demonstrates the prediction-update loop mechanism of an error state Kalman filter centered on an IMU: In the prediction step, the raw IMU data is driven by the system dynamics model to predict the state and covariance, outputting prior estimates including position, velocity, attitude, and geomagnetic deviation. In the update step, the pixel coordinates of visual feature points, the plane distance of the laser point cloud, and the geomagnetic vector measurement and geomagnetic map query values ​​(the difference between B_pred and B_meas) are used as the raw observation geometric information to construct residuals. After Kalman gain calculation, the state vector (including the geomagnetic deviation term) is updated, finally outputting the corrected high-precision navigation state, achieving tight coupling and fusion of visual / laser / geomagnetic and IMU.

[0110] Figure 5 This is a schematic diagram of the workflow in a typical GPS denial and spoofing scenario. High-rise buildings block and deny satellite signals, and ground-based GPS spoofing signal transmitters emit false GPS signals. The drone equipped with this system is in this environment. The left side shows the drone's working process: when the GPS signal weakens or is spoofed, spoofing detection is triggered, and the system switches to fusion navigation mode. It generates a safe navigation trajectory through a multi-source fusion method that combines binocular visual feature tracking, LiDAR scene scanning, and magnetometer matching with geomagnetic maps. Ultimately, the drone can autonomously cross the canyon along the planned path, achieving continuous and reliable navigation in a GPS denial environment.

[0111] Other possible implementations of this application are given below, including:

[0112] 1. Modified map construction method: Geomagnetic vector maps can be constructed online in real time and simultaneously with navigation, using SLAM technology to simultaneously construct environmental geometric maps and geomagnetic maps.

[0113] 2. Sensor variants: The vision sensor can be replaced with an event camera to provide more robust feature tracking under high-speed motion or drastic changes in lighting; the lidar can be a low-cost solid-state model.

[0114] 3. Fusion framework variant: Tightly coupled filters can use factor graph optimization instead of Kalman filtering, which is more suitable for long-term, large-scale global optimization.

[0115] 4. Cooperative navigation variant: In multi-aircraft formations, the leader aircraft can use this invention for high-precision navigation and provide reliable relative or absolute position references to the slave aircraft through inter-aircraft communication, thereby improving the overall anti-spoofing capability of the formation.

[0116] 5. Civil Defense Variant: The system can be simplified into a low-cost GPS spoofing detection and alarm module, containing only a multi-frequency receiver, IMU, and magnetometer, for consumer drones, which will force hovering or landing and alert the user when spoofing is detected.

[0117] The experimental data for this application are given below:

[0118] 1. Deception detection performance: In simulated deception tests, the system can detect and alarm for slow dragging deception within 3 meters of position deviation; for jumping deception, it can detect within 1 second.

[0119] 2. Navigation accuracy: In a 1-kilometer-long urban canyon section where GPS is denied, the final error of pure visual inertial odometry was 15 meters, while after adding geomagnetic tight coupling fusion, the final error was reduced to less than 3 meters.

[0120] 3. Heading maintenance: When flying for 5 minutes without GPS, the heading drift of a pure IMU exceeds 10 degrees, while the heading drift of this system after incorporating geomagnetism is less than 2 degrees.

[0121] 4. System robustness: In street scenarios with 30% dynamic vehicle interference, the navigation success rate is increased from 70% for traditional visual SLAM to 95% for this system.

[0122] This application also provides a method for autonomous navigation and deception defense against GPS denial environments for unmanned aerial vehicles (UAVs), including:

[0123] In the deception detection and isolation thread, the navigation data is subjected to multi-dimensional consistency checks. These multi-dimensional consistency checks include receiver autonomous integrity monitoring, residual checks between IMU / visual predicted position and GPS position, and geomagnetic fingerprint anomaly detection. When the geomagnetic fingerprint anomaly detection determines that the GPS position-geomagnetic reading match is poor while the visual position-geomagnetic reading match is good, the GPS data is isolated from the fusion unit.

[0124] In the tightly coupled multi-source fusion navigation thread, an error state Kalman filter with IMU as the core is adopted. A geomagnetic deviation term is introduced into the state vector. The pixel coordinates of visual feature points and the point-to-surface distance of laser point cloud are directly fed into the filter as the original observation geometric information for tight coupling. At the same time, the difference between the actual measurement value of the magnetometer and the geomagnetic vector obtained by querying the geomagnetic vector map is used as the geomagnetic vector observation data for tight coupling. The navigation source is determined based on the reliability of the sensor, so as to achieve seamless switching and continuous navigation in GPS denied environment.

[0125] For specific limitations on autonomous navigation and deception defense methods for UAVs in GPS-denied environments, please refer to the limitations on autonomous navigation and deception defense systems for UAVs in GPS-denied environments mentioned above, which will not be repeated here.

[0126] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0127] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A system for autonomous navigation and deception defense against GPS denial in unmanned aerial vehicles (UAVs), characterized in that, The system includes a deception detection and isolation module and a tightly coupled multi-source fusion navigation module that operate in parallel, wherein: The deception detection and isolation module is used to perform multi-dimensional consistency checks on navigation data. When the geomagnetic fingerprint anomaly detection determines that the GPS position-geomagnetic reading match is poor while the visual position-geomagnetic reading match is good, the GPS data is isolated from the fusion unit. The multi-dimensional consistency checks include receiver autonomous integrity monitoring, residual checks between IMU / visual predicted position and GPS position, and geomagnetic fingerprint anomaly detection. The tightly coupled multi-source fusion navigation module employs an error state Kalman filter centered on an IMU to introduce a geomagnetic bias term into the state vector. The pixel coordinates of visual feature points, the point-to-surface distance of the laser point cloud, and geomagnetic vector observation data are directly fed into the filter as raw observation geometric information. The filter is then fused and updated by combining the difference between the geomagnetic vector obtained from the geomagnetic vector map query and the actual measurement value of the magnetometer. Based on the sensor reliability decision, the dominant navigation source is determined, enabling seamless switching and continuous navigation in GPS-denied environments.

