A Maneuver Detection Method Based on Model Probability Matching

Abstract

A maneuver detection method based on model probability matching, the method comprising: analyzing different characteristic parameters when a hypersonic reentry vehicle has a maneuver mutation, constructing a maneuver model description matrix, and determining a maneuver model; defining an influence factor and designing a model matching degree function; setting thresholds for the characteristic parameters of different maneuver models, and calculating a matching vector between the current system and each preset model based on the actual characteristic parameters of the current system; substituting the prior influence factor and the calculated matching vector into the model matching degree function to deduce the model matching probability; calculating a model probability correction sequence according to the probabilities of different preset maneuver models corresponding to the true state at the current moment, and finally correcting the model probability vector; and judging whether the preset model switching condition is satisfied according to the model probability vector. According to the present invention, different maneuver models are set according to the characteristics of different maneuver mutations, adapting to the characteristics of the hypersonic reentry vehicle having extremely strong unknown maneuver characteristics.

Description

Technical Field

[0001] The present invention relates to a method for real-time maneuver detection of a hypersonic reentry vehicle, and more particularly to a maneuver detection method based on model probability matching. Background Art

[0002] Hypersonic reentry vehicles have the advantages of large combat radius, high flight speed, and strong maneuverability. It is necessary to vigorously develop interception technologies to address this emerging threat. Rapid and accurate detection and estimation of the target's motion state, and then precise prediction are prerequisites for efficient interception. However, the reentry trajectory of hypersonic reentry vehicles has extremely strong unknown maneuver characteristics. Therefore, it is urgent to study its maneuver characteristics and corresponding maneuver detection technologies to determine the maneuver state of the target, so as to provide real-time information for trajectory prediction, achieve high-precision prediction of the target trajectory, and meet the requirements of interception operations. Summary of the Invention

[0003] The purpose of the present invention is to solve the problem that the reentry trajectory of hypersonic reentry vehicles has extremely strong unknown maneuver characteristics, and to provide a maneuver detection method based on model probability matching. This method uses a maneuver model description matrix to distinguish the states of different known maneuver models, and realizes real-time maneuver detection of hypersonic reentry vehicles through a method based on model probability matching.

[0004] The purpose of the present invention is achieved through the following technical solutions:

[0005] A maneuver detection method based on model probability matching, the method comprising the following steps:

[0006] Step 1: Analyze different characteristic parameters when a hypersonic reentry vehicle has a maneuver mutation, construct a maneuver model description matrix according to the characteristic parameters, and determine the maneuver model;

[0007] Step 2: Define an influence factor according to the maneuver model description matrix and design a model matching degree function;

[0008] Step 3: Set thresholds for the characteristic parameters of different maneuver models, and calculate the matching vectors of the current system and each preset model based on the actual characteristic parameters of the current system;

[0009] Step 4: Substitute the prior influence factor and the calculated matching vector into the model matching degree function to calculate the model matching probability;

[0010] Step 5: Calculate the model probability correction sequence according to the probability that different preset maneuver models match the true state at the current moment, and finally correct the model probability vector;

[0011] Step 6: Determine whether the preset model switching condition is satisfied based on the model probability vector. If it is satisfied, perform model switching and reset the model probability vector according to the model selection result. If the model switching condition is not satisfied, do not perform model switching, and the maneuver detection result model is used as the prior information for subsequent trajectory prediction.

[0012] Further, the specific content of Step 1 is as follows: Analyze the changes of different characteristic parameters when a hypersonic reentry vehicle has a maneuver mutation according to the mission requirements. Assume five preset maneuver models, and select Δax, Δay, Δaz, A K , ΔA k , Δε, Δη as the model characteristic parameters to describe these five models. Among them, Δax is the acceleration difference on the x-axis, Δay is the acceleration difference on the y-axis, Δaz is the acceleration difference on the z-axis, A K is the acceleration direction, ΔA k is the acceleration direction difference, Δε is the velocity sign change, and Δη is the acceleration sign change. Thus, the maneuver model matrix GΒ is obtained;

[0013]

[0014] Each row in GΒ represents a possible model. Among them, GB1 corresponds to the non-maneuver model, and GB2, GB3, GB4, and GB5 correspond to four common maneuver models. In practical applications, the corresponding maneuver models can be designed according to the maneuver characteristics of the aircraft.

