Neural network based method for solving launch parameters of boost-glide vehicle

Abstract

The method for solving launch parameters of a boost-glide vehicle based on a neural network belongs to the field of solving launch parameters of a hypersonic vehicle. The method solves the problem of long solving time of existing methods for solving launch parameters of a vehicle. A sample is constructed, the input data of each sample includes the longitude and latitude of a launch point, a target point and the center longitude and latitude of each no-fly zone, and the output data includes a first maximum negative attack angle, a second maximum negative attack angle, a third constant pitch angle change rate, a launch azimuth angle, a pull-up segment attack angle change time, a pull-up segment end time, a resistance acceleration and a dive attack angle; the sample is used to train a neural network model for solving parameters, and the trained model is used to predict the launch parameters of a planned trajectory of a boost-glide vehicle according to the longitude and latitude of a current launch point, a target point and the center longitude and latitude of each no-fly zone. The application is mainly applied to fast solving of launch parameters under multiple working conditions.

Description

Technical Field

[0001] This invention belongs to the field of hypersonic vehicle launch parameter calculation. Taking boost-glide hypersonic vehicles as the research object, it proposes a fast method for calculating launch parameters under multiple operating conditions. Background Technology

[0002] The rapid response capability of hypersonic vehicles is a crucial indicator of their combat effectiveness. Reducing launch time is key to improving their combat effectiveness and survivability, especially in actual combat where a series of vehicles may be launched to better strike targets, necessitating consideration of planning efficiency. Existing parameter calculation methods employ parametric iteration to sequentially calculate parameters for different flight segments along a trajectory, resulting in lengthy computation times. Therefore, achieving rapid parameter calculation for multiple vehicle launches is a pressing issue that needs to be addressed. Summary of the Invention

[0003] The purpose of this invention is to solve the problem of long calculation time in existing methods for calculating launch parameters of aircraft. This invention provides a method for calculating launch parameters of boost-glide vehicles based on neural networks.

[0004] A method for calculating launch parameters of a boost-glide vehicle based on neural networks, comprising the following steps:

[0005] S1. Obtain the sample set:

[0006] By individually changing the latitude and longitude of the launch point, the target point, and the center latitude and longitude of each no-fly zone, M samples are obtained based on the full trajectory data of the boost-glide vehicle. The full trajectory of the boost-glide vehicle includes the boost phase, the pull-up phase, the glide phase, and the dive-strike phase.

[0007] Simultaneously, the latitude and longitude of the launch point, the target point, and the center latitude and longitude of each no-fly zone are changed, and N samples are obtained based on the full trajectory data of the boost-glide vehicle;

[0008] The input data for each sample includes the latitude and longitude of the launch point and target point, as well as the center latitude and longitude of each no-fly zone. The output data includes the maximum negative angle of attack for the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , pull up the angle of attack to change the timing , End of pull-up segment The drag acceleration represented by the ordinate of the midpoint of the drag acceleration profile. and the angle of attack during the dive strike phase ;

[0009] S2. Select P samples from M+N samples as the training set to train the parameter calculation neural network model and obtain the trained parameter calculation neural network model; wherein, the input data and output data of the samples are used as the input and output data of the parameter calculation neural network model, respectively.

[0010] S3. Input the latitude and longitude of the current launch point, target point, and the center latitude and longitude of each no-fly zone into the trained parameter calculation neural network model. The trained parameter calculation neural network model predicts the launch parameters of the planned trajectory of the boost-glide vehicle, including the maximum negative angle of attack of the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , pull up the angle of attack to change the timing , End of pull-up segment Drag acceleration and the angle of attack during the dive strike phase .

[0011] Preferably, the boost stage employs a three-stage boost system, consisting of a first-stage boost, a second-stage boost, and a third-stage boost, and the boost stage achieves the maximum negative angle of attack for the first stage in each sample. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate and launch azimuth The acquisition process includes:

[0012] Select the end height of the second-stage boost Boost phase end height , heading angle at the end of the boost phase As parameters, the residual function and least squares index function of the boost stage are designed, and the boost stage dynamic equations are combined with iterative solutions to obtain the maximum negative angle of attack of the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate and launch azimuth ;

[0013] During the iteration process, , and The initial value is a given value. The initial value is obtained using the spherical trigonometric formula, which is:

[0014] ;

[0015] in, For the central angle, and These are the longitude and latitude of the launch point, respectively. and These are the longitude and latitude of the target point, respectively.

[0016] Preferably, the residual function of the boost segment for:

[0017] ;

[0018] Least square index function of the boost phase for:

[0019] ;

[0020] in, , , They are respectively , and Expected value It is a 2-norm. These are the variables to be planned.

[0021] Preferably, the timing of the angle of attack change during the pull-up segment in each sample is achieved through the pull-up segment. , End of pull-up segment The acquisition process includes:

[0022] Select the height at the end of the lifting section Track angle at the end of the pull-up phase As parameters, the residual function and least squares index function of the lifting section are designed, and the dynamic equations of the unpowered section are combined for iterative solution to obtain the moment when the angle of attack of the lifting section changes. , End of pull-up segment ;

[0023] During the iteration process, and The initial value is a given value, and the two satisfy the angle of attack profile. ,and

[0024] , For time, Preset angle of attack.

[0025] Preferably, the residual function of the pull-up segment is... for:

[0026] ;

[0027] Least square index function of the pull-up segment for:

[0028] ;

[0029] in, , They are respectively and The expected value of the parameters, It is a 2-norm. These are the variables to be planned.

[0030] Preferably, drag acceleration in each sample is achieved through the gliding phase. The acquisition process includes:

[0031] Through integral relationships and drag acceleration profiles Determine the flight range of the boost-glide vehicle ;

[0032] ;

[0033] in, For the energy of the aircraft, These are the first to third undetermined coefficients, respectively.

[0034] And due to the waiting flight distance The latitude and longitude of the boost-glide vehicle's current position and the target point's position are represented as follows:

[0035] ;

[0036] in, and These are the longitude and latitude of the boost-glide vehicle's current location, respectively. and These are the longitude and latitude of the target point, respectively. This is the distance from the launch point to the Earth's center.

[0037] Given drag acceleration profile Initial values ​​of the x and y coordinates of the upper midpoint, and combined with the above two flight paths. The expression is solved iteratively to obtain the drag acceleration. .