2. The system according to claim 1, characterized in that, The system also includes a multi-source sensor array, through which navigation data is acquired, wherein: The multi-source sensor array specifically includes a multi-band, multi-system satellite receiver, a high-precision tactical-grade IMU, a solid-state lidar, a binocular vision module, and a three-axis high-precision magnetometer, which are used to collect raw satellite navigation observations, inertial measurement data, laser point clouds, visual image sequences, and geomagnetic vector data, respectively.

3. The system according to claim 1, characterized in that, The geomagnetic fingerprint anomaly detection in the deception detection and isolation module specifically includes: The real-time magnetometer readings are compared with the geomagnetic reference values ​​obtained by querying the geomagnetic vector map based on the GPS location, and the real-time magnetometer readings are also compared with the geomagnetic reference values ​​obtained by querying the geomagnetic vector map based on the visual / SLAM estimated location. If the geomagnetic reference value corresponding to the GPS location does not match well with the real-time magnetometer reading, but the geomagnetic reference value corresponding to the visual / SLAM estimated location matches well with the real-time magnetometer reading, then the GPS is determined to be spoofed.

4. The system according to claim 1, characterized in that, The raw observation geometric information in the tightly coupled multi-source fusion navigation module is directly fed into the filter, and then fused and updated by combining the difference between the geomagnetic vector obtained from the geomagnetic vector map query and the actual measurement value of the magnetometer. Specifically, this includes: The pixel coordinates of feature points extracted by binocular vision are directly fed into the error state Kalman filter as the original observation geometric information, and geometrically fused with the feature positions predicted based on IMU. The point-to-surface distance of the laser point cloud plane acquired by the solid-state lidar is directly fed into the error state Kalman filter as the original observation geometric information, and then geometrically fused with the point position predicted by the IMU.

5. The system according to claim 1, characterized in that, The geomagnetic tight coupling in the tightly coupled multi-source fusion navigation module is specifically used for: The ground magnetic bias term is added to the state vector of the error state Kalman filter. The current pose is calculated by the IMU and the geomagnetic vector B_pred is predicted in the current pose by combining the geomagnetic vector map. The difference between the actual measured value B_meas of the magnetometer and the geomagnetic vector B_pred is used as the observation. The filter state, including position, attitude and geomagnetic bias, is updated to provide absolute position observation and heading angle observation to suppress IMU heading drift.

6. The system according to claim 1, characterized in that, The tightly coupled multi-source fusion navigation module also includes an adaptive fusion unit, which dynamically adjusts the noise covariance of the observation quantities corresponding to each original observation geometric information based on the visual feature tracking quality, laser point cloud density, and the degree of local disturbance of the geomagnetic field, thereby achieving adaptive weighted fusion.

7. The system according to claim 1, characterized in that, The system also includes a geomagnetic vector map management module, which is used to pre-store geomagnetic vector maps obtained through prior measurements before flight, or to construct geomagnetic vector maps online in real time using SLAM technology. The geomagnetic vector map includes the total magnetic field strength, components and their spatial gradient, and provides geomagnetic reference value query services for the deception detection and isolation module and the tightly coupled multi-source fusion navigation module.

8. The system according to claim 1, characterized in that, The decision-making process based on sensor confidence specifically includes: when the confidence level of spoofing detection exceeds a threshold, immediately reducing the GPS data weight to zero and seamlessly switching to a tightly coupled fusion navigation mode centered on visual / laser SLAM, geomagnetic vector matching, and high-precision IMU.

9. The system according to claim 8, characterized in that, Once the GPS signal is confirmed to be reliable after flying out of the deception risk zone, it is gradually reintroduced into the fusion filter to restore full-state navigation.

10. A method for autonomous navigation and deception defense against GPS denial environments for unmanned aerial vehicles (UAVs), implemented in the system described in any one of claims 1-9, characterized in that, The method includes parallel execution of a deception detection and isolation thread and a tightly coupled multi-source fusion navigation thread, wherein: In the deception detection and isolation thread, the navigation data is subjected to multi-dimensional consistency checks. These multi-dimensional consistency checks include receiver autonomous integrity monitoring, residual checks between IMU / visual predicted position and GPS position, and geomagnetic fingerprint anomaly detection. When the geomagnetic fingerprint anomaly detection determines that the GPS position-geomagnetic reading match is poor while the visual position-geomagnetic reading match is good, the GPS data is isolated from the fusion unit. In the tightly coupled multi-source fusion navigation thread, an error state Kalman filter with IMU as the core is adopted. A geomagnetic deviation term is introduced into the state vector. The pixel coordinates of visual feature points and the point-to-surface distance of laser point cloud are directly fed into the filter as the original observation geometric information for tight coupling. At the same time, the difference between the actual measurement value of the magnetometer and the geomagnetic vector obtained by querying the geomagnetic vector map is used as the geomagnetic vector observation data for tight coupling. The navigation source is determined based on the reliability of the sensor, so as to achieve seamless switching and continuous navigation in GPS denied environment.