Flight control method and system based on nonlinear model

Abstract

The invention discloses a flight control method and system based on a nonlinear model, and relates to the field of unmanned flight control, and the method comprises the steps: obtaining a BP neural network model; when the target unmanned helicopter executes a single flight task, giving an initial state quantity and a future # imgabs0 # step control quantity sequence, and circularly calling a # imgabs1 # single step state quantity prediction link to obtain a future # imgabs2 # step state quantity response; determining a # imgabs3 # step manipulated variable sequence corresponding to the minimum loss function by adopting a differential evolution algorithm, and taking the # imgabs3 # step manipulated variable sequence as an optimal manipulated variable sequence; applying a first-step manipulated variable in the optimal manipulated variable sequence to the target unmanned helicopter to obtain a real state quantity response of the target unmanned helicopter; and updating the initial state quantity by using the real state quantity response, repeating the process, and continuously performing feedback correction on the state quantity predicted by the BP neural network model until the single flight task of the target unmanned helicopter is finished. According to the invention, the accuracy and the operation efficiency of flight control are improved.

Description

Technical Field

[0001] This application relates to the field of unmanned flight control technology, and particularly to a flight control method and system based on a non-linear model. Background Art

[0002] The low-altitude economy and unmanned aerial vehicle technology are current research hotspots, and the autonomous flight control of unmanned helicopters has been developed for many years. At present, most unmanned helicopters mainly rely on model predictive control, and the results of model prediction directly affect their flight states and efficiency.

[0003] The existing models for controlling the flight of unmanned helicopters are divided into two types, namely linear models and non-linear models. Among them, linear models are mainly applied to linear controlled objects, and usually their prediction accuracy is not high. In order to obtain higher-precision prediction results, a more accurate modeling process is required. To further improve the control accuracy of unmanned helicopters, non-linear models can also be used to achieve predictive control, but the non-linear models require a large amount of computation and are often time-consuming. Therefore, it is difficult for the existing methods that rely on model prediction for flight control to achieve a balance between accuracy and efficiency. Summary of the Invention

[0004] The purpose of this application is to provide a flight control method and system based on a non-linear model, which improves the accuracy and efficiency of the flight control of the target unmanned helicopter.

[0005] To achieve the above purpose, the following solutions are provided in this application.

[0006] In the first aspect, this application provides a flight control method based on a non-linear model. The flight control method based on a non-linear model includes: S1. Obtain a BP neural network model; the BP neural network model is used to predict the state quantity at the next discrete moment according to the current state quantity and the current manipulation quantity of the target unmanned helicopter; S2. When the target unmanned helicopter executes a single flight mission, give the initial state quantity and the future step manipulation quantity sequence, and obtain the future step state quantity response by circularly calling times the single-step state quantity prediction link; the single-step state quantity prediction link includes: inputting the current state quantity and the current manipulation quantity into the BP neural network model, and using the state quantity predicted by the BP neural network model as the state quantity at the next discrete moment; where ; S3. Construct a loss function based on the deviation between the future step state quantity response and the desired state quantity, and use the differential evolution algorithm to determine the Step manipulation quantity sequence and use it as the optimal manipulation quantity sequence; S4. Apply the first step manipulation quantity in the optimal manipulation quantity sequence to the target unmanned helicopter to obtain the true state quantity response of the target unmanned helicopter; S5. Update the initial state quantity using the true state quantity response; S6. Repeat steps S2 - S5, continuously perform feedback correction on the state quantity predicted by the BP neural network model until the single flight mission of the target unmanned helicopter ends.

[0007] In a second aspect, the present application also provides a computer system, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the flight control method based on a non - linear model described in the first aspect.

[0008] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application.

