Method for controlling disturbed allosteric aircraft based on switching model prediction

Abstract

The invention discloses a disturbed variable-structure aircraft control method based on switching model prediction, and solves the problem that a conventional non-optimization control method of a disturbed variable-structure aircraft lacks the capability of generating an optimal control surface control instruction according to different performance optimization indexes under the flight capability constraint. And the stability of a closed-loop system under small disturbance cannot be ensured. According to the disturbed variable-structure aircraft control method provided by the invention, robust optimization control of the variable-structure aircraft under multiple constraints can be realized, the feasibility of an optimization control algorithm and the stability of a closed-loop system are ensured, and the flight performance of the variable-structure aircraft is favorably improved.

Description

Technical Field

[0001] The invention belongs to the field of optimized control of conformational aircraft, and relates to a control method for a subsonic small-disturbance variable-span aircraft, which is used to realize optimized control of autonomous control surface deflection on a conformational aircraft. Background Art

[0002] Traditional fixed-configuration aircraft can only meet specific mission requirements due to their simple structure and single flight mode. With the continuous development of science and technology, in order to cope with complex flight environments and diverse mission requirements, future aircraft development will be unmanned, autonomous, and intelligent. As a major development direction for the next generation of aircraft technology, variable-configuration aircraft have a variety of variable-configuration methods to adjust the partial or overall shape. Compared with fixed-configuration aircraft, variable-configuration aircraft have a wider flight speed range and better aerodynamic shape layout, which can greatly expand the existing flight capability boundaries and have efficient flight capabilities in a wider flight airspace, wider flight altitude, and longer flight time.

[0003] Generally, a disturbed conformational aircraft can construct multiple discrete configurations (ignoring the conformational process) according to the flight performance requirements. Therefore, a subsystem for each discrete configuration can be established according to the dynamic parameters of different discrete configurations, and the aircraft conformational switching system can be modeled through switching logic. At present, the control methods for the conformational switching system of disturbed discrete aircraft mainly include traditional linear control methods and emerging intelligent control methods. Although traditional linear control methods (PID, anti-disturbance control methods, etc.) can complete the aircraft control task through multi-level segmentation or decoupling methods, they still lack specific control strategies in the face of the flight capability constraints of the aircraft under different configurations of small disturbances and different performance optimization indicators. Although intelligent control methods can achieve control according to optimization goals under constraints through neural networks, reinforcement learning and other means, intelligent control methods often rely on massive flight data and related learning strategies lack theoretical guarantees.

[0004] Therefore, designing a robust optimization control strategy for the aircraft's conformational switching system to achieve the optimization control task of a specific flight target under multiple constrained small disturbances is one of the current difficulties and urgent problems to be solved. It is necessary to propose a control method for disturbed conformational aircraft based on switching model prediction. Summary of the invention

[0005] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art, the present invention proposes a control method for a disturbed allostatic aircraft based on switching model prediction, to solve the problem that the control method of the existing disturbed aircraft allostatic switching system lacks the ability to process constraints under small disturbances without relying on flight data to achieve specific optimization goals.

[0006] The technical solution of the present invention is:

[0007] A method for controlling a disturbed allostatic flying vehicle based on switching model prediction, the steps are as follows:

[0008] (1) determining a linearized switching dynamics model of the disturbed aircraft under different configurations and determining variable constraints under each configuration; the variable constraints include disturbance variable constraints, flight state constraints, and control surface saturation constraints;

[0009] (2) constructing a switching disturbance invariant set for all configurations of the aircraft according to the disturbance variable constraints, and obtaining the switching configuration-dependent residence time constraints of different configuration subsystems; the different configuration subsystems refer to the aircraft dynamics model under different configuration variables;

[0010] (3) calculating a flight state and control surface saturation constraint tightening set under different aircraft configurations based on the flight state constraint and control surface saturation constraint determined in step (1) and the switching interference invariant set determined in step (2);

[0011] (4) Design the optimal control objective function under different aircraft configurations;

[0012] (5) establishing a switching model predictive control problem for each configuration of the aircraft for the nominal system of the transformable aircraft according to the constraint tightening set and the optimization control objective function;

