Method and system for large scale low altitude vehicle deviation warning

Abstract

The application discloses a large-scale low-altitude aircraft deviation early warning method and system, which firstly integrates aircraft onboard sensors and external environment data through multi-source heterogeneous data fusion technology, realizes high-precision estimation of flight state, then predicts the future three-dimensional trajectory of the aircraft based on deep learning, then quantitatively evaluates and classifies multiple risk factors such as deviation distance, airspace density, weather conditions and airspace properties, and finally issues corresponding warning or control instructions to the aircraft, ground control party or supervision department according to the risk level through the air-ground collaborative network. The application solves the dynamic deviation monitoring and early warning problem of large-scale and high-density aircraft groups in low-altitude traffic, realizes the transformation from passive lag response to active prediction and early warning, and improves the safety, control accuracy and airspace operation efficiency of low-altitude flight.

Description

Technical Field

[0001] This invention relates to unmanned aerial vehicle (UAV) technology, low-altitude airspace management, artificial intelligence, and flight safety monitoring technology, and in particular to a method and system for early warning of large-scale low-altitude aircraft deviations. Background Technology

[0002] In low-altitude airspace, particularly for general aviation and the emerging low-altitude economy, traditional air traffic control primarily employs procedural control and area control models. Procedural control relies on air-to-ground communication between controllers and pilots, managing traffic by analyzing pilot-reported positions to maintain safe separations between aircraft. This model is still suitable for low-density, small-scale general aviation activities.

[0003] With the rapid development of drone technology and the increasing diversification of its applications, such as logistics delivery, urban inspection, emergency rescue, and geographic mapping, low-altitude airspace is facing the challenge of large-scale, high-density, and heterogeneous flight traffic. To address this, several emerging technological solutions have been proposed to resolve the issue of drone trajectory deviation during flight. For example, existing technologies include cloud-based drone monitoring systems that record the drone's heading trajectory and use methods such as straight-line and circular interpolation to control and correct its specific trajectory. Furthermore, some research is exploring the use of Long Short-Term Memory (LSTM) neural networks to predict drone flight trajectories, pre-calculating deviations and reducing power consumption and accidents. These technologies offer preliminary automated solutions for drone monitoring.

[0004] Traditional low-altitude air traffic control relies on manual intervention. The basic process involves the captain submitting a flight plan before flight, including the route and estimated time. During flight, controllers analyze the distances between aircraft based on periodically reported position information from pilots and use air-to-ground communication for command and coordination to avoid collisions and ensure safety. This model is characterized by a high dependence on human intervention and is a non-real-time, passive management approach. Its core is a "human-in-the-loop, ground-based, distributed node-centric" control model.

[0005] The workflow of cloud-based trajectory correction typically follows these steps:

[0006] Data Acquisition: The cloud-based data acquisition module records the drone's heading and trajectory data in real time during flight.

[0007] Deviation calculation: The cloud server compares the drone's actual flight trajectory with the preset planned trajectory and calculates the deviation between the two using algorithms such as linear and circular interpolation. This method is easy to implement, but its algorithm is simple and cannot capture complex nonlinear motion patterns.

[0008] Command issuance: Based on the calculated deviation, the early warning module issues timely correction commands to the drone so that the drone flies along the predetermined trajectory.

[0009] Another working principle focuses on autonomous prediction on the drone's onboard terminal:

[0010] Data acquisition: The drone acquires and stores its own sensor data in real time, such as basic flight data like position, speed, and attitude.

[0011] Trajectory prediction: A pre-trained Long Short-Term Memory (LSTM) neural network model is used to learn features from historical flight data to infer the drone's trajectory over a future period. This method is the first to achieve prediction based solely on existing sensors, reducing the cost of adding additional sensors.

[0012] Route fine-tuning: Based on the predicted future trajectory, the drone can calculate potential deviations in advance and fine-tune its flight path on the airborne end to reduce power consumption and accident risks. This method does not require establishing a rigorous drone dynamics physical model.

[0013] The aforementioned existing technical solutions have significant limitations in addressing the regulation of large-scale low-altitude aircraft:

[0014] Passive and Delayed: Traditional procedural control models are entirely passive, providing only post-event or immediate corrections, unable to anticipate potential deviations or conflicts. Similarly, correction methods based on geometric interpolation are also post-event response mechanisms, unable to provide early warnings before deviations occur. This passive response model is ill-suited to the conflict-free traffic trajectory operation requirements of environments with multiple strongly coupled constraints.

[0015] Scalability bottleneck: With the opening of low-altitude airspace and the surge in the number of drones, the traditional "human-in-the-loop, ground-based distributed node-centric control model" will struggle to cope with the global optimization and control of future high-volume traffic flows, potentially falling into a dilemma of "strict control leading to stagnation, and lax control leading to congestion." This model lacks the ability to finely control high-density, heterogeneous flight traffic.

[0016] Insufficient Data Robustness and Accuracy: Existing drone monitoring solutions heavily rely on data from single airborne sensors, particularly GPS and inertial navigation systems (INS). However, INS can introduce errors of up to 1 meter within 10 seconds, and GPS is susceptible to interference, spoofing, or signal obstruction. This results in insufficient accuracy and robustness of positioning and prediction results from a single data source.