[0009] The present application first obtains a BP neural network model and effectively captures the highly non - linear flight dynamics response relationship, enabling the model to effectively predict the state quantity at the next discrete moment based on the current state quantity and current manipulation quantity of the target unmanned helicopter. The acquisition of this model not only avoids redundant iterative calculations but also accurately captures the non - linear relationship between manipulation and response. Secondly, the present application also realizes the non - linear control of the target unmanned helicopter. That is, in order to further reduce the difference between the state quantity predicted by the model and the true state quantity response of the target unmanned helicopter, the present application uses a differential evolution algorithm to determine the step manipulation quantity sequence when the difference is the smallest, avoiding the gradient solution problem in conventional non - linear optimization problems. This algorithm does not require constructing the gradient descent of the state quantity and has higher operation efficiency. In addition, the present application only applies the first step manipulation quantity in the step manipulation quantity sequence to the target unmanned helicopter. This approach can ensure that the state quantity predicted by the model is corrected in a timely manner according to the true state quantity response of the target unmanned helicopter, ensuring the accuracy of the model prediction result. Therefore, the present application improves the accuracy and efficiency of the flight control of the target unmanned helicopter. Description of the Drawings

[0010] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0011] Figure 1 It is a flowchart of the flight control method based on a non - linear model provided by an embodiment of the present application.

[0012] Figure 2 This is the control logic diagram of the flight control method based on the nonlinear model provided by the embodiment of the present application.

[0013] Figure 3 This is the basic architecture diagram of the BP neural network provided by the embodiment of the present application.

[0014] Figure 4 This is the neuron structure diagram provided by the embodiment of the present application.

[0015] Figure 5 This is the control logic diagram of the single-step state quantity prediction link provided by the embodiment of the present application.

[0016] Figure 6 This is the flowchart of the differential evolution algorithm provided by the embodiment of the present application.

[0017] Figure 7 This is the speed response diagram of the side shift process provided by the embodiment of the present application.

[0018] Figure 8 This is the position response diagram of the side shift process provided by the embodiment of the present application.

[0019] Figure 9 This is the heading response diagram of the side shift process provided by the embodiment of the present application.

[0020] Figure 10 This is the structural schematic diagram of the computer system provided by the embodiment of the present application. Detailed implementation manners

[0021] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0022] The purpose of the present application is to provide a flight control method and system based on a nonlinear model, which improves the accuracy and efficiency of the flight control of the target unmanned helicopter.

[0023] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0024] Example 1, as Figure 1 and Figure 2 shown, this embodiment provides a flight control method based on a nonlinear model, and the flight control method based on the nonlinear model includes the following steps.

[0025] S1. Obtain a BP neural network model. The BP neural network model is used to predict the state quantity at the next discrete moment according to the current state quantity and the current control quantity of the target unmanned helicopter; the state quantity includes: x The velocity in the y axis direction, z The velocity in the x axis direction, roll angle, pitch angle, heading angle, y The angular velocity in the z axis direction,

[0026] In this embodiment, S1 is specifically as follows.

[0027] S11. When the target unmanned helicopter does not have actual flight data, starting from the force balance equation, moment balance equation and motion coordination equation, according to the principle of force generation, conduct mechanism modeling on each aerodynamic component of the target unmanned helicopter, obtain the response of applying control under different reference states at the mechanism level, and obtain mechanism modeling data.

[0028] Among them, the actual flight data includes the control quantity applied by the target unmanned helicopter in single or multiple flight missions and the corresponding state quantity; the mechanism modeling data includes t The state quantity at the t moment, t The control quantity at the

[0029] .

[0030] .

[0031] .

[0032] In the formula, is the aerodynamic force received by the main rotor in the x axis direction, is the aerodynamic force received by the horizontal tail in the x axis direction, is the aerodynamic force received by the vertical tail in the x axis direction, is the aerodynamic force received by the fuselage in the x axis direction, is the total weight of the target unmanned helicopter, is the pitch angle, is the aerodynamic force received by each aerodynamic component in the x axis direction, is the total mass of the target unmanned helicopter, is the first derivative of the velocity of the target unmanned helicopter in the x axis direction, is the aerodynamic force on the main rotor in the y axis direction, is the aerodynamic force on the horizontal tail in the y axis direction, is the aerodynamic force on the fuselage in the y axis direction, is the roll angle, is the aerodynamic force on each aerodynamic component in the y axis direction, is the first derivative of the velocity of the target unmanned helicopter in the y axis direction, is the aerodynamic force on the main rotor in the z axis direction, is the aerodynamic force on the tail rotor in the z axis direction, is the aerodynamic force on the vertical tail in the z axis direction, is the aerodynamic force on the fuselage in the z axis direction, is the aerodynamic force on each aerodynamic component in the z axis direction, is the first derivative of the velocity of the target unmanned helicopter in the z axis direction.

[0033] The above moment balance equations are as follows.

[0034] .

[0035] .

[0036] .

