An Optimization Method for Active Disturbance Rejection Composite Control of High-Speed Aircraft Based on Adaptive Dynamic Programming

Abstract

The present invention discloses an optimization method for active disturbance rejection composite control of a high-speed aircraft based on adaptive dynamic programming, belonging to the field of attitude control of high-speed aircraft. The composite controller designed by this method includes two parts: a steady-state controller and a dynamic controller. First, a second-order tracking differentiator and a neural network disturbance observer are designed based on the nominal model, which are used to obtain differential commands and compensate for uncertain disturbances respectively. Then, a sliding mode surface is constructed according to the tracking error and its derivative, and a steady-state controller is designed to ensure the balance stability of the attitude system. Finally, a dynamic controller is designed based on the action-dependent adaptive dynamic programming method. This controller consists of an action network and a critic network, which is used to approximate the optimal solution of the Hamilton equation, can ensure that the system can quickly track dynamic commands and adaptively optimize the parameters of the neural network controller online, and effectively improve the tracking performance of the attitude system.

Description

Technical Field

[0001] The present invention belongs to the field of attitude control of high-speed aircraft, and more specifically, relates to an active disturbance rejection composite control optimization method for high-speed aircraft based on adaptive dynamic programming. Background Art

[0002] As a new type of aircraft with high speed, large range, and fast response, high-speed aircraft can not only cruise at high speed within the atmosphere but also cross the atmosphere as a space transportation vehicle, and have broad application prospects in both military and civilian fields. The design of the attitude controller is very important in the entire flight control loop, which is an important guarantee for the flight safety and stability of high-speed aircraft, and is also a key link for responding to guidance commands and improving flight performance. However, due to the advanced aerodynamic configuration of high-speed aircraft, their dynamic and kinematic systems exhibit complex non-linearity, which poses new challenges to the optimal design of the controller.

[0003] Existing control optimization methods are usually effectively applied to linear systems with accurate models and constant parameters. For the complex flight environment with disturbances and highly non-linear dynamic models of high-speed aircraft, traditional control optimization methods need to first linearize the model near the operating point and then design the optimal controller for the attitude system. However, high-speed aircraft usually need to perform aerospace cross-domain flight missions in a large airspace and wide speed range. Linearization expansion from a single point cannot completely solve the optimal control problem of high-speed aircraft. Therefore, in engineering applications, traditional control optimization methods are usually combined with methods such as gain pre-scheduling and controller switching, which results in the need to consume a great deal of time and labor costs to produce a gain scheduling table within a large flight envelope. Considering the problem of the inconsistency of aerodynamic parameters between space and ground, if the aerodynamic data relied on for offline design deviates from the actual flight, it will seriously affect the control quality and flight safety. In addition, the modeling error and disturbance expanded at the operating point may cause mutations in the switching mechanism, resulting in flight control instability. Therefore, it is necessary to develop a robust active disturbance rejection composite optimization control scheme. Summary of the Invention

[0004] Aiming at the above defects or improvement requirements of the prior art, the present invention provides an active disturbance rejection composite control optimization method for high-speed aircraft based on adaptive dynamic programming. For the attitude tracking control process of the pitch channel of high-speed aircraft, a composite controller based on adaptive dynamic programming and active disturbance rejection method is constructed to realize online parameter optimization, achieving the effect of improving the control performance of the aircraft.

[0005] To achieve the above object, according to the first aspect of the present invention, there is provided an active disturbance rejection composite control optimization method for high-speed aircraft based on adaptive dynamic programming, including:

[0006] S1. Based on the second-order attitude control system of a high-speed aircraft, construct a second-order tracking differentiator to calculate according to x d1 Calculate and x d2 , construct a disturbance observer to obtain the observed value of the disturbance of the second-order attitude control system

[0007] where x d1 is the reference command of the aircraft angle of attack α, and x d2 is the reference command of the first derivative of α , and is the first derivative of x d2 ;

[0008] S2. According to and construct a steady-state controller

[0009] where ρ is the atmospheric density, V is the velocity, and c respectively represent the wing reference area and the mean aerodynamic chord, I yy is the moment of inertia, e2 is the derivative of the angle of attack tracking error, λ > 0, is the second derivative of the flight path angle, and C M,α , C M,q , are all aerodynamic parameters;

[0010] S3. Construct a dynamic controller based on an action neural network According to the composite controller control the attitude of the high-speed aircraft, where S is the sliding mode surface of the second-order attitude control system;

[0011] where φ a (S) are respectively the weights and activation functions of the action neural network.