[0017] Limitations of simple geometric models: While linear and circular interpolation algorithms are easy to implement, their simplicity makes them unable to capture complex nonlinear motion patterns or adapt to complex environmental factors such as wind, airflow, and dynamic obstacles. This results in insufficient accuracy and adaptability in dynamic and complex low-altitude environments.

[0018] It should be noted that the information disclosed in the background section above is only for understanding the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0019] The main objective of this invention is to overcome the deficiencies in the aforementioned background technology and provide a method and system for early warning of large-scale low-altitude aircraft deviation.

[0020] To achieve the above objectives, the present invention adopts the following technical solution:

[0021] In a first aspect of the present invention, a method for early warning of large-scale low-altitude vehicle deviation includes the following steps:

[0022] S1. Multi-source heterogeneous data fusion and flight status estimation: Real-time acquisition of multi-source heterogeneous data from aircraft onboard sensors and external environmental data sources, and the data fusion algorithm is used to fuse the data and estimate the flight status to generate high-precision real-time aircraft status information.

[0023] S2. Flight trajectory prediction based on deep learning: Based on the real-time status information of the aircraft, the future three-dimensional trajectory of the aircraft is predicted using a prediction method based on deep learning.

[0024] S3. Multi-factor quantitative risk assessment and classification: The predicted future three-dimensional trajectory is compared with the preset planned route, and multiple risk factors are quantitatively calculated to obtain a comprehensive risk score. Different levels of early warning are triggered based on the comprehensive risk score.

[0025] S4. Tiered Early Warning and Command Issuance: Based on the aforementioned early warning level, issue corresponding level early warning information or control commands to the target aircraft, ground controllers, or regulatory authorities through the air-ground collaborative communication network.

[0026] In a second aspect of the invention, a large-scale low-altitude vehicle deviation early warning system includes:

[0027] The cloud-based early warning server is used to predict the trajectory and quantify the risk assessment of aircraft based on multi-source fusion data, and to generate early warning or control commands.

[0028] The ground communication base station and control center are connected to the cloud-based early warning server for data and command transmission and interaction with low-altitude aircraft, and also have ground control functions.

[0029] The low-altitude aircraft is communication-coupled to the ground communication base station and control center, and is equipped with an onboard unit for receiving and executing instructions from the cloud-based early warning server.

[0030] The cloud-based early warning server, the ground communication base station and control center, and the low-altitude aircraft constitute an air-ground collaborative early warning architecture to achieve closed-loop processing from data collection, fusion prediction, risk assessment to the issuance of graded instructions.

[0031] The present invention has the following beneficial effects:

[0032] This invention provides a method and system for large-scale low-altitude aircraft deviation early warning, applicable to high-density, large-scale low-altitude aircraft swarms, and based on air-ground collaboration and multi-source data fusion for dynamic deviation early warning. It aims to address the technical bottlenecks and safety challenges of existing low-altitude airspace management models in dealing with the large-scale, heterogeneous flight traffic brought about by future low-altitude economic development, providing key technical support for the transformation of low-altitude traffic control models, and enabling traffic management to evolve from traditional manual, passive intervention to a data-link-based air-ground autonomous intent negotiation and intelligent dispatch model.

[0033] This invention fundamentally solves the problems of passivity and lag, insufficient data robustness, and low regulatory credibility in existing technologies when dealing with large-scale, high-density low-altitude traffic flows. By deeply integrating cutting-edge technologies such as air-ground collaborative architecture, multi-source data fusion, deep learning prediction, and blockchain notarization, this solution constructs a proactive, global, and reliable intelligent early warning system.

[0034] Specifically, this invention primarily addresses the scalability and efficiency issues under large-scale traffic conditions. Recognizing that existing control models are inefficient, lack global optimization and control, and are prone to the dilemma of "strict control leading to stagnation, and lax control leading to congestion," this invention constructs an "air-ground collaborative" system based on cloud computing and a distributed architecture. This system can simultaneously process and analyze massive amounts of aircraft data, achieving a shift from passive response to proactive prediction, thereby effectively optimizing the utilization efficiency of airspace resources.

[0035] Secondly, this invention provides high-precision and robust dynamic prediction and early warning capabilities. By innovatively employing a multi-source data fusion algorithm, integrating data from the aircraft's own sensors (such as IMU, vision, and barometers) and external environmental data (such as ground base stations, meteorological data, and 5G-A networks), and utilizing deep learning-based trajectory prediction, this method achieves accurate and dynamic prediction of the aircraft's future trajectory. This fundamental shift from relying on simple geometric models for correction to data-driven prediction significantly improves the accuracy and environmental adaptability of early warnings.

[0036] Furthermore, this invention also addresses the issues of flight data integrity and reliability. To tackle the regulatory challenges posed by illegal activities such as unauthorized flights, blockchain technology is introduced to encrypt and store key flight data, early warning logs, and corrective instructions. This ensures data integrity, immutability, and traceability, providing a reliable basis for subsequent liability determination and establishing the data trust foundation necessary for efficient regulation.