[0013] The nominal system of the allosteric aircraft refers to a linearized switching dynamics model without disturbance;

[0014] (6) for the switching model predictive control problem described in step (5), determining a configuration-dependent residence time feasibility condition for staying in a certain configuration;

[0015] (7) for the switching model predictive control problem described in step (5), determining a configuration-dependent residence time stability condition for staying in a certain configuration;

[0016] (8) determining a maximum lower bound constraint on the configuration-dependent residence time according to the switching configuration-dependent residence time constraint determined in step (2), the configuration-dependent residence time feasibility condition determined in step (6), and the configuration-dependent residence time stability condition determined in step (7);

[0017] (9) Based on the maximum lower bound constraint, according to the aircraft configuration at each sampling moment, solve the switching model predictive control problem under the corresponding configuration in step (5) to obtain the optimal control surface instruction at the current moment; repeat step (5) to obtain the optimal control surface instruction at the next moment.

[0018] Furthermore, the linearized switching dynamics model of the disturbed aircraft under different configurations in step (1) is specifically:

[0019] After linearization, the state space equation of the longitudinal vehicle's conformational switching is:

[0020]

[0021] Where Δα k is the angle of attack deviation at time k, Δω k is the pitch angular velocity deviation at time k, Δδ k is the control input of the rudder deflection at time k; the allosteric switching law σ(k) is a piecewise constant function of time k, Σ is the set of all possible switching configurations of the aircraft, Σ: = {1,2,…,m,…,M}, m represents the code of the dynamic system of the aircraft with different configurations; M represents the number of configurations that the aircraft can switch, A σ(k) represents the system matrix under different allosteric switching laws, B σ(k) represents the input matrix under different allosteric switching laws, d k Represents the disturbance variable of the aircraft;

[0022] For simplicity, let Δq k =[Δα k Δω k ] T is the state of the aircraft at time k;

[0023] Determine the aircraft state, control surface offset and disturbance variable constraint set under different configurations:

[0024] Δq k ∈Θ σ(k) ,Δδ k ∈U σ(k) ,d k ∈D

[0025] Where Θ σ(k) , U σ(k) are the constraint sets under different configurations of aircraft state and control surface deflection, and D is the disturbance variable constraint set.

[0026] Furthermore, in step (2), a switching interference invariant set for all configurations of the aircraft is constructed, specifically:

[0027] Step (2a): Assume j = 1 and calculate v = τ for each subsystem with different configurations m ,τ m +1,…,2τ m -1,m∈Σ step reachable set

[0028]

[0029] In the formula, τ m Indicates the switching configuration dependence dwell time under different configuration aircraft dynamic system codes, K m , m∈Σ is the configuration-dependent feedback control gain, Am Represents the system matrix under the code of the aircraft dynamics system of different configurations, B m represents the input matrix under the code of the aircraft dynamics system of different configurations, O 0 = {0} and the zero-step reachable set is j is the number of algorithm iterations, j is a positive integer, starting from j=1;

[0030] Step (2b): Calculate the convex hull of the v-step reachable set of each subsystem with different configurations

[0031] Ω j =co{O j}

[0032]

[0033] co means to find the convex hull operation;

[0034] Step (2c): If Ω j =Ω j-1 If it is not satisfied, set j = j + 1, that is, the value of j increases by 1, and repeat steps (2a) to (2b) until Ω j =Ω j-1 , obtain the switching interference invariant set Ω ∞ ;

[0035] The minimum configuration for the existence of a switching interference invariant set depends on the residence time τ m ″ is the switching configuration obtained that depends on the dwell time constraint.

[0036] Furthermore, the calculation method of the flight state and control surface saturation constraint set under different aircraft configurations in step (3) is as follows:

[0037]

[0038] In the formula are the constraints set of the aircraft state and the control surface deflection under different configurations after compression; Θ m , U m They are respectively the aircraft state and control surface deflection constraint sets under the aircraft dynamic system codes of different configurations.