[0015] Further, in Step 2, in order to realize the detection of maneuvers, a model matching degree function is defined,

[0016] g i (t) = K axi Β i (Δax) + K ayi Β i (Δay) + K azi Β i (Δaz) + K Ai Β i (A K )

[0017] + K AΔi Β i (ΔA k ) + K εi Β i (Δε) + K ηi Β i (Δη), i = 1…5

[0018] Among them, K jiis the influence factor of each parameter in the five models, where i = 1...5 represents five different models, and j = 1...7 corresponds to Δax, Δay, Δaz, A K , ΔA k , Δε, Δη, these seven different parameters, where Δax is the acceleration difference in the x-axis, Δay is the acceleration difference in the y-axis, Δaz is the acceleration difference in the z-axis, A K is the acceleration direction, ΔA k is the acceleration direction difference, Δε is the velocity sign change, and Δη is the acceleration sign change; from the matching vector Β i (t) and the corresponding influence factor K ji calculate the matching degree function g i (t) of the current system for the five preset models.

[0019] Furthermore, B i (t) = [Β i (Δax), Β i (Δay), Β i (Δaz), Β i (A K ), Β i (ΔA k ), Β i (Δε), Β i (Δη)]| t 。

[0020] Furthermore, in step three, the specific method for obtaining the matching vector is as follows:

[0021] Step three - one: Set corresponding thresholds for each model feature parameter:

[0022] Set the threshold vector H = [H(Δax), H(Δay), H(Δaz), H(A K ), H(ΔA k ), H(Δε), H(Δη)];

[0023] Step three - two: Obtain the current system description vector GΒ(t):

[0024] Compare the estimated values of all model feature parameters with the corresponding items in the threshold vector. When it is less than the threshold, take 0; when it is greater than or equal to the threshold, take 1, that is, the corresponding current system description vector GΒ(t) is obtained. GΒ(t) is a seven - dimensional vector with all values being 0 and 1;

[0025] Step three - three: Obtain the matching vector Β i (t):

[0026] Compare GB(t) with each row of GB bit by bit. When they are the same, take it as 1, and when they are different, take it as 0, so as to obtain the matching vector Β of the current system and each preset model i (t).

[0027] Further, in step five, define the model probability vector N = [p1 p2 p3 p4 p5] T , this vector represents the probability that the five preset models at the current moment match the real model. That is, p1 represents the probability that the preset GB1 model at the current moment matches the real model, p2 represents the probability that the preset GB2 model at the current moment matches the real model, p3 represents the probability that the preset GB3 model at the current moment matches the real model, p4 represents the probability that the preset GB4 model at the current moment matches the real model, and p5 represents the probability that the preset GB5 model at the current moment matches the real model, satisfying ∑pi = 1, i = 1, 2... 5;

[0028] Define the probability correction sequence {dp1, dp2, dp3, dp4, dp5}, where dp1 represents the correction amount of p1, dp2 represents the correction amount of p2, dp3 represents the correction amount of p3, dp4 represents the correction amount of p4, and dp5 represents the correction amount of p5. The role of the probability correction sequence is to correct the model probability vector according to the estimation results of each step of simulation.

[0029] Further, in step six, the specific method of model switching is as follows:

[0030] Step six one: Given the initial value of the model probability vector N, set the relevant thresholds and model switching conditions;

[0031] Step six two: Judge the model characteristic parameters after each step of filtering and calculate the model matching function;

[0032] Step six three: Determine the matching degree order of the current system and the five preset models according to the size order of g i (t);

[0033] Step six four: Set the values of the probability correction sequence {dp1, dp2, dp3, dp4, dp5} according to the model matching order;

[0034] Step six five: Add the probability correction sequence to the model probability vector N for correction;

[0035] Step 66: For the filtering algorithm that requires a specific maneuver model, determine whether the preset model switching condition is satisfied. If it is satisfied, model switching can be performed, and the model probability vector N is reset according to the model selection result. If the model switching condition is not satisfied, model switching is not performed. For the IMM interactive multiple model filtering algorithm that requires the probability of each maneuver model, directly use the maneuver model probability as the maneuver detection output;

[0036] Step 67: Repeat the above process until the filtering ends.

[0037] The beneficial effects of the present invention compared with the prior art are as follows:

[0038] (1) By introducing a maneuver model description matrix, an influence factor, a model matching degree function, a model matching vector, and a model probability vector, the maneuver characteristics and the detection of maneuver mutations of a hypersonic reentry vehicle are mathematically modeled and quantified.