[0038] Preferably, the angle of attack during the dive-strike phase in each sample is achieved through the dive-strike phase. The acquisition process includes:

[0039] Select target point location As parameters, the residual function and least squares index function of the dive-strike phase are designed, and the angle of attack of the dive-strike phase is obtained by iterative solution in combination with the dynamic equation of the unpowered phase. ;

[0040] During the iteration process, the angle of attack during the dive strike phase and tilt angle The initial value is a given value.

[0041] Preferably, the residual function of the dive-strike phase is... for:

[0042] ;

[0043] Least square index function of the dive strike phase for:

[0044] ;

[0045] in, for Expected value It is a 2-norm. For variables to be planned, To solve for the magnitude of the vector.

[0046] Preferably, the detour strategy for the no-fly zone is to first determine the minimum cost detour route, and then determine each waiting waypoint based on the minimum cost detour route, with a one-to-one correspondence between the waiting waypoints and the no-fly zone.

[0047] Preferably, the specific process for determining the minimum-cost detour path includes:

[0048] A1. Define the no-fly zone as a circular area with a fixed radius;

[0049] A2. In the latitude and longitude plane, draw a circle with a constant radius centered on the current flight position of the boost-glide vehicle. Calculate the cost value of each point on the circle using a cost function. Obtain, and

[0050] ;

[0051] in, The first circle The value of each point This represents the total number of no-fly zones. For the first circle The geocentric angle between the point and the target point For the first The threat level of a no-fly zone Greater than or equal to 1 For the first The radius of a no-fly zone For the first circle The point and the first The geocentric angle between the no-fly zone dots;

[0052] A3. Take the point with the minimum replacement value on the circle as the minimum cost point;

[0053] A4. The line connecting the minimum cost points corresponding to each flight position of the boost-glide vehicle from the launch point to the target point constitutes the minimum cost bypass path.

[0054] The methods for determining each waypoint based on the minimum cost detour route include:

[0055] B1. Make a passage through the first... The equation of a straight line perpendicular to the line connecting the launch point and the target point, drawn from the center of a no-fly zone, on a latitude-longitude plane is:

[0056] ;

[0057] in, and These are the x and y coordinates of the equation of the line, respectively. and The first The latitude and longitude of the center of the no-fly zone, The slope of the line connecting the launch point and the target point;

[0058] B2. Select the point on the least-cost detour route that has the shortest straight-line distance to the destination, and designate this point as the first... The waiting waypoints corresponding to each no-fly zone.

[0059] Advantages of this invention:

[0060] This invention provides a method for calculating launch parameters of a boost-glide vehicle based on a neural network. The method first acquires samples, with each sample's input data including the latitude and longitude of the launch point and target point, and the center latitude and longitude of each no-fly zone. The output data includes the maximum negative angle of attack at stage one. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , pull up the angle of attack to change the timing , End of pull-up segment The drag acceleration represented by the ordinate of the midpoint of the drag acceleration profile. and the angle of attack during the dive strike phase This invention utilizes randomly selected samples to train a neural network model for parameter calculation, and then uses the trained model for calculation. This method is fast and efficient, avoiding the shortcomings of existing methods that can only provide one trajectory at a time and require iterative parameter calculation for each change of launch scenario, resulting in long computation times and low efficiency. Compared to traditional iterative methods, the proposed neural network-based fast parameter calculation and planning method has a significant advantage in efficiency. It can meet the need for rapid trajectory planning and avoidance of target points in no-fly zones.

[0061] The feasibility of the proposed method has been verified by demonstrating that the aircraft can successfully fly around no-fly zones and strike targets in all scenarios of this invention, while satisfying all constraints. Attached Figure Description

[0062] Figure 1 This is a flowchart of a method for solving launch parameters of a boost-glide vehicle based on neural networks;

[0063] Figure 2 and Figure 3 All of these are schematic diagrams showing the minimum cost paths and waypoints obtained by making different changes to the parameters in Table 3.

[0064] Figure 4 This is a schematic diagram of the internal structure of the neural network model for parameter calculation;

[0065] Figure 5 This is a schematic diagram illustrating the principle of the neural network model for solving parameters.

[0066] Figure 6 It is the heat flux curve during the gliding phase;

[0067] Figure 7 It is the dynamic pressure curve during the gliding phase;

[0068] Figure 8 It is the overload curve during the gliding phase;

[0069] Figure 9 It is a full-length elevation curve;

[0070] Figure 10 It is the velocity curve for the entire segment;

[0071] Figure 11 It is a trajectory in latitude and longitude coordinates;

[0072] Figure 12 This is the maximum negative angle of attack error curve for level one;

[0073] Figure 13 This is the maximum negative angle of attack error curve for the second stage;

[0074] Figure 14 It is a pitch error curve diagram;

[0075] Figure 15 This is a graph showing the error curve of the rate of change of pitch angle at level three constant values;

[0076] Figure 16 It is a graph showing the error curve when the angle of attack changes during the pull-up phase;

[0077] Figure 17 This is the error curve at the end of the pull-up segment;

[0078] Figure 18 This is a curve showing the drag acceleration error.

[0079] Figure 19 This is a curve showing the angle of attack error during the dive attack phase;

[0080] Figure 20 It is a multi-scenario trajectory planning curve for parameter calculation; Figure 20 From bottom to top, the no-fly zones are designated as 1 to 3.

[0081] Figure 21 This is the flight path curve for No-Fly Zone 1;

[0082] Figure 22 This is the flight path curve for No-Fly Zone 3; Detailed Implementation

[0083] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0084] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0085] Detailed Implementation Method 1, see [link / reference] Figure 1 The method for calculating launch parameters of a boost-glide vehicle based on a neural network, as described in this embodiment, includes the following steps:

[0086] S1. Obtain the sample set:

[0087] By individually changing the latitude and longitude of the launch point, the target point, and the center latitude and longitude of each no-fly zone, M samples are obtained based on the full trajectory data of the boost-glide vehicle. The full trajectory of the boost-glide vehicle includes the boost phase, the pull-up phase, the glide phase, and the dive-strike phase.