[0039] Furthermore, the optimization control objective function designed in step (4) is

[0040]

[0041] Where Δq k+i|k =[Δα k+i|k Δω k+i|k ] T , Δδ k+i|k are respectively the state of the transformable aircraft and the deflection of the control surface at time k+i predicted at time k, Nm is the length of the predicted time domain under different configurations, Q m With R m They are respectively the weight matrix of the aircraft state variables and the positive definite matrix of the control surface deflection weight under different configurations; is the terminal cost function, where P m is the terminal state weight positive definite matrix.

[0042] Furthermore, step (5) establishes the switching model predictive control problem for each configuration of the aircraft for the nominal system of the variable configuration aircraft as follows:

[0043]

[0044] In the formula is the nominal aircraft conformational switching system, where is the nominal rudder deflection, For the nominal flight state,

[0045] is the optimal rudder deflection sequence predicted at time k, is the initial value of the state, m It is the terminal constraint set under different configurations of the aircraft.

[0046] Furthermore, the configuration-dependent residence time feasibility condition for staying in a certain configuration in step (6) is:

[0047]

[0048] Where τ m The minimum value of the configuration-dependent residence time for the aircraft to meet the feasibility conditions, the initial feasible domain that makes the switching model predictive control problem continuously solvable under the m configuration is

[0049]

[0050] Furthermore, the configuration-dependent residence time stability condition for a certain configuration in step (7) is:

[0051]

[0052] Where τ m ′ is the minimum configuration-dependent residence time for the aircraft to meet the feasibility conditions, E ∞ To shrink the terminal set, the shrink terminal set is calculated as follows:

[0053] Step (7a): Define the v-step reverse reachable set of set Z as:

[0054] Step (7b): Calculate v = τ for each subsystem with different configurations m ,τm +1,…,2τ m -1,m∈Σ steps of intersection of reverse reachable sets

[0055]

[0056] Where λ∈(0,1) is the shrinkage factor, Ε 0 =∩ m∈Σ Ξ m , j is a positive integer representing the number of algorithm iterations;

[0057] Step (7c): Calculate E j =Ε j-1 ∩Λ;

[0058] Step (7d): If E j =Ε j-1 If not satisfied, j=j+1, repeat steps (7a) to (7b) until E j =Ε j-1 , obtain the contraction terminal set E ∞ .

[0059] Furthermore, step (8) determines the maximum lower bound constraint of the configuration-dependent residence time as:

[0060] τ m ≥τ m ,τ m ≥τ m ′ and τ m ≥τ m ″,m∈Σ

[0061] Where τ m It is the maximum lower bound constraint of the configuration-dependent residence time under different configurations of the aircraft.

[0062] Furthermore, in step (9), according to the aircraft configuration at each sampling moment, the model predictive control problem under the corresponding configuration in step (5) is solved to obtain the optimal control surface instruction at the current moment Specifically

[0063]

[0064] In the formula The optimal control sequence of the rudder obtained by solving the model predictive control problem under the corresponding configuration in step (5) at time k is Repeat step (5) to obtain the optimal control surface command at the next moment.

[0065] The advantages of the present invention compared with the prior art are:

[0066] 1. Compared with the non-optimized traditional control method, the model predictive control method for the variable configuration switching system of the disturbed aircraft proposed in the present invention can design the optimization objective function according to the heterogeneous optimization indicators under different configurations, and realize the optimal control of the control surface for different configurations. At the same time, compared with the traditional control method, the control method of the present invention can effectively handle the flight state and the control surface deflection constraints, and realize the optimal control of the control surface under the complex constraints of the variable configuration aircraft.

[0067] 2. Under bounded small disturbances, the control method of the present invention can realize small disturbance robust model predictive control under switching of subsystems of different configurations of the aircraft, compared with the model predictive robust control method without dynamic model switching, while ensuring the feasibility of the switching model prediction algorithm and the stability of the closed-loop system. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 is a flow chart of the method of the present invention;

[0069] Figure 2 Schematic diagram of the structure of fixed configuration aircraft and variable configuration aircraft. DETAILED DESCRIPTION