[0039] (2) By designing a model matching degree function, the actual state quantity of the target is converted into a model matching probability, and different maneuver models can be set according to the characteristics of different maneuver mutations, adapting to the characteristics of the hypersonic reentry vehicle with extremely strong unknown maneuver characteristics.

[0040] (3) Design a model switching condition to convert the maneuver detection result into a determined maneuver model or a determined maneuver model probability, meeting the requirements of subsequent trajectory prediction. Brief Description of the Drawings

[0041] Figure 1 It is a structural diagram of the maneuver detection method based on model probability matching of the present invention. Detailed Embodiment

[0042] The technical solutions of the present invention will be further described below in conjunction with the drawings and embodiments, but are not limited thereto. Any modification or equivalent replacement of the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention shall be covered by the protection scope of the present invention.

[0043] Embodiment 1:

[0044] The present invention provides a maneuver detection method based on model probability matching. By introducing a maneuver model description matrix, an influence factor, a model matching degree function, a model matching vector, and a model probability vector, the maneuver characteristics and the detection of maneuver mutations of a hypersonic reentry vehicle are mathematically modeled and quantified. By designing a model matching degree function, the actual state quantity of the target is converted into a model matching probability, and finally a model switching condition is designed to obtain a determined maneuver model or a determined maneuver model probability result. The specific steps are as follows:

[0045] Step 1: Analyze the changes of different characteristic parameters when the hypersonic reentry vehicle has a maneuver mutation according to the mission requirements. Assume that five maneuver models are preset, and Δax, Δay, Δaz, A K , ΔA k , Δε, Δη can be selected as the model characteristic parameters to describe these five models, where Δax is the acceleration difference in the x-axis, Δay is the acceleration difference in the y-axis, Δaz is the acceleration difference in the z-axis, and A K is the acceleration direction, ΔA k is the acceleration direction difference, Δε is the velocity sign change, and Δη is the acceleration sign change. Then the maneuver model matrix GΒ can be obtained.

[0046]

[0047] Each row in GΒ represents a possible model.

[0048] Step 2: Define the influence factor according to the maneuver model description matrix and design the model matching degree function. To achieve the detection of maneuvers, the following model matching degree function is defined.

[0049] g i (t) = K axi Β i (Δax) + K ayi Β i (Δay) + K azi Β i (Δaz) + K Ai Β i (A K ) + K AΔi Β i (ΔA k ) + K εi Β i (Δε) + K ηi Β i (Δη), i = 1…5

[0050] where K ji is the influence factor of each parameter in the five models, i = 1…5 represents five different models, and j = 1…7 corresponds to Δax, Δay, Δaz, A K , ΔA k , Δε, Δη these seven different parameters. From Β i (t) of each model at the current moment and the corresponding influence factor K ji , the value of the matching function g i (t) of the current system for the five preset models can be calculated. During the simulation process, K ji is given in advance according to the characteristics of the model, and Β i(t) is the matching vector between the current system and the i-th type of model at time t.

[0051] Step 3: Set thresholds for the characteristic parameters of different maneuver models, and calculate the matching vector between the current system and each preset model based on the actual characteristic parameters of the current system. Let B i (t) = [Β i (Δax), Β i (Δay), Β i (Δaz), Β i (A K ), Β i (ΔA k ), Β i (Δε), Β i (Δη)]| t .

[0052] Step 4: Substitute the prior influence factor and the calculated matching vector into the model matching degree function to calculate the model matching probability.

[0053] Step 5: Calculate the model probability correction sequence according to the probabilities of different preset maneuver models matching the true state at the current moment, and finally correct the model probability vector.

[0054] Define the model probability vector N = [p1 p2 p3 p4 p5] T , and this vector represents the probabilities of the five preset models matching the true model at the current moment.

[0055] Define the probability correction sequence {dp1, dp2, dp3, dp4, dp5}. The role of the probability correction sequence is to correct the model probability vector according to the estimation results of each simulation step. The magnitude of the values of {dp1, dp2, dp3, dp4, dp5} determines the speed of model judgment. When the values of {dp1, dp2, dp3, dp4, dp5} are too large, it will cause "jitter" in the system.

