Shape-flight path integrated obstacle avoidance planning method based on Q-learning and prediction correction

Abstract

The invention relates to a profile-flight path integrated obstacle avoidance planning method based on Q-learning and prediction correction, and belongs to the technical field of flight path planning. Longitudinal guidance is designed by adopting an analytic prediction correction guidance method, so that the calculation speed and real-time performance are ensured; lateral guidance adopts a segmented guidance method, a heeling angle amplitude increasing strategy is switched to avoid when the vehicle is close to the no-fly zone, and the original lateral control logic is switched back when the vehicle flies away from the no-fly zone, so that real-time obstacle avoidance can be realized. According to the lateral guidance method, a transverse range corridor which converges along with the flying distance is designed by adopting a transverse range formula, so that the guidance precision is ensured, and the problem of frequent reversal of a heeling angle is avoided. The control quantity is corrected in real time by predicting the terminal state online in real time, and the terminal state error of the aircraft is eliminated; and by correcting the variable configuration coefficient of the aircraft, the optimal lift-drag ratio adapting to the current flight state can be given, and the striking precision of the high-speed aircraft is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of flight trajectory planning, and relates to a shape-flight trajectory integrated obstacle avoidance planning method based on Q-learning and prediction correction, which can realize real-time intelligent trajectory planning under different configuration conditions based on given tasks and constraints; and is applied to the guidance and control field of intelligent deformable high-speed gliding aircraft. Background Art

[0002] The intelligent trajectory planning technology of high-speed gliding vehicles is mainly used to plan the optimal trajectory for the reentry phase. The reentry phase includes the gliding phase and the dive phase. The performance requirements of different phases are different. The gliding phase needs to fly with the optimal lift-to-drag ratio to increase the range; the dive phase needs to reduce drag, and the lateral available overload needs to be increased to enhance the penetration capability. The optimal aerodynamic shape is different in different flight phases. However, the aerodynamic shape of existing aircraft is generally the optimal shape under specific working conditions, which cannot take into account the performance requirements of each guided flight phase. The variable configuration aircraft can adaptively change its shape according to the flight environment and flight performance requirements during the guided flight, thereby affecting the flight lift-to-drag ratio, so that the aircraft can maintain the optimal flight state. The introduction of deformation technology in high-speed aircraft can alleviate the performance limitations of traditional fixed-shape aircraft and further improve the flight capability of high-speed gliding aircraft.

[0003] In view of the strong nonlinearity, strong coupling and multiple constraints of the dynamic equations of the reentry phase of high-speed variable configuration aircraft, the pseudo-spectral method is often used to model the reentry trajectory planning problem as an optimal control problem, and then directly match the discrete control variables and state variables. This method has the advantages of low initial value sensitivity, easy convergence and high precision. However, when the model is too complex, the requirements and sensitivity of the pseudo-spectral method for initial values ​​will increase. At the same time, the reentry trajectory planning of aircraft based on the pseudo-spectral method is essentially an offline planning that relies on prior information of the environment and is difficult to cope with changes in the dynamic environment, such as various unknown obstacles that appear during flight. The prediction and correction method can correct the control quantity in real time by predicting the terminal state online, thereby eliminating the terminal state error of the aircraft, and can correct the guidance instructions in advance to strictly control the relevant constraints of the prediction to meet the mission requirements. Therefore, the prediction and correction method is easier to handle process constraint problems and has higher precision and robustness. Summary of the invention

[0004] The technical problems to be solved by the present invention are:

[0005] In order to solve the shape decision and trajectory planning problems of a high-speed variable configuration aircraft in the reentry phase, the present invention proposes a shape-flight trajectory integrated obstacle avoidance planning method based on Q-learning and prediction correction.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A shape-flight trajectory integrated obstacle avoidance planning method based on Q-learning and prediction correction, characterized in that it includes a longitudinal guidance strategy and a lateral guidance strategy;

[0008] The longitudinal guidance strategy:

[0009] The analytical prediction correction guidance method is adopted, and the characteristics of quasi-balanced gliding flight are utilized. Through appropriate simplification, the analytical mathematical relationship between the remaining range and the control amount roll angle and variable configuration coefficient is established based on energy, thereby obtaining an analytical prediction of the remaining range, as shown in the following formula:

[0010]

[0011] Where E0 is the energy factor of the current point; E f is the energy factor expected at the terminal moment; s0 is the remaining range of the aircraft at the current point; s f The energy factor reaches E f The remaining distance at , that is, the deviation from the expected target point; R0 is the radius of the earth; is the drag acceleration, is the lift acceleration; K_morph is the aircraft's variable configuration coefficient;

[0012] The correction algorithm based on Newton iteration method is used to adjust the size of the roll angle and the variable configuration coefficient:

[0013] Adjust the tilt angle σ abs To satisfy the following conditions

[0014] s f =s(E f )=0

[0015] The roll angle correction formula is:

[0016]

[0017] In the formula, It can be approximated by the finite difference method, where the subscript i represents the i-th iteration;

[0018] The correction formula for the isomorphic coefficient is:

[0019]

[0020] In the formula, K_morph i The subscript i indicates the i-th iteration;

[0021] The lateral guidance strategy:

[0022] A lateral corridor that converges with the distance to be flown is designed using the lateral formula, so that the lateral trajectory gradually converges to the predetermined position.

[0023] The calculation formula of the horizontal distance is

[0024] Z = -arcsin(sins g ·sin(ψ-ψ LOS ))

[0025] Where ψ is the heading angle, ψ LOS is the sight azimuth;

[0026] The following horizontal corridor with convergent waiting distance is adopted

[0027]

[0028] Among them, Δψ verf =(Δψ vd0 -Δψ vdf )s g / s g0 +Δψ vdf is the variation law of the designed reference heading angle deviation, Δψ vd0 , Δψ vdf are the initial and final predetermined heading angle deviations, s g0 is the distance to be flown at the initial moment of gliding, s g is the remaining range after clipping;

[0029] Based on the lateral control boundary Z, design the lateral guidance law S σ (t) is

[0030]

[0031] In the formula, τ is the calculation period, K is the control parameter, is the rate of change of the lateral distance deviation, which plays a damping role;

[0032] Combined with the roll angle command size σ determined by the prediction correction method in longitudinal guidance abs The final roll angle command σ can be obtained c for

[0033] σ c =S σ (t)σ abs

[0034] In order to enable the high-speed variable configuration aircraft to avoid the no-fly zone, a segmented lateral guidance method based on predictive correction guidance is designed; when the aircraft approaches the no-fly zone, the original lateral control logic is abandoned and the no-fly zone avoidance lateral control logic is switched; an avoidance strategy is designed for the cylindrical no-fly zone, and the strategy of increasing the amplitude of the roll angle is used for avoidance;

[0035] When the aircraft approaches a no-fly zone, the condition for enabling the no-fly zone avoidance lateral control logic is given by the following formula:

[0036] S n ≤n0R n ,λ≤λ n

[0037] In the formula, S n is the surface distance from the variable configuration aircraft to the center of the no-fly zone, R n is the radius of the no-fly zone, λ n is the longitude of the center of the no-fly zone, and n0 is a set constant;

[0038] The avoidance conditions of the no-fly zone are:

[0039] ψ>ψ max orψ<ψ min

[0040] In the formula, ψ max is the maximum sight angle that can be used when flying over the no-fly zone, ψ min The minimum sight angle that can be used when flying under a no-fly zone;

[0041] When S n ≤n0R n ,λ≤λ n and ψ>ψ max orψ<ψ min When both conditions are met, increase the roll angle amplitude to avoid the no-fly zone

[0042] σ|=|σ c |+σ n

[0043] In the formula, σ n >0;

[0044] The Q-learning algorithm is used to obtain the optimal variable configuration coefficient interval [K min ,K max]; The Q-learning algorithm approximates the optimal value function by optimizing and iteratively calculating the Q(state, action) function, where state represents the current state, action represents the use of the ε-greedy strategy to select the best action to obtain the reward R, and the Q value of Q(state, action) is modified according to the update rule:

[0045] Q(state,action)=(1-η)Q(state,action)+η[R+γmaxQ(state′,action′)]

[0046] Where η is the learning rate, which determines the magnitude of each update; γ is the discount factor, which is used to weigh the importance of immediate rewards and future rewards; state′ is the next state, and action′ is the optimal action in the next state;

[0047] Through the continuous iteration of the Q-learning algorithm, the Q value table for airfoil decision can be obtained to select the optimal variable configuration coefficient interval [K min ,K max ]; The process of optimizing and updating the variable configuration coefficient is shown in the following formula:

[0048]

[0049] A computer system, characterized in that it comprises: one or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned method.