[0037] In the formula, is the aerodynamic moment of the main rotor in the x axis direction, is the aerodynamic moment of the tail rotor in the x axis direction, is the aerodynamic moment of the vertical tail in the x axis direction, is the aerodynamic moment of the fuselage in the x axis direction, is the aerodynamic moment of each aerodynamic component in the x axis direction, is the moment about the x axis direction, is the second derivative of the roll angle, is the main rotor in the yThe aerodynamic moment in the axis direction is the aerodynamic moment in the y axis direction received by the tail rotor, is the aerodynamic moment in the y axis direction received by the vertical tail, is the aerodynamic moment in the y axis direction received by the fuselage, is the aerodynamic moment in the y axis direction received by each aerodynamic component, is the moment of inertia about the y axis direction, is the second derivative of the pitch angle, is the aerodynamic moment in the z axis direction received by the main rotor, is the aerodynamic moment in the z axis direction received by the tail rotor, is the aerodynamic moment in the z axis direction received by the horizontal tail, is the aerodynamic moment in the z axis direction received by the fuselage, is the aerodynamic moment in the z axis direction received by each aerodynamic component, is the moment of inertia about the z axis direction, is the second derivative of the heading angle.

[0038] The above motion coordination equations are as follows.

[0039] .

[0040] In the formula, is the first derivative of the pitch angle, is the first derivative of the roll angle, is the first derivative of the heading angle, is the angular velocity of the target unmanned helicopter in the x axis direction, is the angular velocity of the target unmanned helicopter in the y axis direction, is the angular velocity of the target unmanned helicopter in the z axis direction.

[0041] S12. Determine the basic architecture of the BP neural network.

[0042] As Figure 3 shown, the basic architecture of the BP neural network in this embodiment includes: an input layer, a hidden layer, and an output layer. Among them, the input layer contains 13 neurons, corresponding to the state quantities at t moment (the velocity in the x axis direction,y The velocity in the axis direction, z The velocity in the axis direction, roll angle, pitch angle, heading angle, x The angular velocity in the axis direction, y The angular velocity in the axis direction and z The angular velocity in the axis direction) and t The control amount at the moment (total rotor pitch, longitudinal cyclic pitch, lateral cyclic pitch, and tail rotor pitch); The hidden layer contains 60 neurons; The output layer contains 9 neurons, corresponding to the state quantities at time t+1 ( x The velocity in the axis direction, y The velocity in the axis direction, z The velocity in the axis direction, roll angle, pitch angle, heading angle, x The angular velocity in the axis direction, y The angular velocity in the axis direction and z The angular velocity in the axis direction). As Figure 4 shown, each neuron itself contains an activation function and a bias , The input of each neuron is the weighted sum of the outputs of the previous neurons, and the output of each neuron is the superposition response of the above data, that is, it satisfies: , is the output of the j th neuron, is the output of the i th neuron (also known as the input of the j th neuron), is the connection weight between the i th neuron and the j th neuron, is the bias of the j th neuron, is Sigmoid function, that is, it satisfies: .

[0043] S13. Use the data of mechanism modeling to train the basic architecture of the BP neural network to obtain the BP neural network model.

[0044] In this embodiment, the training of the basic architecture of the BP neural network includes two processes: forward propagation and backward propagation. Forward propagation is to use the direct mapping relationship between input and output under the current connection weights and biases of each neuron. The hidden layer and output layer in the basic architecture of the BP neural network both adopt the Figure 4 neuron structure shown. The initial connection weights and biases of each neuron are randomly generated, and then the prediction deviation of the BP neural network is calculated, is the state output corresponding to the current input of the training sample k ​ is the training output under the current neural network parameters. Using the learning rule to learn the neural network parameters, by solving the gradient and , and using the gradient descent method to update the neural network parameters. When the prediction accuracy requirement is met, stop learning and save the current neural network parameters to obtain the BP neural network model.

[0045] As a preferred implementation, due to the inevitability of errors between the above-mentioned mechanism modeling process of the target unmanned helicopter and the actual flight, there must also be a certain error between the state quantity response predicted by the BP neural network model and the actual state quantity response. In order to improve the accuracy of the flight control of the target unmanned helicopter, it is also necessary to collect the control quantities applied by the target unmanned helicopter during single or multiple flight tasks and the corresponding state quantities after the single or multiple flight tasks of the target unmanned helicopter are completed to obtain the actual flight data of the target unmanned helicopter, and use the actual flight data to update and correct the weights and bias quantities of the BP neural network model, so as to obtain a more accurate BP neural network model and realize the training upgrade of the BP neural network model.