[0088] Simultaneously, the latitude and longitude of the launch point, the target point, and the center latitude and longitude of each no-fly zone are changed, and N samples are obtained based on the full trajectory data of the boost-glide vehicle;

[0089] The input data for each sample includes the latitude and longitude of the launch point and target point, as well as the center latitude and longitude of each no-fly zone. The output data includes the maximum negative angle of attack for the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , pull up the angle of attack to change the timing , End of pull-up segment The drag acceleration represented by the ordinate of the midpoint of the drag acceleration profile. and the angle of attack during the dive strike phase ;

[0090] S2. Select P samples from M+N samples as the training set to train the parameter calculation neural network model and obtain the trained parameter calculation neural network model; wherein, the input data and output data of the samples are used as the input and output data of the parameter calculation neural network model, respectively.

[0091] S3. Input the latitude and longitude of the current launch point, target point, and the center latitude and longitude of each no-fly zone into the trained parameter calculation neural network model. The trained parameter calculation neural network model predicts the launch parameters of the planned trajectory of the boost-glide vehicle, including the maximum negative angle of attack of the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , pull up the angle of attack to change the timing , End of pull-up segment Drag acceleration and the angle of attack during the dive strike phase .

[0092] Boost-glide vehicle trajectory planning: The vehicle trajectory is divided into four parts, each part is planned separately, and then combined to reduce planning complexity. The boost, pull-up, gliding, and dive-strike phases are planned sequentially. The planning approach involves selecting nine core variables and iteratively solving to achieve rapid trajectory planning for the entire vehicle.

[0093] The core variables to be planned are selected as follows: for the boost phase, the maximum negative angle of attack for the first and second stages, the constant rate of change of pitch angle for the third stage, and the launch azimuth angle are selected; for the pull-up phase, the moment of angle of attack change and the moment of end are selected; for the gliding phase, the drag acceleration value at the average value is selected; for the dive attack phase, the constant angle of attack and constant roll angle are selected.

[0094] The preparatory work before specific application, this part is based on existing technology:

[0095] Step 1: Coordinate system definition and dynamic modeling, which can be achieved using existing technologies;

[0096] Step 1-1: Coordinate System Definition and Transformation

[0097] Step 1-1-1: Coordinate System Definition

[0098] Geocentric fixed coordinate system:

[0099] origin For the Earth's core, The axis is the direction of the prime meridian. The axis points in the direction of the Earth's rotation axis. shaft and , The axis forms a right-handed system.

[0100] Ground launch coordinate system:

[0101] origin As the launch point, The axis points to the target point. Axis perpendicular to Axially upward. shaft and , The axis forms a right-handed system.

[0102] Half-velocity coordinate system:

[0103] origin Center of mass, The axis points in the direction of velocity. The axis is located in the ballistic plane and perpendicular to Axially upward, shaft and , The axis forms a right-handed system.

[0104] Projectile coordinate system:

[0105] origin Center of mass, The axis is along the longitudinal direction. The axis is perpendicular in the longitudinal plane of symmetry Axially upward, shaft and , The axis forms a right-handed system.

[0106] Velocity coordinate system:

[0107] origin Center of mass, The axis represents the direction of velocity. The axis is perpendicular in the longitudinal plane of symmetry Axially upward, shaft and , The axis forms a right-handed system.

[0108] Northeast coordinate system:

[0109] origin For the Earth's core, The axis points from the Earth's center to the spacecraft. The axis lies in the equatorial plane and is parallel to the latitude where the spacecraft is located. shaft and , The axes form a right-handed coordinate system. Step 1-1-2: Coordinate system transformation

[0110] The definitions of the angles used in this section are shown in Table 1 below.

[0111] Table 1. Definition of Angle Symbols

[0112]

[0113] 1) From the geocentric fixed coordinate system To the launch system Transformation matrix The relational expression is:

[0114] ;

[0115] in, , and They are respectively around Transformation matrix for axis rotation;

[0116] 2). From the launching system missile system The transformation matrix relation expression is as follows: ;

[0117] 3). From the launching system To half-velocity system The transformation matrix relation expression is as follows:

[0118] ;

[0119] 4). From the half-velocity coordinate system to velocity coordinate system The transformation matrix relationship expression using the tilt angle is as follows:

[0120] ;

[0121] 5). From the velocity coordinate system missile system Considering the angle of attack and sideslip angle, the transformation matrix adopts a transposed order. The expression for the transformation matrix is:

[0122] ;

[0123] 6) From the geocentric fixed coordinate system To the northeast coordinate system The transformation matrix is ​​in transposition, and its expression is:

[0124] ;

[0125] 7). From the half-velocity system To the northeast coordinate system The transformation matrix expression:

[0126] ;

[0127] We can directly transform between the two coordinate systems, or we can perform a simultaneous transformation, resulting in the following equations:

[0128] ;

[0129] This equation contains eight angles. Given five of these angles, the remaining three can be derived using this equation. Taking the solution for the attitude angles as an example, we have:

[0130] ;

[0131] The three attitude angles can then be calculated:

[0132] ;

[0133] Steps 1-2: Aircraft Dynamics Model

[0134] The following assumptions are made: the Earth is an ellipsoid, and considering the Earth's rotation and the accuracy of the gravitational model to [value missing]. The perturbation phase; and the unpowered phase adopts the BTT tilt turn control mode with a sideslip angle of zero.

[0135] Step 1-2-1: Dynamic equations of the boost phase

[0136] The dynamics equations of the powered rocket booster stage are established in the launch frame. According to Newton's second law, the equations of motion for the center of mass have the following form:

[0137] ;

[0138] The state equation in vector form for the booster segment can then be written as:

[0139] ;

[0140] The definitions of each quantity are as follows:

[0141] : The position vector of the aircraft in the launch system.

[0142] : Velocity vector.

[0143] Engine thrust vector.

[0144] Aerodynamic vector.

[0145] Earth's gravitational acceleration vector

[0146] : Earth's rotational angular velocity vector.

[0147] :Bi Chong.

