A method and device for adaptive neural network predictive control of a rotor unmanned aerial vehicle under uncertain disturbance

Abstract

This invention discloses an adaptive neural network predictive control method and computer device for rotary-wing unmanned aerial vehicles (UAVs) under uncertain disturbances. The method includes establishing model predictive control for the UAV, obtaining the reference state and reference control quantity of the model predictive control output, constructing a feedback linearized control law based on the reference state to obtain an ideal control law, approximating the parameters of the control affine dynamics model using a neural network, adaptively updating the weight parameters of the neural network online based on Lyapunov stability to obtain the state error, obtaining the weighted sum of the reference control quantity, the ideal control law, and the state error as the task control law, and controlling the flight of the rotary-wing UAV according to the task control law. This invention can effectively cope with sudden disturbances caused by instantaneous changes in wind fields and complex turbulence faced by rotary-wing UAVs, and can achieve multi-rotor path-following control under uncertain disturbances. This invention has wide applications in the field of UAV technology.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a predictive control method and computer device for an adaptive neural network for a rotary-wing UAV under uncertain disturbances. Background Technology

[0002] Rotary-wing drones, due to their small size and high speed and flexibility, are widely used in complex mission scenarios such as transportation, rescue, and exploration. However, rotary-wing drones are susceptible to various environmental disturbances that can lead to a decline in control quality. For example, when performing exploration missions in unknown and confined spaces such as caves and tunnels, there will be various uncertain disturbances in the environment, such as external interference, near-ground effects, and inter-drone airflow. This requires rotary-wing drones to not only be equipped with powerful sensing systems, but also to have robust control systems to cope with complex and uncertain environmental disturbances.

[0003] On the one hand, rotorcraft unmanned aerial vehicle (UAV) systems can be suddenly subjected to disturbances such as near-ground effects and external airflow. These instantaneous disturbances can cause significant interference to rotorcraft UAVs, leading to large positional deviations and posing significant safety risks. Therefore, stable and rapid tracking performance is crucial for rotorcraft UAVs to accurately reach their target positions. On the other hand, when multi-rotor UAV systems enter an unknown and confined space to perform exploration tasks, the small spacing between rotorcraft UAVs will generate complex turbulence that exerts unmodelable forces on nearby UAVs, making it difficult for the controller to obtain an accurate system model. Summary of the Invention

[0004] In view of the technical problems of rotorcraft drones being prone to deviating from normal operating conditions when subjected to instantaneous and sudden disturbances, the purpose of this invention is to provide an adaptive neural network predictive control method and computer device for rotorcraft drones under uncertain disturbances.

[0005] On one hand, embodiments of the present invention include a predictive control method for an adaptive neural network of a rotary-wing unmanned aerial vehicle under uncertain disturbances, comprising:

[0006] A model predictive control is established for a rotary-wing UAV, and the reference state and reference control quantity output by the model predictive control are obtained;

[0007] Based on the reference state, a feedback linearized control law is constructed to obtain the ideal control law, and a neural network is used to approximate the control affine dynamics model parameters satisfied by the rotary-wing UAV.

[0008] Based on Lyapunov stability, the weight parameters of the neural network are updated adaptively online to obtain the state error;

[0009] The weighted sum of the reference control quantity, the ideal control law, and the state error is obtained as the task control law;

[0010] The flight of the rotary-wing UAV is controlled according to the mission control rate.

[0011] Furthermore, the model-based predictive control of the rotary-wing UAV includes:

[0012] Model predictive control based on integrator system dynamics is established using the following formula:

[0013]

[0014]

[0015]

[0016] x(k)∈X, u(k)∈U

[0017] in, To control the target, x mpcc For the reference state, u mpcc Here, k represents the time step, and N is the total number of time steps. For θ k The approximation of θ k The point on the nominal trajectory of the rotorcraft drone that is closest to the rotorcraft drone. for The corresponding hysteresis error, for The corresponding contour error, q l As the weight of the lag error, q c ρ is the contour error weight, and ρ is the system progress weight. Let x be the predicted velocity of the rotary-wing UAV at time k, and x1 and x2 be the states of the rotary-wing UAV, x = [x1, x2]. T The components in the equation are: u is the ideal control law of the rotary-wing UAV, t represents time, X is the set of state variables of the rotary-wing UAV, and U is the set of control laws of the rotary-wing UAV.