[0037] In summary, this invention upgrades low-altitude aircraft management from a traditional, passive, decentralized, and unreliable model to a proactive, global, and trustworthy intelligent early warning system by integrating multi-source heterogeneous data, employing deep learning-based prediction algorithms, and constructing an air-ground collaborative distributed architecture. The key to this system lies in upgrading traditional, singular, and passive deviation detection into a multi-source, collaborative, and proactive prediction and early warning system, thereby fundamentally improving the safety assurance capabilities and operational efficiency of low-altitude airspace.

[0038] Other beneficial effects of the embodiments of the present invention will be further described below. Attached Figure Description

[0039] Figure 1 This is a technical roadmap for a large-scale low-altitude aircraft deviation early warning method according to an embodiment of the present invention.

[0040] Figure 2 This is a flowchart illustrating the overall process of the large-scale low-altitude vehicle deviation warning method according to an embodiment of the present invention.

[0041] Figure 3 This is a schematic diagram of a large-scale low-altitude aircraft deviation warning device according to an embodiment of the present invention. Detailed Implementation

[0042] The embodiments of the present invention will be described in detail below. It should be emphasized that the following description is merely exemplary and not intended to limit the scope and application of the present invention.

[0043] This invention aims to solve the problem of deviation warning for large-scale low-altitude aircraft swarms in complex environments. It proposes an active intelligent warning method and system based on air-ground collaboration, fusion of multi-source data and deep learning prediction. This transforms the traditional, single, and passive deviation detection into a multi-source, collaborative, and proactive prediction and warning system, significantly improving warning accuracy, system reliability, and overall operational efficiency of low-altitude airspace.

[0044] See Figure 1 and Figure 2 This invention provides a method for early warning of large-scale low-altitude vehicle deviation, comprising the following steps:

[0045] Step S1, Multi-source heterogeneous data fusion and flight status estimation: Multi-source heterogeneous data from the aircraft's onboard sensor A and external environmental data sources are collected in real time, and the data is fused and processed using a data fusion algorithm to estimate the flight status, generating high-precision real-time aircraft status information;

[0046] In some embodiments, the external environment data source includes a ground communication base station and a 5G-A remote sensing integrated base station that provide supplementary location information, a real-time meteorological service that provides wind speed and wind direction data, and an air traffic management platform that provides dynamic airspace status information.

[0047] In some embodiments, step S1, which involves using a data fusion algorithm to fuse the data and estimate the flight state, specifically involves: employing an extended Kalman filter algorithm to take a multidimensional state vector containing the quaternions of the aircraft's position, velocity, and attitude, as well as the sensor's zero-bias estimate, as the system state; predicting the state based on a nonlinear state transition function, wherein the state transition function updates the velocity, position, and attitude by integrating the calibrated inertial measurement unit readings; and then using measurements from multiple sensors and data sources, iteratively updating and correcting the predicted state and error covariance by calculating the measurement residual, innovation covariance, and Kalman gain, thereby achieving a high-precision estimation of the aircraft's state and estimating and compensating for the sensor's zero bias online to suppress error accumulation.

[0048] Step S2: Flight trajectory prediction based on deep learning: Based on the real-time status information of the aircraft, the future three-dimensional trajectory of the aircraft is predicted using a prediction method based on deep learning.

[0049] In some embodiments, in step S2, the deep learning-based prediction method employs a long short-term memory neural network or a convolutional long short-term memory network; the input to the network is a sequence of historical flight state data after data fusion and preprocessing, wherein the preprocessing includes differentiating and segmenting the original data to normalize the data; the output of the network is a sequence of three-dimensional trajectories of the aircraft in the future prediction time domain.

[0050] Step S3, Multi-factor quantitative risk assessment and classification: The predicted future three-dimensional trajectory is compared with the preset planned route, and multiple risk factors are quantitatively calculated to obtain a comprehensive risk score. Different levels of early warning are triggered based on the comprehensive risk score.

[0051] In some embodiments, step S3 specifically includes:

[0052] S31. Deviation risk factor calculation: Calculate the three-dimensional Euclidean distance between the predicted trajectory and the planned route at each predicted time step in the future, and quantify the maximum distance value in all time steps as the deviation risk factor.

[0053] S32. Calculation of airspace traffic density risk factor: Based on the global airspace information provided by the cloud air traffic control platform, with the predicted trajectory as the center, calculate the number density of other aircraft in the preset space volume around the airspace it passes through in the future, and quantify the density into an airspace traffic density risk factor through a preset mapping function.

[0054] S33. Calculation of the risk factor of adjacent restricted airspace: Calculate the minimum distance from the predicted trajectory to all preset restricted airspace boundaries, and quantify the risk factor of adjacent restricted airspace based on the minimum distance through an exponential decay function;

[0055] S34. Calculation of adverse meteorological risk factors: Based on real-time acquired meteorological data, one or more meteorological parameters among wind speed, gust intensity and precipitation rate are quantified into adverse meteorological risk factors through a weighted linear combination.