[0148] Earth's gravitational acceleration You can follow direction and Directional decomposition, at this time It can be represented as:

[0149] ;

[0150] Only considering Perturbation, then:

[0151] ;

[0152] Earth's rotational angular velocity vector and geocentric vector In the emission system, it can be further decomposed into the following form:

[0153] ;

[0154] ;

[0155] The items include:

[0156]

[0157] represent Projection in the emission frame:

[0158]

[0159] in Defined as the geographical latitude of the launch point With geocentric latitude The difference between them It can be obtained from the following formula:

[0160] ;

[0161] Therefore, gravity in the launching frame can ultimately be expressed in the following form:

[0162] ;

[0163] The relevant equations that are also needed are as follows:

[0164] Geocentric latitude:

[0165] ;

[0166] Geographic latitude:

[0167] ;

[0168] The distance between the Earth's center and the launch system of the spacecraft It can be represented as:

[0169] ;

[0170] Geocentric distance at a point on the surface of the ellipsoid for:

[0171] ;

[0172] The distance from the spacecraft's current position to the Earth's surface can be expressed as:

[0173] ;

[0174] speed:

[0175] ;

[0176] The expressions for the trajectory inclination angle and trajectory deviation angle are:

[0177] ;

[0178] The values ​​of various Earth parameters involved in the equation are shown in Table 2:

[0179] Table 2 Common Earth Parameters

[0180]

[0181] The gravitational constant of Earth, For the Earth's semi-major axis, As the Earth's semi-minor axis, This is the Earth's rotational angular velocity. For perturbation.

[0182] Step 1-2-2: Dynamic equations of the unpowered segment

[0183] By constructing a dynamic model of the unpowered gliding segment in a half-velocity frame, the longitudinal and lateral motion characteristics of the aircraft can be described more accurately, and the functional expressions are also relatively simple. The specific form of the unpowered segment equations is as follows:

[0184] State variables are This includes speed, track angle, heading angle, geocentric distance, longitude, and latitude; the control variables are... ,

[0185] , These represent the aerodynamic lift and drag experienced by the glider, respectively.

[0186] Steps 1-2-3: Inter-segment state transformation

[0187] The state variables differ between the boost and glide phases. At the end of the boost phase, the state variables under the launch system must be transformed to serve as the initial values ​​for the glide phase. That is, from... Turn to .

[0188] track angle This refers to the angle between the direction of the aircraft's velocity and the local horizontal plane, which can be determined using the geocentric radius vector and the aircraft's velocity.

[0189] ;

[0190] For heading angle The velocity vector is obtained by projecting it onto the northeast coordinate system:

[0191] ;

[0192] longitude Using the geocentric radius vector under fixed conditions:

[0193] .

[0194] The entire trajectory of a boost-glide vehicle includes the boost phase, pull-up phase, glide phase, and dive-strike phase:

[0195] (I) Boosting Phase Planning:

[0196] The boost phase can use a three-stage booster rocket, which is designed in six parts. The first-stage booster consists of three parts: vertical takeoff, first-stage negative angle of attack turn, and gravity turn; the second-stage booster includes second-stage negative angle of attack turn and fixed-axis flight.

[0197] The pitch angle during the first-stage boost process can be designed as follows:

[0198] ;

[0199] In the formula, , , These represent the moments at the end of vertical takeoff, first-stage negative angle-of-attack turn, and gravity turn, respectively. It can be represented as:

[0200] ;

[0201] For gravity, For thrust;

[0202] To determine the angle of attack profile, considering that in the negative angle of attack turn segment, when the aircraft velocity approaches the speed of sound, it is usually necessary to ensure that the angle of attack is zero. Therefore, taking the aircraft velocity as the independent variable, the angle of attack function for the first-order negative angle of attack turn segment can be designed as a trigonometric function in the following form:

[0203] ;

[0204] In the formula, The speed at the end of vertical takeoff. The speed at the end of a turn with a negative angle of attack. This is the maximum negative angle of attack for a first-order vehicle. For the speed of the aircraft, Given a constant value, This is the angle of attack function for a first-order negative angle of attack turning segment.

[0205] The second-stage boost process includes a second-stage negative angle-of-attack turn and fixed-axis flight, with the pitch angle designed as follows:

[0206] ;

[0207] In the formula, and These represent the end times of the two stages of the second-stage boost. The moment of the end of a second-order negative angle of attack turn. The end time of fixed-axis flight. For the trajectory inclination angle, The pitch angle at the end of the second-order negative angle of attack turn segment.

[0208] Angle of attack function for a second-order negative angle of attack turning segment The pattern of change is designed in sinusoidal form, specifically as follows:

[0209] ;

[0210] in, This is the second-highest negative angle of attack. It is a constant. It is a function in sinusoidal form.

[0211] The third-stage boost is for the constant pitch rate flight segment, and the pitch angle design for this segment is as follows:

[0212] ;

[0213] In the formula, For a moment and Rate of change of pitch angle between At the end of the boost phase, This marks the start time of the third-stage boost phase.

[0214] Four variables to be planned in the boost phase Choose from the following:

[0215] ;

[0216] Achieve the above-mentioned maximum negative angle of attack. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate and launch azimuth The specific process is as follows:

[0217] Select the end height of the second-stage boost Boost phase end height , heading angle at the end of the boost phase As parameters, the residual function and least squares index function of the boost stage are designed, and the boost stage dynamic equations are combined with iterative solutions to obtain the maximum negative angle of attack of the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate and launch azimuth The iteration stops when the index is less than a certain value or no longer changes significantly, and the solution is considered to meet the accuracy requirements.

[0218] During the iteration process, , and The initial value is a given value. The initial value is obtained using the spherical trigonometric formula, which is:

[0219] ;

[0220] in, For the central angle, and These are the longitude and latitude of the launch point, respectively. and These are the longitude and latitude of the target point, respectively.

[0221] residual function of the boost phase for:

[0222] ;

[0223] Least square index function of the boost phase for:

[0224] ;

[0225] in, , , They are respectively , and Expected value It is a 2-norm. These are the variables to be planned.

[0226] (II) Planning of the lifting section:

[0227] This is a transition phase. When the boost phase ends, the aircraft's engines shut down. To ensure the aircraft meets gliding standards, its behavior needs to be adjusted during this phase. Due to the complex atmospheric environment and limited aerodynamic forces and adjustment capabilities during this phase, the control variables must be designed to be as simple as possible. The initial velocity during the pull-up phase is high, requiring a larger angle of attack to ensure safe reentry. As the velocity decreases, the influence of heat flux diminishes, and overload and dynamic pressure become the primary influencing factors, necessitating adjustments to the angle of attack to meet corresponding constraints. Therefore, the angle of attack profile for the pull-up phase is designed.