[0070] A variable configuration aircraft can obtain a better aerodynamic shape layout by adjusting the partial or overall shape to adapt to different flight mission requirements. Therefore, compared with fixed configuration aircraft, it has a wider flight capability boundary and has received widespread attention. Figure 2 Shown is a schematic diagram of the structure of a fixed configuration aircraft and a transformable aircraft. However, the traditional non-optimization control method for a disturbed transformable aircraft lacks the ability to generate optimal rudder control instructions according to different performance optimization indicators under flight capability constraints, and cannot guarantee the stability of the closed-loop system under small disturbances. Although intelligent control methods can achieve control according to optimization objectives under constraints through neural networks, reinforcement learning and other means, intelligent control methods often rely on massive flight data and related learning strategies lack theoretical guarantees for feasibility and stability. In response to this problem, the present invention proposes a control method for a disturbed transformable aircraft based on switching model prediction, which can achieve robust optimization control of the transformable aircraft under multiple constraints, and ensure the feasibility of the optimization control algorithm and the stability of the closed-loop system, which is helpful to improve the flight performance of the transformable aircraft.

[0071] like Figure 1 As shown, the present invention designs a control method for a disturbed allostat based on switching model prediction, which realizes robust optimization control of the allostat under small disturbance and multiple constraints. In theory, sufficient conditions are proposed to ensure the feasibility of the optimization control algorithm and the stability of the closed-loop system. The specific steps are as follows:

[0072] (1) The linearized switching dynamics model of the disturbed aircraft under different discrete configurations is given and the variable constraints under each configuration are determined. The variable constraints include disturbance variable constraints, flight state constraints, and control surface saturation constraints;

[0073] The optimization control method of the present invention is aimed at the longitudinal dynamics model of a subsonic small perturbation allostat aircraft. The dynamics model in step (1) is the longitudinal dynamics model of the allostat aircraft based on the small perturbation linearization method, and the longitudinal dynamics model ignores the yaw motion of the aircraft. The system matrix A of different configurations in the linearized model m With B m ,m∈Σ can be obtained by taking the partial derivative of the Jacobian matrix in the equilibrium flight state.

[0074] The linearized switching dynamics model of the disturbed aircraft under different configurations is specifically:

[0075] After linearization, the state space equation of the longitudinal vehicle's conformational switching is:

[0076]

[0077] Where Δα k is the angle of attack deviation at time k, Δω k is the pitch angular velocity deviation at time k, Δδ k is the control input of the rudder deflection at time k; the allosteric switching law σ(k) is a piecewise constant function of time k, Σ is the set of all possible switching configurations of the aircraft, Σ: = {1,2,…,m,…,M}, m represents the code of the dynamic system of the aircraft with different configurations; M represents the number of configurations that the aircraft can switch, A σ(k) represents the system matrix under different allosteric switching laws, B σ(k) represents the input matrix under different allosteric switching laws, d k represents the disturbance variable of the aircraft; for simplicity, let Δq k =[Δα k Δω k ] T is the state of the aircraft at time k.

[0078] Determine the aircraft state, control surface offset and disturbance variable constraint set under different configurations:

[0079] Δq k ∈Θ σ(k) ,Δδ k ∈U σ(k) ,d k ∈D

[0080] Where Θ σ(k) , U σ(k) are the constraint sets under different configurations of aircraft state and control surface deflection, and D is the disturbance variable constraint set.

[0081] (2) According to the disturbance variable constraints, the switching disturbance invariant set for all configurations of the aircraft is constructed to obtain the switching configuration-dependent residence time constraints of different configuration subsystems.

[0082] The different configuration subsystems refer to: aircraft dynamics models under different variable configurations (such as different spans, chord lengths, etc.).

