Aircraft cluster unit intention identification method based on multi-model mechanism

Abstract

The invention discloses an aircraft cluster unit intention identification method based on a multi-model mechanism. The method comprises the following steps: constructing an aircraft dynamics model and a measurement model; based on the current measurement result of the our aircraft, identifying the constructed model to obtain the target of the enemy aircraft; in the identification process, a volume Kalman filtering method and an interactive multi-model method are fused for identification. According to the method disclosed by the invention, the dynamic characteristics and the guidance characteristics of the aircraft are fully considered, and the accuracy and the real-time performance of the identification process are effectively enhanced through a multi-model cooperative working mode.

Description

Technical Field

[0001] The present invention relates to an aircraft cluster unit intention identification method based on a multi-model mechanism, and belongs to the technical field of aircraft control. Background Art

[0002] Aircraft cluster unit intention recognition can clearly identify the intention of each target in the target aircraft cluster, so as to carry out targeted defense, interception, avoidance and other actions.

[0003] The use of multi-model filtering technology can simultaneously consider multiple possible target dynamic models and adaptively select the model that best suits the current situation based on the observed data. Its characteristic is that it can flexibly respond to the diversity and uncertainty of target behavior, which makes it a reliable choice for solving unknown models and dynamic change problems such as intent recognition.

[0004] Existing multi-model filtering technology mainly uses Multiple Model Adaptive Estimation (MMAE) for filtering scenarios where the model is uncertain but does not change over time. It filters all possible unknown models separately, runs all filters synchronously, uses the Gaussian density function to determine the degree of match between the current model and all possible unknown models, and finally outputs a mixed estimation result. For filtering scenarios where the model is uncertain and changes over time, the main method is Interactive Multiple Model (IMM). Its basic principle is to interactively input based on the known model probability transfer matrix, synchronously run multiple model filtering units, and interactively output the filtering results based on the updated model probabilities of the filtering units.

[0005] However, existing multi-model filtering techniques are effective in relatively simple, general target intention recognition scenarios (such as determining a car's destination). However, their effectiveness is significantly reduced in complex spatial environments. In aerial environments, aircraft have unique dynamic and guidance characteristics. Because generalized kinematic models fail to fully account for these unique characteristics, they are unable to accurately capture the complex and ever-changing trajectory of an aircraft in the air. This significantly reduces the accuracy of target intention recognition, making it difficult to meet the practical needs of rapid and accurate target intention judgment in the air.

[0006] Therefore, it is necessary to conduct a more in-depth study on the existing aircraft cluster unit intention identification methods to solve the above problems. Summary of the Invention

[0007] To overcome the above problems, an in-depth study was conducted and a method for identifying the intention of aircraft cluster units based on a multi-model mechanism was proposed, which includes the following steps:

[0008] Construct aircraft dynamics models and measurement models;

[0009] Based on the current measurement results of our aircraft, the constructed model is identified to obtain the target of the enemy aircraft.

[0010] In a preferred embodiment, the aircraft dynamics model is expressed as:

[0011]

[0012]

[0013] Where, the subscript i represents different friendly aircraft, i = 1,...,n, n represents the total number of friendly aircraft, the subscript m represents the enemy aircraft, ρ i represents the distance between the enemy aircraft and our aircraft i, λ i represents the azimuth angle between the enemy aircraft and our aircraft i, v m represents the speed of the enemy aircraft, γ m Indicates the flight path angle of the enemy aircraft, a m Indicates the acceleration of the enemy aircraft, u m Indicates the acceleration command of the enemy aircraft, v t,i represents the speed of our i-th aircraft, γ t,i represents the flight path angle of our i-th aircraft, a t,i represents the acceleration of our i-th aircraft, u t,i represents the acceleration command of our i-th aircraft, τ represents the time constant, u m,i It represents the acceleration command generated by the guidance law used when the enemy aircraft attacks our aircraft i, and δ(T,i) is the selection function.

[0014] In a preferred embodiment, the state quantity x of the measurement model is set to:

[0015] x i =[x i y i γ i a i V i ]

[0016] Among them, for any friendly aircraft or enemy aircraft, x i 、y i Indicates its position, γ i represents the flight path angle, a i Represents its acceleration, V i Indicates its speed.