[0012] According to the second aspect of the present invention, there is provided an electronic device, including: a computer-readable storage medium and a processor;

[0013] The computer-readable storage medium is used to store executable instructions;

[0014] The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the method described in the first aspect.

[0015] According to a third aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to execute the method described in the first aspect.

[0016] Generally speaking, compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects:

[0017] The method provided by the present invention realizes the active disturbance rejection composite control of a high-speed aircraft based on adaptive dynamic programming by designing a composite controller. The designed composite controller includes two parts: a steady-state controller and a dynamic controller. First, a second-order tracking differentiator and a neural network disturbance observer are designed based on a nominal model to obtain differential commands and compensate for uncertain disturbances respectively. Then, a sliding mode surface is constructed according to the tracking error and its derivative, and a steady-state controller is designed to ensure the balance stability of the attitude system. Finally, a dynamic controller is designed based on the action-dependent adaptive dynamic programming method. This controller consists of an action network and an evaluation network, and is used to approximate the optimal solution of the Hamilton equation, which can ensure that the system can quickly track dynamic commands and adaptively optimize the parameters of the neural network controller online, effectively improving the tracking performance of the attitude system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a block diagram of the active disturbance rejection composite control of a high-speed aircraft based on adaptive dynamic programming provided by an embodiment of the present invention.

[0019] Figure 2 It is a differential effect diagram of the tracking differentiator processing the tracking command provided by an embodiment of the present invention.

[0020] Figure 3 It is a curve diagram of the angle of attack command tracking control effect provided by an embodiment of the present invention.

[0021] Figure 4 It is a curve diagram of the tracking error response provided by an embodiment of the present invention.

[0022] Figure 5 It is a curve diagram of the control input rudder deflection angle response provided by an embodiment of the present invention.

[0023] Figure 6 It is a curve diagram of the disturbance observer compensation response provided by an embodiment of the present invention.

[0024] Figure 7 It is a curve diagram of the adaptive dynamic programming controller response provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] To make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to 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 used 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.

[0026] An embodiment of the present invention provides an active disturbance rejection composite control optimization method for a high-speed aircraft based on adaptive dynamic programming, as Figure 1 shown, including:

[0027] S1. Based on the second-order attitude control system of the high-speed aircraft, a second-order tracking differentiator is constructed to calculate d1 and based on x d2 , and a disturbance observer is constructed to obtain the observed value of the disturbance of the second-order attitude control system where x d1 is the reference command of the aircraft angle of attack α, x d2 is the first derivative of α of the reference command, is the first derivative of x d2 .

[0028] In step S1, a tracking differentiator and a disturbance observer are constructed based on the second-order attitude control system of the high-speed aircraft (S1.1 - S1.3).

[0029] S1.1 Establishment of the high-speed aircraft attitude control system model

[0030] Considering the longitudinal second-order attitude dynamics of the unpowered aircraft, the second-order attitude control system of the high-speed aircraft is constructed as follows

[0031]

[0032] where the system states α, are the angle of attack and the derivative of the angle of attack respectively, the control input u of the system is u = δ e is the rudder deflection angle. Δf(x) is the disturbance of the second-order attitude control system, that is, the disturbance related to the aircraft state,

[0033]

[0034] where represents the dynamic pressure, ρ is the atmospheric density, V is the speed, and c respectively represent the wing reference area and the mean aerodynamic chord, I yy is the moment of inertia, is the second derivative of the flight path angle. q is the pitch angular velocity of the aircraft are aerodynamic parameters.

