Method for identifying aircraft flight trajectory parameters based on inertial devices
By employing an IMU-based deep learning method, the flight trajectory parameters of fire extinguishing projectiles are identified online using prior data and theoretical models. This solves the problems of system complexity and cost caused by relying on external signal sources in existing technologies, and realizes low-cost flight trajectory analysis and intelligent triggering of the fuze system.
Patent Information
- Application Number
- CN202610937546.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-26
- Publication Date
- 2026-11-03
AI Technical Summary
Existing fire extinguishing projectile trajectory identification technology relies on external signal sources, which increases system complexity and cost, and makes it impossible to achieve low-cost ballistic analysis and security system signal triggering.
A flight trajectory parameter identification method based on inertial devices (IMUs) is adopted. By using deep learning to utilize prior data and theoretical models, combined with IMU measurement data, flight trajectory parameters are identified online, enabling passive identification and intelligent triggering of the fuse system to disarm the safety device.
It enables low-cost flight trajectory parameter analysis and security system signal triggering, precisely controls the timing of safety release, and ensures the safety and intelligence of the fuse system.
Smart Images
Figure CN123028583A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aircraft trajectory recognition technology, specifically to a method for identifying flight trajectory parameters based on low-cost inertial measurement unit (IMU) measurement and intelligent fitting algorithms. Background Technology
[0002] In the structural application of fire extinguishing grenades, proximity fuses generally consist of an ignition control system, a safety system, a detonation sequence, and an energy device. Its main structural difference from other fuses lies in the complexity of its ignition control system. The ignition control system of a proximity fuse typically comprises a sensing device, a signal processing device, and an execution device. When the fire extinguishing grenade approaches the target, the fuse's inductive sensing device senses target information based on certain inherent characteristics of the target and surrounding environment's physical fields (such as electromagnetic fields, light intensity fields, sound fields, electrostatic fields, pressure fields, and magnetic fields), or changes in the physical fields around the target caused by the target's presence. This sensed information is then transmitted to the signal processing device (signal processing circuit). The signal processing device processes the received signal and, after meeting certain requirements, outputs an activation signal to the execution device. The execution device then outputs an initiation signal to the detonation sequence, igniting the electric detonating element and detonating the payload. Clearly, regardless of whether the detection principle is based on radio waves, magnetic fields, or sound, it requires sensing signal sources outside the aircraft to generate the crucial signal that triggers the contact fuse safety. In this context, to achieve the proximity detonation function of fire extinguishing bombs, it is necessary to install a special detection device on their structure to receive external signals, making the fuse system more complex and expensive.
[0003] In recent years, flight trajectory recognition technology has developed into one of the core technologies of navigation, guidance, and control systems. Currently, the effective identification characteristics of fire extinguishing projectiles at various stages of flight are mainly determined by active detection or independent equipment detection. For example, radar is used to detect ballistic data to determine its trajectory type and key positions, or navigation information provided by Global Positioning System (GPS) is relied upon to calculate trajectory information. In fact, based on the identification of flight vehicle parameters, such as real-time coordinates, velocity, and corresponding time, it is also possible to determine the moment before a specific projectile or rocket lands and thus determine the triggering conditions.
[0004] However, regardless of whether radar detection or GPS positioning is used, equipment still needs to be mounted on the fire extinguishing projectile to receive external signals. Specifically, radar ballistic measurement requires ground-based radar detection equipment to continuously transmit pulses to obtain real-time ballistic parameters; while using a GPS positioning system requires an onboard GPS receiver to acquire spread spectrum signals transmitted by GPS satellites, perform positioning calculations to obtain real-time coordinates and velocity information, and then complete ballistic data processing and measurement.