[0017] The measurement model is expressed as:

[0018]

[0019] Among them, z m (k) represents the measurement value of the state quantity by our aircraft, ρ m (k) represents the measured value of the relative distance between our aircraft and the enemy aircraft, λ m (k) represents the measured value of the relative angle between our aircraft and the enemy aircraft, ν(k) represents the measurement noise, represents Gaussian distribution, Q m is the intermediate variable, σ ρ is the standard deviation of the measurement noise of the relative distance, σ λ is the standard deviation of the azimuth measurement noise.

[0020] In a preferred embodiment, the volumetric Kalman filter method is integrated with the interactive multiple model method to identify the constructed model.

[0021] In a preferred embodiment, the fusion comprises the following steps:

[0022] S21, decoupling each individual in the cluster, taking the state estimation and probability value of different individuals at time k as input, and using the interactive input method to obtain the interactive input results of different individuals at time k;

[0023] S22. Based on the interactive input result at time k and the measurement result at time k+1, perform state prediction, measurement update, and state estimation on different individuals to obtain state estimates of different individuals at time k+1;

[0024] S23, using the state estimates of different individuals at time k+1 as input to perform probability update, performing interactive output based on the updated probability values, and using the interactive output result as the identification result at time k+1;

[0025] S24. Repeat S21-S23 to obtain the recognition result at the subsequent time.

[0026] In a preferred embodiment, in S21, the estimation result of the cubature Kalman filter is used to output the k-time mixed state estimation based on the interactive input method of the interactive multi-model;

[0027] In S22, the state estimation and covariance of the model are updated using the cubature Kalman filter method.

[0028] In S23, the probability update method in the interactive multi-model is used to perform probability update.

[0029] In S23, the interactive output method in the interactive multi-model is used for interactive output

[0030] The beneficial effects of the present invention include:

[0031] (1) The dynamic and guidance characteristics of the aircraft are fully considered, and the accuracy and real-time performance of the identification process are effectively enhanced through the collaborative working mode of multiple models;

[0032] (2) It solves the shortcomings of traditional methods in capturing motion characteristics and adapting to dynamic changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic flow chart of a method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to a preferred embodiment of the present invention is shown;

[0034] Figure 2 A schematic diagram of the attack scenario in Example 1 is shown;

[0035] Figure 3 The identification results in Example 1 are shown;

[0036] Figure 4 A schematic diagram of the attack scenario in Example 2 is shown;

[0037] Figure 5 The identification results in Example 2 are shown. DETAILED DESCRIPTION

[0038] The present invention will be described in further detail below with reference to the accompanying drawings and examples, through which the features and advantages of the present invention will become more clearly understood.

[0039] The word "exemplary" is used exclusively herein to mean "serving as an example, example, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0040] According to the present invention, a method for identifying the intention of an aircraft cluster unit based on a multi-model mechanism is provided. Figure 1 As shown, the following steps are included:

[0041] Construct aircraft dynamics models and measurement models;

[0042] Based on the current measurement results of our aircraft, the constructed model is identified to obtain the target of the enemy aircraft.

[0043] The aircraft dynamics model is expressed as:

[0044]

[0045]

[0046] Where, the subscript i represents different friendly aircraft, i = 1,...,n, n represents the total number of friendly aircraft, the subscript m represents the enemy aircraft, ρ i represents the distance between the enemy aircraft and our aircraft i, λ i represents the azimuth angle between the enemy aircraft and our aircraft i, v m represents the speed of the enemy aircraft, γ m Indicates the flight path angle of the enemy aircraft, a m Indicates the acceleration of the enemy aircraft, u m Indicates the acceleration command of the enemy aircraft, v t,i represents the speed of our i-th aircraft, γ t,i represents the flight path angle of our i-th aircraft, a t,i represents the acceleration of our i-th aircraft, u t,i represents the acceleration command of our i-th aircraft, τ represents the time constant, u m,i It represents the acceleration command generated by the guidance law used when the enemy aircraft attacks our aircraft i, and δ(T,i) is the selection function.

[0047] Preferably, the selection function is set to:

[0048]

[0049] Wherein, T represents the target of the enemy aircraft, and T=i represents that the target of the enemy aircraft is our aircraft i.