[0056] Step 6: Determine whether the preset model switching condition is satisfied based on the model probability vector. If it is satisfied, model switching can be performed and the model probability vector can be reset according to the model selection result. If the model switching condition is not satisfied, no model switching is performed, and the maneuver detection result model is used as the prior information for subsequent trajectory prediction.

[0057] Furthermore, the specific method for obtaining the matching vector in Step 3 is as follows:

[0058] Step 3.1: Set corresponding thresholds for each model characteristic parameter:

[0059] Let the threshold vector H = [H(Δax), H(Δay), H(Δaz), H(A K ), H(ΔAk ), H(Δε), H(Δη)]]。

[0060] Step 3.2: Obtain the current system description vector GΒ(t):

[0061] Compare the estimated values of all model feature parameters with the corresponding items in the threshold vector. When it is less than the threshold, take 0, and when it is greater than or equal to the threshold, take 1. Then the corresponding current system description vector GΒ(t) can be obtained. GΒ(t) is a seven-dimensional vector with all values being 0 and 1.

[0062] Step 3.3: Obtain the matching vector Β i (t):

[0063] Compare GΒ(t) with each row of GΒ bit by bit. When they are the same, take 1, and when they are different, take 0. Then the matching vector Β i (t) of the current system and each preset model can be obtained. For example, when GΒ(t) = [1 0 1 0 1 0 0], the result of comparing with the first row of the matrix GΒ is Β1(t) = [0 1 0 0 0 1 1].[[]END]

[0064] Further, the specific method of model switching in Step 6 is as follows:

[0065] Step 6.1: Given the initial value of the model probability vector N, set the relevant thresholds and model switching conditions.

[0066] Step 6.2: Judge the model feature parameters after each filtering step and calculate the model matching function.

[0067] Step 6.3: Determine the matching degree order of the current system and the five preset models according to the magnitude order of g i (t).

[0068] Step 6.4: Set the values of the probability correction sequence {dp1, dp2, dp3, dp4, dp5} according to the model matching order.

[0069] Step 6.5: Add the probability correction sequence to the model probability vector N to correct it.

[0070] Step 6.6: For the filtering algorithm that requires a specific maneuver model, judge whether the preset model switching conditions are satisfied. If satisfied, the model can be switched, and at the same time, reset the model probability vector N according to the model selection result. If the model switching conditions are not satisfied, the model is not switched; for the IMM interactive multiple model filtering algorithm that requires the probabilities of each maneuver model, directly take the maneuver model probability as the maneuver detection output.

[0071] Step 6.7: Repeat the above process until the filtering ends.

Claims

1. A maneuver detection method based on model probability matching, characterized in that: The method includes the following steps: Step 1: Analyze different characteristic parameters when a hypersonic reentry vehicle experiences a maneuver mutation, construct a maneuver model description matrix based on the characteristic parameters, and determine the maneuver model; specifically, in Step 1: Analyze the changes in different characteristic parameters when a hypersonic reentry vehicle experiences a maneuver mutation according to the mission requirements, preset five maneuver models, and select Δax, Δay, Δaz, A K , ΔA k , Δε, Δη as the seven quantities of model characteristic parameters to describe these five models. Among them, Δax is the acceleration difference in the x-axis, Δay is the acceleration difference in the y-axis, Δaz is the acceleration difference in the z-axis, A K is the acceleration direction, ΔA k is the acceleration direction difference, Δε is the change in the velocity sign, Δη is the change in the acceleration sign, and thus the maneuver model matrix GΒ is obtained; Each row in GB represents a type of model, where GB1 corresponds to the non-maneuvering model, and GB2, GB3, GB4, and GB5 correspond to four maneuvering models; Step 2: Define influence factors according to the maneuvering model description matrix and design a model matching degree function; Step 3: Set thresholds for the characteristic parameters of different maneuvering models, and calculate the matching vectors of the current system with each preset model based on the actual characteristic parameters of the current system; Step 4: Substitute the prior influence factors and the calculated matching vectors into the model matching degree function to deduce the model matching probability; Step 5: Calculate the model probability correction sequence according to the probabilities of different preset maneuvering models matching the true state at the current moment, and finally correct the model probability vector; Step 6: Determine whether the preset model switching condition is satisfied based on the model probability vector. If it is satisfied, perform model switching and reset the model probability vector according to the model selection result. If the model switching condition is not satisfied, do not perform model switching, and use the maneuver detection result model as the prior information for subsequent trajectory prediction.