[0056] S35. Comprehensive risk score calculation: The deviation range risk factor, airspace traffic density risk factor, adjacent restricted airspace risk factor, and adverse weather risk factor are weighted and aggregated according to preset weights to calculate the comprehensive risk score.

[0057] In some embodiments, step S3 further includes setting tiered warning thresholds: a first threshold, a second threshold, and a third threshold are preset, the thresholds being set based on security case analysis and regulatory standards; when the comprehensive risk score is greater than or equal to the first threshold and less than the second threshold, a level one warning is triggered; when the comprehensive risk score is greater than or equal to the second threshold and less than the third threshold, a level two warning is triggered; when the comprehensive risk score is greater than or equal to the third threshold, a level three warning is triggered.

[0058] Step S4, Tiered Early Warning and Command Issuance: Based on the warning level, issue corresponding level warning information or control commands to the target aircraft, ground controllers or regulatory authorities through the air-ground cooperative communication network;

[0059] In some embodiments, step S4, specifically, involves:

[0060] When a Level 1 warning is triggered, an automatic trajectory fine-tuning command is issued to the target aircraft;

[0061] When a Level 2 warning is triggered, an audible and visual alarm is issued to ground controllers or operators, and the optimal correction scheme generated based on forecast data and airspace status is provided at the same time.

[0062] When a Level 3 warning is triggered, the highest level alarm will be issued to the relevant regulatory authorities, and a forced return or forced landing order may be issued.

[0063] In some embodiments, when the automatic trajectory fine-tuning command is issued to the target aircraft, the command is received by the aircraft's onboard unit and executed by the onboard unit through a super-control coupler integrated with the flight control system, thereby achieving automatic adjustment of the flight attitude and trajectory without human intervention.

[0064] Step S5, Key Data Storage and Traceability: Use blockchain technology to encrypt and store key data, early warning logs, and issued instructions during the flight process to ensure data integrity and immutability.

[0065] In some embodiments, the method is performed by an air-ground coordinated early warning device, the device comprising: a cloud-based early warning server for performing data fusion, trajectory prediction, risk assessment, and command generation; a ground communication base station and control center communicatively connected to the cloud-based early warning server for interacting with the aircraft in terms of data and commands; and an airborne unit deployed on the aircraft for receiving commands and coupling with the flight control system to perform adjustments.

[0066] See Figure 3 The present invention also provides a large-scale low-altitude aircraft deviation early warning system, including a cloud-based early warning server, a ground communication base station and control center, and a low-altitude aircraft;

[0067] The cloud-based early warning server is used to predict the trajectory and quantify the risk assessment of aircraft based on multi-source fusion data, and to generate early warning or control commands.

[0068] The ground communication base station and control center are connected to the cloud-based early warning server for data and command transmission and interaction with low-altitude aircraft, and also have ground control functions.

[0069] The low-altitude aircraft is communication-coupled to the ground communication base station and control center, and is equipped with an onboard unit for receiving and executing instructions from the cloud-based early warning server.

[0070] The cloud-based early warning server, the ground communication base station and control center, and the low-altitude aircraft constitute an air-ground collaborative early warning architecture to achieve closed-loop processing from data collection, fusion prediction, risk assessment to the issuance of graded instructions.

[0071] The large-scale low-altitude aircraft deviation early warning method and system proposed in this invention have the following main technical advantages: They achieve a fundamental shift from passive, delayed response to proactive prediction and early warning. Through deep learning-based trajectory prediction, the system can predict aircraft trajectories in advance and proactively intervene, effectively overcoming the control bottleneck of "controlling to death and releasing to cause congestion" under large-scale traffic. At the data level, the system integrates multi-source data from airborne sensors, ground base stations, and meteorological data, significantly improving the positioning accuracy and robustness in complex environments and overcoming the limitations of traditional solutions that rely on a single data source. At the architecture level, the system adopts a cloud-machine... The distributed air-ground collaborative design of the carrier system possesses strong scalability, supporting global optimization and efficient control of massive amounts of aircraft data. Furthermore, the introduction of blockchain technology establishes a reliable data storage mechanism, ensuring the integrity and immutability of key flight data and commands, enhancing regulatory credibility and post-event traceability capabilities. In addition, multi-dimensional quantitative risk assessment comprehensively considers various variables such as deviation distance, airspace density, environmental factors, and airspace characteristics, achieving a refined balance between safety management and airspace operational efficiency, thereby comprehensively improving the safety assurance capabilities and resource utilization efficiency of low-altitude airspace.

[0072] The following further describes specific embodiments of the present invention and examples of its algorithm implementation.

[0073] A large-scale low-altitude aircraft deviation early warning device and method is proposed, which constructs an active, global, and reliable intelligent early warning system. It fundamentally solves the problems of passivity and lag, insufficient data robustness, and low regulatory credibility of existing technologies when dealing with large-scale, high-density low-altitude traffic flows.