[0046] In addition, both the process of training the basic architecture of the BP neural network based on the mechanism modeling data mentioned above and the process of correcting the BP neural network model using the actual flight data belong to the offline processing process.

[0047] S2. When the target unmanned helicopter executes a single flight task, given the initial state quantity and the future step control quantity sequence, and by repeatedly calling times the single-step state quantity prediction link, obtain the future step state quantity response; .

[0048] As Figure 5 shown, the single-step state quantity prediction link includes: inputting the current state quantity and the current control quantity into the BP neural network model, and using the state quantity predicted by the BP neural network model as the state quantity at the next discrete moment. In addition, when , during the process from the rd step to the th step, the applied control quantity needs to continue to maintain the control quantity at the th step.

[0049] S3. Based on the deviation between the future step state quantity response and the desired state quantity, construct a loss function, and use the differential evolution algorithm to determine the step control quantity sequence corresponding to the minimum of the loss function, and use it as the optimal control quantity sequence.

[0050] In this embodiment, the smaller the deviation between the actual state quantity response of the target unmanned helicopter and the desired state quantity, the smaller the loss function. Therefore, the minimization of the loss function is taken as the optimization goal. The loss function is as follows.

[0051] 。

[0052] In the formula, J is the loss function value composed of the state quantity and the manipulation quantity after applying different manipulation quantity sequences, Y is the state quantity response, Y ref is the desired state quantity, U is the input manipulation quantity, Q 、 R are diagonal positive definite matrices (selected according to the flight state requirements of the target unmanned helicopter).

[0053] Furthermore, the function of the optimization link is to calculate the step manipulation quantity sequence with the minimum loss function. Its implementation method is the differential evolution algorithm. At each discrete moment, an initial sample is randomly generated (i.e., the step manipulation quantity sequence sample), and then the loss function of each sample is calculated to evolve and generate the optimal value of the population. Among them, the individual evolution process is a process of evolving each sample. For the specific optimization process, see Figure 6 。

[0054] S4. Apply the first step manipulation quantity in the optimal manipulation quantity sequence to the target unmanned helicopter to obtain the true state quantity response of the target unmanned helicopter.

[0055] In this embodiment, due to the inevitable existence of factors such as measurement environmental noise and inaccurate modeling, after applying the first step manipulation quantity in the optimal manipulation quantity sequence, the true response of the target unmanned helicopter will inevitably deviate from the response predicted by the BP neural network model. Therefore, the response at this time needs to be fed back to the BP neural network model and used to correct the prediction optimization at the next discrete moment.

[0056] S5. Update the initial state quantity using the true state quantity response.

[0057] S6. Repeat the above steps S2~S5, continuously perform feedback correction on the state quantity predicted by the BP neural network model until the single flight mission of the target unmanned helicopter ends.

[0058] In addition, the above S2~S6 is an online processing process.

[0059] In summary, in view of the difficulties in accurately modeling the target unmanned helicopter and its characteristics of complex strong coupling and strong nonlinearity, a free wake method with higher accuracy (different from the commonly used blade element theory) is adopted in this embodiment. Through iterative calculation, a converged mechanism model is obtained, and then a highly nonlinear flight dynamics response relationship is obtained. The BP neural network is used to learn this response relationship to obtain a response relationship with very low error, avoiding iterative calculation and obtaining a BP neural network model with better accuracy. At the same time, the differential evolution algorithm is adopted in this embodiment to avoid the gradient solution problem in conventional nonlinear optimization problems. As the most efficient algorithm among evolutionary algorithms, the differential evolution algorithm only needs to calculate the response propagation of future state variables during actual processing, without constructing the gradient descent of state variables, and has higher efficiency. In addition, the differential evolution algorithm can also execute control by substituting a sub-optimal solution for the optimal solution according to the limitation of discrete time.