[0018] Further, obtaining the reference state and reference control quantity of the model predictive control output includes:

[0019] Through formula Obtain the reference state x of the model predictive control output mpcc ;

[0020] Through formula a mpcc =u mpcc Obtain the reference control quantity a of the model predictive control output.mpcc .

[0021] Furthermore, the step of constructing a feedback linearized control law based on the reference state to obtain an ideal control law includes:

[0022] Through formula Tracking the reference state x mpcc and the reference control quantity a mpcc ; where (t) represents the functional relationship with time t;

[0023] Through formula Determine the ideal control law u; where f(x) and g(x) are the control affine dynamics models satisfied by the rotary-wing UAV. The parameters in Let f(x) be the nominal model, and a be the pseudo-control variable.

[0024] Furthermore, the pseudo-control quantity a satisfies a = a mpcc +K p ·(x 1mpcc -x1)+K d ·(x 2mpcc -x2)+a rbf ;

[0025] Among them, K p K is the proportional gain of the proportional-derivative controller. d This is the derivative control gain matrix of the proportional-derivative controller.

[0026]

[0027] Furthermore, the approximation of the control affine dynamics model parameters satisfied by the rotary-wing UAV using a neural network includes:

[0028] Obtain the actual weights of the neural network and the implicit function output h(E) of the neural network;

[0029] Based on the actual weights The processing result determined by the implicit function output h(E) Approximating the nominal model

[0030] Furthermore, the online adaptive updating of the weight parameters of the neural network based on Lyapunov stability to obtain the state error includes:

[0031] According to the formula

[0032]

[0033]

[0034] Determine the state error E; where a pd W is the output of the proportional-derivative controller. T ε represents the estimated ideal weights of the neural network, and ε is the system model error caused by uncertain disturbances.

[0035] Furthermore, the online adaptive updating of the weight parameters of the neural network based on Lyapunov stability to obtain the state error includes:

[0036] Define the system error coefficient matrix l1, l2;

[0037] Define the sliding surface function:

[0038]

[0039] Differentiating the sliding surface function, we obtain:

[0040]

[0041] Let the Lyapunov function be:

[0042]

[0043] Differentiating the Lyapunov function, we get:

[0044]

[0045] make Then, the adaptive weights of the neural network are obtained:

[0046]

[0047] Furthermore, the neural network is a radial basis function neural network.

[0048] On the other hand, embodiments of the present invention also include a computer device, including a memory and a processor, the memory being used to store at least one program, and the processor being used to load the at least one program to execute the predictive control method for an adaptive neural network of a rotary-wing unmanned aerial vehicle under uncertain disturbances as described in the embodiments.

[0049] The beneficial effects of the present invention are as follows: The predictive control method of the adaptive neural network for rotorcraft UAVs under uncertain disturbances in the embodiments, by coupling the high-level controller (model predictive control) with the low-level controller (feedback linearization control based on radial basis neural network), controls the flight of the rotorcraft UAV according to the mission control law, which can effectively cope with the sudden disturbances caused by instantaneous change in wind field and complex turbulence interference faced by rotorcraft UAVs, and can realize multi-rotor path following control under uncertain disturbances. Attached Figure Description

[0050] Figure 1 This is a flowchart of the predictive control method of an adaptive neural network for a rotary-wing UAV under uncertain disturbances, as described in the embodiment.

[0051] Figure 2 This is a schematic diagram of the predictive control method of an adaptive neural network for a rotary-wing UAV under uncertain disturbances, as described in the embodiment.

[0052] Figure 3 A schematic diagram of contour error and hysteresis error in the embodiment;

[0053] Figure 4 The simulation results of the predictive control method for an adaptive neural network of a rotary-wing UAV under uncertain disturbances are shown in the embodiment. Detailed Implementation

[0054] The classic approach to controlling rotorcraft unmanned aerial vehicle (UAV) systems under uncertain disturbances is adaptive control. Adaptive control considers a system model with known structure and unknown parameters, and designs adaptive laws based on Lyapunov stability theory to update the model or control parameters. Adaptive control can enable mobile robots to adaptively improve control performance under certain conditions. For example, the L1 adaptive control algorithm improves the trajectory tracking accuracy of quadcopter UAVs. Furthermore, model reference adaptive control can be used to address the impact of parameter uncertainties caused by disturbances on control performance. However, under real disturbances and noise, the unmodeled dynamic statistical characteristics of rotorcraft UAV systems under complex uncertain disturbances are far more complex than the pre-designed conditions. The system operation often fails to meet ideal conditions such as continuous excitation, which leads to the invalidation of the stability assumptions of adaptive control algorithms. Moreover, when encountering severe disturbances, these methods still attempt to track unreachable trajectories rather than adjusting the original trajectory.