[0074] 1. Multi-source heterogeneous data acquisition and fusion

[0075] This invention first addresses the inherent limitations of a single data source. Existing drone monitoring systems heavily rely on GPS and inertial navigation systems (INS), but INS can introduce errors of up to 1 meter within a mere 10 seconds, and GPS is susceptible to signal interference, spoofing, or obstruction, resulting in insufficient accuracy and robustness in positioning and prediction. To overcome this bottleneck, this solution establishes a multi-source data fusion engine, whose data input includes:

[0076] Airborne sensor data: Real-time data acquired from the aircraft's flight control system (FCS), including attitude, velocity, angular velocity, acceleration, magnetometer readings, barometer readings, etc. This data provides the foundation for real-time perception and precise control of flight status.

[0077] External environmental data: Real-time data acquired through cloud platforms and ground communication base stations, including airspace traffic density, meteorological data (wind speed, wind direction), dynamic no-fly / restricted-altitude zone information, and supplementary location information provided by 5G-A integrated sensing base stations. 5G-A integrated sensing technology enables precise monitoring of low-altitude airspace, providing a powerful technical guarantee for efficient management. Furthermore, this solution can utilize photoelectric / infrared imaging, acoustic signals, terrain-related and scene-related technologies to enhance GPS positioning capabilities in complex environments, and improve dynamic navigation accuracy through navigation error detection and correction.

[0078] This solution employs advanced multi-sensor data fusion algorithms, such as the Extended Kalman Filter (EDF), to fuse information from different sensors and data sources in real time. This algorithm can dynamically adjust the weights of different data sources, ensuring that even when a signal from one sensor (such as GPS) is lost or interfered with, high-precision positioning and attitude information can still be continuously provided from other data sources (such as INS, ground base stations, and barometers), thus greatly improving the system's robustness.

[0079] (1) State vector

[0080] To achieve high-precision navigation, the system state is described by a state vector containing 15 elements, which is a common choice in the field of integrated navigation. in: It is the three-dimensional position of the aircraft in a navigation coordinate system (such as NED, North-East-Ground). It refers to its three-dimensional velocity. It is a unit quaternion representing the body's attitude, avoiding the gimbal lock problem of Euler angles. These are the zero-bias estimates of the three-axis accelerometer and the two-axis (or three-axis) gyroscope, respectively.

[0081] Incorporating sensor zero bias into the state vector is a crucial design element. It reflects a deep understanding of the main failure modes (i.e., drift) of low-cost IMUs. By estimating and compensating for these zero biases online, EKF can suppress the unchecked accumulation of errors, thus elevating a simple state tracker into a self-calibrating system. This self-calibration capability is essential for maintaining the long-term integrity of time-series data input to subsequent deep learning models and is the technological cornerstone for achieving the goal of "proactive prediction."

[0082] (2) Nonlinear state transition function

[0083] The state transition function, based on the principles of rigid body kinematics, describes the evolution of the system's state over time. It is determined by the state at the previous moment. and the current IMU measurement (as control input) To predict the state at the current moment. Specifically, the function updates velocity and position by integrating accelerometer readings (after attitude transformation, gravity compensation, and zero bias correction), and updates the attitude quaternion 24 by integrating gyroscope readings (after zero bias correction).

[0084] (3) Implementation of Extended Kalman Filter (EKF)

[0085] EKF applies the standard Kalman filter framework by performing a first-order Taylor expansion (linearization) on the nonlinear system at the current state estimation point.19 The entire process follows a recursive "prediction-update" loop.

[0086] This step utilizes a system dynamics model, from... State estimation at time 1 Predict the state at time k.

[0087] State prediction: .

[0088] Error covariance prediction:

[0089] Measurement model A state vector was established. Compared with sensor measurement value The relationship between them. For example, GPS provides location and speed measurements, while a barometer provides altitude measurements. It is the measurement noise, assumed to be Gaussian white noise.

[0090] (4) Update step

[0091] This step utilizes the actual sensor measurements at time k-1. To correct the predicted state .

[0092] Calculate the new information (measure the residual): Calculate the new information covariance: Calculate the Kalman gain: Kalman gain is essentially an optimal weighting between model predictions and sensor measurements to minimize posterior estimation errors.

[0093] EKF is not merely a "filter," but also an engine for generating high-quality time-series data for subsequent deep learning models. Deep learning trajectory prediction models typically assume clean input data, and EKF, through explicit modeling of noise and uncertainty, precisely satisfies this premise, playing a crucial role in preprocessing and feature enhancement. The synergy between model-driven EKF and data-driven LSTM is one of the core elements of this invention's technical solution; the output quality of EKF directly determines the upper limit of LSTM prediction performance.

[0094] (5) Integrating the covariance matrix (Q and R)

[0095] The performance of EKF is highly dependent on the accurate setting of the process noise covariance matrix Q and the measurement noise covariance matrix R.

[0096] The R-matrix is ​​typically determined based on the sensor's datasheet and offline calibration experiments. Its diagonal elements represent the variance of each measurement. The Q-matrix reflects the level of confidence in the dynamic model and requires fine-tuning based on the actual flight environment and aircraft characteristics to achieve a balance between system response speed and noise suppression capability.

[0097] 2. Flight trajectory prediction based on deep learning

[0098] Unlike existing methods that rely on simple geometric interpolation or physical models, this invention employs a data-driven prediction method.