[0050] A computer-readable storage medium, characterized in that it stores computer-executable instructions, which are used to implement the above method when executed.

[0051] A computer program product, characterized by comprising computer executable instructions, wherein the instructions are used to implement the above method when executed.

[0052] The beneficial effects of the present invention are:

[0053] The present invention is aimed at the trajectory planning problem of high-speed variable configuration aircraft. Based on the Q-learning algorithm and the predictive correction guidance method, an obstacle avoidance trajectory planning method integrating shape and flight trajectory is designed. This method can select the optimal configuration that adapts to the current mission according to the flight state of the aircraft when a flight mission is given, so as to obtain the best flight performance. The longitudinal guidance is designed using the analytical predictive correction guidance method to ensure the speed and real-time performance of the calculation; the lateral guidance adopts the segmented guidance method, switching to the strategy of increasing the amplitude of the bank angle for avoidance when approaching the no-fly zone, and switching back to the original lateral control logic when leaving the no-fly zone, so as to achieve real-time obstacle avoidance. The lateral guidance of the present invention adopts the lateral range formula to design a lateral range corridor that converges with the distance to be flown, which ensures the guidance accuracy and avoids the problem of frequent reversal of the bank angle. Compared with the traditional pseudo-spectral method, the method proposed in the present invention can better adapt to changes in dynamic environment. It can correct the control quantity in real time by predicting the terminal state online, thereby eliminating the terminal state error of the aircraft. And by correcting the variable configuration coefficient of the aircraft, it can give the best lift-to-drag ratio that adapts to the current flight state, greatly improving the strike accuracy of high-speed aircraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0055] Figure 1 Flow chart of the method of the present invention.

[0056] Figure 2 No-fly zone avoidance flight trajectory.

[0057] Figure 3 Comparison of real-time no-fly zone avoidance trajectories between the prediction-correction method and the pseudospectral method.

[0058] Figure 4 The curve of the change of the conformational coefficient.

[0059] Figure 5 Prediction of landing point error change curve.

[0060] Figure 6 Remaining range change curve. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0062] The present invention studies the shape-flight trajectory integrated planning technology of a variable configuration aircraft based on Q-learning and a prediction correction guidance method, and considers the obstacles that may appear in dynamic scenes, designs an obstacle avoidance strategy, and avoids cylindrical obstacles that appear during gliding in real time. First, a reentry dynamics model of a variable configuration aircraft is established, and relevant constraints are given. Based on energy, an analytical mathematical relationship between the remaining range and the control amount roll angle is established, and the remaining range is predicted in real time. The landing point error is continuously corrected by correcting the roll angle, thereby improving the guidance accuracy; and the Q-learning algorithm can be used to obtain the optimal flight configuration according to the flight state and flight mission. In order to achieve obstacle avoidance, the present invention designs a lateral segmented guidance strategy, which executes the roll angle reversal logic when away from the no-fly zone, and switches to the no-fly zone avoidance lateral control logic when it is determined to be close to the no-fly zone, thereby planning a safe and reliable flight trajectory. The trajectory planning scheme designed by the present invention can enhance the safety and reliability of high-speed gliding variable configuration aircraft, and also improve the guidance accuracy of the system.

[0063] like Figure 1 As shown, the following steps are included:

[0064] The aerodynamic model of a morphable aircraft is related to the angle of attack and the Mach number, and is also related to the deformation. Considering the morphing coefficient K_morph∈[1.0,1.4] of a morphable aircraft, when the aircraft shape changes, assuming that the drag coefficient C D Unchanged, lift coefficient C L Change to C L ′, then C L ′=K_morph×C L .