[0083] The purpose of calculating the switching disturbance invariant set is to determine the maximum impact of the accumulation of disturbance variables on the aircraft switching conformational system, and to prevent the constraints of each variable from being violated due to the existence of disturbance variables. The specific calculation method is as follows:

[0084] Construct the switching interference invariant set for all configurations of the aircraft, specifically:

[0085] Step (2a): Assume j = 1 and calculate v = τ for each subsystem with different configurations m ,τ m +1,…,2τ m -1,m∈Σ step reachable set

[0086]

[0087] In the formula, τ m Indicates the switching configuration dependence dwell time under different configuration aircraft dynamic system codes, K m , m∈Σ is the configuration-dependent feedback control gain, A m Represents the system matrix under the code of the aircraft dynamics system of different configurations, B m represents the input matrix under the code of the aircraft dynamics system of different configurations, O 0 = {0} and the zero-step reachable set is j is the number of algorithm iterations, j is a positive integer, starting from j=1;

[0088] Step (2b): Calculate the convex hull of the v-step reachable set of each subsystem with different configurations

[0089] Ω j =co{O j}

[0090]

[0091] co represents the convex hull operation.

[0092] Step (2c): If Ω j =Ω j-1 If it is not satisfied, set j = j + 1, that is, the value of j increases by 1, and repeat steps (2a) to (2b) until Ω j =Ω j-1, obtain the switching interference invariant set Ω ∞ ;

[0093] The minimum configuration for the existence of a switching interference invariant set depends on the residence time τ m ″ is the switching configuration obtained that depends on the dwell time constraint.

[0094] The v-step reachable set calculated in step (2a) is the error system constructed by the actual dynamic model of the disturbed allosteric vehicle (the model in step (1)) and the nominal dynamic model ignoring the disturbance variables (the model in step (5)).

[0095]

[0096] The flight state error

[0097] The number of steps to calculate the reachable set in step (2a) is designed to be v = τ m ,τ m +1,…,2τ m -1,m∈Σ. The reason is that the aircraft configuration switching sequence cannot be enumerated exhaustively, but can be divided into sub-sequence units that satisfy the configuration-dependent residence time constraints. Therefore, only the sub-switching sequence units are considered when calculating the reachable set.

[0098] The reason why the convex hull of the v-step reachable set of each subsystem with different configurations is calculated in step (2b) instead of the union is that the union calculation cannot guarantee the convexity of the set. The calculation of the convex hull can guarantee the convexity of the interference invariant set, which facilitates the subsequent design of predictive controllers based on convex optimization.

[0099] (3) Calculate the flight state and control surface saturation constraint tightening set under different aircraft configurations.

[0100] To facilitate the switching model predictive control problem in the subsequent nominal system design step (5), based on the switching disturbance invariant set Ω calculated in step (2), ∞ , the Pontryagin set subtraction is used to calculate the different flight states and the control surface constraint set. Configuration dependent feedback gain K m It can be obtained by solving the Riccati equation of the linear quadratic regulator (LQR).

[0101] The calculation method is as follows:

[0102]

[0103] In the formula are the constraints set of the aircraft state and the control surface deflection under different configurations after compression; Θ m , U mThey are respectively the aircraft state and control surface deflection constraint sets under the aircraft dynamic system codes of different configurations.

[0104] (4) Design the optimal control objective function under different aircraft configurations.

[0105] Different from the single objective function of traditional model predictive controller design, the optimization objective function in step (4) can be designed according to different aircraft configurations, including the following parameters: prediction time domain N m , positive definite weight matrix Q m , R m and P m , terminal constraints m . Prediction time domain N m , the weight matrix Q m , R m It can be designed according to the predictive control performance requirements of the aircraft, P m It can be obtained by solving the LQR Riccati equation.

[0106] The designed optimization control objective function is

[0107]

[0108] Where Δq k+i|k =[Δα k+i|k Δω k+i|k ] T , Δδ k+i|k are respectively the state of the transformable aircraft and the deflection of the control surface at time k+i predicted at time k, N m is the length of the predicted time domain under different configurations, Q m With R m They are respectively the weight matrix of the aircraft state variables and the positive definite matrix of the control surface deflection weight under different configurations;

[0109] is the terminal cost function, where P m is the terminal state weight positive definite matrix.

[0110] (5) Based on the obtained constraint tightening set and objective function, a switching model predictive control problem is established for each configuration of the aircraft for the nominal system of the transformable aircraft.