[0055] Furthermore, for nonlinear systems that can be linearized by feedback, feedback or feedforward linearization techniques can be used to transform the nonlinear dynamics into a linear integrator model, which can then serve as the predictive model for model predictive control. One approach utilizes the differential planeness characteristics to combine model predictive control with feedforward linearization control, solving the trajectory tracking problem of rotary-wing UAVs. For the path tracking problem of rotary-wing UAVs, a method combining path-following control techniques with a nonlinear dynamic inverse acceleration controller can be employed. However, the performance of these model-based methods is still limited by the discrepancy between the nominal model and the actual system with unknown environmental disturbances.

[0056] Another approach is to improve trajectory tracking accuracy through learning-based methods. For example, robust controllers for Euler-Lagrange systems can be designed using Gaussian processes to estimate uncertain disturbances and their corresponding confidence bounds. However, control constraints are not considered in these controllers. To satisfy control constraints, some methods combine model predictive control with Gaussian processes and consider the constraints in model predictive control; however, Gaussian processes involve large computational costs, making it difficult to guarantee real-time system performance.

[0057] Based on the above principles, this embodiment provides a predictive control method for an adaptive neural network for a rotary-wing unmanned aerial vehicle under uncertain disturbances. (Refer to...) Figure 1 The method includes the following steps:

[0058] S1. Establish model predictive control for the rotary-wing UAV and obtain the reference state and reference control quantity of the model predictive control output;

[0059] S2. Construct a feedback linearized control law based on the reference state to obtain the ideal control law, and use a neural network to approximate the control affine dynamics model parameters satisfied by the rotary-wing UAV.

[0060] S3. Based on Lyapunov stability, the weight parameters of the neural network are updated adaptively online to obtain the state error;

[0061] S4. Obtain the weighted sum of the reference control quantity, ideal control law, and state error as the task control law;

[0062] S5. Control the flight of the rotary-wing UAV according to the mission control rate.

[0063] The following section uses the example of a multi-rotor unmanned aerial vehicle (UAV) system in an environment with unknown interference to explain steps S1-S5 in more detail.

[0064] The principle of steps S1-S5 is as follows: Figure 2 As shown.

[0065] In step S1, firstly, it is assumed that a given θ-parameterized geometric reference trajectory is used.

[0066]

[0067] Where, x ref Let X be the reference trajectory, and let X be the state space of a compact set. The parameterized reference trajectory coordinates can be represented as P(θ) = [x...]. p (θ),y p (θ),z p (θ)] T The parameter θ determines the target point being tracked along the path, and T represents the matrix transpose operation. The actual position coordinates of the system at each time step k can be represented as p. k =[x k ,y k ,z k ] T .

[0068] The established Model Predictive Control (MPC) attempts to minimize the time difference from the current position p. k The projected distance to the desired position P(θ) is calculated while maximizing the progress θ along the straight line. Based on this, the designed cost function can be expressed as:

[0069]

[0070] Among them, e c It is the contour error, e l This is the hysteresis error. The profile error is e. c and hysteresis error e l like Figure 3 As shown. q and γ represent weighting coefficients, θ N Represents progress. Because e c It is itself a solution to an optimal problem, therefore e c This can be considered unsolved, and cannot be directly solved based on the above cost function for the time being. θ k This is the nominally closest trajectory point to the drone, but this point cannot be directly obtained. Therefore, we introduce... It is for θ k An approximation, like Figure 3 As shown. By Figure 3 It can be known symbol

[0071] Hejin

[0072] Similar integrator dynamics:

[0073]

[0074] in, It is a virtual control quantity determined by the controller at each time step. The optimal point for the next trajectory is obtained, where Δt is the sampling time. Define e... l This is a hysteresis error, but e l It cannot be easily obtained, therefore we take it here. For e l An approximation of . Therefore, it is easy to conclude that when When it approaches 0, e c and They can be considered equivalent.