[0065] The three-degree-of-freedom dynamic equation of the high-speed variable-configuration aircraft is:

[0066]

[0067] Wherein, r is the distance from the center of the earth; λ and φ are respectively the longitude and latitude of the projection point of the aircraft on the surface, both in radians; V is the speed; θ is the track angle; ψ is the heading angle; σ is the roll angle; L and D are respectively the lift and drag; m is the mass. In the present invention, the superscript symbol · represents the first-order derivative of the corresponding quantity with respect to time t.

[0068] There is a constraint on the roll angle amplitude, as shown in formula (2):

[0069]

[0070] Where σ(τ) is the current roll angle command, σ c is the maximum value of the limited roll angle amplitude.

[0071] Considering the initial thermal protection requirements and subsequent gliding capability, the angle of attack is taken as a function of speed, as shown in formula (3):

[0072]

[0073] In the formula, α max is a larger constant angle of attack; α LD is the angle of attack corresponding to the maximum lift-to-drag ratio of the variable configuration aircraft; V1 and V2 are given constant speeds.

[0074] The longitudinal guidance of the present invention adopts an analytical prediction correction guidance method. By utilizing the characteristics of quasi-balanced gliding flight, through appropriate simplification, an analytical mathematical relationship between the remaining range and the control amount roll angle and variable configuration coefficient is established based on energy, thereby realizing analytical prediction of the remaining range.

[0075] The energy factor per unit mass is defined as

[0076] E=(V / V c ) 2 / 2-R0 / r (4)

[0077] Where V is the speed, g0=9.81m / s 2 , R0 is the radius of the earth. Taking the derivative of (4) and combining it with the mathematical model of the gliding phase of the variant aircraft, we can get

[0078]

[0079] in, is the drag acceleration, D is the drag, and m is a known quantity. Obviously, the energy factor decreases monotonically with time, so it can be used as the independent variable of the differential equation instead of time.

[0080] The remaining distance s is defined as the great circle connecting the current position and the target position, and its analytical calculation formula is:

[0081] coss=sinφsinφ f +cosφcosφ f cos(λ f -λ) (6)

[0082] Where, λ and φ are the current longitude and latitude of the aircraft respectively, λ f ,φ f are the longitude and latitude of the target endpoint, respectively.

[0083] At the same time, the variation law of s over time is

[0084]

[0085] Where r is the distance from the center of the earth, θ is the track angle, and Δψ is the difference between the heading angle and the line of sight azimuth.

[0086] Since periodic rolling motion is required during the energy management process, Δψ will show positive and negative periodic changes near 0° and is small. For the convenience of analytical solution, it can be considered that Δψ≈0, so the above formula can be simplified to

[0087]

[0088] Using equations (5) and (8), the derivative of the remaining range s with respect to the energy factor E can be obtained as:

[0089]

[0090] The quasi-balanced gliding condition is established as

[0091] m(g 2 / r)-Lcosσ=0 (10)

[0092] Where σ is the tilt angle.

[0093] From formula (10), we can get

[0094]

[0095] In the formula, is the lift acceleration.

[0096] Convert the drag acceleration into a function of lift acceleration and lift-to-drag ratio, and combine (11) to obtain

[0097]

[0098] According to the energy factor expression (4), we can know

[0099] V 2 =2(E+R0 / r)V c 2 (13)

[0100] Substituting (13) into (12), we can get

[0101]

[0102] Assuming that the quasi-balanced gliding process is flying in a certain configuration, substitute equation (14) into equation (9) and integrate to solve the remaining range to the gliding terminal:

[0103]

[0104] Where E0 is the energy factor of the current point; E fis the energy factor expected at the terminal moment; s0 is the remaining range of the aircraft at the current point; s f The energy factor reaches E f The remaining distance at that time is the deviation from the expected target point.

[0105] Next, the roll angle should be corrected in real time to eliminate the remaining range deviation. The longitudinal guidance of the present invention uses a correction algorithm based on the Newton iteration method to adjust the size of the roll angle. The specific implementation method is as follows:

[0106] According to the goal of reentry glide guidance, the waiting distance of the glide terminal should meet

[0107] s f =s(E f )=0 (16)

[0108] The above formula shows that when the energy of the aircraft reaches the set terminal energy, the distance to be flown should be zero, that is, the unity of range and energy is achieved.