2. The maneuver detection method based on model probability matching according to claim 1, characterized in that: In Step 2, in order to achieve maneuver detection, a model matching degree function is defined. g i g(t) = K axi Β i Β(Δax) + K ayi Β i Β(Δay) + K azi Β i Β(Δaz) + K Ai Β i (A K ) + K AΔi Β i (ΔA k ) + K εi Β i Β(Δε) + K ηi Β i Β(Δη), i = 1…5 Among them, K ji is the influence factor of each parameter in the five types of models. i = 1…5 represents five different models, and j = 1…7 corresponds to Δax, Δay, Δaz, A K , ΔA k , Δε, Δη, these seven different parameters. Among them, Δax is the acceleration difference in the x-axis, Δay is the acceleration difference in the y-axis, Δaz is the acceleration difference in the z-axis, A K is the acceleration direction, ΔA k is the acceleration direction difference, Δε is the velocity sign change, and Δη is the acceleration sign change; from the matching vector Β i (t) of each model at the current moment and the corresponding influence factor K ji calculate the value of the matching degree function g i (t) of the current system for the five types of preset models.

3. A maneuver detection method based on model probability matching according to claim 2, characterized in that: B i (t) = [Β i (Δax), Β i (Δay), Β i (Δaz), Β i (A K ), Β i (ΔA k ), Β i (Δε), Β i (Δη)]| t 。 4. A maneuver detection method based on model probability matching according to claim 1 or 2, characterized in that: In Step 3, the specific method for obtaining the matching vector is as follows: Step 3-1: Set corresponding thresholds for each model characteristic parameter: Set the threshold vector H = [H(Δax), H(Δay), H(Δaz), H(A K ), H(ΔA k ), H(Δε), H(Δη)]; Step 3-2: Obtain the current system description vector GB(t): Compare the estimated values of all model characteristic parameters with the corresponding items in the threshold vector. When it is less than the threshold, take 0; when it is greater than or equal to the threshold, take 1, so as to obtain the corresponding current system description vector GB(t). GB(t) is a seven-dimensional vector with all values being 0 and 1. Step Three: Obtain the matching vector Β i (t): Compare GB(t) with each row of GB bit by bit. When they are the same, take it as 1, and when they are different, take it as 0, that is, obtain the matching vector B of the current system and each preset model i (t).

5. A maneuver detection method based on model probability matching according to claim 1 or 3, characterized in that: In step five, define the model probability vector N = [p1 p2 p3 p4 p5] T , this vector represents the probabilities that the five types of preset models at the current moment match the true model, that is, p1 represents the probability that the preset GB1 model at the current moment matches the true model, p2 represents the probability that the preset GB2 model at the current moment matches the true model, p3 represents the probability that the preset GB3 model at the current moment matches the true model, p4 represents the probability that the preset GB4 model at the current moment matches the true model, p5 represents the probability that the preset GB5 model at the current moment matches the true model, and it satisfies Σpi = 1, i = 1, 2... 5; Define the probability correction sequence {dp1, dp2, dp3, dp4, dp5}, where dp1 represents the correction amount of p1, dp2 represents the correction amount of p2, dp3 represents the correction amount of p3, dp4 represents the correction amount of p4, and dp5 represents the correction amount of p5. The role of the probability correction sequence is to correct the model probability vector according to the estimation results of each step of simulation.

6. The maneuver detection method based on model probability matching according to claim 1, wherein: In Step 6, the specific method of model switching is as follows: Step 6-1: Given the initial value of the model probability vector N, set relevant thresholds and model switching conditions; Step 6-2: Judge the model characteristic parameters after each step of filtering and calculate the model matching function; Step Six Three: Determine the matching degree order between the current system and the preset five types of models according to the magnitude order of g i (t); Step 6-4: Set the values of the probability correction sequence {dp1, dp2, dp3, dp4, dp5} according to the model matching order; Step 6-5: Add the probability correction sequence to the model probability vector N for correction; Step 6-6: For the filtering algorithm that requires a specific maneuvering model, judge whether the preset model switching condition is satisfied. If it is satisfied, model switching can be performed, and at the same time, reset the model probability vector N according to the model selection result. If the model switching condition is not satisfied, do not perform model switching; for the IMM interactive multiple model filtering algorithm that requires the probabilities of each maneuvering model, directly use the maneuvering model probabilities as the maneuver detection output; Step 6-7: Repeat the above process until the filtering ends.

Images