[0005] Both of these methods require externally provided information to complete the calculations and measurements. Obviously, when used for ballistic measurements of aircraft flight vehicles, this will increase the complexity of the system and the overall cost. Summary of the Invention
[0006] To address the above problems, this invention proposes a ballistic trajectory parameter identification method based on inertial devices. It utilizes a passive IMU device to acquire the acceleration sequence in the carrier coordinate system, and performs deep learning based on experimental data and theoretical model calculation results to summarize the characteristics of key position points. This enables online flight trajectory parameter analysis and triggers the actuator at a pre-selected ballistic point to disarm the security device, achieving the functions of ballistic analysis and security system signal triggering at extremely low cost.
[0007] The technical solution of this invention is as follows: Considering that different aircraft, such as different types of fire extinguishing bombs, typically have diverse flight trajectories or ballistic trajectories, but for a specific type of aircraft, its flight trajectory remains basically fixed; accordingly, this invention requires collecting prior data based on historical and past flight trajectory data of various aircraft and meteorological characteristic parameters. Then, deep learning is performed using the collected prior data and the calculation results of the theoretical flight trajectory model. A prediction model is obtained by training with historical data, and the prediction model is input into the aircraft's trajectory sensing unit. Then, flight trajectory parameters are identified, including acceleration, attitude, time, flight speed, and current position. The flight trajectory parameter identification specifically includes the following steps:
[0008] Step 1: Transmit the action time or target distance to the trajectory sensing unit through a preset setting method;
[0009] Step 2: After the aircraft launches and takes off, the system power is triggered to control the trajectory sensing unit to initialize. Then, based on the time recorded by the microcontroller of the measurement unit in the aircraft and the IMU measurement data, the measurement unit includes a microcontroller and an IMU device, which includes a three-axis gyroscope, a three-axis accelerometer, and a magnetometer.
[0010] Step 3: The trajectory sensing unit transforms the real-time acceleration obtained from the IMU measurement data to obtain the aircraft's overload value;
[0011] Step 4: Based on the real-time roll rate and overload value of the aircraft, the trajectory sensing unit performs normalization calculations and merges them with the prediction model pre-input into the trajectory sensing unit to quickly match the specific model and meteorological characteristic parameters of the aircraft.
[0012] Step 5: Based on the matched specific model and meteorological characteristic parameters, the trajectory perception unit, combined with real-time IMU measurement data, extrapolates the aircraft's flight trajectory and obtains key point information, including the aircraft's highest point and estimated time before landing.
[0013] Step six: Repeat steps three through five to repeatedly verify the specific model and meteorological characteristic parameters of the aircraft until the moment when the setting conditions are met, the trajectory sensing unit sends a trigger signal, and the fuze system completes the predetermined release action and function.
[0014] During the training of the predictive model, a multi-source data fusion framework is first established using prior data on the aircraft, including:
[0015] Prior flight trajectory data is acquired and spatiotemporally processed; a six-degree-of-freedom flight trajectory model is used for numerical solution, and deep learning based on physical information neural networks is carried out; meteorological random disturbance factors, i.e. meteorological characteristic parameters, are considered and input in the form of numerical sampling during training.
[0016] The implementation logic of the trajectory sensing unit is as follows:
[0017] (1) Data acquisition layer;
[0018] IMU devices were used to acquire triaxial acceleration data, angular velocity data, and magnetometer data.
[0019] (2) Preprocessing module;
[0020] Based on real-world data from a triaxial accelerometer, axis normal separation is performed, and the vector form calculation process is as follows:
[0021] , ;
[0022] in, For real-time acceleration vectors, For axial acceleration, For tangential acceleration, It is real-time speed;
[0023] (3) Feature matching models and methods;
[0024] The key point judgment logic for the measured flight trajectory parameters is as follows: when the axial acceleration is continuously greater than the threshold for 50ms, it is determined to be the initial point; when the overload value reaches an inflection point, it is determined to be the highest point; when the similarity with the preset trajectory is greater than 90%, it is determined to be at the end position.
[0025] During flight, the aforementioned trajectory sensing unit quickly identifies the aircraft's specific model and meteorological characteristic parameters, and continuously verifies them online to improve accuracy.