[0228] The variable to be planned is the moment when the angle of attack changes during the pull-up phase. and the end of the pull-up phase :

[0229] ;

[0230] Select the height at the end of the lifting section Track angle at the end of the pull-up phase As parameters, the residual function and least squares index function of the lifting section are designed, and the dynamic equations of the unpowered section are combined for iterative solution to obtain the moment when the angle of attack of the lifting section changes. , End of pull-up segment ;

[0231] During the iteration process, and The initial value is a given value, and the two satisfy the angle of attack profile. ,and

[0232] , For time, Preset angle of attack.

[0233] Residual function of the pull-up segment for:

[0234] ;

[0235] Least square index function of the pull-up segment for:

[0236] ;

[0237] in, , They are respectively and The expected value of the parameters, It is a 2-norm. These are the variables to be planned.

[0238] The solution can be obtained using an iterative algorithm similar to that used for the boost phase. The iteration stops when the index is less than a certain value or no longer changes significantly, and the solution is considered to meet the accuracy requirements.

[0239] (III) Gliding Segment Planning:

[0240] By adjusting aerodynamic forces, gliders can achieve long-duration, extended-range gliding. Assuming the aircraft employs a BTT (Bill-Tail Turn) mode, meaning no sideslip throughout the process, the control variables for this segment are angle of attack and roll angle. The planning approach for this segment is also relatively mature: design a drag acceleration profile that satisfies all process constraints, and control the glider's longitudinal motion by tracking this profile. For lateral errors, a heading error corridor is designed to eliminate them.

[0241] The specific process for solving the drag acceleration profile can be found using existing technology, as follows:

[0242] 1. Design the drag acceleration profile as It is in the form of a quadratic function. As a variable;

[0243] 2. Based on the initial and desired final velocity and altitude information of the boost-glide vehicle. , Solve for the initial and final states (energy, drag acceleration) of the drag acceleration profile. , Meanwhile, assume the midpoint If it is also located on this cross-section, then the cross-section can be determined by three points, among which... It is an unknown quantity.

[0244] The initial velocity of the aircraft. The initial altitude of the aircraft. The velocity of the aircraft at its desired final state. The altitude of the aircraft's desired final state. The initial state energy of the drag acceleration profile is... The initial state of the drag acceleration profile is the drag acceleration. The final state energy of the drag acceleration profile. The drag acceleration at the final state of the drag acceleration profile. The drag acceleration is represented by the ordinate of the midpoint of the drag acceleration profile.

[0245] Through integral relationships and drag acceleration profiles Determine the flight range of the boost-glide vehicle ;

[0246] ;

[0247] in, For the energy of the aircraft, These are the first to third undetermined coefficients, respectively.

[0248] And due to the waiting flight distance The latitude and longitude of the boost-glide vehicle's current position and the target point's position are represented as follows:

[0249] ;

[0250] in, and These are the longitude and latitude of the boost-glide vehicle's current location, respectively. and These are the longitude and latitude of the target point, respectively. This is the distance from the launch point to the Earth's center.

[0251] Given drag acceleration profile Initial values ​​of the x and y coordinates of the upper midpoint, and combined with the above two flight paths. The expression is solved iteratively to obtain the drag acceleration. .

[0252] Among them, drag acceleration profile The x-coordinate of the upper midpoint is the coordinate of the average x-coordinate of the drag acceleration profile, and the x-coordinate of the midpoint represents energy, while the y-coordinate of the midpoint represents drag acceleration. .

[0253] (iv) Planning of the dive strike phase:

[0254] The dive strike process is simplified, considering only the fixed point of impact on the ground. The variables to be planned are the angle of attack and the slope angle of the dive strike phase. Both are constants, and the performance index is chosen as the position difference between the actual target point and the desired target point. The calculation method is the same as for the boost and pull-up phases, using least squares iteration.

[0255] The angle of attack of the dive-strike phase in each sample is achieved through the dive-strike phase. The acquisition process includes:

[0256] Select target point location As parameters, the residual function and least squares index function of the dive-strike phase are designed, and the angle of attack of the dive-strike phase is obtained by iterative solution in combination with the dynamic equation of the unpowered phase. ;

[0257] During the iteration process, the angle of attack during the dive strike phase and tilt angle The initial value is a given value.

[0258] residual function of the dive strike phase for:

[0259] ;

[0260] Least square index function of the dive strike phase for:

[0261] ;

[0262] in, for Expected value It is a 2-norm. For variables to be planned, To solve for the magnitude of the vector.

[0263] Methods to avoid no-fly zones:

[0264] In actual combat, aircraft often need to consider evading enemy anti-missile interception weapons deployed in certain areas, such as circumventing known no-fly zones. In this case, trajectory planning must not only meet process constraints such as overload, dynamic pressure, and heat flux, but also take into account no-fly zone constraints.

[0265] The solution for avoiding no-fly zones in this invention is to find a series of waypoints so that the planned trajectory passes through the waypoints to bypass the no-fly zone. The no-fly zone constraint has been simplified to a cylindrical region of infinite altitude. At this point, altitude is ignored, and the aircraft's movement is simplified to a latitude and longitude plane, that is, the no-fly zone is a circular region of fixed radius.

[0266] The flight detour strategy for no-fly zones is to first determine the minimum cost detour route, and then determine each waiting waypoint based on the minimum cost detour route, with a one-to-one correspondence between the waiting waypoints and the no-fly zone.

[0267] The specific process of determining the minimum cost detour path includes:

[0268] A1. Define the no-fly zone as a circular area with a fixed radius;

[0269] A2. In the latitude and longitude plane, draw a circle with a constant radius centered on the current flight position of the boost-glide vehicle. Calculate the cost value of each point on the circle using a cost function. Obtain, and

[0270] ;

[0271] in, The first circle The value of each point This represents the total number of no-fly zones. For the first circle The geocentric angle between the point and the target point For the first The threat level of a no-fly zone Greater than or equal to 1 For the first The radius of a no-fly zone For the first circle The point and the first The geocentric angle between the no-fly zone dots; The higher the value, the greater the threat posed by the no-fly zone, and the more likely aircraft should avoid it.

[0272] A3. Take the point with the minimum replacement value on the circle as the minimum cost point;

[0273] A4. The line connecting the minimum cost points corresponding to each flight position of the boost-glide vehicle from the launch point to the target point constitutes the minimum cost bypass path.