[0060] Further, taking the sideward maneuver of the target unmanned helicopter (i.e., the maneuver flight subject specified in GJB-920B-2017) as an example, after executing according to the above flight control method based on the nonlinear model, the control response result of the target unmanned helicopter is shown in Figures 7 to 9 . As Figures 7 to 9 shown, the simulation result meets the requirements for satisfactory performance of the sideward maneuver in GJB-920B-2017, and can make the speed v of the target unmanned helicopter fluctuate between -40 m / s and 40 m / s during the sideward movement, and the longitudinal displacement X and the vertical displacement Z do not exceed the range limit of ±3 m, and the heading angle does not exceed the range limit of ±10°.

[0061] Embodiment 2 This embodiment provides a computer system, which can be a server or a terminal, and its internal structure diagram can be as shown in Figure 10As shown. The computer system includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer system is used to provide computing and control capabilities. The memory of the computer system includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer system is used to store mechanism modeling data and / or actual flight data of the target unmanned helicopter. The input / output interface of the computer system is used to exchange information between the processor and external devices. The communication interface of the computer system is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a flight control method based on a non-linear model.

[0062] Those skilled in the art can understand that Figure 10 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer system to which the solution of this application is applied. The specific computer system may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0063] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0064] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0065] In summary, the present application mainly has the following advantages.

[0066] (1) The present application has a total of two offline processing processes (the process of determining the BP neural network model based on mechanism modeling and the process of updating the BP neural network model based on actual flight data) and one online processing process (the flight control process of the target unmanned helicopter).

[0067] (2) The offline processing process can effectively reduce the computational amount of the target unmanned helicopter in actual flight control, and at the same time continuously improve the prediction accuracy of the BP neural network model for the response of state variables.

[0068] (3) This application uses machine learning methods to obtain a high-confidence non-linear response relationship throughout the flight, improves the model accuracy in the prediction process, and establishes a control method directly for the non-linear model, thereby enhancing the control effect of maneuvering flight.

[0069] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other.

[0070] Specific examples are used in this article to elaborate on the principles and implementation methods of this application. The descriptions of the above embodiments are only used to help understand the method and its core idea of this application. At the same time, for those of ordinary skill in the art, based on the idea of this application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A flight control method based on a non-linear model, characterized in that, The flight control method based on a non - linear model includes: S1. Obtain a BP neural network model; the BP neural network model is used to predict the state quantity at the next discrete moment according to the current state quantity and the current manipulation quantity of the target unmanned helicopter. S2. When the target unmanned helicopter performs a single flight mission, given the initial state variables and the future step control variable sequence, and by circularly calling times of the single-step state variable prediction link, the future step state variable response is obtained; the single-step state variable prediction link includes: inputting the current state variable and the current control variable into the BP neural network model, and using the state variable predicted by the BP neural network model as the state variable at the next discrete moment; wherein, ; S3. Based on the future Construct a loss function based on the deviation between the response of the step state quantity and the desired state quantity in the future step, and use the differential evolution algorithm to determine the step manipulation quantity sequence corresponding to the minimum of the loss function, and use it as the optimal manipulation quantity sequence; S4. Apply the first manipulation quantity in the optimal manipulation quantity sequence to the target unmanned helicopter to obtain the real - state quantity response of the target unmanned helicopter. S5. Update the initial state quantity by using the real - state quantity response. S6. Repeat steps S2 - S5 continuously to perform feedback correction on the state quantity predicted by the BP neural network model until the single - flight mission of the target unmanned helicopter ends.

2. The flight control method based on a non-linear model according to claim 1, characterized in that, The flight control method based on a non - linear model further includes: After the single - flight mission or multiple - flight missions of the target unmanned helicopter end, collect the manipulation quantity applied and the corresponding state quantity during the single - flight mission or multiple - flight missions of the target unmanned helicopter to obtain the actual flight data of the target unmanned helicopter. Update and correct the weights and bias quantities of the BP neural network model by using the actual flight data.

3. The flight control method based on a non-linear model according to claim 1, characterized in that The state variables include: x The velocity in the axis direction, y The velocity in the axis direction, z The velocity in the axis direction, roll angle, pitch angle, heading angle, x The angular velocity in the axis direction, y The angular velocity in the axis direction, and z The angular velocity in the axis direction; The control variables include: total rotor pitch, longitudinal cyclic pitch, lateral cyclic pitch, and tail rotor pitch.