[0099] Choosing a Long Short-Term Memory (LSTM) neural network or a Convolutional Long Short-Term Memory (CNN-LSTM) network as the core prediction model offers significant advantages in processing time-series data and capturing complex nonlinear motion patterns. During training, the model's input is a fused sequence of historical flight data (including position, attitude, velocity, acceleration, and environmental data), and its output is the predicted 3D trajectory of the aircraft over a future period. By performing differencing and segmented standardization preprocessing on the original flight data, the data can be effectively normalized, improving the model's training efficiency and prediction accuracy.

[0100] Instead of relying on complex and difficult-to-accurate physical dynamics models of drones, it infers future trajectories by learning from historical motion states, effectively predicting potential deviations caused by complex external factors such as wind, airflow, and traffic congestion, thereby achieving a fundamental shift from passive response to proactive prediction.

[0101] 3. Deviation Calculation and Risk Classification Assessment

[0102] The early warning mechanism of this solution is not just a simple deviation detection, but an intelligent system that comprehensively evaluates multiple factors.

[0103] Deviation calculation: The cloud-based early warning server compares the predicted trajectory with the preset flight path in real time and calculates the distance deviation between the two in three-dimensional space.

[0104] Risk Assessment Model: This invention establishes a comprehensive risk assessment model, the output of which is a risk level score. Assessment factors include not only the deviation distance and speed of the predicted trajectory, but also real-time airspace traffic density, meteorological conditions (such as wind speed), airspace characteristics (whether it is a no-fly zone or restricted area), and the aircraft's own condition. This multi-dimensional assessment can balance the relationship between safety and efficiency, avoiding inefficient airspace utilization due to overemphasis on safety, and effectively solving the dilemma of "strict control leading to stagnation, and lax control leading to congestion."

[0105] 4. Issuance of tiered early warning and corrective instructions

[0106] Based on the risk assessment results, the system will trigger different levels of early warning and intervention measures, thus achieving refined management of security.

[0107] Level 1 Warning (Slight Deviation): When the risk score is low, it indicates that the aircraft has a slight deviation, but the risk is manageable. The cloud-based warning server will immediately send fine-tuning instructions to the aircraft via ground base stations. After receiving the instructions, the onboard unit integrates with the flight control system (FCS) supercontrol coupler to achieve automatic fine-tuning of flight attitude and trajectory. This process requires no manual intervention, greatly improving response speed and system efficiency.

[0108] Level 2 Warning (Significant Deviation): When the risk score reaches a moderate level, it indicates a potential collision or safety risk. The cloud-based warning server will automatically issue audible and visual alarms to ground controllers or operators, while providing real-time deviation data and the optimal correction plan, awaiting manual confirmation or intervention.

[0109] Level 3 Warning (High-Risk Deviation or Unauthorized Flight): When the risk score is extremely high, or unregistered flight behavior is detected, the system will immediately issue the highest level of alert to the relevant regulatory authorities and may issue a forced return or landing order to ensure public safety.

[0110] 5. Quantitative Risk Assessment and Tiered Early Warning Model

[0111] The abstract concept of "risk" is transformed into a concrete and quantifiable mathematical model, which defines in detail the inputs, calculation process and output of risk assessment services.

[0112] (1) Multifactor risk assessment framework

[0113] This model goes beyond simple judgments based solely on deviation distance. By aggregating multiple weighted risk factors, it calculates a comprehensive risk score R, reflecting the complexity and context-dependent nature of flight risks. This quantitative model is key to intelligent decision-making, transforming the output of predictive models into actionable intelligence. It is this refined measurement of risk, rather than a simple binary judgment, that enables the system to take intervention measures commensurate with the risk level. This effectively avoids the "overreaction to minor deviations" problem, achieving a balance between safety and efficiency.

[0114] (2) Quantification of individual risk factors

[0115] (3) Predicted deviation magnitude ( )

[0116] This factor quantifies the severity of the predicted trajectory deviating from the planned route.

[0117] in, It is a prediction of the three-dimensional position at a future time step t. T represents the planned location, and T represents the prediction time domain.

[0118] (4) Airspace traffic density This factor assesses the risk of collision with other aircraft. The density is calculated within a cylindrical airspace volume around the predicted path of the aircraft.

[0119]

[0120]

[0121] in, This refers to the number of other aircraft within that volume. This refers to the volume size, and f is a scaling function that maps density to a risk value. Data on surrounding aircraft is provided by a cloud-based air traffic control platform.

[0122] (5) Adjacent restricted airspace ): This factor measures the risk of intruding into a no-fly zone (NFZ) or other restricted area.

[0123]

[0124]

[0125] in, It is the set of all restricted airspaces, and dist is the minimum distance from the predicted trajectory to the nearest region boundary. It is a scaling constant. When the predicted trajectory is very close to the danger zone, this function will produce a very high risk value.

[0126] (6) Unfavorable weather index ( )

[0127] This factor integrates real-time meteorological data obtained from cloud platforms, such as wind speed, gust intensity, and precipitation rate.