[0274] The methods for determining each waypoint based on the minimum cost detour route include:

[0275] B1. Make a passage through the first... The equation of a straight line perpendicular to the line connecting the launch point and the target point, drawn from the center of a no-fly zone, on a latitude-longitude plane is:

[0276] ;

[0277] in, and These are the x and y coordinates of the equation of the line, respectively. and The first The latitude and longitude of the center of the no-fly zone, The slope of the line connecting the launch point and the target point;

[0278] B2. Select the point on the least-cost detour route that has the shortest straight-line distance to the destination, and designate this point as the first... This method identifies the waiting waypoints corresponding to each no-fly zone. Clearly, each waiting waypoint determined by this method corresponds to a no-fly zone. Furthermore, it has general applicability to the functional form of no-fly zones and boasts high search efficiency.

[0279] The following section calculates and draws the minimum cost path for different no-fly zone settings, with the target point's latitude and longitude set to [30°, 30°], to demonstrate the rationality and effectiveness of the detour strategy. The no-fly zone settings for each group are shown in Table 3.

[0280] Table 3 No-Fly Zone Parameter Settings

[0281]

[0282] Table 3 lists four parameters for each no-fly zone: center longitude, center latitude, radius, and threat coefficient. By adjusting these parameters to change the complexity of the no-fly zone constraints, the selection of paths and waypoints is observed, thus verifying the feasibility of the detour strategy. The minimum cost paths and waypoints for each group are as follows. Figures 2 to 3 As shown:

[0283] A comparison of the simulation results shows that changes in various parameters of the no-fly zone affect path selection, and consequently, the obtained waypoints. With other parameters remaining constant, the threat level of the no-fly zone controls the distance between the waypoints and the no-fly zone, thus influencing the trajectory of the aircraft as it bypasses the no-fly zone. Furthermore, a certain distance is maintained between the waypoints and the no-fly zone to ensure the aircraft can safely bypass it.

[0284] Design of a neural network model for fast calculation of parameters:

[0285] Because neural networks possess parallel computation and powerful nonlinear function approximation capabilities, they can theoretically meet the requirement of rapid calculation of launch parameters. Therefore, based on the aforementioned trajectory planning process principle, training samples for the neural network are prepared, and the input and output parameters (i.e., launch parameters) of the neural network are determined. A BP network model is selected, network parameters are set, a launch parameter calculation model is established and trained, and the training results are analyzed and verified through simulations under multiple scenarios. Simultaneously, the speed of traditional trajectory planning and the trajectory planning speed using parameter calculation trained by the neural network are examined to verify the feasibility of the proposed method.

[0286] Step 3-1: Basic Structure of Feedforward Neural Network

[0287] A neural network model consists of an input layer, hidden layers, and an output layer, such as... Figure 4 The diagram shows a classic feedforward network structure. The specific training process of the neural network is explained using the case of only one hidden layer. Figure 4 The parameters used in the following steps are defined as follows:

[0288] Input layer The input values ​​of each node.

[0289] Input layer The node is connected to the hidden layer. The weight of each node.

[0290] Hidden layer The input values ​​of each node.

[0291] Hidden layer Threshold for each node.

[0292] Hidden layer activation function.

[0293] Hidden layer Output value of each node

[0294] Hidden layer The node to the output layer The weight of each node.

[0295] Output layer The input values ​​of each node.

[0296] Output layer Threshold for each node.

[0297] : Output layer activation function.

[0298] Output layer The output value of each node.

[0299] : The actual value of the output quantity.

[0300] (1) Forward propagation of signal

[0301] As above Figure 4 As shown, the network input is to The number of hidden layer nodes is Then the hidden layer is the first The input for each node is:

[0302] ;

[0303] The hidden layer function is set to Then the hidden layer is the first The output of each node is:

[0304] ;

[0305] Passed to the output layer The input for each node is:

[0306] ;

[0307] Then the first Output of each node for:

[0308] ;

[0309] Let be the excitation function. It is the independent variable;

[0310] (2) Backpropagation of error

[0311] The backpropagation process of the error is as follows: After subtracting the actual value from the current output value, this value is passed back to the neural network. The neural network then adjusts the weights and thresholds of each node accordingly to correct the deviation and uses this information for the next training round to update the output value. This process is repeated iteratively until the output error meets the requirements. The mean squared error function of the samples is as follows:

[0312] ;

[0313] In the above formula The number of output data points in a sample; As a variable;

[0314] The total deviation of the sample can then be expressed as:

[0315] ;

[0316] In the above formula The total number of samples, and As a variable;

[0317] The neural network should be updated in the direction that maximizes its partial derivative with respect to the error. Specifically, the correction method proceeds forward from the output as follows:

[0318] Output layer weights Correction formula:

[0319] ;

[0320] This represents the change in the weights of the output layer. Indicates the transmission coefficient;

[0321] Output layer threshold Correction formula:

[0322] ;

[0323] in: This indicates the change in each threshold of the output layer;

[0324] ;

[0325] Hidden layer weights Correction formula:

[0326] ;

[0327] The changes in weights of the hidden layer;

[0328] Hidden layer threshold Correction formula:

[0329]

[0330] in: This represents the change in threshold values ​​for each hidden layer.

[0331]

[0332] The final formulas are as follows:

[0333] ;

[0334] ;

[0335] ;

[0336] ;

[0337] Step 3-2: Design of the Neural Network Model for Parameter Calculation

[0338] Step 3-2-1: Input / Output Parameter Design

[0339] Based on trajectory planning, the input and output parameters of the neural network are determined. Nine variables are selected for iterative solution during the trajectory planning process, including the first-order maximum negative angle of attack. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , pull up the angle of attack to change the timing , End of pull-up segment Drag acceleration Angle of attack during dive strike phase Tilt angle These variables can all be used as output parameters for the neural network, but the tilt angle was ultimately chosen as the only one selected. The eight quantities other than those specified are used as output parameters of the neural network. The input quantities of the neural network are the latitude and longitude of the spacecraft's launch point. , latitude and longitude of the target point , And the center latitude and longitude of each no-fly zone , ; , The first The latitude and longitude of the center of the no-fly zone.

[0340] Step 3-2-1: Parameter Calculation of Neural Network Model

[0341] Considering the severe rollover problem of the gliding phase tilt angle in the simulation results, and the large variation in the constant tilt angle of the dive-strike phase across different sample cases, network prediction would require a large number of samples, resulting in high time costs. Therefore, the planning method from the previous step is still used to solve for the tilt angle of the dive-strike phase. Retaining this iteration will not significantly impact the planning time. Three different no-fly zones are still selected, hence the network is a ten-input, eight-output model, with its general structure as follows: Figure 5 As shown:

[0342] The hidden layer is set to a single layer with 30 nodes, and the activation function is set to... In this form, a linear function is selected for the output layer.