[0109] Since the distance to be flown is s f is a nonlinear function of the roll angle. The present invention uses a correction algorithm based on the Newton iteration formula to adjust the roll angle σ abs To satisfy formula (16), the correction formula is:

[0110]

[0111] In the formula, It can be approximated by the finite difference method, where the subscript i represents the i-th iteration.

[0112] In each guidance cycle, equations (15) and (17) are used to iteratively solve. When equation (16) is satisfied, the roll angle value is output as one of the longitudinal guidance instructions. Since the guidance instructions are always adjusted in real time according to the current flight status during the flight, the influence of initial deviation and interference on the guidance accuracy is effectively reduced, and the adaptability of the guidance is improved.

[0113] Next, we design a lateral guidance method for the variable configuration aircraft. In order to enable the high-speed variable configuration aircraft to avoid the no-fly zone, a segmented lateral guidance method based on predictive correction guidance is designed. When the aircraft approaches the no-fly zone, the original lateral control logic is abandoned and the no-fly zone avoidance lateral control logic is switched. After confirming that it has flown away from the no-fly zone and reached a safe area, the original lateral control logic is switched back to use. The specific method is as follows:

[0114] Traditional lateral guidance is usually designed using the heading angle deviation corridor method. Considering that this method has low accuracy and is prone to the deficiency of too frequent roll angle reversal times, and the lateral range changes more slowly and predictably when approaching the target point, the present invention directly uses the lateral range formula to design a lateral range corridor that converges with the distance to be flown, so that the lateral trajectory gradually converges to a predetermined position.

[0115] The calculation formula of the horizontal distance is

[0116] Z = -arcsin(sins g ·sin(ψ-ψ LOS )) (18)

[0117] In the formula, ψ v is the heading angle, ψ LOS It is the sight azimuth, that is, the angle between the line connecting the current point and the target point and the north direction. Its direction is defined in the same way as the heading angle, that is, it is positive when rotating counterclockwise from the north direction. g is the remaining range after clipping, that is

[0118]

[0119] In the formula, s ε is the set threshold value to avoid s g =0.

[0120] Line of sight azimuth ψ LOS The calculation formula is

[0121]

[0122] Using the traverse calculation formula (18), the present invention adopts the following traverse corridor for convergence of the waiting distance:

[0123]

[0124] Among them, Δψ verf =(Δψ vd0 -Δψ vdf )s g / s g0 +Δψ vdf is the variation law of the designed reference heading angle deviation, Δψ vd0 , Δψ vdf are the initial and final predetermined heading angle deviations, s g0 It is the distance to be flown at the initial moment of gliding.

[0125] Obviously, Δψ vd0 , Δψ vdf The opening sizes at the beginning and end of the horizontal corridor are determined respectively, and Δψ is adjusted vd0 , Δψ vdfA funnel-shaped lateral corridor is formed, and the lateral corridor converges with the distance to be flown. Since equation (21) directly uses the lateral formula to form the lateral corridor, the physical meaning is clear, and the parameter design is simple and intuitive.

[0126] Based on lateral control boundary The lateral guidance law S can be further designed σ (t) is

[0127]

[0128] In the formula, τ is the calculation period, K is the control parameter, It is the rate of change of the lateral distance deviation and plays a damping role.

[0129] Combined with the roll angle command size σ determined by the prediction correction method in longitudinal guidance abs The final roll angle command σ can be obtained c for

[0130] σ c =S σ (t)σ abs (twenty three)

[0131] The present invention designs an avoidance strategy for cylindrical no-fly zones, and adopts a strategy of increasing the roll angle amplitude to avoid them. When the aircraft approaches the no-fly zone, the condition for enabling the no-fly zone avoidance lateral control logic is given by equation (24):

[0132] S n ≤n0R n ,λ≤λ n (twenty four)

[0133] In the formula, S n is the surface distance from the variable configuration aircraft to the center of the no-fly zone, R n is the radius of the no-fly zone, λ n is the longitude of the center of the no-fly zone, and n0 is a set constant.