4. The flight control method based on a non-linear model according to claim 3, wherein The determination process of the BP neural network model specifically includes: When the target unmanned helicopter does not have actual flight data, starting from the force balance equation, moment balance equation, and motion coordination equation, based on the principle of force generation, mechanism modeling is carried out on each aerodynamic component of the target unmanned helicopter to obtain the response of applying maneuvers under different reference states at the mechanism level, and mechanism modeling data is obtained; the actual flight data includes the control inputs applied by the target unmanned helicopter during single or multiple flight missions and the corresponding state variables; the mechanism modeling data includes t the state variables at time t the control inputs at time t and the state variables at time + 1; Determine the basic architecture of the BP neural network. Train the basic architecture of the BP neural network by using the mechanism - modeling data to obtain a BP neural network model.

5. The flight control method based on a non-linear model according to claim 4, characterized in that The basic architecture of the BP neural network includes an input layer, a hidden layer, and an output layer; the input layer contains 13 neurons; the hidden layer contains 60 neurons; the output layer contains 9 neurons.

6. The flight control method based on a non-linear model according to claim 1, characterized in that, The loss function is: ； In the formula, J is the loss function value composed of the state quantity and the manipulation quantity after applying different manipulation quantity sequences, Y is the state quantity response, Y ref is the desired state quantity, U is the input manipulation quantity, Q and R are positive definite diagonal matrices.

7. The flight control method based on a non-linear model according to claim 4, characterized in that, The force balance equation includes: ； ； ； Wherein, is the aerodynamic force on the main rotor in the x axis direction, is the aerodynamic force on the horizontal tail in the x axis direction, is the aerodynamic force on the vertical tail in the x axis direction, is the aerodynamic force on the fuselage in the x axis direction, is the total weight of the target unmanned helicopter, is the pitch angle, is the aerodynamic force on each aerodynamic component in the x axis direction, is the total mass of the target unmanned helicopter, is the first derivative of the velocity of the target unmanned helicopter in the x axis direction, is the aerodynamic force on the main rotor in the y axis direction, is the aerodynamic force on the horizontal tail in the y axis direction, is the aerodynamic force on the fuselage in the y axis direction, is the roll angle, is the aerodynamic force on each aerodynamic component in the y axis direction, is the first derivative of the velocity of the target unmanned helicopter in the y axis direction, is the aerodynamic force on the main rotor in the z axis direction, is the aerodynamic force on the tail rotor in the z axis direction, is the aerodynamic force on the vertical tail in the z axis direction, is the aerodynamic force on the fuselage in the z axis direction, is the aerodynamic force on each aerodynamic component in the z axis direction, is the first derivative of the velocity of the target unmanned helicopter in the z axis direction.

8. The flight control method based on a non-linear model according to claim 4, characterized in that The moment balance equation includes: ； ； ； Wherein, is the aerodynamic moment of the main rotor in the x axis direction, is the aerodynamic moment of the tail rotor in the x axis direction, is the aerodynamic moment of the vertical tail in the x axis direction, is the aerodynamic moment of the fuselage in the x axis direction, is the aerodynamic moment of each aerodynamic component in the x axis direction, is the moment of inertia about the x axis, is the second derivative of the roll angle, is the aerodynamic moment of the main rotor in the y axis direction, is the aerodynamic moment of the tail rotor in the y axis direction, is the aerodynamic moment of the vertical tail in the y axis direction, is the aerodynamic moment of the fuselage in the y axis direction, is the aerodynamic moment of each aerodynamic component in the y axis direction, is the moment of inertia about the y axis, is the second derivative of the pitch angle, is the aerodynamic moment of the main rotor in the z axis direction, is the aerodynamic moment of the tail rotor in the z axis direction, is the aerodynamic moment of the horizontal tail in the z axis direction, is the aerodynamic moment of the fuselage in the z axis direction, is the aerodynamic moment of each aerodynamic component in the z axis direction, is the moment of inertia about the z axis, is the second derivative of the yaw angle.

9. The flight control method based on a non-linear model according to claim 4, wherein The motion coordination equation includes: ； Wherein, is the first derivative of the pitch angle, is the first derivative of the roll angle, is the first derivative of the heading angle, is the angular velocity of the target unmanned helicopter in the x axis direction, is the angular velocity of the target unmanned helicopter in the y axis direction, is the angular velocity of the target unmanned helicopter in the z axis direction.

10. A computer system, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the flight control method based on a non - linear model according to any one of claims 1 - 9.

Images