[0050] Preferably, the lateral acceleration of the enemy aircraft and our aircraft is set to be bounded, expressed as:

[0051] |u v |≤a v,max

[0052] Among them, u v Indicates the lateral acceleration instructions of the enemy aircraft and our aircraft, a v,max is the maximum lateral acceleration.

[0053] The aircraft dynamics model proposed in the present invention is a dynamic model. Compared with the traditional dynamics model, the model is more adaptable and consistent with the movement characteristics of the aircraft.

[0054] Furthermore, existing system identification models generally include structural and parameter identification, resulting in low accuracy. Research has found that guidance laws currently used in practical applications typically employ a design approach that nulls the line-of-sight angular rate, and their structures are equivalent to proportional guidance laws with different navigation coefficients. Based on this, the dynamic model in this invention uses the proportional guidance law as the structural model of the guidance law, allowing the identification of the dynamic model to focus solely on parameter identification.

[0055] The state quantity x of the measurement model is set as:

[0056] x i =[x i y i γ i a i V i ]

[0057] Among them, for any friendly aircraft or enemy aircraft, x i 、y i Indicates its position, γ i represents the flight path angle, a i Represents its acceleration, V i Indicates its speed.

[0058] The measurement model is expressed as:

[0059]

[0060] Among them, z m (k) represents the measurement value of the state quantity by our aircraft, ρ m (k) represents the measured value of the relative distance between our aircraft and the enemy aircraft, λ m (k) represents the measured value of the relative angle between our aircraft and the enemy aircraft, ν(k) represents the measurement noise, represents Gaussian distribution, Q m is the intermediate variable, σ ρ is the standard deviation of the measurement noise of the relative distance, σ λ is the standard deviation of the azimuth measurement noise.

[0061] In the present invention, a multi-model mechanism is constructed by building a dynamic model and a measurement model, and then identifying them, which fully considers the dynamic characteristics and guidance characteristics of the aircraft. Through the collaborative working mode of multiple models, the accuracy and real-time performance of the identification process are effectively enhanced.

[0062] In the present invention, the CKF (Cubature Kalman Filter) method is integrated with the interactive multi-model method to identify the constructed model.

[0063] CKF is a commonly used nonlinear model filtering method with high robustness, the ability to maintain accuracy in a wide range of system dynamics, clear volume point selection rules and integration rules, and is easy to implement. It includes state prediction, measurement update, and state update processes. In this invention, the specific process of CKF is not described in detail.

[0064] Interactive multi-model is a commonly used multi-model filtering method that has been widely used in many fields. It includes interactive input, filtering, probability update and interactive output processes. In this invention, the specific process of interactive multi-model is not described in detail.

[0065] Specifically, the fusion includes the following steps:

[0066] S21, decoupling each individual in the cluster, taking the state estimation and probability value of different individuals at time k as input, and using the interactive input method to obtain the interactive input results of different individuals at time k;

[0067] S22. Based on the interactive input result at time k and the measurement result at time k+1, perform state prediction, measurement update, and state estimation on different individuals to obtain state estimates of different individuals at time k+1;

[0068] S23, using the state estimates of different individuals at time k+1 as input to perform probability update, performing interactive output based on the updated probability values, and using the interactive output result as the identification result at time k+1;

[0069] S24. Repeat S21-S23 to obtain the recognition result at the subsequent time.

[0070] According to the present invention, the measurement results are obtained by real-time measurement by our aircraft.

[0071] In S21, the specific method of decoupling is not limited, and those skilled in the art can perform it based on experience.

[0072] In a preferred embodiment, the interactive input method refers to an interactive input process in an interactive multi-model.

[0073] In a preferred embodiment, in S22, a cubature Kalman filter method is used to perform state prediction, measurement update, and state estimation on different individuals.

[0074] In a preferred embodiment, in S23, the probability update method in the interactive multi-model is used to perform the probability update.

[0075] In a preferred embodiment, in S23, an interactive output method in an interactive multi-model is used to perform interactive output.