[0026] The verification method uses Kalman filtering to predict the flight trajectory and compares it with the actual measurement to ensure that the predicted data and the actual situation are within the allowable error. If the selected flight trajectory parameters are matched with the actual measured data at the next moment within a certain range after filtering, the matching parameters are considered successful.
[0027] Considering that existing flight trajectory parameter identification methods typically rely on radar measurements and GPS satellite positioning systems, requiring external measurement systems or dedicated built-in receivers to collect, analyze, and calculate measurement information, and are susceptible to interference, this invention utilizes deep learning based on prior data and theoretical flight trajectory model calculations to summarize the characteristics of key location points. This enables online flight trajectory parameter analysis and triggers actuators at pre-selected locations to disarm the safety device, achieving flight trajectory parameter analysis and safety system signal triggering functions at extremely low cost. The feature point signals implemented in this invention are highly correlated with the entire flight trajectory, allowing for precise control of the safety disarming timing and ensuring the safety of the fuze throughout the entire process. This provides a more intelligent and cost-effective way to achieve the fuze system's safety disarming function and action signal acquisition function. Attached Figure Description
[0028] Figure 1 It is an online system for identifying and predicting the trajectory and evaluating the match between the trajectory and selected parameters. Detailed Implementation
[0029] To clearly illustrate the technical features of the present invention, the present invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings.
[0030] This invention addresses the unique ballistic parameters of specific flight vehicles (such as a fire extinguishing bomb). Considering available meteorological parameters, it employs deep learning based on prior data and theoretical flight trajectory model calculations. By training a classification model with historical data to improve recognition efficiency and summarizing the characteristics of key locations, it proposes a method for identifying flight trajectories by using a passive inertial device (IMU) to acquire overload data online. This method is integrated into a hardware module (measurement unit) in the form of an algorithm, merging with the aircraft's existing fuze security system. At a pre-selected location, the actuator is triggered to disarm the security device, thus achieving both fuze system disarming and signal acquisition functions in a more intelligent and cost-effective manner.
[0031] This invention pre-programs the trajectory sensing unit with input action time or target distance, enabling it to deduce characteristic points of the flight trajectory (such as the highest point or a specific time before landing), and then deactivate the fuse at a specified time. Based on a deep learning-based intelligent fitting model, it can quickly match flight parameters (such as overload characteristics). The intelligent fitting model integrates an atmospheric model, which can deduce key parameters in the atmospheric model based on continuous drag and overload changes identified during flight, and then use these parameters for fitting the flight trajectory.
[0032] First, prior data is collected based on the historical flight trajectory data of various aircraft and meteorological characteristic parameters. Then, deep learning is carried out using the collected prior data and the calculation results of the theoretical flight trajectory model, and a prediction model is obtained by training with historical data.
[0033] When in use, the action time or target distance is transmitted to the trajectory sensing unit through a preset setting method;
[0034] After the aircraft takes off and flies, the system power is triggered to control the trajectory sensing unit to initialize. Then, based on the time recorded by the microcontroller of the measurement unit in the aircraft and the IMU measurement data, the measurement unit includes a microcontroller and an IMU device, and the IMU device includes a three-axis gyroscope, a three-axis accelerometer, and a magnetometer.
[0035] Based on the real-time roll rate and overload value of the aircraft, the trajectory sensing unit performs normalization calculations and merges them with the prediction model pre-input into the trajectory sensing unit to quickly match the specific model and meteorological characteristic parameters of the aircraft.
[0036] Based on matching specific aircraft model and meteorological characteristic parameters, combined with real-time IMU measurement data, the trajectory sensing unit extrapolates the aircraft's flight trajectory and obtains key information, including the aircraft's highest point and estimated landing time.
[0037] Throughout the entire flight, the aircraft acquires IMU data in real time, repeatedly verifies the model and weather characteristics, until the moment when the set conditions are met, the trajectory sensing unit sends a trigger signal, and the fuse system completes the predetermined safety release action and function.