[0111] The dynamic model of the allostatic aircraft designed in the switching model predictive control problem in step (5) is the model of the nominal dynamic system of the aircraft. The error accumulation effect of the disturbance variable in step (1) on the flight of the allostatic aircraft has been eliminated by the tightening constraint set design in step (3). Therefore, solving the optimization control problem (5) for the nominal dynamic system of the aircraft avoids the constraint violation problem caused by the accumulation error of the disturbance variable.

[0112] The switching model predictive control problem is:

[0113]

[0114] In the formula is the nominal aircraft conformational switching system, where is the nominal rudder deflection, For the nominal flight state,

[0115] is the optimal rudder deflection sequence predicted at time k, is the initial value of the state, m It is the terminal constraint set under different configurations of the aircraft.

[0116] (6) Give sufficient conditions to ensure that (5) the configuration switching moment of the aircraft in the optimization control problem and the feasibility of staying in a certain configuration depend on the residence time constraint.

[0117] In step (6), the sufficient condition for ensuring the optimal control problem (5) that the aircraft configuration switching time and the feasible configuration dependence residence time constraint of staying in a certain configuration is:

[0118]

[0119] The right side of the formula ∩ n∈Σ,n≠m F n is the intersection of the feasible domains of the initial values ​​of the configurations of other configurations of the aircraft, except for the configuration subsystem m. This design can ensure the solvability of the optimization problem (5) after the configuration m is switched to other configurations. τ m That is, the reachable set that can make the configuration m feasible reaches ∩ n∈Σ,n≠m F n The minimum number of steps. The initial feasible region that makes the switching model predictive control problem continuously solvable under the m configuration is

[0120]

[0121] (7) Give sufficient conditions to ensure that the configuration switching moment of the aircraft in the optimization control problem (5) and the stability configuration dependence of staying in a certain configuration are constrained by the residence time.

[0122] In step (7), the sufficient condition for ensuring the stability of the aircraft configuration switching and the configuration-dependent residence time constraint of the optimization control problem (5) is:

[0123]

[0124] Where the contraction terminal set E ∞ For 0 =∩ m∈Σ Ξ m Obtained through reverse reachable set calculation.

[0125] The contraction terminal set is calculated as follows:

[0126] Step (7a): Define the v-step reverse reachable set of set Z as:

[0127] Step (7b): Calculate v = τ for each subsystem with different configurations m ,τ m +1,…,2τ m -1,m∈Σ steps of intersection of reverse reachable sets

[0128]

[0129] Where λ∈(0,1) is the shrinkage factor, Ε 0 =∩ m∈Σ Ξ m , j is a positive integer representing the number of algorithm iterations;

[0130] In the process of calculating the intersection of the reverse reachable sets, the switching sequence units of all configurations under the dependent residence time are considered, and the introduction of the contraction factor λ∈(0,1) can ensure that the closed-loop state trajectory of the aircraft enters E ∞ After the convergence. Aircraft status trajectory enters E ∞ After that, the state feedback control law will be adopted, so the closed-loop system state is calculated to enter the state for The residence time constraint τ of the contraction terminal set m ′ can ensure the convergence of the state flight trajectory.

[0131] Step (7c): Calculate E j =Ε j-1 ∩Λ;

[0132] Step (7d): If E j =Ε j-1 If not satisfied, j=j+1, repeat steps (7a) to (7b) until E j =Ε j-1 , obtain the contraction terminal set E ∞ .

[0133] (8) Obtain the maximum lower bound constraints on the existence of the switching disturbance invariant set, the feasibility of the predictive control algorithm and the stability of the closed-loop system configuration dependence on the residence time in (2), (6) and (7).

[0134] In step (8), in addition to verifying whether the feasibility and stability constraints in steps (6) and (7) are satisfied, the constraint on the existence of the switching interference invariant set in (2) should also be verified.

[0135] The maximum lower bound constraint of the configuration-dependent residence time is:

[0136] τ m ≥τ m ,τ m ≥τ m ′ and τ m ≥τ m ″,m∈Σ

[0137] Where τ m It is the maximum lower bound constraint of the configuration-dependent residence time under different configurations of the aircraft.

[0138] (9) According to the aircraft configuration at each sampling moment, solve the model predictive control problem under the corresponding configuration in (5) to obtain the optimal control surface command at the current moment. Repeat step (5) to obtain the optimal control surface command at the next moment.