[0076] In the present invention, through the above preferred embodiment, the cubature Kalman filter method and the interactive multi-model method are effectively integrated. Through the integration, the updated filter state can represent the optimal state estimate of the target at the current moment. The obtained filter covariance can more accurately quantify the uncertainty of the estimation result and more accurately reflect the discreteness of the estimated value. As a result, the residual v output after filtering is j (k) and the residual covariance matrix S j (k) It can evaluate the difference between observations and predictions, which is an important indicator in the probability update stage and provides a more accurate basis for model optimization and selection.

[0077] Example

[0078] Example 1

[0079] A simulation experiment is conducted to identify enemy aircraft targets. Three enemy aircraft are set up and a constant-speed proportional guidance law is adopted. The initial conditions are shown in Table 1.

[0080] Table 1

[0081]

[0082] We have three aircraft, set to constant speed sinusoidal motion, and their acceleration instructions are set to:

[0083] u t,i =Asin(ωt)+B

[0084] Where A is the maneuver amplitude, ω is the maneuver frequency, and B is the acceleration constant. The initial conditions are shown in Table 2.

[0085] Table 2

[0086]

[0087] The identification process includes the following steps:

[0088] Construct aircraft dynamics models and measurement models;

[0089] Based on the current measurement results of our aircraft, the constructed model is identified to obtain the target of the enemy aircraft.

[0090] The aircraft dynamics model is expressed as:

[0091]

[0092] The state quantity x of the measurement model is set as:

[0093] x i =[x i yi γ i a i V i ]

[0094] The measurement model is expressed as:

[0095]

[0096] The cubature Kalman filter method is integrated with the interactive multi-model method to identify the constructed model, and the integration includes the following steps:

[0097] S21, decoupling each individual in the cluster, taking the state estimation and probability value of different individuals at time k as input, and using the interactive input method to obtain the interactive input results of different individuals at time k;

[0098] S22. Based on the interactive input result at time k and the measurement result at time k+1, perform state prediction, measurement update, and state estimation on different individuals to obtain state estimates of different individuals at time k+1;

[0099] S23, using the state estimates of different individuals at time k+1 as input to perform probability update, performing interactive output based on the updated probability values, and using the interactive output result as the identification result at time k+1;

[0100] S24. Repeat S21-S23 to obtain the recognition result at the subsequent time.

[0101] In S21, the interactive input method refers to an interactive input process using an interactive multi-model;

[0102] In S22, the cubature Kalman filter method is used to perform state prediction, measurement update and state estimation on different individuals;

[0103] In S23, the probability update method in the interactive multi-model is used to perform probability update, and the interactive output method in the interactive multi-model is used to perform interactive output.

[0104] The simulation starts at t = 0s. During the simulation, the enemy aircraft's proportional guidance coefficient N = 4, and the maximum accelerations of the enemy aircraft and our aircraft are a m,max =30g and a t,max =10g, first-order time constant τ = 0.1, sampling frequency f = 20Hz, and the Gaussian white noise σ ρ =50m,σ λ =0,001rad.

[0105] During the simulation, the enemy aircraft Vehicle1 is set to attack our aircraft Target2, the enemy aircraft Vehicle2 is set to attack our aircraft Target3, and the enemy aircraft Vehicle3 is set to attack our aircraft Target1. Figure 2 shown.

[0106] The identification results are as follows Figure 3 As shown, Figure 3 (a) shows the attack probability of the enemy aircraft Vehicle 1 on our aircraft. Figure 3 (b) shows the attack probability of the enemy aircraft Vehicle 2 on our aircraft. Figure 3 (c) shows the attack probability of the enemy aircraft Vehicle 3 on our aircraft. Figure 3 As can be seen from the image, within the 1.9-second interval, the attack probability for the inferred target exceeded 0.5, indicating that the target was the enemy aircraft's true attack target. Subsequently, as the attack progressed, the inferred attack probability for the true target increased rapidly, and the system was able to consistently and correctly identify the target before the attack ended, demonstrating high stability.

[0107] Example 2

[0108] The same experiment as in Example 1 was conducted, except that the number of our aircraft was set to 5, and the initial conditions of our aircraft were shown in Table III.

[0109] Table 3

[0110]

[0111] The target of the enemy aircraft is set to switch during the attack, as shown in Table 4. The switching time is 6 seconds after the attack. The attack scenario is as follows: Figure 4 shown.