[0075] The nonlinear dynamics of a quadcopter unmanned aerial vehicle system satisfy the typical control affine dynamics system:

[0076]

[0077] The drone's state is represented as x = [x1, x2]. T x1 and x2 are the components of x. In this embodiment, unless otherwise specified, a dot above the symbol represents the derivative of the physical quantity represented by that symbol with respect to time. For example, Let represent the derivative of x1 with respect to time. The system control variable (the ideal control law of the rotary-wing UAV) is denoted by u. In actual UAV systems, f(x) and g(x) may not be known with complete accuracy.

[0078] In step S1, based on the feedback linearization result of step two, a model predictive control scheme based on integrator system dynamics is reconstructed. The model predictive control can be expressed as follows:

[0079]

[0080]

[0081]

[0082] x(k)∈X, u(k)∈U

[0083] in, To control the target, x mpcc For reference state, u mpcc Here, k represents the number of time steps, and N is the total number of time steps, serving as the reference control value. For θ k The approximation of θ k This refers to the trajectory point that is closest to the rotorcraft in the nominal trajectory of the rotorcraft. for The corresponding hysteresis error, for The corresponding contour error, q lAs the weight of the lag error, q c ρ is the contour error weight, and ρ is the system progress weight. Let x be the predicted velocity of the rotary-wing UAV at time k, and x1 and x2 be the states of the rotary-wing UAV, x = [x1, x2]. T In the equation, u represents the ideal control law of the rotary-wing UAV, t represents time, X is the set of state variables of the rotary-wing UAV, and U is the set of control laws of the rotary-wing UAV. c and q l These are the weights for contour error and hysteresis error, respectively, and ρ is the system progress weight.

[0084] In summary, in step S1, the model predictive control output fed to the bottom feedback linearized reference state can be expressed as:

[0085]

[0086] The model predictive control output to the bottom feedback linearized reference control quantity can be expressed as:

[0087] a mpcc =u mpcc .

[0088] In step S2, in order to achieve the reference state x of the upper-layer MPCC mpcc ={x 1mpcc ,x 2mpcc} and control quantity a mpcc Accurate tracking requires the following relationship to be satisfied during tracking:

[0089]

[0090] The control law is designed based on the concept of feedback linearization, and the ideal control law is set as follows:

[0091]

[0092] Where, g(x) -1 Denotes the inverse of g(x), For the nominal model of f(x), the nonlinear system is transformed into an approximate linear double integrator model by the following equation:

[0093]

[0094] Where 'a' is the pseudo-control variable. The model error is caused by uncertain disturbances.

[0095] Specifically, the pseudo-control quantity 'a' is:

[0096] a = a mpcc +K p ·(x 1mpcc -x1)+Kd ·(x 2mpcc -x2)+a rbf

[0097] in K p and K d These are the proportional and derivative control gain matrices of the proportional-derivative controller, respectively.

[0098] In step S2, a radial basis function neural network (RBF) is used to approximate f(x), so that the output of the neural network can replace f(x).

[0099] When the output of the implicit function of the radial basis function neural network is h(E), the weights of the radial basis function neural network are... Then the output of the radial basis function neural network is: By adjusting the weights of the radial basis function neural network This makes the output of the radial basis neural network... Nominal model approximating f(x)

[0100] Based on the feedback linearization control law obtained in step S2, step S3 is executed to further adjust the weights of the radial basis function neural network, thereby adaptively updating the weight parameters of the neural network online.

[0101] In step S3, firstly, the state error is obtained based on the system state using the following formula:

[0102]

[0103]

[0104] Among them, a pd W is the output of the proportional-derivative controller. T Let ε be the estimate of the ideal weights of the neural network, and ε be the system model error caused by uncertain disturbances. The state error E can be obtained by performing the calculation using the above formula.

[0105] In step S3, the system error coefficient matrices l1 and l2 are then set, and the sliding surface function is defined.

[0106]

[0107] Differentiating the sliding surface function, we get:

[0108]

[0109] Let the Lyapunov function be:

[0110]

[0111] Differentiating the Lyapunov function, we get:

[0112]

[0113] Since radial basis function neural networks can achieve arbitrary approximations, and the error ε is sufficiently small, therefore let Then, the adaptive weights of the radial basis function neural network are obtained:

[0114]

[0115] Adjust the weights of the radial basis function neural network to This enables online adaptive updating of the weights in a radial basis function neural network.