[0139] Get the optimal rudder command at the current moment Specifically

[0140]

[0141] In the formula The optimal control sequence of the rudder obtained by solving the model predictive control problem under the corresponding configuration in step (5) at time k is The first element of

[0142] Step (9) is the same as the traditional model predictive control method, solving (5) to obtain the optimal control sequence of the control surface The first element in That is, the optimal control input of the variable-configuration aircraft is predicted at time k. Repeating (5) can obtain the optimal control surface command at the next moment.

[0143] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for controlling a disturbed allosteric aircraft based on switching model prediction, characterized in that Here are the steps: (1) determining a linearized switching dynamics model of the disturbed aircraft under different configurations and determining variable constraints under each configuration; the variable constraints include disturbance variable constraints, flight state constraints, and control surface saturation constraints; (2) constructing a switching disturbance invariant set for all configurations of the aircraft according to the disturbance variable constraints, and obtaining the switching configuration-dependent residence time constraints of different configuration subsystems; the different configuration subsystems refer to the aircraft dynamics model under different configuration variables; (3) calculating a flight state and control surface saturation constraint tightening set under different aircraft configurations based on the flight state constraint and control surface saturation constraint determined in step (1) and the switching interference invariant set determined in step (2); (4) Design the optimal control objective function under different aircraft configurations; (5) establishing a switching model predictive control problem for each configuration of the aircraft for the nominal system of the transformable aircraft according to the constraint tightening set and the optimization control objective function; The nominal system of the allosteric aircraft refers to a linearized switching dynamics model without disturbance; (6) for the switching model predictive control problem described in step (5), determining a configuration-dependent residence time feasibility condition for staying in a certain configuration; (7) for the switching model predictive control problem described in step (5), determining a configuration-dependent residence time stability condition for staying in a certain configuration; (8) determining a maximum lower bound constraint on the configuration-dependent residence time according to the switching configuration-dependent residence time constraint determined in step (2), the configuration-dependent residence time feasibility condition determined in step (6), and the configuration-dependent residence time stability condition determined in step (7); (9) Based on the maximum lower bound constraint, according to the aircraft configuration at each sampling moment, solve the switching model predictive control problem under the corresponding configuration in step (5) to obtain the optimal control surface instruction at the current moment; repeat step (5) to obtain the optimal control surface instruction at the next moment.

2. A method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 1, characterized in that: The linearized switching dynamics model of the disturbed aircraft under different configurations in step (1) is specifically: After linearization, the state space equation of the longitudinal vehicle's conformational switching is: Where Δα k is the angle of attack deviation at time k, Δω k is the pitch angular velocity deviation at time k, Δδ k is the control input of the rudder deflection at time k; the allosteric switching law σ(k) is a piecewise constant function of time k, Σ is the set of all possible switching configurations of the aircraft, Σ: = {1,2,…,m,…,M}, m represents the code of the dynamic system of the aircraft with different configurations; M represents the number of configurations that the aircraft can switch, A σ(k) represents the system matrix under different allosteric switching laws, B σ(k) represents the input matrix under different allosteric switching laws, d k Represents the disturbance variable of the aircraft; For simplicity, let Δq k =[Δα k Δω k ] T is the state of the aircraft at time k; Determine the aircraft state, control surface offset and disturbance variable constraint set under different configurations: Δq k ∈Θ σ(k) ,Dd k ∈U σ(k) ,d k ∈D Where Θ σ(k) , U σ(k) are the constraint sets under different configurations of aircraft state and control surface deflection, and D is the disturbance variable constraint set.