[0128]

[0129] The advanced nature of this risk model lies in its reliance on a cloud-based architecture. Risk factor data such as traffic density, dynamic no-fly zones, and real-time weather cannot be obtained by the airborne unit itself; they must be provided by a cloud-based UTM service with global situational awareness. This indicates that the core advantage of this invention lies not only in its superior airborne algorithm but also in its system-level capability to leverage information from the entire network to make more intelligent decisions for individual aircraft.

[0130] (7) Tiered early warning threshold

[0131] The system uses continuous Scoring is used to trigger discrete warning levels, enabling refined management.

[0132] Level 1 Warning (Slight Deviation): When At that time, an automated trajectory fine-tuning command is triggered. Level 2 warning (significant deviation): When In such cases, an alert is issued to the ground operator, along with suggested corrective measures.

[0133] Level 3 Warning (High-Risk Deviation): When In such cases, the highest level of alert is issued to regulatory agencies, and instructions such as mandatory return to base or landing at the nearest airport can be given.

[0134] These thresholds (T1, T2, T3) will be set based on security case analysis and relevant regulatory standards.

[0135] Specific weights can be determined based on historical accident data using methods such as the Analytic Hierarchy Process (AHP) or entropy weighting. For example, prioritizing life safety can assign higher weights to risk factors in adjacent restricted airspace and airspace traffic density. Classification thresholds can be set after statistical analysis of risk scores from a large number of historical safe flight and accident cases.

[0136] like Figure 3 As shown, a large-scale low-altitude aircraft deviation early warning system includes:

[0137] Ground communication base station and control center: includes an early warning / command issuance module, a 5G-A communication module, and a ground control platform, which are responsible for data / command transmission and ground control-related functions.

[0138] The cloud-based early warning server, which includes a trajectory prediction engine, a risk assessment module, a database (storing historical trajectories / airspace information), and blockchain evidence storage nodes, is the core processing unit for trajectory prediction, risk assessment, and data storage.

[0139] Low-altitude aircraft: Equipped with an airborne control unit H, an airborne communication module, and an airborne sensor module J, enabling communication and interaction with the outside world through the airborne communication module.

[0140] The user terminal / regulatory operation platform synchronizes data with the blockchain evidence storage nodes.

[0141] The processing mechanism is as follows:

[0142] Data interaction link: The low-altitude aircraft transmits data / commands to the ground communication base station and control center through the airborne communication module and 5G-A communication module; the ground communication base station and control center simultaneously interact with the cloud-based early warning server.

[0143] Cloud-based core processing: After the trajectory prediction engine generates the trajectory, it is transmitted to the risk assessment module to complete the risk assessment and issue instructions; the database stores historical trajectory / airspace information, and the blockchain notarization node is responsible for data notarization and synchronization to the user terminal / operation platform.

[0144] Early warning and command flow: Commands generated in the cloud, combined with the early warning / command distribution modules of ground communication base stations and control centers, are ultimately transferred to low-altitude aircraft or relevant parties.

[0145] In summary, this invention proposes a large-scale low-altitude aircraft deviation early warning method and system. By fusing multi-source heterogeneous data, employing a deep learning-based trajectory prediction method, constructing an air-ground collaborative distributed architecture, and introducing a blockchain-based evidence storage mechanism, a proactive, global, and reliable intelligent early warning system is established. The significant innovative contribution of this invention lies in upgrading the traditional, single, and passive deviation detection mode to a multi-source collaborative, data-driven, and proactive prediction early warning paradigm, achieving a fundamental shift from "post-event correction" to "pre-event early warning."

[0146] Compared with existing technologies, the significant technical advantages of this invention are reflected in the following aspects: its core principle realizes a leap from passive response to active prediction, enabling it to predict trajectories in advance and intervene proactively based on deep learning models, effectively breaking through the control bottleneck under high traffic conditions; by integrating airborne sensor data and external environmental data, it significantly improves the accuracy and robustness of positioning and prediction, overcoming the limitations of a single data source; adopting a cloud-airborne collaborative distributed architecture, it possesses strong scalability, supporting global processing and optimization control of massive amounts of aircraft data; introducing blockchain technology ensures the integrity and immutability of key data, enhancing regulatory credibility and traceability; in addition, the constructed multi-dimensional quantitative risk assessment model integrates multiple variables such as deviation distance, airspace density, environmental factors, and airspace properties, achieving a balance between safety and efficiency, and improving the overall utilization efficiency of airspace.

[0147] This invention also provides a storage medium for storing a computer program, which, when executed, performs at least the methods described above.

[0148] This invention also provides a control device, including a processor and a storage medium for storing a computer program; wherein the processor executes the computer program by performing at least the method described above.

[0149] This invention also provides a processor that executes a computer program, at least performing the methods described above.

[0150] The storage medium can be implemented by any type of non-volatile storage device, or a combination thereof. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc or CD-ROM; magnetic surface memory can be disk storage or magnetic tape storage. The storage media described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0151] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0152] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0153] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0154] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0156] The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0157] The features disclosed in the several product embodiments provided by this invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0158] The features disclosed in the several method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0159] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various equivalent substitutions or obvious modifications can be made without departing from the concept of the present invention, and all such modifications, achieving the same performance or application, should be considered within the scope of protection of the present invention.