[0116] Reference Figure 2 In step S4, the reference control quantity a obtained in step S1 is... mpcc The ideal control law u obtained in step S2 and the state error E obtained in step S3 are assigned corresponding weights w1, w2, and w3, respectively. Then, the weighted sum S = w1a is calculated. mpcc +w2u+w3E, obtain the task control rate S.

[0117] In step S5, the mission control rate S is sent to the low-level attitude controller of the rotary-wing UAV (e.g., Figure 2 The inner loop controller (in the system) is used in the mission control of rotary-wing UAV systems to control the flight of the rotary-wing UAV according to the mission control law S.

[0118] The adaptive neural network predictive control method for rotorcraft UAVs under uncertain disturbances in this embodiment works by coupling a high-level controller (model predictive control) with a low-level controller (feedback linearization control based on radial basis function neural network). The low-level controller uses a neural network to approximate uncertain disturbances online, constructs an integrator model through feedback linearization, and designs adaptive network weights based on system stability to accurately track the reference state. Simultaneously, the high-level controller uses the ideal control law obtained from the feedback linearization control law to optimize the path reference target, achieving proactive response to instantaneous disturbances and providing the low-level controller with a reference state and reference control quantity. Finally, the task control law obtained from the reference control quantity, the ideal control law, and the state error controls the flight of the rotorcraft UAV. This method can effectively cope with sudden disturbances caused by instantaneous wind fields and complex turbulence, enabling multi-rotor path-following control under uncertain disturbances.

[0119] In this embodiment, the simulation results of executing steps S1-S5 once are as follows: Figure 4As shown. According to Figure 4 It can be seen that under the control of steps S1-S5, the rotorcraft UAV can fly along the real trajectory when facing instantaneous sudden wind disturbance deviation. The deviation between the real trajectory and the planned reference trajectory is small, indicating that steps S1-S5 can effectively realize multi-rotor path following control under uncertain disturbances.

[0120] A computer program can be written to execute the predictive control method of the adaptive neural network for a rotary-wing UAV under uncertain disturbances as described in this embodiment. This computer program can be written into a storage medium or computer device. When the computer program is read out and run, the predictive control method of the adaptive neural network for a rotary-wing UAV under uncertain disturbances as described in this embodiment can be executed, thereby achieving the same technical effect as the predictive control method of the adaptive neural network for a rotary-wing UAV under uncertain disturbances in the embodiment.

[0121] It should be noted that, unless otherwise specified, when a feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to the other feature, or indirectly fixed or connected to the other feature. Furthermore, the descriptions of "upper," "lower," "left," and "right" used in this disclosure are only relative to the relative positional relationships of the various components of this disclosure in the accompanying drawings. The singular forms "a," "described," and "the" used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. Moreover, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this embodiment specification is only for describing particular embodiments and is not intended to limit the invention. The term "and / or" as used in this embodiment includes any combination of one or more of the associated listed items.

[0122] It should be understood that although the terms first, second, third, etc., may be used to describe various elements in this disclosure, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, a first element may also be referred to as a second element without departing from the scope of this disclosure, and similarly, a second element may also be referred to as a first element. The use of any and all instances or exemplary language (“e.g.,” “such as,” etc.) provided in this embodiment is intended only to better illustrate embodiments of the invention and, unless otherwise required, does not impose a limitation on the scope of the invention.

[0123] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium. The method can be implemented using standard programming techniques—including a non-transitory computer-readable storage medium configured with a computer program, wherein such a storage medium causes the computer to operate in a specific and predefined manner—according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit (ASIC).

[0124] Furthermore, the procedures described in this embodiment can be performed in any suitable order unless otherwise indicated by this embodiment or clearly contradicted by the context. The procedures (or variations and / or combinations thereof) described in this embodiment can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0125] Furthermore, the method can be implemented in any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices. Aspects of the invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. Furthermore, the machine-readable code, or portions thereof, can be transmitted via wired or wireless networks. The invention described in this embodiment includes these and other different types of non-transitory computer-readable storage media when such media comprises instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor. When programmed according to the methods and techniques described in the invention, the invention also includes the computer itself.