[0343] This invention employs a standard backpropagation (BP) neural network model and applies the gradient descent algorithm. The gradient is calculated using the backpropagation algorithm, thus ensuring training efficiency even with a small number of neural networks. Training is performed using the Bayesian Regularization (BR) algorithm. When the error performance function has a sum-of-squares form, A matrix has the following form:

[0344]

[0345] The gradient calculation formula is:

[0346]

[0347] In the formula, Represents the Jacobian matrix. This is the network error vector.

[0348] The loss function is defined as the error between the predicted value and the true value. For example, the mean squared error function:

[0349]

[0350] In the formula, The total number of samples, The neural network calculates the predicted output value. Output the true value for the sample.

[0351] Step 3-3: Training the neural network for parameter calculation

[0352] The trajectory data obtained from the boost phase, pull-up phase, gliding phase, and dive-strike phase are used to provide trajectory samples for neural network training. During training, the number of samples required is proportional to the complexity of the relationship between input and output; therefore, when selecting samples, it is essential to ensure the data is distributed as evenly as possible. Figure 5 A total of 10 input parameters were used. The latitude and longitude of the launch and landing point and each no-fly zone were changed individually, and 50 sets of samples were selected. Then, multiple sets of latitude and longitude were changed simultaneously to determine 50 sets, for a total of 100 sets of samples. The trajectory was planned to obtain the actual values ​​of eight output quantities. 70 sets were selected for neural network training, and the other 30 sets were used for testing experiments.

[0353] The full trajectory planning simulation is as follows:

[0354] The simulation conditions are set as follows:

[0355] The launch point has latitude and longitude of [0°, 0°], and the target point has latitude and longitude of [30°, 30°].

[0356] The no-fly zone parameters are set to [15°, 18°, 1°, 1]; [20°, 20°, 1°, 1]; [25°, 30°, 1°, 1].

[0357] The angle of attack during the gliding phase is tentatively set at 10°, and the entire flight process is simulated using an aerodynamic fitting function. The index parameters and the results of the variables to be planned are shown in Table 4, and the simulation results are as follows: Figures 6 to 11 As shown.

[0358] Table 4 Optimization Results for the Boost Phase, Pull-up Phase, and Dive Strike Phase

[0359]

[0360] Simulation curves constrained by process Figures 6 to 11 It can be seen that under the simulation conditions, the aircraft satisfies all process constraints throughout the gliding phase, and there is still a large margin compared with the given process constraint amplitude, which can ensure the safe operation of the aircraft. At the same time, the drag acceleration profile tracking effect during the gliding phase is good. When the trajectory curve is observed in the latitude and longitude coordinate system, the trajectory is relatively smooth, and it can completely bypass the no-fly zone and successfully hit the target point. The selection of the waiting waypoint also conforms to the theoretical situation.

[0361] The trained neural network model for calculating emission parameters was validated. The training results were good, with a confidence level close to 1. Further observation of the neural network prediction values ​​for each emission parameter yielded the following errors: Figures 12 to 19 As shown:

[0362] From error Figures 12 to 19 It can be seen that the error of the constant angle of attack in the dive attack phase is relatively large, with the two largest errors reaching 3 degrees. The errors of the other variables to be planned are not large, because their functional relationships with the input parameters are relatively simple and clear, easy to solve, and the training effect is relatively ideal. However, the control variables in the dive attack phase have more complex relationships with each input parameter, and are more nonlinear. When the sample size is small, the training results are not ideal. Since the current sample size is small, the error of the constant angle of attack is considered to be within a reasonable range.

[0363] Simulation results of fast parameter calculation for multiple scenarios:

[0364] Trajectory planning methods are typically designed for single-vehicle processes. However, in actual combat, a series of weapons may be launched to strike targets more effectively, necessitating consideration of planning efficiency. Traditional iterative planning methods take an average of about 25 seconds per planning iteration, and require replanning every time parameters are changed. Assuming the requirement is to launch nine aircraft from different launch points to strike a target, traditional iterative planning would take at least 250 seconds to complete trajectory planning, and even without considering no-fly zones, it would still take around 200 seconds.

[0365] Therefore, a trained neural network was used to solve for the launch parameters, and a multi-scenario trajectory planning simulation was performed. The simulation parameters were set as follows: nine launch points were selected near 0 degrees latitude and longitude, spaced 1 degree apart; the target point's latitude and longitude were [30°, 30°]; three no-fly zones were set along the route, with specific parameters [16°, 16°, 1°, 1]; [20°, 20°, 1°, 2]; [28°, 28°, 1°, 1]; and the gliding angle of attack was set to 10°. The input parameters were fed into the neural network to calculate the corresponding launch parameters. The average time for a single planning operation was 10 seconds, and for nine points, it was approximately 90 seconds. The simulation results after planning are as follows. Figure 20 A magnified view of the no-fly zone, such as Figure 21 and 22 .

[0366] Simulation results show that the aircraft can successfully navigate around no-fly zones and strike targets in all scenarios, while satisfying all constraints, thus verifying the feasibility of the proposed method. Traditional iterative parameter calculation methods for trajectory planning take an average of about 25 seconds per plan, and require replanning every time parameters are changed. For example, launching nine boost-glide vehicles from different launch points to strike a target point would take at least 200 seconds to complete trajectory planning without considering no-fly zones, resulting in low efficiency. However, using a neural network to calculate launch parameters and then plan a faster trajectory takes an average of 10 seconds per plan, completing the above plan in just 90 seconds. Therefore, the neural network-based boost-glide vehicle launch parameter calculation method proposed in this invention has a significant efficiency advantage over traditional iterative methods. It can meet the requirement of rapid trajectory planning and avoiding no-fly zones when striking targets.