3. A method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 2, characterized in that: In step (2), a switching interference invariant set for all configurations of the aircraft is constructed, specifically: Step (2a): Assume j = 1, and calculate v = τ for each subsystem with different configurations m ,τ m +1,…,2τ m -1,m∈Σ step reachable set In the formula, τ m Indicates the switching configuration dependence dwell time under different configuration aircraft dynamic system codes, K m , m∈Σ is the configuration-dependent feedback control gain, A m Represents the system matrix under the code of the aircraft dynamics system of different configurations, B m Represents the input matrix under the code of the aircraft dynamic system with different configurations, O0={0} and the zero-step reachable set is j is the number of algorithm iterations, j is a positive integer, starting from j=1; Step (2b): Calculate the convex hull of the v-step reachable set of each subsystem with different configurations Oh j =what{O j } co means to find the convex hull operation; Step (2c): If Ω j =Ω j-1 If it is not satisfied, set j = j + 1, that is, the value of j increases by 1, and repeat steps (2a) to (2b) until Ω j =Ω j-1 , obtain the switching interference invariant set Ω ∞ ; The minimum configuration for the existence of a switching interference invariant set depends on the residence time τ m ″ is the switching configuration obtained that depends on the dwell time constraint.

4. A method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 3, characterized in that: The calculation method of the flight state and control surface saturation constraint set under different aircraft configurations in step (3) is as follows: In the formula are the constraints set of the aircraft state and the control surface deflection under different configurations after compression; Θ m , U m They are respectively the aircraft state and control surface deflection constraint sets under the aircraft dynamic system codes of different configurations.

5. The method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 1, characterized in that: The optimization control objective function designed in step (4) is: Where Δq k+i|k =[Δα k+i|k Δω k+i|k ] T , Δδ k+i|k are respectively the state of the transformable aircraft and the deflection of the control surface at time k+i predicted at time k, N m is the length of the predicted time domain under different configurations, Q m With R m They are respectively the weight matrix of the aircraft state variables and the positive definite matrix of the control surface deflection weight under different configurations; is the terminal cost function, where P m is the terminal state weight positive definite matrix.

6. A method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 5, characterized in that: Step (5) establishes the switching model predictive control problem for each configuration of the aircraft for the nominal system of the variable configuration aircraft as follows: In the formula is the nominal aircraft conformational switching system, where is the nominal rudder deflection, For the nominal flight state, is the optimal rudder deflection sequence predicted at time k, is the initial value of the state, m It is the terminal constraint set under different configurations of the aircraft.

7. A method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 6, characterized in that: The feasibility condition for the configuration-dependent residence time of staying in a certain configuration in step (6) is: Where τ m The minimum value of the configuration-dependent residence time for the aircraft to meet the feasibility conditions, the initial feasible domain that makes the switching model predictive control problem continuously solvable under the m configuration is 8. The method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 7, characterized in that: The configuration-dependent residence time stability condition for a certain configuration in step (7) is: Where τ m ′ is the minimum configuration-dependent residence time for the aircraft to meet the feasibility conditions, E ∞ To shrink the terminal set, the shrink terminal set is calculated as follows: Step (7a): Define the v-step reverse reachable set of set Z as: Step (7b): Calculate v = τ for each subsystem with different configurations m ,τ m +1,…,2τ m -1,m∈Σ steps of intersection of reverse reachable sets Where λ∈(0,1) is the shrinkage factor, Ε0=∩ m∈Σ Ξ m , j is a positive integer representing the number of algorithm iterations; Step (7c): Calculate E j =Ε j-1 ∩Λ; Step (7d): If E j =Ε j-1 If not satisfied, j=j+1, repeat steps (7a) to (7b) until E j =Ε j-1 , obtain the contraction terminal set E ∞ .

9. A method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 8, characterized in that: Step (8) determines the maximum lower bound constraint of the configuration-dependent residence time as: τ m ≥τ m ,τ m ≥τ m ′ and τ m ≥τ m ″, m ∈ Σ Where τ m It is the maximum lower bound constraint of the configuration-dependent residence time under different configurations of the aircraft.

10. The method for controlling a disturbed allostatic aircraft based on switching model prediction according to claim 1, characterized in that: In step (9), according to the aircraft configuration at each sampling moment, the model predictive control problem under the corresponding configuration in step (5) is solved to obtain the optimal control surface instruction at the current moment. Specifically In the formula The optimal control sequence of the rudder obtained by solving the model predictive control problem under the corresponding configuration in step (5) at time k is Repeat step (5) to obtain the optimal control surface command at the next moment.

Images