[0126] A computer program can be applied to input data to perform the functions described in this embodiment, thereby transforming the input data to generate output data stored in non-volatile memory. The output information can also be applied to one or more output devices, such as a display. In a preferred embodiment of the invention, the transformed data represents physical and tangible objects, including specific visual depictions of physical and tangible objects generated on the display.

[0127] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention, as long as they achieve the technical effects of the present invention by the same means, should be included within the scope of protection of the present invention. Within the scope of protection of the present invention, the technical solutions and / or implementation methods can have various modifications and variations.

Claims

1. A predictive control method for a rotary-wing unmanned aerial vehicle (UAV) under uncertain disturbances using an adaptive neural network, characterized in that, The predictive control method for an adaptive neural network for a rotary-wing unmanned aerial vehicle under uncertain disturbances includes: A model predictive control is established for a rotary-wing UAV, and the reference state and reference control quantity output by the model predictive control are obtained; Based on the reference state, a feedback linearized control law is constructed to obtain the ideal control law, and a neural network is used to approximate the control affine dynamics model parameters satisfied by the rotary-wing UAV. Based on Lyapunov stability, the weight parameters of the neural network are updated adaptively online to obtain the state error; The weighted sum of the reference control quantity, the ideal control law, and the state error is obtained as the task control law; The flight of the rotary-wing UAV is controlled according to the mission control rate. The model-based predictive control of the rotary-wing UAV includes: Model predictive control based on integrator system dynamics is established using the following formula: in, In order to control the target, This is the reference state. The reference control quantity, Indicates the number of time steps. The total number of time steps. To Approximation, The point on the nominal trajectory of the rotorcraft drone that is closest to the rotorcraft drone. for The corresponding hysteresis error, for The corresponding contour error, As the weight of the lag error, For contour error weights, As the system progress weight, For the rotary-wing UAV in The predicted speed and The state of the rotary-wing UAV The components in The ideal controllability of the rotary-wing UAV is... Indicates time, This is the set of state variables for the rotary-wing unmanned aerial vehicle. This is the set of control laws for the rotary-wing unmanned aerial vehicle; The step of obtaining the reference state and reference control quantity of the model predictive control output includes: Through formula Obtain the reference state of the model predictive control output. ; Through formula Obtain the reference control quantity of the model predictive control output. ; The step of constructing a feedback linearized control law based on the reference state to obtain an ideal control law includes: Through formula Track the reference state and the reference control quantity ;in, Indicates time The functional relationship; Through formula Determine the ideal control rate ;in, and The control affine dynamics model satisfied by the rotorcraft UAV The parameters in for The nominal model, It is a pseudo-control quantity; The process of using a neural network to approximate the control affine dynamics model parameters satisfied by the rotary-wing UAV includes: Obtain the actual weights of the neural network and the hidden function output of the neural network ; Based on the actual weights and the implicit function output Determined processing result Approximating the nominal model ; The method of online adaptively updating the weight parameters of the neural network based on Lyapunov stability to obtain the state error includes: According to the formula Determine the state error ;in, The output of the proportional-derivative controller, This is an estimate of the ideal weights for the neural network. This refers to the system model error caused by uncertain disturbances.

2. The predictive control method for an adaptive neural network of a rotary-wing unmanned aerial vehicle under uncertain disturbances according to claim 1, characterized in that: The pseudo-control quantity satisfy ; in, For the proportional-derivative controller, This is the derivative control gain matrix of the proportional-derivative controller. .

3. The predictive control method for an adaptive neural network of a rotary-wing unmanned aerial vehicle under uncertain disturbances according to claim 1, characterized in that, The method of online adaptively updating the weight parameters of the neural network based on Lyapunov stability to obtain the state error includes: Set the system error coefficient matrix ; Define the sliding surface function: Differentiating the sliding surface function, we obtain: Let the Lyapunov function be: Differentiating the Lyapunov function, we get: make Then, the adaptive weights of the neural network are obtained: 。 4. The predictive control method for an adaptive neural network of a rotary-wing unmanned aerial vehicle under uncertain disturbances according to any one of claims 1-3, characterized in that, The neural network is a radial basis function neural network.

5. A computer device, characterized in that, The system includes a memory and a processor, the memory being used to store at least one program, and the processor being used to load the at least one program to execute the predictive control method for an adaptive neural network for a rotary-wing unmanned aerial vehicle under uncertain disturbances as described in any one of claims 1-4.

Images