[0038] Since the algorithm model has certain prior data, on the other hand, the established six-degree-of-freedom flight trajectory model (as shown below) introduces the necessary motion description coordinate system to describe three translational degrees of freedom (position change) and three rotational degrees of freedom (attitude change).
[0039]
[0040]
[0041]
[0042]
[0043]
[0044]
[0045] in:
[0046] m is the mass of the projectile, and V is the velocity. The trajectory inclination angle;
[0047] , and These represent the projected components of the force acting on the projectile along each axis of the ballistic coordinate system.
[0048] , and These represent the moments of inertia of each axis of the projectile.
[0049] , and The components of the projectile's angular velocity along each axis of the projectile's coordinate system;
[0050] , and These are the components of the torque acting on the projectile along each axis of the projectile's coordinate system.
[0051] The aerodynamic parameters are derived from CFD models or wind tunnel test results. The aircraft's body data (such as mass, center of mass, and moment of inertia) can be used to extract key ballistic parameters under specific operating conditions (weather, altitude, etc.), such as acceleration, angular velocity, trajectory curvature, and rate of change of velocity.
[0052] It is worth noting that different types of aircraft have their own flight trajectories under specific operating conditions (weather, altitude, etc.), such as the damping characteristics at the highest point. In order to adapt to a series of common aircraft types, these will be extracted and introduced as key features for flight trajectory recognition.
[0053] In addition, prior data from different aircraft operating conditions can be used to establish a multi-source data fusion framework, including:
[0054] Prior flight trajectory data (from actual measurements) is acquired and spatiotemporally processed; a six-degree-of-freedom ballistic model numerical solution is used to conduct deep learning based on a physical information neural network (PINN); considering random weather disturbances, the data is input in the form of numerical sampling during training.
[0055] This architecture achieves joint optimization of experimental data and ballistic equations by embedding physical constraints into a neural network structure. Its innovation lies in the dynamic nature of the parameters, which can adapt to environmental changes such as wind speed and humidity in real time.
[0056] Taking the actual flight process of a fire extinguishing projectile as an example, the IMU collects acceleration (overload) and angular velocity information, decomposes information such as angular and axial acceleration, and identifies the initial point, highest point, and (set) terminal position. Based on the above theory and method, the implementation logic of the entire ballistic sensing unit is as follows:
[0057] (1) Data acquisition layer;
[0058] A 9-axis IMU (such as MPU-9250) was used to acquire triaxial acceleration (±16g), angular velocity (±2000° / s) and magnetometer data, with a sampling rate of 1kHz36.
[0059] (2) Preprocessing module;
[0060] Based on accelerometer data, axis-normal separation (damping identification) is performed. The vector form calculation process is as follows:
[0061] , ;
[0062] in, For real-time acceleration vectors, For axial acceleration, This is the tangential acceleration.
[0063] (3) Feature matching models and methods;
[0064] The key criteria for determining the measured ballistic parameters are as follows.
[0065]
[0066] During flight, the aforementioned training model is used to quickly identify the type of fire extinguishing bomb, and continuous online verification is performed to improve accuracy. The verification method involves using Kalman filtering (as shown below) to predict the flight trajectory, comparing it with actual measurements to ensure that the predicted data (including speed, overload, etc.) remains within the allowable error range. See attached figure. Figure 1 If the filtered prediction of the selected projectile parameters matches the measured data at the next moment within a certain range, the parameter matching is considered successful.
[0067] The above verification is to confirm that the most suitable projectile model has been matched. Since the trajectory parameter recognition model of the ballistic sensing unit can be applied to a series of projectile objects, it relies on online detection data and matching with the previously trained model to complete the process. Considering factors such as error, continuous online verification and confirmation are necessary through the above process.
[0068] Once the specific fire extinguishing bomb model and corresponding operating conditions are confirmed, the flight trajectory time can be deduced. Then, according to the set parameters, a serial port message can be sent at a specific time and location to trigger the system mechanism to complete its operation, thus achieving the response action of the security mechanism. This method solves the problems of high cost and complexity associated with current systems that rely on external measuring equipment or built-in receiving devices.