Claims

1. A method for early warning of large-scale low-altitude aircraft deviation, characterized in that, Includes the following steps: S1. Multi-source heterogeneous data fusion and flight state estimation: Real-time acquisition of multi-source heterogeneous data from aircraft onboard sensors and external environmental data sources, and the data fusion algorithm is used to fuse and process the data for flight state estimation, generating high-precision real-time aircraft state information; the data fusion algorithm for fusing and processing the data for flight state estimation specifically involves: using an extended Kalman filter algorithm, taking a multi-dimensional state vector containing quaternions of aircraft position, velocity, attitude, and sensor zero-bias estimates as the system state, and performing state prediction based on a nonlinear state transition function, wherein the state transition function updates velocity, position, and attitude by integrating calibrated inertial measurement unit readings; then using measurement values ​​from multiple sensors and data sources, the predicted state and error covariance are iteratively updated and corrected by calculating measurement residuals, innovation covariance, and Kalman gain, thereby achieving high-precision estimation of the aircraft state, and online estimation and compensation of sensor zero bias to suppress error accumulation; S2. Flight trajectory prediction based on deep learning: Based on the real-time status information of the aircraft, the future three-dimensional trajectory of the aircraft is predicted using a prediction method based on deep learning. S3. Multi-factor quantitative risk assessment and classification: The predicted future three-dimensional trajectory is compared with the preset planned route, and multiple risk factors are quantitatively calculated to obtain a comprehensive risk score. Different levels of early warning are triggered based on the comprehensive risk score. Specifically, it includes: Calculate the three-dimensional Euclidean distance between the predicted trajectory and the planned route at each predicted time step in the future, and quantify the maximum distance value in all time steps as the deviation magnitude risk factor. Based on the global airspace information provided by the cloud-based air traffic control platform, and taking the predicted trajectory as the center, the number density of other aircraft in the preset space volume around the airspace it passes through in the future is calculated, and the density is quantified into an airspace traffic density risk factor through a preset mapping function. Calculate the minimum distance from the predicted trajectory to all preset restricted airspace boundaries, and quantify this minimum distance into a risk factor for adjacent restricted airspace using an exponential decay function; Based on real-time acquired meteorological data, one or more meteorological parameters, including wind speed, gust intensity, and precipitation rate, are quantified into adverse meteorological risk factors through a weighted linear combination. The deviation range risk factor, airspace traffic density risk factor, adjacent restricted airspace risk factor, and adverse weather risk factor are weighted and aggregated according to preset weights to calculate the comprehensive risk score. S4. Tiered Early Warning and Command Issuance: Based on the warning level, corresponding warning information or control commands are issued to the target aircraft, ground controllers, or regulatory authorities through the air-to-ground cooperative communication network. When a Level 1 warning is triggered, an automatic trajectory fine-tuning command is issued to the target aircraft. When the automatic trajectory fine-tuning command is issued to the target aircraft, the command is received by the aircraft's onboard unit and executed by the onboard unit through the super-control coupler integrated with the flight control system, thereby achieving automatic adjustment of the flight attitude and trajectory without manual intervention.

2. The large-scale low-altitude vehicle deviation early warning method according to claim 1, characterized in that, In step S1, the external environment data sources include ground communication base stations and 5G-A remote sensing integrated base stations that provide supplementary location information, real-time meteorological services that provide wind speed and wind direction data, and air traffic management platforms that provide dynamic airspace status information.

3. The large-scale low-altitude vehicle deviation early warning method according to claim 1, characterized in that, In step S2, the deep learning-based prediction method employs a long short-term memory neural network or a convolutional long short-term memory network. The network input is a sequence of historical flight state data after data fusion and preprocessing. The preprocessing includes differentiating and segmenting the original data to normalize the data. The network output is a three-dimensional trajectory sequence of the aircraft in the future prediction time domain.

4. The large-scale low-altitude vehicle deviation early warning method according to claim 1, characterized in that, Step S3 also includes setting tiered warning thresholds: preset a first threshold, a second threshold, and a third threshold, the thresholds being set based on security case analysis and regulatory standards; When the comprehensive risk score is greater than or equal to the first threshold and less than the second threshold, a level one warning is triggered. When the comprehensive risk score is greater than or equal to the second threshold and less than the third threshold, a level-two warning is triggered; When the comprehensive risk score is greater than or equal to the third threshold, a level three warning is triggered.

5. The large-scale low-altitude vehicle deviation early warning method according to claim 4, characterized in that, In step S4, the tiered early warning and instruction issuance specifically also includes: When a Level 2 warning is triggered, an audible and visual alarm is issued to ground controllers or operators, and the optimal correction scheme generated based on forecast data and airspace status is provided at the same time. When a Level 3 warning is triggered, the highest level alarm will be issued to the relevant regulatory authorities, and a forced return or forced landing order may be issued.

6. The method for early warning of large-scale low-altitude aircraft deviation according to claim 1, characterized in that, It also includes the following steps: S5. Key Data Storage and Traceability: Utilize blockchain technology to encrypt and store key data, early warning logs, and issued instructions during flight to ensure data integrity and immutability.

Images