[0035] S1.2 Construct a second-order differential tracker

[0036] For the reference command x of x1 d1 , construct the second-order command differential for the aircraft attitude dynamics system as follows:

[0037]

[0038] where and are the command differentials generated by the differential tracker. The differentiator parameters R, a1, a2, b1, b2 > 0, and arsinh is the inverse hyperbolic sine function.

[0039] It can be understood that a second-order differential tracker can also be constructed based on a linear function and an inertial filter. The specific expression of the above second-order differential tracker is only an example, and the present invention does not make a unique limitation on this.

[0040] S1.3 Construct a disturbance observer based on a neural network

[0041] For the disturbance Δf(x) of the aircraft attitude system, construct a disturbance observer for the aircraft attitude dynamics system as follows:

[0042]

[0043] where and are the observed values of the system state, and e o is the observation error. The observer parameters β, Λ o , K o > 0. is the neural network weight, and the network activation function is where and are the activation function parameters. The observed value of the disturbance Δf(x) is

[0044] It can be understood that a disturbance observer can also be constructed based on a linear function and the Luenberger method. The specific expression of the above disturbance observer is only an example, and the present invention does not make a unique limitation on this.

[0045] S2. Construct a steady-state controller according to and where ρ is the atmospheric density, V is the velocity, and c respectively represent the airfoil reference area and the mean aerodynamic chord, I and c respectively represent the airfoil reference area and the mean aerodynamic chord, Iyy is the moment of inertia, and e2 is the tracking error of λ > 0,

[0046] is the second derivative of the flight path inclination angle, and C M,α , C M,q , are all aerodynamic parameters.

[0047] In step S2, a steady-state controller based on the sliding mode surface is designed.

[0048] According to the system second-order dynamics and the differential commands generated by the tracking differentiator, construct the sliding mode surface as follows

[0049]

[0050] where the parameter λ > 0, and e1 = x1 - x d1 is the tracking error of the angle of attack. The dynamics of the sliding mode surface is

[0051]

[0052] where e2 = x2 - x d2 is the tracking error of the derivative of the angle of attack, Compound control strategy where the steady-state controller u s is designed to keep the state always on the sliding mode surface, and the dynamic feedback controller is used to guide the state to the sliding mode surface. The steady-state controller u s is designed as:

[0053]

[0054] where e2 is obtained from the system feedback state and the tracking differentiator, F(x) is the system parameter term, is obtained from the disturbance observer.

[0055] S3. Construct a dynamic controller based on the action neural network According to the compound controller control the attitude of the high-speed aircraft, and S is the sliding mode surface of the second-order attitude control system; where, φ a (S) are the weights and activation functions of the action neural network respectively.

[0056] In step S3, establish a dynamic controller based on action-dependent adaptive dynamic programming.

[0057] Examine the robust optimal control problem of the sliding mode surface system:

[0058]

[0059] Among them, is the cost function of the dynamic control process, is the admissible control law. The quadratic utility function where the optimization parameters Q, R > 0.

[0060] To estimate the Construct an evaluation network:

[0061]

[0062] Among them, are the network weights, and φ c (S) is the neural network activation function, is the observation value constructed by the evaluation network, and its partial derivative with respect to the sliding mode surface is

[0063]

[0064] Among them, is to optimize the neural network control law weights According to the Hamilton equation of the robust optimization problem:

[0065]

[0066] To minimize e c , let the objective function The neural network weights The update law is

[0067]

[0068] Among them, the update parameter β c > 0, is to update the control strategy according to the evaluation result of the cost function and construct a neural network controller:

[0069]

[0070] Among them, are the network weights, and φ a (S) is the neural network activation function. According to the predicted value of the evaluation network Design the training objective Let the objective function The neural network weights The update law is

[0071]

[0072] Among them, the update parameter βa >0, the design of the dynamic controller is now completed.