[0069] There are many specific ways to implement this invention. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.
Claims
1. A method for identifying aircraft flight trajectory parameters based on inertial devices, characterized in that, Prior data is collected based on historical flight trajectory data and meteorological parameters of various aircraft. Then, deep learning is performed using the collected prior data and theoretical flight trajectory model calculations. A prediction model is trained using historical data and input into the aircraft's trajectory sensing unit. Flight trajectory parameters are then identified, including acceleration, attitude, time, flight speed, and current position. The flight trajectory parameter identification specifically includes the following steps: Step 1: Transmit the action time or target distance to the trajectory sensing unit through a preset setting method; Step 2: After the aircraft launches and takes off, the system power is triggered to control the trajectory sensing unit to initialize. Then, based on the time recorded by the microcontroller of the measurement unit in the aircraft and the IMU measurement data, the measurement unit includes a microcontroller and an IMU device, which includes a three-axis gyroscope, a three-axis accelerometer, and a magnetometer. Step 3: The trajectory sensing unit transforms the real-time acceleration obtained from the IMU measurement data to obtain the aircraft's overload value; Step 4: Based on the real-time roll rate and overload value of the aircraft, the trajectory sensing unit performs normalization calculations and merges them with the prediction model pre-input into the trajectory sensing unit to quickly match the specific model and meteorological characteristic parameters of the aircraft. Step 5: Based on the matched specific model and meteorological characteristic parameters, the trajectory perception unit, combined with real-time IMU measurement data, extrapolates the aircraft's flight trajectory and obtains key point information, including the aircraft's highest point and estimated time before landing. Step six: Repeat steps three through five to repeatedly verify the specific model and meteorological characteristic parameters of the aircraft until the moment when the setting conditions are met, the trajectory sensing unit sends a trigger signal, and the fuze system completes the predetermined release action and function.
2. The method for identifying aircraft flight trajectory parameters based on inertial devices according to claim 1, characterized in that, During the training of the predictive model, a multi-source data fusion framework is first established using prior data on the aircraft, including: Prior flight trajectory data is acquired and spatiotemporally processed; a six-degree-of-freedom flight trajectory model is used for numerical solution, and deep learning based on physical information neural networks is carried out; meteorological random disturbance factors, i.e. meteorological characteristic parameters, are considered and input in the form of numerical sampling during training.
3. The method for identifying aircraft flight trajectory parameters based on inertial devices according to claim 1, characterized in that, The implementation logic of the trajectory sensing unit is as follows: (1) Data acquisition layer; IMU devices were used to collect triaxial acceleration data, angular velocity data, and magnetometer data. (2) Preprocessing module; Based on real-world data from a triaxial accelerometer, axis normal separation is performed, and the vector form calculation process is as follows: ， ； in, For real-time acceleration vectors, For axial acceleration, For tangential acceleration, It is real-time speed; (3) Feature matching models and methods; The key point judgment logic for the measured flight trajectory parameters is as follows: when the axial acceleration is continuously greater than the threshold for 50ms, it is determined to be the initial point; when the overload value reaches an inflection point, it is determined to be the highest point; when the similarity with the preset trajectory is greater than 90%, it is determined to be at the end position.
4. The method for identifying aircraft flight trajectory parameters based on inertial devices according to claim 1, characterized in that, During flight, the aforementioned trajectory sensing unit quickly identifies the aircraft's specific model and meteorological characteristic parameters, and continuously verifies them online to improve accuracy.
5. The method for identifying aircraft flight trajectory parameters based on inertial devices according to claim 1, characterized in that, The verification method uses Kalman filtering to predict the flight trajectory and compares it with the actual measurement to ensure that the predicted data and the actual situation are within the allowable error. If the selected flight trajectory parameters are matched with the actual measured data at the next moment within a certain range after filtering, the matching parameters are considered successful.