[0367] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for calculating launch parameters of a boost-glide vehicle based on neural networks, characterized in that, The method includes the following steps: S1. Obtain the sample set: By individually changing the latitude and longitude of the launch point, the target point, and the center latitude and longitude of each no-fly zone, M samples are obtained based on the full trajectory data of the boost-glide vehicle. The full trajectory of the boost-glide vehicle includes the boost phase, the pull-up phase, the glide phase, and the dive-strike phase. Simultaneously, the latitude and longitude of the launch point, the target point, and the center latitude and longitude of each no-fly zone are changed, and N samples are obtained based on the full trajectory data of the boost-glide vehicle; The input data for each sample includes the latitude and longitude of the launch point and target point, as well as the center latitude and longitude of each no-fly zone. The output data includes the maximum negative angle of attack for the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , Pull up the angle of attack to change the timing , End of pull-up segment The drag acceleration represented by the ordinate of the midpoint of the drag acceleration profile. and angle of attack during dive strike ; The boost phase employs a three-stage boost system: a first-stage boost, a second-stage boost, and a third-stage boost. The boost phase also enables the maximum negative angle of attack for the first stage in each sample. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate and launch azimuth The acquisition process includes: Select the end height of the second-stage boost Boost phase end height , heading angle at the end of the boost phase As parameters, the residual function and least squares index function of the boost stage are designed, and the boost stage dynamic equations are combined with iterative solutions to obtain the maximum negative angle of attack of the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate and launch azimuth ; During the iteration process, , and The initial value is a given value. The initial value is obtained using the spherical trigonometric formula, which is: ； in, For the central angle, and These are the longitude and latitude of the launch point, respectively. and These are the longitude and latitude of the target point, respectively. The timing of the angle of attack change during the pull-up segment is achieved in each sample. , End of pull-up segment The acquisition process includes: Select the height at the end of the lifting section Track angle at the end of the pull-up phase As parameters, the residual function and least squares index function of the lifting section are designed, and the dynamic equations of the unpowered section are combined for iterative solution to obtain the moment when the angle of attack of the lifting section changes. , End of pull-up segment ; During the iteration process, and The initial value is a given value, and the two satisfy the angle of attack profile. ,and , For time, Preset angle of attack; The drag acceleration in each sample is achieved through the gliding phase. The acquisition process includes: Through integral relationships and drag acceleration profiles Determine the flight range of the boost-glide vehicle ; ； in, For the energy of the aircraft, These are the first to third undetermined coefficients, respectively. And due to the waiting flight distance The latitude and longitude of the boost-glide vehicle's current position relative to the target point's position are represented as follows: ； in, and These are the longitude and latitude of the boost-glide vehicle's current location, respectively. and These are the longitude and latitude of the target point, respectively. This is the distance from the launch point to the Earth's center. Given drag acceleration profile Initial values ​​of the x and y coordinates of the upper midpoint, and combined with the above two flight paths. The expression is solved iteratively to obtain the drag acceleration. ; The angle of attack of the dive-strike phase in each sample is achieved through the dive-strike phase. The acquisition process includes: Select target point location As parameters, the residual function and least squares index function of the dive-strike phase are designed, and the angle of attack of the dive-strike phase is obtained by iterative solution in combination with the dynamic equation of the unpowered phase. ; During the iteration process, the angle of attack during the dive strike phase and tilt angle The initial value is a given value; S2. Select P samples from M+N samples as the training set to train the parameter calculation neural network model and obtain the trained parameter calculation neural network model; wherein, the input data and output data of the samples are used as the input and output data of the parameter calculation neural network model, respectively. S3. Input the latitude and longitude of the current launch point, target point, and the center latitude and longitude of each no-fly zone into the trained parameter calculation neural network model. The trained parameter calculation neural network model predicts the launch parameters of the planned trajectory of the boost-glide vehicle, including the maximum negative angle of attack of the first stage. Maximum negative angle of attack at level 2 Level 3 constant pitch angle change rate Launch azimuth , Pull up the angle of attack to change the timing , End of pull-up segment Drag acceleration and angle of attack during dive strike .

2. The method for calculating launch parameters of a boost-glide vehicle based on a neural network according to claim 1, characterized in that, residual function of the boost phase for: ； Least square index function of the boost phase for: ； in, , , They are respectively , and Expected value It is a 2-norm. These are variables to be planned.

3. The method for calculating launch parameters of a boost-glide vehicle based on a neural network according to claim 1, characterized in that, Residual function of the pull-up segment for: ； Least square index function of the pull-up segment for: ； in, , They are respectively and The expected value of the parameters, It is a 2-norm. These are variables to be planned.

4. The method for calculating launch parameters of a boost-glide vehicle based on a neural network according to claim 1, characterized in that, residual function of the dive strike phase for: ； Least square index function of the dive strike phase for: ； in, for Expected value It is a 2-norm. For variables to be planned, To solve for the magnitude of the vector.

5. The method for calculating launch parameters of a boost-glide vehicle based on a neural network according to claim 1, characterized in that, The flight detour strategy for no-fly zones is to first determine the minimum cost detour route, and then determine each waiting waypoint based on the minimum cost detour route, with a one-to-one correspondence between the waiting waypoints and the no-fly zone.

6. The method for calculating launch parameters of a boost-glide vehicle based on a neural network according to claim 5, characterized in that, The specific process of determining the minimum cost detour path includes: A1. Define the no-fly zone as a circular area with a fixed radius; A2. In the latitude and longitude plane, draw a circle with a constant radius centered on the current flight position of the boost-glide vehicle. Calculate the cost value of each point on the circle using a cost function. Obtain, and ； in, The first circle The value of each point This represents the total number of no-fly zones. For the first circle The geocentric angle between the point and the target point For the first The threat level of a no-fly zone Greater than or equal to 1 For the first The radius of a no-fly zone For the first circle The point and the first The geocentric angle between the no-fly zone dots; A3. Take the point with the minimum replacement value on the circle as the minimum cost point; A4. The line connecting the minimum cost points corresponding to each flight position of the boost-glide vehicle from the launch point to the target point constitutes the minimum cost bypass path. The methods for determining each waypoint based on the minimum cost detour route include: B1. Make a passage through the first... The equation of a straight line perpendicular to the line connecting the launch point and the target point, drawn from the center of a no-fly zone, on a latitude-longitude plane is: ； in, and These are the x and y coordinates of the equation of the line, respectively. and The first The latitude and longitude of the center of the no-fly zone, The slope of the line connecting the launch point and the target point; B2. Select the point on the least-cost detour route that has the shortest straight-line distance to the destination, and designate this point as the first... The waiting waypoints corresponding to each no-fly zone.

Images