[0134] First, solve the distance S from the aircraft to the center of the no-fly zone n :

[0135]

[0136] In the formula, λ n ,φ n are the longitude and latitude of the center of the no-fly zone, Ω n YesS n The corresponding geocentric angle.

[0137] The central sight angle of the no-fly zone is defined as:

[0138] ψ n=arcsin(sin(λ n -λ)cosφ / sin(φ n -φ)) (26)

[0139] In order to decide whether to avoid the no-fly zone from above or below, ψ n Compare with the current heading angle ψ of the aircraft, if ψ≤ψ n , then avoid the no-fly zone; if ψ>ψ n , then avoid from below the no-fly zone; then solve the maximum sight angle ψ that can be used to fly above the no-fly zone max and the minimum sight angle ψ that can be used when flying under the no-fly zone min :

[0140]

[0141] When the north direction is taken as the reference, the amplitude of each sight angle satisfies ψ max |<ψ<ψ min |.

[0142] For the no-fly zone, the bank angle amplitude is increased to avoid it. The avoidance condition of the no-fly zone is:

[0143] ψ>ψ max orψ<ψ min (28)

[0144] When conditions (24) and (28) are met simultaneously, increase the roll angle amplitude

[0145] σ|=|σ c |+σ n (29)

[0146] In the formula, σ n >0.

[0147] Finally, the optimal variable configuration coefficient under the current flight state is obtained by using the Q-learning algorithm. The corresponding optimal airfoil is selected according to the main mission and flight situation of the current aircraft, so that the optimal flight state can be maintained throughout the flight. The present invention considers the three most common flight missions of high-speed variable configuration aircraft, including gliding flight, maneuver penetration and dive strike.

[0148] Set reward functions according to the characteristics of different flight missions.

[0149] (1) Gliding flight: R = -0.07D + 0.3L + 0.2V;

[0150] In the gliding mission segment, the reward function increases the penalty of drag and reduces the rewards of lift and speed.

[0151] (2) Mobile penetration: R = -0.05D + 0.5L + 0.5V;

[0152] The reward function reduces the penalty of drag and increases the reward of speed in the maneuver penetration phase compared to the gliding flight phase.

[0153] (3) Dive strike: R = -0.03D + 0.7L + 0.7V.

[0154] In the dive attack phase, the proportion of resistance in the reward function is further reduced. In order to ensure the lethality during the dive attack, the proportion of speed in the reward function is increased.

[0155] The Q-learning algorithm approximates the optimal value function by optimizing and iteratively calculating the Q(state, action) function, where state represents the current state, action represents the optimal action selected using a certain strategy to obtain the reward R, and the Q value of Q(state, action) is modified according to the update rule. The ε-greedy strategy is used as the action selection strategy of Q-learning:

[0156]

[0157] Wherein, the subscript i represents the i-th step of the iterative algorithm; ε represents the proxy parameter, which is a positive number less than 1.

[0158] The update rules are:

[0159] Q(state,action)=(1-η)Q(state,action)+η[R+γmaxQ(state′,action′)](31)

[0160] Among them, η is the learning rate, which determines the magnitude of each update; γ is the discount factor, which is used to weigh the importance of immediate rewards and future rewards; state′ is the next state, and action′ is the optimal action in the next state.

[0161] By using the above Q-learning algorithm to continuously iterate, we can obtain the Q value table for airfoil decision making, which is used to select the best variable configuration coefficient that suits the current flight mission. At the same time, considering the influence of factors such as air density and Reynolds number in the actual flight of the aircraft, and the time required for the aircraft to change its configuration, the best variable configuration coefficient selected by the Q value table is classified as an interval [K min ,K max ]. The process of optimizing and updating the conformational coefficient is shown in formula (32):

[0162]

[0163] The following is a detailed description of the effects of this application in conjunction with simulation:

[0164] Considering the problem of high-speed variable-configuration aircraft avoiding no-fly zones when performing missions, a set of strike combat scenarios with no-fly zones are designed for the variable-configuration aircraft dynamics model in equation (1). The starting latitude and longitude coordinates of the variable-configuration aircraft are set to (15°, 30°), and the end coordinates are set to (23.2°, 33.6°). The no-fly zone is set on the necessary path of the original trajectory of the aircraft, with its center coordinates at (22.6°, 33°) and a radius of 25,000m. Assuming that the aircraft glides at a fixed angle of attack α = 10° and the turning method is a banked turn (BTT), the sideslip angle is always 0° during the flight. The variable-configuration aircraft can change the lift-to-drag ratio of the aircraft by changing the wing shape, and the variable-configuration coefficient satisfies K_morph∈[1.0,1.4]. The remaining initial conditions are given in the following table:

[0165]

[0166] By adopting the trajectory planning strategy provided by the present invention to avoid the no-fly zone, the flight trajectory can be obtained as follows: Figure 2 As shown in the figure. As can be seen from the figure, the no-fly zone is set on the necessary path of the original trajectory of the high-speed variable configuration aircraft. The segmented prediction and correction guidance strategy considering the avoidance of the no-fly zone is adopted. When the aircraft approaches the no-fly zone, the dynamic avoidance strategy is turned on. By calculating the geometric relationship between the no-fly zone, the aircraft and the target, the avoidance strategy of bypassing the left or right is obtained and implemented. As shown in the figure, under this condition, the aircraft dynamically avoids from the "bottom" of the no-fly zone, switches back to the original lateral guidance logic of the bank angle reversal after the no-fly zone section ends, and turns to reach the target landing area. The simulation results show the effectiveness of the analytical prediction and correction guidance strategy considering the avoidance of the no-fly zone.

[0167] In order to verify the effectiveness of the trajectory planning strategy given by the present invention, the results are compared with those of the traditional pseudo-spectral trajectory planning under the same conditions. Figure 3 As shown in the figure, it can be seen that the trajectory pre-planned by the pseudo-spectral method cannot avoid obstacles that suddenly appear during flight, while the prediction and correction guidance method can well cope with the dynamic environment during flight and adjust the trajectory in real time to avoid the no-fly zone.

[0168] The change curve of the conformational coefficient after optimization using Q-learning is as follows: Figure 4 The landing point error prediction curve and the remaining range change curve are shown in Figure 5 and Figure 6The simulation results show that the prediction correction guidance method is used to design the flight trajectory. The predicted landing point error and the remaining range are continuously reduced and eventually approach 0. The error between the final landing point and the target point is only 1429m, which shows that it has a high strike accuracy.

[0169] In summary, the feasibility and effectiveness of the predictive correction guidance method for high-speed variable-configuration aircraft considering no-fly zone avoidance proposed in the present invention can be verified.

[0170] The present invention designs a shape-flight trajectory integrated planning method for multi-task configurations based on Q-learning and predictive correction guidance methods. At the same time, considering that traditional reentry trajectory planning methods are difficult to avoid constraints such as obstacles that appear during the flight of an aircraft, the present invention designs a predictive correction guidance strategy that can perform real-time obstacle avoidance, thereby planning a feasible reentry trajectory and reducing dependence on prior knowledge of the environment. For longitudinal guidance, the present invention designs an energy-based analytical prediction method, and corrects the roll angle and variable configuration coefficient online in real time, thereby continuously correcting errors. This method does not require real-time integral calculation of the trajectory equation, which significantly reduces the amount of calculation and time consumption, and also has higher accuracy. For lateral guidance, the present invention designs a segmented guidance method that can use a strategy of increasing the roll angle amplitude to avoid obstacles when approaching them, thereby improving the safety and reliability of trajectory planning.

[0171] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be included in the protection scope of the present invention.