[0073] from Figures 2 - 7 It can be seen that the method provided by the present invention has the characteristics of fast dynamic response, small static error, and the ability to track a wide range of jump instructions.

[0074] An embodiment of the present invention provides an electronic device, characterized by comprising: a computer-readable storage medium and a processor;

[0075] The computer-readable storage medium is used to store executable instructions;

[0076] The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the method described in the above embodiment.

[0077] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to execute the method described in the above embodiment.

[0078] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A high-speed aircraft active disturbance rejection composite control optimization method based on adaptive dynamic programming, characterized in that: include: S1, based on the second-order attitude control system of high-speed aircraft, build a second-order tracking differentiator to use according to x d1 calculate With x d2 , construct a disturbance observer to obtain the observed value of the disturbance of the second-order attitude control system Among them, x d1 is the reference instruction of the aircraft attack angle α, x d2 is the first derivative of α Reference instructions, For x d2 The first derivative of ; S2, according to and Building a Steady-State Controller in, ρ is the atmospheric density, V is the velocity, and c represent the airfoil reference area and mean aerodynamic chord, respectively. yy is the moment of inertia, e2 is the derivative of the angle of attack tracking error, λ>0, is the second-order derivative of the track inclination, C M,α , C M,q , All are pneumatic parameters; S3, building a dynamic controller based on action neural network According to the composite controller The attitude of the high-speed aircraft is controlled, and S is the sliding surface of the second-order attitude control system; in, φ a (S) are the weights and activation functions of the action neural network respectively; In step S3, the dynamic controller is constructed based on the sliding surface of the second-order attitude control system and is used to guide the system state to the sliding surface, including: A1, constructing the cost function of the dynamic control process based on the sliding surface in, is an admissible control law, and the quadratic utility function The optimization parameters Q and R are both greater than 0, S(t) is the sliding surface of the second-order attitude control system at time t, S(t)=e2(t)+λe1(t), e1(t) is the tracking error of the angle of attack at time t; e2(t) is the derivative of the tracking error of the angle of attack at time t; A2, building an evaluation neural network Used to obtain the observed value of the cost function in, and φ c (S) are the weights and activation functions of the evaluation neural network, The update law is the sliding surface dynamics of the second-order attitude control system at time t, Δf(x) is the disturbance of the second-order attitude control system; A3, based on Update law, construct a dynamic controller based on action neural network in, The update law 2. The method according to claim 1, characterized in that In step S32, The update law is to minimize the training objective function of the evaluation neural network get; in, 3. The method according to claim 1, characterized in that In step S33, The update law is to minimize the training objective function of the action neural network get; in, 4. The method according to claim 1, characterized in that In step S2, the steady-state controller is constructed based on the sliding surface and the sliding surface dynamics, so as to keep the system state on the sliding surface; Wherein, the sliding surface is: S(t)=e2(t)+λe1(t), e1(t) is the tracking error of the angle of attack; e2(t) is the derivative of the tracking error of the angle of attack at time t; The sliding surface dynamics are: Δf(x) is the disturbance of the second-order attitude control system.

5. The method according to claim 1, characterized in that The second-order attitude control system of the high-speed aircraft is: in, α、 are the angle of attack and the derivative of the angle of attack respectively. The control input of the system is u = δ e , δ e is the rudder deflection angle, Δf(x) is the disturbance of the second-order attitude control system.

6. The method according to claim 1, characterized in that The second-order tracking differentiator is: in, For x d1 The first-order derivative of , the differentiator parameters R, a1, a2, b1, b2 are all greater than 0, and arsinh is the inverse hyperbolic sine function.

7. The method according to claim 1, characterized in that The disturbance observer is: in, and are α and Observed value, e o is the observation error, the observer parameters β, Λ o ,K o are greater than 0, is the neural network weight, and the activation function of the neural network is and is the activation function parameter, is the observed value of x, α、 are the angle of attack and its derivative, respectively.

8. An electronic device, characterized in that: include: A computer readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to execute the method according to any one of claims 1 to 7.

Images