[0112] Table 4

[0113]

[0114] The identification results are as follows Figure 5 As shown, Figure 5 (a) shows the attack probability of the enemy aircraft Vehicle 1 on our aircraft. Figure 5 (b) shows the attack probability of the enemy aircraft Vehicle 2 on our aircraft. Figure 5 (c) shows the attack probability of the enemy aircraft Vehicle 3 on our aircraft. Figure 5It can be seen that the targets of enemy cluster attacks can be effectively identified, and after the enemy aircraft switches its attack targets, it can still successfully achieve accurate identification of the enemy aircraft targets.

[0115] The present invention has been described above with reference to preferred embodiments, but these embodiments are merely exemplary and serve only as illustrations. On this basis, various replacements and improvements can be made to the present invention, all of which fall within the scope of protection of the present invention.

Claims

1. A method for identifying the intention of aircraft cluster units based on a multi-model mechanism, characterized in that: The following steps are involved: Construct aircraft dynamics models and measurement models; Based on the current measurement results of our aircraft, the constructed model is identified to obtain the target of the enemy aircraft.

2. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 1, characterized in that: The aircraft dynamics model is expressed as: Where, the subscript i represents different friendly aircraft, i = 1,...,n, n represents the total number of friendly aircraft, the subscript m represents the enemy aircraft, ρ i represents the distance between the enemy aircraft and our aircraft i, λ i represents the azimuth angle between the enemy aircraft and our aircraft i, v m represents the speed of the enemy aircraft, γ m Indicates the flight path angle of the enemy aircraft, a m Indicates the acceleration of the enemy aircraft, u m Indicates the acceleration command of the enemy aircraft, v t,i represents the speed of our i-th aircraft, γ t,i represents the flight path angle of our i-th aircraft, a t,i represents the acceleration of our i-th aircraft, u t,i represents the acceleration command of our i-th aircraft, τ represents the time constant, u m,i It represents the acceleration command generated by the guidance law used when the enemy aircraft attacks our aircraft i, and δ(T,i) is the selection function.

3. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 1, characterized in that: The state quantity x of the measurement model is set as: x i =[x i y i γ i a i V i ] Among them, for any friendly aircraft or enemy aircraft, x i 、y i Indicates its position, γ i represents the flight path angle, a i Represents its acceleration, V i Indicates its speed; The measurement model is expressed as: Among them, z m (k) represents the measurement value of the state quantity by our aircraft, ρ m (k) represents the measured value of the relative distance between our aircraft and the enemy aircraft, λ m (k) represents the measured value of the relative angle between our aircraft and the enemy aircraft, ν(k) represents the measurement noise, represents Gaussian distribution, Q m is the intermediate variable, σ ρ is the standard deviation of the measurement noise of the relative distance, σ λ is the standard deviation of the azimuth measurement noise.

4. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 1, characterized in that: The volumetric Kalman filtering method is integrated with the interactive multi-model method to identify the constructed model.

5. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 4, characterized in that: The fusion comprises the following steps: S21, decoupling each individual in the cluster, taking the state estimation and probability value of different individuals at time k as input, and using the interactive input method to obtain the interactive input results of different individuals at time k; S22. Based on the interactive input result at time k and the measurement result at time k+1, perform state prediction, measurement update, and state estimation on different individuals to obtain state estimates of different individuals at time k+1; S23, using the state estimates of different individuals at time k+1 as input to perform probability update, performing interactive output based on the updated probability values, and using the interactive output result as the identification result at time k+1; S24. Repeat S21-S23 to obtain the recognition result at the subsequent time.

6. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 5, characterized in that: In S21, the interactive input process in the interactive multi-model is used to obtain the interactive input results of different individuals at k moments.

7. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 5, characterized in that: In S22, the cubature Kalman filter method is used to perform state prediction, measurement update and state estimation on different individuals.

8. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 5, characterized in that: In S23, the probability update method in the interactive multi-model is used to perform probability update.

9. The method for identifying the intention of aircraft cluster units based on a multi-model mechanism according to claim 5, characterized in that: In S23, the interactive output method in the interactive multi-model is used for interactive output.

Images