Claims

1. A shape-flight trajectory integrated obstacle avoidance planning method based on Q-learning and prediction correction, characterized in that: Including longitudinal guidance strategy and lateral guidance strategy; The longitudinal guidance strategy: The analytical prediction correction guidance method is adopted, and the characteristics of quasi-balanced gliding flight are utilized. Through appropriate simplification, the analytical mathematical relationship between the remaining range and the control amount roll angle and variable configuration coefficient is established based on energy, thereby obtaining an analytical prediction of the remaining range, as shown in the following formula: Where E0 is the energy factor of the current point; E f is the energy factor expected at the terminal moment; s0 is the remaining range of the aircraft at the current point; s f The energy factor reaches E f The remaining distance at , that is, the deviation from the expected target point; R0 is the radius of the earth; is the drag acceleration, is the lift acceleration; K_morph is the aircraft's variable configuration coefficient; The correction algorithm based on Newton iteration method is used to adjust the size of the roll angle and the variable configuration coefficient: Adjust the tilt angle σ abs To satisfy the following conditions s f =s(E f )=0 The roll angle correction formula is: In the formula, It can be approximated by the finite difference method, where the subscript i represents the i-th iteration; The correction formula for the isomorphic coefficient is: In the formula, K_morph i The subscript i indicates the i-th iteration; The lateral guidance strategy: A lateral corridor that converges with the distance to be flown is designed using the lateral formula, so that the lateral trajectory gradually converges to the predetermined position. The calculation formula of the horizontal distance is Z=-arcsin(sins g ·sin(ψ-ψ LOS )) Where ψ is the heading angle, ψ LOS is the sight azimuth; The following horizontal corridor with convergent waiting distance is adopted Among them, Δψ verf =(Δψ vd0 -Δψ vdf )s g / s g0 +Δψ vdf is the variation law of the designed reference heading angle deviation, Δψ vd0 , Δψ vdf are the initial and final predetermined heading angle deviations, s g0 is the distance to be flown at the initial moment of gliding, s g is the remaining range after clipping; Based on lateral control boundary Design lateral guidance law S σ (t) is In the formula, τ is the calculation period, K is the control parameter, is the rate of change of the lateral distance deviation, which plays a damping role; Combined with the roll angle command size σ determined by the prediction correction method in longitudinal guidance abs The final roll angle command σ can be obtained c for s c =S σ (t)s abs In order to enable the high-speed variable configuration aircraft to avoid the no-fly zone, a segmented lateral guidance method based on predictive correction guidance is designed; when the aircraft approaches the no-fly zone, the original lateral control logic is abandoned and the no-fly zone avoidance lateral control logic is switched; an avoidance strategy is designed for the cylindrical no-fly zone, and the strategy of increasing the amplitude of the roll angle is used for avoidance; When the aircraft approaches a no-fly zone, the condition for enabling the no-fly zone avoidance lateral control logic is given by the following formula: S n ≤n0R n ,λ≤λ n In the formula, S n is the surface distance from the variable configuration aircraft to the center of the no-fly zone, R n is the radius of the no-fly zone, λ n is the longitude of the center of the no-fly zone, and n0 is a set constant; The avoidance conditions of the no-fly zone are: ψ>ψ max orψ<ψ min In the formula, ψ max is the maximum sight angle that can be used when flying over the no-fly zone, ψ min The minimum sight angle that can be used when flying under a no-fly zone; When S n ≤n0R n ,λ≤λ n and ψ>ψ max orψ<ψ min When both conditions are met, increase the roll angle amplitude to avoid the no-fly zone σ|=|σ c |+s n In the formula, σ n >0; The Q-learning algorithm is used to obtain the optimal variable configuration coefficient interval [K min ,K max ]; The Q-learning algorithm approximates the optimal value function by optimizing and iteratively calculating the Q(state, action) function, where state represents the current state, action represents the use of the ε-greedy strategy to select the best action to obtain the reward R, and the Q value of Q(state, action) is modified according to the update rule: Q(state,action)=(1-η)Q(state,action)+η[R+γmaxQ(state′,action′)] Where η is the learning rate, which determines the magnitude of each update; γ is the discount factor, which is used to weigh the importance of immediate rewards and future rewards; state′ is the next state, and action′ is the optimal action in the next state; Through the continuous iteration of the Q-learning algorithm, the Q value table for airfoil decision can be obtained to select the optimal variable configuration coefficient interval [K min ,K max ]; The process of optimizing and updating the variable configuration coefficient is shown in the following formula:

2. A computer system, characterized in that include: One or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method of claim 1.

3. A computer-readable storage medium, characterized in that Computer executable instructions are stored, and when the instructions are executed, they are used to implement the method of claim 1.

4. A computer program product, characterized in that The invention comprises computer executable instructions, which are used to implement the method of claim 1 when being executed.

Images